16/44 vs. 24/96 Format Comparison

[pfarrell]

opaqueice wrote:
> SoftwireEngineer;230562 Wrote:

>> If you look at it that way, 44.1khz is good enough by Nyquist
>> criterion.
>
> What do you mean by "Nyquist criterion" if not that frequencies above
> 22kHz (or 18 for that matter) are inaudible?

I'm not SE, but 44.1 is how we get perfect sound forever. Too bad it
wasn't 55kHz@18, we might have received what they claimed. But then, CD
were aimed at replacing casettes, not LPs.

[Timothy Stockman]

44.1 is how we get perfect sound forever. Too bad it
wasn't 55kHz@18, we might have received what they claimed. But then, CD
were aimed at replacing casettes, not LPs.

You probably know this, but the reasons we got 16/44.1:

As I recall, Philips wanted 14-bits; that was about the limit of good DAC chips that could be produced for a reasonable price at the time. It was also what the EIAJ digital audio standard specified. However, Sony held out for 16-bits and eventually got Philips to accept it.

So far as 44.1, the compact disc standard was developed before the days when computers (and hard drives) were up to the task of digital audio. They wanted a recording medium that could be edited, and so they adapted existing U-matic video recorder technology to the task with the PCM1610 processor. They used the European 625-line, 50-field video rate, and after they split it up into bits, they ended up with 44.1KHz. (Most PCM adapters with also run at 525/60, and this ends up with a sample rate of 44.056.) I guess they could have done what Thomas Stockham did, I believe he used an instrumentation recorder for his system and got a somewhat higher sample rate, but I think Sony and Philips prefered to use existing video recorder technology since it was already widely available in the entertainment industry. Also, they probably had an eye toward integrating digital audio with video, and achieving sync would have been more difficult at a oddball sample rate.

As for which format replaced which, yes, I guess it could be said that CDs replaced cassettes, because cassette sales had already overtaken LP sales before CD sales became a significant part of the equation. Since then, LPs have never been (and probably will never be) more than a niche market. In fact, minidisc, not CD, was actually aimed at replacing cassette, but Sony's usual terrible marketing coupled with the rise of MP3 players before MD got a significant foothold put the nail in MD's coffin.

[pfarrell]

Timothy Stockman wrote:

> As I recall, Philips wanted 14-bits; that was about the limit of good
> DAC chips that could be produced for a reasonable price at the time.
> It was also what the EIAJ digital audio standard specified. However,
> Sony held out for 16-bits and eventually got Philips to accept it.

16 bits is nice in that its 96dB snr, making it easy to claim perfect
sound. I think a lot of the early DACs were 12 bits back then.


> So far as 44.1, the compact disc standard was developed before the days
> when computers (and hard drives) were up to the task of digital audio.

So true. If they were a bit more computer aware, they could have used
computer disk technology to read the disks rather than the crude stuff
that RedBook has.

There is some argument about when the design was done, there were PR
releases in 1981 or so. In 1980, I bought a 200 MB disk drive, it cost
$33,000 and was the size of a washing machine. In about 1984, I bought
my first PC hard disk, it was 5MB, cost $5000, and was full height 5-1/4".



> already widely available in the entertainment industry. Also, they
> probably had an eye toward integrating digital audio with video, and
> achieving sync would have been more difficult at a oddball sample rate.

Altho odd ball rates are still with us, see all the 3/2 pulldown for
video of movies, etc.

[kifysara]

Yes. Why?

Perhaps you would like to check the Hydrogen Audio Forums. In the "Listening Tests" sub-forum there is a debate going on concerning this AES paper which - at least for me - clarified some points.

For example, a quote from the guy that opened the thread (hope he doesn't mind):

"The paper concludes with the note that the high res releases sounded much, much better than the same music on CD, for well-known mastering reasons."

[darrenyeats]

For example, a quote from the guy that opened the thread (hope he doesn't mind):

"The paper concludes with the note that the high res releases sounded much, much better than the same music on CD, for well-known mastering reasons."

I saw that hydrogenaudio thread not long ago. When you say "clarified some points" I am a bit worried since I thought your quoted statement in particular was an unclear one. And in fact I still don't know which way you interpreted the quote. :-)

AndyH-ha explains it just a little later in the thread: "If so, then the tests say there is nothing at all about SACD per se that is better -- except the mastering standards, which are quite irrelevant to the tests. Those differences from CD are arbitrary processing decisions and have nothing per se to do with the different technologies."

So SACD (or DVD-A) may well be mastered better than CD but these are commercial or political decisions and nothing to do with the format qualities.

