TR41.3.5/99-05-07
Standard Method for Measuring
Transmission Performance of
Hands-Free Telephone Sets
P1329 Draft 22
April 1999
Copyright  1999 by the Institute of Electrical and Electronic Engineers, Inc.
345 East 47th Street, New York, NY 10017 USA
All rights reserved.
This is an unapproved draft of a proposed IEEE standard, subject to change. Permission is hereby granted
for IEEE Standards Committee participants to reproduce this document for purposes of IEEE
standardization activities. Permission is also granted for member bodies and technical committees of ISO
and IEC to reproduce this document for purposes of developing a national position. Other entities seeking
permission to reproduce portions of this document for these or other uses, must contact the IEEE Standards
Department for the appropriate license. Use of information contained in this unapproved draft is at your
own risk.
IEEE Standards Department
Copyright and Permissions
445 Hoes Lane, PO Box 1331
Piscataway, NJ 08855-1331 USA
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
1
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
IEEE Standards documents are developed within the Technical Committees of the IEEE Societies and the
Standards Coordinating Committees of the IEEE Standards Board. Members of the committees serve
voluntarily and without compensation. They are not necessarily members of the Institute. The standards
developed within IEEE represent a consensus of the broad expertise on the subject within the Institute as
well as those activities outside of IEEE that have expressed an interest in participating in the development
of the standard.
Use of an IEEE Standard is wholly voluntary. The existence of an IEEE Standard does not imply that there
are no other ways to produce, test, measure, purchase, market, or provide other goods and services related
to the scope of the IEEE Standard. Furthermore, the viewpoint expressed at the time a standard is approved
and issued is subject to change brought about through developments in the state of the art and comments
received from users of the standard. Every IEEE Standard is subjected to review at least once every five
years for revision or reaffirmation. When a document is more than five years old and has not been
reaffirmed, it is reasonable to conclude that its contents, although still of some value, do not wholly reflect
the present state of the art. Users are cautioned to check to determine that they have the latest edition of any
IEEE Standard.
Comments for revision of IEEE Standards are welcome from any interested party, regardless of membership
affiliation with IEEE. Suggestions for changes in documents should be in the form of a proposed change of
text, together with the appropriate supporting comments.
Interpretations: Occasionally questions may arise regarding the meaning of portions of standards as they
relate to specific applications. When need for interpretation is brought to the attention of IEEE, the Institute
will initiate action to prepare appropriate responses. Since IEEE Standards represent a consensus of all
concerned interests, it is important to ensure that any interpretation has also received the concurrence of a
balance of interests. For this reason IEEE and the members of its technical committees are not able to
provide an instant response to interpretation requests except in those cases where the matter has previously
received formal consideration.
Comments on standards and requests for interpretations should be addressed to:
Secretary, IEEE Standards Board
445 Hoes Lane
P.O. Box 1331
Piscataway, NJ 08855-1331
USA
IEEE Standards documents are adopted by the Institute of Electrical and Electronics Engineers without
regard to whether their adoption may involve patents on articles, materials, or processes. Such adoption
does not assume any liability to any patent owner, nor does it assume any obligation whatever to parties
adopting the standards documents.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
2
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
INTRODUCTION
(This introduction is not a part of IEEE Standard P1329-199x, IEEE Standard Methods for Measuring Transmission
Performance of Hands-Free Telephone Sets.)
This standard has been developed in response to a widely expressed need by the telecommunications
industry for a standard, comprehensive method for testing the transmission performance of hands-free
telephone sets. This standard is also in agreement with the ITU-T (formerly CCITT) test arrangements and
calibration procedures.
The IEEE will maintain this standard current with the state of the technology. Comments on this standard
and suggestions for the additional material that should be included are invited. Comments should be
addressed to: Secretary, IEEE Standards Board, The Institute of Electrical and Electronics Engineers, Inc.,
445 Hoes Lane, Piscataway, NJ 08855.
This new standard began in 1992 and was prepared by the Subcommittee on Telephone Instrument Testing
of the Transmission and Access Systems Committee of the IEEE Communications Society (formerly the
IEEE Communications Technology Group).
At the time this standard was approved, the members of the Working Group of the Subcommittee on
Telephone Instrument Testing, were as follows:
John Bareham
Roger Britt
Chandru Butani
Cliff Chamney
Paul Coverdale
Steve Graham
James Gurnavage
Roger Gutzwiller
Glenn Hess
Roger Hunt
Frederick M. Kruger
Ron Magnuson
Stephen Rittmueller
Terry Spencer
Joseph Sternalio
Christopher Struck
Stephen Whitesell
Robert Young
At the time this standard was approved, the members of the Subcommittee on Telephone Instrument
Testing, were as follows:
John Bareham, Chair
Glenn Hess, Vice Chair
Steve Graham, Secretary
Roger Britt
Chandru Butani
Cliff Chamney
Paul Coverdale
James Gurnavage
Roger Gutzwiller
Frederick M. Kruger
Ron Magnuson
Stephen Rittmueller
Terry Spencer
Christopher Struck
Stephen Whitesell
Robert Young
When the IEEE Standards Board approved this standard on ... it had the following membership:
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
3
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
TABLE OF CONTENTS
1. Overview................................................................................................................................................ 8
1.1. Scope. ................................................................................................................................... 8
1.2. Purpose. ................................................................................................................................ 8
1.3. Contents of Standard............................................................................................................. 8
1.4. How To Use This Standard................................................................................................... 9
2. References.............................................................................................................................................. 10
3. Definitions ............................................................................................................................................. 12
4. Acronyms and Abbreviations................................................................................................................. 16
5. Test Methods ......................................................................................................................................... 16
5.1 General. ................................................................................................................................. 16
5.2 Fast Fourier Transform (FFT) and Cross Spectrum Analysis. ............................................... 16
5.2.1 Dual-Channel FFT................................................................................................ 17
5.2.2 Single-Channel FFT. ............................................................................................ 17
5.2.3 Maximum Length Sequence (MLS) Analysis....................................................... 17
5.3 Real-Time Filter Analysis (RTA). ......................................................................................... 17
5.3.1 Dual-Channel Real-Time Filter Analysis. ............................................................ 17
5.3.2 Single-Channel Real-Time Filter Analysis. .......................................................... 17
5.4 Sine-Based Analysis. ............................................................................................................. 18
5.4.1 Discrete Tone (Stepped Sine)............................................................................... 18
5.4.2 Swept Sine............................................................................................................ 18
5.4.3 Time Delay Spectrometry (TDS). ........................................................................ 18
5.5 Free Field Techniques. .......................................................................................................... 18
5.6 Method Comparative Summary. ............................................................................................ 19
Signal Type ................................................................................................................................................ 19
6. Test Signals............................................................................................................................................ 20
6.1 General. ................................................................................................................................. 20
6.2 Classifications........................................................................................................................ 20
6.3 Modulation Types.................................................................................................................. 20
6.3.1 Square Wave Modulation..................................................................................... 20
6.3.2 Sine Wave Modulation......................................................................................... 20
6.3.3 Pseudo-Random Modulation. ............................................................................... 20
6.4 Deterministic Signals............................................................................................................. 21
6.4.1 Sine Wave. ........................................................................................................... 21
6.4.2 Pseudo-Random.................................................................................................... 21
6.5 Random Signals. .................................................................................................................... 21
6.5.1 White Noise.......................................................................................................... 21
6.5.2 Pink Noise. ........................................................................................................... 21
6.6 Speech-Like Signals. ............................................................................................................. 21
6.6.1 Simulated Speech. ................................................................................................ 21
6.6.1.1 P.50 Artificial Voice. ..................................................................................... 22
6.6.1.2 P59 Artificial Conversational Speech............................................................. 22
6.6.1.3 Simulated Speech Generator (SSG). ................................................................. 22
6.6.2 Synthesized Speech. ............................................................................................. 22
6.6.3 Real Speech. ......................................................................................................... 22
6.7 Compound Signals................................................................................................................. 22
6.7.1 Sequential Presentation. ....................................................................................... 22
6.7.1.1 Composite Source Signal. ................................................................................. 23
6.7.2 Simultaneous Presentation.................................................................................... 23
6.7.2.1 TDS Sweep with P.50 Noise Bursts. ................................................................. 23
6.7.2.2 TDS Sweep with Artificial Voice...................................................................... 23
6.7.2.3 TDS Sweep with Real Speech ........................................................................... 23
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
4
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
6.7.2.4 TDS Sweep with Random or Pseudorandom Noise .......................................... 23
6.7.2.5 Pseudorandom Noise with P.50 Noise Bursts. .................................................. 23
6.7.2.6 Pseudorandom Noise with Artificial Voice. ...................................................... 23
6.7.2.7 Pseudorandom Noise with Real Speech ............................................................ 23
6.7.2.8 Pseudorandom Noise with Random or Pseudorandom Noise............................ 23
6.7.2.9 Sine Wave with Notched Real Speech. ............................................................. 24
6.8 Test Signal Bandwidth ............................................................................................ 24
6.9 Signal Parameter Summary.................................................................................................... 24
Test Signal......................................................................................................................... 24
6.10 Signal Comparative Summary. ............................................................................................ 24
7. Test Equipment, Environment, and Impairments ................................................................................... 25
7.1 General. ................................................................................................................................. 25
7.2 Test Equipment...................................................................................................................... 25
7.2.1 Measuring Microphones....................................................................................... 25
7.2.2 Mouth Simulator................................................................................................... 26
7.2.3 Head And Torso Simulator (HATS)..................................................................... 26
7.2.4 Standard Circuits for Transmission and Voice Switching Measurements. ........... 26
7.2.5 Standard Circuits for Acoustic Echo Canceller Measurements. ........................... 26
7.3 Test Environment................................................................................................................... 27
7.3.1 Background Noise Level. ..................................................................................... 27
7.3.2 Anechoic Chamber. .............................................................................................. 27
7.3.3 Simulated Free-Field. ........................................................................................... 29
7.3.4 Test Table............................................................................................................. 29
7.3.5 Test Room Characteristics.................................................................................... 30
7.4 Impairments. .......................................................................................................................... 30
7.4.1 Network Impairments. .......................................................................................... 30
7.4.1.1 Test Loops......................................................................................................... 30
7.4.1.2 Loop Current. .................................................................................................... 30
7.4.1.3 Termination Impedance..................................................................................... 30
7.4.1.4 Network Noise................................................................................................... 30
7.4.2 Acoustic Impairments........................................................................................... 30
7.4.2.1 Nearby Reflecting Surfaces. .............................................................................. 30
7.4.2.2 Hoth Room Noise.............................................................................................. 31
7.4.2.3 Room Reverberation.......................................................................................... 31
7.5 Post Processing...................................................................................................................... 31
8. Test Calibration...................................................................................................................................... 31
8.1 Measurement Bandwidth and Resolution. ............................................................... 31
8.2 Send. ...................................................................................................................................... 31
8.2.1 Acoustic Test Spectrum........................................................................................ 31
8.2.2 Acoustic Test Level.............................................................................................. 32
8.2.3 Mouth Simulator Calibration Procedure............................................................... 32
8.2.4 HATS Calibration Procedure ............................................................................... 33
8.3 Receive and Echo. ................................................................................................................. 34
8.3.1 Electrical Test Spectrum. ..................................................................................... 34
8.3.2 Electrical Test Level............................................................................................. 34
9. Transmission Measurements .................................................................................................................. 35
9.1 General. ................................................................................................................................. 35
9.1.1 Measurement Bandwidth and Resolution. ............................................................ 35
9.1.2 Recommended Test Methods. .............................................................................. 35
9.2 Positioning of HFT and Test Transducers. ............................................................................ 37
9.2.1 Desktop Hands-Free. ............................................................................................ 37
9.2.2 Open Listening. .................................................................................................... 39
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
5
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
9.2.3 Non Desk-Top Hands-Free................................................................................... 40
9.2.4 HATS Positioning. ............................................................................................... 40
9.3 Send. ...................................................................................................................................... 41
9.3.1 General. ................................................................................................................ 41
9.3.2 Special Instructions For Microphones.................................................................. 41
9.3.3 Frequency Response............................................................................................. 41
9.3.4 Noise. ................................................................................................................... 41
9.3.5 Input-Output Linearity.......................................................................................... 42
9.3.6 Distortion.............................................................................................................. 42
9.3.6.1 Test Signal......................................................................................................... 42
9.3.6.2 Suitability Test .................................................................................................. 42
9.3.6.3 Distortion Measurement .................................................................................... 43
9.3.7 Loudness Rating Applications.............................................................................. 43
9.3.8 Conversion of Measured Data to ISO R 10 Format.............................................. 43
9.3.9 Mid-Band Average Send Sensitivity. ................................................................... 44
9.3.10 Send Directionality............................................................................................ 44
9.3.10.1 Measurement Procedure. ................................................................................. 44
9.3.10.2 Data Presentation............................................................................................. 45
9.4 Receive. ................................................................................................................................. 46
9.4.1 General. ................................................................................................................ 46
9.4.2 Frequency Response............................................................................................. 46
9.4.3 HATS DRP to ERP Correction. ........................................................................... 47
9.4.4 Noise. ................................................................................................................... 47
9.4.5 Input-Output Linearity.......................................................................................... 47
9.4.6 Distortion.............................................................................................................. 47
9.4.6.1 Test Signal......................................................................................................... 48
9.4.6.2 Suitability Test .................................................................................................. 48
9.4.6.3 Distortion Measurement .................................................................................... 48
9.4.7 Loudness Rating Applications.............................................................................. 49
9.4.7.1 Corrected Receive Loudness Rating Using a Free-Field Microphone............... 49
9.4.7.2 Corrected Receiver Loudness Rating When Using HATS ................................ 49
9.4.8 Mid-Band Average Receive Sensitivity................................................................ 49
9.4.9 Receive Directionality. ......................................................................................... 50
9.5 Digital Only. .......................................................................................................................... 50
9.5.1 Echo Path Response. ............................................................................................ 50
10. Voice Switching Measurements.......................................................................................................... 51
10.1 General. ............................................................................................................................... 51
10.2 Classification. ...................................................................................................................... 51
10.3 Switching Parameters. ......................................................................................................... 51
10.4 Test Conditions.................................................................................................................... 52
10.4.1 Signal Levels. ..................................................................................................... 52
10.4.2 Loop Lengths...................................................................................................... 52
10.4.3 Noise Levels....................................................................................................... 53
10.5 Test Parameters. .................................................................................................................. 53
10.5.1 Threshold Level.................................................................................................. 53
10.5.2 Build-Up Time. .................................................................................................. 53
10.5.3 Hang-Over Time................................................................................................. 54
10.5.4 Switching Time and Thresholds Between Two Active States. ........................... 56
10.5.5 Take-Over (Break-Through) Time. .................................................................... 60
11. Acoustic Echo Canceller Measurements .............................................................................................. 60
11.2 Test Signals. ........................................................................................................................ 61
11.3 Test Conditions.................................................................................................................... 61
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
6
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
11.4 Round Trip Echo Path Delay (EPD).................................................................................... 61
11.5 Echo Return Loss (ERLST) – Single Talk. ........................................................................... 62
11.5.1 Echo Return Loss, Temporally Weighted – Single Talk (ERLTST)..................... 62
11.5.2 Echo Return Loss, Segmental – Single Talk (ERLSST). ...................................... 62
11.5.3 Weighted Terminal Coupling Loss – Single Talk (TCLWST).............................. 62
11.6 Convergence Time (Tc)....................................................................................................... 63
11.7 Echo Return Loss, Temporally Weighted – Double Talk (ERLTDT).................................... 63
11.8 Send Speech Attenuation During Double Talk (ADT_S). ...................................................... 63
11.8.1 Send Speech Attenuation During Double Talk vs Time (ASTDT). ....................... 64
11.8.2 Send Speech Attenuation During Double Talk, Conversational Average
(ASADT)........................................................................................................................... 64
11.9 Receive Speech Attenuation During Double Talk (ARDT) ................................................... 64
11.9.1 Receive Speech Attenuation During Double Talk vs Time (ARTDT)................... 64
11.9.2 Receive Speech Attenuation During Double Talk, Conversational
Average (ARADT)............................................................................................................ 65
11.10 Send Speech Front End Clipping Time During Double Talk (TSFDT). ............................... 65
11.11 Receive Speech Front End Clipping Time During Double Talk (TRFDT)........................... 65
Annex A Simulated Speech Generator....................................................................................................... 67
Annex B Composite Source Signal ............................................................................................................ 69
Annex C ITU-T Recommendation P.50 Noise Bursts Over TDS Sweep .................................................. 71
Annex D Hoth Room Noise ....................................................................................................................... 72
Annex E Useful Conversion Procedures .................................................................................................... 75
Annex F Recommended Test Bed.............................................................................................................. 77
Annex G Detailed Test Methodology For Temporally Weighted ERL...................................................... 82
Annex H ERLt Test Algorithm................................................................................................................... 86
Annex I Double Talk Testing..................................................................................................................... 91
Annex J Acoustic Echo Path Tutorial ........................................................................................................ 99
Annex K HFT Microphones ...................................................................................................................... 100
Annex L 1/3 Octave Passbands.................................................................................................................. 101
Annex X DRP TO ERP Corrections For HATS Receive Measurements................................................... 102
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
7
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Standard Method for Measuring
Transmission Performance of
Hands-Free Telephone Sets
1. Overview
Objective or subjective methods can be used to measure hands-free telephone (HFT) transmission
performance. This standard discusses objective procedures utilizing a sound source, laboratory
microphone, and test instruments to characterize transmission performance. Subjective procedures are
particularly applicable for rating overall communication connections involving the real voice and real ear of
human subjects. Hands-free telephones can be evaluated by purely objective methods provided they agree
with the desirable performance characteristics of subjective testing.
Hands-free telephones present a complex problem for accurate and repeatable measurements of the device
characteristics. In order to meet these measurement goals the reader must be aware of the non-linear and
time-variant characteristics of an HFT. This standard describes test signals and corresponding analysis
methods, which can be chosen to ensure the HFT is in a well defined operating state during testing.
Execution of this standard provides a means of determining the operational characteristics of an HFT in
conditions encountered during normal operation.
1.1. Scope.
This standard provides the techniques for objective measurement of electroacoustic and voice switching
characteristics of analog and digital hands-free telephones. Due to the various characteristics of HFTs and
the environments in which they operate, not all of the test procedures in this standard are applicable to all
HFT’s. Application of the test procedures to atypical HFTs must be determined on an individual basis.
1.2. Purpose.
The purpose of this standard is to provide practical methods for making laboratory measurements of the
transmission and voice switching characteristics of HFTs so that their performance may be evaluated on a
standardized basis.
1.3. Contents of Standard.
This is a brief summary of the sections contained in the standard. The primary measurement procedures
appear in the sectioned portion of the document. Attached Annexes contain additional information or
details of procedures referred to from within the relevant section.
Sections 2, 3 and 4 provide references, definitions, and acronyms, which will be useful in executing the tests
of this standard. These sections provide a background in the terminology used for HFT testing.
Section 5 details the most common analysis techniques used to make the electroacoustic and switching
measurements. The section explains the advantages and disadvantages of each technique in relation to the
stimulus signals chosen for testing.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
8
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Section 6 details the test signals used to place the HFT in a well-defined state for the measurements. This
section explains the use of these signals to maintain a stable HFT state as well as provide a suitable
measurement signal.
Section 7 details the test equipment and the test environment. The test equipment includes the interfaces,
analyzers and transducers. The test environment include the room characteristics and furniture used during
testing.
Section 8 details the calibration procedure needed to ensure that the equipment is in a known state.
Calibration of the electrical interfaces and acoustic transducers is explained.
Section 9 details the various send and receive transmission test procedures. This section includes
procedures for positioning the HFT for testing and generating reports of the transmission characteristics.
Section 10 details the procedures and parameters involved in testing the voice switching characteristics of
half-duplex and full-duplex HFTs. Various environmental conditions are explained in reference to the
operation of a voice switched HFT.
Section 11 details the procedures and parameters unique to testing an HFT equipped with an acoustic echo
canceller (AEC). Characteristics unique to AEC HFTs are presented. Also, environment and interface
conditions specific to AECs are described.
1.4. How To Use This Standard.
The following sequence is suggested when referencing the contents of this standard prior to setting up a test
sequence:
1. Choose Acoustic Environment [section 7]. Electroacoustic measurements on hands-free telephones must
be made in an environment with a low background noise level. In addition, either the test environment or
the test method should minimize or eliminate the influence of reflections. Section 7.6 provides guidance in
the selection and set-up of the test environment. Under certain circumstances, measurements may have to be
made in a "real room environment." That is, in a room having acoustic characteristics similar to those
expected to exist in a room in which the hands free telephone may be used.
Be sure to record for the test report the test room environmental conditions of temperature, humidity, and
barometric pressure, in addition to the ambient acoustic environment ( overall noise level, and 1/3 Octave
band sound levels are recommended).
2. Choose test signal based on HFT behavior [section 6]. Selection of the correct test signal(s) is critical
since different hands-free telephones may respond differently to many of the possible test signals. The
choice will be a balance between one that correctly stimulates the detection circuits of the HFT, and one
that is suitable for the specific measurement.
3. Choose analysis method (based on test signal), [section 5]. Each analysis technique has inherent
advantages and disadvantages. A particular method may be better suited for use with certain stimulus
signals. In fact, certain methods rely upon the use of a synchronized or otherwise unique signal. Please note
that within a given test, different test signal types (i.e., continuous sine, pulsing, alternating tones, pink
noise, tone burst, simulated speech, etc.) may be required.
4. Setup hardware, fixtures, etc. [section 7]. Different equipment will be required depending upon whether
analog or digital sets are being tested and which test signals and analysis methods are chosen. The first part
of section 7 provides guidance in the selection of the test equipment required to generate and process the
selected test signal(s).
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
9
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
5. Calibrate the system [section 8]. Follow the checkout and calibration procedures provided by instrument
manufacturer(s), or existing relevant standards, to verify proper operation and calibration of all equipment,
as appropriate.
6. Perform transmission measurements [section 9]. Procedures are provided for measuring parameters
affecting the send and receive performance of hands-free telephone sets. Procedures for measuring electrical
characteristics also are provided in this section.
7. Perform voice switching measurements [sections 10 & 11]. Refer to section 10 to select the procedures
and parameters to be used in testing the voice switching characteristics of hands-free telephones. The
manner in which the HFT handles the transmission of signals in both directions can be classified into
different groups. This section provides guidance in test set-up, test environment, test equipment and
procedures. Section 11 provides guidance for the testing of the unique characteristics of hands-free
telephones with acoustic echo cancellation (AEC) circuitry. The methods described will effectively evaluate
AEC performance.
For each test, describe the test environment, position of HFT, test signal, analysis method, frequency range
and resolution, and other relevant conditions.
2. References
This standard shall be used in conjunction with the following publications. When the following standards
are superseded by an approved revision, the revision shall apply, but the impact on results should be
determined.
[1] ANSI S1.1-1994, American National Standard Acoustical Terminology (Including Mechanical Shock
and Vibration)1
[2] ANSI S1.4-1983, American National Standard Specification for Sound Level Meters
[3] ANSI S1.6-1984 (R1990), Preferred Frequencies, Frequency Levels, and Band Numbers for Acoustical
Measurements
[4] ANSI S1.11-1986, Specification for Octave-Band and Fractional-Octave-Band Analog and Digital
Filters
[5] ANSI S1.12-1967 (R1986), American National Standard Specifications for Laboratory Standard
Microphones
[6] ANSI S3.36-1985, American National Standard Specification for a Manikin for Simulated In-Situ
Airborne Acoustic Measurements.
[7] ANSI/IEEE Standard 100-1988, IEEE Standard Dictionary of Electrical and Electronics Terms2
1ANSI and ISO publications are available from the Sales Department, American National Standards Institute, 11 West 42nd Street,
13th Floor, New York, NY 10036 USA. (Tel: 212-642-4900 Fax: 212-302-1289)
2IEEE publications are available from the Institute of Electrical and Electronics Engineers, Service Center, 445 Hoes Lane, P.O. Box
1331, Piscataway, NJ 08855-1331, USA. (Tel: 908-981-0060 Fax: 908-981-9667)
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
10
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
[8] ANSI/IEEE Standard 269-1992, IEEE Standard Methods for Measuring Transmission Performance of
Analog and Digital Telephone Sets
[9] ANSI/IEEE Standard 661-1992, IEEE Standard Method for Determining Objective Loudness Ratings of
Telephone Connections
[10] ANSI/IEEE Standard 743-1984, IEEE Standard Methods and Equipment for Measuring the
Transmission Characteristics of Analog Voice Frequency Circuits
[11] British Standard 6789, Apparatus with One or More Particular Functions for Connection to the British
Telecommunications Public Switched Telephone Network, Part 2, Specification for Apparatus with
Loudspeaking Facilities, 1984
[Need a reference, similar to footnote 2 & 3, for obtaining this publication.]
[12] European Telecommunications Standards Institute Draft pr1-ETS 300 245-3, Integrated Services
Digital Network (ISDN); Technical Characteristics of Telephony Terminals, Part 3, PCM A-Law
Loudspeaking and Handsfree Function, 1994
[Need a reference, similar to footnote 2 & 3, for obtaining this publication.]
