Transient Intermodulation Distortion (TIM)

Transient Intermodulation Distortion (TIM)
A Method
for Measuring
Transient
Distortion
(TIM)*
EERO
Technical Research
LEINONEN,
MATTI
Centre of Finland,
OTALA,
Electronics
Intermodulation
AND JOHN
Laboratory,
CURL
SF-90101
Oulu, Finland
A description of the square-sine intermodulation measuring method is given. The practical
measuring procedure is described, and measurement results of commercial audio power
amplifiers and integrated operational amplifiers are presented, in many cases showing
appreciable transient intermodulation in units which have excellent total harmonic distortion
and SMPTE intermodulation distortion specifications. The correlation of these measurements
with results obtained using other methods, such as the CCIF-IM and the noise-transfer
method, is treated, and the interdependence of these with the slew rate and the power
bandwidth is discussed.
INTRODUCTION:
The lack of correlation
ventional amplifier distortion measurements
tests has been noted by many designers of
ment. Modem amplifiers often measuring
total harmonic distortion at 1 kHz or below
modulation
distortion,
as measured
with
between
con-
which has been described
in detail
elsewhere
[1 ]-[3].
and listening
audio equipunder 0.01%
0.1% interthe SMPTE
So far TIM has been measured by determining the onset
of suppression of the internal drive signal in an amplifier
[4], but this has only yielded a qualitative limit beyond
which TIM can be expected. The aim of the present work
has been to quantitatively
measure this form of dynamic
method, may sound completely unacceptable.
It appears
therefore
that the audible difference between
various
intermodulation
distortion without the necessity of dealing
with the internal circuitry of an amplifier. The method
amplifiers is not due to these static distortion figures alone,
but some other reason must be involved.
One possible
mason is dynamic intermodulation
distortion,
which is
created by the frequency rather than the amplitude characteristics of the signal. One form of dynamic intermodulation distortion
is the transient
intermodulation
(TIM)
proposed here will measure both static and dynamic
intermodulation
distortions and can be used without any
knowledge of the out-of-band behavior of the unit to be
measured.
* Presented October 30, 1976, at the 55th Convention of the '
Audio Engineering Society, New York.
signal
consists . of a low-pass filtered i square
wave
and a
·
. '
.
.
smusold, having a peak-to-peak amPhtude raUo of 4:1.
i
JOURNALOF THE AUDIOENGINEERINGSOCIETY
MEASUREMENT
The measurement
170
METHOD
setup
is shown
in Fig.
1. The test
The sinusoid should have as high a frequency as possible
within the audio band, and 15.00 kHz is chosen as it is the
highest specified frequency for broadcast use. Furthermore, it allows measurements of tuners, radio links, etc.,
as well as lower class audio equipment. For AM broadcasting and certain tape recorders, a lower frequency, for
instance 6 kHz, can be chosen,
The selection of the square-wave frequency must be
done so that the harmonics of it do not overlap with the
sinusoid nor with the intermodulation products,
It has been shown elsewhere [5] that an optimum
separation is obtained when the frequencies of the sinusoid
and the square wave are related as
down relative to the fundamental of the 3.18-kHz square
wave, without any low-pass filter. Therefore, the signal
should not be too severe as far as amplifier stress is concemed.
To measure the intermodulation products, a selective
voltmeter or a spectrum analyzer can be used. The
distortion spectrum is outlined in Table I and shown
graphically in Figs. 2 and 3. Fig. 2 shows the input
frequency spectrum and Fig. 3 the measurement result of
the popular operational amplifier /.cA 741, which represents a case of very strong dynamic intermodulations,
where also second-order products such as 2f2 -+ nf_, 3f2 -+
nf_, etc. are generated. Traces of these can be found in
f2/f_ = {(Y + 1)[y(3' + 1)]_}_
(1)
Fig. 3.
The separation of the intermodulation products in frequencyis about 1 kHz.Psychoacoustic
investigations
[7]
f2/fx = {3'[3'(3' + 1)]_}_
(2)
seem to point out that less than 0.2% rms TIM is audible
under no-masking conditions, indicating that the spectrum
analyzer should have at least an 80-dB dynamic range and
500-Hz, -60-dB selectivity for reliable measurements.
