210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
1
Advanced Voiceband
Modem Configuration and
Troubleshooting
Session 210
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
2
Agenda
• Understanding the Anatomy of a V.90
Modem Connection
• What V.90 Wants and Does Not Want
• Hooking up Things Right and Proving It
• Living with Imperfection
• Hand-to-Hand Combat with Individual
Connections
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
3
Anatomy of a V.90
Connection
What a Modem Is
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
4
Communications Link—
“Old World” Model
Modems Are an “Old World” Artifact—
They Deliver Data over Circuits
DCE
DCE
DTE
DTE
‘The Network’
The Circuit
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
5
Modem Protocol Chart—
Transmission Plane
Layer
Protocols
Application
8 Bit Data
Stream
Async
Framing
PPP
SLIP
1 Startbit, 7–8 Databits, 0–1 Parity Bits (Even, Odd, Mark, Space), 1–2 Stopbits
Normally 10 Bits Per Word: 1 Startbit, 8 Databits, No Parity, 1 Stopbit
DCE
Async
V.42: MNP4, LAP-M (EC Uses Sync Framing with 8-Bit Words) Framing
Compression MNP5, V.42bis
Modem
(DCE)
TTY (Async Terminal)
Error Control
Modulation
(Compression Requires EC)
Analog Carriers
Bell103, Bell212A, V.21, V.22, V.22bis, Client->Server: Analog
V.23, V.32, V.32bis, V.FC, V.34
V.34-Like Modulation
PSTN
Circuit
POTS Circuit—3–4 KHz
Analog (~200–~3600Hz)
PSTN Link
Loopstart Analog Local Lines
E&M Analog Trunks
(Lines and
Trunks)
Physical
210
1081_06F9_c1
2-Wire Copper, 4-Wire Copper
© 1999, Cisco Systems, Inc.
Half-Digital: K56Flex, X2, V.90
Server->Client:
Digital PCM Words
PCM Layer—
A-law
8-bit Words 8000/sec (NA)
• -law
(World)
DS0 Layer—
CAS (RBS) 64Kbps
Sync Serial
Channel
Clear
Digital Channel 56—63Kbps Channel
CAS
Channelized
T1s
PRI, BRI
R2 E1
Clear-Channel Trunks
in SS7 Networks
SONET, HDSL, 2B1Q, etc.
6
Modem Operations
TxD
RxD
Buffer
Flow Control
Buffer
RTS
CTS
Compressor
Compress
Decompress
Compressor
Packetizer
Checksum
Retransmit
Packetizer
Modulator/
Demodulator
Modulator/
Demodulator
CP
DSP
Modulation
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
7
Anatomy of an Analog Modem
Serial
Port
Binary Data
from User’s
Application
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
U
A
R
T
RX
A/D
DSP
+
TX
RJ11
Jack
D/A
Digitally
Encoded Analog
Samples
Analog
Signal
8
Modem Standards
Modem Standard Key Features
Bell 103/V.21/V.23
FSK Modulation
Bell 212A
DPSK Modulation
V.22
Data Scrambler, International Standard
V.22bis
QAM Modulation, Channel Equalization
V.32
Trellis Coded Modulation, Echo Canceller
V.32bis
Multiple Bit Rates, Deeper Trellis
V.32terbo
Higher Bit Rate
V.34
Pre Emphasis, Multiple Baud Rates, Multiple
Carrier Frequencies, Shell Mapping, Precoding,
Constellation Shaping, Power Control, Line
Probing, Non-Linear Encoding, 4D-trellis
V.34+
Higher Bit Rate
V.90/KFlex/X2
PCM Modulation, Spectral Shaping, Digital
Impairment Probing, Dynamic Constellations
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
9
Modem Standards (Cont.)