[13] ISO 3: Preferred Numbers, Series of Preferred Numbers2
[14] ISO/DIS 266 Acoustics, Preferred Frequencies for Measurements (Revision of ISO 266:1975)
[15] ITU-T Recommendation G.122
[16] ITU-T Recommendation G.167
[17] ITU-T Recommendation P.34, Transmission Characteristics of Hands-Free Telephones, 19933
[18] ITU-T Recommendation P.50, Artificial Voices, Telephone Transmission Quality, 1989
[19] ITU-T Recommendation P.51, Artificial Mouths, 1993
[20] ITU-T Recommendation P.56, Objective Measurement of Active Speech Level, 1993[21] ITU-T
Recommendation P.57, Artificial Ears, 1993
[21] ITU-T Recommendation P.58, Head and Torso Simulator for Telephonometry, 1993,
[22] ITU-T Recommendation P.59, Artificial Conversational Speech, 1993
[23] ITU-T Recommendation P.76, Determination of Loudness Ratings; Fundamental Principles,
Telephone Transmission Quality, Geneva, 1989
[24] ITU-T Recommendation P.79, Calculation of Loudness Ratings for Telephone Sets, 1993
[25] ITU-T Recommendation P.340
3ITU-T (formerly CCITT) publications are available from the ITU-T General Secretariat, International Telecommunications Union,
Sales Section, Place des Nations, CH-1211, Geneve 20, Switzerland/Suisse. (Tel: +41-22-730-5285 Fax: +41-22-730-5194) ITU-T
publications are also available in the United States from National Technical Information Service, Department of Commerce, 5285
Port Royal Road, Springfield, VA 22161. (Tel: 703-487-4650 Fax: 703-321-8547)
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
11
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
[26] ITU-T Recommendation P.501
3. Definitions
These definitions apply specifically to measurements of the transmission performance of hands-free
telephone sets, and may not be applicable to other disciplines. For definitions not covered, see ANSI S1.11994 [1] and ANSI/IEEE Standard 100-1988 [7].
3.1
acoustic echo canceller (AEC). A circuit or algorithm designed to eliminate acoustic echoes and
prevent howling due to acoustic feedback from loudspeaker to microphone.
3.2
acoustic input. The free field sound pressure level developed by a mouth simulator at the mouth
reference point. Also see sound pressure level.
3.3
acoustic output. The sound pressure level developed at the measuring microphone. Also see sound
pressure level.
3.4
attenuation range (aH). The difference in level in dB between maximum inserted switched loss and
the full removal of that switch loss in a particular transmission direction.
3.5
automatic gain control (AGC). A circuit or algorithm that varies gain as a function of the input
signal amplitude.
3.6
build-up time (TR). Time from the input signal going above the threshold level up to 50% of the
complete removal of the insertion loss.
3.7
convergence time (Tc). The time required to reach within 3 dB of maximum echo return loss, or 25
dB loss, which ever occurs first.
3.8
3.9
double-talk (DT). Two talkers speaking simultaneously in opposite transmission directions.
echo path delay (EPD). The total delay of the echo path from the receive electrical test point to the
send electrical test point, excluding any delay in the test equipment.
3.10 echo return loss (ERL). The loss in the echo path from the receive electrical test point to the send
electrical test point. ERL can be weighted in time, frequency or both.
3.11 echo return loss, segmental double talk (ERLSTD). As per annex I with receive electrical test point
and Sin simultaneously active. [This definition will need clarified without using the annex reference.]
3.12 echo return loss, segmental single talk (ERLSST). The echo return loss from the receive electrical
test point to the send electrical test point as measured in section 11.6.2.2. The acoustic echo
canceller is in normal operation and the mouth simulator is inactive. The algorithm features a fixed
32 ms frame size averaged over 320 ms. [This definition will need clarified without using the section
reference.]
3.13 echo return loss, temporally weighted double talk (ERLTDT). As per section 11.6.4.1 with the
receive electrical test point and mouth simulator simultaneously active. [This definition will need
clarified without using the section reference.]
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
12
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
3.14 echo return loss, temporally weighted single talk (ERLTST). The echo return loss from the receive
electrical test point to the send electrical test point as measured in annex G. The acoustic echo
canceller is in normal operation, and there is no signal coming from 50TP. The algorithm features a
perceptually based non-fixed window size. [This definition will need clarified without using the
annex reference.]
3.15 echo path response. The output at the send electrical test point due to an input at the receive
electrical test point. It is a measure of acoustic, vibration, and electrical coupling from the receive
circuit to the send circuit.
3.16 feed circuit. An arrangement for supplying DC power to a hands-free telephone set and an AC path
between the hands-free telephone and a terminating circuit.
3.17 frequency response. Electrical, acoustic, or electroacoustic sensitivity (output/input) or gain as a
function of frequency.
3.18 full duplex. An operating condition which allows simultaneous communication in both send and
receive directions with 3 dB or less switched loss in either direction. Classified as Type 1 in ITU-T
Recommendation P.340 [25].
3.19 half duplex. An operating condition which allows communication in either send and receive
directions with more than 20 dB switched loss in either direction. Classified as Type 3 in ITU-T
Recommendation P.340.
3.20 hands-free reference point (HFRP). The point on the reference axis of the mouth simulator, 50 cm
in front of the lip plane.
3.21 hands-free telephone (HFT). A device for connection to a telephone network capable of two-way
voice communication without close coupling to the user’s mouth or ear.
3.22 hands-free telephone test circuit. An assembly consisting of a hands-free telephone set(s) and
interface(s) as may be required to realize simulated partial telephone connections.
3.23 hang-over time (TH). Time from the input signal going below the threshold level up to 50% of the
complete insertion of the switched loss.
3.24 head and torso simulator (HATS). A device that accurately reproduces the sound transmission and
pick-up characteristics of the median head and torso of adult humans. See ANSI S3.36-1985 [6],
ITU-T Recommendation P.58 [22]: 1993, and IEC 959 (1990).
3.25 interface. A device placed between the line output of a digital hands-free telephone set and the test
equipment. The device performs at least one of the following functions: simulation of a normal
network connection control of the hands-free telephone, or access for the reference codec to the
digital voice signal.
3.26 Loop. See recursive.
3.27 loudspeaking telephone. A telephone with hands-free receive but not hands-free send capability.
3.28 mouth reference point (MRP). The point on the reference axis of the mouth simulator, 25 mm in
front of the lip plane.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
13
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
3.29 mouth simulator. See ITU-T Recommendation P.51 [19], “Artificial Mouth.”
3.30 near-field test point (NFTP). The acoustical measurement point located 1 cm directly above the
center of the HFT’s loudspeaker along the axis of the speaker.
3.31 noise discriminator/noise guard/noise monitor. A circuit or algorithm intended to discriminate
between speech and noise. It can affect switching, transmission, and/or noise performance.
3.32 non-linear devices. Nonlinear processing including, AGC circuits, noise guards, etc.
3.33 open listening. A mode of telephone communication in which a handset is used in the normal
position for send. The incoming signal is received simultaneously by the handset and loudspeaker.
3.34 partial duplex. An operating condition which allows simultaneous communication in both send and
receive directions with 3 to 20 dB switched loss in either direction. Classified as Type 2 in ITU-T
Recommendation P.340.
3.35 receive. The acoustic output of a hands-free telephone due to an electrical input.
3.36 receive attenuation during double talk (ARDT). Attenuation in the receive path, seen at 50TP,
inserted during double talk. The send talker initiates the double talk.
3.37 receive electrical test point (RETP). The electrical measurement point of a battery feed circuit,
reference codec, or wireless reference base station for signals applied to the hands-free telephone in
the receive direction. For further discussions on electrical interfaces, see IEEE Standard 269-1992
[8]. For further discussions on wireless interfaces, see appropriate wireless standards.
3.38 Receive Loudness Rating Directionality (RLRD). Receive loudness rating versus angle around
the HFT, normalized to the loudness rating at 50TP.
3.39 receive speech front end clipping during double talk (CRDT). The length of time that speech
undergoes syllabic clipping, as seen at 50TP, just after the onset of double talk. The receive talker
initiates the double talk.
3.40 Recommended Test Position (RTP). An acoustic test point other than 50TP, that corresponds to
the most appropriate user position for non-standard desktop and non-desktop applications. This may
be specified by the HFT manufacturer.
3.41 Recursive. See Loop.
3.42 reference codec. For the purposes of this standard, a well-defined analog-to-digital and digital-toanalog converter for testing digital telephones using analog test equipment. See Standard 269-1992
section 5 for details on a reference codec.
3.43 reference volume control setting. The volume control position resulting in a nominal receive
loudness rating.
3.44 send. The electrical output of a hands-free telephone due to an acoustic input.
3.45 send attenuation during double talk (ASDT). Attenuation in the send path, seen at the send
electrical test point, inserted during double talk. The receive talker initiates the double talk.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
14
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
3.46 send electrical test point (SETP). The electrical measurement point of a battery feed circuit,
reference codec, or wireless reference base station for signals coming from the hands-free telephone
in the send direction. For further discussions on electrical interfaces, see IEEE Standard 269-1992.
For further discussions on wireless interfaces, see appropriate wireless standards.
3.47 send front-end syllabic clipping during double talk (CSDT). The length of time that speech
undergoes syllabic clipping, as seen at the send electrical test point, just after the onset of double
talk. The send talker initiates the double talk.
3.48 send noise level. The send noise level of an HFT is measured in units of dBm, psophometrically
weighted (dBmp), according to the method described in ITU-T 0.41 (10/94). See Annex E for
details.
3.49 Send Loudness Rating Directionality (SLRD). Send loudness rating versus angles around the
HFT, normalized to the loudness rating at 50TP.
3.50 single-talk (ST). One talker speaking while the opposite transmission direction is silent.
3.51 sound pressure level. The sound pressure level, in decibels, of a sound is 20 times the logarithm to
the base 10 of the ratio of the pressure of the sound to the reference pressure. For this standard, the
reference pressure is normally 1 Pascal (Pa), and sound pressure levels are expressed in dB re 1 Pa
(dBPa). When a reference pressure of 20 uPa is used, the sound pressure level will be expressed as
dBSPL. Unless otherwise indicated, RMS values of pressure are used. Most telephony acoustic
measurements are referenced to 1 Pa. However, measurements such as receive noise and room noise
are generally referenced to 20 uPa.
Note: 0 dB Pa = 94dBSPL, 0 dBSPL = 20 microPascals, 1 Pa = 1 N/m^2. An A-weighted [2] sound
pressure level in dB (dBSPL, A-weighted) is often abbreviated as dBA.
3.52 spectrum analyzer. An instrument that measures the power of a signal in multiple frequency bands.
The frequency bands may be constant bandwidth (i.e. FFT analyzer), or constant percentage
bandwidth (real-time filter analyzer).
3.53 switching time (TS). Time taken to switch from one transmission direction to the other in
alternating single talk conversation.
3.54 take-over time (TT). Time taken to switch from one transmission direction to the other in doubletalk conversation. The signal in the first direction is continuously applied while the interrupting
signal is applied in the opposite direction. TT is measured from the application of the interrupting
signal to 50% removal of loss in that direction.
3.55 threshold level (ITH). The minimum signal level necessary for removing insertion loss.
3.56 two-wire transmission. A transmission scheme where the send and receive signals are carried in
one pair of wires.
3.57 weighted terminal coupling loss (TCLw). Long term, time averaged echo return loss, weighted in
the frequency domain per ITU-T Recommendation G.122 [15].
3.58 50 cm test point (50TP). The acoustic test point 50 cm from the HFT.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
15
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
4. Acronyms and Abbreviations
For frequency response, use the commonly accepted letter “H” as follows:
HS (f) = send frequency response (in dB V/Pa)
HR (f) = receive frequency response (in dB Pa/V)
HEP (f) = echo path frequency response (in dB V/V)
For spectra, use the letter “G.” This corresponds to common usage, especially in two-channel FFT analysis
literature. The analysis bandwidth must be specified:
GERP (f) = spectrum at Ear Reference Point (in dBPa)
GHFRP (f) = spectrum at Hands-Free Reference Point (in dBPa)
G50TP (f) = spectrum at 50 cm Test Point (in dBPa)
GMRP (f) = spectrum at Mouth Reference Point (in dBPa)
GRETP (f) = spectrum at Receive Electrical Test Point (in dBV)
GSETP (f) = spectrum at Send Electrical Test Point (in dBV)
For levels measured over a wide band, with the bandwidth to be specified, use the letter “L.” This
corresponds to common usage in sound level measurements, as specified in ANSI S1.1:
LERP = level at Ear Reference Point (in dBPa)
LHFRP = level at Hands-Free Reference Point (in dBPa)
L50TP = level at 50 cm Test Point (in dBPa)
LMRP = level at Mouth Reference Point (in dBPa)
LRETP = level at Receive Electrical Test Point (in dBV)
LSETP = level at Send Electrical Test Point (in dBV)
For specially calculated sensitivities:
SR = average receive sensitivity (in dB Pa/V)
SR0 = normalized receive sensitivity (in dB SPL/V at 1 meter)
Ss = average send sensitivity (in dB V/Pa)
5. Test Methods
5.1 General.
Various analysis techniques are available for electroacoustic measurements. Each technique has inherent
advantages and limitations. A particular method may be better suited for use with certain stimulus signals.
Certain methods, in fact, rely upon the use of a synchronized or otherwise unique stimulus signal. The
following section describes the most common techniques and their application to measurements of
loudspeaking and hands-free telephones.
5.2 Fast Fourier Transform (FFT) and Cross Spectrum Analysis.
The Fourier Transform is a mathematical operation that decomposes a time signal into its complex
frequency components. The Inverse Fourier Transform reverses the process, reconstructing the time signal
from its Fourier components. By applying the FFT algorithm to a sampled time signal, a spectrum can be
computed. This is a parallel analysis resulting in a narrow band (constant bandwidth) frequency spectrum.
Low frequency resolution may be limited. Here, blocks of time data are analyzed. Care must be taken in the
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
16
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
proper windowing of the data (i.e., Hanning, Flat-top, etc.), overlap processing, and the number of averages
to ensure an accurate analysis. The number of spectral lines and the record length determines the frequency
resolution. The frequency range and time resolution are inversely related. Because the data is discrete, the
highest frequency that can be measured is determined by the sampling frequency. Some degree of data
processing is usually available in both the time domain and in the frequency domain. An FFT analyzer may
also have a zoom capability, for increased frequency resolution across a restricted bandwidth.
5.2.1 Dual-Channel FFT.
A dual-channel FFT analyzer performs simultaneous measurements of the hands-free telephone input and
output. This type of measurement is optimized for system analysis. Most FFT analyzers calculate the
frequency response from the cross spectrum and either the input or output autospectrum. In this way,
different response estimators can be used to minimize noise at the system input or output. This also enables
other functions such as coherence, phase, group delay, coherent power and non-coherent power to be
computed. Extensive data processing is normally available in both the time domain and in the frequency
domain. It is possible to improve measurement S/N by averaging and delay compensation, however, this
method is limited in measuring harmonic distortion.
5.2.2 Single-Channel FFT.
Without cross spectrum capabilities, the system input and output must be measured separately. These
response measurements require control of the excitation spectrum and/or a two-pass analysis. The
measurement S/N due to noise at the system input or output is therefore not improved. Any post-processing
features available will apply only to the directly measured spectra, not to the response function. This
method is also limited in measuring harmonic distortion.
5.2.3 Maximum Length Sequence (MLS) Analysis.
The MLS technique4 employs a large (typically 16K), single-channel FFT and a well-defined pseudorandom pulse excitation. The length of the excitation signal is equal to the correlation length, eliminating
leakage. The MLS excitation and analysis are inherently synchronized. The received response signal is
cross-correlated with the MLS signal to obtain the time response. An FFT is then used to obtain the
frequency response. This also enables computation of coherence, phase, group delay, coherent power and
non-coherent power. Some non-linear analysis capabilities and post-processing are available. This method
does allow improvements of measurement S/N.
5.3 Real-Time Filter Analysis (RTA).
Although usually implemented digitally, real-time analysis is essentially a parallel filter bank, usually
implemented digitally. This results in a constant percentage (logarithmic) frequency resolution. The
analysis is carried out in parallel and the signal is processed continuously. The filters should be 1/12 or
1/24 octave and fulfill the ANSI S1.11-1986 [4] standard. The statistical accuracy of real-time
measurements is usually determined by the averaging time or confidence level. This type of analysis is
optimized for single-port acoustical measurements (i.e., no control of the system input).
5.3.1 Dual-Channel Real-Time Filter Analysis.
Two channels enable simultaneous measurements of the system input and output and direct computation of
the frequency response (output/input). This method does provide limited harmonic distortion capability,
and some direct post-processing of the data.
5.3.2 Single-Channel Real-Time Filter Analysis.
4 D. D. Rife and J. Vanderkooy, “Transfer-Function Measurement with Maximum Length Sequences”, J. Audio Eng. Soc., Vol. 37,
No. 6, (June 1989).
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
17
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
A single-channel real-time analyzer requires separate measurements of the system input and output.
Response measurements will require control of the excitation spectrum and/or a two-pass analysis. This
method requires the test system to be time invariant, and is limited in measuring harmonic distortion. Some
direct post-processing of the data may be available.
5.4 Sine-Based Analysis.
Sinusoidal excitation provides a high measurement S/N ratio and high degree of frequency selectivity. The
analysis is performed serially using either a quadrature or RMS detector. This often includes a tracking
filter for noise suppression and selective measurements of distortion components. The quadrature detector
multiplies the response signal by a synchronized (and appropriately delayed) sine and cosine signal. This
enables measurement of the complex, steady-state frequency response (i.e., magnitude and phase, real and
imaginary parts). Complex averaging algorithms can be employed to improve the measurement S/N ratio.
The use of an RMS detector requires a separate phase meter to obtain phase information.
5.4.1 Discrete Tone (Stepped Sine).
Discrete tone testing allows a measurement to be performed at precisely defined frequencies. These
frequencies can be at the ANSI/ISO [3,13,14] preferred numbers or in other user-defined formats. In this
case, the frequency interval (not resolution) should be stated. In addition to frequency response
measurements, intermodulation and difference frequency distortion testing are usually carried out using this
method. Phase and group delay information is also provided. These tests normally require an anechoic
room, although tone-burst techniques can be used with gating to obtain simulated free field results, and
measurement S/N can be improved using complex averaging.
5.4.2 Swept Sine.
This technique is similar to discrete tone testing, but instead employs a continuous linear or logarithmic sine
sweep excitation. The measurement is typically slow due to sweep rate limitations. Swept sine testing is
usually performed using analog instrumentation. In addition to frequency response measurements,
harmonic distortion is well suited to this method. This method requires an anechoic room, although toneburst techniques can be used with gating to obtain simulated free field results. The frequency may not be
constant throughout the analysis.
5.4.3 Time Delay Spectrometry (TDS).
TDS5 utilizes a linearly swept sine excitation signal that is synchronized to the measuring instrument. With
this signal, a one-to-one relationship between time and frequency is established and simulated free field
measurements can be performed. The measured response signal is multiplied with an appropriately delayed
version of the excitation. This, in turn, is fed to a filter and detector. The frequency resolution of a TDS
measurement is narrow band (constant bandwidth) with frequency. This is, by design, a selectable
measurement parameter. Therefore, low frequency resolution may be limited. Like other simulated free
field techniques, the effective time window determines this, i.e. the time between the arrival of the direct
sound and the arrival of the first reflection. This method is also well suited for harmonic distortion, and
provides phase, group delay, and time response information. TDS may be implemented using analog or
digital processing. In the later case, refinements and corrections for deterministic errors in the measurement
process may also be incorporated. It is possible to improve measurement S/N through complex averaging
or delay compensation. This method allows post-processing of the data.
5.5 Free Field Techniques.
Gating employs a time capture during the measurement, effectively windowing the measured response. The
frequency resolution is the reciprocal of the gating time.
5 R. Heyser, “Acoustical Measurements by Time Delay Spectrometry”, J. Audio Eng. Soc., Vol. 15, No. 10, (October 1967).
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
18
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Post-process time windowing enables the direct sound in a measurement to be separated from reflection,
producing a simulated free field condition. In this case, the frequency resolution is the reciprocal of the
applied time window. Both gating and post-process windowing can be used on measurements in ordinary
rooms.
As discussed previously, MLS and TDS are inherently simulated free field techniques. The time windowing
may be performed as a part of the data collection or as a post-processing window operation.
The frequency resolution available in an anechoic chamber is largely determined by the measurement
technique employed. Anechoic chambers are limited at low frequencies by the size of the open space
available and the depth of the absorptive material on the walls, floor and ceiling.
5.6 Method Comparative Summary.
The table below identifies various test methods described previously. The corresponding test signals and
conditions are shown for each method. The various signal classifications will be described in section 6.
Signal Type
Ref.
Anechoic
Chamber
Needed?
Deterministic
Signal
Random
Signal
Speech-Like
Signal
Compound
Signal
5.2.1
5.2.2
5.2.3
FFT/Cross Spectrum
Dual-Channel FFT
Single-Channel FFT
Max. Length Seq.
Y
Y
R
Y
Y
N
Y
Y
N
Y
Y
Y
*
Y
*
5.3.1
5.3.2
Real-Time Filter
Dual-Channel RTA
Single-Channel RTA
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
5.4.1
Sine-Based
Discrete Tone
Y
N
N
N
Y
Y
Y
Y
*
Test Method
5.4.2 Swept Sine
Y
N
N
5.4.3 Time Delay Spectr.
R
N
N
Y Test method is appropriate for this signal.
N Should not be used.
R Required signal with this test method.
* Anechoic chamber is required unless simulated free field methods are used.
Table 1 – Signal Compatibility with Test Method
Note, a “Yes” entry indicates the test method is appropriate for the specific test signal, while “No” implies it
should not be used. The “Yes/No” entry depends on the test signal used. The “Optional” entry may be
required depending on the test method, and the column entries for the chamber indicate it’s need.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
19
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
6. Test Signals
6.1 General.
The test signal must place the hands-free telephone in a well-defined reproducible state for the period of the
measurement. It must insure that the transfer function of the unit remains stable during the measurement
period, and yet provide a suitable signal for the specific measurement. The choice of the signal will be a
balance between one that correctly stimulates the detection circuits in the hands-free telephone, and one that
is suitable for the specific measurement.
6.2 Classifications.
The various types of signals are divided into several groups as discussed below. The classical measurement
signals can be separated into deterministic signals and continuous random signals. More complex random
signals include modulated random signals and speech-like signals that characterize human speech. Finally,
there are compound signals composed of two sources, one for biasing the unit into a stable state, and the
other being the actual test signal itself.
6.3 Modulation Types.
Several types of modulation may be applied to deterministic or random signals. This is done in order to
approximate the syllabic rhythm of real speech.
Test signals may be modulated in various ways to correctly stimulate a hands-free phone, depending on the
signal processing actually used in the phone. For example, a modulated noise signal is often an appropriate
stimulus for a send circuit with a noise-guard feature. In the presence of a continuous signal over a few
hundred milliseconds in duration, the noise-guard process reduces gain substantially. On the other hand, a
continuous noise signal is often an appropriate stimulus for a receive circuit with automatic gain control
(AGC).
6.3.1 Square Wave Modulation.
Square wave modulation is an on-off pattern. The recommended pattern is 250 ms ON and 150 ms OFF,
±10 ms. This pattern is common in many international hands-free testing methods. It is close to the
modulation rate of real speech. Other timing patterns may be used after confirming that the maximum
measured response has been reached.
In some cases, a periodic pulse pattern of this type will not correctly activate the telephone circuit. In such
cases, a randomly varied pulse pattern may be used. The average “on” and “off” times should approximate
250 ms and 150 ms respectively.
With this type of modulation, all measurements are to be performed during the “on” part of the pattern. For
other types of modulation, the signal is to be measured during the entire presentation time.
6.3.2 Sine Wave Modulation.
Sine wave modulation may be used to produce a simple and smooth speech amplitude envelope. The
recommended rate is 4 Hz. Modulation depth should be at least 50%, but not so great as to introduce
distortion.
6.3.3 Pseudo-Random Modulation.
Pseudo-random modulation may be used to produce a relatively speech-like amplitude envelope. The
modulation spectrum should cover from approximately 1 to 10 Hz, with the center at approximately 4 Hz.
The extremes of the modulation spectrum should be rolled off gradually.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
20
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
6.4 Deterministic Signals.
Deterministic (periodic) signals can always be used to measure the frequency response of linear, time
invariant telephones. When modulated, they can be used to measure the response of telephones with many,
but not all, speech-processing features. Deterministic signals may be used to measure linearity, some kinds
of distortion, and switching times.
6.4.1 Sine Wave.
In addition to use in measuring the frequency response of linear, time invariant telephones, sine waves are
useful for measurements of harmonic and difference-frequency distortion.
Square wave, sine wave and pseudo-random modulation are all suitable.
