Why does vinyl (analog) sound better?

Lee,

The Nyquist theorem is a perfectly valid and well-proven mathematical construct for digital sampling and subsequent analog recreation of sinusoidal-type waveforms and any combination thereof (which would include audio/music waveforms). Nobody seriously questions the theory itself in the context for which it was formulated. The questions arise, however, concerning it's practical application in the realm of electronic circuits and filters, and its relevance to the entire phenomenon of human hearing/audio perception. I think this is what you are getting at when you say it perhaps doesn't describe all dimensions of the audio event. I'm just attempting a little pre-emptive strike on your behalf. You have to realize that any time or transient event has a corresponding frequency component and vice-versa. It is simply a different mathematical domain for describing the same phenomena. Luke is correct about this. What isn't certain to me is that human perception cuts off at 20 KHz, as some seem to believe. True, you can't hear discreet pitches or tones above that, but that doesn't necessarily mean there is no perception of the energy above that point if it is present in music. This may be the crux of the matter, and the reason why some audio designs continue to incorporate bandwidth capabilities past 20 KHz. Plus there is the matter of correct (undistorted) phase relationships, which to be preserved, require more bandwidth than just 20-20K.

 

That's what I am getting at...something is missing and it's not small. I remain unconvinced (perhaps wrongly, perhaps not) that the formula can describe sharp transients like those found in a live music event.

 

I think you're misreading me, or else confusing me with someone else. I'm not saying 16/44 will in practice necessarily capture everything. I'm saying that if it doesn't capture something, it isn't for the reasons being put forward.

 

Luke,

Let me ask this question. Do you have any evidence that suggests sharp transients are indeed captured by Nyquist?

I think there have been some AES papers that bring into question this point.

 

Define "sharp transients". Sharper than 22kHz?

What evidence do you have to suggest they aren't?

 

The attack of a snare or cymbal or sax note...

Listening tests clearly suggest 16/44 is not good at capturing this.

 

Actually it discusses analog filters too. And I know dekkersj is aware of these things because I've been reading his generally excellent posts on hydrogenaudio.org for quite awhile now.

 

And note that by "evidence", we mean "not conjecture".

The "evidence" we have for the Nyquist theorem capturing sharp transients is that it is mathematically proven. You have a bandlimited signal, you sample it, you reconstruct it, you get the exact same signal back out, in every single degree. Seems fairly airtight to me. The people who don't believe that are the people who don't know math or how to read a proof.

Of course, there are about 15,000 technicalities involved in making that happen acceptably in the real world. But so far, out of all the AES articles I've seen on the subjects, and all the debates I've seen, I have seen a whole lot of conjecture, even a lot of conjecture that people use to build products out of, but no evidence.

 

What is the frequency of such attacks?

 

I would agree but....
1. that was not what I meant by science being intuitive.
2. Nyquist aint science. It's math.

 

No. No. No. No, that's simply not true. Sorry. And that's strange because in the first paragraph you got it right. But here you are seriously mischaracterizing he position of the 'other side', which generally proposes that the inherent sound of the LP/turntable system really can and does sounds better *to those who like added, 'euphonic' effects resulting from the LP format + playback technology*. Examples include midrange phase distortion. No is saying it doesn't or *can't* really sound better to *anyone* -- including you. Because '*sounds* better' is *always* going be a purely subjective call.

Some people feel mono sounds better than stereo. Some people prefer the sound of recorded music, at home, to live performances. No one I've seen, including the most rabid objectivists, says they *can't* have such preferences, or that their preference is 'wrong'.

What can be factually wrong are the reasons people give for their preference, e.g., nonsense like 'analog has infinite resolution' or 'there are spaces between the music in digital'.

Too often what it comes down to -- or starts out as, just as often -- is one or both sides mischaracterizing the claims of the other, due (I suspect) to innocent or willful lack of understanding of the facts behind them. Then ridiculous amounts of time and verbiage are spent trying to correct / inform.

 

As I just pointed out in a previous post. Nyquist isn't science it is math. Nyquist is mathematically indisputable. The beauty of math. But the issue is the real world engineering involved in applying it.

 

Wrong. Both Shannon and Nyquist were investigating practical problems of communication and signalling: engineering problems. It was *never* just an abstract math problem divorced from the material world.

To call it just 'math' requires calling nearly all audio engineering problems, and not to mention theoretical physics, 'math'. Simply because they require math.

H. Nyquist, "Certain topics in telegraph transmission theory," Trans. AIEE, vol. 47, pp. 617-644, Apr. 1928.
C. E. Shannon, "Communication in the presence of noise", Proc. Institute of Radio Engineers, vol. 37, no.1, pp. 10-21, Jan. 1949.

And no, science is not merely 'intuitive'. If it were, it would be vastly easier to do. Intuition has a place, but diligent application of plodding old *reason* is essential.

