Serious Question: How can DAC's have a SOUND SIGNATURE if they measure as transparent? Are that many confused?

[pkane]

This is only true for a large number of averaged samples with a stationary signal.

In effect it is a side-effect of FFT Math and using steady state signals.

In the real world, instantaneously, the actual signal remains hidden and unrecoverable.

In case of a "noise like" signals (e.g. cymbals) or complex melodies there is nothing to average as the signal itself approaches being technically (but not for listening humans) indistinguishable from noise.

Thor

This has nothing to do with FFT. It has to do with correlated quantization distortion vs uncorrelated noise. Our ears don't hear individual samples, so the effect on an individual sample is irrelevant.

 

[Thorsten Loesch]

This has nothing to do with FFT. It has to do with correlated quantization distortion vs uncorrelated noise. Our ears don't hear individual samples, so the effect on an individual sample is irrelevant.

I did not mention "samples".

I mentioned "noise like" (or shall we say semi-stochastic) signals, or shall we say "wavelets".

We had cymbals posted here before. take the tail end before the cymbal fades into the noise floor...

Or use steady state noise like Microphone self noise (which audibly changes tonality with some forms and levels of dither).

Dither has it's own effects, including when using haped (violet/ultraviolet) noise.

I have no problem with dither. It has uses. It is a tool. Use it correctly.

Like all subtractive or additive processes it will alter the signal and reduce objective fidelity and may also impact subjective fidelity in a positive or negative way.

Thor

 

[solderdude]

So you object to presenting fact because these facts may actually make people think that they are facts?

I am not objecting to presenting facts.

Just a word of 'caution' on how 'the general public' is likely to misinterpret the words:

Dither would replace the distorted sinewave with noise

eventhough technically the choice of words is correct is might give the impression that DACs either distort or have noise.

 

[Thorsten Loesch]

I am not objecting to presenting facts.
Just a word of 'caution' on how 'the general public' is likely to misinterpret the words:

Well, "the generl public" has zero interest in audio science.

eventhough technically the choice of words is correct is might give the impression that DACs either distort or have noise.

Shock! Horror!

In my experience ALL DAC's DISTORT AND HAVE NOISE!

Now, the real issue how much distortion and noise and what spectrums and how audible these may be. That is the science.

Thor

 

[solderdude]

This reply is exactly what I meant.
Sorry if you cannot see this.

 

[pma]

Stop feeding Troll, please.

 

[fpitas]

Stop feeding Troll, please.

I'd give this post 10 likes if I could.

 

[IAtaman]

Linear PCM has same size steps for all levels. The level change at the LSB for all LPCM systems is 6dB if only the LSB toggles. The level change for 16 Bit with full scale signal and LSB toggeting is 0.00013dB.

Thank you for the answer, appreciate it. I had to read it a few times to understand enough to be able to ask follow up questions. What I get from you said is that due to linear distribution of levels across the full scale, 16bit allocates very very low resolution to low signal levels and 18bit solves this problem to a degree.

If a linear Amplifier reproduced a sinewave at 90dB below rated power the way shown above, we would consider it broken by design.

This is the part I don't get. Lets take Topping A30Pro for example. It has 120db SNR so that's great. But at normal listening levels with a relatively sensitive headphone, THD+N is already at 85db according to Amir's measurements, so why am I worried about the distortion of -90db signals?

 

[solderdude]

16bit allocates very very low resolution to low signal levels and 18bit solves this problem to a degree.

Yes, and 20 bit is better still so is 24 bit even 'finer' resolution but in practice (with 2V signal levels) those will be buried in noise.
As noise is random it will be shown 'low' in amplitude in FFT with large bins and will show very small signals peaking out above a 'seemingly low' noise floor.
Yet, in real time, as we listen, it will be buried in noise.

Now, in the good old days some morse folks could distinguish 'beeps' buried below a noise floor. In fact normal folks can also hear a constant tone buried in the noise floor, just not as good as a trained person.
The problem is music is not a constant tone. It is dynamic and varies in frequency.
So real world noise floor differs from measured noise floor (a single number) because the frequency spectrum is also important. High frequency noise is more annoying than low frequency noise for instance (vinyl as an example).

So resolution doesn't say everything, recording levels matter as well.

