PCM vs DSD. Why DSD and its variants?

FrantzM

Member Sponsor & WBF Founding Member
Apr 20, 2010
6,455
29
405
Bruce B has a thread on the Weiss converter ... I am of the opinion that most of what that SD carries over 20 KHz has a fair amount of noise. I really would like the experts here to shed some light for me on DSD vs PCM. I am not practitioner of Digital Signal Processing although that was one of my favorite subject in Grad School. I think I understand the basics. From my limited understanding I find the DSD process wasteful. I fail to see what it brings in term of advantages when high bit depth and high sampling digital are almost trivial. It has never been proven in a clear fashion that DSD is any superior to PCM ... So Why DSD at all?. Care has to be taken so serious are its artifacts and noise at higher frequency..Some audiophiles maintain that DSD is superior. This remain an opinion, one that I don't share. Any light please from the experts ? DSD vs PCM? Now that we have high bit depth and higher sampling rate available do we still need DSD or DXD? And what do we get recording in DSD and converting it to PCM ( I know that it can't be a perfect process) .. What is the rational behind recording in DSD or DXD or DXD2?
 

MylesBAstor

Well-Known Member
Apr 20, 2010
11,236
81
1,725
New York City
Bruce B has a thread on the Weiss converter ... I am of the opinion that most of what that SD carries over 20 KHz has a fair amount of noise. I really would like the experts here to shed some light for me on DSD vs PCM. I am not practitioner of Digital Signal Processing although that was one of my favorite subject in Grad School. I think I understand the basics. From my limited understanding I find the DSD process wasteful. I fail to see what it brings in term of advantages when high bit depth and high sampling digital are almost trivial. It has never been proven in a clear fashion that DSD is any superior to PCM ... So Why DSD at all?. Care has to be taken so serious are its artifacts and noise at higher frequency..Some audiophiles maintain that DSD is superior. This remain an opinion, one that I don't share. Any light please from the experts ? DSD vs PCM? Now that we have high bit depth and higher sampling rate available do we still need DSD or DXD? And what do we get recording in DSD and converting it to PCM ( I know that it can't be a perfect process) .. What is the rational behind recording in DSD or DXD or DXD2?

How about few people have actually heard "real" SACD since it is converted to PCM?
 

amirm

Banned
Apr 2, 2010
15,813
37
0
Seattle, WA
How about few people have actually heard "real" SACD since it is converted to PCM?
If you have at least the original SACD players they had DACs that handled the native stream. The notion of converting it to PCM and then playing it never made any sense to me especially if the content started as PCM.
 

amirm

Banned
Apr 2, 2010
15,813
37
0
Seattle, WA
Back to the original question. This is a bit like asking if the fridge light goes off when you close the door :).

SACD (DSD is a more generic mean and can imply other things) uses a single bit to represent your music. Obviously that is not enough at face value. The solution is to push the sampling rate way up to 2.8 Mhz (compared to 0.0441 Mhz for CD). That gives system lots of bandwidth. Then we take the horrendous "quantization noise" (error between the single bit and how much the music changed) and push that up into ultrasnoic area above 20 Khz.

The bit that is like the fridge analogy is that we ostensibly don't hear above 20 Khz. So how the heck do we know that pushing noise up there is good or bad. Or what shape we use to do that with. Do I put a ton at 40 and much less at 30 or equal amount for both? Do we do listening tests? We are not supposed to hear high frequencies but then, we try to listen for them in such a test? The solution I suspect is to wing it. Put a curve there and hope for the best.

And what is the true bandwidth of the system now? The player likely would want to filter out some of that ultrasonic range as to not cook your tweeter. In some sense, you can say the system has as much bandwidth as you like.

One thing you do have now is damn high sampling rate. Maybe that is the appeal.
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
If you have at least the original SACD players they had DACs that handled the native stream. The notion of converting it to PCM and then playing it never made any sense to me especially if the content started as PCM.

Only if the source started out PCM then it doesn't make sense to take a SACD and go back to PCM. But, you have to look at it this way. How many people can play a DSD file? Especially if the SACD is out of print.

