Why Don't High SINAD Receivers Exist?

[cistercian]

Yeah that came out wrong. Im just fustrated that SINAD is being used to judge equipment without any reference to audibility levels of distortion.

SINAD is something you can train your brain to handle. I am a hardcore DXer which means for decades I listened to noisy
radio signals and tried to extract the voice. A friend, now passed, was also a DXer and he and I would listen for a persons
voice at a group gathering and listen to them as they moved around. We would occasionally shout something at them from
across the room to show we could hear them and it drove them nuts. It was hilarious!
That being said, the higher the signal to noise ratio the more pleasant and relaxing the experience. Trying to tunnel through
the noise is tiring. People adapt quickly to stimulus and adapt to it. I mention my experience as an extreme case to highlight
that SINAD will have a subjective component. The gear I used as a young man had an audible noise floor...but I got used to
it and found it fine. Unless you are listening critically at a very low level much of what we have now should be fine and I strongly
suspect the bean counters limit mass market stuff for the benefit of profit and delivering something "good enough".

Testing like Amir does exposes lazy practices so one would hope the makers can improve the specs. But...the point that is adequate
for most people may be lower spec than some here would want. I have come to accept most mass produced consumer stuff is
not very good...but don't tell that to the people who own it and love it.

 

[Aerith Gainsborough]

I don't understand why SINAD measured here for AV receivers is done only in direct mode when all digital processing is turned off, which does not represent practical listening modes. I would like to see AVR tested while room correction is fully active. What would be the SINAD difference for the Denon X6700H with Audyssey off vs on?

While I agree with your sentiment, that it should be tested because it represents the practical use case, I would not expect there to be much of a difference.

For most inputs, the signal is already in the digital domain when arriving at the AVR, so altering it should not degrade SINAD because these alterations would occur before it hits the DAC stage in the first place, no?

I would expect a degradation for any analog signal, because it has to be digitized, altered and then must be piped through the DAC instead of being fed directly to the amp.

 

[lashto]

I was unaware that Amir was personally responsible for all psychoacoustic research on Earth.

Man, he's really been dragging his feet, huh?

I seriously doubt that SINAD or it's THD+N brother will ever be (properly) correlated to audibility. These are engineering metrics and do an excellent job at showing engineering quality. That is what ASR is (mostly) about.

I would like a lot more psychoacoustics around here but in those terms SINAD is just an awful basket of (very different) distortions. Or to put it in more general terms: using EE-metrics to assert audibility is ~ like using a ruler to assert a painting's expressivity

IMO, a totally different metric/number is needed to assert audibility. Many are trying but the progress is quite slow and we still do not have a really good metric (AFAIK).

 

[maverickronin]

I seriously doubt that SINAD or it's THD+N brother will ever be (properly) correlated to audibility. These are engineering metrics and do an excellent job at showing engineering quality. That is what ASR is (mostly) about.

I would like a lot more psychoacoustics around here but in those terms SINAD is just an awful basket of (very different) distortions. Or to put it in more general terms: using EE-metrics to assert audibility is ~ like using a ruler to assert a painting's expressivity

Yeah. That's why Amir pretty much hands out Panthers for DACs based on having SINAD so high a human can't hear it no matter its composition.

Some might say that's too high of a standard, but when small volume manufacturers can do it with 2 channels for ~$100 I don't see why we shouldn't expect something within a couple steps of that from an AVR instead of the current state of affairs where ~90% of them seem to lag behind first generation CD players from nearly 40 years ago...

IMO a totally different metric/number is needed to assert audibility. Many are trying but the progress is quite slow and we still do not have a really good metric (AFAIK).

I think that kind of research would certainly be useful, but the kind of time you'd need to invest into to get kind of widely accepted results would be enormous. Even then I think the results would be a lot more useful in designing speakers than electronics. Conducting tests with actual people just takes too much time and resources. A small team in a lab can design completely transparent electronics far faster than they can conduct tests to figure out what types and levels of distortion and noise are acceptable.

It would be good for the industry/hobby as a whole to have such information, but nobody's willing to spend the money on it. Given that the industry is far more marketing than science, doesn't have any kind of trade organization which conducts research for its member companies, and can't make a good case for government grants, I don't see that changing anytime soon.

The public Harman research is fortunate anomaly. It's an amazing start, but no one's ever attempted to replicate it or follow up on it and I don't think it will ever be done. From upper management's point of view it might have even just been a way to get back at Consumer Reports.

