A FULLY IN-BAND IVIULTITONETEST FOR
TRANSIENT INTERMODULATION
DISTORTION
by
Robert R. Cordell
Bell- Laboratories
HOImdel, NJ 07733
ABSTRACT
test for
A specially deslgned three tone lntermodulatlon
(TIM) 1s proposed.
translent
intermodulation dlstortion
The test employs three tones to simul-taneously produce
beat distortlon
even-order difference and odd-order triple
products near I kHz-, and resuLts 1n extremely simple, Vet
advantage of the
sensltive
lnstrumentati-on.
A signlflcant
test is that the stimulus and response both li-e 1n-band.
INTRODUCTION
lntermodulatlon
The heightened awareness of translent
(TIM)L,2,3 brlngs wlth lt the need for efflcient
distortion
and economical testlng technlques which can obJectively
measure this dlstortion
while providlng good correlation
Three tests are
wlth subJectlve l1stenlng test results..
currently
in use:
the slne-square DIMr4 20-kTz THD,2 and
t h e n o ' ia r i t v - m o d u l a t e d s a w t o o t h . )
Each of these tests
has various lndivldual
disadvantages (e.g., expensive
j-nstrumentatlon, unrealistic
test slgnal characterlstlcs,
insensi-tivity
to I'softil TIM, etc. ) , but they all suf fer
from the fact that elther much of the stlrnufus (test
products)
signal) or much of the response (dlstortion
Thls can cause
spectra lie considerably out-of-band.
r e s u f t s 1 n m a n yb a n d - } l m i t e d
testing problems or unreallstic
good examples are tape
Particufarly
or nonflat systems.
Another example of
recorders and phono preampllflers.
growing imporLance is dlgltal
audio systems, where antla'l i asi nc
cutoff
f i I ter"
just
ero
omn'lnrrarl WhlCh
intrOdUCe
a
Very
Sharp
above 20 kHz.
of the CCIF
The test proposal here i-s a modification
The CCIF test 1s
frequency IM test.
two-tone difference
adequately sensltlve and does not require expensive
but has the fatal disadvantage of aftnost
instrumentation,
total lnsensitlvity
to odd-order distortlon
mechanlsms. To
retain the advantages and el_lminate the odd-order lnsenslproblem, a thlrd ton,e is added. The frequencies
tivlty
-l-
are selected to simultaneousry prod.uce a triple-beat
product
(A-!-C) sllehtIy
below I k]Hz and a difference-frequenby
product (B-C) slightly
above I kHz. We will refer to the
new test as the I'multitone intermodulatlon
test'r (MIM).
Because the peak slgnal slope of the test ls not grossly
1n excess of that of worst-case program signals, and belause
the measured product fal1s 1n the most sensltive portion of
;he audible spectrum, good subjective correlation
can be
expee!ed.
THE NATURE OF TIM
In discussing audio amplifier
dlstortlon
issues, it is
convenient to divlde dlstortion
types lnto two categories:
statlc and dynamlc.4 static
dlstortion
fevels depend only
on the amplltude of the si-gnaI, while dynamic dlstortlon
levels depend on both the amplltude and frequency of the
slgnal.
Arternatlvely,
one may say that dynamlc dlstortlon
ls a function of the rate-of-change of the signal.2
In
general, 1ow-frequency dtstortion
t e s t s , s u c h a s S M P T EI M
and 1-kHz THD, measure static
ampllfier
dlstortions,
whlle
high-frequency distortlon
tests, such as ZO-kHz THD and
CCIF (e.g., f9- and 20-kHz dlfference
tone) measure dynamic
ampl1f1er distortions.
Transient lnternodulation
dj-stortion
(TIM) is a type of dynamic intermodulation
prodlstortion
duced by a partlcular
mechanlsm.
In any casej distortlon
produced by a nonis inevitably
linearlty
sornewhere in the ampllfler,
and a given nonlinearity
will produce measurable dlstortlon
products (elther
harmonic or intermodulation)
with any test which adequately
exerclses the nonlinearity.
As an example, a simple high
frequency sinusold adequately exerclses the TfM mechanism
1n an anplifier
a n d p r o d u c e s a n e a s 1 1 y m e a s u n a b l - eT H D
product indlcatlng
the presence of TfM.2
The TIM mechanism is quite simple and is easily understood
by referring
to the highly simpllfied
power amplifier
schematlc shown 1n Fig. 1.
The differential
pair conslsting
of Qt and Q2 acts as a transconductance amplifler.
Its
a'rrnan'F
nrr{-n',+u
C ri f !i V
u
v se !S)
tuhr er s
bu ao S
o se
n
u .f ir ur a
u rf r i
tu
n
f
va
n In' pg iu. Il 7 , fi rvr ge Irp
O
k {?J ,
whlch ln turn drives the output stage.
At m1d and high
frequencies the predriver
stage acts as an integrator
due
to the action of compensating capacitor Cl; the slgnal
current from Ql is integrated bo form the drive voltage
for the output stage.
As a result,
the slgnal level in
the first
stage is proportional
to frequency for a given
n
rrrnr'F
vqvy*v
is
r
Tn
the time
rr-ra time
derivative
domain
the
sisnal
of the output
-2-
'in
the
fJrst
s tvaq 6r ur e
u
sI-gna1, and so 1ts
Disto output rate-of-change.
amplitude 1s proportional
produced by ampllfier
stages such as these which
torfion
are situated prior to the point of frequency compensati-on,
and whlch 1s a funcflon of the rate-of-change as a result '
(TIM).
intermodulabion distortion
is calLed transient
pair is characterlzed by a well-behaved
The dlfferentlal
(the tanh functlon).