MODEM
DUPLEX
RATE (kbps)
MODULATION
V.21/B103
V.23/B202
V.22/B212
V.26bis
V.22bis
V.27ter
V.29
V.17
V.32
V.32bis
V.32terbo
V.34+
V.90/Kflex
F/D FDM
F/D FDM
F/D FDM
F/D FDM
F/D FDM
H/D
H/D
H/D
F/D EC
F/D EC
F/D EC
F/D EC
F/D EC
0.3
0.6, 1./0.075
1.2
2.4
2.4, 1.2
4.8, 2.4
9.6, 7.2, 4.8
14.4, 12.0, 9.6, 7.2
9.6, 4.8
14.4, 12.0, 9.6, 7.2, 4.8
4.8 to 19.2
2.4 to 33.6
28 to 56
FSK
FSK
DPSK
DPSK
QAM
QAM
QAM
QAM
TCM/QAM
TCM/QAM
TCM/QAM
FSK/DPSK/TCM/QAM
PCM-PAM
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
10
QAM (V.34) Modulation
• Bit stream is divided into N-bits per symbol sequence
Each bit sequence specifies an amplitude and phase change of
a sinusoidal carrier
90
2 Phase Bits 2 Amplitude Bits
= 4 Phases = 4 Amplitudes
... :1100:1010:1111
(4 Bits Per Symbol)
180
Modulate
Cosine Carrier
0
270
QAM Signal
A[j]*cos(fc*t + ph[k])
j = 0, 1 (4 Discrete Amplitudes)
k = 2, 3 (4 Quadrature Phase Changes)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
11
Some Standards Illustrated
V.22 bis
2400 bps QAM
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
V.27
4800 bps DPSK
V.32 bis
9600 bps QAM with Trellis
12
Anatomy of a V.90
Connection
The Voiceband (Analog) Circuit
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
13
The Old World Circuit
What an Analog Modem Has to Work with:
Level (dBm)
• A passband
between 2.4 and
4 Khz wide
• Some amount of
near-end and
far-end echo
• SNR between 25
and 40 dB
• Some amount of
attenuation,
hopefully < 20dB,
increasing at higher
frequencies (slope)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
-10
-20
-30
-40
-50
-60 0
1
2
3
4
5
Frequency (KHz)
14
V.34 Speeds as a Function of
Bandwidth
Symbol Rate Carrier Frequency Bandwidth Requirements Maximum Bit Rate
2400
2743
2800
3000
3200
3429
1600 Hz
1800 Hz
1646 Hz
1829 Hz
1680 Hz
1867 Hz
1800 Hz
2000 Hz
1829 Hz
1920 Hz
1959 Hz
400-2800 Hz
600-3000 Hz
274-3018 Hz
457-3200 Hz
280-3080 Hz
467-3267 Hz
300-3300 Hz
500-3500 Hz
229-3429 Hz
320-3520 Hz
244-3674 Hz
21600 bps
21600 bps
26400 bps
26400 bps
26400 bps
26400 bps
28800 bps
28800 bps
31200 bps
28800 bps
33600 bps
Note that a Classical 300-3400Hz Voiceband Yields at Best 28800 bps
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
15
Example Chart of SNR
Performance
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
16
An Analog Modem Call
POTS (Analog) circuit
Digital
Analog
Digital
• Traditionally, a 3KHz analog channel (300 to 3300Hz)
• Signal/noise ratio of 25dB (?)
• Suitable for V.34 analog modem—maybe 19200 bps?
• If you’re lucky, you get 3.5KHz and 38dB SNR, and
reach 33600 bps
This is very near the theoretical information capacity
(Shannon) limit for such a channel
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
17
Anatomy of a V.90
Connection
The Digital PSTN: PCM and V.90
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
18
Pulse Code Modulation—
Nyquist Theorem
Voice Bandwidth =
300 Hz to 3400 Hz
v
t
Analog Audio Source
Sampling Stage
= Sample
8 kHz (8,000 Samples/Sec)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
19
Pulse Code Modulation—
Analog to Digital Conversion
A—Law (Europe)
Quantization Noise
10010011 11011001
µ—Law (USA–Japan)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
Quantizing Stage
Note: PCM
Quantization
Imposes a
~38dB SNR
Noise Floor
20
POTS Call Thru the Modern PSTN
Telco Network
DS0
Trunks
Switch
Switch
Analog
2 Wire
DS0
Trunks
Switch
Analog
2 Wire
Signaling
Network
Analog
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
Digital
Analog
21
Analog Call Through the
Digital Network
TA
PSTN
TA
DTE
DTE
Digital
Digital—64Kb DS0
Digital
3.1KHz Analog
• Analog modems modulate a digital signal (up to 33.6Kb)
in a 3.1KHz analog channel which is transmitted through
the network in a 64Kb DS0, silly, isn’t it?
• How to take advantage of the digital nature of the PSTN to
achieve faster speeds between the DTEs?
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
22
Using the Digital Network: ISDN
TA
PSTN
TA
DTE
DTE
Digital—64Kb DS0
• How to take advantage of the digital nature of the PSTN to
get faster speeds between the DTEs?
• The best way: ISDN—run pure digital end-to-end, and use
the full 64Kb DS0 (but costs more)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
23
Digital Modems in the PSTN
Analog
V.90 Modem
PSTN
Digital
V.90 Modem
DTE
NAS
Analog
Local
Loop
Digital—64Kb DS0
Digital PCM Signal (up to 56Kb)
Analog V.34 Signal (up to 33.6Kb)
How to Take Advantage of the Digital Nature of the
PSTN to Get Faster Speeds between the DTEs?
• The cheap, complicated way: PCM modems
Analog (client) PCM modem transmits an analog signal
210
1081_06F9_c1
Analog modem receives a digital (PCM) signal
© 1999, Cisco Systems, Inc.
