SINAD Measurements

[March Audio]

It is better only in a quantitative manner.

In a qualitative manner, the statement is undefined and without moorings.

That's the essence of this entire conversation -- those who think lower distortion numbers necessarily mean better audio quality.

Not science.

You know, this exact conversation has been going on since I was a teenager building audio gear in my garage. And still people are being duped by the "lower THD is better" imposture.

That's not what has been stated by ASR. Saying that below a certain figure it becomes transparent is absolutely moored in science.

 

[restorer-john]

To recapture scientific rigor, I suggest the following ==> simply add a new test to ASR’s procedures. The new test is simple: a pure noise reading. No signal, broadband (10Hz - 22kHz), no weighting, in dBu (uVrms). This is an absolute baseline test in our lab.

I and others have been calling for a baseline residual noise in uV forever. If S/N is presented, we can work back to it, but not from SINAD. Considering it's just a uV window on the AP's dashboard, I can't see the problem.

 

[March Audio]

I and others have been calling for a baseline residual noise in uV forever. If S/N is presented, we can work back to it, but not from SINAD. Considering it's just a uV window on the AP's dashboard, I can't see the problem.

I also think its a useful metric to show.

 

[restorer-john]

Mine is fully differential and does .003% THD at 2V line level out.

Mine had a limited slip differential and did 0-100 in....

 

[Xulonn]

This is a science-based high-fidelity consumer audio forum website that has more audio science and industry experts and luminaries visiting than other audio websites where I have participated since the 1990's. With rare exception, they post under their real identities, and most quickly gain the respect of the regular visitors to ASR. I get a bit of an Alice in Wonderland feeling when an anonymous commercial audio business figure arrives here and starts whining, pontificating, and calling people "dudes". It's probably the passive-agressive "nice and polite on the surface, while cutting and insulting underneath" style that irritates me.

I much prefer @restorer-john 's approach of simply asking that noise data from the AP dashboard be included with each review, a bit of data that is apparently is quite useful to him and some of our other resident experts. This simple approach is much preferable to ranting for pages about the unscientific nature of Amir's methods and decisions, as if insults will achieve a stated goal of publishing component noise measurements.

Can you do that, Amir? It would likely eliminate a lot of anguish!

Regarding the usefulness of SINAD, I have always accepted Amir's logic for including SINAD...

@amirm sez: SINAD or THD+N is not perceptually correlated measure. So it is hard to make an argument for audibility of any value regardless of conditions selected. You can have two amps with identical SINAD and one may have audible distortion and the other not. The reason to measure SINAD is to detect engineering excellence as a general measure.

...and appreciated his admission that although it can demonstrate evidence of good, thorough engineering - that the engineers have done a good job unhindered by marketing department demands or lack of skill and experience - it still might not be the perfect master benchmark for rankings

There are many manufacturers who have achieved high SINAD that seems to parallel the overall excellent performance of their components. To me, high SINAD indicates that the designer has paid attention too all of the details related to keeping the signal clean, and I have been impressed by the ability of some very small companies to achieve spectacular performance that matches a high SINAD.

Anyone care to do a correlation study for SINAD vs overall sonic excellence of the hardware in the ASR review archive?

 

[scott wurcer]

Mine had a limited slip differential and did 0-100 in....

I got my AMC Gremlin up to a 100 a couple of times.

 

[pma]

A small reminder - both SINAD and THD+N are pointless numbers if the measurement bandwidth is not provided with the numbers. Same applies to noise in uV, however I concur it would be nice to show it, together with measurement bandwidth.
How about SFDR, might be nice to cover as well.

 

[amirm]

@amirm does ASR have the funds for you to visit @signalpath, or for him to visit you and perform the test I suggested earlier?

I am happy to put forward $1000 reward for him to show that a 60 dB SINAD amp sounds different than a 100 dB one in controlled, double blind test that is recorded for us to see. We have others that can help raise this bounty if needed. I am that confident he can't tell the difference let alone have a reliable preference.

 

[amirm]

A small reminder - both SINAD and THD+N are pointless numbers if the measurement bandwidth is not provided with the numbers.

Measurement bandwidth is always provided in my tests. It is either in big letters on top of the graph, or the bottom of the dashboard. It is also documented in my tutorial: https://www.audiosciencereview.com/...derstanding-digital-audio-measurements.10523/

Couldn't be more clear than this. So what is the basis for the complaint?

 

[amirm]

If S/N is presented, we can work back to it, but not from SINAD.

I show dynamic range for DACs and SNR for analog devices. Have you not seen them?

 

[amirm]

Chart says "Distortion and Noise (SINAD) Higher Better". That means 120dB SINAD is BETTER than 20dB SINAD. There’s nowhere said that Higher SINAD is Better perceptual audio quality!

