I want to discuss the Nyquist theorem, anyone interrested?

[ingrast]

...Another flaw with CD is the number of bits.
As a digital format, music is represented in a finite resolution in small steps(against the analog's infinite resolution).
....

Unfortunately a recurrent misconception.

A man by the name Claude Shannon demonstrated elegantly and beyond doubt that a physical communication channel with finite signal to noise ratio does indeed implicitly quantify the information conveyed.

His little opus known as Information Theory marked some conceptual turning point that is the foundation of modern communications technologies.

Putting in other words, the effective quantization step in an analoge channel, strange as it may sound, is nothing more than the minimum difference in signal levels that can be distinguished in spite of the noise.

Even an entirely analog communication channel has an inherent equivalent information transport capacity, resolution as said above, the other constraint being the bandwith or maximum transmission bit rate to put it simply.

The sample word size, be 16, 18, 20 or 24 bits simply expresses the maximum dynamic range available. If it is less than the downstream reproduction chain, it is the limiting factor, otherwise not (for example if your preamplifier - amplifier has a lower dynamic range).

Rodolfo

 

[AndyN]

I want to discuss the Nyquist theorem

Ok, but you'll need to talk twice as fast as the rest of us so we can get the full meaning of what you're saying.

😀 Sorry, couldn't resist..

 

[Mr Evil]

...Nobody. You need to know what it sayd basically, but a suitable design may be better attained by actually comparing different sampling rates...

Sure, you can use a DAC or whatever without knowing how they work, just like you can put together a chip amp without ever even knowing what a transistor is, maybe ever a good one, but that's not the same as designing one.

16 bits is perhaps not enough for mixing(no experience),but with recording I only hear improvements from a higher samplingrate.
an 18 or 20 bit adc yields no improvement to my ears.

It may not be a big difference, but it is there. You need a reproduction system with good enough SNR for the difference not to be lost. As I mentioned earlier there have been good, scientific, tests done showing the audibility of higher than CD resolutions. As you say though, the difference for mixing is much bigger. I wouldn't go without my 24bit card for music work!

 

[hummhoom]

Technology has come a long way since the original Red Book CD. Now that DVD-Audio is becoming more common place, we can all enjoy 96KHz/24bit sound.

 

[Werner]

In this case, the Nyquist theorem states that we can perfectly reconstruct the *band limited* signal.

In other words, the Nyquist theorem says absolutely nothing about the sonic damage that results from the brickwall lowpass filter.

On the other hand, if you sample audio at 192 kHz, a low-pass filter isn't really even needed.

Mr Hansen posts a little gem with three very relevant points in it, and it still goes largely ignored.

Tasting unfiltered malt, as well as emailed discussions with Mr Qvortrup and Mr Kobayashi, led me to investigating the unfiltered recording side, only to find out later that Mr Faulkner was there (long) before me.

Some of you may want to read this:


http://www.stereophile.com/features/104law/index.html

When archiving LPs to CD I nowadays record at 88.2kHz, filter out above 30kHz with a reasonable slope, and then subsample by simple averaging. It sounds interesting enough and some of you are probably equipped to try it yourself ...

 

[phase_accurate]

Thorsten wrote:

No-one criticises Nyquist. I personally merely challenge an obvious and blatant misapplication.

My opinion is also that the sampling theorem isn't a GUARANTEE but a MINIMUM requirement and should therefore be treated as such.

Carlos wrote:

What you surely don't need is 1 bit at 2.8Mhz.

Anyway, I always simplify analog filters and what always sounds clearly better to me is a 1st order lowpass filter at some hundreds of khz , with a small cap across the feedback resistor on the buffer/gain stage.

Are you aware of the fact that the bitstream (that you seem to dislike) is the ideal candidate to be converted in your preferred way ?

16 bits is perhaps not enough for mixing(no experience),but with recording I only hear improvements from a higher samplingrate.

During a discussion a famous sound engineer once mentioned that the 32bit @ 352.8 ksamples/s of their mixing consoles were what he regrads as sonically transparent ! But one has to be aware that the requirements regarding resolution are higher for processing of signals than they are for simple storage and this is due to computational errors.

Regards

Charles

 

[carlosfm]

Are you aware of the fact that the bitstream (that you seem to dislike) is the ideal candidate to be converted in your preferred way ?

Yes, and it can sound quite good, very detailed.
What bitstream (and delta-sigma) dacs lack is that dynamics of the best multibits.

 

[Bernhard]

In the CD-DA format we have 16 bits, not enough to represent music, although it can sound quite good.

Some very simple maths:😎
"In a 16-bit audio format, we can represent a sinusoidally varying voltage audio signal by 216 or 65,536 discrete levels. It is apparent then that quantization is limiting performance factor in the overall digital audio system, by the number of bits allowed to the quantizing system. The system designer is faced with determining how many bits create a sufficient model of the original signal."

Unfortunately most DACs lack sufficiant linearity to repoduce 16 bits.

So what do we need 18 or 20 bits for ?

 

[carlosfm]

Unfortunately most DACs lack sufficiant linearity to repoduce 16 bits.

So what do we need 18 or 20 bits for ?

True.
I was talking about the recordings.
You just need a 16bit dac to play 16bit recordings.
No need to play tricks with the bits.
The best dacs for CD audio are 20+ years old.😎
My oppinion, of course...

 

[mr_push_pull]

Some of you made some remarks about me not participating in the discussion. I think the title of the thread is unfortunate, I shouldn't have used the word "discuss". What I wanted was someone to explain where the confusion comes from with this Nyquist thing.
If you expect to read something like "multibit DACs sound better" from me, I'm not your man.
Nevertheless, I read all the replies and I think that the discussion is very educational.