Thus,
symmetrical statlc nonlinearlty
mechanism made dynamic in
TIM ls a static distortion
1s situated in
behavlor by the fact that the nonlinearity
where slgnal amplltudes are proa part of the circuit
in the
portlonal- to frequency (or output rate-of-change
stage is driven
ff the dlfferential
tlme domain sense).
to clipping by the slgnal amplltudes lt must handle, then
occurs,
the well-known phenomenon of slew rate llmiting
If the slgnal ampli-tudes
and so-ca1led 'rhard" TIM resul-ts.
but not
in this stage are enough to cause distortlon
subsfewlng TIM 1s
then so-calledrtsofttror
clipplng,
produced.
The mechanism 1s simple, well behaved, and well
understood.
There is no mystery about it.
Because TIM depends on the rate-of-change of the program
appralsal of TIM mechanlsms and measurslgnal, a reallstlc
1ng technlques requlres some conslderatlon of expected
program characteristics.
Based on discussi-ons elsewhere,
it can be concl-uded that program sources are not as fast
(peak
as some mlght thlnk, a normallzed rate-of-change
(
v
/vs)/v
p
e
a
k
t
o
0
.
9
7
5
amplltude ratlo)-ol
slope to
9.05
This corresponds to
seeming to be the upper l-lmlt .6 r'[ 'd
a power bandwidth of l-ibtle more than 10 kHz or so, and
l - e s s t h a n ' a 3 V , / U Sr a t e - o f - c h a n g e f o r a 1 0 0 - w a t t a m p l - l f l e r .
even on
slew rate ltmltlng'
This suggests that outright
extremely
T
I
M
,
i
s
a
n
l
e
v
e
l
s
o
f
h
l
g
h
w
l
t
h
f
a
i
r
l
y
ampllflers
rare evenL. The real problem is clearly subsfewing TIM.
out that what we hear as TIM
Tt is afso worth polntlng
products whlch show up in
1s undoubtedly intermodulation
the midband as the resul-t of complex lntermodulatj-on
interactions
among several hlgh-frequency components.
Thls observation is lmportant when evalua"tlng the effect
of feedback factor on TIM, because feedback factor is a
products
functlon of frequency and reduction of dlstortion
by feedback depends on the feedback factor at the product
f z ra ne rY rq av rn. nv Jr r
a
.
It is tempting at thls point to draw sone concluslons about
in
ahead of power amplifiers
the use of low-pass fllters
f
ilter
s
u
c
h
a
q
u
e
s
t
A
l
t
h
o
u
g
h
for TIM-free reproduction.
the
when the input signal is,
can prevent slew rate limiting
and may improve
s&V, a square wave wlth a l0-ns rlsetlme,
lt will do
performance with certain TIM tests,
amplifier
-3-
little,
if anything, toward prevention of subslewlng TIM
under normal program condltions
because it witl not significantl-y affect the rate-of-ehange of a slgnal characterized by a maximum power bandwldth of only about 10 kHz.
The LPF seems only useful for eliminating
high-level
ultrasonic
disturbances whlch should not exist ln the first
place.
We can conelude from thls that so-called 'rnon-sl_ew
llmitedrr ampllfler
deslgns (whlch often incorporate an
lnput LPF to guarantee that slewing cannot occur, regardless
of the input slgnal rlsetime)
do not necessarily achieve
subjectively
better TTM performance than ordi"nary deslgns.
EXISTING TESTS FOR TTM
Three tests now exist for measuring bhe TIM performance of
an ampllfier.
They are the s14e-square DIM test proposed
by Otala,q hlgh-frequency THD,z and the polarity-modul_ated
sawtooth recently proposed. by Sansui.5
ft should be
polnted out that none of these tests makes any dlstinction
between TIM and any other form of dynamic intermodulatlon
dlstortlon;1t
is not llkeIy
that the ear does so, either.
An important characteristic
of all dynami-c lntermodulatlon
.tests is peak rate-of-change of the test slgnal, since this
is the real- functlon drlving the nonl-inearity,
rather than
amplltude alone.
As we wll-l- see shortly,
some .of the
tests resort to rather hlgh rates-of-change to achi_eve
acceptable sensi-tivlty.
The measurement fLoor of a test is just as important in
assessing the merlt of a test as the sensltlvity.
A test
whlch produces large numbers but whlch also has a high
measurement f100r is not really
more sensitive
or merltorious than a test which produces smarler numbers and has a
correspondingly
lower measurement floor.
fn a sense, what
we are really
concerned about is the dynamlc range of the
test.
ft must be recognized that the varlous distortion
tests
produce a number which ls merely a metrlc for actual performance, and whose absolute value 1s of limlted
significance
when compared to a number generated by a different
dlstortion
test.
r f a n l n d i v i d u a l - c l - a i m s t o b e a b l e t o h e a r 0 . 0 1 _n e r e e n t
THD, he ls merely saying that he can hear the effect of the
distortion
products produced. und.er program conditions
by an
amplifier
whlch measured 0.01 percent rHD in the laboratory.
pIM-30
The slne-square test was,,pfgposed specifieally
for testlng
TIM in audio ampllfiers.Ll,10
T h e t e s t c o r n bj . n e s a 3 . f 8 - X H Z
-4-
4:1 peak, amplitude
square wave and a 15-kHz sine wave ln a
b
y
a single pole
f
i
l
t
e
r
e
d
'
i
s
w
a
,
t
e
s
q
u
a
i
"
t
h
e
and.