And that is all it says. It is so common to post graphs online and leave it to the poor reader to figure out if taller is better, or shorter. The indication on the graph says exactly that: as a number, higher SINAD is better than lower SINAD. It always is. It is a metric just like voltage is.

 

[amirm]

Because THD is not necessarily an indicator of perceived sonic performance. There is no peer-reviewed, scientific correlation between THD and perceptual quality (beyond a certain threshold). A higher-THD product may indeed perform far better than a lower THD product. There's no scientific correlation (beyond a certain threshold).

This is true and why you don't want to buy products with middling THD numbers. You would then have to figure out the psychoacoustics of which one has more audible distortion.

Get the SINAD low enough though, and every psychoacoustic analysis points to inaudibility so you are golden. You don't need to become an expert in science. Again this is stated cleanly in my published article on measurements: https://www.audiosciencereview.com/...derstanding-digital-audio-measurements.10523/

The beauty of this is that it is not very expensive to get such DACs or headphone amps. The latter comes in for just $99!

It is a bit like not washing your hands when you go to bathroom. You may not get sick from it. But we do know that if you wash them extremely well, you are not going to get sick from it.

 

[amirm]

1) ASR states that “120dB SINAD guarantees transparency.” This is not a scientific statement. It is a belief, or at best a “postulate.” It has not been proven that -120dB SINAD creates identical perceptual (ABX) performance in all audio designs, under all circumstances. Only after sufficient and rigorous ABX testing is completed and peer-review-documented on a vast range of -120dB SINAD audio gear (professional micamps, ADCs, DACs, power amplifiers, individual components, etc.) can such a statement be scientifically made. Until then, it is nothing more than a personal belief, a conjecture, a postulate.

That's like saying calculous is not math. Audio science is not just about ABX tests. The core of it in this topic is about psychoacoustics. That science is developed on the basis of countless listening tests in controlled experiments to find the detection thresholds for various types of signals. The bible of that is Fastl & Zwicker text: Psychoacoustics: Facts and Models

https://www.amazon.com/Psychoacoustics-Models-Springer-Information-Sciences/dp/3540650636

But you don't even need to read that text. Fletcher and Munson who were AT&T researchers, performed listening tests which determined threshold of hearing:

The threshold curve can also be derived objectively based on self-noise of the hearing system.

If we now keep distortion and noise below that curve, it is absolutely inaudible. You can jump up and down all you want but you cannot make a case for audibility of either.

Go in your studio and close the door. Do you hear bass frequencies rumbling? Likely not. Now get a SPL meter and you will absolutely measure low frequencies likely coming from miles away (freeway noise, etc.). You can see this in Fielder's analysis: https://www.audiosciencereview.com/forum/index.php?threads/dynamic-range-how-quiet-is-quiet.14/

We see that at 20 Hz, skywalker Scoring Stage has noise level of nearly 40 dB. But because it is below threshold of hearing, you cannot hear it.

This is psychoacoustic analysis. It is standard practice in peer reviewed articles on audibility of noise and distortion. It is referenced subh in my published article above:

References
“Noise: Methods for Estimating Detectability and Threshold, ” Stuart, J. Robert, JAES Volume 42 Issue 3 pp. 124-140; March 1994
“Dynamic-Range Issues in the Modern Digital Audio Environment, ” Fielder, Louis D., JAES Volume 43 Issue 5 pp. 322-339; May 1995

Fielder was the president of AES for a while and worked at both Ampex and Dolby. Please read the papers. Ask questions if not clear. But for heaven's sake don't say we don't follow science. We absolutely do.

 

[restorer-john]

I show dynamic range for DACs and SNR for analog devices. Have you not seen them?

OK, Amir. Just remind us all again, what units we measure residual noise amplitude with.

Not everything needs to be dB referenced to something else.

Residual noise is a directly comparable metric, unlike dynamic range or SNR which are not.

And which is easier to understand, SNR of 117dB with respect to 2.0V or a residual of 2.82uV? The bandwidth is already 22.4KHz on your dashboard anyway isn't it? I do not understand the resolute indifference to what is a standard metric, is quoted by numerous manufacturers (Benchmark et al) and is available on your dashboard in real time. Your gear can measure noise into unheard of levels and yet people are left to break out the calculator every time to get a noise figure in uV.

 

[signalpath]

I get a bit of an Alice in Wonderland feeling when an anonymous commercial audio business figure arrives here and starts whining, pontificating, and calling people "dudes". It's probably the passive-agressive "nice and polite on the surface, while cutting and insulting underneath" style that irritates me ... This simple approach is much preferable to ranting for pages about the unscientific nature of Amir's methods and decisions, as if insults will achieve a stated goal of publishing component noise measurements.