 

I don't think the issue is frequency. I don't doubt that digital sources can produce all the audible frequencies, and probably some that aren't audible (at least by me). But in order to make a realistic sound of attack, the source and playback system have to accurately reproduce the "suddenness" of the attack, the dynamics of the attack (how quickly does it peak, how quickly does it decay, and the correct slopes of the peaks and decay), and the entire length of the event. I've never heard CDs reproduce these elements as well as decent vinyl.

 

You didn't bother to read Mr. Lavry's nice little paper, did you? What frequencies are being left out, exactly?

Continous does not necessarily mean more accurate.

 

Nyquist theorum itself is pure math. Also I never said science was "merely" intuitive. I simply said it was intuitive. It can also be quite deductive. Rest assured Einstein was right when he claimed intuitive leaps are needed for science to advance. he was speaking from experience. But mostly I suggest you think more about the issue of math verses science. Math is a language used describe things. Science is a method used to find out what and how things work. Nyquist theorum is pure math. That is why it is indisputable. It is not subject to falsification via new evidence. No evidence can disprove something that is pure math. There is no evidence involved in the proof for Nyquist theorum. It is proven mathematically.

 

This was *never* even close to the scientific consensus, and AFAIK appeared in exactly *one* book about insect flight, attributed to a third party. The claim was the offspring of back-of-the napkin calculations by a Swiss aerodynamicist talking to a biologist, back in the 30's.

see

http://www.math.niu.edu/~rusin/known-math/98/bee

for the rest of the story.

And of course, in the case of the bee clearly it *does* fly.

Actually, the proper analogy to the bee story would be: 'Vinyl shouldn't sound good, yet it does'.

 

Let's assume you're right and that everyone literally believed advertising slogans (including the audio mags that raved about some digitally-recorded LPs, then CD themselves), thus feeling horriby betrayed when any CD didn't sound good 'enough'. (One wonders why they, even 'audiophiles', were so primed for something better than LP....)

So, it's 2006..and digiphobes *are still saying it*.

 

OK, I dig what you're saying now. Sorry for the misread or erroneous attribution of opinion. Is it your position that no one has yet proposed or demonstrated a valid reason or cause?

 

Yes "digiphobes" are still saying it. Some are still claiming that all redbook CDs and players are completely transparent too. Don't believe me just read up on Howard Ferstler's claims. Some people say the darndest things. Perhaps there would have been a lot less digiphobes had there been some critical listening and honest public evaluation of the media back in the day on the par t of the advocates. Perhaps there would have been a lot less wild speculation on wy so many LPs sound better than their CD counterparts had there been more honest and openminded discussion from the CD advocates early on. I'm not excusing the wild things people say as to why LPs often sound better than LPs but I do understand it.

 

I believe this is basically correct. However, once we apply the theorem to the question of physical electrical circuits which exhibit non-ideal behavior, along with human hearing and perception, then we have taken the definitive step from the abstract to the "real", not when we crossed the border from pure mathematics to signal processing theory as such. Signal theory may be engineering but it is still a type of mathematics, or applied math if you will. You can write it all out on paper, though perhaps certain assumptions are made to simplify calculations and equations along the way.

 

Bingo.

As for SACDs, unlike CDs, the spec for them *forbids* digital 'overs' (seen as digital level slamming up and 'beyond' 0 dB). So the only way to get the sort of dynamic range compression seen on too many modern CDs is to pass the SACD through a PCM stage, do the dirty work there, then bump the digital file back over to DSD (lowering the levels first). In the half-dozen or so SACD's I've tried to 'look at' , I haven't seen that done, which is a good sign.

So and SACD stands a good chance of sounding better than its cognate modern CD remaster, simply by virtue of what remastering engineers aren't allowed to do to them.
That of couse has nothing to do with the higher sample rate and fewer bits of DSD.

Nope, though of course to take theoretical advantage of digital one wants a mic with a good frequency response out to 22 kHz, and speakers to match. The nonnegotiable requirements would be to level match and tim-synch, and of course to do the comparison double blind. Then all you have to worry about it whether you are comparing the filters, rather than the formats

 

The thing is, that *is* frequency. If that "suddenness" as you call it is less than 22 kHz, 16/44 won't have a problem with it. Frequency=time domain.

There might very well be some issue causing what you are hearing. It's just incorrect to assume that there's some time based voodoo that 16/44 isn't picking up.

 

Strongly disagree. Even with a perfect mastering job, many things like cymbal decay are not realistic in 16/44.

Nope. You simply set up the same recording session and feed the mics into hard drive decks recording at 16/44, 24/96, and DSD (and analog tape if you want) and compare. You use the best filters can find within a certain reasonable price point and you have a sufficient test for practical implementation of each format.

 

The problem Luke is that you don't know that it isn't time-based related. It might very well be. At best, it's an open question for everyone.