Why SINAD doesn't matter

Discussion in 'Blind Testing and Psychoacoustics' started by purr1n, Aug 28, 2021.

  1. Pancakes

    Pancakes Friend

    Pyrate Contributor
    Joined:
    Aug 13, 2020
    Likes Received:
    1,417
    Trophy Points:
    93
    Location:
    Atl
    ASR is a cult of lazy. Period. And Amir is the laziest f**k of all.

    You'll never hear anything even remotely close to the following type of understanding and nuance (taken from the FirstWatt F6 manual):

    "One of the most fascinating things is the whole thing about absolute phase – how much can you hear absolute phase? Not whether your speakers are in or out of phase with each other, but absolute phase. It has become quite apparent that it matters. It matters in the context of how you look at the second harmonic structure of the amplifier as it relates to the speakers and to what comes out of the recording process. I mentioned that positive-going phase for second harmonic has a particular sound. If you flip that characteristic you've got a different sound.
    I'm not here to tell you what you like, I've noticed that when you get reasonably experienced listening to that effect you can then go through your record collection, often deciding which recordings are in-phase and out-of-phase. I find that totally fascinating.
    It relates to something I can talk about briefly, one of my favorite soap-box subjects. If you go into the literature of psychoacoustic perception, there is a very good book by Diana Deutsch at UC San Diego called “The Psychology of Music” and in it there are several chapters talking about how the low level neural networks of the brain take the data from the ear and what they do with it like the bureaucracy at the DMV, and you have an army of these things and for each of them the job it to make a decision – what goes with what, and these are called grouping mechanisms. Each bit of the network takes disparate bits of audio information and decides whether they go together or not. The system is sensitive to such things as loudness, timing, pitch, harmonic structure and phase. Decisions are made at very low levels and then get passed upwards for increasingly more abstract decisions until the final result, the “executive summary” is handed to the guy who sits behind your eyes at the control panel and imagines he's in charge.
    So what are we doing when we play with the distortions of an amplifier? Well, we're just fooling ourselves, fooling the ear and the brain. And sometimes that's a good thing. It's plausible to me that if you tag the sound with a particular characteristic (I'm not claiming that expertise) it seems to slip more easily through these neural systems like poop through a goose, and the decision-making process is easier. There is a lot of work going on in the brain when we are talking about listening – a vast army of neurons working this thing, and if you make their life easier, they aren't working as hard.
    We are talking about listener fatigue, talking about people who get tired after a half hour and shut the music off versus guys who go through their entire record collection all night long. We are literally talking about fatigue – the brain gets tired.
    So why do we try to fool the ear? It makes people happy. It helps them to relax while they listen to music and try to forget all the terrible problems in the world. I'm not here to deliberately create distortion, but if my simple little circuits are going to have some distortion anyway, I can at least try to organize it the way I want."
     
    • Like Like x 11
    • Epic Epic x 4
    • List
  2. Brause

    Brause Friend

    Pyrate Contributor
    Joined:
    Nov 25, 2016
    Likes Received:
    467
    Trophy Points:
    63
    Location:
    Calgary, AB
    Objective does not exist. Any measurement setup is subjective, and the parameters within such setup are, too. He uses a modified $69 soundcard and freeware.

    Next, Amir does not establish a relationship between his measurements ("quantity") and sound quality ("quality").

    There is no direct relationship between quantity and quality - and if there was, his handful of graphs would not be sufficient to characterize a device.

    His measurements are biased and his listening is biased towards his measurements ("expectation bias"). His listening is reduced to "clear, distortion-free signal" and "powerful sound".

    In summary, this is bias on top of another bias. A second degree bias so to speak.
     
  3. Thad E Ginathom

    Thad E Ginathom Friend

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    14,134
    Trophy Points:
    113
    Location:
    India
    @Shredicus, don't simply shit on the word, which is, sadly, how the two ends of the hifi spectrum treat each other. (they also make up their own definitions of each others' labels and then attack those.)

