advertisement


Can you tell the difference between 96/24 and 44/16 and mp3?

One could argue that it doesn't matter. :)

The "ringing" that some people make such an issue out of only occurs with artificial test signals that violate the Nyquist criteria, and is at the filter cutoff frequency, so inaudible anyway.

The first part of this is a total misunderstanding on your part - there is no such thing as an artificial test signal in the digital domain that doesn't meet the Nyquist criteria. By definition, *any* series of sample values are bandlimited, the bandlimiting is the thing that has occurred to get the signal into the digital domain!

So, the shape of an impulse response (the sinc function you see in the above examples) is *exactly* and mathematically correct, and is a real input signal (one with an equal amount of every frequency up to but not including nyquist), and is the fundamental building block of all DAC functionality. It's easiest to think of a DAC is producing sinc functions scaled and shifted in time for each input sample (scaled so that they peak at the amplitude of the input impulse). Weirdly, and maybe counter-intuitively, the sinc functions cancel out with each other, so that what remains from a series of smoothly varying samples does not include any obvious visual ringing.

The second part - it's at Nyquist. Well, yes, you aren't going to hear the frequency itself assuming you are sampling high enough, but the modulation of the amplitude gives you the energy across the frequency spectrum, which is the bit you hear - it's almost like the amplitude change is the bit you hear, not the frequency itself.

Possibly a better way of saying it would be, if you can't hear the nyquist frequency, you could filter it out and it would make no difference, but that would just give you silence :)
 
The first part of this is a total misunderstanding on your part - there is no such thing as an artificial test signal in the digital domain that doesn't meet the Nyquist criteria. By definition, *any* series of sample values are bandlimited, the bandlimiting is the thing that has occurred to get the signal into the digital domain!

I have to disagree. A step response is an artificial signal that violates the Nyquist criteria.
 
The problem occurs earlier in the chain, when down-sampling to 44kHz (e.g. your typical over-sampling ADC captures HF like cymbals, or maybe software down-samples electronic sounds like square[ish] waves from a much higher sampling rate). Namely, the anti-aliasing filter is generally a very steep one somewhere about 20-22kHz, causing ringing in the presence of HF input in this band.

The DAC can only reproduce the ringing already baked into its 44kHz input samples about 20-22kHz. Using DAC filters you then can shift the ringing to be more pre or post, or to allow imaging (which is plain wrong to my mind).
 
Last edited:
I have to disagree. A step response is an artificial signal that violates the Nyquist criteria.
Not as I understand the point to which you are responding.

A step in a digital audio file, reconstructed through the Shannon-Whittaker interpolation formula, produces a band-limited analogue audio step that "rings" [1] at the step. The corollary is that an analogue audio step when band-limited, sampled and quantized can produce a step in the digital audio file as long as the sampling points correspond to the step-height crossings of the band-limited ringing (unlikely in practice but theoretically possible).

The same argument applies in the case of an impulse in the digital audio file and its both-ways correspondence with the band-limited sinc function in the analogue audio.

The only case I can see for a violation of the Nyquist criterion lies in the strict requirement that sampling frequency/2 [Fs/2] > maximum signal frequency content [Fmax]. That is, any frequency content exactly at Fs/2 gets aliased to Fs/2, so content at exactly Fs/2 is not allowed and would violate the Nyquist criterion.

However, I think that for a finite length digital audio file with finite quantization I can always choose Fs/2 = Fmax + epsilon for a value of epsilon that is as small as I like but non-zero. So I think that any sequence of sample values in a real-world audio file always reconstructs to meet the strict inequality of the Nyquist criterion.

Even if this is wrong (and I would be happy to be convinced it's wrong), aliasing from Fs/2 to Fs/2 is not a practical problem. So I always have difficulty understanding the practical significance of assertions I see at times that illegal sequences of samples are possible in a digital audio file.

[1] I don't like the word "ringing" in this context - it isn't ringing in the sense I normally use - it's just a resemblance.
 
The only case I can see for a violation of the Nyquist criterion lies in the strict requirement that sampling frequency/2 [Fs/2] > maximum signal frequency content [Fmax]. That is, any frequency content exactly at Fs/2 gets aliased to Fs/2, so content at exactly Fs/2 is not allowed and would violate the Nyquist criterion.

Indeed.

However, I think that for a finite length digital audio file with finite quantization I can always choose Fs/2 = Fmax + epsilon for a value of epsilon that is as small as I like but non-zero. So I think that any sequence of sample values in a real-world audio file always reconstructs to meet the strict inequality of the Nyquist criterion.