24
PCM—Pulse Amplitude
Modulation
• PCM modems do not modulate a carrier
• Information is carried as amplitude pulses
encoded in the mu/A-law digitally encoded
samples (PCM data)
• This is possible only because the transmit
modem is directly connected to the
digital network
PCM Modems Use a Technique Called Pulse Amplitude
Modulation (PAM); Information Is Carried in the
Amplitude of Each PCM Digital Sample
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
25
Anatomy of an Analog Modem
~
U
A
R
T
Router
Interface
Binary Data
From
Internet
RX
DSP
TX
TDM
Mux
Digitally Encoded
Analog Samples
T1/E1
Framer
Telco
Interface
E1/T1
Digital
Signal
(and Modem Management Commands)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
26
PCM Modulations
• X2—pre-standard PCM modulation
from 3COM (USR)
• K56Plus—pre-standard modulation
from Rockwell (now Conexant)
• K56Flex—Rockwell and Lucent’s
update to K56Plus
• V.90—ITU-T PCM standard
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
27
X2
• 3COM (USR)’s pre-standard PCM
modulation
• Never supported by Cisco
• Supports 32–64k in 1333 bps steps
• Uses V.8 to communicate capabilities
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
28
K56Plus
• Rockwell’s original pre-standard
• Uses V.8 for capabilities
• 32000 to 60000 in 2000 bps steps
• Supported by Microcom 56k
modems, not by MICA
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
29
K56Flex
• Rockwell and Lucent defacto
“standard” (V1.0 is Rockwell only)
• Uses V.8bis for capabilities
• 32000–6000 in 2000 bps steps
• Microcom 56K supports V1.0 and
above; MICA supports V1.1
• Detects 0, 3 and 6dB Lucent digital
pads; 0 and 6 dB Nortel pads
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
30
The V.90 Standard
• Ratified by the ITU-T in 9/98
• V.8 for capabilities; V.8bis is optional
• 28 to 56k in 1333 bps steps
• 3200/3000 Hz uplink support required
(31200 bps); 3429 Hz optional
• Fractional digital pads detected via
DIL (Digital Impairment Learning)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
31
What V.90 Wants
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
32
What V.90 Wants…Piece by Piece
NAS
PSTN
DTE
Client
DTE
210
1081_06F9_c1
Client House
DCE Wiring
© 1999, Cisco Systems, Inc.
Local
Loop
DS0
Circuit
through
PSTN
Digital
N
Line
Digital
A Modem
S
33
What V.90 Wants: Digital Modems
• MICA modems running good
portware (2.6.2.0)
Turn on digital pad compensation
(S52=1) for higher speeds
• Microcom modems running good
firmware (5.1.20)
&F (digital pad compensation is on
by default)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
34
What V.90 Wants: NAS
• NAS model and Cisco IOS® version
basically irrelevant to modem
connections (NAS just passes
DS0 through)
• Minimum Cisco IOS versions for V.90:
MICA: 12.0(1+)* 11.3(5+){T,AA,NA}
11.2(16+)P (5300/3600)
Microcom: 11.3(5+){T,AA,NA} 12.0(1+)*
11.2(14+)P (5200)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
35
What V.90 Wants: Digital Line
(T1, E1, BRI)
• Should have very low BER to switch
• PRI is better than RBS CAS CT1 (RBS
damages 1333 bps per DS0)
If CAS (RBS or R2), be sure NO signaling
noise during call setup!
• Pure digital straight into digital
switch: NO channel bank!
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
36
What V.90 Does Not Want:
Digital Line
Phone
Switch
Trunk Side Service
Phone
Switch
Channel
Bank
Line Side Service
No Channel Banks!
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
37
What V.90 Does Not Want:
Channel Banks
• Since V.90 requires a digital path
from the digital modem to the last
D/A on the client’s POTS line, a
channel bank on the NAS access line
will destroy V.90
• MICA is not supported in a channel
bank application (because it doesn’t
want to see near-end echo)
• Microcom is limited to mediocre V.34
in a channel bank topology
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
38
What V.90 Wants: PSTN Circuit
• Very few slips or BERs (fairly
rare nowadays)
• No analog trunks (rare unless the
NAS is connected to a PBX)
• Each RBS link destroys 1333 bps
(unless they happen to align)
• No sub-64k coding (e.g. 32k ADPCM
will limit you to 16800 bps V.34)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
39
What V.90 Does Not Want in PSTN
Circuit: Multiple D/A
Must Have Exactly One D/A
Conversion in the Circuit!
NASb
NASa
5ESS
Client B
Channel
Bank
1AESS
Channel
Bank
Channel
Bank
Client A
Channel
Bank
1AESS
1AESS
Client C
Which Digital Modems Can Get PCM to Which Analog Modems?
• NASa to Client A?
• NASa to Client B?
• NASa to Client C?
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
• NASb to Client B?
• NASb to Client C?
• NASb to Client A?
40
PSTN Circuit Digression:
Digital Pads
Reduce Signal Amplitude—Get
“Fixed Loss” across All Circuits
6dB Pad
Lookup
Table
Codec
-19dBm
Analog
Digital Pads Are
Normally—but Not
Necessarily—
in the Line Direction
in the Subscriber’s LE
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
41
Digital Pads and PCM
Modulations
• K56Flex can only cope with 3 and 6
dB digital pads (and has trouble with
Nortel 3dB), K56Flex always does
pad compensation (PCM level boost)
• V.90 can learn any pad value, pad
compensation is optional for V.90
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
42
V.90 Digital Pad Compensation
We Can Boost the PCM Amplitude
if We Know There’s a Pad
6dB Pad
Lookup
Table
“I Can Built a
PCM
Constellation
that
Compensates
for the
Circuit
Padding”
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
Codec
-13dBm
Analog
“OK, well I Hear a 6dB Pad,
So Boost Your PCM Signal”
43
V.90 with and without Pad
Compensation
On This Circuit Path with a 6dB Pad, Turning
on Pad Compensation Earned an Extra 4kbps
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
44
What V.90 Wants: PSTN Circuit
• Basically, all V.90 really wants from
the PSTN is a 64k clear channel data
circuit, for the price of a POTS call!
