DSD comparison to PCM.

Lee

Active Member
Feb 4, 2011
1,402
0
36
Alpharetta, Georgia
Do not agree. I am with Anthony Cordesman on this issue. Read his article on the Burmester MC151 in The Absolute Sound.
FWW, my opinion is based on mic feeds that have been split to PCM and DSD converters and we compared the resulting recording to a live event of acoustic instruments. We have done this several times with chamber music and acoustic guitar. We also have tried over several top DACs.
 
May 30, 2010
13,912
5
38
Portugal
Just looked at the recording shown today in Stereophile (Fischer's New Mahler Symphony 3) https://channelclassics.nativedsd.com/albums/38817-mahler-symphony-no-3-in-dminor

Considering that the technical details shown in the booklet are just "dsd Super Audio/Grimm Audio Pyramix Editing/Merging Technologies" I would suppose that it was recorded in DSD and mixed/mastered in DXD, probably all other stereo formats coming from the DXD master. However, looking at the prices, I see DSD128 is more expensive than DXD. Usually sellers have an higher price for DXD, as it is the higher quality. Any special reason for this different price scheme?
 

Audiocrack

Active Member
Aug 10, 2012
1,902
0
36
Just looked at the recording shown today in Stereophile (Fischer's New Mahler Symphony 3) https://channelclassics.nativedsd.com/albums/38817-mahler-symphony-no-3-in-dminor

Considering that the technical details shown in the booklet are just "dsd Super Audio/Grimm Audio Pyramix Editing/Merging Technologies" I would suppose that it was recorded in DSD and mixed/mastered in DXD, probably all other stereo formats coming from the DXD master. However, looking at the prices, I see DSD128 is more expensive than DXD. Usually sellers have an higher price for DXD, as it is the higher quality. Any special reason for this different price scheme?
Have a look at the nativedsd website for the technical (recording) aspects of this recording. Jared explains on this site how this original dsd 64 recording was reworked in dxd and from there to dsd 128 and dsd 256. Have not bought this recording (yet) so do not know how the original dsd 64 recording compares to the other (higher) format files.

The Challenge Classics recordings made by Bert van der Wolf are original dxd recording but are also made available in various dsd formats. His dsd 256 files are (also) more expensive than his dxd files.
 

RWetmore

New Member
Apr 14, 2013
12
0
0
I haven't read the whole thread, but it seems to me it comes down to the fact that every converter, DSD or PCM, will exhibit some signature to the sound, good or bad, desirable or undesirable, likeable or unlikeable.

When DSD first came out there were problems with it, i.e. artificially 'softened' edges and dynamics (a type of distortion), but these problems were (to my ears at least) fully overcome by the Grimm AD1, which managed to achieve amazingly low distortion and great sound with their design. But there is still the problem with it only operating at 64x and the inherent HF noise above 50khz.

The recordings done in DXD (352.8khz/24bit PCM) by 2L and downconverted to DSD do not sound as good to me as ones made with the Grimm AD1. It's because the distortion isn't kept as low and is above audible threshold and I hear some 'softening' and 'rounding' of the edges, i.e. distortion. This seems to be a problem with higher sampling rates, both for PCM and DSD. That is, the faster the sampling, the more distortion and/or the harder it is to keep audible distortion at bay.

I know Jared Sacks has experimented with and tried other DSD converters that operate at 128x and 256x, but he says the Grimm still sounds the best (and I think he still goes out of his way to still use it most of the time). Eelco Grimm says he/they could make a DSD converter that does 128x DSD with equally low distortion as 64x on the AD1, but not one at 256x.

Maybe I'm biased since I'm partial to the Grimm and its ultra low distortion, but I would think a design of theirs operating at 128x would trounce everything else by a wide margin. After all, even 14 years after it's come out, the AD1 still arguably sounds the best. If only they would make an AD2 at 128x. Maybe someday they will.
 

RWetmore

New Member
Apr 14, 2013
12
0
0
Sure do...... Do tape archival all the time, plus have mastering projects in here.
Subjectively 'better' or more accurate to the source? It gives that same tight sound as the Grimm? I think Jared Sacks has tried the Horus and didn't think it was as good as the Grimm, which is why I'm surprised to hear this.
 