    There is nothing wrong with objectivity, just as there is nothing wrong with subjectivity. There is nothing wrong with measurements: there is everything wrong with the misrepresentation of measurements, especially a single measurement. And, as Woland comments, making a cult out of it. Work, the devil, eh, Woland? ;)

    Most of us have done something objective in our audio lives, from comparing frequency response charts to considering how the mind and its biases work. All of us, excepting perhaps Amir's cult, have done a very great deal of something subjective: the simple and wonderful experience of listening to good music.

    Hmmm... I had used the words objectivism and subjectivism, but the ismness of it all is the big problem. I changed the isms to ity. I think it's healthier.
     
    • Like Like x 2
    • Epic Epic x 2
    • List
  4. Woland

    Woland Friend

    Pyrate
    Joined:
    Jan 13, 2021
    Likes Received:
    1,311
    Trophy Points:
    93
    Location:
    a friendly land
    Thanks for your excellent note @Thad E Ginathom

    When I used "cargo cult", it was a very specific reference familiar to people who have studied philosophy of science -- I should have explained it more. It means following the appearance of science, but without having the actual substance.

    The "cargo cult" term was used in this way by physicist Richard Feynman

    I've seen cult-like behavior in ASR threads, but I don't make a broad claim about the site because I don't know it and its user base well enough. I have seen enough to conclude that the "science" is largely "cargo cult" imitation of science.
     
    • Like Like x 4
    • Epic Epic x 1
    • List
    Last edited: Sep 17, 2021
  5. Thad E Ginathom

    Thad E Ginathom Friend

    Pyrate
    Joined:
    Sep 27, 2015
    Likes Received:
    14,134
    Trophy Points:
    113
    Location:
    India
    Fascinating, and thank you. I had heard of cargo cults before, but never explained like that. So all I got was that, for them, big things in the sky brought stuff that they liked and wanted. Now I understand it better, thanks to you and Feynman :)
     
  6. Jerseysam

    Jerseysam New

    Contributor
    Joined:
    Aug 9, 2021
    Likes Received:
    97
    Trophy Points:
    13
    Location:
    Cincinnati, OH
    As a newbie to hi-fi (all things relative)...and this may just be my sheer stupids....does SINAD strictly focus on (or is a measurement approach of) distortion, and if so are folks that hold that as the "key metric" saying that tone, timbre, soundstage, etc. don't exist? That's oft times the vibe I get...."here's how 'X' does with distortion....story over!". There's a whole thought track that would fairly ask "how do human ears really experience SINAD, etc." (that I don't think is ever really explained well).....but beyond that....tone, decay, transients, dynamics, soundstage....those are real things. They are for sure 'subjective' in terms of enjoyment or metrics.....but they exist. And in audio....I kinda', sorta', maybe would think those qualities are big deals.

    Loose analogy, but measuring how an ovens' actual internal temperature matches its setting gauge is valuable when you looking at buying an oven. "Hey, X runs 20 degrees off....Y is always right on the money." Dig it...valuable. But the size of the oven, electric or gas, in-cabinet or under-stove, fit a turkey or ten.....people look at those other things as common sense factors in a wider decision. But....and again, maybe I'm just missing context....a non-trivial amount of conversation on the audio internets seems to stop at "My ZhangBhang BeautyGloryWaifu Amp V3.2 that I got for $100 is great SINAD, end of story....better than everything else". That's odd to me. And not to be a jerk.....it seems maybe a message centered around folks whose budget is more under $1K, (which is most normal people everywhere) and sending the message "hey, that new ZhangBhang is really the best....you're so smart for not spending more to have the best....proven by NUMBERS!". And then I think to myself...."Man, if I had an interest in selling ZhangBhang boxes....bending the sub $1K conversation to my chosen field of battle wouldn't be a bad gig...".
     