D
I
M
-30, denoting
" i i i o:,o k H z . T h e t e s t i s r e f e r r e d t o a s
w
i
t
h
3
0
kHz filtering'
distortlon
" iy n a m i c l n t e r m o d u l a t i o n
d
Amorestringenttestwibhl00-kHzfllterlnghasalsobeen
products of 15 kHz
various lnrermodutation
;";;;;";:4-"Fh;
are measured on
i.Ur--i"d 1rs lnreger multlples
;i;t-3:tg
r
a
s
hion and referred
R
M
s
summed1n an
. ,p"6t"um analyzer,
percentage.
d
l
s
tortlon
a
a
t
a
r
r
l
v
e
t
o
to the I5-kHz alnplriude
(V/vs)/v
high normallzed rate-of-change of 0.32
The fairly
d
i
s
t
o
r
t
i
o
n
d
y
n
a
l
i
c
of the DIM-30 test read'iIy excites
l
s not
t
e
s
t
t
h
e
o
f
s
e
n
s
l
t
l
v
i
t
y
t
h
e
mechanlsms. However,
r
a
t
e
of-change
b
e
c
a
u
s
e
e
x
p
e
c
t
m
i
g
h
t
o
n
e
g
r
e
a
t
a
s
quite as
-the
i
t
s peak
h
a
l
f
tovrl the stope exceeds
is ritrry
d,riv
slgm
o
r
e
A
'
t
i
m
e
"
v
"
r
E
t
h
e
o
f
value for only inorrl 6 percent
h
i
g
!
f
a
i
r
l
y
t
h
e
i
s
t
e
s
t
l
n
i
s
o
f
limitation
nlflcant
a good spectrum anaLyzet wlth 99 dl. ?f
floor:
distortlon
This
dynamic range *irr
viEtd a ftoor of 0.035 percent.o
impalrstheabilityofthetesttomeasuresubslewlngTlM.
heavily on
Because the test signal rate-of-change depends
w
ave' this
s
q
u
a
r
e
ol-rt-of-uu"a (>20 kH;) harmonlcs of the
for
r
e
s
u
lts
o
p
t
l
m
i
s
t
i
c
test can be led to glve somewhat
In
'
f
l
l
t
e
r
s
l
o
w
p
a
s
s
f
r
o
n
t
e
n
d
incorpo"Itfttg
amplifiers
a
nd
useful lnformatlon
general, rroweveri trt" tEst-ylelds
p
e
r
f
o
r
mance.
a
m
p
l
l
f
i
e
r
o
f
spectral display
a very interesiing
TheprimaTydisadvantageof'this.usefultestlstheexpenslve
and exiremely tedlous procedure.
lnstrumentltion
20-kHz THD
appearance ' a slne
Although seemlngly not very dynamic 1n
t
hlgh irequency 1? ttolg..than adequately
wave of sufficrEnLrv
TIM.mechanisms'
amplifler
aynarnic ln terms of-exerclsin[
h
a
ve been suggested
m
e
a
i
u
r
e
m
e
n
t
s
T
H
D
and so htgh_freq"u""v
For singl-e-number tests 2]-kldz
as a good test for fiU.2
the largest
fHO- (fiffO-ZOl ts preferred because it developes
for any
(
0
.
1
2
5
$
/
v
s
)
/
V
)
r
a
t
l
o
peak slope/peak inplitude
in-band slnusoid.
is not quite as
At a glven peak-to-peak amplitude, THD-20
peak sLope;
s
m
a
l
l
e
r
t
h
e
o
f
sensltive as DIM-30 because
an order of
l
e
a
s
t
a
t
i
s
f
l
o
o
r
however, 1ts measirement
a more sensil
s
I
t
magnitude lower than that for DIM-30 '
tivetestforsoftTlMandhasagreateroveralldynamlc
are
with adequate sensltivlty
uiityr""s
;;;;"lg--riiopartlcular
O
f
u
s
e
'
w
l
d
e
s
p
r
e
a
d
1
n
b
u
t
moderatery expen"iuu,
-5-
ls the fact thab most amplifier
slgnlficance
include a quote of THD up to 20 kHz at rated
While the
front-end
bandwidth
a serious
example
specificatlons
power.
wlth
THD-20 test is nob misled by ampliflers
it ls mlsled 1f the overall
low-pass filters,
This can be
of the DUT is not at least I00 kHz.
for
problem in testlng phonograph preampllfiers,
SAWTOOTH
POLARITY-MODULATED
sawtooth measurement technlque
The polarlty-modulated
p
r
o
p
o
s
e
d
b
y
Sansul, its primary advantage
was recently
an
fn practlce'
belng inexpenslve instrumentatlon.)
20-kHz sawtooth waveform has 1ts polarlty
unfiltered
Thus, lts
reversed at approxlmately a 7B-Hz rate.
applireversal,
W
l
t
h
o
u
t
t
h
l
s
r
a
t
e
.
a
t
asymmetry reverses
asymmetrical sawtooth signal
cation of the high-frequency,
even- or od&order
wlll cause the ampllflerfs
to an amplifler
w
h
o
s
e magnitude
g
e
n
e
r
a
t
e
d
c
o
f
f
s
e
t
to
a
nonllnearitles
The dc offset
depends on the severlty of the nonfinearlty.
forned by
ls prlmarily
due to even-order nonlinearltles
of the sawtooth fundaordlnary second harmonic dlstortton
1s
mental-. The dc offset due to odd-order nonllnearitles
between the
prlmarily
due to "2A-8il type intermodulatlon
In the event
sawtoothts fundamental and second harmonlc.
one might
that these offsets are of opposite polarity,
The
readlng could result.
speculate that an optlmlstic
perlodic polarlty
r e v e r s a l m e r e l y r r c h o p s t rt h i s d c o f f s e t
The much
lnto an easily measured low-frequency ac slgnal.
higher frequency test slgnat is el-lmlnated in the analyzer
inexpensive low-pass flltering.
by relatively
from the extremely
Thls test derlves most of its sensitivity
sawtooth edges which it applies to the
sharp unflltered
is
The normal-ized rate-of-change
under test.
ampllfler
(
v
a
n
d
(
p
r
o
b
a
b
l
y
1
0
l
e
a
s
b
a
t
/
V
)
/
v
s
)
extremely high
n^nqprrrpntIv the test sJena'l rate-of-change is very large.
