DSD playback via computer better than from SACD
- Thread starter jkeny
- Start date
Bruce B
WBF Founding Member, Pro Audio Production Member
We've known this for a few years.... We rip SACD's for HDTracks and can playback the SACD via the Playback Designs MPS-5 and the DSD file via Sonoma through the same DAC and they sound different.
OK perhaps someone can correct my deficiencies in my technical knowledge here. My understanding is - PCM arrives as a 16 bit "word" - the whole thing is loaded into the DAC and an output voltage is created. My understanding is that this is more susceptible to jitter because delays in the arrival of the bits can actually cause the frequencies to change. But for DSD - a one bit datastream simply says "louder than the last bit" or "softer than the last bit". It is essentially a 2.8MHz analog waveform, requiring only a low pass filter to output sound. Therefore, DSD should be less susceptible to jitter than PCM.
Is this completely wrong, or partially wrong, or correct? How much does jitter affect DSD?
Is this completely wrong, or partially wrong, or correct? How much does jitter affect DSD?
Keith, is DSD not 1-bit PCM sampled at 2.8224MHz & therefore inherently more susceptible to timing issues? Of course how data is handled, transmitted/received has a lot to do with the effects of jitter on the final sound.
Bruce, why do you think the two identically bit-perfect streams from different sources sound different?
Bruce, why do you think the two identically bit-perfect streams from different sources sound different?
Last edited:
Bruce B
WBF Founding Member, Pro Audio Production Member
It's actually a delta- sigma stream without the decimation. It uses pulse-density modulation encoding.
The only thing I can think of are jitter and filter differences.
Thanks Bruce. Actually looking at some interesting material about quantisation errors that probably have significantly more impact on DSD than on PCM. From the father of the field Bernard Widrow http://oldweb.mit.bme.hu/books/quantization/
It supports Opus111's contention of S-D having issues with quantiser noise modulation issues!
Last edited:
Let me see if I have this correct. You start with the SACD. Either play it via the transport of the Playback Designs MPS-5 or rip the DSD file from the SACD to a server and then play this through the DAC of the Playback Designs MPS-5. How do these two compare? Sorry for the stupid questions.
Bruce B
WBF Founding Member, Pro Audio Production Member
That is correct. Even captured the analog signal from both formats in hi-rez digital and did the null test. Not even close!
Bruce, you've done a null test - very interesting. Would you have any results from this? The difference file would be of great interest.
Could either one of you tell me what a null test is, assuming that it's not too technical. Is it like a null hypothesis (like a t-test) being tested?
It's a comparative test of the two different output streams - if they are both identical it should effectively null to zero i.e no difference.
In this case Bruce has tested the direct SACD playback & the same file copied to disk & played from disk through the same DAC. So where the file is being played back from is different but the bits are the same. His test showed that the analogue output from these two playback scenarios was very different. The "theory" says that the same bits should produce the same analogue waveform but it doesn't in this & other such cases!
That was the point in my starting the thread
!It's a comparative test of the two different output streams - if they are both identical it should effectively null to zero i.e no difference.
In this case Bruce has tested the direct SACD playback & the same file copied to disk & played from disk through the same DAC. So where the file is being played back from is different but the bits are the same. His test showed that the analogue output from these two playback scenarios was very different. The "theory" says that the same bits should produce the same analogue waveform but it doesn't in this & other such cases!
That was the point in my starting the thread
Thanks, John. I wonder if one sounded better than the other.
Is error correction the same for DSD and SACD bit streams?
p.s. The theory says "identical" only if the bits, timing, and filters (processing), and of course DAC and analog chain, are all identical, something not obvious to me.
p.s. The theory says "identical" only if the bits, timing, and filters (processing), and of course DAC and analog chain, are all identical, something not obvious to me.
Last edited:
Yes, the PC based one sounds better according to the Cookie Marenco in the link I gave & I think Bruce says the same!
Don, that's why I put "theory" in quotes
Some, (a lot) of people still state that bit-identical files are identical & any differences heard between two such files is self-delusion! I'm not sure what theory states that the files are only identical if bits, timing & filters (processing) are the same? I would add CM noise & possibly some other factors to this. But this isn't a theory as such, it's refinement of the original bit-identical statement based on experiential learning.That's why I'm interested in Bruce's Null test as it is difficult to do null tests accurately & have any proof that the same bits going into a DAC can produce different analogue out.
How many commercial recordings are available in DSD format? And from those, what percentage is pure DSD, never having been converted to PCM and back to DSD at some point?
There are probably an infinite number of other effects we could add that would make things non-identical. PSRR, CMRR, RFI/EMI, hysteresis, stray magnetic fields, phase noise, all the various jitter effects, anything and everything anyone can think up.
Bit-identical data at the DAC input should produce identical outputs, all else equal. I suspect the processing is different between DSD and SACD and that accounts for the difference. However, determining that all else is equal can be a non-trivial task. I would expect that reasonable shielding and power-supply decoupling would produce identical sonic results for two files IFF the actual bits delivered to the DAC are the same for both files. Maybe not if you live near a big transmitter, high-tension lines, or some such...
This is one of those debates that will likely go nowhere as nobody is likely to change their opinion based on whatever arguments (for either side) are brought forth.
Bit-identical data at the DAC input should produce identical outputs, all else equal. I suspect the processing is different between DSD and SACD and that accounts for the difference. However, determining that all else is equal can be a non-trivial task. I would expect that reasonable shielding and power-supply decoupling would produce identical sonic results for two files IFF the actual bits delivered to the DAC are the same for both files. Maybe not if you live near a big transmitter, high-tension lines, or some such...
This is one of those debates that will likely go nowhere as nobody is likely to change their opinion based on whatever arguments (for either side) are brought forth.
Yes, Don, in essence I agree!
But is DSD not the format that SACD uses & therefore it should encounter the same processing/filter as SACD within the same DAC? However, DSD is delivered via USB & SACD is delivered via ?
Anybody who claims (and there are quiet a few) that delivery of the correct bits is all that is important is blinkered & missing the bigger picture!
But is DSD not the format that SACD uses & therefore it should encounter the same processing/filter as SACD within the same DAC? However, DSD is delivered via USB & SACD is delivered via ?
Anybody who claims (and there are quiet a few) that delivery of the correct bits is all that is important is blinkered & missing the bigger picture!
Delivery where? I do not know if the processing DSD bits see is the same that SACD uses by the time it gets to the DAC; I would be surprised if the ECR and digital filtering is the same but do not know. On the output side, delivery of the correct bits does not mean the analog output is "good", but it should be the same if the input bits are the same. I think I am not following so by definition must be "blinkered".
I don't think I can contribute to this discussion but it is interesting to watch.
I don't think I can contribute to this discussion but it is interesting to watch.
 | Steve Williams Site Founder | Site Owner | Administrator |  | Ron Resnick Site Co-Owner | Administrator |  | Julian (The Fixer) Website Build | Marketing Managersing |