May 30, 2010
13,912
5
38
Portugal
Been there.... done that. Beat this dead horse enough. Had a Grimm for over a year and have done many DSD shoot-outs.

DSD Battle Royale!

To this day, I'm still using the Merging Horus with new A-D cards.
Nice to know - the Jordi Savall recordings I consider some of the best digital recordings I have ever listened, were carried by Manuel Mohino with Merging Horus.
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 26, 2010
6,554
1
38
Seattle, WA
www.pugetsoundstudios.com
Subjectively 'better' or more accurate to the source? It gives that same tight sound as the Grimm? I think Jared Sacks has tried the Horus and didn't think it was as good as the Grimm, which is why I'm surprised to hear this.
Switching between source and converter is seamless with the Horus. The Grimm and Pinot were "better" but not accurate to the source.
 

Audiocrack

Active Member
Aug 10, 2012
1,902
0
36
Nice to know - the Jordi Savall recordings I consider some of the best digital recordings I have ever listened, were carried by Manuel Mohino with Merging Horus.
With you Micro as regards the - generally speaking - high quality of the Alia Vox/Jordi Savall sacd’s. Any particular favourites among them?

However, so far I have only been able to find and download (low rate) pcm files of various Jordi Savall recordings which are in my opinion not great sounding. Do you have other experiences in this regard?
 
Dec 12, 2012
243
0
16
I apologize if the following points were already made somewhere upthread.

Probably, the two most salient performance parameters for any digital signal coding scheme are the Nyquist bandwidth (which, effectively sets the signal's high frequency limit for audio), and the native quantization noise level. DSD (sigma-delta modulation) maximizes Nyquist bandwidth, but sacrifices native quantization noise level to do so. PCM (pulse-code modulation) minimizes quantization noise level, but sacrifices Nyquist bandwidth to do so. DSD seeks to mitigate it's high quantization noise level via noise-shaping, which doesn't remove the high quantization noise, but rather, relocates it above the audio band. PCM mitigates it's lesser Nyquist bandwidth via brickwall bandlimiting of the signal, completely eliminating it as an issue, per pure sampling theory.

I suggest that differences in subjective sound between these schemes are likely due to the fundamental trade-off between Nyqust bandwidth and quantization noise level. More specifically, is due the implementation of their respective performance parameter mitigation technologies - brickwall bandlimiting for PCM, and noise-shaping for DSD - and how the human ear-brain system perceives the net technical performance result via music.
 
Last edited:

RWetmore

New Member
Apr 14, 2013
12
0
0
Switching between source and converter is seamless with the Horus. The Grimm and Pinot were "better" but not accurate to the source.
Hmmmm. I wonder what Jared Sacks is hearing then? What always stuck out for me is the Grimm's sound was tight and crisp, where as other converters sounded artificially 'soft' and 'gooey' by comparison. But of course I haven't actually heard the Horus. Is the Horus a chip or is it made with discrete parts like the Grimm?

I'll be interested to see if Grimm comes out with an AD2 that does 128x. 64x is quite limiting and is probably largely why SACD failed to develop into a much larger niche market.
 

RWetmore

New Member
Apr 14, 2013
12
0
0
I apologize if the following points were already made somewhere upthread.

Probably, the two most salient performance parameters for any digital signal coding scheme the Nyquist bandwidth (which, effectively sets the signal's high frequency limit for audio) and the native quantization noise level. DSD (sigma-delta modulation) maximizes Nyquist bandwidth, but sacrifices native quantization noise level to do so. PCM (pulse-code modulation) minimizes quantization noise level, but sacrifices Nyquist bandwidth. DSD seeks to mitigate it's high quantization noise level via noise-shaping, which doesn't remove the high quantization noise, but rather, relocates it above the audio band. PCM mitigates it's lesser Nyquist bandwidth via brickwall bandlimiting of the signal, completely eliminating it as an issue, per pure sampling theory.