• But, unlike 64k ISDN, V.90 can
tolerate digital impairments, such as
pads, a-law-to-u-law conversion, and
robbed bits
• In the future: bidirectional V.90-like
PCM? (“V.91”)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
45
What V.90 Wants: Local Loop
< 3 Miles Good Old Twisted Pair Copper
PSTN
< 3 Miles UTP
or
PSTN
Digital
Darrier
“Integrated” SLC
(Subscriber Line Concentrator)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
46
What V.90 Does Not Want:
Local Loop
Load Coils (Every 18kft on Long Loops)
PSTN
or
Channel
Bank
“Non-Integrated” AKA
“Universal”SLC
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
PSTN
Digital
Carrier
47
What V.90 Wants: House Wiring
Walljack
Short Cord
Run Straight
to the Modem
Local
Loop
Each Line on a Separate
Unshielded Twisted
Pair, or Only One Line
Per Quad Cable
Note: V.90 does Not Require Any Particular House Wiring
(in Fact, My Wiring Looks a Lot Like the Following Slide—
Yet I Get Good V.90 Nonetheless)
But: Poor House Wiring CAN Make the Difference
between Good V.90 and Mediocre V.34
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
48
What V.90 (or V.34 for that Matter)
Does Not Want: House Wiring
Splitter
4
Walljack
1
NIU Local
Loop
2
3
5
1. Crosstalk from Two Lines in the Same Quad Cable
2. Corroded or Shoddy Connectors
3. Bridge-Tapped Line Running Parallel to
Fluorescent Light
4. Flat Silver Satin Parallel to Power Cable
5. Extraneous Equipment with Dubious Electrical
Characteristics on the Same Line
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
49
Hooking Things up
Right and Proving It
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
50
Order the Right Digital Line
• Get a digital path into a digital switch not into a
channel bank
MICA won’t work right when connected into a channel
bank (no near-end echo cancellers
Microcom can work, but only V.34—no PCM
connects possible
• Use PRI vs. RBS if you can
PRI is 64Kb clear channel, gives good DNIS, ANI, crisp
channel cut-through
RBS is 62-2/3Kb (each additional RBS trunk adds
another 1333Kb of degradation, down to 56K), restricted
functionality, sloppy cut-through
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
51
Order the Right
Digital Line, (Cont.)
• Order the right signaling
If PRI, it usually just works, (We are user-side PRI,
not network-side—some PBXes can only be
user-side)
If RBS, we recommend e&m-fgb (wink-start)
With e&m-fgb, MICA (not Microcom) can
handle incoming DNIS (not ANI), in MF or DTMF
(not pulse)
With e&m-fgd, MICA can handle incoming DNIS
and ANI, but there are some restrictions
If R2…get to know your switch guys real well, and
you’d better get an E2 analyzer ready
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
52
Hook It Up Right—
Prove the Digital Line
• Get the line buildout right, use “cablelength
long” or “cablelength short” (DSX1) if
needed, (no “cablelength short” on AS5200)
• Make sure your T1/E1 is (almost) perfect
• Extended BERT with various patterns
is nice
• T1 errors can show up on the switch side but
not on the NAS side
In some Cisco IOS versions, NAS “show controller”
can fail to show errors that it really receives
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
53
Hook It Up Right—
Prove the Signaling
• Make a handset call and be sure to
hear the good music (watch out for
noise due to RBS/R2 mixups)
debug serial int
show contr t1 [/e1] ! show ABCD state
conf term
service timestamp debug date msec
service internal
modem-mgmt csm debug-rbs ! see ABCD changes
debug modem csm ! see MICA DNIS/ANI decode
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
54
Hook It Up Right—
Prove the DS0s
• If fortunate enough to be doing PRI, then place a
sync data call thru and verify that data moves
flawlessly thru the B channel
This proves the DS0 path from the NAS to the LE (and
hence proves the digital carrier—T1/E1—as well)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
55
Hook It Up Right—Prove Voice
Path Thru the LE
• The modem equivalent of a ping to the
next-hop router
• Place a reverse telnet modem call out to
the LE and back in again—should get a
good 33600 (clear-channel DS0s), 31200
(RBS) or 26400 (channel bank) connection
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
56
Reverse Telnet Test
$ telnet 172.16.24.116 2055 ! NAS IP, 2000+line
Trying... Connected.
