• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). There are daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Extensive study of IMD (and THD) distortions on 4 amplifiers with different topology

pma

Major Contributor
Joined
Feb 23, 2019
Messages
3,212
Likes
6,615
Location
Prague
Extensive study of IMD (and THD) distortions on 4 amplifiers with different topology

THD+N (SINAD) at 1kHz and 5W power has become a comparison measure of amplifier parameters here in ASR. I would like to demonstrate my deep persuasion that much more complex view of amplifiers distortion mechanism must be taken into account when trying to compare amplifiers based on their distortion performance. Distortion measurements methods have been developed and standardized for decades and their comparison survey was published for example by Cabot [1].

1. Explanation of intermodulation test methods used

1.1. SMPTE intermodulation


Intermodulation measurements using the SMPTE method have been around since the 1930s (Hilliard 1941, [2]). The test signal is a low frequency (usually 60Hz) and a high frequency (usually 7kHz) tone, summed together in a 4 to 1 amplitude ratio. Other amplitude ratios and frequencies are used occasionally, like DIN 250Hz+8kHz. This signal is applied to the device under test, and the output signal is examined for modulation of the upper frequency by the low frequency tone. The modulation components of the upper signal appear as sidebands spaced at multiples of the lower frequency tone. The amplitudes of the sidebands are root mean square summed and expressed as a percentage of the upper frequency level. Considering the SMPTE test in the time domain it becomes quite easy to understand its operation. The small amplitude high frequency component is moved through the input range of the device under test by the low frequency tone. The amplitude of the high frequency tone will be changed by the incremental gain of the device at each point, creating an amplitude modulation if the gain changes. This test is therefore particularly sensitive to such things as crossover distortion and clipping. High order nonlinearities create bumps in the transfer characteristic which produce large amounts of SMPTE IM. SMPTE testing is also good for exciting low frequency thermal distortion. The low frequency signal excursions excite thermal effects, changing the gain of the device and introducing modulation distortion. One advantage in sensitivity that the SMPTE test has in detecting low frequency distortion mechanisms is that the components occur at a high frequency. In most audio circuits there is less loop gain at high frequencies and so the distortion will not be reduced as effectively by feedback.

1.2. CCIF intermodulation

The CCIF IM distortion test differs from the SMPTE test in that a pair of signals closely spaced in frequency are applied to the device under test. The nonlinearity in the device under test causes intermodulation products between the two signals which are subsequently measured. For the typical case of input signals at 19kHz and 20kHz the intermodulation components will be at 1kHz, 2kHz, 3kHz, 4kHz, 5kHz, etc. and 18kHz, 21kHz, 17kHz, 22kHz, 16kHz, 23kHz, etc. Even order or asymmetrical distortions produce the low “difference frequency” components while the odd order or symmetrical nonlinearities produce the components (“skirts”) near the input signals. The distortion products generated in this test are usually very far removed from the input signal. This positions them outside the range of the auditory systems masking effects. If a test which measures what the ear might hear is desired, the CCIF test is a good candidate. The question of correlation with audibility for this test is covered very well by Stanley and McLaughlin 1977. However, whether one component of the test is more audible than another (second order more than third or vice versa) is not covered. The totally in band character of the test is one of its most attractive attributes. Unlike the sine-square test, the CCIF test may be used to severely band limited system. Distortion both before and after the band limiting point will be tested. Methods which lose their sensitivity with extreme band limiting can only test the circuitry before the band limiting. Some insight into the performance of the CCIF test as an indicator of SID can be obtained by examining the normalized derivative of the signal. The peak amplitude of the CCIF signal is twice that of a single sine wave. If ω1 and ω2 are the two input frequencies, normalizing the peak derivative by the peak amplitude we find that the normalized peak rate of change is (ω1 & ω2)/2. This is equal to that of a sine wave at the average input frequency. If the sine waves are very close in frequency compared to their mean frequency we find that the signal derivative will have the same general shape as the signal itself. Thus, the CCIF test should be similar to a THD test at the mean frequency in terms of its sensitivity to SID, but it will allow all of the resulting high order components to appear in band.

2. Amplifiers under test

3 different amplifiers were put in the test.
1st of them is class D AIYIMA A07 based on TI TPA3255 IC. My review with measurements is on
https://www.audiosciencereview.com/...s-and-review-lm4562-and-opa2134-option.19155/
It is a small cheap class D amplifier with respectable measurements with standard 1kHz THD+N test. However, more deep intermodulation measurements have not been done.

2nd amplifier is based on 52 years old Sinclair Z-30 project, with better output transistors and some other changes described in
https://www.audiosciencereview.com/...-amplifier-and-40-years-of-my-diy-mods.20198/
However, the basic circuit topology of this class AB amp has been preserved and it may be assumed that the well tuned original module would have similar distortion performance. THD+N of this amp at 1kHz is considerably worse that that of AIYIMA A07.

3rd amplifier is my 2x250W/4ohm DIY project described in
http://pmacura.cz/DIY_250W_4ohm_amplifier.html
It is a class AB amplifier with optimum bias of the output stage based on Douglas Self blameless topology. This amplifier has by far highest continuous output power from the amplifiers under test, 2x250W/4ohm and 2x150W/8ohm. THD at 5W is slightly higher at 5W than that of AIYIMA A07, but at higher power it dominates.

Regarding different output power of the DUT amplifiers, AIYIMA was tested at 20W for SMPTE and DIN IMD and 10W for CCIF 19+20kHz, Z-30 was tested at 15W for SMPTE and DIN IMD and 10W for CCIF 19+20kHz, A250W was tested at 50W for SMPTE and DIN IMD and 25W for CCIF 19+20kHz. Always with 8ohm resistor load. Let's see the results.

4th amplifier is Yamaha AX-396 (added on Sep 29, 2021)
class AB amplifier that I bought in the year 2002

3. SMPTE IMD 60Hz+7kHz

A07_SMPTE_IMD_20W_8R_xfi.png


Z30_SMPTE-IMD_15W_8R_xfi.png


A250W SMPTE IMD 60Hz+7kHz 50W 8ohm xfi.png


Yamaha AX-396 SMPTE IMD 60Hz+7kHz 8ohm.png



4. DIN IMD 250Hz+8kHz

A07_DIN_IMD_20W_8R_xfi.png


Z30_DIN-IMD_15W_8R_xfi.png


A250W DIN IMD 250Hz+8kHz 50W 8ohm xfi.png

Yamaha AX-396 DIN IMD 250Hz-8kHz 8ohm.png




5. CCIF IMD 19+20kHz

A07_CCIF_8R_10W_xfi.png


Z30_CCIF-IMD_10W_8R_xfi.png


A250W_CCIF_8R_25W_xfi.png

Yamaha AX-396 CCIF IMD 19+20kHz 8ohm.png



6. Summaries

Cumulative result was put in the chart and 1kHz THD and THD+N at 5W and at IMD test power were added as well.
IMD_distortion_study3.png


Numbers plotted in the graph are below.
IMD_distortion_study_plots3.png


We can see that AIYIMA scores best at 5W/8ohm/1kHz THD and THD+N, followed closely by Yamaha and A250W, with Z-30 worst by far. In SMPTE and DIN IMD, A250W is best. In CCIF IMD test, A250W is best followed closely by Yamaha, with AIYIMA worst by far. In all-round distortion competition, A250W is the best from the amps under test, followed by Yamaha. It is difficult to me to tell the verdict between AIYIMA and Z-30. However, in a controlled A/B and ABX test AIYIMA was distinguishable from other two amps by listening, most probably because of FR modulation by speaker impedance above 8kHz. But it is not excluded that the bad high frequency linearity of AIYIMA might have been a reason as well.

(Yamaha AX-396 was added on September 29, 2021)

Literature:

[1] Cabot, Richard C.: “A Comparison of Nonlinear Distortion Measurement Methods.”, AES Preprint 1638, May 1980.
[2] Hilliard, John K.: “Distortion Tests by the Intermodulation Method”, Proceedings of the IRE, December1941.
[3] Stanley, Gerald and Dave McLaughlin: “Transient Intermodulation Distortion and Measurement”, AES Preprint 1308, November 1977.
 
Last edited:

Matias

Major Contributor
Forum Donor
Joined
Jan 1, 2019
Messages
3,466
Likes
6,498
Location
São Paulo, Brazil
Very interesting, thanks.
Also for reference, below is the Hypex NCore NC252MP IMD CCIF, with skirts at -110 dB and below at 125W in 4 ohms. Very nice for 500 usd of a stereo amp.

nc252mp imd.jpg
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
3,212
Likes
6,615
Location
Prague
Thanks for sharing, the A07 seems to be the gift that keeps on giving.

Regarding intermodulation distortion, I really do not thing so. In CCIF IMD, it is considerably worse than the 52 years old Sinclair.
 

audio2design

Major Contributor
Joined
Nov 29, 2020
Messages
1,769
Likes
1,785
What this test most shows is just how useless these tests can be if interpreted as a single number.

Weighted for audibility, especially if weighted to typical listener age, the d2l spur at -76db for the Z30 CCIF would dominate distortion performance and render the Z30 worst measuring even though it looks great unweighted.
 

bigjacko

Addicted to Fun and Learning
Joined
Sep 18, 2019
Messages
673
Likes
328
For SMPTE and possibly CCIF the IMD are quite close to the tone, will those IMD easily masked by the tone? Also does the spikes around 16kHz count as IMD as well? They are 2*8kHz +- 250Hz.
 

Miska

Addicted to Fun and Learning
Manufacturer
Joined
Feb 20, 2019
Messages
615
Likes
426
I would also include TIM distortion measurement using the high performance amp test signal version.

This is 15 kHz sine and 3.18 kHz square wave using 4:1 amplitude ratio. Frequency response of the test signal is 1st order low-pass with fc=100kHz. (see AES paper of Leinonen/Otala/Curl from 1976)


(I've started using this test signal for ADC/DAC testing too)
 
Last edited:

zvukofor

Member
Joined
Oct 15, 2018
Messages
19
Likes
21
Location
Armenia
Yeap. It reminds me of a tests for THD, where higher levels of low order harmonics gives much less problems than lower level high order harmonics. We have hearing masking low orders, and we have different IMD patterns as a complications of different harmonics spectrum... so yes, we can’t use single number for THD in measurements.
 
  • Like
Reactions: pma
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
3,212
Likes
6,615
Location
Prague
I have just added Yamaha AX-396 results into the test and post #1 has been updated.
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
3,212
Likes
6,615
Location
Prague
below at 125W in 4 ohms.

I wonder how they calculate power for the CCIF 19+20kHz test (or similar DIF twin-tone). This test tone, if it has the same amplitude (peak value) as the single sine tone, has components F1 and F2 with half amplitude and rms voltage is -3dB compared to a single sine. Power is then half of the power of the single sine wave. For 125W/4ohm CCIF, the single sine power with same amplitude would have 250W/4ohm power. That's the reason why in my chart power in CCIF test is always lower than in the THD test. However, some producers and reviewers use "peak power" in the CCIF test. As power is power, any "peak power" is misleading. Do you have an idea how they define those 125W/4ohm??
 

Matias

Major Contributor
Forum Donor
Joined
Jan 1, 2019
Messages
3,466
Likes
6,498
Location
São Paulo, Brazil
I wonder how they calculate power for the CCIF 19+20kHz test (or similar DIF twin-tone). This test tone, if it has the same amplitude (peak value) as the single sine tone, has components F1 and F2 with half amplitude and rms voltage is -3dB compared to a single sine. Power is then half of the power of the single sine wave. For 125W/4ohm CCIF, the single sine power with same amplitude would have 250W/4ohm power. That's the reason why in my chart power in CCIF test is always lower than in the THD test. However, some producers and reviewers use "peak power" in the CCIF test. As power is power, any "peak power" is misleading. Do you have an idea how they define those 125W/4ohm??
Nope, maybe you can send an email to Hypex asking?
 
