Demonstrations of sampling, quantization, bit-depth, and dither [video] | Hacker News Reader

Demonstrations of sampling, quantization, bit-depth, and dither [video]

xiph.org

28 pointsnathankot6 comments

Excellent video, such a pleasure to watch. Even with all the equipment and setup problems, labs are such a good memory...

(Just a minor thing but I thought that the video suggested that the Gibbs phenomenon was a result of the finite decomposition instead of, unless my memory is misleading me, being a sort of fundamental feature of the piecewise smooth square wave function)

Most xiph videos are of very high educational quality.

There are a lot of basic errors in this. It may impress someone who doesn't know anything about the technology - nice vintage gear collection - but that doesn't keep it from being factually wrong.

Just one example: the true SNR of a professional reel to reel tape master recorder is somewhere around 100dB - e.g. a Studer A series (around 75dB SNR @ 30ips) with Dolby SR noise reduction (about 25dB of noise reduction.)

Anyone who knows so little about recording equipment they think this is equivalent to 13 bits of digital resolution shouldn't be making videos about this subject.

That sounds like a factual error, but could you point out more basic errors in the science of what he's claiming?

I don't know enough to verify your claims, but your argument has enough logical fallacies in it to make me question your seriousness:

> nice vintage gear collection - but that doesn't keep it from being factually wrong.

This is a snide remark. You are implying that the presenter believes his vintage gear collection somehow adds to his factual standing. Of course it doesn't, nor did he imply it was.

> the true SNR of a professional reel to reel tape master recorder is somewhere around 100dB

That doesn't contradict anything he said.

> Anyone who knows so little about recording equipment they think this is equivalent to 13 bits of digital resolution shouldn't be making videos about this subject.

This is equivalent to just saying "he's wrong and he doesn't know what he's talking about". You're not adding anything to the conversation. You may or may not know anything about the subject, but you haven't added it here. If you calculated the bit depth equivalent of a Studer and showed that it was higher than 13 bits that would be interesting.

Well, it is equivalent. Not in terms of what a played back sound is like, because tape and digital both introduce characteristic artifacts, but in terms of "this signal peak is so many dB above the noise floor."

If you actually code some audio DSP(and I do), you quickly learn that there are a lot of ways to color sound that aren't quantified except in a very broad statistical sense, but are measurable in audition. What digital is very good at is the basics of reproduction with a (statistically) uncolored result. At artistic effect, it is somewhat more limited; a lot of compute time and engineering has to be thrown into reproducing some straightforward analog world things and model accuracy still has room for improvement even now.

If you want to quibble with the video, do so on its terms. It isn't saying that digital is better, it's saying that digital is better at a very specific task.

Reply on news.ycombinator.com