access-3 line 26 MICA V.90 modems
atdt 5703932
CONNECT 33600 /V.42/V.42bis
access-3 line 52 MICA V.90 modems
router>term len 0
router>show modem log ! move some data
router>show modem op 1/1 ! µcom: “modem at”
Param #9 TX,RX Bit Rate: 33600, 33600
Param #11 TX,RX Symbol Rate: 3429, 3429
Param #21 Signal Noise Ratio: 41 dB
Param #26 Far End Echo Level: -68 dBm ! <-55
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
57
Hook It Up Right—
26/3600 NM-AM
• In general, things are much simpler when
hooking up an analog modem such as the
analog modem network module (NM-AM)
for the 2600/3600
• However, I have one important point for
hooking up an NM-AM system
Be absolutely sure to use a properly grounded
power source; otherwise your modem calls will
suffer from a horrible buzz
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
58
Getting Things Basically Working
• At this point, you have proven that the NAS
and its connection to the PSTN are healthy:
T1/E1 is proven to be (nearly) flawless
You can make a strong V.34 modem call thru the
T1/E1, to the local switch, and back into yourself
You are running known good Cisco IOS and
firmware/portware
• Now, let the users dial in and see how
they do
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
59
First, Look at the Big Picture
• Perfection is unattainable in this world
Expect a CSR of 95% (V.34) or 92% (V.90) ,
maybe 10% premature drops
• Use:
show modem summary
show modem connect-speeds
show modem call-stats
To get the overall picture
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
60
Data to Gather for Trend Analysis
• Now that everything’s perfect on the NAS and in the
digital line to the switch…it’s time to gather data;
for each problematic call, get the following info if
possible:
show modem op (MICA) / AT@E1 (Microcom) while connected
show modem log for the session of interest (after
disconnect)
ANI (caller’s number)
Time of day
Client modem hardware / firmware revision
Interesting info from client (after disconnect)—ATI6, ATI11,
AT&V, AT&V1, etc.
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
61
show modem summary
Usage
0%
Incoming calls
Succ
Fail Avail
6297
185
64
Outgoing calls
Succ
Fail Avail
0
0
0
Busied
Out
0
Failed
Dial
0
No
Ans
0
Succ
Pct.
97%
• No ans: call came into modem but modem
did not go offhook (CPU too busy?)
• Fail: modem went offhook but modems
failed to train up
• Succ: modems trained up; Cisco IOS saw
“DSR” go high (still doesn’t mean that
PPP negotiated successfully, etc.)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
62
show modem call-stats
compress retrain lostCarr rmtLink
#
%
#
%
#
%
#
%
9
41
271
3277
Mdm
Total
trainup hostDrop wdogTimr inacTout
#
%
#
%
#
%
#
%
7
2114
0
0
• rmtLink is good—it means that EC was in effect, and the
client DTE decided to hang up
• hostDrop (aka dtrDrop) is usually good—it means that
the host DTE (Cisco IOS) decided to hang up
Idle timeout
Circuit clear from the telco
PPP LCP termreq from the client
(It’s hard to figure out dtrDrop reasons without AAA accounting)
• Other reasons are bad—should be < 10% of total
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
63
show modem connect-speeds
as1-icg>show modem connect 33600 0
Mdm
26400 28000 28800 29333 30667
Tot
614
0
1053
0
0
31200
1682
32000
0
33333
0
33600 TotCnt
822
6304
• Look for a healthy distribution of
V.34 speeds
A peak at 26.4 (CT1s in the boonies) up to 31.2
(if using ISDN) is normal
• Look for a smattering of K56Flex,
V.90 speeds
No PCM connects? Network topology problem
• Few connects at impaired V.34 speeds
(21.6, 16.8–19.2 is a special case)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
64
DCE Speed Distribution
Nconections
Rx
Tx
0 1
0
0
0
0
2
0
0
0
03
0
0
0
0
4
0
0
0
0
5
0
0
0
0
6
0
0
0
0
S
p
e
e
d
(
b
p
s
)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
65
modem call-record terse
(11.3AA/12.0T)
*May 31 18:11:09.558: %CALLRECORD-3MICA_TERSE_CALL_REC: DS0 slot/contr/chan=2/0/18,
slot/port=1/29, call_id=378, userid=cisco,
ip=0.0.0.0, calling=5205554099,
called=4085553932, std=V.90, prot=LAP-M,
comp=V.42bis both, init-rx/tx b-rate=26400/41333,
finl-rx/tx brate=28800/41333, rbs=0, d-pad=6.0
dB, retr=1, sq=4, snr=29, rx/tx chars=93501/94046,
bad=5, rx/tx ec=1612/732, bad=0, time=337, finlstate=Steady, disc(radius)=Lost Carrier/Lost
Carrier, disc(modem)=A220 Rx (line to host) data
flushing - not OK/EC condition - locally
detected/received DISC frame -- normal LAPM
termination
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
66
show modem operationalstatus (MICA)
•show modem operational-status dumps
the parameters pertaining to the current (last) connection
http://www.cisco.com/univercd/cc/td/doc/
product/software/ios120/12cgcr/dial_r/dr
prt1/drmodmgt.htm#xtocid10451279
• modem at-mode AT@E1 is the analogous feature
for Microcom
http://www.cisco.com/univercd/cc/td/doc/
product/access/acs_serv/5300/mod_info/at
cmnds1.htm#23729
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
67
Client-Side Performance Data
It’s very important to gather client-side
performance data, to find client-specific
trends:
• Client hardware model, firmware version:
ATI3I7
• Client-reported disconnect reasons:
ATI6 or AT&V1
PC’s modemlog.txt, ppplog.txt
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
68
Analyze the Performance Data
• Now that you have collected and
understood the performance data for
your modem system, it’s time to look
at any remaining patterns/
components that may have room
for improvement
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
69
Bad Modems/Channels?
