THD has a much bigger effect on sound than you think

[Killingbeans]

Sorry. Again, words. This is not the way of ASR.

OK, I'm starting to get that lost cause feeling that pma mentioned earlier. Or maybe you are just on some weird crusade. Either way, I'm not going to waste more energy on this.

 

[AdamG]

Putting a feeler out to the Members involved in this, er, um, discourse. I will decide based on the number of like’s this post gets. Who here wants this thread finished and closed?

 

[Jimbob54]

Putting a feeler out to the Members involved in this, er, um, discourse. I will decide based on the number of like’s this post gets. Who here wants this thread finished and closed?

There is precedent for how this discourse goes if left open.

 

[Pdxwayne]

Putting a feeler out to the Members involved in this, er, um, discourse. I will decide based on the number of like’s this post gets. Who here wants this thread finished and closed?

People react negatively when asked to prove a statement with abx.....Strange that they feel free to ask me to do abx, yet they gave all kinds of excuses when asked to do the same. ASR way?

 

[NTK]

I was doing a little bit of research on this subject and came across this paper by Alex Voishvillo** of Harman.
[**Among the various accomplishments of Dr. Voishvillo, IMHO one may be particular noteworthy to ASRers is that Dr. Voishvillo is the designer of the dual-diaphragm/dual-voicecoil D2 compression tweeter of the JBL M2 speakers. Reference]

The paper is a good introduction to nonlinear distortions of loudspeakers and loudspeaker drivers.

(PDF) Assessment of nonlinearity in transducers and sound systems - From THD to perceptual models

PDF | Research of the audibility of loudspeaker nonlinear distortion has not shown good correlation between traditionally used metrics (harmonics and... | Find, read and cite all the research you need on ResearchGate

www.researchgate.net

Interesting tidbits from his paper was that in 1929 it was already known that intermod distortions are more important than harmonics.

Another one is that clean hard clipped music at 22.6% THD is quite tolerable, whilst, at merely ~1/10 magnitude, 2.9% THD cross-over distortion (when measured with a sine tone at a particular amplitude) was not.


For electronics, this old paper from Richard Cabot (formerly of Audio Precision) is also quite informative.

https://www.researchgate.net/publication/277870128_A_Comparison_of_Nonlinear_Distortion_Measurement_Methods

 

[VintageFlanker]

Who here wants this thread finished and closed?

No pun intended (or maybe there is), but I would pay more attention to any other WTF thread mike7877 would have the good idea to open in the future...

 

[audio2design]

DISTORT app doesn't do anything at 0dBFS -- it's up to the user to decide what signal to feed it as input and at what level. Internally, all computations are in 64-bit floating point, and any results that exceed 0dBFS are reported as clipping.

How are you calculating maximum signal excursion?

 

[Pdxwayne]

OK, I'm starting to get that lost cause feeling that pma mentioned earlier. Or maybe you are just on some weird crusade. Either way, I'm not going to waste more energy on this.

Just download the files and listen. No need to do abx. Just listen and let me know if you can easily hear a difference.

Easy and fast.

 

[audio2design]

If the transfer function for a amp or amp + speaker is Known why don’t we crate an inverse of this in DSP and cancel the distortion?

Seems to work in RF PAs

Single function and often limited signal range, defined loading, etc in RF makes it an easier problem not to mention the acceptable distortion is much higher normally.

 

[audio2design]

OK, I'm starting to get that lost cause feeling that pma mentioned earlier. Or maybe you are just on some weird crusade. Either way, I'm not going to waste more energy on this.

My one word No answers very early in the thread feel prophetic now.

 

[peng]

I was doing a little bit of research on this subject and came across this paper by Alex Voishvillo** of Harman.
[**Among the various accomplishments of Dr. Voishvillo, IMHO one may be particular noteworthy to ASRers is that Dr. Voishvillo is the designer of the dual-diaphragm/dual-voicecoil D2 compression tweeter of the JBL M2 speakers. Reference]

The paper is a good introduction to nonlinear distortions of loudspeakers and loudspeaker drivers.

(PDF) Assessment of nonlinearity in transducers and sound systems - From THD to perceptual models

PDF | Research of the audibility of loudspeaker nonlinear distortion has not shown good correlation between traditionally used metrics (harmonics and... | Find, read and cite all the research you need on ResearchGate

www.researchgate.net

Interesting tidbits from his paper was that in 1929 it was already known that intermod distortions are more important than harmonics.
View attachment 160967
Another one is that clean hard clipped music at 22.6% THD is quite tolerable, whilst, at merely ~1/10 magnitude, 2.9% THD cross-over distortion (when measured with a sine tone at a particular amplitude) was not.
View attachment 160968

For electronics, this old paper from Richard Cabot (formerly of Audio Precision) is also quite informative.

https://www.researchgate.net/publication/277870128_A_Comparison_of_Nonlinear_Distortion_Measurement_Methods

Of course, but if THD is very low such as below 0.01%/-80 dB, chances would be good that IMD would not be so high as to being an audible issue. Same for crossover distortions, that sometimes get cited as by people as a reason why THD is not that important because at low level output it is crossover distortions that is bad. Yet if one takes a look of math, at low level output such as -70 dB at 50 mW for the AVR-X3700H, it may not an audible issue either. A speaker with sensitivity at 90 dB/W/1m at would produce only about 67.5 dB from 3 meters, it would be tough to hear the harmonics due to crossover distortions at -70 dB, even though the harmonics resulted from COD/or XOD if preferred, are not related to the music.