The one advantage I see of DSD is in the image/spatial realm. I feel this is where DSD excels. Yes, I know there is UHF noise above 22k.. but PCM has it as well if you do tape archival.

Recording in DSD128fs is better by pushing that UHF noise out farther from the audible bandwidth. Plus, you don't need any more space than a 24/96 file when recording at DSD64fs.

IMO, the only PCM equivilant to DSD is at least 352.8k and above, and on "some" material, that's still not good enough.
 

FrantzM

Member Sponsor & WBF Founding Member
Apr 20, 2010
6,455
29
405
Bruce

I am somewhat lost here. IT is my understanding that you can't intervene on DSD with DSP. How is it mastered then? How do you EQ, limit and the other stuff? Doesn't mastering anything DSD requires a conversion to PCM? Or am I mistaken ... Even on SACD aren't we eventually listening to a double conversion Step 1 DSD recording
Step 2 DSD to PCM conversion for mastering
Step 3 PCM to DSD to SACD) ???

I am utterly confused. Seems to me there is no pure SACD/DSD since the mastering step seems to require PCM conversion a process which by the very nature of DSD is lossy ... Illuminate me please

Thanks in advance

Frantz
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
Short answer:

When I get a DSD file, and someone wants me to master their material, I stream the DSD file out the DAC and into the analog console to apply any processing in the analog domain and go back in to the ADC capturing in DSD. No PCM conversion.

Long answer:

O'kay.. to continue. If one wanted to just use in the box (ITB) plugins, yes, there is some "conversion". Now let me explain. If you were to make an edit, say take out a tick/pop, just those few samples would be "calculated" as though it were PCM (32/352.8kHz). The DSD file would remain intact. Another workstation that we use adds 8 bits (DSD-wide) to make the calculation. Just like a fade in/out, those few seconds would be calculated as though it were PCM, but it would not change the format of the original file.
Unless you're doing the same process (ie: comp, EQ, limit) on the entire DSD file, the file stays in DSD format.

This is what I understand to be what's going on underneath the hood and could be completely in left field.
 
Last edited:

FrantzM

Member Sponsor & WBF Founding Member
Apr 20, 2010
6,455
29
405
cjfrbw

Thanks .. a few things though. It is not so clear-cut that DSD preserves the pulse response of analog. One of the things Philips and its ally Sony presented as proof of the supeiority of DSD, was a relatively nice square wave at 10 KHz, something Redbook can't because of its response above 22.1 KHz is zero. Subsequent independent anlysis anlysis showed the impossibility of such "clean" square waves considering the noise that is pushed upward of 20 Khz .. In this regard 24/96 is superior.

I'll wait for Bruce answer in the meantime I remain extremely skeptical about DSD claims to superiority , my ears and my understanding of the process don't tell me so...
 

FrantzM

Member Sponsor & WBF Founding Member
Apr 20, 2010
6,455
29
405
Hey! What happened to Carl's post?
 

Mike Lavigne

Member Sponsor & WBF Founding Member
Apr 25, 2010
12,474
11,369
4,410
i will offer my WAG 'wild ass guess' as to one technical reason why DSD sounds better than PCM (besides sampling rate advantage possibilities).

i think it has partially to do with the much simpler math involved with the conversion of the analog wave form to dsd digital and back again, verses the math involved to convert analog to PCM and back again. more of the substance of music is lost in the conversion with PCM. once changed to PCM there is something missing.

maybe at some extreme frequency and bit depth PCM might be equal, but i've not heard it (even at 24/384). i've heard master tapes compared regular dsd copies and to 2xdsd copies of them. i've heard Lps compared to regular dsd and 2xdsd copies of them....as well as high sampling rate, high word length PCM copies of both. dramatically different results in favor of dsd and 2xdsd. music loses body, depth, and space with PCM.....compared to master tape or vinyl.

maybe high level vinyl and master tape is not a worthy reference.

i've also sat in Bruce's studio and heard an analog signal run thru converters at many different formats. same result.

i know Frantz has read me post these ideas multiple times and still seriously doubts my perspectives. i respect he will have to hear it for himself before he changes his mind.

soon Playback Designs will have players and DAC's which will be able to play dsd files, 2xdsd files and 24/384 files over USB. maybe then, when more people are exposed to this stuff some of the skepticism will diminish.
 