This may be obtained with most selective voltmeters and
with some automatic graphic spectrum analyzers. Some
oscilloscope-type spectrum analyzers, however, have too
small a dynamic range and too poor selectivity for this
measurement.
or
where f_ is the frequency of the square wave, f2 is the
frequency of the sinusoid, and 3' is a positive integer.
Within the bounds given by Eqs. (1) and (2), the
square-wave frequency can be chosen freely. A high
square-wave frequency increases the sensitivity of the
method, but a low frequency is a more realistic drive
signal for the amplifier. A very low frequency could also
measure
possible
thermal
phenomena often encountered
adjustment
distortion
in power amplifiers and
Total intermodulation distortion is given by
operational amplifiers [6], but would at the same time
require very selectivemeasuringequipmentowing to the
multiplicity of closely spaced intermodulation products
falling to the audio band. As a compromise, a frequency
of 3.18 kHz has been chosen for the square wave,
corresponding to 3' = 4. If a 6-kHz sinusoid is used, the
preferred frequency of the square wave is 1.27 kHz.
The square wave is filtered with a single-pole low-pass
filter to limit its rise time. The preferred cutoff frequency
is 30 kHz (-3 dB) (1.6 dB down at 20 kHz), which is
roughly equivalent to bandwidths found in many audio
signal sources and lower class amplifiers. However, in
many cases the modem sources furnish ultrasonic signals
up to comparatively high frequencies. Quadraphonic
pickups, for instance, usually have extended frequency
ranges up to 50 kHz, and appreciable signal levels are
produced at higher frequencies owing to record distortion.
A 100-kHz low-pass filter is therefore recommended for
the square wave in the case of measurements of all high
quality equipment. It is to be noted that the fifth harmonic
of the measurement
signal is already more than -12
draM(%) = 100
Input
Signal
Distortion
Component
f2 - 5fx
f2 - 4f_
fl
3f2
4fl
f2
9elective
_'_
under te_t
voltmeter
or
f2 - 7fl
f2- 2f_
f2 - 8A
f2- f_
8.64
9.54
10.44
11.82
f2 - 9f_
13.62
12.72
15.00
f2 - 6fl
f2 -3f_
15.90
f2 - 10f_
f2 +
f_
16.80
18.18
f2 - 1lf_
19.08
19.98
6fl
Fig. I. Measurement setup.
APRIL 1977, VOLUME 25, NUMBER 4
Frequency
(kHz)
0.90
2.28
3.18
4.08
5.46
6.36
7.26
5f 1
analyzer
_._
(3)
Table 1. Signal and distortion components falling into the audio
band.
2fl
htm_lifier
·
where Vnt is the amplitude of each intermodulation componentf2 - nfx, n being a positive integer, and V2 is the
amplitude of the sinusoid.
Each component Vnt has two contributory pans,
1) the dynamic intermodulation component caused by
the rise-time portion of the square wave driving the
amplifier to frequency-dependent nonlinearity, for
instance TIM;
dB
t__°C_
/V2
171
EERO LEINONEN, MATTI OTALA, AND JOHN CURL
2) the static intermodulation component caused by the
amplitude-dependent nonlinearity of the amplifier,
The two components are in principle orthogonal and
_/dd vectorially, the results depending on the exact phase
difference of the components. In order to separate the
static intermodulation component, the square wave may
be changed to a triangular wave of equal peak-to-peak
amplitude. This decreases drastically the rise time, leaving
only the intermodulation components caused by static
nonlinearities. Fig. 4 shows the resulting static intermodulation spectrum for /zA 741 on otherwise the same
conditions as in Fig. 3. As such, this part of the test is
roughly analogous to the SMPTE intermodulation test
method,
MEASUREMENT
PROCEDURE
1) A source of sinusoidal voltage and a source of
square-wave voltage are connected to the input of the
amplifier under test by means of series resistors so that the
sources do not load each other. The output of the square-
A<_)
- l0
- 20
t
-80
1
,[
_
_
_
The resulting peak-to-peak voltage (that is, 1.25 times
the square-wave peak,to-peak voltage), measured at the
.output, is taken as reference output voltage. When documenting the test conditions, the corresponding output
power is measured by replacing the test signal with a
3.18-kHz sinusoidal voltage of equal peak-to-peak
amplitude.
