PCM to DSD...the new PS Audio DAC

DaveyF · Apr 17, 2014

Last night I attended my audio society meeting where we were pleased to greet Paul McGowan of PS audio fame. Paul was demo'ing his new DAC that up samples all PCM discs and data to DSD. Paul believes that DSD is a more musical sounding and analog like medium. He talked about how his new DAC can take a simple redbook CD and convert the PCM data to DSD. One of the things that Paul likes about DSD is that it does NOT truncate the signal vs. PCM. Also, he was stating that digital recording and playback has the advantage of much greater dynamic range than vinyl...I believe he said that vinyl has maximum dynamic range of 65db vs Digital has up to 90+ db.
We listened to his new DAC...priced at $6K and did a few AB's with PCM vs DSD and one track that was PCM upscaled to DSD and then re-scaled to PCM..which I felt was the best sound of the lot; although Paul did state that he felt the treble was slightly distorted leading to a slightly more "pleasing" sound....which was very possible.
After listening to the new PS audio DSD DAC, I think it is definitely a contender.
Question is if this is a new trend in DAC's that will take all standard redbook and upscale/convert to DSD.

BTW, for all those who have never met Paul, he is a super knowledgable guy with a great attitude. A real pleasure to have attended his meeting.

BlueFox · Apr 17, 2014

My Sony XA5400ES SACD player has been converting PCM to DSD for the last 5 years, and CDs sound great. Best $1200 I have spent. The drawback is it doesn't have a digital input.

Ken Newton · Apr 17, 2014

DaveyF said:
...Question is if this is a new trend in DAC's that will take all standard redbook and upscale/convert to DSD....

I do not readily see why that should be. Understand, that DSD and current production PCM application audio converters are all based on sigma-delta technology, with the following three primary differences between the two. Pure DSD converters have 1-bit of native resolution, while PCM application converters have 5-bits or 6-bits of native resolution. This means that PCM application converters have much higher native dynamic range than do DSD converters. The second main difference is that DSD utilizes much stronger noise-shaping to effectively increase audio band dynamic range than do the PCM application converters. DSD must include strong noise-shaping, simply because it's native resolution is so low. However, since both converters are sigma-delta based, both utilize some degree of noise-shaping. The third main difference is oversampling bandwidth, the spectrum of which, sigma-delta converters utilize to relocate the quantization noise stemming from their low native resolutions. DSD sigma-delta needs greater oversampling bandwidth than do PCM application sigma-delta converters, because DSD has much lower native resolution.

This situation is complicated, however, because DSD is necessarily recorded at a much higher native sample rate than PCM for the reasons discussed above. This fact has key implications for the signal reconstruction filter design, which, in my opinion, is what primarily gives DSD it's particular sound character. Taken in total, it's not readily apparent to me why converting PCM to analog via 1-bit sigma-delta (DSD) should sound better than converting PCM to analog via 5-bit sigma-delta. However, subjective performance is usually determined by the implementation details, and so, perhaps, an answer is yet to be found therein.

rbbert · Apr 17, 2014

I continue to be confused by the whole DSD/PCM issue(s); here's why

1) experts here and elsewhere can't seem to agree on whether or not DSD is only a type of PCM, or whether it's PWM, which might be a totally different animal or (again) only a type of PCM

2) the vast majority of current DAC chips use delta-sigma processing, which almost everyone seems to agree is DSD, so PCM is essentially converted to DSD by the DAC chip prior to DAC

3) If PCM and DSD can be interconverted so easily (and they obviously can, purely in the digital computer realm), why is one better than the other

4) Why does it make a difference (if it does) if the computer player (e.g. JRiver), the disc player (e.g. the Sony) or the DAC (PS Audio in the example above, but again according to most experts it occurs in any DAC using delta sigma processing) converts PCM to DSD

Ken Newton · Apr 17, 2014

rbbert said:
I continue to be confused by the whole DSD/PCM issue(s); here's why

1) experts here and elsewhere can't seem to agree on whether or not DSD is only a type of PCM, or whether it's PWM, which might be a totally different animal or (again) only a type of PCM

I'd argue that DSD is not PCM. Math operations, such as addition and multiplication, can be directly performed on PCM data, but not on noise-shaped DSD data, as far as I know.