Amplifiers without input LPFs wlth moderate, but adequate,
sl-ew rate, and which test extremely well in the previous
r.t
vvr.vvYqvrrv
fU.spP vsoht s
e A ? , hv Lc
vsrr
e
u ^v
^ Jnv e
v vn
e sf a d
tn sl cw pafg
limit
on
thls
test
and
poor showlng.
The subjective
glve an unjustlflably
of this test is also questionabfe because its
corretation
risetime ls so much faster than that of music.
because this test depends so heavily on extreme
Flna]ly,
signal stop" (and on out-of-bbnd sawtooth harmonics) 1t
witt ne eailty misled in its assessment of TIM (particularly
which incorporate input
substewlng TIM) by amplifiers
Iow-pass fiLters.
The low-pass fil_ter not only reduces
the peak s1gna1 sIope, but also reduces the waveform
asymmetry and second harmonie l-evel, resultlng
1n a
particularly
serlous l-oss ln sensitivity
to odd-order
nonlinearities.
A MULTTTONETIM TEST
Although each of the existing tests has meri-t, each al-so
has some dlsadvantages.
rb seems that the foll-ow1ng llst
of desirabLe test characteristlcs
w o u l _ ds u f f i c e t o d e scrlbe an almost ldeal TfM test:
(1)
lnexpensive
lnstrumentatlon,
(Z) simple measurement procedure, (3)
good sensitivlty
(particularly
to mldband even- and
odd-order IM products), (4) in-band itlmulus and response,
(5) good subjectlve correl-atlon due to musi-c-l-lke test
slgnal rlsetimes.
None of the exlsting tests meet a1l of
these goa1s.
The multitone IM test to be presented here
represents an attempt to get closer to this ideal-.
The multltone
intermodulation
test (MIM) i-s a variation
of
the well-known CCIF two-tone IM test ln which two hlghfrequency tones spaced apart by a small frequency difference
(e.g., f9 and 2O kTz) are applled to the unlt under test
and the difference
frequency IM product 1s measured. The
tone sources need not be very low ln dlstortlon,
and s1mp1e
low-pass flltering
permits separation of the distortion
product from the stlmuIus, thus permltting
exceJ_lent
sensitlvlty.
The ccrF test has been.used wd-th some success
for measurement of hlgh-frequency
dlstortlon,
but it 1s
sensltlve only to even-order distortion
mechanisms. Its
total insensitivity
to odd-order distortion
mechanisms,
which are the more prominent sources of distortion
in many
contemporary designs, render 1t useless as a rellabl_e
general purpose TrM measurement tool.
rf a spectrum analyzer
or other sharp cutoff filtering
technique is used to l-ook
at tle rr2A-Brrproducts as well_, which in this example lie
at 18 and 2L kHz, then the test will_ also be sensltlve to
odd-order products; under these condltlons,
however, the
lnstrumentation
eost becomes high.tl
In order to retain the advantages of the CCIF test white
incorporating
odd-order sensltlvlty,
a third tone has been
added. The three frequencles are chosen so that the three
tones produce a trlple
beat product (A-B-C) at slightty
below I kHz, whlle two of the tones produce a CCIF-like
difference
frequency product (B-C) at slightly
above I kHz.
Specifically,
the three equal-leve1 tones are at 9.00,
-T
-
10.05 and 20.00 kHz, resulting
i-n odd-order products at
950 Hz and even-order products at 1050 Hz. lt should be
emphasized that the dlfference
frequency product contalns
contrlbutlons
from all even-order mechanlsms, not Just
second order.
s1milarly,
the triple
beat product iontains
contrlbutlons
f r o m a l - ] , o d d - o r d e r m e c h a n i s m l n n rv ri r4rPs rw .urr-r hr rnr 6 l
"v
ordgr.
12
The even- and odd-order products are serected to Lie fairly
cl-ose to each other in frequency so that both products can
be passed through a relatlvery
narrow bandpass fil-ter
centered about--L kHz, lmprovlng immunity tb nolse generated
ln the unlt under test.
The photos 1n Flgures 2-6 ilrustrate
the appearance of
the test in the frequency and tj-me domalns.^ Flgure 2
shows the appearance of the undistorted
test si[nal
on a
spectrum anal.yzer (2 kHz/dj-v, 30-Hz BW), wh1]e Flg. 3
lflustrates
the distortlon
products added to the Jpectrum
when the test slgnal 1s passed through a 741 operatlonal
i r c r " un pncer rs "var rt 1i n6 e
sar.rm
r pnf ' lri r rf e
a vt 4a rnr ri r nr vr vr -c r " tvi - r .nbo bc-4.i e4nir trne r e
q
vnrfs t 1 0
at
an
output
]evel of 4.2 v p-p.
The products of
here are
seen at the extreme 1eft of the photo at aboub I kiHz.
Notlce there are two unequal very closely-spaced products
there; these are the odd- and even-order products.
Flg. 4
1g . l'closeuptt centered on t kHz (eOO Hz/aiv, f0_Hz BWf
clearly
showing the odd-order product on the ]eft
and the
smaLler even-order product on the right.