3) Theamplitudeof theintermodulation
productsat the
outputof theequipmentundertest is measured,and the
total distortion is calculated by root-mean-square
(rms)
summingall the productsfallinginto the specifiedaudio
band. The distortion percentage is calculated by taking the
rms sumof the intermodulation
products,dividingit by
the amplitude of the 15.00-kHz test signal component at
the output,andmultiplyingthe quotientby 100. This is
the totalintermodulation
distortion,composedof thestatic
- 30
- ,0
-so
- 60
- 70
wave source must be filtered with a single-pole low-pass
filter having a cut-off frequency of 30 kHz ( -3 dB) or 100
kHz (-3 dB), depending on quality requirements of the
equipment being measured. When measuring frequencydependent circuits, .for instance, phono preamplifiers
employing
RIAA equalization,
appropriate
counterequalization must be used to create a nominal-level test
signal in the output.
2) The frequency of the sinusoidal source is adjusted to
15.00 kHz and the frequency of the square-wave source to
3.18-kHz7 The amplitudes are adjusted so that the amplifier is operating under desired working conditions and
the ratio between the peak-to-peak voltages of the sources
is4:1.Thisisequivalent
to
a) the ratio of 5.66:1 between the rms amplitudes
b) the ratio of 11.3 : 1 between the peak-to-peak
value of the squarewave and the rms value of the
sinusoid.
f_k.z>
_.
5
oo.o
_ _
Fig. 2. Frequency spectrum of input signal,
andthedynamic
components.
4) To measure
the static intermodulationdistortion
only, the square-wavecomponentisreplacedby a triangular wave of the same frequencyand equal peak-to-peak
amplitude. The distortion is calculated as above.
A(dB)
A(dB)
- l0
- l0
- 20
,,
- 20
- 30
-4o
S
Z_
,
T
_
,
,
_ C
'
_
_
_
'
_
.
..2
- 50
-
70
llj [A,i
I
IJ
- 70
-
Il
80
f(kaz)
Fig. 3. Frequency spectrum of output signal of operational
amplifier /aA 741. Conditions--noninverting circuit; 20-dB
gain; 5-kll load resistance; 5-V output voltage peak to peak; _
5-Vsupplyvoltage,
172
- 60
Fig. 4. Static intermodulation of operational amplifier /.LA
741obtained by replacing the square wave with a triangle wave of
equal frequency and peak-to-peak amplitude. Otherwise same
conditions
asinFig.3.
JOURNAL
OF THE AUDIO ENGINEERING
SOCIETY
A METHOD FOR MEASURING TRANSIENT INTERMODULATION
INTERPRETATION
OF THE RESULTS
Power amplifiers
The dynamic intermodulation distortion products are
often generated by TIM, that is, the rapid rise of the
square wave driving the amplifier near the onset of the
slew
rate limit
[5]. The
the limiting
mechanism
can be[3],deduced
fromtype
the ofintermodulation
spectrum
by comparing
the relative
amplitude
of the
different products as follows,
1) If the amplitudes of the even products (f2 - 2f, f2
- 4f_ ....
) are dominant as, for instance, in Fig. 3, the
limiting mechanism is symmetrical with respect to positire and negative slewing. In the case of perfect symmetry, the odd-order products (f2 -f_,f2 - 3f_, etc.) vanish.
The higher the amplitudes of the odd products are, the
more unsymmetrical the limiting mechanism is. If the
even-order and odd-order products have equal amplitudes,
the limiting is completely one sided.
2) If the amplitudes of the f2 - 2f_ product and the
f2 - 8fx product are about the same, the
mechanism is abrupt.
3) If the f2 - 8f_ product is small compared
f2 - 2ft product, the onset of the slew rate
gradual,
It is commonplace to find abrupt symmetrical
limiting
with the
limit is
limiting
in well-designed operational amplifiers and gradual unsymmetrical limiting in power amplifiers and old operational amplifiers.
Eight popular power amplifiers/receivers were tested.