2) the vast majority of current DAC chips use delta-sigma processing, which almost everyone seems to agree is DSD, so PCM is essentially converted to DSD by the DAC chip prior to DAC

No, not converted to DSD, which is defined as 1-bit sigma-delta, but converted to 5-bit or 6-bit sigma-delta. DSD is a form of sigma-delta processing, sigma-delta processing is not a form of DSD.

3) If PCM and DSD can be interconverted so easily (and they obviously can, purely in the digital computer realm), why is one better than the other

See the comments in my prior post, up above.

4) Why does it make a difference (if it does) if the computer player (e.g. JRiver), the disc player (e.g. the Sony) or the DAC (PS Audio in the example above, but again according to most experts it occurs in any DAC using delta sigma processing) converts PCM to DSD

I don't readily see that it should make a positive difference, at least, not from a technical standpoint. However, subjective perceptions aren't so easily quantified, so, perhaps, some secondary technical or implementation effect is positively affecting the resulting sound of such conversion for some listeners.

jap · Apr 17, 2014

DaveyF,

Were they using a preamp in the demo?

rbbert · Apr 17, 2014

Thank you for your response, Ken, although I must say it confirms my confusion more than clarifies!

Ken Newton · Apr 17, 2014

rbbert said:
Thank you for your response, Ken, although I must say it confirms my confusion more than clarifies!

Then, I've failed, miserably.

Bruce B · Apr 17, 2014

BlueFox said:
My Sony XA5400ES SACD player has been converting PCM to DSD for the last 5 years, and CDs sound great. .

As well as the Playback Designs and EMM Labs.

mep · Apr 17, 2014

Converting PCM to DSD is not going to make a recording sound like a native DSD recording whether that means a new recording that was made with a DSD recorder or a master tape that was converted to DSD.

rockitman · Apr 17, 2014

mep said:
Converting PCM to DSD is not going to make a recording sound like a native DSD recording whether that means a new recording that was made with a DSD recorder or a master tape that was converted to DSD.

I'm more interested in knowing what sounds better...Native PCM playback (16 or 24 bit) or those same files converted to DSD on the fly.

mep · Apr 17, 2014

rockitman said:
I'm more interested in knowing what sounds better...Native PCM playback (16 or 24 bit) or those same files converted to DSD on the fly.

That I couldn't tell you, but there are people here who can render their opinion.

rbbert · Apr 17, 2014

mep said:
Converting PCM to DSD is not going to make a recording sound like a native DSD recording whether that means a new recording that was made with a DSD recorder or a master tape that was converted to DSD.

How do you know this?

Bruce B · Apr 17, 2014

rbbert said:
How do you know this?

Because people record the same data stream in DSD and PCM, upsample the PCM file to DSD and you can hear differences in the 2 DSD files.

Elberoth · Apr 17, 2014

Ken Newton said:
I do not readily see why that should be. Understand, that DSD and current production PCM application audio converters are all based on sigma-delta technology, with the following three primary differences between the two. Pure DSD converters have 1-bit of native resolution, while PCM application converters have 5-bits or 6-bits of native resolution. This means that PCM application converters have much higher native dynamic range than do DSD converters. The second main difference is that DSD utilizes much stronger noise-shaping to effectively increase audio band dynamic range than do the PCM application converters. DSD must include strong noise-shaping, simply because it's native resolution is so low. However, since both converters are sigma-delta based, both utilize some degree of noise-shaping. The third main difference is oversampling bandwidth, the spectrum of which, sigma-delta converters utilize to relocate the quantization noise stemming from their low native resolutions. DSD sigma-delta needs greater oversampling bandwidth than do PCM application sigma-delta converters, because DSD has much lower native resolution.

This situation is complicated, however, because DSD is necessarily recorded at a much higher native sample rate than PCM for the reasons discussed above. This fact has key implications for the signal reconstruction filter design, which, in my opinion, is what primarily gives DSD it's particular sound character. Taken in total, it's not readily apparent to me why converting PCM to analog via 1-bit sigma-delta (DSD) should sound better than converting PCM to analog via 5-bit sigma-delta. However, subjective performance is usually determined by the implementation details, and so, perhaps, an answer is yet to be found therein.