  7. Woland

    Woland Friend

    Pyrate
    Joined:
    Jan 13, 2021
    Likes Received:
    1,311
    Trophy Points:
    93
    Location:
    a friendly land
    Gotta be careful on this one. The locals had a good insight and they wanted the planes to return. It was just their lack of familiarity of the underlying mechanism, their inability to tap theory, that led them astray. Runways don't attract planes and cargo. Yet to the participants or a casual observer, what they were doing looked right.

    So too in ASRs case, they have graphs that look sciency. But, as this thread identifies, AmirNADS is simply not a particularly useful thing to graph. And there is massive inconstency across ASR analyses - reflecting a dogmatic lack of scientific curiosity and understanding.

    Edit: changed SINAD to AmirNADS
     
    Last edited: Sep 16, 2021
  8. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    First of all, there is no such thing as a specific measurement called SINAD. We cannot go around and say that X amp has 123db SINAD or Y DAC has 137 SINAD. That is utterly meaningless. SINAD is not anywhere near the equivalent of well understood measurements such as MPG, max speed, 0-60, 1/4 times, FPS, etc.

    SINAD is the inverse of THD+N. That is total harmonic distortion and noise.

    The next question is THDN of what? For THD+N to be measured, there must be stimulus and measurement parameters. The stimulus in this case is a 1kHz steady state sine wave. The parameters, as defined by ASR are a 300-ohm resistor load. A 2Vrms output (for balanced amps 4Vrms output). And a bandwidth of 20kHz. These parameters are arbitrarily chosen by Amir. A 300-ohm load at 1kHz is typical for Sennheiser HD6** headphones. The 2Vrms (and 4Vrms) output makes no sense because 0.775Vrms gets us 100db on the HD6** headphones.

    Amir is basically measuring the amount of extraneous signal (bandwidth limited to 20kHz) from a 1kHz steady state signal at a level that will eventually smoke most headphones in about 5-10 minutes.

    I propose to not say SINAD when referencing the graphs and first panel of any ASR review. I recommend that we use AmirNADS. That is Amir's Noise and Distortion Standard.
     
    Last edited: Sep 16, 2021
  9. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    Oh, if anyone is curious to the test signal that is used for AmirNADS (and most others including distortion and power)
     
  10. Pancakes

    Pancakes Friend

    Pyrate Contributor
    Joined:
    Aug 13, 2020
    Likes Received:
    1,417
    Trophy Points:
    93
    Location:
    Atl
    1khz tone is one of my faves. I try to listen for at least half hour per day. Not as emotionally satisfying as pink noise though.
     
  11. batriq

    batriq Probably has made you smarter

    Pyrate
    Joined:
    Oct 31, 2015
    Likes Received:
    1,041
    Trophy Points:
    93
    Since the time I worked on wavelet transform for function approximation, I always wondered if they can be used for analyzing audio signals. Of course it’s easy to decompose an audio signal through a wavelet transform but no one has correlated that with what we hear or proprieties of what we like and dislike about an amp or dac. For example plankton is represented by the detailed basis vectors (leaves of the wavelet tree) but these are also the vectors that represent noise. In fact you can smooth a signal by zeroing these out. Anyway, if anyone can correlate a wavelet transformation of an audio signal with good properties of an amp, it’s @purr1n
     
  12. rlow

    rlow A happy woofer

    Pyrate Contributor
    Joined:
    Jul 18, 2017
    Likes Received:
    7,738
    Trophy Points:
    113
    Location:
    Canada
    Wow, that tone brings me back to the days of waking up in the middle of the night with the TV on and a test pattern showing (you know, when stations weren’t on 24/7).

    Ah those were the days.
     
  13. Metro

    Metro Friend

    Pyrate
    Joined:
    Dec 27, 2016
    Likes Received:
    1,590
    Trophy Points:
    93
    Location:
    San Francisco
    "Sinus Tone"? Are you supposed to play it into your nose?
     
  14. Biodegraded

    Biodegraded Friend

    Pyrate Contributor
    Joined:
    May 28, 2017
    Likes Received:
    7,981
    Trophy Points:
    113
    Location:
    Vancouver BC
    [​IMG]
     
  15. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    One more time ...