The amplifiers were Sansui 771, Marantz 2270, Harman
Kardon 230 A, Sony TA 5650, Salora 2000, Dux TA
4000 (= Philips), Pioneer SX 535, and Tandberg TR
2075. The amplifiers were chosen to represent the newest
medium-price generation in the European market. Each
unit was tested using specified output load resistances and
with tone controls carefully adjusted to flat frequency
response. The volume control was set to maximum and the
test signal was fed to the AUX-input of each amplifier.
30-kHz filtering of the square wave was used. For reference, the dynamic intermodulation
was also measured
without this filter, because in a good design, the preamplifier should limit the frequency response to that allowable for the power amplifier. The static intermodulation
was also measured and all the results are shown in Figs. 6
to 9. The amplifiers appear in random order, with the
numbering bearing no correlation to the order of the
amplifiers in the list above.
The results show that some dynamicintermodulationis
present in all the amplifiers tested. The contribution of
dynamic distortion was small or medium in amplifiers 3
and 5 to 8, whereas amplifier 4 is an example of
reasonable static intermodulation results combined with
dramatic dynamic
power levels.
DI'_ (%)
TEST RESULTS
Operational
/
such as LM 318, LF 356, LF 357, MC 1456, HA 2505,
and CA 3140. Measurements were made using recommended operating voltages, a noninverting circuit, 20-dB
gain setting, and 5-kf_ load resistance, unless otherwise
The recommended
compensations
for 20-dB
Theandresults
are gain
shownwere
in Fig.
gain
for unity
used. 5. It shows that operational amplifiers/aA 709,/aA 739,/aA 741, LM 301, and
MC 1456 exhibit strong dynamic intermodulation even at
iow output voltages. This high distortion completely
excludes the use of/aA 741 in audio circuits, and restricts
the use of/aA 709,/aA 739, and LM 301 to gains greater
than 20 dB, and to output voltages less than a few volts
peak to peak. Even then, extreme precaution must be
taken to check possible distortion caused by improper
compensation. Irrespective of the strong dynamic intermodulation, the static intermodulation values and the total
harmonic distortion are extremely good for these units as
will be shown later.
No dynamic intermodulation was found in LM 318, LF
356, LF 357, and HA 2505, although some very weak
static intermodulation was found. Characteristic of these
amplifiers was their high slew rate.
APRIL
1977, VOLUME
25, NUMBER
4
/
1 0
/
/
/
/
/
increase
,20 k}tz.... )
739
,_itv_%/ __01
,z/'
'_'//'
/'//_'
/
The dynamic intermodulation was measured from nine
operational amplifiers. These included the popular types
gA 709, gA 739, gA 741, and LM 301, frequently used
in audio circuitry, and some newer fast and ultrafast types
intermodulation
0 kJlz
/7.........
Amplifiers
.3.0
specified.
DISTORTION (TIM)
for higher
0 Fdl
z
=i....
/ /
I
10kHz
///
//.
7092
//,(//'
//"/
/
//!/
.0 3/
//_.f.
//_//
o /°'/
I/'
/!/
o/
i// o/
/.//d5'
//
0.1 / ../,, ,
/
/
/
/
1_2Z%....-_14%
....__00
_.z
o_y_.O-o.<
of*/*'----*
..3o m=
/
/
/
y
/
,,-_00_
/_
_'?//_,>'_/'G'_
_ _._"_'/_f/5
/'
. 301 20cI_
31:0._"/,_/30,0_,_.
3140
/
×,,-
/
\\ /739
/,,/"_9
//J 5
OCltm
/./
///
,,,,,_./
///
_, Y
/ _10
20dB com_.
Uout
,
15
t_
Fig. 5. Dynamicintermodulation
distortionofdifferentoperational amplifiers as a function of output voltage, with squarewave
filtering
frequency
as a5-.kf_
parameter.
Conditions-noninverting
circuit;
20-dB gain;
load resistance;
_+15-V
supply voltage. Types of amplifiers--gA 709, /aA 741, MC
1456, LM 301,/aA 739, LM 301, and CA 3140.