MylesBAstor

Well-Known Member
Apr 20, 2010
11,236
81
1,725
New York City
i will offer my WAG 'wild ass guess' as to one technical reason why DSD sounds better than PCM (besides sampling rate advantage possibilities).

i think it has partially to do with the much simpler math involved with the conversion of the analog wave form to dsd digital and back again, verses the math involved to convert analog to PCM and back again. more of the substance of music is lost in the conversion with PCM. once changed to PCM there is something missing.

maybe at some extreme frequency and bit depth PCM might be equal, but i've not heard it (even at 24/384). i've heard master tapes compared regular dsd copies and to 2xdsd copies of them. i've heard Lps compared to regular dsd and 2xdsd copies of them....as well as high sampling rate, high word length PCM copies of both. dramatically different results in favor of dsd and 2xdsd. music loses body, depth, and space with PCM.....compared to master tape or vinyl.

maybe high level vinyl and master tape is not a worthy reference.

i've also sat in Bruce's studio and heard an analog signal run thru converters at many different formats. same result.

i know Frantz has read me post these ideas multiple times and still seriously doubts my perspectives. i respect he will have to hear it for himself before he changes his mind.

soon Playback Designs will have players and DAC's which will be able to play dsd files, 2xdsd files and 24/384 files over USB. maybe then, when more people are exposed to this stuff some of the skepticism will diminish.

I am not worthy :)

I totally agree with all you've said. Even the 24/384 copy of a master tape is not a "mirror image." Good but no cigar.
 

amirm

Banned
Apr 2, 2010
15,813
37
0
Seattle, WA
i will offer my WAG 'wild ass guess' as to one technical reason why DSD sounds better than PCM (besides sampling rate advantage possibilities).

i think it has partially to do with the much simpler math involved with the conversion of the analog wave form to dsd digital and back again, verses the math involved to convert analog to PCM and back again. more of the substance of music is lost in the conversion with PCM. once changed to PCM there is something missing.
Technically the reverse is true Mike :). Noise shaping requires math with noise shaping and such. Think of SACD as a PCM with one bit. Indeed, the history of it came from sigma delta PCM converters.

I think the reason SACD sounds better is a mystery. Technically, it has no business sounding better. But I have personally done blind testing of it against content that started as PCM and still liked the SACD version! I have no engineering explanation for what was second generation copy and hence, should have been worse, not better.
 

FrantzM

Member Sponsor & WBF Founding Member
Apr 20, 2010
6,455
29
405
MikeL

I respect your points of view much more than you give me credit for. Yours are of a true music lover and serious, purist audiophiles. I would like to consider myself to be both too.
I however do not subscribe to the trust my ears mantra.. My ears and my senses in general helped by my brain fool me too often. I need a solid ground on which to base my beliefs. In matter of religion only do I go on the "simply believe" and I am not a religious person...
It's been said that SACD is superior to CD , yet many if not most SACD are mastered with a conversion to PCM then reconverted to DSD/SACD .. Even the most mathematics-challenged person would tell me that this can't be a good thing ... you go from something to PCM and back to that something a process that is inherently lossy. I also would add that the maths involved in SACD are not trivial either.. The sampling rates is high because of the bit depth which is only 1 bit, so to compensate you take more samples with each sample less accurate than the equivalent in PCM.. You make up with sheer numbers... PCM is getting there in the meantime 384 is out already...
Now I cna understand one preferring a medium to another... I take the example of a person preferring Tang to Orange juice often to which Audiophiles answer which is the Tang? To ne that's not the point, people like what they like It could be there there are differences between SACD and Redbook CD... The problem that I have had in comparing the two was that the machine on which the comparison were made were of a different level of excellence. I have compared my Burmester DAC with a top of the line Sony machine and I liked the CD better .. might have been different with a DSD DAC of the Burmester caliber ... I know however controlled experience using a protocol in which I believe and you may not (more or less blind) showed the difference to vanish at 24/96 for DSD. I don't know if it was DSD64 or 128. And it should be so . CD can't reproduce a 10 KHz but 24/96 is flat to almost 50 KHz and we can debate all we want of the audibility of UHF but there are very few speakers that get in this region .. The better ones crap out at 40 Khz .. so the 100 Khz bandwidth of SACD ... I doubt we can hear that high but that is an aside.
A mistake not to be made is that PCM shops the signals and therefore it no longer resembles the original .. This is due to the graphics generally used to describe a purely mathematical process... If the signal is band limited the maths say that the signal can be recovered perfectly if you sample at twice the highest frequency in the signal. Physics say we are not there yet but we have made progress... considerable progress and that progress is not DSD IMHO. I could be wrong.