16 bits can be enough for reproduction but is seldom enough for recording where headroom for peaks is needed (one cannot say how high peaks are going to be).
For recording 24 bits makes sense.
When using digital volume control and amps/speakers have to blast at high levels noise floor and resolution starts to matter.

So whether or not a 16 bit resolution is enough in practice depends on what it is used for.

 

[Thorsten Loesch]

Thank you for the answer, appreciate it. I had to read it a few times to understand enough to be able to ask follow up questions. What I get from you said is that due to linear distribution of levels across the full scale, 16bit allocates very very low resolution to low signal levels and 18bit solves this problem to a degree.

Yes and no. Tape noise and Microphone noise act as natural and inherent dither, so if we do a straight analogue to 16 Bit recording with optimum levels, the problem is probably banished to sufficiently low levels to be inaudible.

BUT, you have little margin for error.

Going 18 Bit makes sure we are away from digital clipping (headroom - I believe Bill C called the oval office that) and have some distance to the lower levels.

This is the part I don't get. Lets take Topping A30Pro for example. It has 120db SNR so that's great. But at normal listening levels with a relatively sensitive headphone, THD+N is already at 85db according to Amir's measurements, so why am I worried about the distortion of -90db signals?

This appears to be a Headphone Amplifier, not an "Amplifier" (which commonly means "Speaker Amplifier").I could look up Amir's measurements, but that is meaningless. We are back to my fundamental point that measurements create data, not information. Information requires context.

With headphones the issue is that on one side we have an AKG K1000 with 80dB/1V sensitivity and on the other side something like the Campfire Audio Andromeda with 127dB/1V sensitivity.

In order to get "0dB" noise on the Andromeda's we need 127dB SNR @ 1V plus 0.1V output for 107dB SPL (closest for easy calculation to the THX target of 105dB).

For the AKG K1000 we need to have 26V Output with low distortion and > 80dB SNR @ 1V.

All other headphones fall in-between this huge range of 48dB. So it is not particularly reasonable to state any specific operation level. And it is equally not particularly sensible to make an amplifier that has 127dB SNR @ 1V and + 26dBV output.

If you buy a ****** amplifier, you will have excess noise. So don't buy one like this.

Make sure to get a design that has the required performance to match your headphones.

The same goes BTW for speakers, but less so. My own speakers at the Den are 98dB/2.83V/1m which is 11dB more sensitive than the "average" speaker of 87dB/2.83V and 16dB more sensitive the BBC LS3/5 at 82dB/2.83V.

So an amplifier that is silent on a BBC LS3/5 may not be silent in my system. An Amplifier that delivers 105dB peak @ 3m for a stereo pair on my own speakers (only needs 10W/8R) may not be able to produce acceptable sound levels with a pair of LS3/5. My own amplifier delivers < 0.004% THD @ 10W (9V) and 0.16mV noise (A-Weighted) giving 92dB SNR @ 105dB peak SPL.

Why are we "worried" about distortion at -90dB? Depends on the music you are playing how it is recorded and the system context. In both my speaker systems and my HP system, I think so.

You? Up to you.

Thor

 

[IAtaman]

Yes and no. Tape noise and Microphone noise act as natural and inherent dither, so if we do a straight analogue to 16 Bit recording with optimum levels, the problem is probably banished to sufficiently low levels to be inaudible.

BUT, you have little margin for error.

Going 18 Bit makes sure we are away from digital clipping (headroom - I believe Bill C called the oval office that) and have some distance to the lower levels.

Once we get past the article-gate, and you started explaining instead of arguing, it was quite informative speaking to you. Thank you. And sounds like there was not a very big disagreement after all!

 

[krabapple]

Prof Schneider always held that bit depth is the important thing.
The red book idea was always flawed, because 44.1khz 16 bit (CD), could be proven to truncate the reverb.

Only if done incorrectly.

This was a solved problem by 1990.

See Lipshitz and Vanderkooy in JAES. Prof Schneider should have kept up with the literature.

 

[fpitas]

This thread started with a troll title, and then went downhill

 

[krabapple]

https://www.aes.org/e-lib/browse.cfm?elib=18296

"The overall conclusion is that the perceived fidelity of an audio recording and playback chain can be affected by operating beyond conventional resolution."

