I’m not trolling you (not intentionally anyway). My point is that in theory Shannon and Nyquist might be right - and no one seems do doubt. They have never been able to proof in practice that it works for audio since technology wasn’t advanced enough to do so. If I look at measurements I do on DAC’s I never see coming back what I have put in, regardless brand, price and so on. And I have had over a hundred DAC’s on my AP. So, accepting the Shannon/Nyquist theorem, I can only conclude that we are still not able to build ADC’s and DAC’s well enough to apply the Shannon/Nyquist theorem in practice. Listening tests confirm this. What I did notice is that those manufacturers that go beyond using of the shelf DAC chips and program their own in FPGA’s, using far more taps, sound clearly better. Especially the transient response improves drastically. I also notice that, especially with those of the shelf DAC’s, having the computer upsample a redbook file to 2fs or 4fs, often does improve the sound quality. Some explain this as follows: a computer has far more computing power than the chip in the DAC that does the digital filtering and since the filtering for higher sampling rates doesn’t have to be so drastic, that chip can handle that filtering better. I can’t prove this statement but it more in line with practice than stating that, since there is a Shannon/Nyquist theorem, it’s all ok (now I’m pulling your leg - you never said that). So it’s not the S/N-theorem that is questioned but real world consequences. For that is what we must live with.
A while back we had a discussion about what the Nyquist theorem actually says. My point was that the perfection that is claimed for Nyquist is not practically realizable because it requires processing an infinite amount of data, which is awkward because it requires an infinite amount of time and memory and power and money. Nyquist says that the signal has to be band limited, and a perfectly band limited signal cannot be time limited.
This leads to practical limits. In theory there is no difference between theory and practice, but in practice there is…
Anyway, Audiostream just published an analysis of this. It is not aimed at audiophile stuff, but it is educational – if you wade through the examples it illustrates many of the related challenges and misinterpretations.
OMG, what a big pile of bullshit.
Everybody with two years of college physics education will shake his head in disbelief, how someone can twist Nyquists work in such a perverted way.
And everything only to justify to demand big money for useless so-called improvements.
And that is NOT educational and it IS aimed at audiophile stuff.
No well educated engineer will based his work on this strange theories or spend any money for it. Go in any lab and see with what equipment professional work with sampling is done outside the audiophile surrounding. You will not find any voodoo devices to improve the results.
Are there any particular statements in the article that you think are patently false ? I thought it was a good explanation of aliasing and discussion of misconceptions.
I think we should strive to have a more civilized tone here in the Roon forum than is common in the CA and ASR forums.
If this is voodoo, you are convinced that we get perfect reconstruction as long as we stay below the Nyquist frequency? Cool, enjoy your Redbook content and devices.
You are right. I will keep calm.
I do, thank you.
To be specific, the quote from the article:
"The difficulty with the Nyquist-Shannon sampling theorem is that it is based on the notion that the signal to be sampled must be perfectly band limited. This property of the theorem is unfortunate because no real world signal is truly and perfectly band limited. In fact, if a signal were to be perfectly band limited—if it were to have absolutely no energy outside of some finite frequency band—then it must extend infinitely in time.
What this means is that no system that samples data from the real world can do so perfectly—unless you’re willing to wait an infinite amount of time for your results."
Is a canard. No one has ever claimed that the Redbook samples audio that is limited to 22.05 kHz. The Redbook upper frequency limit is based on the physical limits of human hearing. Humans do not hear above 20kHz even when they are very young and in a perfect listening environment. Depending on the individual, the upper limit is in the 15-18kHz range and at these upper frequencies the energy required for audibility approach the levels that produce physical pain or hearing damage. The Redbook limits are based on practicality and the fact that there is no need for reproducing sound above 22.05kHz because people can’t hear that high.
The low pass filters employed in early digital audio devices were, in a word, nasty. This is not a result of sampling and reconstruction, but rather of economics and manufacturers engineering products to hit price points and have reasonable profit margins. This can be easily illustrated by the work of Frank van Alstine, who for years ripped out the analogue stages of Philips CD players and replaced them with his own circuits which greatly improved the sound quality. no changes were made to the digital stages.
Hello, I was the one who sent the paper to audiostream to ask the author for permission for publication.
Here is a link to the paper
The main quote I’d like to highlight from the paper is
"Measured purely from a sample rate perspective, increasing the signal sample rate
will always increase the signal fidelity. It will often decrease the cost of any analog antialiasing
and reconstruction filters, but it will always increase the cost of the system digital
hardware, which will not only have to do its computations faster, but which will need to
operate on more data"
Before I get into further discussion, I am new to this forum, and I generally agree with AndersVinberg about his take on the Nyquist Shannon sampling theorem.
I don’t know how to private message on here. Is there anyway I can get in contact with you AndersVinberg?
Thanks
To send a private message, click on the circle with a letter or symbol to the left of the name, and then click on Message in the popup.
If you can’t hear the difference, look at a square wave on your scope: The pre- and post ringing is not in the original signal (I mean the analogue sound that was recorded) since in the analogue world pre-echo’s might not exist (note the careful statement, I know too little of Einstein’s work and bending time). The waves over the top of a square wave are the result of the band limiting and are not interesting, but the pre-ringing and post-ringing (and sometimes other artifacts like overshoot) are.