The paper reports they took hi-rez material and down-sampled it to red book appropriately, and people couldn't tell the difference. The point is when you use identically mastered source no difference can be detected between the consumer formats.
Darren

PS: The test involved even more than down-sampling - in fact a CD-quality A/D/A loop after the outputs of a hi-rez player, a neat way to minimise the variables. So we hope the physical equipment was transparent...it seems so because no differences were detected from it (or of course the change in format) according to the report.
Darren

[darrenyeats]

I'm not SE, but 44.1 is how we get perfect sound forever. Too bad it
wasn't 55kHz@18, we might have received what they claimed. But then, CD
were aimed at replacing casettes, not LPs.

Based on my beliefs about human nature, and audiophiles in particular, I feel certain we'd be having an equivalent conversation had it been 18/55 or 24/96 to start with.
Darren

[opaqueice]

Perhaps you would like to check the Hydrogen Audio Forums. In the "Listening Tests" sub-forum there is a debate going on concerning this AES paper which - at least for me - clarified some points.

For example, a quote from the guy that opened the thread (hope he doesn't mind):

"The paper concludes with the note that the high res releases sounded much, much better than the same music on CD, for well-known mastering reasons."

They do make a comment like that, but as darren says, it simply means that SACDs are (in some cases) mastered in such a way that they sound better than the equivalent CD. According to their results, that SACD could be released as a CD and the improvement would still be there.

The test was to compare an SACD playing through an SACD player (so at full resolution) to the output of that same player passed through an A->D->A chain, with the A->D->A at 44.1/16. So listeners heard the same music, but in different resolutions. None could tell the difference.

[jt25741]

I think it is highly fruitless to talk about 'audibility'. This seems like a relative/subjective thing. There are people who cannot tell the difference between a boombox sound and separates sound. If I am in the academic world, I would just devote my time to show which standard recreates an analog waveform accurately (theoretically). If you look at it that way, 44.1khz is good enough by Nyquist criterion. but the bit depth of 16bit can cause quantization errors. So I would always like to go for the wider bits whenever possible.

Indeed. Nyquist says that any frequency above 1/2 the sampling frequency is garbage, and must be thrown out(filtered). It doesn't say that sampling frequencies that are higher do not approximate the analog waveform more precisely! This is a common misunderstanding. Just because 44.1Khz will provide a valid approximate data point from sampling a 20Khz signal -- it will yield approximately 2 samples here. Anything less -- well is not correlated. Again, 2 samples are just barely enough where you will not have noise and produces a very crude curve at that Not as nice as as say -- 4 at 10Khz, 8 at 5Khz -- or 50 or so at 800Hz -- all from Redbook.

The bottom line is, higher sampling rates "can" sound better and "should" be better -- particularly at the higher encoded frequencies(cymbals etc.) The fact that they often don't sound any better can be because of numerous reasons -- many of which have already been suggested.

[pfarrell]

Indeed. Nyquist says that any frequency above 2X the sampling frequency is garbage, and must be thrown out(filtered).

The key word here is *must*. Its not optional, you must do it.

And one of the claims in favor of higher sample rates, and oversampling in general, is that you can use simpler filter circuits. Many early systems had 12dB/octave filters, which really screw up phase. The designers needed the steep slope to avoid causing massive aliasing problems. With 88.2kHz, you can use a 6dB/octave filter and be as free of aliases as a 44.1 with a 12dB/octave filter.

The 16 vs 24 argument is less well grounded. Since recording engineers consider -70 dB to be gone, the 96dB provided by 16 bits is overkill, in theory. But there are established audible tests that show that dithering of the 16 bit signal is important. With a few more bits, dithering becomes less important. And in the days of PCs with RAM enough to hold an entire RedBook CD in memory, adding a few bits is easy and cheap.

[opaqueice]

Indeed. Nyquist says that any frequency above 2X the sampling frequency is garbage, and must be thrown out(filtered). It doesn't say that sampling frequencies that are higher do not approximate the analog waveform more precisely! This is a common misunderstanding. Just because 44.1Khz will provide a valid approximate data point from sampling a 20Khz signal -- it will yield approximately 2 samples here. Anything less -- well is not correlated. Again, 2 samples are just barely enough where you will not have noise and produces a very crude curve at that Not as nice as as say -- 4 at 10Khz, 8 at 5Khz -- or 50 or so at 800Hz -- all from Redbook.

The bottom line is, higher sampling rates "can" sound better and "should" be better -- particularly at the higher encoded frequencies(cymbals etc.) The fact that they often don't sound any better can be because of numerous reasons -- many of which have already been suggested.

You're very confused about this. It's not a "common misunderstanding". Just ask yourself what the difference is between a 20 kHz square wave and a perfectly smooth 20 kHz sine wave.

Answer - a 60 kHz sine wave (plus higher frequencies). So if you can hear 60 kHz, you'll know the difference.