The top osci.l_loscope trace in Fig. 5 i.fl-ustrates the
appearance of the three-tone MrM test slgnal when the sweep
1s synchronlzed to the filtered
and extracted distorti-on
pioducts shown in the lower trace.
Finally,
the dlstortion
products,at 950 and 1050 Hz are shown wlth a slower sweep
in Flg. 6.
The "artplitude-modulatedr! appearance 1s due to
the two products beatlng againsb each other.
If the
products were equal in arnplltude, the apparent modulation
depth would be 1O0 percent.
The distortion
percentage will be deflned here as the value
of the 950-Hz and 1050-Hz distortion
products, measured
together on an average responding ac voltmeter,
referred
to the rms varue of a sine wave of the same peik-to-peak
amplltude as the three-tone MrM test signal.
rn the case
of power amplifiers,
the test power tever wil_t be defined
as the average power corresponding to a slne wave of the
same peak-to-peak amplitude as the MIM test slgnal.
The choice of 1 kHz as the approximate frequency where
products are made to appear i-s somewhat arbit,"^r-n
hrr i s
-8-
choosing a higher
First,
based on two consid.erations.
for a given
frequency band requires sharper fil-tering
T
h
e
s
e
c
o
n
d consideratlon
r
e
j
e
c
t
i
o
n
.
s
l
g
n
a
l
tevet of test
observation that what 1s normally
is based on the earlier
is lnterheard as a result of high-frequency distortlon
rather
m
i
d
b
a
n
d
modulatlon products fotded down into the
most
e
a
r
i
s
t
h
e
B
e
c
a
u
s
e
h
a
r
m
o
n
i
c
s
.
than high-frequency
products of this
near I kHz, the lntermodr:lation
sensitive
with this cholce,
test have been chosen to appear there.
the test accurately takes into account any IM distortion
product reduction at midband frequencies afforded by
negative feedback which may be larger in the midband than
at the upper band edge. THD-20, for example, does not
posslbillty
into account (THD-20 may actually
take thii
p
e
s
s
i
m
i
s
t
i
c
ln thls regard because its secondbe somewhat
amplitudes depend on the feedback
and thlrd-harmonic
factors at 4o tHz and 60 kHz, respectively).
ttimpulsivett 1n
Somemay wonder if this test, not being
nature, wlll adequately exercise and expose TIM dlstortlon
ttyesrrr for the same
mechanisms. The answer is a clear
exercise of the
reasons that 20kHz THD measurements do:
p
eak slew rate of
p
r
l
m
a
r
i
l
y
t
h
e
o
n
d
e
p
e
n
d
s
TIM mechanisms
on the waveshape of the
the test slgnal, and very llttle
The peak slew rate of the three-tone test
test signal.
that of a full-ampli-tude
two-thirds
slgnal Is approximltety
signifi(V/us ) /Y) , stil1
(
0
.
0
8
q
a
b
o
u
t
i
.
e
.
Z}-Lltlz sinusold
,
cantly more than worst-case program (0.05 (V/us)tv) '6'('6
Because of the sma]ler peak slew rate stlmulus compared with
percentages
other TIM tests, somewhat smaller distortlon
o
p
e
r
a
t
i
ng at a given
g
i
v
e
n
n
o
n
l
i
n
e
a
r
l
t
y
a
are produced from
peak-to-peak voltage l-eveL. However, the greatly reduced
measurembnt floor makes up for thls in terms of overall
measurement technique dynamic ran€5e.
Also, because of lts somewhat fower peak test slgnal s1erythan THD-20 or DIM-30
rate', MIM is somewhat l-ess sensltive
'
t
h
a
r
d
"
TIM mechanisms, in whlch an
to certain forms of
smal] slew rate is exceptionally
with a relatlvely
amplifier
./
and then suddenly
r
ate limit
s
l
e
w
t
o
i
t
s
linear rlght up
Such an
generates TIM wtren pushed beyond that 1lmit '
whose slew rate was between 8.4 and t2.5 V/Vs and
inplifier
whose rated output was l-00 V p-p would be a case in point '
Presumably the three-tone test would show no TIM, while
Howe.ver,
THD-20 or DIM-30 wou]d indicate TIM at rated output.
T
I
M
*
p
r
o
ducing
t
h
e
would not enter
wlth program the ampllfier
be
t
h
e
r
e
f
o
r
e
w
o
u
l
d
t
e
s
t
slewing region and the three-tone
Such
T
I
M
f
r
e
e
.
a
s
a
m
p
l
i
f
i
e
r
t
h
e
accuraie ln characterizing
a
d
e
q
u
a
t
e
l
y
t
h
a
n
m
o
r
e
b
e
w
o
u
l
d
a
n
d
u
n
u
s
u
a
l
,
v
e
r
y
a case 1s
handled if the three-tone test resufts are quoted i-n
-9-
Rated slew rate wlll
combination with rated slew rate.
slewi-ng ( i. e. , I'hard" TIM)
always handle the outright
TIM which is what ls usually
condltion;
it is the'rsoft"
heard and for which better measuring techniques are needed.
INSTRUMENTATION
As mentioned earlier,
simplicity
and economy of instrumenLation are lmportant virtues of the MfM distortlon
measuring technique, and this ls iflustrated
by the
simplified
block diagrams of bhe MfM test slgnal generator
:nd
MTM
nrodrrnf.
nnnl
qhor^rn
vz.a'r,
i n
tr'i crnrAs
7
and
fl
rcsnpe-
With the exception of the ac VTVM, the entire
tively.
generator/analyzer
prolobype fits
ln a 3-I/2 by 6 by 10
inch box.