I don't think much progress is going to made in the psychoacoustics of this anytime soon. I mean, nobody's even done conclusive research on hi rez yet...

 

[lashto]

Yeah. That's why Amir pretty much hands out Panthers for DACs based on having SINAD so high a human can't hear it no matter its composition.

Some might say that's too high of a standard, but when small volume manufacturers can do it with 2 channels for ~$100 I don't see why we shouldn't expect something within a couple steps of that from an AVR instead of the current state of affairs where ~90% of them seem to lag behind first generation CD players from nearly 40 years ago...



I think that kind of research would certainly be useful, but the kind of time you'd need to invest into to get kind of widely accepted results would be enormous. Even then I think the results would be a lot more useful in designing speakers than electronics. Conducting tests with actual people just takes too much time and resources. A small team in a lab can design completely transparent electronics far faster than they can conduct tests to figure out what types and levels of distortion and noise are acceptable.

It would be good for the industry/hobby as a whole to have such information, but nobody's willing to spend the money on it. Given that the industry is far more marketing than science, doesn't have any kind of trade organization which conducts research for its member companies, and can't make a good case for government grants, I don't see that changing anytime soon.

The public Harman research is fortunate anomaly. It's an amazing start, but no one's ever attempted to replicate it or follow up on it and I don't think it will ever be done. From upper management's point of view it might have even just been a way to get back at Consumer Reports.

I don't think much progress is going to made in the psychoacoustics of this anytime soon. I mean, nobody's even done conclusive research on hi rez yet...

well, "high fidelity" is still defined as <1% THD. By that fully deprecated EE definition (from ~1960 IIRC and it was never any good) almost any 2020 device should sound magnificent.

You brought up pretty much all the reasons why everything is so slow and incomplete in the PsychoAcoustics domain. Maybe this thread should be renamed "Why don't good sounding devices exist?" .

There is however some usable psycoacoustics research and I find it very annoying that the EE & PA branches do not work more together. To me those two seem to be made for each other but there is almost zero cooperation (or at least nothing visible/mainstream).

 

[3dbinCanada]

Ken Ishiwata legendary Marantz engineer:
Measurement or listening... which takes priority?
All measurements we do are static. Yes, they are essential to find out how amplifiers are working, but they don’t tell you about the sound quality. You have been in my listening room in Eindhoven. We’ve spent so many hours in that room listening and fine tuning all Marantz products. Sometimes that listening forces us to change our basic design, despite the fact that the measurements are okay.

That is nothing but a sales pitch to the gullible audiophile and has no technical content whatsoever.

 

[3dbinCanada]

Yeah. That's why Amir pretty much hands out Panthers for DACs based on having SINAD so high a human can't hear it no matter its composition.

Some might say that's too high of a standard, but when small volume manufacturers can do it with 2 channels for ~$100 I don't see why we shouldn't expect something within a couple steps of that from an AVR instead of the current state of affairs where ~90% of them seem to lag behind first generation CD players from nearly 40 years ago...



I think that kind of research would certainly be useful, but the kind of time you'd need to invest into to get kind of widely accepted results would be enormous. Even then I think the results would be a lot more useful in designing speakers than electronics. Conducting tests with actual people just takes too much time and resources. A small team in a lab can design completely transparent electronics far faster than they can conduct tests to figure out what types and levels of distortion and noise are acceptable.

It would be good for the industry/hobby as a whole to have such information, but nobody's willing to spend the money on it. Given that the industry is far more marketing than science, doesn't have any kind of trade organization which conducts research for its member companies, and can't make a good case for government grants, I don't see that changing anytime soon.

The public Harman research is fortunate anomaly. It's an amazing start, but no one's ever attempted to replicate it or follow up on it and I don't think it will ever be done. From upper management's point of view it might have even just been a way to get back at Consumer Reports.

I don't think much progress is going to made in the psychoacoustics of this anytime soon. I mean, nobody's even done conclusive research on hi rez yet...

Harmon's work started off with Dr Floyd Toole who was hired by Harmon to further his research from his work at Canada's renown National Research Council. His work is ground breaking but unfortunately, not enough research is being done in physcoacoustics. Toole has debunked a lot of audiophoolery. I have posted a question about speaker measurements and asked him if two speakers from different manufacturers measured identically in the same room, would they sound the same? His answer was yes. His measurements are far more extensive than Stereophile and Soundstage, the two best publications for posting speaker measurements.