6.4.2 Pseudo-Random.
A pseudo-random signal has a periodic structure in the time domain. In the frequency domain, almost any
magnitude and phase spectrum is possible. When used with FFT types of analysis the period of the pseudorandom signal is to be matched in length and triggered to the analysis period. These signals require a semicomplex test procedure.
Square wave, sine wave and pseudo-random modulation are all suitable. If square wave modulation is used,
the “on” time must correspond to one complete period of the pseudo-random signal.
6.5 Random Signals.
Random signals can be described by their statistical characteristics, such as the long-term power spectral
density and probability density functions. These signals are not periodic, but are stationary as far as these
statistical characteristics are concerned. When measuring such signals, a sufficient number of averages
must be taken to obtain a given accuracy in estimating the long-term spectrum.
In practice, many practical noise generators produce pseudo-random signals, typically with a very long
period. If the period of such signals is very long compared to the analysis period, and if the analysis period
is not correlated to the generator period, then these signals can be considered random.
6.5.1 White Noise.
White noise has a constant spectral density per Hertz. The amplitude distribution is typically truncated
Gaussian, with a crest factor of 12 dB, ± 2 dB.
Square wave, sine wave and pseudo-random modulation are all suitable.
6.5.2 Pink Noise.
Pink noise has a power spectral density that decreases 3 dB per octave. The amplitude distribution is
typically truncated Gaussian, with a crest factor of 12 dB, ± 2 dB.
Square wave, sine wave and pseudo-random modulation are all suitable.
6.6 Speech-Like Signals.
Speech-like signals include ITU-T Recommendation P.50 [18] artificial voice, ITU-T Recommendation
P.59 [23] artificial conversational speech, simulated speech generator (SSG), as well as synthesized and real
speech signals. Speech-like signals are ideal for determining linearity and transfer characteristics in the
frequency domain. These signals place the hands-free telephone in a well-defined reproducible state, ensure
that the transfer function of the unit remains stable, and provide a suitable signal for the specific
measurement.
6.6.1 Simulated Speech.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
21
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Typical parameters of simulated speech include long-term average spectrum, short-term spectrum,
instantaneous amplitude distribution, speech waveform structure, and the syllabic envelope.
6.6.1.1 P.50 Artificial Voice.
ITU-T Recommendation P.50 defines the temporal and spectral parameters for a test signal which emulates
the characteristics of speech. This artificial voice is a continuous speech signal with a frequency range of
89.1Hz to 8919 Hz. Pauses may need to be inserted to emulate the on-off characteristics of conversational
speech. See section 6.6.1.2 for information on inserting pauses.
6.6.1.2 P59 Artificial Conversational Speech.
Artificial conversational speech is a test signal generated by inserting pauses in the continuous artificial
voice signal described by ITU-T Recommendation P.50. The on-off temporal characteristics of
conversational speech are defined in ITU-T Recommendation P.59. This test signal is useful for evaluating
devices that are sensitive to the on-off nature of conversational speech.
6.6.1.3 Simulated Speech Generator (SSG).
To generate a signal approximating the amplitude distribution of speech, a main signal having a Gaussian
distribution is modulated by a specially tailored modulating signal, and the resultant signal is shaped to
approximate the long-term frequency spectrum of speech. See annex A for details of this signal.
6.6.2 Synthesized Speech.
Speech-like signals may be produced using a digital processing technique rather than applying one of the
signal sources described above. Conversational speech can be sampled, digitized, processed, and
reproduced as synthesized speech. It may also be created from complex multiple tones that simulate the
talk-spurts, pauses, and activity factors associated with speech characteristics.
6.6.3 Real Speech.
Speech-like signals are not limited to signal sources or synthesized digital processing, but may also include
real speech signals. This is often done by recording conversational speech, preferably in a digital format to
avoid signal degradation with use. These real speech recordings are then reproduced using a playback
device as the signal source.
6.7 Compound Signals.
The signals described above rely on one signal source to place the hands-free telephone in a well-defined
reproducible state, insure that the transfer function of the unit remains stable, and provide a suitable signal
for the specific measurement. By applying two signal sources, one can be used specifically for “biasing”
the unit into a stable, reproducible state, while the other is the actual test signal required for measurements.
These compound signals include those where the two sources are applied in sequence, and those where both
sources are applied simultaneously.
Compound test signals can provide extra test flexibility and solve problems which are difficult or
impossible using simple test signals. The bias signal can be a signal that, by itself, is unsuitable or very
inconvenient for the actual measurement. The measurement signal can be a signal that, by itself, is
unsuitable as a bias signal.
If desired, the measurement signal can be presented so as not to have a substantial effect on the action of the
bias signal. This can be done by adjusting the temporal and/or level relationships between the two signals.
The bias signal can be changed to put the telephone in different states with minor or even no change in the
measurement signal.
6.7.1 Sequential Presentation.
This class of test signals is characterized by the separation of the bias and analysis signals in time. The bias
signal is presented until the HFT is in a stable state. Once a stable state is reached, the appropriate analysis
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
22
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
signal is applied and a measurement is performed. The analysis must be completed while the HFT is still in
its stable state. The CSS is one example of this type of signal.
6.7.1.1 Composite Source Signal.
The Composite Source Signal (CSS) is a compound signal using a voiced signal to simulate the voice
properties, followed by a deterministic signal for measuring the transfer functions, and an inserted pause to
provide amplitude modulation. The deterministic signal has either a flat or speech shaped power density
spectrum. It has the advantage of short measurement periods and duplex operation where, using an
uncorrelated double talk signal, the test signals can be applied from the talking and listening directions at
the same time. See annex B for details of this signal.
6.7.2 Simultaneous Presentation.
This class of test signals is characterized by presentation of the bias and analysis signals at the same time.
Some conditioning of the HFT may be required before beginning the analysis. The bias and analysis signals
must be separable by the analysis method. A synchronous analysis method is usually required. The P.50
Burst with Sine Sweep is one example of this type of signal.
6.7.2.1 TDS Sweep with P.50 Noise Bursts.
This compound signal has two components, which are presented at the same time, but not synchronized with
each other. The bias signal is intended to insure that the telephone is in a stable, well-defined operating
state. The measurement signal is intended to ensure a well-defined, reproducible measurement, which is
especially well adapted to simulated free-field techniques. An anechoic room is not necessary when using
this signal. See annex C for details of this signal.
6.7.2.2 TDS Sweep with Artificial Voice.
Similar to section 6.7.2.1, except the bias is a continuous artificial voice signal defined in ITU P.50 (See
section 6.6.1.1).
6.7.2.3 TDS Sweep with Real Speech
Similar to section 6.7.2.1, except the bias is real speech (see Section 6.6.3)
6.7.2.4 TDS Sweep with Random or Pseudorandom Noise
Similar to section 6.7.2.1, except the bias is white or pink random noise as described in 6.5. Pseudorandom
noise (6.4.2) with white or pink spectrum is considered equivalent if the pseudorandom period is not
correlated with the bias.
6.7.2.5 Pseudorandom Noise with P.50 Noise Bursts.
This compound signal has two components, which are presented at the same time, but not synchronized with
each other. The bias signal is intended to insure that the telephone is in a stable, well-defined operating
state. The measurement signal is intended to ensure a well-defined, reproducible measurement, which is
especially well adapted to simulated free-field techniques. An anechoic room is not necessary when using
this signal.
6.7.2.6 Pseudorandom Noise with Artificial Voice.
Similar to section 6.7.2.5, except the bias is a continuous artificial voice signal defined in ITU P.50 (See
section 6.6.1.1).
6.7.2.7 Pseudorandom Noise with Real Speech
Similar to section 6.7.2.5, except the bias is real speech (see Section 6.6.3)
6.7.2.8 Pseudorandom Noise with Random or Pseudorandom Noise
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
23
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Similar to section 6.7.2.5, except the bias is white or pink random noise as described in 6.5. Pseudorandom
noise (6.4.2) with white or pink spectrum is considered equivalent if the pseudorandom period is not
correlated with the bias.
6.7.2.9 Sine Wave with Notched Real Speech.
A sine wave signal is the measurement signal and real speech is the bias signal. A notch filter removes a
band of the speech signal at the sine wave frequency. (See annex I for details.)
6.8 Test Signal Bandwidth
In general, the test signals and analysis methods in this standard cover a frequency range of 100 Hz to
10kHz. Some signals, such as SSG (6.6.1.3), are defined only for a smaller bandwidth, and can only be
used within their defined range.
The minimum bandwidth for this standard is 175 Hz to 4.5 kHz, or the 1/3 octave bands from 200Hz
through 4 kHz.
6.9 Signal Parameter Summary.
The table below defines the bandwidth and minimal analysis interval for the various test signals identified in
section 6. These parameters would be applied to the calibration and test procedures.
Ref.
Test Signal
Usable
Bandwidth
(Hz)
Minimum
Alternative
Analysis
Analysis Interval
Interval
ISO R40 or 1/12 ISO R40 or 1/12
Oct. steps
Oct. steps
6.4.1
Sine Wave
100-10,000
6.4.2
Pseudo-Random
100-10,000
25 Hz bands
1/12 Oct. bands
6.5.1
White Noise
100-10,000
25 Hz bands
1/12 Oct. bands
6.5.2
Pink Noise
100-10,000
1/12 Oct. bands
50 Hz bands
6.6.1. 1
P.50 Artificial Voice
89.1-8,910
1/12 Oct. bands
50 Hz bands
6.6.1.2
P.59 Artificial
Conversational Speech
89.1-8,910
1/12 Oct. bands
50 Hz bands
6.6.1.3
Simulated Speech Generator
100-5,000
25 Hz bands
1/12 Oct. bands
6.6.2
Synthesized Speech
100-10,000
25 Hz bands
1/12 Oct. bands
6.6.3
Real Speech
100-10,000
25 Hz bands
1/12 Oct. bands
6.7.1.1
Composite Source Signal
100-10,000
25 Hz bands
1/12 Oct. bands
6.7.2.1
TDS Sweep with Bias
100-10,000
50 Hz Bands
1/12 Oct. bands
6.7.2.5
Pseudorandom noise with
100-10,000
50 Hz Bands
1/12 Oct. bands
Bias
(Alternates shown in parentheses)
Table 2 - Test Signal Parameters and Analysis Methods
6.10 Signal Comparative Summary.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
24
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
The table below identifies the various test signals described previously. The corresponding test methods
and conditions are shown for each signal. The various method classifications were described in section 5.
Test Signal
Sine
FFT/Cross Real-Time Free Field
Based
Spectrum
Filter
Technique
Method Method
Method
Method
Bias
Signal
Signal
Gating or
Modulate
HFT
Stability
6.4 Deterministic
6.4.1 Sine Wave
6.4.2 Pseudo-Random
Yes
No
No
Yes
No
Yes
Yes
No
Optional
Optional
Yes
Yes
Uncertain
Uncertain
6.5 Random
6.5.1 White Noise
6.5.2 Pink Noise
No
No
Yes
Yes
Yes
Yes
No
No
Optional
Optional
Yes
Yes
Uncertain
Uncertain
6.6 Speech-Like
6.6.1 Simulated Speech
6.6.2Synthesized Speech
6.6.3 Real Speech
No
No
No
Yes
Yes
Yes
Yes
Yes
Yes
No
No
No
No
No
No
Maybe
Maybe
Maybe
Yes
Yes
Yes
6.7 Compound
6.7.1 Sequential
No
Yes
No
No
Yes
Maybe
Yes
6.7.2 Simultaneous
No
Yes
Yes
Yes
Yes
Maybe
Yes
This table may be removed!!! It is somewhat redundant, and suitability of a given test signal is
discussed in appropriate sections. There are too many inter-dependencies.
Table 3 - Test Signals Comparison
Note, a “Yes” entry indicates the test signal is appropriate for the specific test method, while “No” implies it
should not be used. The “Optional” and “Maybe” entries may be required depending on test method.
7. Test Equipment, Environment, and Impairments
7.1 General.
The test equipment, environmental concerns, and impairments recommended to evaluate hands-free
telephone set transmission performance are described in this section. See IEEE Standard 269-1992 for
information concerning other test equipment needed for this standard. Different equipment will be required
depending on whether analog or digital sets are being tested and which test signals and analysis methods are
chosen. For hands-free analog telephone sets, feed circuits and test loops are required. For hands-free
digital telephone sets, interfaces and codecs are required. The test equipment required to generate and
process input and output signals for the sinusoidal method includes a variable frequency generator, a level
recording device, and a spectrum analyzer, harmonic analyzer, or distortion analyzer. For the continuous
spectrum method the required equipment includes a continuous spectrum generator, a spectrum analyzer,
and a processor (optional).
7.2 Test Equipment.
7.2.1 Measuring Microphones.
A half inch (type M) laboratory standard free-field microphone per ANSI S1.12-1967 (R1986) [5] is
required for calibrating the mouth simulator and measuring receive characteristics. The sensitivity of the
microphone should be constant within ± 0.5 dB from 100 Hz to 10,000 Hz.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
25
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
A one inch (type L) laboratory standard free-field microphone per ANSI S1.12-1967 (R1986) is required
for measuring low sound pressure levels, background noise in the test environment, and receive noise from
the HFT under test. The sensitivity of the microphone should be constant within ± 0.5 dB from 100 Hz to
5,000 Hz.
In principle, background noise in the test environment should be measured with a low noise, random field
microphone. In practice, use of a free-field microphone is acceptable provided six measurements are taken
with the microphone aimed in each direction along three mutually perpendicular axes (left/right, front/back,
up/down). The maximum of the six measurements shall be used.
7.2.2 Mouth Simulator.
The mouth simulator shall comply with the specifications given in ITU-T Recommendation P.51.
7.2.3 Head And Torso Simulator (HATS).
A head and torso simulator for measurements of hands-free telephones should conform to ITU-T
Recommendation P.58. The HATS should be equipped with both an ear simulator and a mouth simulator.
These transducers should conform to ITU-T Recommendation P.57 [21]. Electro-acoustic measurements
on hands-free telephones using HATS will differ from measurements performed using a separate
microphone and mouth simulator. This use of HATS should therefore be reported. The HATS
configuration, neck position (if variable), the use of absorptive clothing, and any other relevant parameters
should also be noted.
7.2.4 Standard Circuits for Transmission and Voice Switching Measurements.
The standard circuits for hands-free telephone transmission and voice switching measurements are the same
as those used in IEEE Standard 269-1992 for telephones. For interfacing to wireless systems, see
appropriate wireless standards.
7.2.5 Standard Circuits for Acoustic Echo Canceller Measurements.
A test hybrid is required to perform acoustic echo return loss measurements on analog HFT’s. The test
hybrid is used to cancel or remove the HFT’s hybrid reflection, making a two-wire analog HFT behave as a
virtual four-wire phone. This means that only the acoustic portion of the echo path will remain, and thus
can be measured. There are a number of ways to implement a test hybrid: the use of an echo canceller with
freezable coefficients, measurement post processing, or use of a freezable adaptive analog filter bank.
The requirements for the hybrid are:
• High impedance bridge mode for tapping off lines without affecting impedance.
• DC blocking on the input.
• Known one way delay and round trip echo path (in/out delay). Any delay in the digital hybrid or
acquisition card will impact both the echo path model delay and the synchronization of data
acquisition. The fixed echo path delay inserted by the units must be quantified.
• Noise and distortion performance 10 dB below echo detection thresholds. A full scale SNR of 80 dB is
preferred.
• High echo return loss. G.167 calls for a digital hybrid with an echo return loss of 60 dB.
• Ability to freeze the adaptive hybrid to ensure the canceller does not enhance return loss.
The test hybrid is to be trained with white noise. Fulfillment of the above requirements should be tested at a
nominal level of -16 dBV at RETP with a 600Ω load on the hybrid. Measure the long-term average level.
This can be done using ITU-T Recommendation P.56 (SV6) on the input and output. The difference should
be 60 dB and at least 50 dB. An amplifier with high gain (at least 40 dB) will be required on the echo path
to bring any residual echo up to a level that will trigger the P.56 algorithm's activity detector.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
26
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Available commercial hybrids may not meet the above criteria. However, one possible post-processing
implementation of this technique, which does meet the above specifications, is given in annex F.
7.3 Test Environment.
Electroacoustic measurements on hands-free telephones should be conducted in a test environment, which
will not affect the results beyond the intended influence of the test table and the measurement transducers.
The test environment must have a low background noise level. In addition, either the test environment or
the test method should minimize or eliminate the influence of reflections.
[Need to address simulating a “real room” environment for some measurements.]
7.3.1 Background Noise Level.
The background noise level in the test environment should not exceed the limits shown in table 4. The
overall A-Weighted noise level must not exceed 29 dBSPL [12].
Octave Band
Center Frequency (Hz)
Octave Band
Level
(dBSPL)
63
49
125
34
250
29
500
29
1000
29
2000
29
4000
29
8000
29
Table 4 - Test Room Noise Levels
7.3.2 Anechoic Chamber.
An anechoic chamber should be large enough to comfortably accommodate the test table and transducers
and HFT. Free field conditions should exist throughout the frequency range of interest. Errors due to
deviations from the inverse square law or due to the influence of reflections shall not exceed ± 1.5 dB
below 800 Hz. Errors above 800 Hz shall not exceed ± 1.0 dB.
Figure 1 shows the position of transducers relative to the location of the test table, but free field verification
is done without the test table in the chamber.
A verification of free field conditions can be conducted as follows: Using a ½ inch free field microphone
positioned along the seven axes shown in figure 1, measure at distances of 315, 400, 500, 630, 800 and
1000mm from the mouth simulator. The level should decrease 6 dB for each doubling of distance.
Deviations represent departure from ideal free field conditions [11], [12].
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
27
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
2
3
Virtual Point Corresponding
to Centerpoint of Table
(Point B)
1
4
Top View
Lip Ring
50
C
30
B
Upper Table
40
7
6
5
30
B
Upper Table Surface
40
Dimensions in cm.
Figure 1 - Axes for Determining Free Field Conditions
Note: Axes 1, 2, 3, and 4 are in horizontal plane normally occupied by table surface. Measurements of free
field sound pressure are made in absence of the table.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
28
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
The signal used for this test should have the same frequency format as the subsequent hands-free response
measurements. The tolerances specified above must be met over the entire measurement range in that
format. This verification should also be performed if a simulated free field technique is employed (see
section 7.3.3).
7.3.3 Simulated Free-Field.
Several methods exist which enable reflection-free measurements in a room with ordinary, untreated
boundaries. Tone burst gating, FFT analysis of short noise bursts, Time Delay Spectrometry (TDS),
Maximum-Length Sequence (MLS), and post-process windowing of an impulse response are among the
possibilities which can together be called “simulated free-field methods.”
All these methods create a time window, which is adjusted to be slightly shorter than the interval between
the first arriving signal and the first reflection. Although each method creates the time window differently,
the effective length of the time window is the essential parameter. The lowest frequency that can be
measured (fmin) is the inverse of the length of the time window (T). For simulated free-field results, the time
window can be set as long as desired, as long as no reflections are included. In general, the larger the room,
the longer the time window can be.
When the following conditions are met, simulated free-field methods can be used as a substitute for an
anechoic room:
(1) The signal has a suitable bias effect on the hands-free telephone, so the unit is measured in the
desired state. The simulated free-field signal may be the only test signal used, or it may be
combined with a suitable bias signal.
(2) The effective time window (T) is long enough (> 5.8 ms milliseconds) to result in a lower
frequency limit (fmin) of 175 Hz or lower. This is slightly below the lowest frequency of the 200
Hz 1/3 octave band.
Example: The measurement microphone is 50 cm from the lip ring of the mouth simulator. This direct path
corresponds to a signal delay of about 1.5 ms. The first reflection path from the mouth to the nearest
boundary is 124 cm, and another 124 cm back to the microphone, for a total of 248 cm. This corresponds
to a delay of about 7.3 ms. The difference between the direct and first reflection path is about 5.8 ms. If
the effective time window is set to this amount, the lower frequency limit (fmin) is 175 Hz.
If the test table is positioned midway between floor and ceiling, the above example could be set up in a
room of about 2.5 meters (m) ceiling height, and smallest wall-to-wall dimensions slightly larger. If the
smallest wall-to-wall dimensions are on the order of 5 m, with a ceiling height of 2.5 m, the test table should
be located on the floor. In such a case, the time window can be approximately doubled, and (fmin) halved.
The lower frequency limit (fmin) is also the frequency resolution of the measurement. In the above example,
there is effectively a data point every 175 Hz, although the curve will usually be smooth. This can result in
some smearing of detail at the lower end of the measurement range. However, the true response of many
hands-free telephones is already rather smooth at low frequencies, so this may not always be serious. A
larger untreated room can be used for better resolution.
7.3.4 Test Table.
The test table should be a hard, bare table with a surface area of at least 1 square meter. No horizontal
dimension of the table shall be less than 0.8m. The table should be flat, rigid, horizontal and thick enough
to provide a sound reflecting surface on which the HFT rests. The tabletop should be composed of a high
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
29
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
density, hard surfaced material such as hardwood, polished marine plywood or smooth plastic laminated
high-density particleboard. The table shall be used for all measurements of desktop HFT’s. It may also be
used for measurements of non-desk-top HFT’s where appropriate.
7.3.5 Test Room Characteristics.
The reverberation time (RT60) of the test room should meet ITU-T Recommendation G.167 [16]. When
averaged over the transmission bandwidth, RT60, shall be approximately 500 ms, the reverberation time in
the lowest octave shall be no more than twice this average value; the reverberation time in the highest
octave shall be not less than half this value. The volume of a typical test room shall be of the order of 50
m3 / 1500 ft3.
7.4 Impairments.
It is important to understand how a hands-free telephone performs under less than ideal conditions. For this
reason, acoustic and network impairments may be introduced into the measurement so the effect on the
performance of the HFT may be evaluated. It is beyond the scope of this standard to list all impairments,
but some are listed in the following sections.
7.4.1 Network Impairments.
Network impairments are those that can occur in the path of a telephone connection.
7.4.1.1 Test Loops.
An analog HFT should be tested with various lengths of cable or simulated cable. Recommended loop
lengths for testing North American telephones are 0 km. 2.7 km and 4.6 km (0 feet, 9000 feet and 15, 000
feet) of 26 AWG cable.
7.4.1.2 Loop Current.
Loop current may be varied to determine if there are any detrimental effects. This is especially important if
the HFT is powered from the line rather than from a local power supply.
7.4.1.3 Termination Impedance.
Network termination impedance mismatch can result in the send signal being reflected back toward the
telephone. This is especially true with analog systems but reflections can also occur with a digital system if
some portion of an end-to-end connection is analog. This reflection can effect HFT echo canceller or
switching circuits.
7.4.1.4 Network Noise.
Network circuit noise can effect the HFT in various ways. Many designs sense the nominal noise level on
the line in order to set the switching threshold. Also, network noise can affect other non linear processes
within an HFT. Network noise shall be white in spectrum, and levels measured in dBm, psophometric
weighted (dBmp).
7.4.2 Acoustic Impairments.
Acoustic Impairments often occur due to the environment in which the telephone is placed. Some examples
are given in the following sections:
7.4.2.1 Nearby Reflecting Surfaces.
The HFT should be evaluated with nearby reflecting surfaces. In particular receive signals such as dial tone
may cause a half-duplex telephone to become unstable and oscillate between the receive and send states –
this is commonly referred to as self-interruption.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
30
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
7.4.2.2 Hoth Room Noise.
Hoth noise is random acoustic noise that has a power density spectrum corresponding to that published by
Hoth. This spectrum is designed to simulate typical ambient room noise. See annex D for details.
7.4.2.3 Room Reverberation.
[John will write a contribution for this section.]
7.5 Post Processing.
If post processing is used for acoustic echo testing, the main post processing functions will include:
•
•
•
•
•
•
•
•
•
•
Calculation of segmental ERLs using 32 ms frames and 320 ms average.
Calculation of the ERLt algorithm.
Calculation of rms level with 8 ms sliding average.
Calculation of rms level with 4 ms sliding average.
Calculation of impulse response given source and output data files.
Notch and bandpass filters (see section F.4).
Calibration of Sound card in D/A and A/D.
A filter replicating the Fletcher-Munson response at 30 Phons. This can be
accurately approximated (to within a dB or two below 2500 Hz) by a first order high
pass filter with a -3 dB point of 800 Hz.
File scaling and measurement of rms level.
Activity detection as called for in the various tests.
8. Test Calibration
8.1 Measurement Bandwidth and Resolution.
The same bandwidth shall be used for calibration and measurement. The actual bandwidth or frequency
interval used shall be stated. The calibration and measurement shall be performed using the same
measurement resolution. The measurement resolution shall be stated.
8.2 Send.
8.2.1 Acoustic Test Spectrum.
The acoustic test spectrum is measured at the Mouth Reference Point (MRP). For sinusoidal test signals,
the spectrum shall be flat within ± 1 dB over the actual measurement bandwidth. The electrical input to the
mouth simulator may be equalized to meet this requirement.
For other test signals, the acoustic spectrum shall meet the target spectrum and spectrum tolerance for the
type of signal used, as defined in section 6. The default tolerance is ± 3 dB from 175 -4500 Hz (or the 1/3
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
31
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
octave bands from 200 - 4000 Hz), and +3/-5 dB elsewhere. The electrical input to the mouth simulator
may be equalized to meet this requirement.
