Thread: Sacd bit processing

Posts: 9

Post by systemerror909 October 30, 2007 (1 of 9)
Okay, so i've been a big fan of sacd for a while now. In total I've got 36 sacds, with five more in the mail to me.

While i've enjoyed the format very much and agree that it has obvious sonic advantages to redbook, from a listening standpoint, I have never fully understood the theory behind how the bits are processed.

My point: Sacd is sampled at 1 bit, 64x redbook sampling rate of 44.1khz. Redbook on the other hand is multibit: 16bits at 44.1khz. While it appears that sacd has a tremendous advantage, with roughly 4x the amount of information that a redbook has, I am have been skeptical about how this information is distributed. Assuming that the 1 bit system is linear, wouldn't this mean that the 64x sampling really only equates to the maximum of a 6 bit difference for that period of time (since 2^6=64)? With redbook, arent' we capable of a maximum of 16 bits worth of difference in this same sampling period? The way I am interpreting this 1 bit system to be working is that is shifts all this extra resolution/dynamic range into the lower frequencies. Is this true, is this how sacd sound is processed?

I contend that there is room for greater accuracy in the timing of sound with sacd, since (and this may be imperceptible to human ears) the smallest difference in time between two sounds is 1/44.1k of a second, where as the blazing 2.8mhz sacd sampling rate makes timing much more accurate. I think this would maybe be a factor in for example binaural recordings where the timing between the two sounds is of utmost importance to soundstaging (again maybe differences in timing is already beyond human perception for redbook so this doesn't matter).

Anyway, this ends my rant, but I was just wondering if this makes sense to anyone else, and if I do have an accurate idea of how SACD vs redbook works.

Thanks

Post by zeus October 30, 2007 (2 of 9)
systemerror909 said:

While i've enjoyed the format very much and agree that it has obvious sonic advantages to redbook, from a listening standpoint, I have never fully understood the theory behind how the bits are processed.

You may want to read the FAQ (see the link above). DSD supports a bit depth up to 20-bit (120dB).

Post by Sam October 31, 2007 (3 of 9)
systemerror909 said:

My point: Sacd is sampled at 1 bit, 64x redbook sampling rate of 44.1khz. Redbook on the other hand is multibit: 16bits at 44.1khz. While it appears that sacd has a tremendous advantage, with roughly 4x the amount of information that a redbook has, I am have been skeptical about how this information is distributed. Assuming that the 1 bit system is linear, wouldn't this mean that the 64x sampling really only equates to the maximum of a 6 bit difference for that period of time (since 2^6=64)? With redbook, arent' we capable of a maximum of 16 bits worth of difference in this same sampling period? The way I am interpreting this 1 bit system to be working is that is shifts all this extra resolution/dynamic range into the lower frequencies. Is this true, is this how sacd sound is processed?

I contend that there is room for greater accuracy in the timing of sound with sacd, since (and this may be imperceptible to human ears) the smallest difference in time between two sounds is 1/44.1k of a second, where as the blazing 2.8mhz sacd sampling rate makes timing much more accurate. I think this would maybe be a factor in for example binaural recordings where the timing between the two sounds is of utmost importance to soundstaging (again maybe differences in timing is already beyond human perception for redbook so this doesn't matter).

You're right that SACD trades ultrasonic resolution for audio band resolution. It only makes sense to concentrate on what we can actually hear. :-)

Sampling rate limits the minimum duration of a sound event, but it's irrelevant to the time position accuracy. A sound event doesn't have to "wait" for a sample.

Post by systemerror909 October 31, 2007 (4 of 9)
zeus said:

You may want to read the FAQ (see the link above). DSD supports a bit depth up to 20-bit (120dB).

I have read the FAQ, and I know that DSD claims a theoretical dynamic range of 120db, I guess the question that I am asking is How? Given that this is a 1 bit system, the 64 bits that fall into 1/44100th of a second only add up to a total of 6 bits of range for that period. While SACD seems to be very accurate in the time domain (much faster sampling) it seems like it would be less accurate in terms of this dynamic range. It just seems to me that in order to be as accurate, a 1 bit system would require a much higher sampling rate. Does my argument make any sense?