As far as I understand the technology behind this DAC, the DAC doesn't use a regular SD DAC chipset to conver the data. Instead, it uses a simple lowpass filter.

So the basic block design would look like this:

INPUT RECEIVER -> FPGA (UPSAMPLING) -> LOW PASS FILTER

That is the reason all PCM data has to be upsampled to DSD.

This is not the first DAC using this concept. I believe the first one was Lampizator DSD.

Ken Newton · Apr 17, 2014

Elberoth said:
As far as I understand the technology behind this DAC, the DAC doesn't use a regular SD DAC chipset to conver the data. Instead, it uses a simple lowpass filter...

DSD makes use of noise-shaping, which functions within a sigma-delta processing engine. So, DSD is a form of sigma-delta conversion. After the sigma-delta processing engine, the DSD data eventually arrives at a 1-bit quantizer. This 1-bit quantizer is, in essence, a binary (2-state) driver or logic gate. The 2-state quantizer output could be directly sent to an analog low-pass filter to reconstruct the original signal, as you correctly understand. However, it should also be understood that exactly the same can be done with a PCM converter, the only distinction being the required sharpness of the output filter. The necessary filter sharpness is dependent on the native sample rate of the recorded digital signal.

rbbert · Apr 17, 2014

Bruce B said:
Because people record the same data stream in DSD and PCM, upsample the PCM file to DSD and you can hear differences in the 2 DSD files.

But again, how does anyone know whether this is inherent to the process or dependent entirely on the implementation? Grimm's own white paper says DSD is just a special form of 1-bit PCM, and also says 24/352.8 PCM is transparent to DSD. I know that may be nothing more than advertising copy or wishful thinking, but if so what is the point, especially since it's Grimm's DSD DAC?

DaveyF · Apr 17, 2014

jap said:
DaveyF,

Were they using a preamp in the demo?

That's a very good question, jap. In this instance there was a preamp utilized. According to Paul, the reason was due to the fact that the demo was using the Vandy 5's which have a built in sub amp. Therefore, Paul felt it better to use a preamp, than to drive the amp directly from the output of the DAC. The DAC can drive amps directly...Paul stated that DSD is basically music, wherein PCM is code. BTW, Vandy 5's are NOT among my favorite speaker's, although they did sound fairly good with the set-up and room that we listened in. The Vandy's do NOT have a great tweeter IMO- and with that they were not really able to reproduce the somewhat improved treble response that I have heard from DSD on other speakers.
Paul went over a lot of technical aspects of why the DSD upscaling was superior to the PCM...most of which went over my head. If the admin is ok with this, I will ask him to chime in here to elaborate on the technical aspects of his design.

Mike Lavigne · Apr 17, 2014

my good friend Ted Smith designed the new PS Audio DAC. Ted is one of the smartest people I know. congrats to Ted for the launch of this fine product.

he has had it in my room a number of times over the last five+ years during the development of it from before he hooked up with PS Audio up to just recently. a very nice sounding dac for sure. he used my system (and possibly others?) to check his progress as well as my Playbacks Design MPS-5 in my system as a target.

I'm a large believer in dsd as superior to PCM straight up, and prefer my PCM converted to dsd. my Playback Designs does have 2xdsd as it's native output.

ddk · Apr 18, 2014

This is more hocus pocus than anything else. The source is PCM, 16/24 or whatever, conversion to DSD, up sampling or any other form of manipulation cannot magically make the starting point more than what it is. At best the manipulation can give you a slightly different sound but generally I found the resulting sound worse. This kind of game has nothing to do with superiority or inferiority of either format, just a sleight of hand.
david

PCM to DSD...the new PS Audio DAC

Well-Known Member

Member Sponsor

Well-Known Member

Well-Known Member

Well-Known Member

Banned

Well-Known Member

Well-Known Member

WBF Founding Member, Pro Audio Production Member

Member Sponsor & WBF Founding Member

Member Sponsor

Member Sponsor & WBF Founding Member

Well-Known Member

WBF Founding Member, Pro Audio Production Member

Member Sponsor

Well-Known Member

Well-Known Member

Well-Known Member

Member Sponsor & WBF Founding Member

Well-Known Member

Similar threads