The continuous spectrum default tolerance shall also apply to the bias and measurement parts of the
compound, parallel test signals defined in section 6.7.2.
8.2.2 Acoustic Test Level.
The standard test level for send, Nominal PMRP, is -4.7 dBPa, at the MRP. Total harmonic distortion should
be less than 5% for this test condition.
For sinusoidal test signals, the level at the MRP shall be held constant at all frequencies of test, within the
tolerance specified in section 8.2.1.
For other continuous spectrum test signals, levels shall be measured over the entire spectrum. Out-of-band
signals from 40 to 20,000 Hz shall add no more than 0.5 dBPa to this level.
8.2.3 Mouth Simulator Calibration Procedure.
Mouth simulator calibration requires measurements at both the MRP and the HFRP in a free field, without
the test table.
To calibrate at the MRP, use a ½” free-field microphone oriented at 0 degrees to the mouth axis, with the
center of the protection grid at the MRP (see figure 2). Subtract 0.6 dB from the measurement to give the
actual sound pressure at the MRP (this compensates for the fact that the acoustic center of the microphone
is slightly in front of the protection grid). This method is valid over the entire frequency range covered in
this standard6.
25 mm
Free Feild
Microphone
Lip Ring of
Mouth Simulator
Figure 2 - Calibration of Mouth Simulator at MRP
To calibrate at the HFRP, use a 1/2” free-field microphone oriented at 0 degrees to the mouth axis, as
shown in figure 3.
6 Use of a ½” pressure microphone oriented at 90 degrees to the mouth axis, with the center of the protection grid at the MRP., is
essentially equivalent up to 5000 Hz.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
32
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Lip Ring of
Mouth Simulator
Free-field
Microphone
50 cm
Figure 3 - Calibration of Mouth Simulator at HFRP
To calibrate the mouth, first measure GMRP(f) the spectrum at the MRP. Adjust the mouth equalization to
meet the target spectrum for the signal being used at a total sound pressure of -4.7 dBPa.
Next, measure LHFRP, the total sound pressure level at the HFRP. Adjust the drive to the mouth simulator,
without changing the spectrum shape, so that LHFRP = -28.7 dBPa.
Finally, remeasure GMRP(f), the calibrated spectrum at the MRP. This spectrum is used to calculate the send
frequency response. Remeasure LMRP, the total calibrated sound pressure level at the MRP, should be
measured for reference. In general, LMRP will be close to, but not exactly -4.7 dBPa.
LMRP - LHFRP = Corr, a correction factor relating to the acoustic radiation characteristics of the mouth. For a
mouth simulator exactly meeting the specifications of ITU-T Recommendation P.51, Corr = 24 dB. The
Corr value for an actual mouth simulator will differ slightly, depending on the type of mouth and the
bandwidth of the measurement.
Send stimulus levels specified in this standard are nominal levels, which differ from actual levels at the
MRP as follows:
LMRP = nominal LMRP + Corr – 24 dBPa
8.2.4 HATS Calibration Procedure
Calibration of the HATS mouth simulator is performed using the procedure described in section. 8.2.3 with
the following exceptions:
1. The spectrum at the MRP may be measured by an acoustically equivalent method if provided for by the
manufacturer7.
2. LHFRP is measured with the HATS neck stright as shown in figure 4. The neck position may be
different for measurements.
7 A 1/4 inch microphone at 90° to the mouth axis.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
33
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Equivalent Lip-Plane
50 cm
Free-field
Microphone
Figure 4 - HATS Mouth Simulator Calibration at the HFRP.
The ear simulator(s) is calibrated according the manufacturer’s recommendation.
8.3 Receive and Echo.
Electrical test signals in the receive direction shall be provided for a 900 ohm resistive source impedance.
Calibration is performed across a 900 ohm resistive calibration load. As a result, electrical test signals are
specified under nominally loaded conditions (This is equivalent to one-half the open-circuit voltage).
After calibration, the 900 ohm calibration load is removed. The source is then connected to the HFT
without further adjustment.
8.3.1 Electrical Test Spectrum.
For sinusoidal test signals, the spectrum shall be flat within ± 1 dB over the actual measurement bandwidth.
The electrical input may be equalized to meet this requirement.
For other test signals, the spectrum shall meet the target spectrum and spectrum tolerance for the type of
signal used, as defined in section 6. The default tolerance is ± 3 dB from 175 -4500 Hz (or the 1/3 octave
bands from 200 - 4000 Hz), and +3/-5 dB elsewhere. The electrical input may be equalized to meet this
requirement.
The continuous spectrum default tolerance shall also apply to the bias and measurement parts of the
compound, parallel test signals defined in section 6.7.2.
8.3.2 Electrical Test Level.
The standard test level for receive is -18 dBV. This test level is recommended for measurements at
minimum volume, and -28 dBV at the 900 ohm calibration load is recommended at maximum volume.
Total harmonic distortion should be less than 5% for these test conditions.
For sinusoidal test signals, the level shall be held constant at all frequencies of test, within the tolerance
specified in section 8.3.1.
For other continuous spectrum test signals, levels shall be measured over the entire spectrum. Out-of-band
signals from 40 to 20,000 Hz shall add no more than 0.5 dBV to this level.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
34
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
9. Transmission Measurements
9.1 General.
Procedures are given in the following sections for measuring parameters affecting the send and receive
performance of hands-free telephone sets. These parameters include frequency response, noise, inputoutput linearity, distortion, equalization, and directivity. In addition, procedures are given for measuring
electrical characteristics.
Because hands-free telephone set characteristics are affected by loop impedances, terminations, loop
currents, and operating levels, the measurements should be made under test loop and signal level conditions
representative of those conditions the hands-free telephone set is expected to encounter in use. Records
should be kept of parameters used in all measurements. The sensitivities measured should be presented as
dB Pa/V or dB V/Pa. Input levels should be reported.
All equipment should be calibrated in accordance with the manufacturer's recommendations prior to making
measurements. The location of the mouth simulator and the characteristics of the room should be such that
the level of any extraneous sound is at least 20 dB below that of the lowest signal to be measured. Proper
positioning or sound treatment should be provided to minimize reflected waves and ambient noise. In
general, anechoic chambers are required to perform the tests in this standard.
Test conditions for all transmission measurements are described in section 7 unless otherwise stated. A
range of volume control settings should be used where appropriate, but most measurements should use the
reference volume control setting.
9.1.1 Measurement Bandwidth and Resolution.
The same bandwidth shall be used for calibration and measurement. The actual bandwidth used shall be
stated. The calibration and measurement shall be performed using the same measurement resolution. The
measurement resolution shall be stated.
9.1.2 Recommended Test Methods.
The table below identifies the recommended test method(s) for each measurement parameter. Multiple
methods should be conducted to ensure the HFT is characterized at optimum performance. HFTs with
compression or VOX circuitry should be characterized with multiple methods to ensure a fully operational
state.
Frequency
Response
Noise
InputOutput
Linearity
Distort
5.2 FFT/Cross Spectrum
5.2.1 Dual-Channel FFT
5.2.2 Single-Channel FFT
5.2.3 Max. Length Seq.
Yes
Yes
Yes
No
Yes
No
Yes
Yes
Yes
No
No
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
5.3 Real-Time Filter
5.3.1 Dual-Channel RTA
5.3.2 Single-Channel RTA
Yes
Yes
No
Yes
Yes
Yes
No
No
Yes
Yes
Yes
Yes
Yes
Yes
5.4 Sine-Based
5.4.1 Discrete Tone
Yes
No
Yes
Yes
Yes
Yes
Yes
Test Methods
Loudness Mid-Band Loudness
Rating
Average
Rating
Sensitivity Directivity
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
35
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
5.4.2 Swept Sine
5.4.3 Time Delay Spectr.
5.5 Free Field Tech.
MLS & TDS
Yes
Yes
No
No
Yes
Yes
Yes
Yes
Yes
No
Yes
Maybe
Table 5 - Test Methods Comparison
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Note, a “Yes” entry indicates the test method is appropriate for the specific measurement parameter, while
“No” implies it should not be used. The “Maybe” entry depends on the free field technique used.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
36
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Frequency
Response
Noise
InputOutput
Linearity
Distort
6.4 Deterministic
6.4.1 Sine Wave
6.4.2 Pseudo-Random
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
Yes
Yes
6.5 Random
6.5.1 White Noise
6.5.2 Pink Noise
Yes
Yes
Yes
Yes
Yes
Yes
No
No
Yes
Yes
Yes
Yes
Yes
Yes
6.6 Speech-Like
6.6.1 Simulated Speech
6.6.2 Synthesized Speech
6.6.3 Real Speech
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
No
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Test Signals
6.7 Compound
6.7.1 Sequential
6.7.2 Simultaneous
Yes
Yes
Yes
Yes (1)
Yes
Yes
Yes
Yes (1)
Table 6 - Test Signals Comparison
Loudness Mid-Band Loudness
Rating
Average
Rating
Sensitivity Directivity
Note, a “Yes” entry indicates the test signal is appropriate for the specific measurement parameter, while
“No” implies it should not be used. Also, any signal may be utilized to place HFT in the proper test state
for noise measurements, and reference (1) under distortion measurements must be a sine wave analysis
signal.
9.2 Positioning of HFT and Test Transducers.
9.2.1 Desktop Hands-Free.
The hands-free telephone should be positioned 40 cm back from the front edge of the test table as shown in
figure 5A and 5B. Figure 5A shows the mouth simulator positioned for send measurements. The lip ring is
positioned 50 cm from the front center of the hands-free telephone and 30 cm above the table. The axis of
the mouth simulator is aligned with the 50 cm imaginary line.
Receive measurements are made with the artificial mouth replaced by the measuring microphone with the
center of the microphone grid at point C and the microphone axis aligned with the 50 cm imaginary line.
Hands-free telephones with separate microphone and loudspeaker housings should be positioned as shown
figure 6. Note, this physical test arrangement should be used for all measurements including frequency
response. Effects due to the presence of the table are considered a part of the measured response.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
37
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Mouth Simulator
(or Measuring Microphone)
50TP (Center of Lip Ring)
50 cm
30 cm
40 cm
Figure 5A - Standard Test Position for Hands-free Measurements - Side View
(Mouth Simulator shown for Send Measurements)
Approximately one meter
40 cm
R
M
Center Line of Table
J
M
Measuring Microphone
(or Mouth Simulator)
Figure 5B - Standard Test Position for Hands-free Measurements - Top View
(Microphone shown for Receive Measurements)
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
38
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Approximately one meter
Hands-free
Microphone
40 cm
60 cm
o
48.6
48.6
Center Line of Table
o
Measuring Microphone
(or Mouth Simulator)
40 cm
Hands-free
Loudspeaker
Figure 6 - Standard Test Position for Hands-free
with Detached Microphone and Speaker
9.2.2 Open Listening.
In open listening mode the telephone transmits via the handset microphone and receives through both the
handset receiver and the loudspeaker. The send gain is usually much less than normal hands-free since the
send microphone is at the mouth rather than at a nominal 50 cm distance. This reduction in send gain is
usually not enough to eliminate the need for gain switching or echo canceling to prevent acoustic echo.
Send measurements in open listening mode are made with the handset mounted on a test head and using
standard handset transmission test methods outlined in IEEE-269-1992 although different test signals may
be required if the telephone employs switching or acoustic echo canceling in the open listening mode.
Open listening receive loudspeaker measurements are made using the same test position and test methods
employed in hands-free measurements except the that the handset is taken off the cradle and placed out of
the way during measurement.
Echo path loss measurements during open listening are the same as for hands-free except for the positioning
of the handset. Figure 7 shows a recommended test setup for making echo path loss measurements in the
open listening mode.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
39
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
50TP (Center of Handset Receiver)
30 cm
50 cm
40 cm
Figure 7 - Standard Test Position for
Open Listening Echo Path Loss Measurements
9.2.3 Non Desk-Top Hands-Free
HFT’s designed for other than traditional table top or desktop positioning should be tested with the
appropriate user positioning in mind. This position shall be defined as the “recommended test position”
(RTP). The RTP should be obtained from the manufacturer, and should be based upon the product’s
intended use. For testing purposes, this will dictate the distance and position geometry relationship between
the HFT and the mouth simulator and microphone. Measurements performed at other distances or positions
shall be noted, and in the absence of a RTP, 50TP is recommended.
9.2.4 HATS Positioning.
When using HATS, position HFT as shown in figure 8. The head may be bent at the neck to face directly at
the HFTP if this facility is available. The intersection of the HATS lip plane and the mouth axis should be
positioned at 50TP.
50TP (center of lip plane)
50 cm
30 cm
40 cm
Figure 8 - Position of HATS, Hands-Free Telephone, and Test Table
Note: This physical test arrangement should be used for all measurements including frequency response.
Diffraction effects due to the presence of the table are considered a part of the measured response. The use
of HATS can cause a dip to appear in the measured frequency response due to reflections from the chest.
These effects are also considered a part of the measurement.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
40
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Due to the symmetry of the HATS, the measurement can be performed using only one ear simulator. The
test object, however, may not have a symmetric or centrally located receiver. Therefore, the ear used for
the measurement (right or left) should be reported. Note that the ERP does not occupy the same position as
the MRP. While this is more realistic than a simple microphone, there will be level and response
differences compared to measurements performed using a separate microphone and mouth simulator. The
increased distance to the ERP will decrease the measured response level at low frequencies. The “obstacle
effect” of the head will be part of the measured receive response. The obstacle effect causes an increase in
the measured response of approximately 8 to 12 dB around 3 kHz at the ERP.
9.3 Send.
9.3.1 General.
Send measurements consist of connecting the hands-free telephone to the appropriate interface as covered in
section 7.5. Select the desired test signal to be applied as an acoustic stimulus from the calibrated mouth
simulator, and measure the resultant signal at the Send Electrical Test Point (SETP). Measure the sound
pressure level of the mouth simulator as prescribed in section 8.2.2. Adjust the acoustic output spectrum as
defined in section 8.2.1 for sinusoidal and other signals as defined in section 6. Due to interaction with the
receive path, send measurements should be made at reference and maximum receive volume control
settings.
9.3.2 Special Instructions For Microphones.
Multiple microphones, directional microphones, beam steering microphones, microphones with gating
and/or AGC result in a unique set of measurement challenges. For testing purposes, it may ease the task if
there is prior knowledge of the amount and types of microphones incorporated in the HFT. Users who are
unfamiliar with microphone types and designs may wish to refer to Annex K.
9.3.3 Frequency Response.
The send frequency response HS(f) is given by the equation below:
(Eq. 1a)
G
(f)
H S (f) = 20 log SETP
+ Corr - 24 in dBV/Pa
G MRP (f)
where:
GSETP(f) is the RMS power spectrum at the Send Electrical Test Point (SETP)
GMRP(f) is the RMS power spectrum at MRP.
Note, if the cross-spectrum method is used, the send frequency response becomes:
(Eq. 1b)
H S (f) = 20 log
G (M RP)(SETP) (f)
G MRP (f)
+ Corr -24 in dBV/Pa
where:
G(MRP)(SETP) (f) is the cross spectrum.
9.3.4 Noise.
The hands-free telephone's microphone should be isolated from sound input and mechanical disturbances
that would cause significant error. Measure the electrical output signal at the SETP, averaging over a
minimum period of 5 seconds using the noise meter described in IEEE Standard 269-1992 section 5.15.1.
Single-channel FFT or real-time analysis may also be implemented for measuring noise spectra. However,
the overall weighted average may require some post-processing to comply with the noise meter defined in
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
41
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
IEEE 269-1992. Steps should be taken to ensure the hands-free telephone is fully operating in the send
mode.
9.3.5 Input-Output Linearity.
Measure the hands-free telephone set as described in section 9.3.3 using any of the recommended test
methods prescribed in that section. Apply acoustic input levels representing the total range that the handsfree telephone is expected to encounter in use. For a linear characteristic, the output level should follow an
input level change dB for dB.
9.3.6 Distortion.
Distortion tests for HFTs are derived from standard harmonic distortion measurement techniques; however,
a continuous sine wave signal is frequently not a suitable signal for testing HFTs. Alternative signals are
specified below, as well as a test to determine which signal is suitable for the HFT under test.
In principle, these methods can be extended to difference-frequency distortion measurements. The stimulus
consists of two sine waves (or two narrow-band pseudo-random noise signals) of equal amplitude, with the
stimulus level calculated on a power basis. Analysis is with either a notched weighting filter or by bandpass
filters (or equivalent algorithm).
Difference-frequency distortion tests may be the best way to evaluate an HFT above about 1000Hz, where
the harmonics of a single tone (or narrow-band pseudo-random noise signal) lie above the cutoff frequency
of the HFT.
9.3.6.1 Test Signal
Three types of distortion test signals are recommended. These include continuous sine waves (6.4.1),
modulated sine waves and narrow-band pseudo-random noise.
A square wave (6.3.1), sine wave (6.3.2), or a pseudo-random modulation (6.3.3) can modulate sine wave
signals.
The narrow-band pseudo-random noise (6.4.2) may be used as the default test signal for all distortion
measurements. It should have an effective bandwidth of 25 to 50Hz, and out-of-band signals should add no
more than 0.5dB to the overall level of the test signal. A period of 250ms is recommended for this signal,
since this will provide some modulation at a 4Hz rate. The crest factor should be 9 3dB.
When a narrow-band pseudo-random test signal is not suitable, modulation may be applied in a similar
manner to modulating a sine wave.
Send distortion should be measured at the standard stimulus level of –4.7dBPa, and other levels in the range
from –30dBPa to +6dBPa. Measurements should also be made over a range of frequencies within the
telephone band, such as the ISO R-10 preferred frequencies from 315Hz to ½ the upper frequency limit of
the HFT under test.
Note: Use of Acoustic Reference Level (ARL) is not recommended for send distortion measurements. The
rationale is that send sensitivity is normally not user-adjustable, and the level of the human talker lies in a
reasonably controlled range. The HFT should have low distortion regardless of the level actually sent to the
far end.
9.3.6.2 Suitability Test
To test the suitability of a particular distortion test signal, send frequency response should first be measured
at the standard test level (9.3.3). The proposed distortion test signal should then be applied at each
distortion test frequency, at the standard level, and the send frequency response measured at those
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
42
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
frequencies. If the result is within 2dB of the comparable values previously obtained in the complete send
frequency response, then the proposed distortion test signal is suitable. Distortion does not have to be
measured using the same test signal as send response.
9.3.6.3 Distortion Measurement
The output fundamental is measured at SETP, with a bandpass filter or equivalent algorithm. The distortion
is measured by use of a psophometric filter according to ITU-T recommendation O.41, but with a notch
added to eliminate the test signal. The output of the notched filter includes harmonics as well as noise.
The notch filter output is divided by the fundamental and expressed in percent. The result is signal-to-totaldistortion-and-noise.
The notch must attenuate the test signal by at least 50dB. This will result in a distortion floor of 0.3%,
permitting measurements of distortion from 1% and above with 6% or better accuracy.
The filter shall be compensated for the notch on a power basis. A constant shall be added to each point of
the notched filter frequency response, so that the power sum of all points, on a logarithmic frequency scale,
is equal to the power sum all frequency response points of the original psophometric filter.
Harmonic analysis using bandpass filters, or an equivalent algorithm, is recommended for further diagnostic
information. At each harmonic frequency, the bandpass filter output is divided by the fundamental and
expressed in percent. The result is 2nd, 3rd, etc. harmonic distortion at each test frequency.
9.3.7 Loudness Rating Applications.
The send frequency response defined in section 9.2.4 can be used directly in calculating TOLR according to
IEEE Standard 661-1992 [9]. ISO R10 format data (1/3 octave) is required for calculating SLR according
to ITU-T Recommendation P.79 [25]. Measured data can be converted using the procedures in section
9.3.8.
9.3.8 Conversion of Measured Data to ISO R 10 Format.
Measurements may be performed in various frequency formats, depending upon the analysis method
employed. Response measurements of hands-free telephones can contain numerous peaks and dips. This
conversion, therefore, should be performed using “band-averaging”. The measured points within a
particular 1/3 octave band are “power averaged” according to equation 2.
At each ISO R10 preferred frequency
(Eq. 2)
1
H ′( f ) = 10 log 10 
N
where
H ′( f )
N
∑ 10
i =1
Hi
10



= response at the new preferred ISO R10 frequency
f = preferred ISO R10 frequency
N = number of response values within the 1/3 octave band centered at f
= index for each response value within the 1/3 octave band
i
Hi = measured response value (in dB)
For the lowest frequency within the band, i = 1. For the highest included frequency, i = N. The 1/3 octave
passband limit frequencies can be calculated as:
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
43
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
(Eq. 3)
( n /10 ) ± 0.05
f = 10
where n is the Band Number.
Example: For the 100 Hz band, the Band Number = 20; For the 125 Hz band, the Band Number = 21, etc.
Table L1 in Annex L contains a list of the band numbers, 1/3 octave center frequencies, and corresponding
pass band limits.
For measured data at frequencies coinciding with a band-edge frequency (i = 1 and/or i = N), reduce the
value by 3 dB, and use that data point in both the upper and lower frequency band calculations.
For constant percentage bandwidth measurements, there will always be the same number of points for each
converted band (4 or 8, for 1/12 or 1/24 octaves, respectively). For constant bandwidth data (e.g., FFT) on
a log frequency axis, the measurement data will appear under sampled at low frequencies and over sampled
at higher frequencies.
9.3.9 Mid-Band Average Send Sensitivity.
Unlike a handset or headset, a hands-free telephone is not closely coupled to the mouth and ear during use.
Therefore, a single-number sensitivity calculation more general than loudness rating may be appropriate for
some applications. The mid-band average sensitivity is useful for estimating the electroacoustic transducer
sensitivity and/or output level. The mid-band average send sensitivity is similar to a microphone sensitivity
figure.
The mid-band average send sensitivity is calculated using equation 4:
(Eq. 4)
2500
SS =
∑H
S
(f)
f = 500
8
[dB V/Pa]
where HS(f) are the send response values (in dB V/Pa) at the ISO R10 preferred 1/3 octave frequencies from
500 Hz to 2500 Hz. For responses in other formats, the data should be converted using the method
described in section 9.3.8.
9.3.10 Send Directionality.
The directional characteristics of the HFT are useful in determining how well the device performs when the
user is not directly in front of the telephone - such as in a conference situation. The overall directional
characteristics are reported as Send Loudness Rating Directionality (SLRD). SLRD is send loudness rating
vs. Angle around the HFT, normalized to the loudness rating at 50TP.
Due to variations in the frequency response around the HFT, it is also advisable to investigate each response
at the various measurement points. For example polar plots at various frequencies can be derived from
these responses.
9.3.10.1 Measurement Procedure.
The HFT and mouth simulator are positioned as described in section 9.2, using 50TP or RTP as
appropriate. This position shall correspond to the 0 degree angle.
The send frequency response of the HFT is measured at the 0 degree angle and the loudness rating
calculated from the data. The mouth is then moved counter-clockwise (or the HFT rotated about its
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
44
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
physical center) to the second measurement angle. Take care to keep the center of the lip ring of the mouth
simulator at a constant distance from the physical center of the HFT. A second frequency response
measurement is made and the loudness rating calculated. This process is continued for the remaining
positions of interest. Twelve measurements, at 30-degree intervals, are suggested as a minimum.
9.3.10.2 Data Presentation.
Once the send loudness ratings are calculated the data may be presented in the form of a table of loudness
rating verses angle or may be presented graphically as shown in figure 9, below.
To plot the data graphically the following method is suggested:
1) Take the Loudness Rating at the 0 degree angle as a 0 dB reference.
2) Calculate the delta between loudness ratings at each of the data points and the 0 dB reference. The
more positive the delta the quieter it is.
3) Plot the deltas for each angle around the HFT.
Table 7 and figure 9 illustrated two example data sets. Example 1 is characteristic of a hands-free
telephone with an omnidirectional microphone pattern. Example 2 is more directional with the most
sensitive position being at the front of the telephone.
ANGLE
Degrees
CCW
0
30
60
90
120
150
180
210
240
240
300
330
Example 1
SEND
Delta
LOUDNESS relative to
RATING
SLR at 0
degrees
(SLR)
dB
dB
15
16.9
16.6
17.1
15.0
17.2
17.1
18.2
0.0
+1.4
+1.6
+2.1
0.0
+2.2
+2.1
+3.2
Example 2
SEND
Delta
LOUDNESS relative to
RATING
SLR at 0
degrees
(SLR)
dB
dB
16.6
18.9
18.8
25.9
27.5
24.4
26.0
28.5
0.0
+2.3
+2.2
+9.3
+10.9
+7.8
+9.4
+11.9
COMMENT
Reference Position
Example 2 Send is
11.9 dB quieter at 210
degree position.
16.2
+1.2
26.5
+9.9
17.8
+2.8
23.1
+6.5
17.1
+2.1
19.7
+3.1
16.5
+1.5
18.7
+2.1
Note: The more positive the Loudness Rating the quieter the response.