 

[pma]

Keeping distortion below -80dB (HD or IMD, as you like it) and being composed mostly of low order harmonics means it would be inaudible under any circumstances.

 

[richard12511]

Is that distortion in the bass region audible?

Not even close to audible.

 

[richard12511]

It is in my experience. I guess I have experience with genuinely high fidelity gear, that might be a requisite.

I have good experience with high end, high fidelity gear, and I have a very different experience than yours. In my experience, 0.1% distortion amounts to a difference smaller than the difference between the left and right speaker of the same model(and I don't really worry about that). Can you provide some audibility studies to support your conclusion that this is something we should worry about?

 

[sarumbear]

The sound energy in music in the 1-4khz range is 30-40db lower than bass, 15-30db than midrange.

If you define bass as below 100Hz then your figures are not representative of reality according to this research on LTAS, published by AES. See figure 4 on the paper below.

https://www.researchgate.net/profile/Anders-Elowsson/publication/317098414_Long-term_Average_Spectrum_in_Popular_Music_and_its_Relation_to_the_Level_of_the_Percussion/links/5ecedde3458515294518191d/Long-term-Average-Spectrum-in-Popular-Music-and-its-Relation-to-the-Level-of-the-Percussion.pdf?origin=publication_detail

 

[MRC01]

...
As you can see the oboe and clarinet have 3rd harmonics higher than the fundamental. It's why it sounds different from other instruments.
In case of the oboe harmonics are high even up to the 8th harmonic where in 'audiophool' amps their Harmonics usually only go as high up as 4th or 5th and then are below noise levels.

If you measured that flute in its bottom register, the first harmonic or two would be stronger than the fundamental. More generally, instruments have different harmonic structures in their different octaves, and when played at different volume levels. And musicians can alter the harmonic structure with embouchure changes (for strings, bowing or plucking changes) to produce different tone colors. All this is even more reason that harmonic distortion is masked by actual music.

 

[pkane]

The transfer function is not only a function of level, but also of frequency (sometimes very heavily) and also of load impedance. This would be my question to @pkane, Paul, would you be able to implement frequency as a further variable in the Distort??

It's a lot more complicated. Frequency/time dependency in the transfer function requires much more extensive modeling/simulation. The main issue I have with it is how to make it easy to specify such a function that would resemble some realistic device, without asking the user to enter complex numbers into the app If you have some ideas on how to make this easier to specify, let me know.

There's already a simple, bandwidth-limited feedback simulation built-in, but this has very small frequency dependency. Here's an example of a device with no feedback:

Same device, but now with 20dB of negative feedback:

 

[danadam]

250hz_h2_to_h4_120_h5_60.wav and 1 more file

2 files sent via WeTransfer, the simplest way to send your files around the world

we.tl

Glad that you wish to try. Thx!

Not sure what's this all about but here you go:

Spoiler: ABX log

foo_abx 2.0.6d report
foobar2000 v1.4.3
2021-10-23 23:02:47

File A: 250hz_tone.wav
SHA1: 9bedd7cc84145a54cc2979ec4bbf86d51f0f082a
File B: 250hz_h2_to_h4_120_h5_60.wav
SHA1: b604c1acb2c2cdc3fd9bceba6aef1358de8d1a6e

Output:
ASIO : ASIO MADIface USB
Crossfading: NO

23:02:47 : Test started.
23:03:28 : 01/01
23:03:48 : 02/02
23:04:07 : 03/03
23:04:24 : 03/04
23:04:50 : 04/05
23:04:59 : 05/06
23:05:08 : 06/07
23:05:21 : 07/08
23:05:48 : 08/09
23:06:12 : 09/10
23:06:12 : Test finished.

----------
Total: 9/10
p-value: 0.0107 (1.07%)

-- signature --
750aef1b371029d1a7922ea86e32c3eebd24b940

 

[Pdxwayne]

Not sure what's this all about but here you go:

Spoiler: ABX log

foo_abx 2.0.6d report
foobar2000 v1.4.3
2021-10-23 23:02:47

File A: 250hz_tone.wav
SHA1: 9bedd7cc84145a54cc2979ec4bbf86d51f0f082a
File B: 250hz_h2_to_h4_120_h5_60.wav
SHA1: b604c1acb2c2cdc3fd9bceba6aef1358de8d1a6e

Output:
ASIO : ASIO MADIface USB
Crossfading: NO

23:02:47 : Test started.
23:03:28 : 01/01
23:03:48 : 02/02
23:04:07 : 03/03
23:04:24 : 03/04
23:04:50 : 04/05
23:04:59 : 05/06
23:05:08 : 06/07
23:05:21 : 07/08
23:05:48 : 08/09
23:06:12 : 09/10
23:06:12 : Test finished.

----------
Total: 9/10
p-value: 0.0107 (1.07%)

-- signature --
750aef1b371029d1a7922ea86e32c3eebd24b940

I have a different thread going on now..... Please comment there like if you think "everyone" can hear a difference. Also provide your hardware chain please.

Thanks!

Can you hear a difference? 250hz and 1250hz playing at same time, but 1250 Hz is at 60db down. Vs pure 250hz tone.

I was able to hear a difference using AKG 371, but not easy at all. Got abx results of 8/8. Then one member said "everyone" can hear a difference. I asked the member to give abx result but I got none. The other member chimed in to support the first member and basically said science already...

www.audiosciencereview.com

 

[AdamG]

As previously posted I was considering closing this thread. However, the conversation continues and occasionally a nugget of knowledge and wisdom slips out. As long as the conversation remains respectful, civil and relatively productive it will remain open. I withdraw my proposal to close this thread. Thank you for your understanding.