    Our infamous internut guru has a fondness for testing headphone amps at 4 Vrms (because it is an easy drop down selection on the AP555.) Many headphones have a typical sensitivity of between 95 and 100 dB/mW. Which implies a test voltage of 4 Vrms would produce for many headphones in excess of 110 dB SPL. Now also consider that the SINAD measurement only demonstrates performance at a specific frequency, typically 1 KHz. Now observe the following graph:
    Listener operating area annotated - cropped.png
    The SINAD chart so infamously used indicates performance for the one red dot labeled in the graph above. Outside the reasonable listener operating area. Furthermore it only provides indication of performance at that singular point. What about all the other points inside the actual listener operating area? Woefully inadequate representation by many orders of magnitude. Again I reiterate that using 1 KHz SINAD at 4Vrms (or 2Vrms) as a discriminant for amplifiers is an extraordinarily oversimplification befitting only of audio science simpletons.
     
    • Like Like x 12
    • Epic Epic x 10
    • List
  16. Woland

    Woland Friend

    Pyrate
    Joined:
    Jan 13, 2021
    Likes Received:
    1,311
    Trophy Points:
    93
    Location:
    a friendly land
    Last edited: Sep 17, 2021
  17. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    you might find subtle TLAs interesting.
     
  18. purr1n

    purr1n Desire for betterer is endless.

    Staff Member Pyrate BWC
    Joined:
    Sep 24, 2015
    Likes Received:
    89,777
    Trophy Points:
    113
    Location:
    Padre Island CC TX
    The main reason he did it is because the APx555 maintains a greater advantage over its competitors at 2Vrms and 4vrms than at lower levels. See how the competitive THD+N advantage of the APx555 lessens significantly at the more realistic Vrms levels that we use (-10 and 0dBu in orange -vs- purple 2Vrms and 4Vrms)

    upload_2021-9-18_13-6-42.png

    Most non-technical people don't realize or truly understand the kinds of games that ASR plays. Measuring at the higher voltage levels, although completely unrealistic, assures that ASR retains supremacy in the results / score oriented race.
     
    Last edited: Sep 18, 2021
  19. Woland

    Woland Friend

    Pyrate
    Joined:
    Jan 13, 2021
    Likes Received:
    1,311
    Trophy Points:
    93
    Location:
    a friendly land
    Cute, but we're in a world where many are dying because they listen to other science fakers -- leading them to injest poisons and stay unvaccinated during a pandemic.

    If you really want that three letter acronym, "Amir + Slimy Sychophants" is a more accurate characterisation of the subject matter.
     
    Last edited: Sep 18, 2021
  20. atomicbob

    atomicbob dScope Yoda

    Pyrate BWC MZR
    Joined:
    Sep 27, 2015
    Likes Received:
    18,653
    Trophy Points:
    113
    Location:
    On planet
    Looking at reasonable listener use case levels for dynamic headphones with 97 dB/mW typical sensitivites, measurements performed with stimulus of 25 mVrms to 500 mVrms are more sensible:
    HD650 sensitivity example dBV.png

    Taking data points from the APx555 graph above and overlaying with dScope results produced the following graph:
    THD+N vs amplitude dScope - APx555 comparison - cropped.png
    It would appear at actual listener levels the dScope is able to obtain lower distortion measurements. Above 500mVrms the APx555 has the advantage for this one specific type of measurement.

    Screenshots of dScope loopback THD+N data used to create the graph above from 25 to 500 mVrms:
    20210918 dScope ext loopback 1KHz THD+N 25mVrms.png
    20210918 dScope ext loopback 1KHz THD+N 50mVrms.png
    20210918 dScope ext loopback 1KHz THD+N 100mVrms.png
    20210918 dScope ext loopback 1KHz THD+N 200mVrms.png
    20210918 dScope ext loopback 1KHz THD+N 300mVrms.png
    20210918 dScope ext loopback 1KHz THD+N 500mVrms.png
     
    • Like Like x 11
    • Epic Epic x 1
    • List

Share This Page