dumbo
dumbo's picture
Offline
Last seen: 4 months 1 week ago
Joined: Sep 26 2009 - 6:59pm
Whats the point of a 32bit DAC
John Atkinson
John Atkinson's picture
Online
Last seen: 14 min 30 sec ago
Joined: Nov 7 2010 - 3:31pm
dumbo wrote:
I was recently looking at the new Oppo Universal player with some interest and noticed that it utilizes the latest ESS 9018 Reference DAC chip which boasts a 32bit "Hyperstream" architecture. Now this just seems silly to me but certainly no more silly then the Chord 64bit DAC.

There are no 32-bit media, let alone 64-bit. But the output of the digital filter in a DAC or player could have a 32-bit bit depth, and feeding those data straight to a DAC that can accept them will result in one less processing stage. Does it matter? Probably not.

John Atkinson
Editor, Stereophile

j_j
j_j's picture
Offline
Last seen: 11 years 10 months ago
Joined: Mar 13 2009 - 4:22pm

Ok, let's see, first, the noise floor at your ear is about 6dB SPL white noise, broadband across 20-20K. (the broadband nature is why you can't hear it)

So, if we take that as the lowest level, 32 bits is 32*6.02=192.4dB above that, or an SPL (if it were possible in a linear sense, which it is not) or about 198.4 dB SPL. That's for 32 bits uniform PCM. Floating point would be orders of magnitude worse.

Now, 194dB SPL is 1 atmosphere RMS, which is already impossible, but let's ignore that for a minute.

That would mean a pressure, raw pressure, not above/below atmospheric, from -.6 atmospheres, to 2.6 atmospheres. Now, you can't have below zero absolute pressure in the macroscopic world.

But what does the 2.6 atmosphere peak mean? Well, for one thing, that's 1.6 atmospheres of overpressure, or a pressure of about 1.75 TONS per square foot all over your body, windows, doors, etc.

The proper term for such overpressures (and overpressures, unlike 'under zero pressure' is quite possible) is "military". You would not survive the experience, and the only ways to create such pressure in the atmosphere involved large quantities of unstable materials.

dumbo
dumbo's picture
Offline
Last seen: 4 months 1 week ago
Joined: Sep 26 2009 - 6:59pm

I guess If the benefits of using a higher then 24bit DAC are that it requires less processing of a 24bit song (ie..same concept as using a higher Ghz PC for chugging away on large math calculations) then I can now see the purpose for using it.

If on the other hand the above is not the purpose for using a 32bit bit-depth or higher DAC and instead it is sold under the impression that a human will hear the difference between it and a 24bit DAC then I still don't see the reasoning behind it. Especially since we only have 24bit media to play with and their are no new inventions on the horizon to increase the capabilities of the human ear.

absolutepitch
absolutepitch's picture
Offline
Last seen: 2 months 4 weeks ago
Joined: Jul 9 2006 - 8:58pm

If your 32*6.02 gives 192.64 dB, that seems to mean that it is 6.02 dB/bit? Maybe I don't understand.

j_j
j_j's picture
Offline
Last seen: 11 years 10 months ago
Joined: Mar 13 2009 - 4:22pm
absolutepitch wrote:

If your 32*6.02 gives 192.64 dB, that seems to mean that it is 6.02 dB/bit? Maybe I don't understand.

 

Dynamic range of a uniformly quantized PCM system is 20*log10(2) per bit.

There is an additional constant, so it is, strictly speaking, a bit deeper than 6.02*nbits, it's 6.02*nbits+c for a small constant in dB. That small constant, however, is usually consumed by the reality of circuitry, give or take a dB, so the rule of 6dB/bit (which is really 6.02 but the .02 isn't so important) is a very, very good rule for such a system.

The reason is simple, when you add another bit, you reduce the noise due to (dithered) quantization by a factor of 2 in amplitude, or 4 in power. That's the 6dB.

absolutepitch
absolutepitch's picture
Offline
Last seen: 2 months 4 weeks ago
Joined: Jul 9 2006 - 8:58pm

jj: If we take an example of a 2-bit system, that should be a dynamic range of 12 dB. There are 4 distinct values for the 2-bits. Does that mean that each binary increment in value are in 4 dB steps, from binary 00 to 01 to 10 to 11, representing 0, 4, 8, and 12 dB levels?

j_j
j_j's picture
Offline
Last seen: 11 years 10 months ago
Joined: Mar 13 2009 - 4:22pm

It's 4 levels, at +-.25 and +-.75 of the maximum (for reconstruction). The decision levels are +-.5 and 0.