Thor

You are once again very late to the party . The party being the debate about this paper when it came out. Pick and choose your papers a certain way, and sure, you can conclude this. I daresay no one but the true believers and Bob Stuart find it dispositive at this point.

But even the 'meta' finding, if one find it convincing, is pretty far from the 'even my wife hears it ' or 'the French laugh at those who cannot hear 16 vs 24 bit' , no? You know, the sort of blowhard claims audiophiles make all the time. So Why did it take a meta-analysis to make this 'discovery'? The mountain labored mightily and brought forth a mouse.

 

[krabapple]

Correct. If programmers wrote efficient code 640k would still be enough.

16/44 was never good enough.

Thor

It was the very definition of 'good enough'. Rather than 'a best fit for all purposes'.

 

[SIY]

Only if done incorrectly.

This was a solved problem by 1990.

See Lipshitz and Vanderkooy in JAES. Prof Schneider should have kept up with the literature.

Actually, long before that. Either Herr Professor was being misquoted or he is a crank.

 

[krabapple]

Ok, let's have a few more, shall we? I had a problem finding the right bit's on my phone...

It was amusing to see how everyone kept jumping up and down to claim what I posted did not matter, when in fact this research alone is sufficient to put that baby to be permanently (just as a Meta Analysis of the negative ABX results published by ABX Corporation invalidates ABX resoundingly).

Sampling Rate Discrimination: 44.1 KHz Vs. 88.2 KHz

"Overall, participants were able to discriminate between files recorded at 88.2kHz and their 44.1kHz down-sampled version. Furthermore, for the orchestral excerpt, they were able to discriminate between files recorded at 88.2kHz and files recorded at 44.1kHz."

Effect of High-Resolution Audio Music Box Sound on EEG

"α2 and β1 wave band power significantly increased with high-resolution audio music box sound. α wave band power increased with high-resolution audio music box sound.Conclusions: These results suspect that high-resolution audio music box sound has the effect of increasing of vigilance and relaxation of the brain."

Evaluation of Sound Quality of High Resolution Audio

"In addition, perceptual discrimination in quantization distortion is also focused on in between 16 bits and 24 bits resolution. As the result of the listening test with 28 participants, the difference of quantization resolution could be discriminated with the accuracy of 60.3 %."

High-Resolution Audio with Inaudible High-Frequency Components Induces a Relaxed Attentional State without Conscious Awareness

"The present study shows that high-resolution audio that retains high-frequency components has an advantage over similar and indistinguishable digital sound sources in which such components are artificially cut off, suggesting that high-resolution audio with inaudible high-frequency components induces a relaxed attentional state without conscious awareness."

Now as Scientist in behavioural Psychology I will go get popcorn and watch all the Cargo Cult Scientists try make the papers I posted now go away and somehow claim they don't matter. And I am thinking about the thought process in the minds of Cargo Cult Scientists. I suspect it matches more of less that famous scene in Dirty Harry...

Now do I have more papers left?

Thor

Ah, all the oldies but goodies trotted out again. What a clown show. And now you are a "Scientist in behavioural Psychology"[sic]? Uh huh.

So, Dr., how come you didn't post links to any of the papers that *did not* find evidence for audibility of hi rez? If you were intending a literature review, I mean.

 

[krabapple]

The question HAS BEEN answered unambiguously long before now.

The point is that many vested interests are not in favour of the conclusion. Any fact that is disliked is buried. Look at "Climate Science". The fact that much of the "research" that set off the current "Warmist" stream in climate science was fake is being studiously ignored and it is still being presented and used to promote warmism. This has all the hallmarks of Cargo Cult Science.

And now, you're going beyond mere sophistry into just plain lying. I hold you in contempt.

Actually, long before that. Either Herr Professor was being misquoted or he is a crank.

Yes, it was the earlier 80s, but I wanted to give Prof. Schneider the benefit of doubt.

 

[pkane]

Like all subtractive or additive processes it will alter the signal and reduce objective fidelity and may also impact subjective fidelity in a positive or negative way.

Only the first part is right. It will alter the signal, but when used properly, it'll improve fidelity. This is not subjective, this is mathematics.

 

[Roland]

This thread is a fascinating exposition of bigotry and bear baiting. Measuring hi-fi is terribly first world, so why is ASR so toxic?