While I can understand someone preferring Analog to digital, I almost find it almost amusing for anyone to qualify digital as "unlistenable". The only qualifier for such statement would be an hyperbole, to make a point .. This might have been true 25 years ago.. The earth has circled the sun a few times in the meantime.. This is no longer the case
 

Mike Lavigne

Member Sponsor & WBF Founding Member
Apr 25, 2010
12,474
11,369
4,410
MikeL

I respect your points of view much more than you give me credit for. Yours are of a true music lover and serious, purist audiophiles. I would like to consider myself to be both too.
I however do not subscribe to the trust my ears mantra.. My ears and my senses in general helped by my brain fool me too often. I need a solid ground on which to base my beliefs. In matter of religion only do I go on the "simply believe" and I am not a religious person...
It's been said that SACD is superior to CD , yet many if not most SACD are mastered with a conversion to PCM then reconverted to DSD/SACD .. Even the most mathematics-challenged person would tell me that this can't be a good thing ... you go from something to PCM and back to that something a process that is inherently lossy. I also would add that the maths involved in SACD are not trivial either.. The sampling rates is high because of the bit depth which is only 1 bit, so to compensate you take more samples with each sample less accurate than the equivalent in PCM.. You make up with sheer numbers... PCM is getting there in the meantime 384 is out already...
Now I cna understand one preferring a medium to another... I take the example of a person preferring Tang to Orange juice often to which Audiophiles answer which is the Tang? To ne that's not the point, people like what they like It could be there there are differences between SACD and Redbook CD... The problem that I have had in comparing the two was that the machine on which the comparison were made were of a different level of excellence. I have compared my Burmester DAC with a top of the line Sony machine and I liked the CD better .. might have been different with a DSD DAC of the Burmester caliber ... I know however controlled experience using a protocol in which I believe and you may not (more or less blind) showed the difference to vanish at 24/96 for DSD. I don't know if it was DSD64 or 128. And it should be so . CD can't reproduce a 10 KHz but 24/96 is flat to almost 50 KHz and we can debate all we want of the audibility of UHF but there are very few speakers that get in this region .. The better ones crap out at 40 Khz .. so the 100 Khz bandwidth of SACD ... I doubt we can hear that high but that is an aside.
A mistake not to be made is that PCM shops the signals and therefore it no longer resembles the original .. This is due to the graphics generally used to describe a purely mathematical process... If the signal is band limited the maths say that the signal can be recovered perfectly if you sample at twice the highest frequency in the signal. Physics say we are not there yet but we have made progress... considerable progress and that progress is not DSD IMHO. I could be wrong.

While I can understand someone preferring Analog to digital, I almost find it almost amusing for anyone to qualify digital as "unlistenable". The only qualifier for such statement would be an hyperbole, to make a point .. This might have been true 25 years ago.. The earth has circled the sun a few times in the meantime.. This is no longer the case

Frantz,

thanks, i know we respect each other's perspectives.

to be clear, i don't recall referring to PCM as a format as unlistenable ever. in fact, i likely listen to it 25% to 50% of the time and enjoy it. sometimes i'll listen only to my music server for a week at a time. the Playback Design MPS-5 has great PCM at all levels of sample rate.

i respect that many excellent systems are PCM sourced exclusively and i don't consider those systems invalid because of that.