The generator consists of three simple fixed-frequency
Weln bridge oscillators
whose outputs comblne at an inverting summer amplifler
to form the composite test signal.
ground of the summer operationaf
The virtual
amplifler
due to the
minimlzes lntermodulation
among the oscillators
summlng operatlon.
Metal shlel-ds isolate the oscillators
to prevent
from each other and from the summing amplifier
intermodulaLion due to coupling of other signals lnto the
osclllators.
to
A maJor advantage of the MIM technique ls the ability
utilize
inexpenslve signal sources for each of the three
component frequencles without serlous concern for their
The osclllators
harmonic distortlon
characteristics.
a slngle
used here are extremely slmple in deslgn, utilizing
rms
operational amplifler
and a FET. They produce t voIts,
at about 0.0l-percent
THD; larger val-ues of THD would still
were, howevqr, carefully
The oscillators
be acceptable.
dcsi
onpd
tv nv
n
r n vdu ru rU os F
PL
I n
v vrl ^ r n
I r vnr J
eL se
\vr sar r \ /J
f
,
rs
se
thc
sttm
nf
soltrc.o
-
to the
UUT, and anal-yzer noises is the major contributor
produced by
Dlstortlon
measurement fl-oor of the system.
the summer 1s well below bhe system measurement floor.
Because there is over three octaves of frequency difference
between the product frequency and the lowest test slgnal
rf fn
c ua vr qr Lar nr vaJ r, /
\vruFr Jr \ r
cs rhr 4r r! yn
4r rnr ru l
nroniso
IJrvvfeu
f i ' l t - c r . , i n +or r b
if s v
llnnecFssArV
in the MfM product analyzer.
As shown in Fig. 8, the test
processed by a pair of Z-kHz
slgnal from the UUT is first
bo attenuate the
third-order
Bubterworth low-pass fllters
original
test signal components by over 70 dB while lntroThe
of 1 kHz.
ducing no significant
foss 1n the vicinity
voltage follower
are of the simple unity-gain
filters
ru vl a
p fs
6 l ir b qo9 n
rrii I i .z.ins A
sinsl
e
ooerati-Ona1
-10-
amplifier.rJ
EaCh
filter
is followed by 20 dB of gain to mai-ntaj-na good
signal-to-noise
ratlo.
To minimize the product detecti-on noise bandwidth and to
achieve further
attenuation at the test slgnal frequencies,
the test signal from the UUT is next passed through a
fourth-order
Butterworth bandpass fllter
whose bandwidth
is about L50 Hz and whose gain is approximately 10.
A
s t a n d a r d a c V T V I v ic o m p l e t e s t h e M f M p r o d u c t a n a l y z e r .
The
sain
*n
I .
4
of
the
! ,l .r U _r 1- Iv
analvz.er
yn u6 ua ^l .
c a n e - l f i r r ui iL- vr rL v J
the
ic
a
is
set
d j -h^'t
DfBI.ld.I
+a^#
UCJL
nnnl
r
to
'l
^r,^"t
agVgf
naznonL
vLrlu
Pvl
be
1000
^f
d.U
nan
ver
at
v
rha
UllE
I
L
kHz,.
r L ! ! u t
ihFttt
l]il./L{u!
rillirrnlI
so
e v
fu rhr aL
aS
with
aa rnr aa lLv JqLavr L
1read
ffOm
ac VTVM
The measurement floor of the generator/analyzer
combination is 0.0002 percent with a 1.414-V peak slgnal l-evel
into Lhe analyzer, and is due al-most entirely
to thermalnoise.
A redesign of the first
could push
analyzer filter
this figure befow 0.0001 percent, but in most practlcal
situatlons
the thermal noise of the UUT w111 set the
measurement floor.
For example, a I00-watt amplifier
with
(20 kqz flat)
a 100 dB S,/N ratio
will generate about 28 UV
of noise at its outpuf in the nolse bandwidth of the
At the maxlmum BO-V p-p test l-evel- this will
anal-yzer.
result ln an additlonal
nolse of 1.0 UV af the analyzer
lnput.
This is less than the anal-yzerrs 2 uv of referred
input noise and wil-l therefore only increase the mea.surement floor to about 0.00022 percent.
However, at a much
smaller 2.8-V p-p tesL leve1, the fu1l 28 pV of amplifier
noi se
wi I I
anncar
at
f.he
analvz.cr
s r l u 4 J
a v r
Jnnrrt
4 r r y 4
v
:nri
result
in
a
of about 0.0028 percent.
measurement floor
For comparJ-son, THD-20 aLso has a noise f1oor which may
be dominated by UUT noise.
In this case, however, the
.l
f
oor
i s
hi gher
hv
e
f ao f.nr, nf
20-?O
J v
ment bandwidth must be on the order
a spectrum analyzer 1s used.
heeause
the
measure-
of 100-200 kHz unless
EXPERIMENTALRESULTS
In order to see how the multltone IM test compares with
others 1n practice,
THD-20, DfM-30 and MIM were measured
The resul-ts, representing
and plotted for five circuits.
performance for four operational amplifier
configurations
and one power amplifier,
are shown 1n Flgures 9-13.
A 741 operational
amplifier
operating ln the inverting
measured, and the
unity gaj-n configuration
was first
configuration
results are shown in Fig. 9.