I would also like to add that SINAD has nothing to do with build quality. Yamaha more than any other manufacturer of AVRs has the least returns and repairs in the industry. SINAD does NOT measure build quality.

 

[maverickronin]

I have posted a question about speaker measurements and asked him if two speakers from different manufacturers measured identically in the same room, would they sound the same? His answer was yes.

That's just a tautology for anyone who's not hopelessly lost in audiophile woo.

Of course, basically the only way two speakers from different manufacturers would end up measuring the same (within some margin of error...) is if they were clones of the same design so it's not all that useful to bring up either.

 

[lashto]

That's just a tautology for anyone who's not hopelessly lost in audiophile woo.

Of course, basically the only way two speakers from different manufacturers would end up measuring the same (within some margin of error...) is if they were clones of the same design so it's not all that useful to bring up either.

I'd go even further to say that it's impossible to even define "measuring the same": since we don't know much about the correlation between measurements and sound there is no meaningful "margin of error". And we are back into Toole's circle of audio bullshit (which IMO is his best contribution).

In absolute terms, no two devices measure exactly the same. There are even variations between same brand DACs and two speakers from the exact same production batch might measure very differently. Two speakers from diff manufacturers usually measure so far apart that they might be easily described as different 'species'.

As of today, "if it measures the same, it will sound the same" is just another EE lalala. Might sound great at first (and it's theoretically true) but it's neither defined/definable nor has any practical use. It's not even funny as some tautologies could be

 

[lashto]

While I agree with your sentiment, that it should be tested because it represents the practical use case, I would not expect there to be much of a difference.

For most inputs, the signal is already in the digital domain when arriving at the AVR, so altering it should not degrade SINAD because these alterations would occur before it hits the DAC stage in the first place, no?

I would expect a degradation for any analog signal, because it has to be digitized, altered and then must be piped through the DAC instead of being fed directly to the amp.

the digital theory sounds great but digital audio is not just (lossless) math. In practical terms many screw up that math to start with: either 'simple' math mistakes or they do it at 16-24bit instead of 64 because it's much cheaper, or....

Even if the math was perfect, a small degradation is still to be expected as no electronic circuit is 'perfect'. I bet the first AVR measurement with DSP/roomcorrection enabled would be a big 'surprise'. If those DSP circuits don't waste at least 5dB SINAD and/or alter most Hs, I might just delete all my posts & account

 

[EB1000]

While I agree with your sentiment, that it should be tested because it represents the practical use case, I would not expect there to be much of a difference.

For most inputs, the signal is already in the digital domain when arriving at the AVR, so altering it should not degrade SINAD because these alterations would occur before it hits the DAC stage in the first place, no?

I would expect a degradation for any analog signal, because it has to be digitized, altered and then must be piped through the DAC instead of being fed directly to the amp.

SINAD is sensitive to harmonic distortion and added noise, both of which should ideally be absent from any linear process like applying EQ. If such digital processing reduces SINAD by over 10db, then the implementation is faulty. I can clearly hear an increased hiss when YPAO is active in my Yamaha receiver when I turn the volume to the 0db and nothing is playing...

 

[Aerith Gainsborough]

If those DSP circuits don't waste at least 5dB SINAD and/or alter most Hs, I might just delete all my posts & account

Hmm my knowledge is probably lacking (after all I am not an EE) but I do not understand.

SINAD is the combined noise and harmonic level vs the level of original test signal, yes?
If so, why would an EQ introduce either: noise or different harmonics? All it alters should be the relative volumes of certain frequencies.
In the case of room correction I would expect not much change beyond 1KHz at all, because that's well past the Schroeder Frequency of most rooms.
Dirac would be an exception, since that software does a lot more to the signal than simple EQ. Allegedly.

 

[3dbinCanada]

I'd go even further to say that it's impossible to even define "measuring the same": since we don't know much about the correlation between measurements and sound there is no meaningful "margin of error". And we are back into Toole's circle of audio bullshit (which IMO is his best contribution).

Whats your degree again? Have you studied acoustics? No? What credentials do you possess that you can make an informed evaluation of Dr Toole's work? So this is you pulling yet another df opinion out of your ass?

In absolute terms, no two devices measure exactly the same. There are even variations between same brand DACs and two speakers from the exact same production batch might measure very differently. Two speakers from diff manufacturers usually measure so far apart that they might be easily described as different 'species'.

And you have SOOO much practical experience and theoretical knowledge when it comes to measurements...