173
EERO LEINONEN, MATTI OTALA, AND JOHN CURL
CORRELATION
WITH OTHER METHODS
harmonic frequencies to be measured may lie outside the
Two standardized measurement methods are frequently
used to specify audio amplifier distortion. These are the
total harmonic distortion measurement method and the
SMPTE intermodulation distortion measuring method,
described in detail elsewhere [8]. A third method, the
CCIF
intermodulation
recently
been used
on certain
high-quality method
products,[8],andhassome
reviewers
have
recently used a fourth method, the so-called noise-transfer
method [9]. It is interesting to try to establish the correlation between the results obtained with all these methods.
The totalharmonic distortion method uses a pure 1-kHz
sinusoid as.the input signal. The measured amplitudes of
the harmonic frequencies at'the output of the equipment
under test are rms added and divided by the fundamental
frequency to yield the distortion percentage. This test
measures purely the static harmonic distortion. If higher
test frequencies are used, several uncertainties of correctness are encountered, mostly caused by the fact that the
D_,_,_>
/ .... 2
/
Rated output power
_._-'}30°°
'_
w
/
//
/'
frequency signal, usually in the range of 81-400 Hz, and
a high-frequency signal, usually 7 kHz. The amplitude
modulation of the high-frequency signal is measured and
given as a percentage. This method measures purely static
intermodulation distortion at the frequency of the lowfrequency signal.
The CCIF-IM method uses two closely spaced highfrequency sinusoidal test signals and measures their interference products. It measures both the static and the
dynamic intermodulation,
depending on the test frequencies.
The noise-transfer
method uses as an input signal
band-limited pink noise having a frequency range of
11-20 kHz. The intermodulation products of this noise at
a frequency range of 0-10 kHz are measured. This test
measures mostly the dynamic intermodulation.
Table II shows the measurement results for a number of
popular operational 'amplifiers using the proposed method,
as well as the four methods described above. In addition,
the measured slew rate of the amplifier is shown.
The following conclusions may be deduced from the
results.
1) The total harmonic distortion and the SMPTE
/.
/'
/
....
passband of the equipment under test. Some of these
uncertainties are outlined in the following sections.
The SMPTE-IM method uses two signals, a low-
.........
"/
./'
_DI'I
,/_IH
- 0.03
1
DI/_/I_I(%)
1
_Rated5output
omp.
55W _c_4er
/'/'
1.0
/11'4
6
/
/
/'
tion
for6.commercial
power amplifiers
and 2. Solid lines-Fig.
Measured dynamic
and staticIintermodulation
distorstatic intermodulation; dashed lines--dynamic intermodulation
with30-kHzfilteringof squarewave;dot and dashlines--no
filtering
ofsquare
wave.
o_'w:_
a>
////
--'J';"
./,/'
,
-_....-'_'_
_--'
s
l0
_
TM
Dllq 6
_x
ms
_/._,
_,_'"_
.20
4b
_0_,_t.._
Fig. 8. Distortion characteristics of amplifiers 5 and 6. Conditions as in Fig. 6.
,m,_4
.DI'!
_/0'3
DIM 5
4
1.0
am_.3
30
_ated
,:]rl,_.
4 Out_ut
70 rx_er
1.0
//
O. 3
/ ·/
./?,/,
DI'I 3
//
amp. 8
15 W
/DIM
[
,
.,,' .
8
'
,. DI)_ 8
.×...'
.
114 8 .>_
.,
DIM 7
/
.0.1
2.5
Fig. 7. Distortion characteristics of amplifiers 3 and 4. Conditions
asinFig.6.
174
5
_-0
20
40
Fig. 9. Distortion characteristics of amplifiers 7 and 8. Conditions
asinFig.6.
JOURNAL
OFTHE
AUDIO
ENGINEERING
SOCIETY
A METHOD FOR MEASURING
methods
give very low distortion
quality of the amplifier
completely unacceptable.
distortion
and the
figures, even when the
as judged with other tests is
The use of the total harmonic
SMPTE-IM
tests
is questionable
for
TRANSIENT
correlation
INTERMODU[.ATION
exists between
DISTORTION
them, depending
(TIM)
on their dras-
tically different definitions.
Slew
Rate
audio equipment if dynamic distortion is suspected.