• Are problems associated with particular
NAS modems? (show modem)
• Are problems associated with specific
DS0s on your digital line? (show
controller t1 call-counters)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
70
Problems with Particular
Server Modems
• Use show modem, show modem call-stats and look for
any modems with abnormally high rates of trainup failure or
bad disconnects
• (MICA) If adjacent pairs of modems look hosed, then
probably a hung/dead DSP problem. Use copy flash
modem to the affected HMM to recover. If not using portware
2.6.2.0, please upgrade ASAP
• Verify that all modems are correctly configured. Use modem
autoconfigure type with debug confmodem to ensure
correct settings. May need to reverse telnet to fix up
modems that are badly misconfigured
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
71
Problems with Particular DS0s
• Bad DS0s are rare but possible. (Always the
telco’s fault, never ours :-)
• Use show controller t1 call-counters
• Look for any DS0s with abnormally high TotalCalls
and abnormally low TotalDuration
• Busy out DS0s (isdn service dsl, ds0
busyout) to target specific suspected bad ones
TimeSlot
1
2
3
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
Type
pri
pri
pri
TotalCalls
873
753
4444
TotalDuration
1w6d
2w2d
00:05:22
72
Trend: Bad Circuit Paths
• If:
Long distance calls have problems but
local do not (or vice versa)
Calls at certain times of day have
problems
Calls from specific remote exchanges
have problems
• Then:
210
1081_06F9_c1
You might be getting bad circuit paths
through the PSTN
© 1999, Cisco Systems, Inc.
73
Trend: LD Bad, Local Good
• If long distance calls are bad but
local calls are good, then:
Double-check to make sure that the
digital line connects into a digital switch
not a channel bank
Check with telcos to examine the circuit
paths used for long distance
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
74
Trend: Some Calling Areas
Have Problems
• If calls from specific geographical
regions/exchanges tend to have
problems:
Learn the network topology from the telco
If multiple A/D conversions (nonintegrated SLCs, analog switches) are
used to serve an area, then PCM modem
connects will be impossible, and V.34 may
be somewhat degraded
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
75
Trend: Particular Client Modems
Have Particular Problems
• So now the great majority of calls
connect in the desired modulation
with the desired speed
• …but there are still a few particular client
modems running some particular firmware
on some particular POTS lines calling via
some particular circuits which are failing
to behave as desired…
• So you can now adjust to living with
imperfection, or else move in for
hand-to-hand combat
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
76
Living with Imperfection
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
77
Symptoms of a Suboptimal
Connection
An Individual Modem Connection Will
Manifest Various Symptoms of
Suboptimality:
• Failure to train up
• Train up, but no EC
• Don’t train up in
the desired
modulation
• Inadequate
performance
(throughput)
• Don’t train up in
the desired speed
• Premature
disconnection
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
78
The Underlying Causes
There are Three Fundamental Causes of
Imperfect Modem Connections, the First
Two Causes are Just Corollaries of Each other:
• Poor circuit quality (relative to the
desired modulation/speed)
• Excessive modulation/speed (relative
to the circuit quality)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
79
The Underlying Causes (Cont.)
The Third Underlying Impediment to Perfect
Modem Connections is What One Can Call
Modem Quality. There are Several Aspects
to Modem Quality:
• The ability of the modem optimally to resolve the
data from the analog signal received
• The ability of the modem to select the optimal
receive modulation/speed for the given circuit
• The ability of the modem to interoperate nicely
with the vast and evershifting range of peer
modems (of various quality) encountered in
the field
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
80
Addressing the Underlying
Causes—Circuit Quality
• Circuit quality problems?
• Reengineer the circuit
• This may prove difficult, as it may require
pulling new cable, replacing equipment,
involving multiple providers, paying for
new carrier facilities, etc
• To locate the problematic portion of the
circuit, experiment with various call paths
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
81
Addressing the Underlying
Causes—Modem Quality
• If all modems were of infinite quality, then,
controlling for circuit quality, they would always
select the optimal modulation/speed
• If you hold the peer modem and circuit constant, you
can vary the modem to find another of higher quality
(for that particular circuit and peer)
• If you have quality issues with a modem, you may
contact the vendor for improved firmware, etc. See
http://56k.com for links to client firmware updates
(http://808hi.com/56k/x2-lucent.htm
for LT winmodem)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
82
Testing against Known Good
Modems on a Known Good Circuit
• See if the client modem/circuit is capable of
good V.90 connections by dialing a known
good NAS with digital modems, on a known
good circuit
• Cisco’s test AS5300 (MICA and Microcom
modems) at +1 408 570 3930 / 3932
(http://www.cisco.com/warp/public/471/83.html)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
83
Addressing the Underlying
Causes—Excessive Speed
• Given that circuits are in fact
imperfect, so that they cannot always
attain the ideal speed, and given that
some modems are imperfect, so that
they will not always choose the
optimal speed for the imperfect
circuit…
• The remaining option is to help the
modems choose a better (slower)
speed for the imperfect circuit
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
84
Detuning Modems
• Given a pair of modems and a circuit, they
usually can (and sometimes will) select a carrier
speed that yields an unstable connection
• A nominally higher DCE speed may produce
worse performance than a nominally lower one,
due to retrains, EC retransmits and lost carrier
• If you cap the modems to restrict their range of
modulations and/or DCE speeds, they may be
able to train up (where they had been unable to
do so before), or may achieve more stable
connections
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
85
Detuning Modems (Cont.)