Table 7 - Send Loudness Rating Directionality Examples
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
45
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Example Plot 2
90°
Example Plot 1
60°
120°
30°
150°
0°
180°
12
9
6
3
Front
of Telephone
0 dB
dB
Delta (loss) relative
to SLR at 0 degrees
330°
210°
240°
300°
270°
Note: The HFT is loudest at 0 degrees in this example
Figure 9 - Send Loudness Rating Directionality with Two Sample Plots
9.4 Receive.
9.4.1 General.
Receive measurements are made by connecting the hands-free telephone to the appropriate interface as
covered in section 7.5. Select the desired test signal to be applied as an electrical stimulus at the Receive
Electrical Test Point (RETP), and measure the resultant signal at the 50 cm Test Point (50TP). Measure the
electrical level across a 900 ohm source as prescribed in section 8.3.2. Adjust the electric output spectrum
as defined in section 8.3.1 for sinusoidal and the other signals. These signals are defined in section 6.
9.4.2 Frequency Response.
The receive frequency response HR(f) is given by the equation below:
(Eq. 5a)
G (f)
H R (f) = 20 log 50TP
in dB Pa / V
G RETP (f)
where:
G50TP(f) is the RMS power spectrum at the 50TP
GRETP(f) is the power spectrum at the Receive Electrical Test Point (RETP)
Note: If the cross-spectrum method is used, the receive frequency response becomes:
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
46
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
(Eq. 5b)
H R (f) = 20 log
G (RETP)(50TP) (f)
G (RETP) (f)
in dB Pa / V
where:
G(RETP)(50TP) (f) is the cross spectrum.
These power spectra can be obtained by using discrete measurements of the power at each point as outlined
in sinusoidal methods, or by continuous spectrum methods. Any of the test methods described in section 5
may be used to measure frequency response. The sine-based methods are the simplest procedures to
implement, but require an anechoic chamber. Free field techniques incorporate the most complex
procedures, however, they may be conducted in an ordinary room.
Results should be reported as dB Pa/V.
9.4.3 HATS DRP to ERP Correction.
Due to the construction of the ear simulator in the HATS, resulting receive data will be referred to the ear
drum (DRP). In order to refer the receive response to the ERP, a correction must be applied to the
measured receive response. Depending upon the format of the measured data, the correction values in
annex X should be added to the measured response. Table X.1 is from ITU-T Recommendation P.58.
Table X.2 is the same correction at the ISO R40 preferred frequencies. It may be necessary to convert this
table to other formats. The receive response should always be presented referred to the ERP.
9.4.4 Noise.
The test signal chosen for the receive frequency response measurement (9.4.2), is to be applied for at least
10 seconds prior to the noise measurement. Measure the acoustical output signal at 50TP, averaging over
30 seconds, beginning 500ms after the receive test signal is removed. The measurement may be performed
directly with an A-weighting filter, or the A-weighted result may be calculated from a spectrum
measurement. The result is receive idle noise, measured in dB re 20 Pa, expressed as L50TP(A).
To test the validity of this measurement, de-activate the HFT and repeat the measurement for 5 seconds,
expressed as LROOM(A). If LROOM(A) is at least 10dB less than L50TP(A), the measured L50TP(A) is
considered valid. If not, the receive noise can only be said to be below LROOM(A).
If the above test fails, first be sure the test room and measuring microphone meet the requirements of this
standard. If very low noise levels must be measured and the noise in the test room cannot be further
reduced, an alternate procedure can be used. Reposition the test microphone at half the standard distance
from the HFT. (If 50TP is the appropriate test position, the microphone should be placed 25cm from the
HFT. Otherwise, use ½ RTP.) Repeat the measurement procedure. If the validity test passes, subtract 6dB
from the measured result and report that value.
This alternate procedure is allowable only for HFTs with one loudspeaker, and will give approximate
results. Noise measurements closer than ½ the standard test distance are not recommended.
9.4.5 Input-Output Linearity.
Measure the hands-free telephone set as described in section 9.4.2 using any of the recommended test
methods prescribed in that section. Apply electrical input levels representing the total range that the handsfree telephone is expected to encounter in use. For a linear characteristic, the output level should follow an
input level change dB for dB.
9.4.6 Distortion.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
47
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Distortion tests for HFTs are derived from standard harmonic distortion measurement techniques; however,
a continuous sine wave signal is frequently not a suitable signal for testing HFTs. Alternative signals are
specified below, as well as a test to determine which signal is suitable for the HFT under test.
In principle, these methods can be extended to difference-frequency distortion measurements. The stimulus
consists of two sine waves (or two narrow-band pseudo-random noise signals) of equal amplitude, with the
stimulus level calculated on a power basis. Analysis is with either a notched weighting filter or by bandpass
filters (or equivalent algorithm).
Difference-frequency distortion tests may be the best way to evaluate an HFT above about 1000Hz, where
the harmonics of a single tone (or narrow-band pseudo-random noise signal) lie above the cutoff frequency
of the HFT.
9.4.6.1 Test Signal
Three types of distortion test signals are recommended. These include continuous sine waves (6.4.1),
modulated sine waves and narrow-band pseudo-random noise.
A square wave (6.3.1), sine wave (6.3.2), or a pseudo-random modulation (6.3.3) can modulate sine wave
signals.
The narrow-band pseudo-random noise (6.4.2) may be used as the default test signal for all distortion
measurements. It should have an effective bandwidth of 25 to 50Hz, and out-of-band signals should add no
more than 0.5dB to the overall level of the test signal. A period of 250ms is recommended for this signal,
since this will provide some modulation at a 4Hz rate. The crest factor should be 9 3dB.
When a narrow-band pseudo-random test signal is not suitable, modulation may be applied in a similar
manner to modulating a sine wave.
Receive distortion should be measured at the standard stimulus level of –16dBV, as well as other levels in
the range from –30dBV to 0dBV. Measurements should also be made over a range of frequencies within
the telephone band, such as the ISO R-10 preferred frequencies from 315Hz to 3150Hz. Test frequencies
over ½ the upper frequency limit of the HFT under test may be useful in evaluating the loudspeaker system.
Distortion measurements should be made over a range of volume control settings, including minimum,
reference, and maximum.
Note: It may be advisable to adjust the test signal level to produce a specified sound pressure output at
50TP, similar to the method in TIA-470B for handsets. A good case could be made for this approach, since
users tend to adjust volume controls based on absolute level of the received signal.
9.4.6.2 Suitability Test
To test the suitability of a particular distortion test signal, receive frequency response should first be
measured at the standard test level (9.4.2). The proposed distortion test signal should then be applied at
each distortion test frequency, at the standard level, and the receive frequency response measured at those
frequencies. If the result is within 2dB of the comparable values previously obtained in the complete
receive frequency response, then the proposed distortion test signal is suitable. Distortion does not have to
be measured using the same test signal as receive response.
9.4.6.3 Distortion Measurement
The output fundamental is measured at 50TP, with a bandpass filter or equivalent algorithm. The distortion
is measured by use of a A-weighting filter according to ANSI S1.4, but with a notch added to eliminate the
test signal. The output of the notched filter includes harmonics as well as noise. The distortion is divided
by the fundamental and expressed in percent. The result is signal-to-total-distortion-and-noise.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
48
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
The notch must attenuate the test signal by at least 50dB. This will result in a distortion floor of 0.3%,
permitting measurements of distortion from 1% and above with 6% or better accuracy.
The filter shall be compensated for the notch on a power basis. A constant shall be added to each point of
the notched filter frequency response, so that the power sum of all points, on a logarithmic frequency scale,
is equal to the power sum all frequency response points of the original A-weighting filter.
Harmonic analysis using bandpass filters, or an equivalent algorithm, is recommended for further diagnostic
information. At each harmonic frequency, the bandpass filter output is divided by the fundamental and
expressed in percent. The result is 2nd, 3rd, etc. harmonic distortion at each test frequency.
Distortion measurements should be made over a range of volume control settings, including minimum, midway, and maximum volume.
9.4.7 Loudness Rating Applications.
The receive frequency response defined in section 9.4.2 can be used directly in calculating ROLR according
to IEEE Standard 661-1992. ISO R10 format data (1/3 octave) is required for calculating RLR according to
ITU-T Recommendation P.79. Measured data can be converted using the procedures in section 9.3.8.
9.4.7.1 Corrected Receive Loudness Rating Using a Free-Field Microphone.
Receive loudness ratings calculated according to IEEE Standard 661-1992 and ITU-T Recommendation
P.79 assumes listening with one ear, using a handset. Both ears are used to listen to a hands-free telephone,
resulting in a perceived loudness advantage compared to handset listening. Hands-free receive loudness
rating calculations must therefore be corrected to enable subjectively relevant comparison between handset
and hands-free loudness ratings.
The receive loudness rating is first calculated according to the relevant standard. Then 14 dB is subtracted
from the result giving the hands-free receive loudness rating. The effect of the correction is to make the
hands-free correction appear 14 dB louder than the uncorrected rating. The 14 dB correction is provisional,
and is under continued study.
9.4.7.2 Corrected Receiver Loudness Rating When Using HATS
When using HATS, the correction is 12 dB. Most of this correction is due to binaural versus monaural
listening. Other factors include head diffraction, pinna effects, and the position of real ears relative to
50TP.
9.4.8 Mid-Band Average Receive Sensitivity.
Unlike a handset or headset, a hands-free telephone is not closely coupled to the mouth and ear during use.
Therefore, a single-number sensitivity calculation more general than loudness rating may be appropriate for
some applications. The mid-band average sensitivity is useful for estimating the electroacoustic transducer
sensitivity and/or output level.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
49
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
The mid-band average receive sensitivity is calculated using equation 6:
2500
SR =
∑H
R
(f)
f = 500
8
[dB Pa/V]
(Eq. 6)
where HR(f) are the receive response values (in dB Pa/V) at the ISO R10 preferred 1/3 octave frequencies
from 500 Hz to 2500 Hz. For responses in other formats, the data should be converted using the method
described in section 9.3.8.
The normalized receive output sensitivity for a measurement at 50TP is:
S! RO = S R + 88
[dBSPL at 1 m for 1 Volt input]
(Eq. 7)
The normalized receive output sensitivity is similar to a sensitivity specification for a loudspeaker.
9.4.9 Receive Directionality.
As with the send condition, the receive characteristics of the HFT can vary with position around the
telephone. The overall receive directional characteristics are reported as Receive Loudness Rating
Directionality (RLRD). RLRD is a measure of the Receive Loudness Rating verses angle around the HFT.
Measurement methods and data presentation are the same as used for Send Directionality, Section 9.3.10,
with the exception that receive measurements are made with the measurement microphone at the 50TP or
RTP.
Due to variations in the receive frequency response around the HFT, it is advisable to also investigate each
response at various measurement points. For example, polar plots at various frequencies can be derived
from these responses.
9.5 Digital Only.
9.5.1 Echo Path Response.
The echo path frequency response is the RMS power spectrum at the send electrical test point (SETP)
divided by the spectrum at the receive electrical test point (RETP).
(Eq. 8a)
H EP (f) = 20 log
G SETP (f)
in dB V / V
G RETP (f)
Note, if the cross-spectrum method is used, the echo path frequency response becomes:
(Eq. 8b)
H EP (f) = 20 log
G (RETP)(SETP) (f)
G (RETP) (f)
in dB V / V
where:
G(RETP)(SETP) (f) is the cross spectrum
GRETP (f) is the input autospectrum
These power spectra can be obtained by using discrete measurements of the power at each point as outlined
in sinusoidal methods, or by continuous spectrum methods.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
50
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Results should be reported as dB V/V.
10. Voice Switching Measurements
10.1 General.
Most hands-free telephones contain voice-switched circuits whose main function is to avoid singing through
acoustic feedback. In various ways, such circuits insert a loss in either the sending or receiving direction
relative to full send or receive. Switching from one direction to the other occurs when a signal above a
certain threshold is applied from the opposite direction, or when the control circuit, taking into account the
relative levels and the nature of the signals in both directions, allows the switching. The amount of switch
loss a set employs determines the type of set as described below.
Some designs employ acoustic and or line echo cancellation, and can closely approximate full duplex
performance; however, many current echo canceller designs still need some switching to meet necessary
requirements for echo return loss (ERL). In the case of acoustic echo cancelling hands-free telephones,
voice switching implementation is similar to voice-switch only telephones, but with shallower switched loss
depth.
The following sections define the switching parameters and describe methods for testing them.
10.2 Classification.
Hands-free sets are classified into three types on the basis of duplex capability, and this classification is
determined by the attenuation range (aH) as defined below:
Type 1 has less than 3 dB of switched loss depth and can be considered “Full Duplex”.
Type 2 has between 3 and 20 dB of switched loss depth and can be considered “Partial Duplex”.
Type 3 has more than 20 dB of switched loss depth and has no duplex capability.
Measurements of voice switching characteristics may be divided into two categories:
A: Characteristics for alternate conversation, in which two parties communicate by alternating speech
(single talk) spurts without interrupting each other. In this case, it may be assumed that the voice
switch circuit returns to an idle state before being activated by an input signal in either direction.
B: Characteristics for simultaneous conversation, in which both parties may interrupt each other by
simultaneous talk (double-talk), or where speech at one end of a connection breaks through acoustic or
network noise that is present at the other end.
The first case is of fundamental importance, as its characteristics also affect simultaneous conversation
characteristics, and hands-free telephones should therefore always be checked in that respect. The second
case is a difficult environment for typical switched-loss hands-free telephones, as well as hands-free
telephones employing acoustic echo cancellation.
10.3 Switching Parameters.
There are six fundamental voice switching parameters, threshold level (ITH), build-up time (TR), hang-over
time (TH), switching time (TS), take-over time (TT) and attenuation range (aH). A suitable choice of
switching parameter values can minimize the degradation of speech quality introduced by voice switching.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
51
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Improper choice of parameter values, particularly switching times, may lead to serious clipping effects and
loss of initial or final consonants in speech.
Threshold levels should be chosen so that switching is not interrupted by random environmental noise
sources at either end of the call, and must also allow the user to move about close to the HFT. In addition,
ambient room/network noise effects on threshold should not impair performance. Ambient noise levels can
be used to improve threshold performance, as talkers tend to speak louder in a noisy environment than in a
quiet one.
Build-up time should be short enough so that the initial transient components of speech are not lost, but not
so short that insertion loss removal would be noisy.
Hang-over time should be long enough to cover average pauses in speech so that intermittent unwanted
switching does not occur before the initial talker is finished, but short enough to allow for reasonable breakin from the second talker.
Switching time from one active to active state to the other should balanced to best simulate full duplex
operation. Switching time is also dependent on both build-up time and hang-over time.
Attenuation range can be measured during the measurement of switching time from one active state to
another. The attenuation range is obtained from the difference between the maximum level at full activation
and the minimum level obtained immediately after transmission reversal. This represents the maximum
electrical return loss introduced by switched loss only in the HFT (does not include the loss due to an
acoustic echo canceller). At the reference volume control setting, this value can be used to classify the HFT
as Type 1, Type 2 or Type 3, but be aware that the HFT can, in some cases, have different attenuation
ranges depending upon the volume control setting.
Many of these parameters will be somewhat different in the case of a full or partial duplex hands-free
system (Type 1 & 2) that uses acoustic echo cancellation. If voice switching is employed along with
acoustic echo cancellation, the switching depth may be reduced, thus making measurements more difficult.
10.4 Test Conditions.
Unless stated otherwise in the following procedures, all test equipment, the test environment, impairments,
calibration and positioning are the same as for the transmission tests. Refer to Sections 7,8, and 9.2.2 for all
HFTs. Refer to Sec. 9.3.2 (HFT microphones) and 9.4.3 (HATS) when applicable.
Unless stated otherwise, a test hybrid as described in Section 7.2.5 is necessary for applying the following
tests to an analog HFT.
It is important for the terminal to function properly in the face of impairments such as acoustic noise,
network noise and echo path changes both acoustic and network in nature. It is recommended that testing
be repeated along with the application of combinations of impairment. Refer to section 7.4.
10.4.1 Signal Levels.
The send levels should be variable from -50 to 0 dBPa and the receive levels should be variable from -70 to
0 dBV.
10.4.2 Loop Lengths.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
52
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
It is advisable to test analog hands-free telephone switching parameters with a variety of loop lengths as
defined in section 7.4.2.1, or whatever the intended deployment recommends. In all cases, state the
conditions.
10.4.3 Noise Levels.
Network noise environments of -90 to -50 dBmp, as measured at the SETP, should be employed as
recommended for switching parameter tests above. Network noise is defined in section 7.4.1.4.
A range of 35 to 65 dBA Hoth acoustic noise should be used as recommended for the tests above. ITU
recommends that 50 dBA be used as a Hoth noise level. Hoth room noise is defined in section 7.4.2.2 and
can be generated as described in annex D.
For typical measurement conditions, vary the network noise from -90dBmp to -50 dBmp in 10 dB
increments. Similarly, vary the acoustic Hoth noise [see annex D] from 35 dBA to 65 dBA in 10 dB
increments. If the test chamber permits, also perform measurements in as quiet an ambient room noise
condition as possible.
For measurements that require both electrical and acoustic noise, a matrix combination of the above
network and acoustic noise levels should be performed.
The above matrix can be varied according to need depending on the type of switching parameter. It is
valuable to test threshold levels under a wider variety of conditions as compared to switching times.
10.5 Test Parameters.
10.5.1 Threshold Level.
Threshold level (ITH) is the minimum signal level necessary for removing insertion loss. Due to the unstable
hysteresis-type nature of the voice switch around threshold, the “full on” threshold is defined as the
minimum level for which the voice switch is in the stable state where the insertion loss can be completely
removed.
The envelope of the 1004 Hz periodic tone burst input signal is shown in figure 10. Adjust amplitude I1 to
determine ITH. Adjust amplitude I2 to be zero, i.e. a silent period. Adjust the on time T1 to 100 ms. Adjust
the off time T2 to be greater than the hang-over time.
For Type 2 and 3 sets, send and receive thresholds can be obtained by increasing the amplitude from a low
level until “full on” switching occurs.
The send threshold level is measured in dBPa with respect to the MRP and the receive threshold is
measured at the RETP. The reference volume control setting should be used for these tests, but other
volume control settings should also be used to characterize its effect on threshold.
For both transmission directions, measure the threshold level under quiet conditions, noisy network
conditions and noisy room conditions. See section 7.4 for basic environment conditions.
10.5.2 Build-Up Time.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
53
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Build-up time (TR) is the time from the input signal going above the threshold level up to 50% of the
complete removal of the insertion loss. Note that 50% refers to a representation of the output amplitude on
a linear scale.
The envelope of the 1004 Hz periodic tone burst input signal is shown in figure 10. For receive, it is
recommended that amplitude I1 be adjusted to -30 dBV. For send, it is recommended that amplitude I1 be
adjusted to -4.7 dBPa, amplitude I2 may be set to zero. Adjust the on time T1 to 100 ms. Adjust the off
time T2 to be greater than the hang-over time.
For both transmission directions, measure the build-up time at the reference volume control setting with no
network or acoustic impairments.
10.5.3 Hang-Over Time.
Hang-over time (TH) is the time from the input signal going below the threshold level up to 50% of the
complete insertion of the switched loss. Note that 50% refers to a representation of the output amplitude on
a linear scale.
The envelope of the 1004 Hz periodic tone burst input signal is shown in figure 10. For receive, it is
recommended that amplitude I1 be adjusted to -30 dBV. For send, it is recommended that amplitude I1 be
adjusted to -4.7 dBPa. Adjust amplitude I2 to a level just below ITH or set I2 so that the set will switch to an
idle mode. Adjust the on time T1 to 100 ms. Adjust the off time T2 to be greater than the hang-over time.
For both transmission directions, measure the hang-over time under typical line and room conditions with
the volume control in the nominal position. See section 7.6 for basic environment conditions.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
54
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Threshold
I1 (S)
Send
Input
Pressure
I (S)
ITH (S)
2
T1
T2
TH(S)
T (S)
R
Send
Output
Voltage
time
_1
2
O2(S)
O1(S)
O1(S)
O3(S)
1
_
2
O2(S)
TR(S) = measured build-up time (send)
TH(S) = measured hangover time (send)
Figure 10a - Send Build-up and Hangover Time
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
55
time
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Threshold
I1 (R)
Receive
Input
Voltage
I (R)
ITH (R)
2
T1
T2
TH(R)
T (R)
R
Receive
Output
Pressure
time
_1
2
O2(R)
O1(R)
O1(R)
O3(R)
1
_
2
time
O2(R)
TR(R) = measured build-up time (receive)
TH(R) = measured hangover time (receive)
Figure 10b - Receive Build-up and Hangover Time
10.5.4 Switching Time and Thresholds Between Two Active States.
Switching time (TS) is the time taken to switch from one active state to the other, i.e. from full send to full
receive or from full receive to full send in alternating conversation. The signal in the first direction is
removed 10 to 30 ms after application of the signal in the opposite direction. TS is measured from the
removal of signal in the first direction to 50% removal of loss in the opposite direction. 50% refers to a
representation of the output amplitude on a linear scale. Active-to-active thresholds are measured under the
same conditions, and are determined when full transmission direction reversal no longer occurs.
The following method typically applies to both Type 2 and Type 3 sets. This test method recommends that
the receiving test microphone be placed halfway between the set and mouth simulator to keep relative levels
of send and receive even, facilitating measurements. Mounting the microphone at the 50TP reference
position is recommended for break-through threshold level measurements.
For receive to send, the hands-free telephone is subjected to a periodic 1000 Hz tone burst in the receive
direction. Within a delta (∆) 10 to 30 ms before the receive burst is completed, a send burst is sent to the
set to switch the transmission direction (see figure 11a).
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
56
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
For send to receive, the hands-free telephone is subjected to a periodic 1000 Hz tone burst from the mouth
simulator. Within a delta (∆) of 10 to 30 ms before the send burst is completed, a receive burst is sent to the
set to switch the transmission direction (see figure 11b).
The overlapped time interval before the primary tone burst is completed and the secondary tone burst is sent
should be within the range of 10 to 30 ms. Send and receive tone bursts should have adjustable amplitudes
to facilitate proper switching before the timing measurement is made.
For attenuation range (aH), upon capture of the active-to-active state reversal of the transmission direction,
record the amount of loss removal for both transmit to receive, and receive to transmit, and obtain the
difference between the maximum level at full loss removal and the minimum level obtained after
transmission reversal. If the transmit to receive is not the same as receive to transmit, then the larger of the
two is the attenuation range.
The send levels should be varied from -20 to 6 dBPa and the receive levels should be varied from -55 to -10
dBV to measure break-through switching thresholds where full transmission direction change occurs. This
should be done for typical noise environments of -90to -50 dBmp network noise levels and acoustic Hoth
noise levels of 35, 45, 55, and 65 dBA (see section 7.6).
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
57
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Threshold
I1 (S)
I0 (S)
Send
Input
Pressure
time
T1
∆T
I0 (R)
Receive
Input
Voltage
time
TS (R-S)
T0
1
_
2
Send
Output
Voltage
O1 (S)
O 1 (S)
time
O0 (R)
Receive
Output
Pressure
time
T (R-S) = measured switching time (receive to send)
S
Figure 11a - Send Switching Time
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
58
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Threshold
I1 (R)
I0 (R)
Receive
Input
Voltage
time
T1
∆T
I0 (S)
Send
Input
Pressure
time
TS (S-R)
T0
1
_
2
Receive
Output
Pressure
O1 (R)
O 1 (R)
time
O0 (S)
Send
Output
Voltage
time
T (S-R) = measured switching time (send to receive)
S
Figure 11b - Receive Switching Time
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
59
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
10.5.5 Take-Over (Break-Through) Time.
Take-over time (TT) is the time taken to switch from one transmission direction to another, in double talk
conversation. It is measured from the start of an interrupting signal to the time of 50% removal of the
insertion loss in the interrupting direction. The signal in the first direction is applied continuously during
the test.
Take-over time is measured as shown in figures 12a and 12b. The test microphone is placed at the near
field test point (NFTP).
The test results may be influenced by To, the conditioning time between the start of the first signal and the
application of the interrupting signal. 200ms is the standard value for T0.
Take-over time can vary greatly depending on the relative levels in each direction. Testing over a range of
levels is recommended. Test levels in each direction and To should be repeated along with TT. TTR_S
denotes take-over time from receive to send, TT_SR the reverse.
11. Acoustic Echo Canceller Measurements
An Acoustic Echo Canceller (AEC) is a device or function that aids in improving the duplex performance of
an HFT by reducing the echo perceived by the far end. Even a superior acoustic design will suffer from
room reverberation and acoustic coupling due to the sensitivity of the microphone and the close proximity
of the microphone and loudspeaker. Unlike switched-loss echo suppression, an AEC attempts to remove
the echo signal by estimating the returned echo and subtracting the echo from the transmitted signal. An
AEC can reduce or eliminate the amount of switched-loss necessary to maintain stability of the acoustic
system, resulting in improved HFT performance. Ideally, the AEC reduces the amount of switched-loss
necessary, and results in a Type 1 HFT which allows full-duplex communication. Conversely, a poor AEC
implementation will impair HFT performance.