2) The noise-transfermethodgives
a clearindicationof
dynamic intermodulation
distortion when it is relatively
high. For the low distortion values its sensitivity seems to
be rather poor, probably caused by two factors,
a) the input power density is rather small, resulting
in low amplitude of the intermodulation
products,
b) these intermodulation
products are masked by the
thermal noise,
The interdependence
of the slew rate and the squaresine test results
is shown in Table III. For each
3) The CCIF-IM method gives a reliable indication of
dynamic intermodulation
distortion. However, it seems to
be less sensitive than the square-sine method,
4) The results of the noise-transfer method, the squaresine method, and the CCIF-IM method correlate roughly,
Depending on the
operational amplifier,
of change, measured
component, up to 8.7
individual departures probably depending on the slightly
different sides of the more or less same phenomenon these
finding as a safety factor of 10 for usual signal levels of
operational
amplifiers
and 30-kHz bandwidth,
a safe
minimum
slew rate would be of the order of 10 V//xs,
tests are measuring,
The exact correlation between the different measuring
methods is subject to a more detailed investigation
[11 ].
which only few operational amplifiers are able to attain.
Applied to power amplifiers, the equivalent safe minimum
slew rate should be around 100 V//xs, which only a few of
the amplifiers on the market are able to handle. These
results show that even the fastest present amplifiers must
remain
suspect as far as dynamic
intermodulation
is
OTHER
SPECIFICATIONS
operational
amplifier,
signal levels required to produce
dynamic intermodulation distortion values of 0.3%, 0.1%,
and 0.05% were determined.
The 0.1% value corresponds
to the present knowledge of the threshold of audibility of
TIM. From the signal levels obtained, the maximum rate
of change of output voltage was calculated.
thus confirming
compensation
and the type of the
the distortion begins at output rates
as the rise time of the square-wave
times below the measured slew rate,
earlier
predictions
[2],
[3]. Using this
Amplifiers are often characterized
with the slew rate
and the power bandwidth.
These are basically
related
concerned.
Correspondingly,
slew rate specifications
as
low as 1 V//xs and 20 V//xs will probably lead into troubles
specifications,
in dynamic
although
usually
only a first-order
Table II. Some operational
rough
amplifiers measured
Square-Sine
Distortion (%)
Type
Remarks
/aA 709
/aA739
MC1456
LM301
/aA739
/aA741
LM301
/aA709
20-dB compensation
20-dBcompensation
20-dBcompensation
O-dBcompensation
O-dBcompensation
O-dB
compensation
General conditions: noninverting
voltage. Dash signifies unmeasurable
Table III. Maximum
30-kHz
Filter
100-kHz
Filter
0.039
-0.043
-5.6
3.8
3.5
62
0.28
0.04
0.32
0.03
8.9
6.3
5.8
63
circuit; 20-dB gain; peak-to-peak
distortion.
intermodulation
with different measuring
CCIF-IM,
14.00 kHz+
15.00 kHz
(%)
performance.
methods.
Noise
Test,
SMPTE-IM
Noise
200 Hz +
Level
8.0 kHz
(dB)
(%)
-_
-0.02
0.32
0.44
0.66
26
-----20
--'26
-20
-6
----0.31
0.10
0.10
0.11
signal output rate of change (square-wave component)
distortion (DIM).
/aA741
LM301
MC1456
CA3140
/aA709
/aA739
Remarks
O-dB
compensation
20-dB
compensation
20-dBcompensation
0.3 % DIM
0.29
0.25
1.8
13.5
-2.1
< 0.02
< 0.02
< 0.02
< 0.02
< 0.02
< 0.02
< 0.02
< 0.02
output voltage 10 V; 5-kll load resistance;
for different
0.1% DIM
0.11
0.11
1.6
13.5
3.0
2.1
3.0
2.1
1.8
1.3
0.64
0.61
0.58
0.20
_+15-V supply
levels of total dynamic intermodulation
Maximum Rate of Change in Output (V//as) for
Type
Total Harmonic
Distortion,
Slew
1 kHz
Rate
(%)
(V//as)
0.05% DIM
0.07
0.07
1.2
10.4
2.8
1.7
Measured
Slew Rate
(V//as)
0.61
0.58
1.8
13.5
3.0
2.1
General conditions: Noninverting circuit; 20-dB gain; + 15-V supply voltage. 2-kfl load resistance is used in conformity with slew
rate measurement practice of the amplifiers in question.
