Is it audiophoolia to care about SINAD differences which have no correlation in blind listening tests? H2/H3 distortion 'enriches the sound'?

[BobbyTimmons]

Most of the variability found in DACs and amplifiers wouldn't correlate with findings in blind listening tests. Research even shows lower order distortion can be a good thing for some ('Recent studies have clearly shown the human PREFERENCE for THD distortion of low order or at higher amplitudes. These are viewed subjectively as enriching the sound', Earl L. Geddes & Lidia W. Lee, 2008).

 

[DonR]

It's a preference so there is no right or wrong. Is it hi-fi? No.

 

[Anton D]

The SINAD race beyond the audible threshold is snake oil, of course.

 

[BobbyTimmons]

The Fosi Audio ZA3 has a very good SINAD at 5 Watt/4 Ohm load.

When SINAD is 'bad' enough to be just audible in the tube amps in the red with over 0.1% THD that could be preferred in some blind listening tests when it includes more audible H2/H3 distortion ('Recent studies have clearly shown the human PREFERENCE for THD distortion of low order'). The measurement which might be audible for some young people with sensitive hearing is the load dependent frequency response where the performance is worse than a typical Class A/B amplifier.

The example product has a great price/performance ratio. It looks great for the price. Just from the consumer standpoint the SINAD results won't correlate with results in blind listening tests unless at the bottom end (Luxman tube amps) which might add H2 and H3 distortion. Research says lower order distortion is preferred in listening test.

 

[Deleted member 21219]

Research says lower order distortion is preferred in listening test.

Dr. Geddes didn't say that the preference was a good thing, he just noted that it existed. After all, diabetics prefer sweets.

Was there any data on number of participants or a description of their statistical makeup, protocols for the equipment used for comparisons, or specific degrees of distortions under discussion? I'd be interested.

Jim

Note: Post has been edited.

 

[kemmler3D]

I think caring about 120 vs. 100 SINAD is a little off-base but not totally foolish. On one hand, it gives you margin of error for suboptimal gain structures. And on the other hand, it does show (at least in specific ways) quality of engineering for the dollar.

If your daily driver can go 200 or 300 Km/H doesn't really make a difference in daily life, where you rarely break 100. However, you would probably be happy to get the 300 Km/h car for the same money. I think SINAD-chasing is something like that.

As for wanting more distortion in your playback instead of less... it's a matter of taste / point of view. I say get the most hi-fi gear you can and then use DSP for H2/H3 if you must. For me accuracy is a goal unto itself, because I want to hear "the real thing" - but for others accuracy is an afterthought compared to subjective impressions.

In the end, my POV on "the real thing" or accuracy / authenticity is really just another arbitrary preference, but it has the advantage of being (more) objectively measurable.

 

[Anton D]

After all, diabetics prefer sweets.

Diabetes is a hobby now?

 

[Deleted member 21219]

Diabetes is a hobby now?

No, but "preference" is a force with some power ... its effects need to be taken into consideration.

Jim

 

[BobbyTimmons]

I think caring about 120 vs. 100 SINAD is a little off-base but not totally foolish. On one hand, it gives you margin of error for suboptimal gain structures. And on the other hand, it does show (at least in specific ways) quality of engineering for the dollar.

If your daily driver can go 200 or 300 Km/H doesn't really make a difference in daily life, where you rarely break 100. However, you would probably be happy to get the 300 Km/h car for the same money. I think SINAD-chasing is something like that.

As for wanting more distortion in your playback instead of less... it's a matter of taste / point of view. I say get the most hi-fi gear you can and then use DSP for H2/H3 if you must. For me accuracy is a goal unto itself, because I want to hear "the real thing" - but for others accuracy is an afterthought compared to subjective impressions.

In the end, my POV on "the real thing" or accuracy / authenticity is really just another arbitrary preference, but it has the advantage of being (more) objectively measurable.

The measurement is telling us something if it correlates to preferences in blind listening tests. Has there been research showing that magnitude of differences in SINAD commonly found in amps and DACs does that more than cable lifters?

 

[BobbyTimmons]

Dr. Geddes didn't say that the preference was a good thing, he just noted that it existed. After all, diabetics prefer sweets.

The measurements are only pertinent to sound quality if they can be correlated to consumer preferences in blind listening tests.

 

[kemmler3D]

Has there been research showing that the differences in SINAD commonly found in amps and DACs does that more than cable lifters?

Yes, at least in that it's possible (easy, even) to hear the differences quite clearly, if you amplify the signal enough.

Beyond that - The theoretical concept behind >80dB SINAD (up to 120+) as I understand it is based on the maximum dynamic range of human hearing. Humans can hear a dynamic range of something like 130dB. So, if we want an audio system that can cover the entire audible range without any (even in theory) audible flaws, we need that much SINAD plus flat FR out to 20khz or a bit beyond.

I am not aware that preferences for high SINAD have been established, in this case it's a theoretical approach, not an empirical one.

 

[Curvature]

'Recent studies have clearly shown the human PREFERENCE for THD distortion of low order or at higher amplitudes. These are viewed subjectively as enriching the sound'

You misleadingly copied the text from a DIY audio thread, not a study: https://www.diyaudio.com/community/threads/geddes-on-distortion-perception.121253/

In the two studies Geddes made with his wife, one of them included subjective testing and showed a very low negative correlation to THD. "Clearly shown" is an overstatement. Geddes never returned to the topic seriously.

THD, IMD, MTD, the list goes on, are complex physical measurements, not based in subjective or perceptual criteria. Anything a person can conclude based on available studies of those measurements is tenuous if you want concrete, detailed answers. What they allow is broad generalization. There's no comprehensive model of audibility of nonlinear distortion yet.