It is important that the HFT performs reasonably well in terms of loudness ratings, frequency response and
noise for both transmission directions. By reducing transmit sensitivity to below nominal, it is possible to
enhance echo control and full duplex operation. This design will result, however, in degraded performance
of the HFT. Send and receive transmission measurements (Section 9) should be made before proceeding
with the tests in this section.
Type 1 and Type 2 HFTs may employ switched-loss while the AEC adapts to the environment and speech
signals. The characteristics of the switched-loss operation can be evaluated using the techniques described
in Section 10. However, care must be taken to ensure the AEC does not affect the switched-loss
characteristics as they are evaluated. This can be accomplished by allowing the AEC to fully train in both
transmission directions.
The perceived quality of an AEC is a function of both the physical characteristics of the echo canceling
function and the subjective nature of human hearing. The following test method provides means to evaluate
how the AEC will be perceived as well as how the AEC operates. The use of post processing of the test
results, temporal weighting, specific test signals and real-world environment allow the method to produce
objective measures which correlate to subjective perception of performance.
This section contains procedures for evaluating the performance of an Acoustic Echo Canceller (AEC) as
implemented in a full duplex (Type 1) or partial duplex (Type 2) HFT as defined in section 3. The
measurable quality parameters included are: echo path delay (EPD), echo return loss (ERL), convergence
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
60
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
time (TC), attenuation as a function of time (A), and front end clipping time (TF). Procedures are given for
both single-talk (ST) and double-talk (DT) conditions.
Measurement of the above parameters requires specialized techniques and careful control of environment
and test signals to obtain accurate and repeatable results. Unlike Type 3 HFTs employing only voice
switching,, an AEC device typically employs non-linear processing which results in not only a time varying
device but an adaptive device. That is, the HFTs to which this section applies will respond differently
depending on the environment, the type and level of test signals applied, as well as the order and timing of
the measurement steps. An additional difference between AEC and voice switched HFTs is that the AEC
device is likely to introduce some delay between the applied signal and the transmitted echo. This delay
complicates the ERL measurements and requires some form of delay compensation to allow for accurate
measurements.
11.2 Test Signals.
Real speech (6.6.3) and a sine wave embedded in real speech (6.7.2.9) are recommended for most of the
tests in this section, and could be used for conformance once a standardized speech sample set is agreed
upon. The Composite Source Signal (CSS) (6.7.1.1) has been proposed as an alternative, though it can
provide an overly optimistic ERL if the AEC has a non-linear processor (NLP). The artificial voice
according to ITU-T recommendation P.50 (6.6.1.1) has not yet been fully investigated for use in these tests.
Random or pseudorandom noise (6.4.2 and 6.5) are recommended for EPD, while CSS is recommended for
TC. For a larger discussion on test signals see section 6.
Unless otherwise noted, receive test signals should be presented at -18 dBV at RETP, and send test signals
should be -4.7 dBPa at MRP. Since HFT performance may change with signal levels presented to it, it is
advisable to investigate the HFT using different test signal levels. For example, a range of -26 to -6 dBV is
suggested for receive, and -6 to +6 dBPa for send.
11.3 Test Conditions.
Unless stated otherwise in the following procedures, all test equipment, the test environment, impairments,
calibration and positioning are the same as for the transmission tests. Refer to Sections 7,8, and 9.2.2 for all
HFTs. Refer to Sec. 9.3.2 (HFT microphones) and 9.4.3 (HATS) when applicable.
Unless stated otherwise, a test hybrid as described in Section 7.2.5 is necessary for applying the following
tests to an analog HFT.
It is important for the terminal to function properly in the face of impairments such as acoustic noise,
network noise and echo path changes both acoustic and network in nature. It is recommended that testing
be repeated along with the application of combinations of impairment. Refer to section 7.4.
11.4 Round Trip Echo Path Delay (EPD).
Echo audibility is dependent upon the round trip delay in the echo path. This delay is measured by
determining the impulse response of the HFT. Continuous white or pink noise, random or psuedorandom,
are recommended test signals (6.4.2 and 6.5). The compound signal of Section 6.7.2.4 or 6.7.2.8 may also
be used. The HFT can be placed in nearly any reverberant or non-reverberant environment, as the first
acoustic echo will be due to direct coupling in or near the HFT.
The test signal is applied at RETP at –18dBV for 10 seconds so that the acoustic echo canceller reaches full
convergence. No signal other than the acoustic return from the loudspeaker(s) is applied to the
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
61
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
microphone(s). After 10 seconds, echo is measured at SETP, and the impulse response is then calculated
between RETP and SETP.
Echo path delay is the delay of the primary peak in the impulse response. The first peak is the start of the
impulse unless a subsequent peak is at least 10 dB greater.
11.5 Echo Return Loss (ERLST) – Single Talk.
Echo return loss from the network interface perspective is measured. Three methodologies are described, in
descending order of preference. Only one method need be conducted, with the first strongly recommended.
Real speech (6.6.3) is the recommended test signal for all three methods.
11.5.1 Echo Return Loss, Temporally Weighted – Single Talk (ERLTST).
Test signal is applied at RETP for 30 seconds so that the different functional units (in particular the acoustic
echo canceller) reach their steady states. No other signal than the acoustic return from the loudspeaker(s) is
applied to the microphone(s).
Record the electrical signals at RETP and SETP for the next 1 minute. Align the RETP and SETP
recordings in time by adding delay equal to EPD to the RETP signal. The time dependent value ERLTST is
the difference (in dB) between the signal level at RETP and SETP calculated using the algorithm in annex
H. ERLTST is named ALERLt in Annex H. Background for the algorithm is in annex G.
11.5.2 Echo Return Loss, Segmental – Single Talk (ERLSST).
Segmental ERL, ERLs, is described in the Freetel methodology 8,9. Echo and source powers are broken
into 32 ms segment averages, then echo return loss is calculated for each segment. An averaging window
ten segments long slides every 32 ms to smooth the results. It is more a measure of canceller behavior than
user reaction.
ERLs should not be calculated for segments in which the stimulus at RETP is below a threshold 10dB less
than the long term rms level of the signal at RETP, so that source inactivity is not falsely interpreted as
high echo return loss. With continuous speech, this will statistically occur less than 3% of the time (ITU-T
Recommendation P.50, figure 4). Instead, the current segment's ERLs is made equal to the previous
segment, providing smoothing.
The test method is the same as 11.5.1, except that the time-dependent value ERLSST is the difference (in dB)
between the signal level at RETP and SETP calculated as shown in annex H, except using 32 ms power
averaged segments, averaged over 10 segments. Background for the algorithm is in annex G.
11.5.3 Weighted Terminal Coupling Loss – Single Talk (TCLWST).
TCLw is defined in ITU-T G.167 and ITU-T G.122; however, TCLw is not recommended, as it does not
adequately account for audible temporal fluctuations in TCL. ITU-T SGXII reports show that segmental
TCL can vary by 10 dB or more over time.
8
Enhancements of hands-free telecommunications, Esprit Consortium, Annals of telecommunications, 49 no 7-8 1994.
9
Methodology of Evaluation and Standards, Deliverable 1.2, Freetel, July 29 1993.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
62
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
The test method is the same as 11.5.1, except that no time alignment is needed. The time independent value
TCLWST is the difference (in dB) between the 1-minute averages of the signals at RETP and SETP,
calculated as shown in G.122, trapezoidal method.
11.6 Convergence Time (Tc).
The test procedure is an extension of 11.5.2, (ERLSST), except CSS is recommended as the test signal.
Convergence measurement is an extension of the segmental ERL measurement, with further post
processing.
Test signal is applied at RETP for 90 seconds so that the different functional units (in particular the acoustic
echo canceller) reach their steady states. No other signal than the acoustic return from the loudspeaker(s) is
applied to the microphone(s).
Record the electrical signals at RETP and SETP for the entire 90 seconds. Align the RETP and SETP
recordings in time by adding delay equal to EPD to the RETP signal. Calculate the long term ERLSST over
the final 60 seconds, using the method of section 11.5.2, and averaging all 320 ms segments.
For the first 30 seconds, calculate the time dependent value ERLSST as the difference in dB between the
signal level at RETP and SETP, calculated as in section 11.5.2 using 32 ms power averaged segments,
averaged over 10 segments. However, in this case ERLSST should not be calculated for segments in which
the stimulus at RETP is below a threshold 25dB less than the long term rms level of the signal at RETP.
Determine the time required for the time dependent value ERLSST to reach within 3 dB of the long-term
average ERLSST. This is the convergence time, Tc
11.7 Echo Return Loss, Temporally Weighted – Double Talk (ERLTDT).
Real speech (6.6.3) is the recommended test signal. See further comments in Annex I, section I.2.
The canceller is trained as described in Annex I, section I.1, with the activity mask “talker active just before
onset of double talk” applied to RETP. The “talker initiating double talk” mask is applied at the mouth
simulator. Carry out the test described in Annex I, section I.5.
Record the electrical signals at RETP and SETP during the 20-second tone application. Align the RETP and
SETP recordings in time by adding delay equal to EPD to the RETP signal. The time dependent value
ERLTDT is the difference (in dB) between the test signal level at RETP and the echo signal level at SETP,
calculated using the algorithm in annex H. ERLTST is named “ALERLt” in Annex H.
11.8 Send Speech Attenuation During Double Talk (ADT_S).
Two methods are given. The first measures the attenuation vs time after entering doubletalk for a specific
frequency. The result of this method may depend greatly on the exact nature of the speech signal used,
particularly as doubletalk is begun. There may also be a dependence on the frequency of the measurement
tone, which is a sine wave embedded in the real speech creating doubletalk.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
63
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
The second method measures a long-term conversational average. The result may depend greatly on the
nature of the speech signals used, especially their temporal characteristics and degree of correlation. The
result is the send response during doubletalk, which can be compared to the single-talk response measured
in Section 9.
11.8.1 Send Speech Attenuation During Double Talk vs Time (ASTDT).
A sine wave embedded in real speech (6.7.2.9) is the recommended test signal. See further comments in
Annex I, section I.2.
The canceller is trained as described in Annex I, section I.1, with the activity mask “talker active just before
onset of double talk” applied at MRP. The “talker initiating double talk” mask is applied at RETP. Carry
out the test described in Annex I, section I.3.
Using the 8 ms sliding averaging window on the sine signal measured at SETP, the time dependent value
ASTDT is the difference (in dB) between the first 8 ms average before double talk and each 8 ms average
after double talk.
11.8.2 Send Speech Attenuation During Double Talk, Conversational Average (ASADT).
When applying this test to an analog HFT, a test hybrid is not strictly necessary. However, use of a test
hybrid will improve the signal-to-noise ratio of the measurement, resulting in lower averaging time.
TDS sweep or pseudorandom noise with real speech as the bias (6.7.2.3 or 6.7.2.7) is the recommended test
signal. See further comments in Annex I, section I.2.
The canceller is trained as described in Annex I, section I.1, with the activity mask “talker active just before
onset of double talk” applied at MRP. The “talker initiating double talk” mask is applied at RETP.
Immediately after the training period is completed, uncorrelated but similar real speech bias signals are
applied continuously in both directions at standard levels, until the measurement is completed with
sufficient averaging of the TDS sweep for a clean measurement. The TDS measurement sweep is applied in
the send direction. Send speech attenuation is the send response obtained during doubletalk divided by the
standard send response measured in Section 9 under the same conditions, except in single-talk mode. Send
attenuation may also be expressed as the increase in SLR (loudness reduction) between the single-talk and
double-talk cases.
11.9 Receive Speech Attenuation During Double Talk (ARDT)
Two methods are given. The first measures the attenuation vs time after entering doubletalk for a specific
frequency. The result of this method may depend greatly on the exact nature of the speech signal used,
particularly as doubletalk is begun. There may also be a dependence on the frequency of the measurement
tone, which is a sine wave embedded in the real speech creating doubletalk.
The second method measures a long-term conversational average. The result may depend greatly on the
nature of the speech signals used, especially their temporal characteristics and degree of correlation. The
result is the receive response during doubletalk, which can be compared to the single-talk response
measured in Section 9.
11.9.1 Receive Speech Attenuation During Double Talk vs Time (ARTDT)
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
64
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
A sine wave embedded in real speech (6.7.2.9) is the recommended test signal. See further comments in
Annex I, section I.2.
The canceller is trained as described in Annex I, section I.1, with the activity mask “talker active just before
onset of double talk” applied at RETP. The “talker initiating double talk” mask is applied at MRP. Carry
out the test described in Annex I, section I.3, substituting receive for send and vice-versa. Therefore,
receive and send signals are swapped and results monitored at 50TP.
Using the 8 ms sliding averaging window on the sine signal measured at 50TP, the time dependent value
ARTDT is the difference (in dB) between the first 8 ms average before double talk and each 8 ms average
after double talk.
11.9.2 Receive Speech Attenuation During Double Talk, Conversational Average (ARADT)
When applying this test to an analog HFT, a test hybrid is not strictly necessary. However, use of a test
hybrid will improve the signal-to-noise ratio of the measurement, resulting in lower averaging time.
TDS sweep with real speech or pseudorandom noise as the bias (6.7.2.3 or 6.7.2.7) is the recommended test
signal. See further comments in Annex I, section I.2.
The canceller is trained as described in Annex I, section I.1, with the activity mask “talker active just before
onset of double talk” applied at RETP. The “talker initiating double talk” mask is applied at MRP.
Immediately after the training period is completed, uncorrelated but similar real speech bias signals are
applied continuously in both directions at standard levels, until the measurement is completed with
sufficient averaging of the TDS sweep for a clean measurement. The TDS measurement sweep is applied in
the receive direction. Receive speech attenuation is the receive response obtained during doubletalk
divided by the standard receive response measured in Section 9 under the same conditions, except in singletalk mode. Receive attenuation may also be expressed as the increase in RLR (loudness reduction) between
the single-talk and double-talk cases.
11.10 Send Speech Front End Clipping Time During Double Talk (TSFDT).
A sine wave embedded in real speech (6.7.2.9) is the recommended test signal. See further comments in
Annex I, section I.2.
The canceller is trained as described in Annex I, section I.1, with the activity mask “talker active just before
onset of double talk” applied at RETP. The “talker initiating double talk” mask is applied at MRP. Carry
out the test described in Annex I, section I.4.
Using the 4 ms sliding averaging window on SETP, measure the rms level of the tone. Determine the time
at which the tone level rises to within 3 dB of its average level at 200 ms into double talk. Determine the
difference in time between this point and the cessation of the tone as seen at SETP. The difference between
200 ms and this length of time is the send speech front end clipping time during double talk, (TSFDT ):
If you have managed to read this far, you have won a vacation for two in Maui, compliments of Kruger and
Associates Inc. 37 Somerset Dr., Commack, N.Y. 11725-1636, phone (516) 543-5392. This is subject to
completion of a skill testing question administered by Kruger and Associates Inc.
11.11 Receive Speech Front End Clipping Time During Double Talk (TRFDT)
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
65
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
A sine wave embedded in real speech (6.7.2.9) is the recommended test signal. See further comments in
Annex I, section I.2.
The canceller is trained as described in Annex I, section I.1, with the activity mask “talker active just before
onset of double talk” applied at MRP. The “talker initiating double talk” mask is applied at RETP. Carry
out the test described in Annex I, section I.4, substituting receive for send and vice-versa. Therefore,
receive and send signals are swapped and results monitored at receive output.
Using the 4 ms sliding averaging window on 50TP, measure the rms level of the tone. Determine the time
at which the tone level rises to within 3 dB of its average level at 200 ms into double talk. Determine the
difference in time between this point and the cessation of the tone as seen at SETP. The difference between
200 ms and this length of time is the receive speech front end clipping time during double talk, (TRFDT):
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
66
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Annex A
Simulated Speech Generator
Main Signal The main signal consists of eight 1024-point pseudo-random noise segments. Each segment has the same
magnitude spectrum but a different phase spectrum with the phase randomized within and between the
segments uniformly from 0 to 360 degrees, in order to randomize the interaction between the
intermodulation products of the harmonically related spectral components. The duration of each segment is
80 ms and they are merged with each other through a raised cosine window with an additional 80 ms
merging segment between them. The simultaneous fade-out of the previous segment and the fade-in of the
following segment eliminate the transients, which would occur at the segment boundaries. The complete
main signal thus consists of eight pseudo-random segments interleaved with eight merging segments, each
of 80 ms duration having a total length of 1.28 seconds. A simple filter at the output provides the desired
frequency shaping to approximate an average speech spectrum.
Modulating Signal Measurements show that a Gamma distribution with parameter m = 0.545 provides a good approximation to
the instantaneous amplitude distribution of continuous speech. The syllabic characteristics can be
represented by a low pass response that is practically flat up to about 4 Hz (the -3 dB point) followed by -6
dB per octave roll-off.
The final wave shape of the modulating signal was derived empirically from the Gamma distribution.
Varying the period of this pulse in a pseudo-random manner and adjusting its rise and fall time ratio results
in a satisfactory approximation to the spectrum of the modulation envelope of real speech.
Combined Signal In order to extend the repetition time of the final signal and to spread more evenly the maxima of the
modulating signal over the repeated sequence of the Gaussian signal, the ratio between the sampling clock
frequencies of both signals was chosen to be 4/255. Thus the clocking frequency of the main signal is
12,800 Hz, and the clock frequency for the modulating signal is about 200.8 Hz. The repetition times are:
1.28 seconds for the Gaussian signal, 10.2 seconds for the modulating signal and 326.4 seconds for the final
modulated signal.
Main
Signal
Source
(Gaussian)
Shaping
Filter
Output
Modulating
Signal
Source
(Gamma)
Figure A - Block Diagram of Simulated Speech Simulator
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
67
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Gaussian Signal Generator The Gaussian signal is made up of sixteen segments. The odd number segments are generated by filling a 2
by n array with zeros and then filling in the desired real and imaginary spectrum components using
equations one and two. The first entry is zero i.e. no DC component and there are no components above
5500 Hz.
X r ( ω ) = cos ( 2 π α − π )
(Eq. A1)
X i ( ω ) = sin ( 2 π α − π )
(Eq. A2)
where:
α is a random number with uniform distribution 0 ≤ α ≥ 1
The inverse FFT is then taken to transfer the signal to the time domain.
xi ( n ) ⇔ X i( ω ) X r ( ω )
The even number segments S(n) are:
Si(n) =Si(n-1)*0.5(1+cos((π (i-0.5)/1024) + Si(n+1)*0.5(1-cos((π (i-0.5)/1024)
i = 1 to 1024
n = 2, 4...,16 for n+1>16 use n+1-16
(Eq. A3)
Gamma Function For the Gamma function the 2048 samples are divided into 21 random length pulse periods (number of
samples). The periods are 167, 43, 63, 119, 48, 57, 78, 88, 93, 107, 51, 71, 259, 60,67, 207, 143, 54, 130,
45, 98. Each period is divided into rise time of one third and a fall time of two thirds, i.e. rise and fall times
are in 1:2 ratio.
The cubic interpolating spline function is used to model the rising and falling section of each segment.
First calculate10 the coefficients B(I), C(I), D(I) for I =1 to 60 for a cubic interpolating spline. The number
of points (knots) is 60. The abscissas of the knots, in increasing order, range in value from 0.05648176 to
0.983219. Y is the ordinate of the knots. Y (I) =I-0.5.
where:
n = number of samples in the rising (or falling) section
s(i) is the value of the ith data point in the period
For the rising time period:
s(i) = spline value at abscissa (-0.5/n)+(1/n*i)
For the falling time:
s(i) = spline value at abscissa (-0.5/n)+(1/n*(n+1-i)
10 Refer to G.E. Forsythe, M.A. Malcolm, C.B. Moler, “Computer Methods for Mathematical Computations”, Prentice Hall, Inc.
1977 for additional information.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
68
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Annex B
Composite Source Signal
The composite source signal consists of the following components:
ITU-T Recommendation P.50 Artificial Voice Component The voiced signal part of the CSS is the conditioning signal intended to activate possible speech detectors in
hands-free telephone. The reason a voiced signal has been chosen is that presumably all devices designed
for speech transmission will quickly respond to a voiced sound. The signal is to activate the hands-free
telephone for the direction of transmission to be measured. As the duration, beginning and end of the
voiced signal are known exactly, this signal can also be used to measure the switching time. By means of
the signal shape in the time domain, the switching time and delay time of the entire system can be
determined according to ITU-T Recommendation P.34 [17]. The duration of the signal amounts to 50 ms.
Pseudo Noise Signal Component The measurement signal is the pseudo noise signal presented after the voiced artificial speech sound. The
signal has certain noise-like features. The magnitude of its Fourier transform is constant with frequency
while the phase is changing. For measurements, only the magnitude of the transfer function is of interest.
The phase is not that important but can be determined as well.
The signal is a complex spectrum produced in the frequency domain according to the following equation:
H ( k ) = W ( k )∗ e j∗ik ∗Π
where:
(Eq. B1)
k = − M 2 ,..., M 2, without 0
i k ∈ {+1, 0}, i k = −i − k , random
The index M is adjusted to the chosen FFT size (i.e. 2048 points). The equation shows that the amount of
the produced complex spectrum is constant for all frequencies if W(k) is chosen equal to 1 for all
frequencies, whereas the phase may be + or - for each frequency, corresponding to a random sequence.
However, to produce a different weighting in the frequency domain, W(k) can easily be adjusted in order to
produce different spectra for the duration of the PN-sequence. Then, the spectrum is transformed into the
time domain by means of the inverse Fourier transform producing the following signal:
S ( n) =
where:
1
M
M 2
∑ H ( k )∗ e
j 2 Π∗n∗ k M
(Eq. B2)
k = M 2k ≠0
n = − M 2 ,..., M 2
i k ∈ {1, 0}, random
Thus, a signal is produced which is limited in time (corresponding to the chosen length of the Fourier
transform) and which is adjusted to the chosen FFT size correctly. If a longer time sequence is wanted, the
signal can be cycled. This method permits time sequences of any length.
The duration of this measurement signal may amount to about 200 ms by appropriate choice of M and the
sampling rate.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
69
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Pause Component The pause has two purposes. An initial pause before applying any measurement signal is necessary to put
hands-free telephones with time-variant transfer functions into a defined initial state. To this end, the pause
should be as long as possible (>1 s). If, however, the unit is to be put into a constantly activated state
(running speech-like), the intermediate pauses should be shorter (about 100 ms) to provide suitable
amplitude modulation to the composite signal. The pause of the composite source sequence is 150 ms.
Use of Composite Source Signal for Double Talk Measurements By choosing the voiced part of the CSS with a different pitch frequency than the single talk signal, and by
using random noise rather pseudo-noise, a double talk signal can be created that has a low correlation to the
single talk signal. Refer to ITU-T Recommendation P.501 [27] for additional information.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
70
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Annex C
ITU-T Recommendation P.50 Noise Bursts Over TDS Sweep
The bias signal consists of random noise with a spectrum and spectrum tolerance conforming to ITU-T
Recommendation P.50. For send measurements, it is presented in bursts at a 4 Hz rate and 50% duty cycle
(125 ms "ON", 125 ms "OFF"). The bias is presented at the standard test level during the "ON" bursts.
For receive measurements, the bias may be presented either continuously or in the burst pattern.
Continuous presentation may be the most appropriate bias of a telephone with a simple AGC function, but
burst presentation may be better for telephones with more complex functions. Ideally, both ways should be
measured to determine which gives the most typical results. The telephone will be measured in its average
state during the entire measurement.
The measurement signal is a series of sine sweeps from 100 to 10,000 Hz, at any rate suitable for Time
Delay Spectrometry (TDS) measurements. The sweeps are not synchronized with the bias pulses. The
sweep spectrum is the P.50 spectrum times a 3 dB-per-octave rising characteristic. (This is equivalent to
passing a flat sweep through a filter designed to shape pink noise into the P.50 spectrum.) At 315 Hz, the
level of the measurement signal is 15 dB below the overall level of the bias signal.
The measurement is performed by TDS. The sweep length and number of averages are adjusted to obtain a
satisfactory signal-to-noise ratio in the measurement. Typically, a measurement time (sweep length times
number of averages) in the range of 16 to 128 seconds gives good results.
The true frequency resolution of the TDS measurement will be determined by the time window chosen, not
by the frequency interval in the analyzer. The minimum effective time window is 5.7 ms, which
corresponds to a frequency resolution (lowest measurable frequency) of 175 Hz (see section ??? ).
In principle, this method can be used with any desired bias signal, including any of the speech-like signals
described in section 6.6.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
71
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Annex D
Hoth Room Noise
Hoth noise can be described as acoustic random noise that has a power density spectrum corresponding to
that published by Hoth1. The spectrum of Hoth noise is designed to simulate typical ambient room noise
over time.
Test Table
1m
Loudspeaker
50cm
Test
Figure D1 - Hoth Noise Test Setup
Hoth noise can be reproduced using two non-correlated white noise generators and two equalizers in order
to produce the required spectrum through four loudspeakers positioned radially 50 cm above the table, 1
meter away from HFRP and 45 degrees apart (see figure D1). Each one of the two uncorrelated noise
signals are delivered to two loudspeakers in alternated fashion.
