INVITATION: Can you hear high harmonic distortion?

[Archimago]

Hi guys and gals, just wanted to invite folks here to a new test to try out in early 2020! It's based on Paul K's Distort software.

You know that measurements of audio gear typically will include harmonic distortion tests. We see numbers like 1% THD, 0.1% THD+N, etc... all the time. Typically also included with graphs in DAC, amp, preamp, phono cartridge results.

But what does this mean when it comes to listening and with real music?

Here's an opportunity to try for yourself and let me know what you hear - preferably using the best sound system you have! As usual, the results will be collected, analyzed, and published once the data gather is done by the end of April 2020. Plenty of time to listen and let me know what you hear.

Remember that I have used rather high amounts of distortion in some of these samples... Theoretically, this should be nowhere as difficult as "can you hear 16-bit vs. 24-bit?" or "24/96 vs. 16/44.1" tests.

Test: Is high Harmonic Distortion in music audible?

Thanks in advance...

 

[daftcombo]

Hi Archimago,

And thanks for your invitation.

I might be interested (a lot, actually) but your website doesn't appear properly in Firefox and is hardly readble (at least for me). See by yourself:

 

[pma]

It depends on music samples used, but, in general, non-linear distortion is usually inaudible up to high levels, especially low order distortion. I intentionally use the term non-linear and not harmonic, because it is a universal term. I made a lot of tests of this kind. As a result, SINAD is telling almost nothing, if it is not absolutely horrible.

 

[maty]

...As a result, SINAD is telling almost nothing, if it is not absolutely horrible.

I agree. The reality is too complex, both in amplification and loudspeakers.

As for the Internet tests, the vast majority have not optimized the O.S. (Win, Linux...) for multimedia, which is an extremely limiting factor.

 

[threni]

Hi Archimago,

And thanks for your invitation.

I might be interested (a lot, actually) but your website doesn't appear properly in Firefox and is hardly readble (at least for me). See by yourself:

View attachment 48494

Firefox has a reader view button which gets rid of a lot of annoying ads and terrible web design decisions.

 

[Fluffy]

I did the test. The results were very inconclusive.

Hi Archimago,

And thanks for your invitation.

I might be interested (a lot, actually) but your website doesn't appear properly in Firefox and is hardly readble (at least for me). See by yourself:

View attachment 48494

Looks fine for me on firefox:

 

[Tks]

It depends on music samples used, but, in general, non-linear distortion is usually inaudible up to high levels, especially low order distortion. I intentionally use the term non-linear and not harmonic, because it is a universal term. I made a lot of tests of this kind. As a result, SINAD is telling almost nothing, if it is not absolutely horrible.

SINAD in tandem with measured 'noise' and THD and others like IMD or SNR's separately I think adequately packages things into another sort of easy-to-grasp quick-look on what generally expected performance can be...well, expected - in relative terms.

But sure, I can somewhat see what you're saying in the same way if a display maker advertises "very good HDR" or "VESA HDR 600 certified". That only tells of some minimum spec performance of high dynamic range the display is capable of, but not much of that SKU's particular performance. Likewise when I see a device with 120dB SINAD, I'm never going to think "hmm maybe this device has horrendous THD performance, something to the tune of old garbage tube amps with distortion artifacts that could be heard -23db down or something" or noise level at max output being something like -15db down. That's just not happening, nor have I ever seen such a thing.

SINAD seems like an adequate summation, (unless you have a singular value that better demonstrates as much informative value as SINAD in a quick glance, I can't see why you would say something like 'absolutely terrible').

 

[Wes]

We will never know what is an adequate summation without data. Otherwise it is just guessing, tho educated guessing.

 

[Tks]

We will never know what is an adequate summation without data. Otherwise it is just guessing, tho educated guessing.

I mean if we're going to get THAT philosophical. I think you may have serious work cut out coming to a consensus with others on what level constitutes, or satisfies the definition of 'adequate'. My usage of it was within contextual relativism - that being, relative to other values taken singularly. You may have different ideas of what adequate mean though, so I don't know if that is enough for you as a definition.

 

[Wes]

we don't do science by consensus

 

[Tks]

Sure but we "do communication" with agreed upon definitions for words that are phonetically or textually produced.. I'm saying unless you establish what you're talking about when you say something, then -by definition- no one will know what you're even talking about.. literally. So unless you have an interpretation on what you feel "adequate" constitutes/means, then I won't have a clue on what it is you're trying to communicate, let alone move on to scientific or empirical efforts to classify things under such a word.

 

[pkane]

This blind test is designed to get a sample of people, who are presumably interested in quality audio reproduction, to report their sensitivity to non-linear distortion through their favorite/best/most expensive audio system. The goal is to get a sense of both, audibility of such distortions and individual preferences to see if larger patterns emerge.

Many more tests are possible, for example, testing for specific shape of non-linearity at the same or different SINAD level, different levels of noise, and even different shapes of the noise floor itself. All of these tests can be performed today by each of us, individually. What @Archimago is doing here is gathering (what I hope will be) a lot more statistical information from a larger group. While far from being proof-positive, this is a start of a process to move the needle in that direction, IMHO.

 

[Wes]

tks, do you have a specific question?

or are you in search of a philosophy forum?

 

[PierreV]

I still haven't fully recovered from listening to "Wild World" repeatedly, and here comes another test... I think I know who I should blame
Wonderful blog btw.

 

[danadam]

Results:

• part I - Procedures & Settings
• part II - Respondent Results

 

[Blumlein 88]

Kind of a mixed bag of results. Looks like 3% is not obvious with music over speakers. Is maybe notable among those using headphones. As headphones can usually be lower distortion than speakers that isn't a problem. And apparently speakers often being a 1% or so distortion is not a problem either. Since transducers are about the only thing that could have that level of distortion normally if fits with the idea distortion pretty much is a non-issue in modern audio.

So give me low noise and flat FR. We are good.

Of course arch came to the opposite conclusion. I think he is really reaching with the idea the 2nd and 3rd order distortion at such levels is an explanation for his results.

 

[flipflop]

Even if sample C was preferred over sample B, which I doubt due to the lack of randomization of the test files and lack of confidence intervals in the article, it's not a reason for consumers to go out and buy poorly performing products. That piece of information is mainly relevant to the people who participate in the creative process (i.e., artists, mixing- and mastering engineers, etc.).

Archimago's own conclusions are sensible.

 

[Blumlein 88]

Even if sample C was preferred over sample B, which I doubt due to the lack of randomization of the test files and lack of confidence intervals in the article, it's not a reason for consumers to go out and buy poorly performing products. That piece of information is mainly relevant to the people who participate in the creative process (i.e., artists, mixing- and mastering engineers, etc.).

Archimago's own conclusions are sensible.

I'm not being critical of Archimago. He is doing something to get some results for everyone to look over. I wouldn't even say his conclusions are not sensible. Yet, I think they aren't very much proof of the lower order distortion being preferred. More of a possibility to investigate further. One which might be true.

 

[pkane]

The comment about low order distortion being preferred was suggested by me as a possible explanation for the apparent preference for the -75dB THD sample. It’s not supported by the data from this particular test, but does make for an interesting follow-up