Thinking of these as "4 dB steps" is not right, it's hard to explain why, but basically, a dB scale would mean logarithmic quantization, which is not what uniform PCM is. Uniform PCM has equal sized 'step sizes'.

 

Digital Signal Processing, by Proakis and Manolakis, pages 32 and on for a few pages, or Discrete Time Processing by Oppenheim and Schaffer, pates 192 and on a few, have reasonably accessable explainations.

Having looked through some texts, it's time for somebody to make a nice, graphic explaination. It's not that hard. Hmm. Well, yeah.

paulsax
paulsax's picture
Offline
Last seen: 11 years 5 months ago
Joined: Feb 25 2010 - 3:09pm

Not my field but I thought the 6db (when refering to quantizeation)  rule was a signal to quantized noise ratio thing instead of a sound pressure level thing.   Been horribly wrong before but something is not adding up to me on this one.  Been a few years though......

j_j
j_j's picture
Offline
Last seen: 11 years 10 months ago
Joined: Mar 13 2009 - 4:22pm
paulsax wrote:

Not my field but I thought the 6db (when refering to quantizeation)  rule was a signal to quantized noise ratio thing instead of a sound pressure level thing.   Been horribly wrong before but something is not adding up to me on this one.  Been a few years though......

 

You're right, it's a signal to quantization noise issue. 

 

The SPL comments somewhat far above are assuming you match the quantization noise to the SPL of the noise of the atmosphere at your eardrum.

absolutepitch
absolutepitch's picture
Offline
Last seen: 2 months 4 weeks ago
Joined: Jul 9 2006 - 8:58pm

Well jj, looks like I need to do some reading to understanding what you mean. I'm not up on the theory so need to learn some fundamentals first. Thanks.

amblin
amblin's picture
Offline
Last seen: 12 years 9 months ago
Joined: May 12 2011 - 11:26am

hmmm.. i'm still using my almost 20years old 16bit philips LHH1000 (dynamic range = 96dB, S/N ratio= 101dB, seperation= 100dB,  THD @ 1kHz = 0.0015) ... And i still consider it one of the best.....And it's quite heavy.. around 60 pounds, well, heavy is good..especially when most weight goes to internal parts, not ultra thick bullet proof metal box..in fact i've seen cd players feature aluminium faceplate thicker than the front armor of M113 APC..(3.8cm/1.5in) .......

More bits, more detail, more this and that does not necessarily mean "better sound".  If you only look into the technical numbers on paper, sure,those fancy, brand new machines will be better, but music is more than just 'bits' or 'dBs'.  ( and no, i'm not one of those vinyl crazies that refuse to accept new things. . actually i sold all my vinyl collection for my new system based on cd.....

   

John Atkinson
John Atkinson's picture
Online
Last seen: 14 min 30 sec ago
Joined: Nov 7 2010 - 3:31pm

amblin wrote:
i'm still using my almost 20years old 16bit philips LHH1000 (dynamic range = 96dB, S/N ratio= 101dB, separation= 100dB,  THD @ 1kHz = 0.0015) ... And i still consider it one of the best...

The LHH1000 was a great player in its day, but its technical performance is exceeded by the current generation of DAC chips. You can find my 1989 review at www.stereophile.com/content/philips-lhh1000-cd-player, though I should note that my measurements predate the magazine's acquisition of an Audio Precision test system.

John Atkinson
Editor, Stereophile

amblin
amblin's picture
Offline
Last seen: 12 years 9 months ago
Joined: May 12 2011 - 11:26am

Wow,John,  what a pleasure that my little 2 cents got your attention.

Actually back in the mid-early 90s,  i used to own a translated copy of your original review but for some reason they left out the technical measurements  (translated by a magazine based in taiwan). And that's the one article that made up my mind buying the philips. ha ha ha.

I totally agree that the new chips are more sophisticated and more capable  converting all those 1's and 0's, they have to be. But maybe it's because i got my ears so 'converted' by the philips over the years so i found most new machines (the ones i had the pleasure to hear, of course) a little bit too 'bright' for my listening taste.  ( my more up to date cdp is the marantz SA7) . I don't know, maybe to me, it's like comming across a nice old nikon F3 in the age of highly digitalized DSLRs. ( i was about to say leica m3, but i think the philips is not that timeless. lol)

Log in or register to post comments
-->
  • X