Using a free field microphone placed at the HFRP in absence of the test table, the 1/3 octave spectrum can
be calibrated. Once the spectrum is within ± 2 dB in each band of the Hoth specification, replace the table
and hands-free telephone to the correct position. The overall A weighted level can now be set with a probe
microphone located close to the microphone on the telephone, with the probe microphone configured to
measure dBSPL with A weighting (dBA).
Table D [CCITT Series P 1989, Supplement No. 13] below gives the spectrum density adjusted in level to
produce a reading of 50 dBA on an IEC recommended sound level meter. Figure D2 shows a plot of this
spectrum. The spectrum below is independent of level and shifts equally for each 1/3 octave band.
Frequency (Hz)
Spectrum Density
(dB SPL/Hz)
Bandwidth 10 log Total power in each
_ƒ (dB)
1/3 Octave Band
(dBSPL)
Tolerance (dB)
1 HOTH (D. F.): Room noise spectra at subscribers' telephone locations, J.A.S.A., Vol. 12, PP.499-504, April 1941
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
72
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
100
125
160
200
250
315
400
500
630
800
1000
1250
1600
2000
2500
3150
4000
5000
6300
8000
32.4
30.9
29.1
27.6
26.0
24.4
22.7
21.1
19.5
17.8
16.2
14.6
12.9
11.3
9.6
7.8
5.4
2.6
-1.3
-6.6
13.5
14.7
15.7
16.5
17.6
18.7
19.7
20.6
21.7
22.7
23.5
24.7
25.7
26.5
27.6
28.7
29.7
30.6
31.7
32.7
Table D - Hoth Noise Parameters
45.9
45.5
44.9
44.1
43.6
43.1
42.3
41.7
41.2
40.4
39.7
39.3
38.7
37.8
37.2
36.5
34.8
33.2
30.4
26.0
±3
±3
±3
±3
±3
±3
±3
±3
±3
±3
±3
±3
±3
±3
±3
±3
±3
±3
±3
±3
Spectrum Density Vs Frequency
35
30
Spectrum Density
25
20
15
10
5
0
-5
-10
100
1000
10000
Frequency (Hz)
Figure D2 - Hoth Noise Spectrum
Typical Hoth noise levels range from 35 dBA to 65 dBA, and switching parameter and speech detection
tests should be performed at these levels in 10 dB increments.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
73
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Note that at low frequencies, sound levels are somewhat difficult to control due to both the size of the test
chamber (attenuates poorly), and the introduction of external noise (air-conditioning/heating etc.). The test
chamber should be designed to minimize undesirable low frequency sound levels.
For optimum ambient noise simulation in the test chamber, it is best to have a diffuse source for Hoth noise.
This can best be achieved by having somewhat reflective walls, and multiple sound sources. A compromise
can be made with either the room or the number of sound sources.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
74
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Annex E
Useful Conversion Procedures
E.1 Conversions for dBV to dBm for 600 and 900 Ω
0 dBm is accepted as 1 mW, typically using a circuit impedance of 600 ohms or 900 ohms.
0 dBm = 10 log 1(mW)
dBV = 10 log V2
= 20 log V
For R = 600 ohms:
P = V2/R, therefore
dBm = 10 log V2/R * 1000
= 10 log V2/600 * 1000
= 10 log V2/0.600
So, V = 774.6 mV or 0 dBm = -2.22 dBV
For R = 900 ohms:
P = V2/R, therefore
dBm = 10 log V2/R * 1000
= 10 log V2/900 * 1000
= 10 log V2/0.900
So, V = 948.7 mV or 0 dBm = -0.46 dBV
To change from 600 ohms to 900 ohms or vice versa, for a constant voltage:
Correction (dB) = -10 log (0.600/0.900) = 10 log (0.900/0.600) = 1.76 dB
Correction (dB) = 10 log (|Z1| / |Z2|), i.e., the log of the ratio of the magnitude of the impedances, when
converting from impedance Z1 to Z2.
If converting from "Z1 = 600 ohms" to "Z2 = 900 ohms", the correction factor is -1.76 dB, thus we subtract
1.76 dB from the measurement.
Depending on the impedance being used, conversion factors can be applied dB for dB to the measured or
calculated result. Example 1: To convert a 600 ohm -20 dBm signal to dBV, simply subtract 2.22 to get 22.2 dBV. Example 2: -20 dBm is measured across 600 ohms. To find the level across 900 ohms, add a
correction of -1.76 dB to get -21.76 dBm (since the larger load dissipates less power).
E.2 Conversions for dBmp to dBrnC for Electrical Noise Measurements
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
75
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Two weighted noise measurement units have typically been used in telephony, dBmp and dBrnC. The main
differences between these two measurement units are the shape of the weighting filter and the reference
unit. The weighting filter for dBrnC is described in IEEE 743.
The differences in the weighting functions are extremely slight, as to be insignificant; thus the conversion
between the two units can be expressed as:
–
dBrnC = dBmp + 90
E.3 Loudness Rating Conversions
Conversion from IEEE 661 to ITU-T P.79 as specified by EIA/TIA-579A is as follows:
SLR (P.79) = TOLR (IEEE 661) + 57
RLR (P.79) = ROLR (IEEE 661) - 51
STMR (P.79) = SOLR (IEEE 661) + 9
The above conversions should be used as an approximation only. These conversions are based upon
approximated frequency response curves as specified in TIA-579A. Proper conversion may depend upon
actual measurements being made with each measurement standard where frequency responses deviate
significantly from the norm.
E.4 Acoustic Sound Pressure Conventions
dBPa (dB Pascals)
dBSPL (dB Sound Pressure Level)
Where,
0 dBPa = 94dBSPL, and 0 dBSPL = 20 microPascals, 1 Pa = 1 N/m2
An A weighted sound pressure level in dB (dBSPL, A weighted) is often abbreviated to “dBA”.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
76
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Annex F
Recommended Test Bed
F.1 Example Implemented Test Hybrid
The test hybrid can be constructed from a passive 2-4 wire hybrid, mated to a digital echo canceller as
described in figure F1 below. The digital echo canceller is described below.
P as sive two - f o ur wir e hybr id
L ine o ut
Tip R in g
inte rf ace &
D C Fe e d
S o und c ar d
with P r o g.
D SP
L ine in
P os t P roc es sing o n P C
f or N L M S E c ho
C anc e lle r
H FT
Figure F1
The passive 2-4-wire hybrid is a transformer setup that can be balanced to provide a simulated 4-wire signal
path that is then interfaced to the digital echo canceller. Figure F3 below describes the passive hybrid
circuit.
The digital echo canceller can be implemented through a post processing or real time application. In either
case, simultaneous (full duplex) record-playback capability is needed. The algorithm is a “normalized least
mean square” (NLMS), that has 249 regular coefficients and 1 DC coefficient, which cancels any possible
DC component in the incoming signal.
F.2 Example Adaptive Echo Canceller
An NLMS (normalized least mean square) adaptive echo canceller is used to model the actual echo path
formed by the 4 to 2 wire hybrid and the HFT. The block diagram is shown in figure F2.
x (n)
D/A
Echo
canceller
Sound
card
zz-D-D
e(n)
DU T
A/D
y (n)
Figure F2 - Block diagram of 2-to-4 wire conversion
The adaptive echo canceller has 249 regular coefficients and 1 DC coefficient, which cancels any possible
DC component in y(n). The delay of D samples inserted in the y(n) path, named as delay in the program, is:
1.
2.
a delay of 20 samples to assure that the adaptive filter can correctly model the echo path, minus
the measured delay jointly caused by the sound card, the 4-to-2 wire hybrid, and the HFT. This
part is about 15 samples with the current implementation, resulting in D=5.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
77
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
The convolution and subtraction are performed as per the following equations
sum (n ) =
N _ LMS − 2
∑ h24 (n, j ) ⋅ x (n − j ) + h 24 (n, N _ LMS − 1)
j=0
e ( n ) = y (n ) − sum ( n )
where N_LMS=250 is the total number of tap coefficients, h24 is an array containing the coefficients.
The adaptation step size µis determined by
1

Ex ( n ) =  1 −

N _ LMS
 ⋅ Ex ( n − 1 ) + 2 ( n )
x

α
Ex ( n ) + ε
µ =
In steady state, Ex(n) is an estimate of the energy in x(n), with a scaling factor N_LMS. α =0.05 is the
normalized step size, and ε =0.1 is a small constant to put an upper limit to µ.
The coefficients are then updated according to
β = µ ⋅ e (n )
h 24 ( n + 1, j ) = h 24 (n, j ) + β ⋅ x ( n − j ),
j = 0, 1, ..., N _ LMS − 2
h 24 ( n + 1, N _ LMS − 1 ) = h24 (n, N _ LMS − 1 ) +
α ⋅ e (n )
N _ LMS + ε
The training signal x(n) is white, with a probability distribution function
p ( x )
1

 2 M


0

=
≤
,
x
,
otherwise
M
where M is the maximum magnitude. Such a uniformly distributed x(n) is obviously zero-mean, with a peakto-RMS ratio.
Peak
≡
RMS
M
∫
∞
−∞
x p(x)dx
2
=
M
1 M 2
x dx
2M ∫−M
=
M
1 3 x= M
x
3M x=0
= 3
In order to estimate the performance of the 2-to-4 wire conversion, we need to calculate its echo return loss
(ERL). the formula used is
∑x (n)
∑e (n)− ∑ν (n)
2
ERL=10⋅ log10
n
2
n
2
n
where ν (n) is the noise present in e(n) and is caused mainly by the HFT and also by the 4-to-2 wire hybrid.
By using the above formula, we assume the two components of e(n), the residual echo and the noise, to be
uncorrelated.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
78
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
The timing for the application of the training signal x (n), for updating coefficients, and for estimation of
echo+noise e(n), and estimation of noise ν (n), is shown in figure F3.
T rain in g
s ig n al x (n )
On
C o efficien ts
U p d ated
C alcu latin g
en erg y in
x (n ) an d e(n )
O ff
C alcu latin g
en erg y in
x (n )
O ff
F ro zen
On
O ff
O ff
0
On
9
11
13
s
Figure F3 - Timing diagram
The adaptive filter converges in the first 9 seconds, after which it is frozen. Then the training signal stays on
for two more seconds in order to get an estimate of the energy in x(n) and e(n) between 9 and 11 seconds.
During the period of 11 to 13 seconds, the training signal is absent so that an estimate of the energy in noise
ν (n) can be obtained. These three energy estimates are then used to calculate the ERL as per the above
formula.
Cross-correlation calculation:
For this implementation of the echo canceller, it is recommended to compensate for any built-in delay that
may be present in the hardware being used. A cross-correlation calculation may be used to identify this
delay.
The cross-correlation calculation is performed between x(n) and y(n-Dsc), where Dsc is the base value of the
sound card delay, which has been obtained in the sound card setup stage.
11 cross-correlation values are calculated as follows:
Rxy ( m) =
3999
∑ x (n + m) ⋅ y(n − D ) ,
sc
m = 0,±1,±2,...,±5
n= 0
If the magnitude of a certain Rxy(m) is larger than that of any other cross-correlation value, m is deemed to
be the variation of sound card delay.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
79
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Sample C source code for NLMS Adaptive Echo Canceller:
/* Heading */
define
define
define
N_LMS
250
alfa
0.05
epsilon 0.1
/* Variable declaration */
int
double
double
j;
beta,mu,mu_for_DC,ex,sum;
y,e,beta,h24[N_LMS],x[N_LMS];
/* Initialization (executed once on start-up only) */
ex=32768.0*N_LMS;
beta=1.0-1.0/(double)N_LMS;
for (j=0; j<N_LMS; j++) {
h24[j]=0; x[j]=0;
}
mu_for_DC= alfa/((double)N_LMS+epsilon);
/* Convolution (executed once every sampling interval before and */
/* after coefficients are frozen) */
/* Acquire input data */
y=...;
/* Get y(n), already delayed by D samples */
x[0]=...;
/* Get x(n) */
/* Do convolution to obtain sum(n) */
sum=h24[N_LMS-1];
for (j=N_LMS-2; j>0; j--) {
sum+=h24[j]*x[j]; x[j]=x[j-1];
}
sum+=h24[0]*x[0];
/* Convolution complete */
/* Derive error e(n) */
e=y-sum;
/* Output result e(n) */
...=e;
/* Coefficient updating (executed once every sampling interval before */
/* coefficients are frozen) */
/* Determine step size mu */
ex=beta*ex+x[0]*x[0];
mu=alfa/(ex+epsilon);
/* Derive beta */
beta=mu*e;
/* Update coefficients h24[0:N_LMS-2] */
h24[N_LMS-1]+=mu_for_DC*e;
for (j=N_LMS-2; j>=0; j--)
h24[j]+=beta*x[j+1];
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
80
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Room Noise
(if required)
Test Room as
per Section 8.6.4.2
Test
Table
Handsfree
Terminal
Electrical equivalent
of acoustic Rout
Artificial Mouth or
Meas. Mic (if required)
Electrical equivalent
of acoustic Sin
Head and Torso
Simulator or
equivalent
Sources
Digital Set
Analog Set
2 Wire/4 Wire
Digital Hybrid
Either
Rin
Data
Aquisition
Sout
ISDN/Reference
Codec
Rin
V
D.C. Feed
Source
Figure F4 - Test Bed Block Diagram
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
81
Sout
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Annex G
Detailed Test Methodology For Temporally Weighted ERL
G.1 Echo Return Loss Algorithm
The temporally weighted echo return loss ERLt measurement method is described. This method requires
that the echo and the source signal be recorded over the duration of the measurement, and post processing
be used. Real-time measurement techniques are possible, but are not described in this Standard.
Freezing the canceller is not recommended for ERL tests. Freetel results1,2 with non-stationary signals
have shown that convergence times and subsequent converged ERL when "thawed" depend upon the point
in time at which the canceller was frozen.
G.1.1 Echo Return Loss, Temporally Weighted (ERLt)
Temporally weighted ERL, ERLt, is intended to:
•
Provide a measure of time dependent echo return loss with peaky behavior, psycho-acoustically
weighted; the ERLt.
•
Provide an estimate of the number of potentially objectionable echo bursts, and the
acoustically weighted echo return loss during the bursts.
psycho-
The echo signal is first filtered to model the frequency selectivity of human hearing at loudness levels of 30
Phons, as described in section G.1.2. This weights the echo power in a way that the human hearing
response would.
Noise reduction may then be applied and the echo and stimulus files synchronized. Noise reduction is
where the noise is measured and subtracted from the echo plus noise to arrive at a better estimate of the
echo alone. Such a measurement should occur for at least two seconds after all stimulus activity has
stopped. Echo and source are converted into 4 ms power averaged frames allowing adequate resolution and
immunity to synchronization errors.
If the stimulus is inactive, the algorithm simply skips that frame, and moves on to the next echo and
stimulus frames. If the stimulus is declared active, the echo frame is compared with a threshold to
determine if an echo event occurs. The period of echo activity between inactive echo states is termed an
echo "event". These events are then weighted using psycho-acoustic modeling.
By using a threshold of 65dB (5 dB above ulaw noise floor), ERLt can be determined. The actual test
algorithm in pseudo code is detailed in annex H.
G.1.2 Modeling Echo Audibility
[Instantaneous Loudness according to Zwicker?]
In modeling echo audibility, the algorithm accounts for 3 fundamental aspects of human hearing behavior:
(1) The frequency selectivity of human hearing at a loudness levels of 30 Phons ("Fletcher-Munson"
response equivalent to 30 dB at 1 kHz) 11.
11Hearing, Gulick, Gescheider, Frisna, Oxford University Press, 1989.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
82
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Thirty Phons was chosen as it represents echo levels that result from terminals that just fail handset
terminals coupling loss specifications (determined using loss planning analysis). Variance from 20 to 50
Phons provide essentially the same weighting within the telephony band. An A weighted filter is used.
Note that the use of this exact weighting characteristic assumes headphone/handset type listening, or "mean
audible pressure" (MAP) response. Free-field listening such as over a hands-free would require the
Robinson and Dadson "mean audible field" (MAF) weighting, but the difference is slight. MAP weighting
will be used to better reflect the more common use of handset.
The average loss of the filter with white noise is 1.3 dB when measured using ERLs or ERLt. With non
stationary signals, the loss will be time dependent.
(2) The ear's tendency to combine the loudness of sequential signals even though they may be discrete in
time ("temporal combination"). This typically occurs when the two signals are separated by a silent period,
which is less than 20 ms 12,13,14. If two bursts of echo are separated by a period of inactivity less than 20
ms, they are considered as one longer echo event as far as loudness is concerned. This continues until the
gap between events is at least 20 ms, at which time the echo event is declared over. This can be thought of
as a 20 ms hangover for the current echo event. During this hang-over period, echo and stimulus powers
are not included as part of the event. An example of temporal combination follows in figure G1 below:
New Echo
Duration of Echo = 100 ms
Echo
Amplitude
(Power)
Activity
Threshold
Temporal
Combination
20
40
60
Time (20 ms/div)
80
100
120
Figure G1
(3) The duration of the total echo event after temporal combination is measured based on the ear's natural
temporal integration behavior. The total duration includes any gap(s) between events that are captured by
temporal combination, but not the final 20 ms hangover.
If the duration is less than 750 ms, the level of the event is reduced to account for the temporal integration
behavior of human hearing. An equation describing the relationship was derived based upon audition
studies with noise15:
Temporal integration weighting = - 23 + 8log(t) in dB
12The LEDE Concept, D. Davis, C. Davis, JAES, 1985.
13The Detection of Reflections, S. Olive, JAES, 1987.
14Modification of Timbre by Resonance, S. Olive, JAES, 1988.
15Auditory Demonstrations, ASA, Philips CD set 1126-061.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
83
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
where t = total duration of echo event (ms), t_750 ms
Note that tones result in a slightly different relationship, but it was felt that noise was a much closer
approximation to the true nature of the echo than a sine.
Relative Loudness Level (dB)
A graphical representation of temporal weighting is shown below in figure G2.
0
-10
-20
Broadband Noise
10
100
1000
Duration (ms)
Figure G2
If the duration is longer than 750 ms, the level of the total event is left unweighted. Test results have shown
echo bursts less than 750 ms to be common occurrences from cancellers.
G.1.3 Expressing ERLt Results
Traditional ERL methods refer the echo power during the duration of measurement to the source power
during the duration of measurement to arrive at the echo return loss. In this method, the final weighted
power of echo during each event is referred to the power of the source signal during the same event, to
arrive at the "Active ERLt", AERLt, of each event. The echo is referred to the source signal during the
event only, as this is the way in which our ear would compare the echo.
A long term average of the weighted active echo return loss is found by summing the power of all weighted
echo during active events, and comparing to the power of the source as seen during all events only. The
result is the "Active Long Term ERLt".
For comparison with traditional ERL methods, the power of all weighted echo during events is summed,
then referred to the total source power as measured for the entire duration of the measurement. The result is
the "Long Term ERLt".
Note that the terminology for ERLt results was chosen to be consistent with P.56 nomenclature.
Other statistics compiled include minimum and maximum AERLt, standard deviation ("sigma") of AERLt,
the mean of AERLt, the total number of echo events (combined events due to the "Haas" effect are
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
84
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
considered one total event), the number of echo events per minute, the percentage of echo event free
speech, number of events < 750 ms, the average length of an event and the duration of source inactivity.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
85
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Annex H
ERLt Test Algorithm
ERLt is a newly proposed method for evaluating the echo return loss of a terminal using psychoacoustic
modeling and for predicting the occurrences of potentially objectionable echoes. It incorporates 3
fundamental aspects of human audition:
•
•
•
frequency selectivity of human hearing ("Fletcher Munson" response);
temporal addition of level for events within 20 ms of each other ("Haas" effect);
temporal integration for stimuli below 750 ms.
Audition details leading to the algorithm are available in annex G. The implementation details of the
algorithm follow.
A source signal as described in section 6 may be used. Speech based stimulus signals are recommended as
their results are most representative of real world usage. The system output is always some echo or noise
making its way through the system uncancelled.
The stimulus and echo should be recorded and made available in digital format. User inputs regarding set
type (analog or digital), EPDn and double talk or single talk tests should be available. Calibration
parameters should be used to scale echo and stimulus frames to absolute values, and hybrid processing
should have removed hybrid echo for 2 wire analog sets.
The stimulus and the echo files will be processed as power values averaged over 4 ms frames. The
successive stimulus file frames will be termed xi, the echo frames will be denoted yi, where i = 1, 2, 3.... is
the actual frame index. Intermediate frames conforming to an "echo event" will be noted as xk, and yk,
where k = 1, 2, 3... is the echo event index, and is reset when the event ends a new one commences.
Statistics compiled during the ERLt measurement include the Active Long Term ERLt (ALERLt), Long
Term ERLt (LERLt), minimum and maximum Active ERLt (MINERL, MAXERL), its sigma and mean, the
total number of echo events (combined events due to the "Haas" effect are considered one total event)
(NEVENTS), the number of echo events per minute (NEVMIN), the percentage of echo event free speech
(PER), number of events < 750 ms (N750), the average length of an event (AVGEVENT), and the duration
stimulus was inactive (DUR). The terminology for ERLt results was chosen to be consistent with P.56
nomenclature. The duration of stimulus inactivity is not included in the time-based results.
H.1 ERLt Algorithm
Step 1 (Optional but recommended)
Calculate the correlation of stimulus and echo file to fine tune EPDn. Use the criteria that the
present correlation peak occurs at EPDn unless a following correlation peak has a magnitude at
least 10 dB greater. This approximate guideline is based upon subjective studies in the JAES on
delay detection with multiple impulses.
Step 2
Align the echo and stimulus files in time by removing delay equal to EPDn from the echo file.
Step 3
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
86
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
The individual echo samples are processed through a filter approximating the mean audible
pressure equal loudness contour for 30 Phons. This can be accurately approximated (within ± 1
dB from 200 Hz to 2500 Hz) by a first order high pass filter with a -3 dB point of 800 Hz.
Step 4
If it can be assumed that the noise in the echo path is stationary and uncorrelated with the echo,
the noise is measured for 2 seconds after the stop of source and echo activity. The noise is then
subtracted from the echo plus noise to arrive at a better estimate of the echo alone.
Step 5
Samples are converted to absolute numbers using the calibration data. The stimulus samples are
combined into 4 ms power averaged frames denoted as xi. The weighted, noise filtered echo
samples are combined into 4 ms power averaged frames denoted as yi.
Step 6 Begin Echo Return Loss Calculations
Initialize variables:
i = 0 (frame counter)
j = 0 (frame counter for inactive signal duration)
nk=0 = 0 (number of frames in current echo event)
NSAMPS = 0 (accumulated number of frames for all events)
HAAS = 0 (counter up to 20 ms)
ei=0 = 0 (running summation of all echo power for all events after weighting, as seen at frame
counter i)
pi=0 = 0 (running summation of all stimulus power during the measurement, as seen at frame
counter i)
ek=0 = 0 (running summation of echo power during the particular echo event after weighting, as
seen at event frame counter k)
sk=0 = 0 (running summation of stimulus power during the particular echo event after weighting,
as seen at event frame counter k)
WEIGHT = 0 (temporal based weight of most recent event)
LEVENT = 0 (echo return loss level of most recent event, after weighting)
NEVENT = 0 (total number of echo events)
N750 = 0 (total number of echo events < 750 ms)
MINERL = 75 (minimum echo return loss level of all events)
MAXERL = 0 (maximum echo return loss level of all events)
EVENT[NEVENT] = 0 (initialize array for all event loss levels (in dB) to zero; used to calculate
sigma)
TEMPSK = 0 (running sum of stimulus power during all events)
SUM = 0 (used in calculating sigma)
SQ = 0 (used in calculating sigma)
Step 7
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
87
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Increment frame counter and read in 4 ms averaged echo power yi, and 4 ms averaged stimulus
power, xi; if there are no more valid inputs and either measurement file is complete, go to step 8.
1
i = i +1 (unless last i, then go to step 8)
Sum stimulus powers
pi = pi + xi
Is stimulus loud enough for a valid echo loss calculation? If not, disregard present frame and
move to next frame.
4
If xi < (long term stimulus rms level - 25 dB)
j=j+1
i=i+1
Go to 4
Else
Test echo against threshold
If yi _ -65 dB
{5 dB above ulaw noise floor}
Increment frame event counter
k = k +1
Increment frame event length including any gaps < 20 ms
nk = nk +1 + HAAS
Reset "Haas kicker"
HAAS = 0
Accumulate echo power of event
ek = ek + yi
Accumulate stimulus power during event
sk = sk + xi
Go to 1
Else
Has there been no event within last 20 ms?
If k=0
HAAS = 0
Go to 1
Else
There has been an event within the last 20 ms
HAAS = HAAS + 1
Has 20 ms without an event elapsed after a recent event?
If HAAS*4 < 20
Go to 1
Else
An event is over, add an event to the event counter
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
88
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
NEVENT = NEVENT + 1
Increment the total events duration counter by adding the duration in frames of the most recent
event
NSAMPS = NSAMPS + nk
Was the most recent event duration < 750 ms?