That is not what I was taught - but it is way too early in the day for me to try to do the math to verify it.

I don't like the word "ringing" in this context - it isn't ringing in the sense I normally use - it's just a resemblance.

I agree. The term "ringing" is misleading.
 
Unlike Santa, the improved accuracy of hi-res agrees with theory and it's measurable, and the question is just about perceivability.

As far as I know, there's no hard evidence it's perceivable, but it isn't inconceivable.
I have some published papers in my files that suggest the temporal resolution of the human hearing system - detectable differences in time of arrival - are several times smaller than would be expected from the ear's frequency response. There are two apparently independent sources (neither is Oohashi, BTW) but it's a field so far away from my expertise that I can't evaluate the quality (or otherwise) of the research or be confident in my interpretation.

So it may be possible that the band-limiting of audio to the conventional 20 kHz frequency limit has some perceptual consequence other than from the frequency range.
 
I have some published papers in my files that suggest the temporal resolution of the human hearing system - detectable differences in time of arrival - are several times smaller than would be expected from the ear's frequency response. There are two apparently independent sources (neither is Oohashi, BTW) but it's a field so far away from my expertise that I can't evaluate the quality (or otherwise) of the research or be confident in my interpretation.

Hopefully avoiding the mistakes from the Oohashi study.
 
I don't like the word "ringing" in this context - it isn't ringing in the sense I normally use - it's just a resemblance.

Right. The problem with the word "ringing" is that it suggests some extra signal on top of the original waveform, when, in reality, it is actually an absence of something - an absence of higher-order harmonics (above the filter cut-off frequency). This is not limited to digital signals, but something that happens with any bandwidth limiting (low-pass filtering) - in our amps, speakers and even ears. This animation illustrates the effect quite well:

https://en.wikipedia.org/wiki/Gibbs_phenomenon#/media/File:SquareWave.gif
 
I have to disagree. A step response is an artificial signal that violates the Nyquist criteria.

Well, you can disagree if you like, but you really need to do some more reading as you're misleading yourself. Honestly, I know what i'm talking about, I do this for a living, and it's not open to debate or interpretation, it's just a fact about what the Sampling Theorem means.

I'll repeat, the digital signal does not contain frequencies above nyquist, it's a requirement of sampling in the first place :(
 
I'll repeat, the digital signal does not contain frequencies above nyquist, it's a requirement of sampling in the first place :(

A *sampled* signal does not contain frequencies above Nyquist, but, as I wrote, a step response is not a signal we have sampled from an input - it is an artificial signal created in the digital domain.
 
Good Afternoon,

I've gone through the files supplied by Mark Waldrup and selected A or B and reported my findings accordingly.

I have no idea how successful I was - I wouldn't describe any of them as being a night over day difference.

Regards

Richard
 
comparing some rips yes I can hear differences , in the early days of hi-res I purchased some, now rarely do so. the most cost effective method is buying very cheap cds from any source and ripping them enjoy more vinyl stock than digital .first new record was elvis first release in 1961 .
 
From Mark’s latest email.

So, if you haven't yet asked to participate, downloaded the files, done some comparative listening, and submitted your selections, PLEASE take this opportunity to sign up and submit your results. I would like to especially encourage anyone under the age of 40 to take the challenge. It seems most of those that have submitted the final form are somewhat older. It would be great to get some younger ears involved.

So those of you lucky bastards with bat like under 40s ears, now is your chance to shine. Or if you’re one of “the rest” you might just add to the statistical robustness of the study by adding to it.

.sjb
 
From Mark’s latest email.

So, if you haven't yet asked to participate, downloaded the files, done some comparative listening, and submitted your selections, PLEASE take this opportunity to sign up and submit your results. I would like to especially encourage anyone under the age of 40 to take the challenge. It seems most of those that have submitted the final form are somewhat older. It would be great to get some younger ears involved.

So those of you lucky bastards with bat like under 40s ears, now is your chance to shine. Or if you’re one of “the rest” you might just add to the statistical robustness of the study by adding to it.

.sjb

I take it, as I've not heard back from Mark, that you don't get a 'personal' result back then?

Regards

Richard
 
Good Afternoon All,

I see an EM popped up in my Inbox this morning from Mark Weldrup. Although not directly aimed at myself, as one of the apparently 350 people who have tried his challenge I still await my specific result, but I see his is stating that there was, on average, only a 40% success rate in determining which track was which.

It is disappointing that so few people bothered even attempting the challenge in the first place but the results are hardly a resounding endorsement for so called Hi-Res audio.

Regards

Richard
 


advertisement


Back
Top