APRIL 1977,VOLUME25, NUMBER4
175
EERO LEINONEN, MATTI OTALA, AND JOHN CURL
Power Bandwidth
Power bandwidth is specified to be that frequency at
which the amplifier is capable of delivering half of its
rated output power with a specified maximum total harmonic distortion, usually 1% [10]. It is widely used as a
criterion for the high-frequency capability of an amplifier,
Although this is true to a certain extent, there are a
number of effects which make this criterion less consistent
in the case of audio amplifiers. All these effects stem from
the fact that in order to specify the power bandwith one
must measure distortion components which are outside the
passbandof the amplifier.
As an illustration, consider an amplifier which has a
measured power bandwidth of 20 kHz. If now an ideal
35-kHz low-pass filter is added to the output, the total
harmonic distortion falls to zero per definition. The
"resulting" power bandwidth in the range of 20-35 kHz
then only depends on the amplifier topology and dimensioning,
The problem lies in the definition, because for real-life
signals, such as music,the audible frequency band would
be crowded with intermodulation products, and the sound
would most probably be very bad, in spite of the fact that
the total harmonic distortion would be very low and the
power bandwidth would be high.
It is important to notice that it would also be impossible
to use the SMPTE intermodulation measuring method to
detect this intermodulation, because the SMPTE method
only measures static intermodulation.
This illustration shows how a little trick could be used
to fool the usual measuring methods. It is not implied that
designers would use such tricks deliberately, but there
exist a number of "built-in" mechanisms which perform
the sameoperation.
1) It is usual to incorporate an RLC network in the
output of power amplifier to ensure stability during capacirive output loading. This acts as a filter network above 50
kHz, and it usually decreases the total harmonic distortion
some -2 to -6 dB at 20-30 kHz, depending on the
harmonic spectra,
2) The amplifier feedback phase margin is usually not
exactly -90 degrees. This departure may affect the
amplitude of a given harmonic up to -10 to +20 dB.
3) Some of the harmonics may be outside the closedloop bandwidth of the amplifier, above which an. added
attenuation of -6 to -18 dB per octave is generated. This
may be rephrased by stating that most high-frequency
distortion is generated in the driver circuits. The ultrasonic
harmonic components may then be attenuated'in
the
intrinsically slow power output stages.
This effect may result in some -2 to -20-dB reduction
of the 20-30-kHz total harmonic distortion value, depending on the amplifier closed-loop bandwidth.
As can be seen, a number of effects decrease the
reliability of the power bandwidth specification, and
therefore no attempts to correlate power bandwidth with
the CCIF-IM or square-sine method have been made in
thisproject.
The definition of the power bandwidth would be much
more reliable if it were not specified with 1% total
176
harmonic distortion but, say, 0.1% CCIF-IM for any
combination.of amplifier passband signals. As is shown in
Table II and by experience, a low value of total harmonic
distortion combined with reasonable power bandwidth
may in some cases indicate the presence of strong dynamic
intermodulation.
CONCLUSION
A new audio distortion test method has been proposed
and applied. Its measurement results and correlations with
known methods have been discussed showing that
1) many amplifiers having excellent total harmonic
distortion and SMPTE-IM data show high values of
distortion as measured with the proposed method,
2) the proposed method seems to yield qualitative
correlation with other methods measuring the dynamic
intermodulation distortion,
3) the proposed method seems to be more sensitive
thantheseexisting
methods.
It is believed that the proposed test gives a stringent, but
realistic, test signal to an amplifier with the option of
convenient passband adjustment to simulate different sighal sources and their capability to produce various levels
of responses.
REFERENCES
[1] M. Otala, "Transient Distortion in Transistorized
Audio Power Amplifiers," IEEE Trans. Audio and Electroacoust., vol. AU-18 pp. 234-239 (1970).
[2] M. Otala, "Circuit Design Modifications for
Minimizing Transient Intermodulation Distortion in Audio
Amplifiers," J. Audio Eng. Soc., vol. 20 pp. 396-399
(June 1972).