• You can cap the DCE modulations/speeds at
the server side or the client side
• Server side caps will improve stability across
the board, without requiring configuration
effort on the client side, the downside is that
this must be a lowest-common-denominator
setting, and will therefore reduce performance
for some {modem, modem, circuit} tuples that
could handle a higher speed
• Client side caps can be used to achieve
optimal performance on a link-by-link basis,
the downside: greater configuration effort is
required (often by unsophisticated users)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
86
Detuning Modems (Cont.)
• Normally, the DTE speed will not directly affect
the stability of a modem connection
• A properly implemented DTE will be able to
handle the full speed (e.g. 115200 bps), and will
use flow control as needed if it can’t keep up.
• However, it has been reported that that some
Windows systems can prematurely disconnect
unless the DTE speed is reduced and/or the
async serial buffer sizes are reduced.
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
87
Server Side Caps (Cont.)
• Consider (carefully!) configuring the server
modems so as to be less aggressive and
more persistent:
210
1081_06F9_c1
MICA:
S19 (EC retransmission limit), S34 (Fall-forward
timer), S38 (Lost carrier timer), S34 (fall-forward
timer), S40 (retrain threshold)
S54=26 (rate-cap overaggressive clients, enable
Signal Quality Checking) (2.6.2.0)
Increase S32 (SQ threshold) to 3 (more conservative
Rx speeds across the board) (pre-2.6.2.0)
Microcom: Increase S10 (Lost carrier timer), disable
rate renegotiates (:t110=0), increase EQM trip
threshold (:t34)
© 1999, Cisco Systems, Inc.
88
Client Side Caps
• This is the same idea as server side caps:
At a gross level, you can successively disable
modulations till you find one that works
At a fine level, you can keep stepping down through the
DCE speeds supported by the top modulation, till you
find a stable one
Remember: always test using a terminal program (e.g.,
Hyperterminal)…PPP just gets in the way
• Rockwell (ITU V.25ter): use the +MS command to
control the modulations and DCE speeds
Example: AT+MS=12,1,300,36000
Allow V.90 and lower modulations; min speed 300;
max speed 36000
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
89
Client Side Caps (Cont.)
• TI (USR/3COM): use &U to set minimum DCE
speed, &N for maximum
Example: AT&U4&N11
Min speed 4800; max speed 21600
• PCTEL: use S37 with N0 for max DCE speed
Example: ATN0S37=12
Max speed 44k if V.90, 56k if K56Flex, 31200 if V.34
• Mwave: use S28 to cap DCE speed
Example: ATS28=15
Max speed 16800
• Lucent and Hayes: check docs
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
90
Hand-to-Hand Combat
with Individual
Connections
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
91
So You Really Want to Wrestle
with a Modem Problem
• If you really want to get to the bottom
of an individual modem problem,
you’ll want to:
Get your hands to the AT prompt at the
client modem, while it’s attached to the
POTS line of interest
Be prepared to gather a .wav file of the
trainup music
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
92
V.90 Training
Data
V.8 bis
V.8 (Phase I)
Line Probe
(Phase II)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
Half Duplex
Echo Cancel and
Equalizer Training
(Phase III)
Full Duplex
Training
Digital Impairment
Learning (DIL)
93
Negotiation Overview
Rest of V.8 Bis.
K56Flex
Rest of V.8 Training
V.90
V.90 + V.34
K56Flex
Phase Reversals
and 15 Hz Modulation
Phase Reversals
V.34
V.22 bis
2100 Hz ABT
Off V.8 Bis
Hook Tone
V.42
(LAP-M)
Optional
MNP-3/4
Optional
V.22 bis
2100 Hz
Tone
V.32
Async.
mode
Optional
V.32
FSK
MNP-2/4
Complete
Training
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
94
V.34 Training
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
95
K56Flex Training
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
96
V.90 Training
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
97
Listening to the DILs
• The DIL (Digital Impairment Learning sequence)
is the musical score (PCM sequence) that the
V.90 analog modem tells the digital modem to
play back to it, so that it can discern any
digital impairments in the circuit; (such as
multiple D/A conversions, a-law/μ-law,
robbed bits, digital pads)
• If you don’t hear the DIL, then the modems
did not negotiate V.90 in V.8/V.8bis (i.e., a
modem compatibility issue; if you DO hear
the DIL, but then a retrain in V.34, then the
analog modem decided, on the basis of the
DIL playback, that V.90 was infeasible
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
98
Some V.90 DILs
Pctel 7.55
USR (3COM)
LT winmodem
5.28
Rockwell (Conexant) 2.2
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
99
Symptom: Trainup Failure
• Is the symptom a failure to train?
Does the music have noise in it? If so,
then clean up the circuit
Does the client give up quickly, without
running V.34 training? E.g., perhaps it
flips out when it hears V.8bis Cre; in
which case, try disabling V.8bis (hence
K56Flex) on the server (if acceptable), or
get new client firmware, or swap it out
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
100
Symptom: Trainup Failure (Cont.)