Klippel did a good job presenting certain thresholds:

 

[BobbyTimmons]

Yes, at least in that it's possible (easy, even) to hear the differences quite clearly, if you amplify the signal enough.

Beyond that - The theoretical concept behind >80dB SINAD (up to 120+) as I understand it is based on the maximum dynamic range of human hearing. Humans can hear a dynamic range of something like 130dB. So, if we want an audio system that can cover the entire audible range, we need that much SINAD plus flat FR out to 20khz or a bit beyond.

I am not aware that preferences for high SINAD have been established, in this case it's a theoretical approach, not an empirical one.

Detecting even unmasked distortion at -60db (0.1%) is almost impossible. Perceiving the difference between 0.001% and 0.0001% THD+N doesn't seem likely even if the distortion wasn't lower order harmonics which are masked.

On the Klippel site a lot of people were having trouble perceiving down to -60db (0.1%). That's when they use audible types of distortion. Unlike in this test real world THD+N has a lot of inaudible lower order harmonics which are masked by the music. Perceiving -60db (0.1%) in THD+N in typical equipment should be a lot more difficult.

(Not my result, just an illustration)

 

[BobbyTimmons]

You misleadingly copied the text from a DIY audio thread, not a study: https://www.diyaudio.com/community/threads/geddes-on-distortion-perception.121253/

In the two studies Geddes made with his wife, one of them included subjective testing and showed a very low negative correlation to THD. "Clearly shown" is an overstatement. Geddes never returned to the topic seriously.

THD, IMD, MTD, the list goes on, are complex physical measurements, not based in subjective or perceptual criteria. Anything a person can conclude based on available studies of those measurements is tenuous if you want concrete, detailed answers. What they allow is broad generalization. There's no comprehensive model of audibility of nonlinear distortion yet.

Klippel did a good job presenting certain thresholds:

I haven't 'misleadingly' anything. I'm using @MarnixM's citation format.

 

[Curvature]

I haven't 'misleadingly' anything. I'm using @MarnixM's citation format.

Which is equally misleading.

World of difference between a thread and published research. Onus is on you, as OP, to understand the sources and present them with the right weight.

 

[Chrispy]

I more look at it as just one might be technically very slightly better and more desireable in that respect. Not particularly meaningful on its own in any case, not something I use as a major indicator. I more think of ranges of spec as to how suitable the dut is for possibilities of audible differences....

 

[BobbyTimmons]

Which is equally misleading.

World of difference between a thread and published research. Onus is on you, as OP, to understand the sources and present them with the right weight.

If you think text quotes with all capital letters like 'human PREFERENCE for THD distortion' would be used in published research then reading 'misleading' citation formats on an internet forum is the least of your worries. It's a text written by Geddes with the date given. He's writing informally, he uses a more exaggerated writing style than in an academic paper. He's writing with ALL CAPS to show the weight he gives to the finding. He found a preference for lower order distortion. We review with different assumptions as SINAD includes lower order distortion. Distortion with those components below 0.1% is not going to be audible. DACs/amplifiers with more than 0.1% THD+N don't seem common enough to be an issue.

 

[Deleted member 21219]

The measurements are only pertinent to sound quality if they can be correlated to consumer preferences in blind listening tests.

You've got it completely backwards.

Sound quality can only be determined by tests and measurements. Sound quality is determined b y a comparison of what comes out of a particular part of the reproduction chain (pre-amp, amp, DAC or speaker) with what went in. The phrase "sound quality" is a relative assessment of accuracy ... the closer the output is to the input, the higher the sound quality.

Perhaps you're thinking of the term "sound qualities", as in, "What sort of sound qualities does this violin have?", or, "I like the sound qualities of that 1950 Philco radio."

So a more accurate statement would be ...

"Consumer preferences in blind listening tests are only of statistical value if they are consistently able to be correlated to measurements."

Under no circumstances does this statement assign value other than statistical value to preferences. It says nothing about sound "quality". Why? Because preferences are only pertinent to the person who has them. Preference is a personal, or IOW, subjective thing, and on top of that, it may be totally unique.

As for what is "pertinent", and to whom ... for the population at large, tests and measurements are the only things that are "pertinent".

Jim

 

[BobbyTimmons]

You've got it completely backwards.

Sound quality can only be determined by tests and measurements. Sound quality is determined b y a comparison of what comes out of a particular part of the reproduction chain (pre-amp, amp, DAC or speaker) with what went in. The phrase "sound quality" is a relative assessment of accuracy ... the closer the output is to the input, the higher the sound quality.

Perhaps you're thinking of the term "sound qualities", as in, "What sort of sound qualities does this violin have?", or, "I like the sound qualities of that 1950 Philco radio."

So a more accurate statement would be ...

"Consumer preferences in blind listening tests are only of statistical value if they are consistently able to be correlated to measurements."

Under no circumstances does this statement assign value other than statistical value to preferences. It says nothing about sound "quality". Why? Because preferences are only pertinent to the person who has them. Preference is a personal, or IOW, subjective thing, and on top of that, it may be totally unique.

As for what is "pertinent", and to whom ... for the population at large, tests and measurements are the only things that are "pertinent".

Jim

The measurements are pertinent to perceived sound quality if they can be correlated to preferences in blind listening tests (actually double blind listening tests). That's how the whole field works. That's the groundwork for researchers like Sean Olive.

 

[BobbyTimmons]

I more look at it as just one might be technically very slightly better and more desireable in that respect. Not particularly meaningful on its own in any case, not something I use as a major indicator. I more think of ranges of spec as to how suitable the dut is for possibilities of audible differences....

Me too. Just it doesn't seem convincing that fussing over the inaudible sonic qualities of equipment designed for music reproduction is more rational than doing it over the color of the cardboard box it was delivered in. According to the science they are qualities which should have no humanly perceivable effect on the product's function.