The inverting
-11
-
due to high-frequ9n9{
was chosen bo eliminate distortion
the amplifier
c
o
n
f
i
guratlon
t
h
i
s
In
c o m m o n - m o d ee f f e c t s . 2
slew rate'
t
0
'
6
5
V
/
v
s
a
a
n
d
b
a
n
d
w
i
d
t
h
exhibited a 500-kHz
indicate
c
u
r
v
e
s
d
l
s
t
o
r
t
l
o
n
t
h
e
t
h
r
o
u
g
h
l
i
n
e
s
d
a
s
h
e
d
The
p
e
a
k
r
a
t
e
o
f
change
the polnt at which the test slgnal-ts
onrreIs the device slew rate.
ev4s^v
A11 three tests indicate a simllar trend of distortl0n
asthetestlevellsincreased,wlthasomewhatfaster
Below
is approached'
raLe of rise as the slew rate limit
the
s
l
e
w
i
n
g
,
c
a
u
s
e
s
s
i
g
n
a
l
D
I
M
3
0
t
h
e
the level at whlch
tests track each other quite well wlth a more or less
THD-20 lies 4-7 dB befow
constanb declbel difference:
DIM-30,whileMIMlles30-35dBbelowDIM-30'Although
uru ries 30_35 dB below DIM-30, its measurement floor
for the test is-only 0.00022 percent (not shown), 38 ag
below that for DIM.
the results for the 'same 741 operaFigure 10 illustrates
galn of 10 '
operated at an invertlng
amplifier
tiSnal
Here the closed-lobp bandwldth 1s approximately 75 kHz'
about tO'65 V/us'
while the slew rate is stilt
fevels are somewhat
with the exception that the distortlon
higherduetothesma]Ierfeedbackfactor,therelative
behavlor of the three tests is about the same, wlth MIM
27 dB below DIM-30. In this case, the MIM
typically
mLisurembnt ft-oor 1s only 23 dB below the DIM-30 floor.
The increase in the MIM me€rsurementffoor here is due to
the lncreased u*pri-rr"r noise produced by the 74t in trre
X10 configuratlon.
Figure 11 shows the same data for..a /48 operational ampllcompensated with 4 pF and operated at a
fier
externally
to the 741 with the
firfs ampt-i-fier ls similar
giin of I0.
compensated. Because the
Exception that lt is-externally
compensati-on increases the slew rate to about
lighter
!2.g V/vs, greatly reduced TIM 1s expected, and this is
conflrmed bY Lhe curves.
further reduction in TIM when the
Flgure 12 illustrates
is operated *+tl ,t snaller 2 pF
implifier
?46-;p";ationai
w
h
i
c
h yields-a--t4'8-VUs slew rafe'
conp"ns4ting capa"itor
m
e
a
s
ures of TIM show reductions.
t
h
r
e
e
a
l
1
As expectedl
TestresultsforapowerampliflerareshownlnFig.13.
o f a -pcpbrnocxei m
r na t e l y . L 9 7 0 v l n t a g e ' w ? s
This 70-watt amplifier,
for issues of TIM and
designed without particular
slow quasl(4 V/us).
It also uses a falrly
t
h
e only
n
o
t
i
s
T
I
M
" "pt "l itm
" entary output stage, so
c" io" m
A
gain' MIM 1s
d
l
s
t
o
r
t
i
o
n
'
d
y
i
a
m
i
c
,otr""L of
significant
-12-
typicall_y 27 dB below DIM-30 for levels less than those
At higher
wfricn cause slewing on the DIM-30 slgnal.
The
e
v
ident'
1
s
greater separatlon
levefs, a slightly
s
h
a
l
l
o
w sIope,
r
a
t
h
e
r
1
t
s
w
i
t
h
c
u
r
v
e
,
f
f
t
O
l
Z
O
behavlor of the
a
n
o
m
a
f
o
u
s
.
r
a
b
h
e
r
seems
CONCLUSION
intermodulation
A new method for measuring transient
dlstortion
and other forms of dynamlc ampllfier
distortlon
test are
t
h
e
o
f
p
r
i
m
a
r
y
advantages
has been proposed.. The
p
r
o
cedure,
m
e
a
s
u
r
e
m
e
n
t
s
l
m
p
l
e
inexpensive tnstrumentatlon,
advanl
a
t
t
e
r
T
h
"
r
e
s
p
o
n
s
e
.
a
n
d
s
t
l
m
u
l
u
s
i
n
b
a
n
d
f
u
l
l
y
and
b
a
n
d
l
l
n
i
t
e
d
m
a
n
y
in
tage allows the test to be utlltzed
tests mlght not
syJtu*s where other dynamic distortLon
p
r
o
p
e
r
l
y
.
work
Although the test is not as stringent, as others in terms
and as a result ylelds smalfer
oi p"a[ rate-of-change,
with other dynanlc
correlatlon
p
e
r
c
e
n
t
a
g
e
s
,
dlstortion
tests 1s good and' dynamlc range ls slmilar '
distortlon
high
Because the best does not resort to unreallstically
good
t
e
s
t
,
u
n
d
e
r
a
m
p
l
i
f
i
e
r
rates-of-change to stress the
e
x
p
e
c
t
e
d
'
b
e
c
a
n
subJective correlatlon
-
1? -
REFERENCES
t.
2.
I'The Theory of Transient
M. Ota1a, E. Lelnonen,
IEEE Transactlons on
Intermodul_ation Dlstortlon,rr
Acoustlcs, Speech and Signil Processlng, Vol' ASSP-25'
No. 1, pp. 2-8, Feb. I97T '
f r A nO v e r v i e w
W. G. Jung, M. L. Stephens, C' C' Todd,
6
8
, June - Aug''
N
o
s
'
6
3
,
of SrD and'TiM," Audl;, voi.
10?o
L )
3.
lr.
|
t .
''
i n F e e d b a c k A m p . l i fi e r S ,
P. Garde, ,,TranSient Dlstortion
J. Audio Eng. Soc., vol. 26, No. 5, pp' 314-321, May,
1978.
t'A Method for Measuring
E. Lelnonen, M. Otala, J. Cur1,
(TIM)," J' -Audio
Transient Intermodul-atlon Distortlon
q
,
p
p
.