I suggest that differences in subjective sound between these schemes are likely due to the fundamental trade-off between Nyqust bandwidth and quantization noise level. More specifically, is due the implementation of their respective performance parameter mitigation technologies - brickwall bandlimiting for PCM, and noise-shaping for DSD - and how those human ear-brain system perceives the net technical performance result via music.
Yes, but 352.8Khz/24bit PCM doesn't require brickwall bandlimiting, and would seem to be the optimal solution. The problem seems to be audible distortion at that high PCM sample rate. I at least hear it with the device Morton Lindberg is using to record at that sample rate.

This is why I like Grimm's philosophy, which is to make low distortion the design priority. Once you've reached a frequency response and dynamic range to capture the 'full sound', distortion is the enemy and going faster isn't going to be better if distortion can't be kept at bay.
 
Dec 12, 2012
243
0
16
Yes, but 352.8Khz/24bit PCM doesn't require brickwall bandlimiting, and would seem to be the optimal solution. The problem seems to be audible distortion at that high PCM sample rate. I at least hear it with the device Morton Lindberg is using to record at that sample rate.

This is why I like Grimm's philosophy, which is to make low distortion the design priority. Once you've reached a frequency response and dynamic range to capture the 'full sound', distortion is the enemy and going faster isn't going to be better if distortion can't be kept at bay.
However, there also is high rate DSD. The trade-offs are the same, relative to a given information rate. In other words, for a given data rate regardless whether PCM or DSD.

Distortion isn't a function of the coding scheme. It is a function of the A/D and D/A quantizer design located at either end of a given coding chain. For example, while 1-bit coding (two data points; 0 or 1) inherently has the highest quantization noise, it also has an inherently linear transfer function because two points can only define a straight line (as we know from baisc geometry). PCM, on the other hand, is an multibit coding scheme, and has great difficulty in delivering highly linear native operation because each of it's possible data points must lay exactly on the same line. A 16-bit PCM DAC, for example, has over 65,000 possible such data points.
 

RWetmore

New Member
Apr 14, 2013
12
0
0
However, there also is high rate DSD. The trade-offs are the same, relative to a given information rate. In other words, for a given data rate regardless whether PCM or DSD.
Well yes, but I think you have to go too high with PCM in order to not have to brickwall bandlimit, which is the seeming advantage of DSD (and perhaps why it seems to sound better). But distortion is big issue with DSD, even at 64x. There were papers written about this when DSD first came out, and I still hear it in essentially all of the DSD recorded SACDs I have, save for those done with the Grimm AD1. The distortion I hear is an artificial 'softening' or 'rounding' of the 'edges'. Somehow, Grimm was able to engineer this out of their converter, which at the time at least was a big breakthrough for DSD.

But yes, I know distortion is harder to keep a bay with higher sampling rates with DSD too. In fact, Eelco Grimm says they can't built a DSD converter with equally low distortion as the AD1 at 256x. He's pretty sure they could at 128x, but not 256x. But it seems to me that 128x should be enough since with similar noise shaping, even at 60Khz there is still about a -100 dB noise floor and even at 100Khz it's still about -70 dB. It's exceedingly rare for anything over 50khz to even make it to a microphone. So it seems 128x should give one the 'full sound' without compromise, provided distortion is kept below audible threshold.
 
Well yes, but I think you have to go too high with PCM in order to not have to brickwall bandlimit, which is the seeming advantage of DSD (and perhaps why it seems to sound better). But distortion is big issue with DSD, even at 64x.
That depends on the DAC. With my ODSX, you can select the 192 filter even when playing 44.1 Solves the digital filter issue nicely. The analog filters are all that is needed. They tell me that it's one of the best 44.1 DAC's. Won a shootout on Audiogon for just this reason.

IMO, it's very difficult to do a fair A/B of DSD and PCM because of the amount of logic and clocking that are different, even in the same playback device. It is fairly easy to do a good job on the DSD design and a lousy job on the PCM design, or visa-versa.

I suppose you could compare the best DSD DAC to the best PCM DAC, but you would have to have the lowest jitter source for both. Even then, are you really comparing DSD to PCM? These probably have different power subsystems, filters and output stages.... The result of these differences could be even larger than DSD versus PCM.

Steve N.
Empirical Audio
 

Members online