[3] M. Otalaand E. Leinonen, "The Theoryof Transient Intermodulation Distortion," Monitor-Proc. IREE,
vol. 37, pp. 53-59 (Mar. 1976) and IEEE Trans. Acoust.,
Speech, Signal Processing,'vol. ASSP-25, pp. 2-8 (Feb.
1977).
[4] M. Otala and R. Ensomaa, "Transient Intermodulation Distortion in Commercial Audio Amplifiers," J.
Audio Eng. Soc., (Project Notes/Engineering Briefs), vol.
22, pp. 244-246 (May 1974).
[5] M. Otala and E. Leinonen, "Possible Methods for
the Measurement of Transient Intermodulation Distortion," presented at the 53rd Convention of the Audio
Engineering Society, Zilrich, 1976. Available as Publication 16/76, Electrical and Nuclear Technology Series,
Technical Research Centre of Finland.
[6] J. E. Solomon, "The Monolithic Op Amp: A
Tutorial Study," IEEE J. Solid State Circ., vol. SC-9,
314-332 (1974).
[7] M. Petri-Larmi,
"Investigations
on the
Psychoacoustics of Transient Intermodulation Distortion"
(to be published).
[8] IEC Publ. 268-3, Clause 20.
[9] F. M. Huges, "Seventeen Amplifiers," Hi-Fi for
Pleasure, pp. 56-63 (Mar. 1976).
[10] Deutsche Industrielle Norm DIN 45500 B1.6
1.73.
[11] E. LeinonenandM. Otala, "Correlationof Audio
Distortion Specifications,"
presented at the 56th Convention of the Audio Engineering Society, Paris, 1977.
JOURNAL OF THE AUDIO ENGINEERING SOCIETY
A METHOD FOR MEASURING
TRANSIENT
INTERMODULATION
DISTORTION
(TIM)
THE AUTHORS
E. Leinonen
M.Otala
Eero Leinonen was born in Paltamo, Finland, on
January 24, 1951. He received the M. Sc. degree from the
University of Oulu, Oulu, Finland, in 1974.
He is presently employed by the Technical Research
Centre of Finland, Oulu, where his work is concerned
with audio amplifiers and microprocessors,
Mr. Leinonen is a member of the Audio Engineering
Society and the Finnish Society of Electronics Engineering.
·
Matti Otala was born in Oulu, Finland, in 1939. He
receivedthe M.Sc. and Lic. Techn. degrees from the
Technical University, Helsinki, Finland, in 1963 and
1967, respectively, and the degree of Dr. Techn. from the
University of Oulu, Oulu, Finland, in 1969.
From 1962 to 1966 he worked for Oy Helvar, Helsinki,
Finland designing stereophonic equipment and later as
technical manager. From 1966 to 1968 he worked for Oy
Nokia Ab Electronics, microwave link division, as a
project leader. In 1967 he joined the University of Oulu as
Professor of Electronics, specializing in process instrumentation. From 1972 to 1973 he worked at the
Philips Research Laboratories, Eindhoven, The Netherlands, on magnetic bubble technology, and in 1974 in the
APRIL 1977,VOLUME25 NUMBER4
J. Curl
Centre National d'Etudes des Telecommunications, Paris,
France, on time sharing telephone exchanges. In 1975 he
was appointed Director of the Electronics Laboratory of
the Technical Research Centre of Finland, Oulu. He has
published some 60 scientific papers on audio, superconductivity, metal physics, instrumentation, magnetic bubbles, and computer memory organization. He holds a
number of international patents.
Dr. Otalais a memberof manyFinnishandinternational societies.
·
John Curl was bom in San Francisco, California, in
1942. He graduated from San Francisco State University
in 1966 with a B.A. in physics and was employed by
Ampex Corporation from 1967 to 1969 in the professional
audio and research departments. There, he worked primarily in low noise research and servo design.
Since 1971 he has been an independent cgnsultant with a
number of firms, including: Mark Levinson Audio Sysrems, Gale Electronics and Design (London), I.H.E.M.
(Switzerland), and Alembic lnc. (San Francisco).
He is a member of the Audio Engineering Society and
the Institute of Electrical and Electronics Engineers.
177
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement