• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

45kHz BW distortion study of 3 amplifier topologies (2xAB, 1xD) + A/B test

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
Following the discussion in the Audiophonics HPA-S400ET thread on Purifi HF non-linearity, I have decided to make a comparative measurement of 3 of my amplifiers. THD vs. power into 4ohm with @45kHz bandwidth. The amplifiers under test are:

1) A250W4R, 250W/4ohm class AB amplifier based on Douglas Self "blameless" topology, with simple 1-pole dominant pole frequency compensation

2) PMA-NC252MP, 250W/4ohm class D amplifier with Hypex NC252MP module

3) SAB small class AB amplifier, 40W/4ohm, based on vintage Kenwood KA-5010 amplifier with 2-pole frequency compensation (this allows for higher feedback at high frequencies). The amplifier has recently been provided with SMPS power supplies.

The results:

1) A250W4R
A250W_20-15k_thdpower.png

Distortion is frequency dependent, best result is for 1kHz, from 1W to 20W we can see contribution of class AB crossover distortion which worsens with frequency.


2) NC252MP
NC_20-15k_thdpower.png

Very low distortion at 20Hz and 1kHz, however gets worse at 5kHz and higher. In spectrum analysis, I can see that the distortion is composed mostly of high order harmonics above 20kHz.


3) SAB small class AB with 2-pole compensation
SAB_20-15k_thdpower.png

The distortion is almost independent on power and frequency. The best result by far, from my point of engineering view.

We can also see why reviewers mostly use 1kHz plots. 1kHz gives the "nicest" results.

4) UCD180HG result added. Remarkably independent of frequency.
UcD_20-15k.png
 
Last edited:

LTig

Master Contributor
Forum Donor
Joined
Feb 27, 2019
Messages
5,760
Likes
9,442
Location
Europe
Following the discussion in the Audiophonics HPA-S400ET thread on Purifi HF non-linearity, I have decided to make a comparative measurement of 3 of my amplifiers. THD vs. power into 4ohm with @45kHz bandwidth. The amplifiers under test are:

1) A250W4R, 250W/4ohm class AB amplifier based on Douglas Self "blameless" topology, with simple 1-pole dominant pole frequency compensation

2) PMA-NC252MP, 250W/4ohm class D amplifier with Hypex NC252MP module

3) SAB small class AB amplifier, 40W/4ohm, based on vintage Kenwood KA-5010 amplifier with 2-pole frequency compensation (this allows for higher feedback at high frequencies). The amplifier has recently been provided with SMPS power supplies.

The results:

1) A250W4R
View attachment 195911
Distortion is frequency dependent, best result is for 1kHz, from 1W to 20W we can see contribution of class AB crossover distortion which worsens with frequency.


2) NC252MP
View attachment 195914
Very low distortion at 20Hz and 1kHz, however gets worse at 5kHz and higher.
The curves for both 5 and 10 kHz up to 100W are below those for the AB amp. Only the 15 kHz curve is higher which wouldn't bother me much as the harmonics are inaudible. Might be interesting to measure IMD using 19+20 kHz signals.
 

dc655321

Major Contributor
Joined
Mar 4, 2018
Messages
1,597
Likes
2,235
Might be interesting to measure IMD using 19+20 kHz signals.

And keep the power axes all the same length, say 100W.
The SAB is surely well-behaved though...
 

Matias

Master Contributor
Forum Donor
Joined
Jan 1, 2019
Messages
5,029
Likes
10,796
Location
São Paulo, Brazil
We can also see why reviewers mostly use 1kHz plots. 1kHz gives the "nicest" results.
You got the order wrong. 1kHz and its harmonics is where the ear is most sensitive, therefore designers prioritize this region for lowest distortion, and reviews show it as well.

Distortion above 10 kHz is irrelevant because:
1. Musical signal is way below full scale, so its harmonics even lower than on mids and bass; and
2. Harmonics are above absolute hearing threshold (20+ kHz) and below low level audibility.

Good old Fletcher-Munson curves say it all.

fm.jpg
 
Last edited:
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
I would like to see a valid evidence (a proper set of tests) that the class D distortion and mess above 20kHz is unimportant. Not because of thresholds of hearing (everyone knows Fletcher-Munson curves for some 70 years at least) but to make sure that the class D ultrasound distortion cannot create audible intermodulations in tweeters. Is this possibility excluded? Is there any proof?
 

julbo

Member
Joined
Jan 10, 2022
Messages
33
Likes
33
Location
Mediterranean
If this creates an excessive excitation of a tweeter in the breakup band (say 25K for a Seas DXT 27TBCD) will it impact the audible band simply because the shape of the tweeter changes ?
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
One more measurement - Hypex UcD180HG. Remarkably independent of frequency, though higher distortion. Very predictable, no tricks.

UcD_20-15k.png



THD and Noise plotted separately at 5kHz test frequency, UcD180HG. We can see what is noise and what is distortion (normally messed up in THD+N/SINAD plots). We can see that there is no dramatic rise of ultrasound noise (up to 45kHz) and all is harmonic distortion.
UcD_5k_THD_N_seprtd.png
 
Last edited:

Matias

Master Contributor
Forum Donor
Joined
Jan 1, 2019
Messages
5,029
Likes
10,796
Location
São Paulo, Brazil
I would like to see a valid evidence (a proper set of tests) that the class D distortion and mess above 20kHz is unimportant. Not because of thresholds of hearing (everyone knows Fletcher-Munson curves for some 70 years at least) but to make sure that the class D ultrasound distortion cannot create audible intermodulations in tweeters. Is this possibility excluded? Is there any proof?
It is listed on the FAQ, including a link to calculations of the carrier power on the tweeter, it is super low.

Also if there are suspicions of tweeters intermodulating high frequency distortion, let us see the measurements, mic on tweeter. Until then it is all supposition, nothing more.
 

Soniclife

Major Contributor
Forum Donor
Joined
Apr 13, 2017
Messages
4,500
Likes
5,417
Location
UK
I would like to see a valid evidence (a proper set of tests) that the class D distortion and mess above 20kHz is unimportant. Not because of thresholds of hearing (everyone knows Fletcher-Munson curves for some 70 years at least) but to make sure that the class D ultrasound distortion cannot create audible intermodulations in tweeters. Is this possibility excluded? Is there any proof?
Wasn't it you that tested this and didn't find any problem? There was a thread on this forum about it.

Edit, found the thread.
 
Last edited:

ctrl

Major Contributor
Forum Donor
Joined
Jan 24, 2020
Messages
1,616
Likes
6,086
Location
.de, DE, DEU
I would like to see a valid evidence (a proper set of tests) that the class D distortion and mess above 20kHz is unimportant. Not because of thresholds of hearing (everyone knows Fletcher-Munson curves for some 70 years at least) but to make sure that the class D ultrasound distortion cannot create audible intermodulations in tweeters. Is this possibility excluded? Is there any proof?
It is not clear to me how the break-up frequency >22kHz of a tweeter is supposed to be excited?
In the normal 44kHz music signal the frequency range >22kHz is not included.

Are you saying that the amplifier always produces "noise" above 22kHz regardless of the music signal?



Could we measure IMD which would be especially caused by the tweeter break-up frequency >22kHz? Sure, no problem at all.

Here a tweeter with a very pronounced break-up frequency at 25.5kHz.
First the frequency response (FR) of the tweeter measured in the near field:
1648648210412.png

Let's consider four cases with a two tone sinusoidal signal of 4kHz + 25.5kHz (tweeter break-up resonance).

The yellow graph shows a pure sine tone measurement of 4.1kHz (slightly shifted, so that nothing is covered) as an overlay for each measurement, in order to better distinguish IMD from HD.
(ignore the noise floor, it depends on the number of averages and the traffic outside my window.)

1. the 4kHz and 25.5kHz signals have about the same SPL.
2. the signal at 25.5kHz is attenuated about -22dB.
3. the signal at 25.5kHz is about -30dB attenuated.
4. the signal at 25.5kHz is about -35dB attenuated.
1648648652811.png 1648648685985.png1648648702593.png 1648648725665.png

Already at -35dB attenuation of the 25.5kHz signal at the tweeter break-up resonance (which is the worst case for signals >22kHz) the resulting IMD hardly matters anymore.

What attenuation compared to the music signal would the "ultrasound distortion" of the amplifier be? -30dB, -40dB, -50dB,...?
 
  • Like
Reactions: pma

Kijanki

Active Member
Joined
Dec 16, 2021
Messages
106
Likes
69
I would like to see a valid evidence (a proper set of tests) that the class D distortion and mess above 20kHz is unimportant. Not because of thresholds of hearing (everyone knows Fletcher-Munson curves for some 70 years at least) but to make sure that the class D ultrasound distortion cannot create audible intermodulations in tweeters. Is this possibility excluded? Is there any proof?
Modulation (creation of new, audible signals) requires non-linear element. Tweeter's membrane is likely to have some non-linearity, but only when it moves - not likely at 450kHz.
 

Mnyb

Major Contributor
Forum Donor
Joined
Aug 14, 2019
Messages
2,640
Likes
3,609
Location
Sweden, Västerås
Thanks for doing this interesting comparison. Contributing to the learning part of this forum :)
 
Last edited:
  • Like
Reactions: pma
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
A/B test
I have made a comparative listening A/B test between the SAB class AB amplifier and Hypex NC252MP class D amplifier. I have a switch-box with remote control. Output voltage level was matched within 0.03dB at the speaker input. The listening level was 75-80dBA, up to 82dBA, measured by a SPL meter.

The music used:
  • Santana Space Shifter
  • Rolling Stones Bridges to Babylon
  • Lynyrd Skynyrd
  • Mozart Colloredo Serenade (Linn Records)
Listening impressions: the sound difference is very small, but there is some -

NC252MP: more dry, narrower space, less depth, subjectively less dynamic
SAB: more punchy, more space, subjectively deeper bass

Seamless switching, no pops-clicks.
 

dc655321

Major Contributor
Joined
Mar 4, 2018
Messages
1,597
Likes
2,235
A/B test
I have made a comparative listening A/B test between the SAB class AB amplifier and Hypex NC252MP class D amplifier. I have a switch-box with remote control. Output voltage level was matched within 0.03dB at the speaker input. The listening level was 75-80dBA, up to 82dBA, measured by a SPL meter.

The music used:
  • Santana Space Shifter
  • Rolling Stones Bridges to Babylon
  • Lynyrd Skynyrd
  • Mozart Colloredo Serenade (Linn Records)
Listening impressions: the sound difference is very small, but there is some -

NC252MP: more dry, narrower space, less depth, subjectively less dynamic
SAB: more punchy, more space, subjectively deeper bass

Seamless switching, no pops-clicks.

Do you intend to remove the peeking and add some iterations to your testing program?
 

ctrl

Major Contributor
Forum Donor
Joined
Jan 24, 2020
Messages
1,616
Likes
6,086
Location
.de, DE, DEU
Listening impressions: the sound difference is very small, but there is some -

NC252MP: more dry, narrower space, less depth, subjectively less dynamic
SAB: more punchy, more space, subjectively deeper bass
Assume that you can hear a difference and that the frequency responses of the amplifiers are almost identical (with constant impedance load). Don't know much about electronics, so excuse my naive questions.

What kind of speakers are you using for listening tests? Do they have a highly variable impedance?
Since loudspeakers can have impedance peaks of 50 ohms (or more), my question would be to make sure that this does not affect the frequency response or distortion of the used amplifier.

Have you tested the amplifiers with variable impedance load? Optimal would be a 50 Ohm peak around 40Hz and around 2-3kHz.
 

pogo

Major Contributor
Joined
Sep 4, 2020
Messages
1,241
Likes
382
NC252MP: more dry, narrower space, less depth, subjectively less dynamic
SAB: more punchy, more space, subjectively deeper bass
These results were actually to be expected. The SAB based on the Kenwood design should has a DF=160 (50Hz, 8Ohm) and does not keep this constantly high over the entire FR, compared to an NC252MP which is generally much higher. See also here @4Ohm:
Link

subjectively deeper bass = The room modes are more strongly excited by the chassis swing-out behavior, which are not contained in the original signal.
 
Last edited:
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
These results were actually to be expected. The SAB based on the Kenwood design should has a DF=160 (50Hz, 8Ohm) and does not keep this constantly high over the entire FR, compared to an NC252MP which is generally much higher. See also here @4Ohm:

See here the real-world. Less than +/- 0.1dB.

SAB with_without quadral.png


You have forgotten to mention the cable impedance effect:


It makes more than the power amplifier DF in this case.
 
Last edited:

pogo

Major Contributor
Joined
Sep 4, 2020
Messages
1,241
Likes
382
See here the real-world.
I cannot confirm this and there must be more, especially in the time domain.
Here is an interesting video in german from a sound engineer, who explains which technical data are actually relevant and audible nowadays:
Link
You can also add a subtitle in your native language.

It doesn't seem to be that simple and especially with potent speakers like Magico there are differences to be heard, see also here:
Link
 
Top Bottom