it's only when comparing it to higher level analog vinyl and RTR tape, or SACD and dsd done properly, where PCM falls short.

i think it's a mistake to use the multiple ways SACD's might be mastered as evidence that SACD or DSD is flawed. of the first 10,000 CD's ever mastered, how many sounded like crap? most simply due to poor sloppy mastering. Bruce finds redbook sourced SACD's regularly. you'll note my specific comments refer mostly to dsd where i personally know about the source used.

i recommend considering one of the new Playback Designs players/dacs with the 'super' hirez USB capability. they are reasonably priced, and you can get those files with info on the source and compare things for yourself.

i agree that in the short term trusting your ears can be less than reliable. however, over time, when you can listen in many states of mind and awareness the truth will 'out'. your body does not lie to you while your mind is fickle. music is one of those things to be enjoyed and savored, not dissected and proved.

and yet if it's important to you to get that proof then that's the way you gotta play it.

take care,

Mike
 

DaveyF

Well-Known Member
Jul 31, 2010
6,129
181
458
La Jolla, Calif USA
Frantz,


i agree that in the short term trusting your ears can be less than reliable. however, over time, when you can listen in many states of mind and awareness the truth will 'out'. your body does not lie to you while your mind is fickle. music is one of those things to be enjoyed and savored, not dissected and proved.



Mike

That is so very true Mike. In the pro audio world, you rarely if ever, see anyone dissect or try and prove what they are hearing....everyone's too busy enjoying the music. This would be a very good thing if it could permeate to the high-end audio world.
 

Vincent Kars

WBF Technical Expert: Computer Audio
Jul 1, 2010
860
1
0
I must admit I do think DSD is a kind of paradox.
By design it has a tremendous bandwidth up to 100 kHz so you can record without using a low pass filter. Might be beneficial.
However from 22 kHz on the noise becomes very strong, any possible music signal will probably be drowned in this noise.


DSD quantisation noise with a sampling rate of 64 times 44.06 kHz.
Malcolm Hawksford, Essex University, UK
Source: https://www.hdtracks.com/files/DSD_to_LPCM.pdf
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
I must admit I do think DSD is a kind of paradox.
By design it has a tremendous bandwidth up to 100 kHz so you can record without using a low pass filter. Might be beneficial.
However from 22 kHz on the noise becomes very strong, any possible music signal will probably be drowned in this noise.


DSD quantisation noise with a sampling rate of 64 times 44.06 kHz.
Malcolm Hawksford, Essex University, UK
Source: https://www.hdtracks.com/files/DSD_to_LPCM.pdf

I have to take exception to this graph. The black line representing "CD-audio" is not correct. CD has a noise floor of -96dB.. not -144dB.

Also this is being resampled to Redbook. Going to a higher resolution, the noise is not near that much, as you can see from the graphs that I've posted elsewhere.
 

garylkoh

WBF Technical Expert (Speakers & Audio Equipment)
Sep 6, 2010
5,599
225
1,190
Seattle, WA
www.genesisloudspeakers.com
I have to take exception to this graph. The black line representing "CD-audio" is not correct. CD has a noise floor of -96dB.. not -144dB.

Also this is being resampled to Redbook. Going to a higher resolution, the noise is not near that much, as you can see from the graphs that I've posted elsewhere.

The graph says 20bit/44.1kHz....
 

About us

  • What’s Best Forum is THE forum for high end audio, product reviews, advice and sharing experiences on the best of everything else. This is THE place where audiophiles and audio companies discuss vintage, contemporary and new audio products, music servers, music streamers, computer audio, digital-to-analog converters, turntables, phono stages, cartridges, reel-to-reel tape machines, speakers, headphones and tube and solid-state amplification. Founded in 2010 What’s Best Forum invites intelligent and courteous people of all interests and backgrounds to describe and discuss the best of everything. From beginners to life-long hobbyists to industry professionals, we enjoy learning about new things and meeting new people, and participating in spirited debates.

Quick Navigation

User Menu

Steve Williams
Site Founder | Site Owner | Administrator
Ron Resnick
Site Co-Owner | Administrator
Julian (The Fixer)
Website Build | Marketing Managersing