Post by Sam October 31, 2007 (5 of 9)
systemerror909 said:

I have read the FAQ, and I know that DSD claims a theoretical dynamic range of 120db, I guess the question that I am asking is How?

The 120dB is only for the audio band (< 20Khz), not the entire 1.4Mhz bandwidth.

Post by zeus October 31, 2007 (6 of 9)
systemerror909 said:

I have read the FAQ, and I know that DSD claims a theoretical dynamic range of 120db, I guess the question that I am asking is How? Given that this is a 1 bit system, the 64 bits that fall into 1/44100th of a second only add up to a total of 6 bits of range for that period.

I've seen your mistake before. In the interval between redbook PCM samples 64 bits have passed. This gives you 2^64 permutations, not 2^6. You can't just think in terms of the bits getting aggregated as PCM samples. The reconstruction process is different (and also somewhat proprietary).

Post by mwagner1962 November 1, 2007 (7 of 9)
You know, I never understood the SCIENCE behind any of the audio recording stuff...period. Does this in ANY way affect how I love SACD?? Let's see...I now have about 615 SACDs and despite a few dogs (long since gone) and some less than stunning sounding SACDs (long since traded or sold), I can safely say that SACD sounds superb, and I do not have to dwell on the science behind what I am hearing.

Here is what is important for ME for any recording, either redbook OR SACD: does it sound realistic?? Does it sound natural??? Is there a sense of space and depth to the ensemble?? SACDs from Pentatone, Harmonia Mundi, Alia Vox, Caro Mitis and numerous others offer a rang and depth of sound I have never heard from most of my 1000+ redbook recordings, though in recent months I have bought some truly amazing redbook CDs (from Hyperion and Soli Deo Gloria as a few examples)...

If I see a potentially killer redbook disc (that I suspect will not come out on SACD) I will buy it and enjoy. If a recording is available as both a redbook AND an SACD, I will definitely buy the SACD.

In closing, you did not mention what types of music are on the SACDs you have already bought. If you love classical, then you will not go wrong with SACDs from Pentatone, Telarc (with Atlanta over Cincinnati), Harmonia Mundi, Caro Mitis, Channel Classics, Tudor, CPO and others...SO, just sit back and enjoy the music and take the science out of the equations...or, listen with your HEART instead of your HEAD

Cheers,

Post by systemerror909 November 1, 2007 (8 of 9)
zeus said:

I've seen your mistake before. In the interval between redbook PCM samples 64 bits have passed. This gives you 2^64 permutations, not 2^6. You can't just think in terms of the bits getting aggregated as PCM samples. The reconstruction process is different (and also somewhat proprietary).

Yes! This is what I am talking about and confused by. I've studied a good deal of mathematics/physics/optics/electronics in college and think I can understand most of the rational behind the pcm/dsd formats. What I was asking about specifically was the reconstruction process and how it is different. Maybe that is too complex of a question to be asking on these forums, but I figured someone might know.

Btw, my listening tastes include a wide range of music. Currently I have a wide range of SACDs ranging from classical, jazz, pop, rock. Yes, I hear the difference between sacd and redbook, I want to know how it works.

Post by zeus November 1, 2007 (9 of 9)
systemerror909 said:

Yes! This is what I am talking about and confused by. I've studied a good deal of mathematics/physics/optics/electronics in college and think I can understand most of the rational behind the pcm/dsd formats. What I was asking about specifically was the reconstruction process and how it is different. Maybe that is too complex of a question to be asking on these forums, but I figured someone might know.

My understanding is that DSD is a form of Pulse Density Modulation, where the density of the bits mirrors the waveform. More info here:

http://en.wikipedia.org/wiki/Pulse-density_modulation

Where DSD differs is in the n-order filtration. It's never specified though what the order is (maybe it varies from implementation to implementation). All this is beyond me anyway and I just trust my ears.

Closed