1
7
0
1
7
7, April 1977'
N
o
.
2
5
,
Eng. Soc., Vo].
5.S.Takahashl,S.Tanaka,'tAMethodofMeasuringTransient The
63rd Conventlon of
Intermodulation Dlstortionr"
#
L
\
7
8, May I979'
p
r
e
p
r
l
n
t
Audio Eng. Soc.,
6.
7.
8.
g.
T. HoIman, "Dynamic Range Requirements of Phonographic
Audio, pp' 72-79, July L977'
Preamplifi"rs,r'
I'A method f or Measuring
R . R . C o r d e l - 1 , t t C o m m e n t so n
Transient lntermodulatlon Distortionr , " J. Audio'Eng'
S o c . , V o I . 2 7 , N o . 4 , p p . 2 9 5 - 2 9 6, A p r i l 1 9 7 9 '
of The
J. Lammasriiemi,K. Karinieminen, "Dlstrlbution
C
o
n
v
e
n
t
ion of
6
2
n
d
phonograpir-Signal Rate of Changer"
L
9
7
9
'
The Audlb fng. Soc., March
t ' C o m m e n t so n t C o r r e f a t i o n A u d i o D i s R. R. .Corde11,
27'
Iti"ii"""tents'
tortion
,tr J ' Audio Eng' Soc' , Vol '
1
9
7
9
.
No. 4, pp. 297-298, APril
l0.E.Leinonen,M.otala,||CorrelationAudioDistortion
Measurem"r',ii,;'1. Audio Eng' soc', Vol' 26' No' L/2'
pp. ]r2-l-g, Jan ./Feb - 1978'
ll.J.S.Aagaardr"ADlmprovedMethodforThelvleasurement
IRE Trans, Audlo'
of Nonlinear Audio Diitortioflr"
N o v . , / D e c. L 9 5 9 .
12.
'rRelations Between
M. V. Callendar, S. Matthews,
AmplitudesofHarmonicsAndlntermodulationFrequenElectronlc Engineer, June 1951'
;r;4,"
13.P.R.Geffe,r'HowtoBulldHigh-QualityFiltersout
Nov' 11, I976'
of Low-Q""iitv--parts,tr Electr6nlcs,
-14-
Ql
Q2
0.33
FIGURE1:
A Highly
Sinplified
power Amplifier
F I G U R E2 : S^P( e
Z ck tHr uz m
/ d or f t h e M I I { , 3 O H zB W )
i"rt- iig"aI
MIM.spectrun-inclucling
FIGURE 3:
aiiiottion
Produtts Produced !{at- a galn
a 74L op-anp operating
4'2VP-P'
o
f
i
O
1
e
v
e
1
a
i
t
of
C l o s e u P of the MII'{
F I G U R E4 :
c
e
ntered near lkHz for
products
APPearance of the
FIGURE5:
s
i
g
nlL (toP trace)
MIl"{test
and the extracted l{IM distortion
Products (bottoin
SweeP synchron:.zeo
trace).
to bottom trace.
Extracted I'{IM
FIGURE 6:
distortion
Products as viewed
with a slower sweep than in
Fig. 5 and sYnchronized to
Note "anPlitude
enielope.
modulated" aPPearance due to
presence of both even- and
irdd-order distortion
Products
Fl osc.
9.00kHz
F I G U R E7 :
MIM Test Signal
Source
2-kHzLPF
3rd ORDER
BUTTERWORTH
l-kHz BPF
4th ORDER
150 Hz BW
F I G U R E8 :
MIM product Analyzer
100
Drll-30
10
THD-20
1.0
z
o
H
0.1
H
tt
o
E{
(a
H
a
It
0. 01
0.001
0.0001
0.5
10
I '
OUTPUT,V P-P
F I G U R E9 : D i s t o r t i o n c h a r a c t e r i s t i c s o f a 7 4I o p e r a t i o n a l
amplifier operating as a unity-gain inverter (vs = i12v) .
3o
1.00
10
TED-z0
1.0
a
o
H
€{
G
0.1
o
F
ul
H
a
tr.
0.01
0.001
0.0001
Q.5
10
1.0
3o
OUIPUT' V p-p
F I G U R E1 0 : D i s t o r t i o n c h a r a c t e r i s t i c s f o r a 7 4 1 o p e r a t i o n a L
arnplifier operating at an inverting gain of l-0'
100
10
1.0
DIM-30
v
o
H
H
cd
o
H
v)
0'.1
THD-20
H
a
\R
0.01
l{ru
0. 001
0. 0001
0.5
1.0
10
OUTPUT, V p-p
characteristics
Distortion
F I U G R E1 1 :
conpensated with
externally
amplifier
gain of 10.
an inverting
for a 748 operational
4 pF and operated at
30
100
10
1.0
zo
H
H
0.1
DII'I-30
o
H
ct,
H
THD.2O
a
\R
0.01
0. 001
0. 0001
0.5
.0
10
3o
OUTPUT, V p-p
FIGURE12:
Distortion
characteristics
amplifier
externally
compensated with
an inverting
gain of t0.-
for a 748 operational
2 pF and operated at
100
Dil'r-30
10
1.0
a
o
I'TIM
H
E{
0.1
o
H
6
H
o
\R
0.01
0. 001
0.0001
10
3o
100
OUTPUT,V p-p
F I G U R E1 3 : D i s t o r t i o n c h a r a c t e r i s t i c s
anplifier with substantial TIM.
for a 70-watt power