• Does the client try training in an
advanced modulation (V.90, K56Flex,
V.34), but eventually give up? Then
try capping the client at successively
lower modulations/DCE speeds
• (Or improve the circuit, or improve
the modem code, of course)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
101
Symptom: Not the Desired Speed
• Does the client train up in the desired modulation,
but at a slower than yearned-for speed?
• If so, then simply coercing the modems to
connect at a faster nominal DCE rate would
nearly always be a grave mistake
• The best solution to this problem will normally be
to improve the circuit; improving the modem code
will rarely yield results here (the exception being a
modulation with much recent churn, such as V.90)
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
102
Symptom: Trainup, but No EC
• Does the client train up in the desired
modulation, fail to negotiate error control?
• If so, then the probable cause is high BER in
the selected modulation, resulting in data
damage or retrain during the sensitive EC
negotiation phase, in this case, the useful
approach will be to detune the client for a
slower modulation/DCE speed
• Another possibility is modem bugs in EC
negotiation or late modulation trainup; (seen
with some popular client firmware), in which
case, pursue improved client firmware
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
103
Symptom: Inadequate Throughput
• Is inadequate throughput observed thru the
modem link (given the chosed modulation/speed)?
• If so, then first rule out the possibility of higherlayer protocol deficiencies, (e.g., do a simple ping
thru the link)
• Monitor the EC retransmits (block errors) on the
link, if they are too high (> 5% of frames), then
modem detuning is in order, similarly, if the
slowness is due to excessive retraining, then
detuning should help
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
104
Symptom: Premature Disconnection
• Does the modem link terminate before the
user wanted it to?
• If so, then first ascertain whether the
disconnection was initiated by one of the
DTEs, if so, then fix the DTE application
• If neither DTE initiated the disconnect,
then (as always), modem detuning, circuit
improvement, and/or a firmware upgrade,
should yield good results
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
105
Symptom: Premature
Disconnection (Cont.)
DCE
DTE
DCE
DTE
• Monitor the DTE links to see who initiates the disconnect
• Using the old RS-232/AT interface model, a DTE-initiated
disconnect will show as a DTR drop or as “+++ATH<CR>” on
the Tx lead, a DCE-initiated disconnect will show as a DCD drop
(which Cisco IOS calls “DSR”)
• The ideal is for all disconnects to be voluntary—i.e., initiated by
one or the other of the DCEs
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
106
Symptom: Premature
Disconnection (Cont.)
Tracking Whether DCE or DTE Initiated
the Drop - from the Cisco IOS Side
• “Debug modem” will tell the truth, however, if the
drop was initiated by a PSTN circuit clear, this
will show as a DTR drop (because, on a system
with digital modems, the DTE handles the PSTN
interface), so also check the PSTN signaling
(“debug isdn q931”, “modem-mgmt csm
debug-rbs”)
• Use “modem call-record terse” to get all the
good info in one nice package, if the disconnect
was a pure Cisco IOS-side decision, then this will
give the Cisco IOS explanation
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
107
Symptom: Premature
Disconnection (Cont.)
Tracking Whether DCE or DTE Initiated
the Drop - from the PC Side:
• Turn on the TAPI modem log (control panel ->
modems -> properties -> connection ->
advanced, or some such)
• After the disconnect, see whether the disconnect
was initiated by the PC or by the modem; (don’t
take TAPI’s interpretation of events very
seriously, it tends to lie)
• If it was the PC that initiated the disconnect, next
contact the friendly folks at Microsoft who will be
eager to help solve the problem
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
108
Symptom: Premature
Disconnection (Cont.)
DTE-Initiated Example: the PC User (Me)
Clicked the DUN Disconnect Button
• “modem call-record terse” on Cisco IOS showed:
disc(radius)=User Request/Received Terminate
disc(modem)=A220 received DISC frame -normal LAPM termination
• PC’s modem log showed:
Hanging up the modem.
Hardware hangup by lowering DTR.
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
109
Symptom: Premature
Disconnection (Cont.)
DCE-initiated example: the PC user (me) unplugged the
phone cord while in the midst of transferring data
• “modem call-record terse” on Cisco IOS showed:
disc(radius)= (n/a)/(n/a
disc(local)=9 DTR Drop
Alas this doesn’t tell the REAL reason why Cisco IOS dropped the
call—that the caller line went loop-open, (note to self: DDTS this)
• PC’s modem log showed:
Remote modem hung up.
Recv: <cr><lf>NO CARRIER<cr><lf>.
Alas this shows that TAPI has no way of knowing what’s really
going on—it’s default behavior when it sees a network-side
disconnect is to blame the remote modem!
• If you’d been listening to the voiceband at this time, then
you would have known what happened!
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
110
For Further Insight
• Gilbert Held, The Complete Modem Reference, 3rd Ed.
• The cisco-nas mailing list (email
cisco-nas-request@datasys.net; archives at
http://cisco-nas.datasys.net)
• John A. C. Bingham, The Theory and Practice of
Modem Design
• John G. Proakis, Digital Communications
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
111
Please Complete Your
Evaluation Form
Session 210
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
112
210
1081_06F9_c1
© 1999, Cisco Systems, Inc.
113