I am not anonymous.

I have not insulted Amir. In fact, just the opposite.

Dude.

 

[amirm]

And which is easier to understand, SNR of 117dB with respect to 2.0V or a residual of 2.82uV?

117 dB every day of the week and twice on sunday. No one knows what to do with a microvolt number. But everyone can understand that if you play at peak of 120 dB, your noise floor will be -3 dB SPL so at threshold of audibility. What on earth can you do with a 2.82 microvolt?

For psychoacoustic analysis, dB is the right number. This is why I don't like to use percentages for THD+N either. Who is going to remember 0.05% versus 0.009%?

For the same reason I hate using dBV, dBu, etc.

 

[March Audio]

117 dB every day of the week and twice on sunday. No one knows what to do with a microvolt number. But everyone can understand that if you play at peak of 120 dB, your noise floor will be -3 dB SPL so at threshold of audibility. What on earth can you do with a 2.82 microvolt?

For psychoacoustic analysis, dB is the right number. This is why I don't like to use percentages for THD+N either. Who is going to remember 0.05% versus 0.009%?

For the same reason I hate using dBV, dBu, etc.

I know where John is coming from, its a useful number to guage if you are likely to hear any amp hiss. The sensitivity of the speaker dictates if its audible, not the SNR dB value. We can use it as a number to judge how relatively noisy the amp is. Yes with SNR we can calculate backwards if we know the max output voltage, but it is most definitely simpler to see just the residual noise voltage.

EDIT: This also would be very useful with headphone amps as we have some extremely sensitive IEMs out there.

 

[restorer-john]

What on earth can you do with a 2.82 microvolt?

OK Amir. Your ficticious D/A converter has the above numbers in my post. You input 500mV (typical) from my example above into your amplifier with 28dB of gain and a 100uV noise floor. Which dominates, the amplifier's residual or the D/A converters residual? What is the total noise floor and is it audible on your speakers which are 91dB@1M/2.83V? (assume A wtd)

 

[restorer-john]

I know where John is coming from, its a useful number to guage if you are likely to hear any amp hiss.

Exactly. Thanks.

 

[signalpath]

I am happy to put forward $1000 reward for him to show that a 60 dB SINAD amp sounds different than a 100 dB one in controlled, double blind test that is recorded for us to see. We have others that can help raise this bounty if needed. I am that confident he can't tell the difference let alone have a reliable preference.

Absolutely. With one caveat. I get to pick the amplifiers under test. I say this because with some amplifiers, there is no detectable difference. With others, there are marked sonic signatures. And the test will be done at our lab / studio using our blinded AB technique and reference chain. One neutral observer attending.

Amir, I'm not sure why you think this test will fail. You've already presented evidence of an AES amplifier listening study in which at least some ABX differences were in fact repeatably detected by blinded listeners. This is all the evidence you should need. The answer is in your lap. Differences were detected. Mission accomplished, Done. Science wins. The premise that "IC differences cannot be detected" was proven invalid by one lone peer-reviewed AB test. Why do you continue to insist otherwise in the face of peer-reviewed evidence?

What would be even more instructive is to put together 2 or 3 top-end -130EIN / -120dBTHD industry micamps and do some DPA-based recording of difficult sources (rattling keys, bell tree, etc.), align the tracks on AB test DAW (Sequoia), and do the same blinded comparison testing. Easily detected differences, likely due to input capacitor transient behavior rather than periodic / harmonic distortions of the active stages.

In fact, we recently did a similar test for the most advanced micamp we've ever developed (premiering at NAMM in 2 weeks!). The test was to AB the sonic signature of various candidate input capacitors. At least two or three were rejected almost immediately. 10/10. A number were in the middle, OK but iffy. And two parts were clearly more transparent to the source, at least 8/10. We are using those qualified parts in the new product. (keep in mind, none of the capacitors under test impacted the baseline noise or THD performance of the signal path -- the source of detectable distortion is something other than THD -- which is another conversation that goes to root of why -120dB THD does not "guarantee transparency")

Anyway, if you let me pick the amplifiers myself, the AB differences will be even more pronounced than those detectable IC's in the AES Journal test. It's sort of not fair (to you).

But I'm puzzled about the $1,000 reward. If I do get a reward, I'll donate it back to the ASR community.

Now to the more important issue.

None of this is about me. Testing my AB chain, claims, ears ... is a silly side-show. Audio claims aren't settled by one test on one person. Rather, it takes a large population of samples and trials and participants to achieve peer-review acceptance (like the AES paper you highlighted). But you know that.

So, I'll make this deal. If I can score minimum 8/10 on a simple blinded comparison of two amplifiers, you will make my suggested changes to the ASR community site.

Night night. I'll address all your other recent comments tomorrow.