If nk*4 < 750
Calculate temporal integration weighting for most recent echo event
WEIGHT = 8*log10(nk*4) - 23
Increment the counter for the number of events that were temporally weighted
N750 = N750 +1
Else
Calculate weighted echo return loss of the most recent event in dB
LEVENT = 10*log10(sk/ek) - WEIGHT
Store the minimum and maximum echo return losses in dB
IF LEVENT < MINERL; MINERL = LEVENT
IF LEVENT > MAXERL; MAXERL = LEVENT
Store the echo return loss of the most recent event in dB for future sigma calculation
EVENT(NEVENT) = LEVENT
Reconvert the echo return loss of the most recent event into linear; recalculate weighted linear
echo power
ek = sk/(10**(LEVENT/10))
Accumulate all the echo event powers for future use in calculating ALERLt and LERLt
ei = ei + ek
Accumulate all the stimulus powers during events for future use in calculating ALERLt
TEMPSK = TEMPSK + sk
Reset echo event variables
k=0
nk = 0
WEIGHT = 0
HAAS = 0
ek = 0
sk = 0
Go to 1
Step 8
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
89
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Calculate Active Long Term ERLt (ALERLt), Long Term ERLt (LERLt), the number of echo
events per minute (NEVMIN), the percentage of echo event free speech (PER), the average length
of an event (AVGEVENT) and duration during which speech was inactive (DUR).
Note: Zero check ei before computing; if ei = 0, set ALERLt and LERLt to 100 dB.
ALERLt = 10*log10(TEMPSK/ei)
LERLt = 10*log10(pi/ei)
NEVMIN = 60*NEVENT/((i-j)*0.004) {number of events per minute}
PER = 100*((i-j) - NSAMPS)/(i-j) {percentage of echo free speech)
AVGEVENT = NSAMPS*4/NEVENT {average length of an event in milliseconds}
DUR = j**0.004
Calculate sigma by analyzing the EVENT array which contains the echo return loss of each event;
each event, regardless of duration, is given equal weighting in the sigma calculation; the
suggestion is that it is the transition between discreet events and not their duration that is most
objectionable.
Loop j from 1 to NEVENT
SUM=SUM+EVENT(j)
SQ=SQ+EVENT(j)**2
ENDLOOP
SIGMA = SQRT(SQ/NEVENT - [SUM/NEVENT}**2)
Calculate mean of the events
MEAN = SUM/NEVENT
Step 9
Output statistics
Print ALERLt, LERLt, MINERL, MAXERL, NEVENT, NEVMIN, PER, N750,
AVGEVENT, DUR,
SIGMA, MEAN
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
90
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Annex I
Double Talk Testing
I.1 General
Testing in double talk mode is somewhat complex in that the test signal required has to allow for
discrimination between both transmission directions. The method also has to do this in such a way as to
allow the hands-free phone to operate in as normal a fashion as possible. The following methods detail how
this can be done as referenced by sections ???.
I.1 Canceller Training prior to Double Talk
Before double talk testing is entered, all cancellers must be fully converged using training signals for both
transmission directions. The test signals should be uncorrelated, and the training period should allow at
least 5 seconds of single talk in each direction. The amplitude mask statistics for the training period are
shown below:
Parameter
Rate (%)
Over 60 seconds (s)
Talk Spurt
38.53
23.12 ~ 24
Pause
61.47
36.88 ~ 36
Double Talk
6.59
3.95 ~ 0
Mutual Silence
22.48
13.49 ~ 12
Table I1 - Temporal Parameters Of Conversational Speech
I.1.1 Double Talk Training Activity Masks
The exact amplitude masks will now be specified along with signal amplitude characteristics during double
talk. Each type of double talk test has special requirements for signal duration and amplitude during double
talk. For echo return loss, the duration must be long enough to capture any divergence, but not so long as to
be a burden on test system memory resources or so long as to result in an unacceptable computation time.
Tests have shown a 20 second double talk duration to be acceptable for double talk echo return loss testing.
Once double talk has ended (the talker initiating double talk becomes inactive), the echo return loss
measurement may continue for 10 seconds (the talker active just before double talk remains active) to
measure recovery after double talk. After that time, two seconds of silence should be played. In this way,
the noise in the echo path can be measured. If it can be assumed that the noise and echo are uncorrelated,
and that the noise is stationary, the noise measured in the two seconds may be subtracted from the echo plus
noise during double talk to arrive at a more precise measure of the echo during double talk.
The duration of double talk during double talk attenuation and clipping tests may be much shorter. As all
time constants under study should be less than 200 ms, the double talk duration is set at 200 ms. Analysis is
continued (the talker active just before double talk remains active) for one second after the end of double
talk for the attenuation tests in order to measure any loss removal as single talk is re-entered. There is no
need to estimate and correct for noise in the double talk attenuation and clipping tests.
The recommended masks are below.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
91
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Activity Mask
Talker Active Just before Double Talk
Silence (2 sec)
Optional: 10+2
sec for recovery
after double talk
Double
Talk
Time (s)
0
18
30
48
60
90
Activity Mask
Talker Initiating Double Talk
Double
Talk
Time (s)
0 3
80
33
45
60
Figure I1 - Echo Return Loss Test Activity Mask
15
Activity Mask
Talker Active Just before Double Talk
Time (s)
0
18
30
48
60 61.2
Activity Mask
Double Talk
Talker Initiating Double Talk (200 ms)
Time (s)
0 3
15
33
45
60
Figure I2 - Attenuation Test and Clipping Test Activity Mask
I.1.2 Synchronizing the Double Talk Training Activity Masks
The timing of the masks must be synchronized to avoid pre-mature double talk. This involves accounting
for the 1.5ms air path delay between the artificial mouth and the HFT. The mouth simulator signal should
be initiated 1.5 ms before the RETP signal by delaying the stimulus file used on receive by 1.5 ms. The
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
92
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
start of double talk is defined as occurring when the microphone location sees valid send activity and RETP
sees valid receive activity.
In the activity mask diagrams, the t=0 starting point refers to the beginning of the file applied at RETP. The
starting point for the mouth simulator signal can be thought to be t=-1.5 ms, but the terminal will see them
synchronized. At the 60 second mark, double talk is entered and double talk testing begins.
I.1.3 Compensating for Measurement Filters
The double talk test methods require the use of filters injected in the audio path (for details, see I.6). These
filters will have an impact on the time domain resolution and the precise moment at which double talk
testing can begin. By using filters of known ringing time, the measurement can be put in a wait state while
the filter ringing settles. An example implementation and further explanations can be found in annex F,
section F.4.
I.2 Double Talk Source Signals
Double talk testing often imposes conflicting restraints on the type of test signal used.
For echo return loss tests, the double talk signal presented to the canceller at RETP (or mouth simulator)
should be as similar to the training signal as possible. Cancellers typically freeze adaptation during double
talk. For example, if the double talk signal at RETP differed from the training signal, residual echo would
typically be unrealistically high. This constraint indicates that the double talk signal at RETP should be the
same as the training signal at RETP for echo return loss measurements.
Unfortunately, the use of the speech files alone is not acceptable during double talk. The correlated parts
between the two "talkers" would invalidate some test results: parts of the one talker's speech may look like
echo of the other talker if the parts are correlated. Another problem is that double talk onset must be very
accurately detected for attenuation and clipping tests. This would be very difficult to define over repeated
tests using different speech files, but is very easy with tones.
To overcome these issues, both signal types are used; speech as per the training signal, and tones to
accurately define the start of double talk. How they are used depends upon the specific test. These
concepts are best explained in the following individual test sections, but will be briefly described here.
Speech signals used during training are continued during double talk, as required. A sinusoidal tone is
mixed in with the speech or injected on its own to provide an easily measurable reference for attenuation
tests or an easily definable start of double talk for clipping tests. By using notch or band-pass filters at
SETP (or receive output) at the tone frequency, either just the tone or just the speech can be monitored.
When the tone is mixed in with the speech, the power of the tone must be representative of the long term
average power of speech at its frequency, so as to not impact the canceller with any gross deviations in
spectral energy from that of the training signal. ITU-T Recommendation P.50 specifies an average spectral
relationship (third octave values used). The following tones are recommended. Their power is defined as
the number of dB below the average active speech energy in the speech file, when measured as per ITU-T
Recommendation P.56 (see section 5.2.2.5 g):
Tone Frequency
Relative Tone Level (dB) below
Nominal Speech Level
500
1000
1750
2500
9
14
18
22
I.3 Double Talk Attenuation Testing
The example shown determines double talk attenuation in the send direction. The concept is easily
extended to the receive direction by reversing signals and monitoring at the receive output.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
93
Amplitude
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Rin
Rout
Echo
Canceller
Frequency
Adaptive
Model
echo
path
-
Sout
Sin
Frequency
Echo
Frequency
Amplitude
Amplitude
Amplitude
+
Frequency
The methodology is explained below:
The set is reset, and trained as described in section I.1. The "talker active before double talk" is the mouth
simulator. The "talker initiating double talk" is RETP.
Notice that the signals at both the mouth simulator and RETP are shown notch filtered at the tone
frequency. This notch filter is not present during the entire training period, but only just before double talk
and for the remainder of the measurement. The idea is to mix in a tone at the mouth simulator just before
double talk (still in single talk), monitor its rms level, have RETP enter double talk, and continue
monitoring the tone level.
The double talk attenuation is the difference in tone level before double talk and during double talk. The
tone is discriminated by applying a band-pass filter at the tone frequency at SETP. By continuing
measurement during double talk, the switching characteristics including rate of insertion and depth can be
determined. The rate of attenuation removal can also be determined by making the activity mask for the
"talker initiating double talk" low again after the attenuation depth has stabilized.
Characteristics of the notch filter will now be described. The notch is required on RETP to ensure that echo
at the tone frequency does not impact the measurement of the tone. The notch filter must show enough
attenuation to ensure speech at the tone frequency is adequately repressed so as not to impact the level of
the tone at the mouth simulator. The notch filter bandwidth must be tight enough to minimize impact on
surrounding frequencies so that the signals are not significantly different than the training signals. Example
filter types are described in annex F, F.4.
The band-pass filter has similar constraints. Taken with the notch filter, it must have enough out of band
attenuation to ensure that speech or echo does not impact the tone level. It must also have a short enough
impulse response that the time domain impacts is minimized, as the rate of attenuation insertion is also
being measured. The exact filter type is described in annex F, F.4.
The impulse response of the notch filter does not impact the measurement as the tone mixed in at the mouth
simulator will be large enough in level to swamp any residual ringing of the notch's.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
94
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
The rms level of the tone is to be measured using an 8 ms sliding rectangular window for smoothing. The
window is slid in 4 ms increments for 4 ms of overlap between adjacent points to smooth the results.
Amp litude
The timing of the measurement must be fine-tuned to account for any ringing of the bandpass filter. The
tone should be injected at 60 seconds, minus the bandpass filter's ringing time (< 20 ms assumed), minus 8
ms. The tone reference measurement during single talk is taken starting at 60 seconds minus 8 ms, after the
filter ringing has ended. Set delay in the direction of measurement leads to partial measurement during the
end of the bandpass filter's ringing. As long as set delay is low (provisionally <5 ms), the amount of ringing
effects encountered will be slight and should not impact the measurement.
Tone
injection
for duration of
Measurement
Double
Talk
Starts
60
< 20 ms
time (s)
8 ms
Filter Ringing
Reference
Measurement
When sub-banding techniques are used in the AEC, or the technique is unknown, it is advisable to repeat
the test for each frequency shown in section I.2. In many cases, the attenuation test results at any one
frequency may not be indicative of subjective quality. As voice has dominant spectral energy in the lower
frequency range, we would expect that the switched loss to be more audible there. If the depth of
attenuation is frequency independent, it is advised to use higher test frequencies, as the required filters will
have less of an impact on the over-all voice levels.
I.4 Double Talk Front End Syllabic Clipping Testing
The example shown determines double talk front-end syllabic clipping testing in the send direction. The
concept is easily extended to the receive direction by reversing signals and monitoring at the receive output.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
95
Amplitude
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
R in
Rout
Echo
Canceller
Frequency
Adaptive
Model
echo
path
-
Sout
Sin
Frequency
Echo
Amplitude
Measure Duration
Amplitude
Amplitude
+
Frequency
Duration Known
Frequency
The set is reset, and trained as described in section I.1. The "talker active before double talk" is RETP.
The "talker initiating double talk" is the mouth simulator.
The notch filter is applied at RETP as described in section I.2. Double talk is entered by applying a tone at
the mouth simulator. The tone is injected for only 200ms. The duration of front-end syllabic clipping is
200ms minus the duration of tone activity at SETP. Using a tone ensures that the precise moment that
double talk is entered is known.
The duration of tone activity at SETP is determined by monitoring its level at SETP, after passing the signal
through a band-pass filter to remove any residual echo. Activity is declared when the level of the tone is no
quieter than 3 dB below the tone level 200 ms after double talk is entered. It is assumed that all double talk
attenuation is fully inserted before 200 ms into double talk. The impact of set and bandpass filter ringing
times on duration are minimized by setting the threshold at a high 3 dB.
The rms level of the tone is to be measured using a 4 ms sliding window, slid in 2 ms increments for 2 ms of
overlap between adjacent points.
The use of a 1.75 kHz or 2.5 kHz tone is recommended, at the relative level recommended in section I.2.
The frequency is chosen to minimize the impact of the notch filter on the RETP signal.
I.5 Double Talk Echo Return Loss Testing
The example shown determines echo return loss looking towards the terminal from the network. The
concept is easily extended to talker echo path loss by reversing signals and monitoring at the receive output.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
96
Amplitude
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Rin
Rout
Echo
Canceller
Frequency
Adaptive
Model
echo
path
-
Sout
Sin
Frequency
Echo
Amplitude
Echo
Amplitude
Amplitude
+
Frequency
Frequency
The set is reset, and trained as described in section I.1. The "talker active before double talk" is RETP.
The "talker initiating double talk" is the mouth simulator.
The notch filter is applied at RETP as described in section I.2. Double talk is entered by applying a tone at
the mouth simulator. The tone is injected for 20 seconds. The echo return loss is found by first notching
out the tone and then measuring the residual echo using one of the techniques in section G.1. Using a tone
ensures that the precise moment that double talk is entered is known.
Once double talk has ended, the echo return loss measurement is to continue for 10 seconds to measure
recovery after double talk. After that time, one second of silence should be played. In this way, the noise in
the echo path can be measured. If it can be assumed that the noise and echo are uncorrelated, and that the
noise is stationary, the noise measured in the last second may be subtracted from the echo plus noise during
double talk to arrive at the echo during double talk.
The timing of the measurement must be fine tuned knowing the echo path delay. This delay properly aligns
the source and echo. For the example shown, this value is determined by the EPDn test of section 11.6.1.
For talker echo path loss, the echo path delay off the network interface (for 2-wire terminals) is found by the
EPDa test of section 11.6.1.
As per the clipping test, it is recommended that a 1.75 kHz or 2.5 kHz tone be used at the relative level
recommended in section I.2.
I.6 Double Talk Measurement Filters
Double talk testing requires the use of notch and bandpass filters at various frequencies. A recommended
implementation is tabulated below.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
97
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
The terms described are:
o fpl: lower frequency at which the bandpass or bandstop is at - 3 dB.
o fpu: upper frequency at which the bandpass or bandstop is at - 3 dB.
o fsl: lower frequency at which the bandpass or bandstop is at - atten dB.
o fsu: upper frequency at which the bandpass or bandstop is at - atten dB.
o atten: the specified full attenuation of the filter.
o atten: the actual full attenuation of the filter.
o ripple: ripple of the filter in dB (+/-).
o gain: gain of the bandpass filter (linear) in the pass band.
o order: filter order in taps (8 kHz sample rate) for the bandpass. The bandpass ringing time is order
times 125 us. For the bandstop, order refers to the order of the biquad (elliptical).
Filter
Type
fpl
fpu
fsl
fsu
atten
atten
(actual)
ripple
gain
order
500 Hz
1 kHz
1.75 kHz 2.5 kHz
500 Hz
1 kHz
1.75 kHz 2.5 kHz
FIR
FIR
FIR
FIR
IIR
IIR
IIR
IIR
Bandpass Bandpass Bandpass Bandpass Bandstop Bandstop Bandstop Bandstop
495
990
1733
2475
400
800
1450
2100
505
1010
1767
2525
610
1250
2100
2950
435
900
1611
2302
435
900
1610
2300
570
1100
1900
2715
570
1100
1900
2715
30
30
30
30
30
30
30
30
31
29.5
34
34
30
40
1
0.92
160
1
.9
100
3
.78
80
1
.99
60
1.5
1.5
1.5
1.5
6
6
6
6
The bandpass filter's ringing time will impact the measurement if not accounted for. Measurements must
commence only after the filter had stopped ringing due to initial application. This is necessary so that a
clean reference measurement can be made for attenuation tests and clipping tests. The longest ringing time
is 20 ms for the 500 Hz bandpass filter. Since the averaging window for measurement in attenuation testing
is 8 ms, the bandpass filter must be inserted at 20 + 8 = 28 ms before the onset of double talk, or 60 - 0.028
= 59.972 ms in the activity mask of section I.1.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
98
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Annex J
Acoustic Echo Path Tutorial
J.1 Acoustic Echo Path
The use of a real room with appropriate acoustic characteristics is recommended. Simulated echo paths are
not recommended as they will not capture any non-linearities, which limit cancellation ability in real world
use.
The reverberation time (RT60) of the test room should meet ITU-T Recommendation G.167. When
averaged over the transmission bandwidth, RT60, shall be approximately 500 ms; the reverberation time in
the lowest octave shall be no more than twice this average value; the reverberation time in the highest
octave shall be not less than half this value. The volume of a typical test room shall be of the order of 50 m3
/ 1500 ft3.
Acoustic canceller works mainly on early room reflections, which show a low density of reflections.
Residual echo power is dominated by the reverberant tail of the acoustic system’s room impulse response,
with a high density of modes and long duration as shown below. Specification should include both
reverberation time as per ITU-T Recommendation G.167 and early/late ratio of room. This is an item
worthy of study.
Direct Path
Discrete Early Reflections
10
Relative Amplitude
Reverberant Tail
0
Time
-10
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
99
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Annex K
HFT Microphones
This annex briefly describes a variety of microphone types and applications as they may relate to HandsFree Telephones (HFTs).
The frequency response can be measured as a function of distance, and a function of polar position around
an imaginary “front” or “frontal axis” of the hands-free telephone. Similarly, it may be desirable to measure
the frequency response for both a number of vertical and horizontal polar positions.
Omni directional microphones have a uniform output regardless of the actual position of the microphone
itself. The electrical output of an omni directional microphone decreases by 6 dB with every doubling of
distance from the sound source in a simulated free field or anechoic environment.
Cardioid, hyper cardioid, super cardioid, and other directional microphones generally have both frequency
response and polar characteristics which vary as a function of position and distance from the sound source.
Directional microphones have two or more input ports. It cannot be assumed that the output of such a
microphone will decrease by 6 dB with every doubling of distance from a sound source. For the most part,
the frequency characteristics of directional microphones are both polar and distance dependent. The sound
source should be placed at a distance and direction that approximates real use conditions or the RTP.
The microphone system can be implemented with electronic (automatic) gain control (AGC). To reduce
background noise, some hands-free telephone sets have send circuits with variable sensitivity based on the
voice signal's presence. (When voice is not present, the send sensitivity is reduced, and when voice is
present, this sensitivity is restored to normal). It is particularly important that such microphones be tested at
normal use input levels to obtain the proper sensitivity.
Some HFT’s, frequently known as "teleconferencers," feature multiple microphone elements, typically 2-4,
all within a common housing. This type of assembly may be sensitive to both distance and orientation, and
may also feature gating, or the ability to turn the microphone on and off
Teleconferencers frequently have optional auxiliary microphones intended to extend the coverage area or
reach of the teleconferencer. These auxiliary microphones may be either directional or omni-directional,
and may therefore be sensitive to both distance and orientation.
If the microphone is a separate device, such as a stand, boundary or a lapel microphone, it should be so
tested and so described with the appropriate user position as a reference. This configuration is frequently
used in PCs and automobiles.
Some HFT’s feature multiple microphones which are combined with electronic processing to produce
highly directional fixed or variable polar patterns. Testing and describing such systems may require
techniques which differ from traditional measurement methods. For instance, beam steering reacts to
sounds above a given level. One purpose for using such a method is to improve the acoustic signal to noise
ratio, which may be a useful way to describe the test results.
It is especially advisable to perform send loudness rating directionality (SLRD) on HFTs with directional
microphone systems.
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
100
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Annex L
1/3 Octave Passbands
Band No.
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
Nominal Center Frequency
[Hz]
100
125
160
200
250
315
400
500
630
800
1000
1250
1600
2000
2500
3150
4000
5000
6300
8000
10000
Table L1
1/3 Octave Passband [Hz]
89.1
112
141
178
224
282
355
447
562
708
891
1120
1410
1780
2240
2820
3550
4470
5620
7080
8910
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
101
112
141
178
224
282
355
447
562
708
891
1120
1410
1780
2240
2820
3550
4470
5620
7080
8910
11200
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Annex X
DRP TO ERP Corrections For HATS Receive Measurements
Prior to the calculation of Receive LR on hands-free telephones, section 8.3.x provisionally recommends
the addition of 14 dB to the measured receive response to account for the binaural effect of two ear
listening. Because a portion of this effect is due the obstacle effect of the head (which is included as a part
of the HATS measurement), a provisional correction of {12 dB ?} to HATS receive measurements is
recommended. This is to be added to HATS receive measurements prior to any single figure calculations
(i.e., LR, Mid-band Average Sensitivity, etc.). This correction should not be applied to the presented
frequency response.
Frequency
(Hz)
92
97
103
109
115
122
130
137
145
154
163
173
183
193
205
218
230
244
259
274
SDE
(dB)
0.1
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
-0.1
-0.1
0.0
0.1
0.0
-0.1
-0.2
-0.3
-0.3
SDE
SDE
Frequency
Frequency
Frequency
(Hz)
(dB)
(Hz)
(dB)
(Hz)
290
-0.3
917
-1.3
2901
307
-0.2
972
-1.4
3073
325
-0.2
1029
-1.8
3255
345
-0.2
1090
-2.0
3447
365
-0.4
1155
-2.3
3652
387
-0.5
1223
-2.4
3868
410
-0.4
1296
-2.6
4097
434
-0.6
1372
-3.1
4340
460
-0.3
1454
-3.3
4597
487
-0.7
1540
-3.9
4870
516
-0.6
1631
-4.4
5158
546
-0.6
1728
-4.8
5464
579
-0.6
1830
-5.3
5788
613
-0.6
1939
-6.0
6131
649
-0.8
2054
-6.9
6494
688
-0.8
2175
-7.5
6879
729
-1.0
2304
-8.1
7286
772
-1.1
2441
-9.1
7718
818
-1.1
2585
-9.5
8175
866
-1.2
2738
-10.4
8659
Table X1 - 1/12 Octave Filter Center Frequencies
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
102
SDE
(dB)
-11.0
-10.5
-10.2
-9.1
-8.0
-6.9
-5.8
-5.0
-4.2
-3.3
-2.7
-2.4
-2.4
-2.5
-3.3
-4.5
-5.9
-9.0
-14.2
-20.7
STANDARD METHOD FOR MEASURING TRANSMISSION
PERFORMANCE OF HANDS-FREE TELEPHONE SETS
Frequency
(Hz)
100
106
112
118
125
132
140
150
160
170
180
190
200
212
224
236
250
265
280
SDE
(dB)
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
-0.1
-0.1
0.0
0.1
0.0
-0.1
-0.1
-0.2
-0.3
-0.3
Frequency
(Hz)
335
355
375
400
425
450
475
500
530
560
600
630
670
710
750
800
850
900
950
SDE
(dB)
-0.2
-0.3
-0.4
-0.4
-0.5
-0.4
-0.5
-0.7
-0.6
-0.6
-0.6
-0.7
-0.8
-0.9
-1.1
-1.1
-1.2
-1.3
-1.4
Frequency
(Hz)
1120
1180
1250
1320
1400
1500
1600
1700
1800
1900
2000
2120
2240
2360
2500
2650
2800
3000
3150
SDE
(dB)
-2.1
-2.3
-2.5
-2.8
-3.2
-3.6
-4.2
-4.7
-5.2
-5.8
-6.5
-7.2
-7.8
-8.5
-9.3
-9.9
-10.6
-10.7
-10.4
300
-0.2
1000
-1.6
3350
-9.6
315
-0.2
1060
-1.9
3550
-8.5
Frequency
(Hz)
3750
4000
4250
4500
4750
5000
5300
5600
6000
6300
6700
7100
7500
8000
8500
9000
9500
10000
Table X2 - ISO R40 Preferred Frequencies
Copyright  1999 IEEE. All rights reserved.
This is an unapproved IEEE Standards Draft, subject to change.
103
SDE
(dB)
-7.5
-6.3
-5.3
-4.5
-3.7
-3.0
-2.6
-2.4
-2.5
-2.9
-4.0
-5.3
-7.5
-12.2
-18.6
*
*
*