  • Like
Reactions: ENG
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
3,212
Likes
6,615
Location
Prague
I don't think that test has any real audible implications, the d5L is already at my threshold of audibility at full amplitude.

The test indicates to real circuit design issues. I do not know how old are you. There was a school in the mid-seventies that also supported importance of only 1kHz THD+N, similar as ASR does now, almost 50 years later. However, some of the amps that were perfect in the 1kHz sine test were not good in listening tests and sometimes had technical issues during operation. The problem was found and indicated as SID (slew induced distortion). This SID is also indicated by the CCIF 19+20kHz method or THD 20kHz, however the later has distortion components out of the audio band. AIYIMA is poor in the 19+20kHz and is the only one from the amps under test I can tell in the ABX test. It has "specific" sound and this is wrong. If you change the distance between the test tones or shift them lower in the audio band, the issue is still here. The inaudibility under specific test signal conditions does not mean the issue is always inaudible. The test just indicates there is an issue.
The 1kHz sine test would be sufficient only in case of frequency independent non-linearity with transfer function according to polynomial function. None of those is true in the real amplifier world. Non-linearities are frequency dependent and are not smooth, they may have dead zones and hysteresis in the transfer function.
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
3,212
Likes
6,615
Location
Prague
Some time ago I posted the listening ABX test with files recorded through AIYIMA A07 (which did not meet any attraction, though the amp is popular here), the link is here:

It was by now the only test with power amplifier (of those that I have performed and posted) in the loop when the amp was distinguishable in the ABX test. I have just tested the files with Paul's @pkane Deltawave and yes, the pk metrics result indicates that it is audible. Much "higher" score than even in case of the tube preamplifier.

A07_pkmetrics.png
 
Last edited:

AnalogSteph

Major Contributor
Joined
Nov 6, 2018
Messages
2,268
Likes
2,167
Location
.de
They're all basically fine, even if the A07 has the comparatively ugliest IMD spectra. I bet it's just the FR deviation that makes it distinguishable.

The comparatively high IMD2 (and H2) levels in the Z30 evolution most likely suggest an issue in the input stage symmetry or VAS department (I'm assuming nos. 5, 6 and 7 have already been taken care of adequately). The input stage is looking a bit too clever for its own good (what's the deal with the current reference?), I'd strip this back to pretty much a plain "blameless" and optimize from there. Nothing wrong with having a trimpot for fine-tuning currents, of course, just not sure about the rest. If Douglas Self can get something like 0.0006% out of a tweaked Blameless I don't see why you shouldn't be able to at least make it to substantially below 0.01%. Output stage bias looks good as-is.
 

Matias

Major Contributor
Forum Donor
Joined
Jan 1, 2019
Messages
3,466
Likes
6,498
Location
São Paulo, Brazil
They're all basically fine, even if the A07 has the comparatively ugliest IMD spectra. I bet it's just the FR deviation that makes it distinguishable.
Don't forget output impedance, it can also make FR changes, in some extreme cases like below quite significant changes.
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
3,212
Likes
6,615
Location
Prague
So, this is my judgement. I made a "single number" measure as a square root of sum of powers of distortion numbers of the amps in the individual tests. The formula is:

Score = SQRT(B3^2+B4^2+B5^2+B6^2+B7^2+B8^2+B9^2)

And this is the resulting chart
results.png


From best to worst:

1. A250W
2. Yamaha AX-396
3. AIYIMA A07
4. Sinclair Z-30 modified
 
Last edited:

audio2design

Major Contributor
Joined
Nov 29, 2020
Messages
1,769
Likes
1,785
Without psychoacoustic weighting I don't find IM measurements useful and the same could be said about THD but normally the product spurs are low enough not to be an issue.
 
Top Bottom