As of today, "if it measures the same, it will sound the same" is just another EE lalala. Might sound great at first but it's neither defined/definable nor has any practical use. It's not even funny as some tautologies could be

What's your experience again? You dont have any? I see...

 

[lashto]

here's how a DSP board looks like. Lots of components/circuits on that board = lots of possibilities to waste SINAD . Even the smallest resistor/capacitor/etc will introduce some D because there is no perfect circuit/component. EQ/DSP software can be done ~lossless, hardware cannot.
One can build/use a good DSP board that only 'wastes' a few dBs of SINAD but those are not exactly cheap. And everything I read about those AVRs indicates that the main mantra is "build them cheap", not "build them well".

@EB1000 posted that "DSP/YPAO hiss" example above, it is probably the most often reported effect of enabling DSP.

 

[3dbinCanada]

@EB1000 posted that DSP hiss example, it is probably the most often reported effect of enabling DSP.

In correct yet again. I studied noise generation in semiconductors and that hiss you are hearing is MAINLY caused by the output stages of the amplifiers and is called generation-recombination noise and not DSP hiss. All semiconductors suffer from this effect.

 

[Aerith Gainsborough]

here's how a DSP board looks like. Lots of components/circuits on that board = lots of possibilities to waste SINAD .

Sure, if you need to use an analog signal, digitize, process and re analogize™ it, that makes sense.
But when the signal is digital (PCM) to begin with and remains digital until it reaches the same DAC of the AVR it would reach in pure direct mode, I don't see how SINAD could be lowered.

 

[3dbinCanada]

Sure, if you need to use an analog signal, digitize, process and re analogize™ it, that makes sense.
But when the signal is digital (PCM) to begin with and remains digital until it reaches the same DAC of the AVR it would reach in pure direct mode, I don't see how SINAD could be lowered.

.

I cannot speak for other AVR manufacturers but if Yamaha receives an analog signal at its line inputs or phono input and Pure Direct Mode is enabled, then the signal path remains analog and does not get digitized.

 

[lashto]

Sure, if you need to use an analog signal, digitize, process and re analogize™ it, that makes sense.
But when the signal is digital (PCM) to begin with and remains digital until it reaches the same DAC of the AVR it would reach in pure direct mode, I don't see how SINAD could be lowered.

Measurements will only tell you how much was wasted but not why/how. Only the manufacturer can clarify that. Good luck asking SoundUnited/Yamaha/NAD/etc. All that stuff abut the pure digital signal is just 'beautiful theory'. Why would AVRs provide an extra "pure direct" option if the DSP-enabled "auto" was same as "pure"?

If you think that ASR will ever do DSP-enabled measurements and wanna hold a bet about the amount of "DSP waste", you know where to find me . I'll maintain my (optimistic) stance of "at least 5dB" SINAD difference.

 

[3dbinCanada]

Measurements will only tell you how much was wasted but not why/how. Only the manufacturer can clarify that. Good luck asking SoundUnited/Yamaha/NAD/etc.
All that stuff abut the pure digital signal is just 'beautiful theory'. Why would AVRs provide an extra "pure direct" option if the DSP-enabled "auto" was same as good/pure?

If you think that ASR will ever do DSP-enabled measurements and wanna hold a bet about the amount of "DSP waste", you know where to find me . I'll maintain my (optimistic) stance of "at least 5dB" difference.

Here's your proof.. that Yamaha keeps analog in the analog domain while in Pure Direct Mode. Anything else you need to know about AVRs?

https://www.avsforum.com/threads/how-many-of-us-are-2-channel-only.2778617/post-59985582

 

[Aerith Gainsborough]

I cannot speak for other AVR manufacturers but if Yamaha receives an analog signal at its line inputs or phono input and Pure Direct Mode is enabled, then the signal path remains analog and does not get digitized.

If you listen in "pure direct", a digital PCM signal will go through the DAC stage but not the DSP stage, if you want to listen to a room corrected analog signal of e.g. a turntable, you are no longer listening via "PureDirect".
Keep in mind, I was talking about digital input only.

All that stuff abut the pure digital signal is just 'beautiful theory'. Why would AVRs provide an extra "pure direct" option if the DSP-enabled "auto" was same as good/pure?

Because Audio folks are paranoid and many detest any processing whatsoever.
Also: it does make sense to provide it when you work with analog paths and want to make it clear to the user that you can bypass the built in DSP if you choose to do so.

So yes, they will implement a "pure direct" button, whether it actually does result in a cleaner signal path for digital signals or not.