Login / Register
 
DSD vs PCM
New Reply
Subscribe
blissmusic
Thread Starter
#1
3rd December 2005
Old 3rd December 2005
  #1
Gear interested
 
Joined: Sep 2005
Location: New York City
Posts: 24

Thread Starter
blissmusic is offline
DSD vs PCM

Is there any advantage to DSD technology over PCM based systems?

What are the pro's and cons particularly in terms of sound quality in each?
#2
3rd December 2005
Old 3rd December 2005
  #2
Gear Guru
 
Sounds Great's Avatar
 
Joined: Jan 2005
Location: Twin Cities, MN
Posts: 16,239

Sounds Great is offline
Quote:
Originally Posted by blissmusic
Is there any advantage to DSD technology over PCM based systems?

What are the pro's and cons particularly in terms of sound quality in each?

DSD sounds way better to me. Many people agree. Of course y.m.m.v.
__________________
-Rob

Ice in her eyes, frozen tears could never be a surprise
You can't erase a dream you can only wake me up
#3
3rd December 2005
Old 3rd December 2005
  #3
Lives for gear
 
Curve Dominant's Avatar
 
Joined: Oct 2002
Location: Control Room
Posts: 1,958

Curve Dominant is offline
Quote:
Originally Posted by blissmusic
Is there any advantage to DSD technology over PCM based systems?
First and foremost, you have to understand that DSD is in fact a form of PCM.

DSD is simply 1-bit PCM.

Quote:
What are the pro's and cons particularly in terms of sound quality in each?
DSD is 1-bit PCM.

Other forms of PCM include 24-bit PCM.

Surely, you can deduce the advantages between 1-bit recording and 24-bit recording.

It's that simple, really.
__________________
Eric Vincent
http://www.studioericvincent.com
blissmusic
Thread Starter
#4
3rd December 2005
Old 3rd December 2005
  #4
Gear interested
 
Joined: Sep 2005
Location: New York City
Posts: 24

Thread Starter
blissmusic is offline
Quote:
Originally Posted by Curve Dominant
First and foremost, you have to understand that DSD is in fact a form of PCM.

DSD is simply 1-bit PCM.

Other forms of PCM include 24-bit PCM.

Surely, you can deduce the advantages between 1-bit recording and 24-bit recording.

It's that simple, really.
Never thought of it like that.

DSD has a high sampling rate (2.87 MHz) at 1-bit which I am aware of.

Wait a second . . . DSD is simply 1 bit PCM? I thought there was a completely different architecture from PCM to a delta-sigma modulation scheme such as DSD? Am I right?

Tony
#5
3rd December 2005
Old 3rd December 2005
  #5
3 + infractions, forum membership suspended.
 
Joined: Jul 2003
Posts: 1,716

jbuntz is offline
I dont think DSD classifies as pulse code modulation for the simple reason that the 1 or 0 represents the differential (or delta sigma) and not the individual voltage values for every sample. 1's and 0's as PCM would sound massively distorted and would have a 6 dB per octave rise to the top end.
#6
3rd December 2005
Old 3rd December 2005
  #6
Gear Guru
 
Sounds Great's Avatar
 
Joined: Jan 2005
Location: Twin Cities, MN
Posts: 16,239

Sounds Great is offline
DSD is not PCM.
#7
3rd December 2005
Old 3rd December 2005
  #7
Lives for gear
 
djui5's Avatar
 
Joined: Mar 2004
Location: Phoenix
Posts: 6,661

Send a message via Yahoo to djui5
djui5 is offline
Quote:
Originally Posted by blissmusic
Wait a second . . . DSD is simply 1 bit PCM? I thought there was a completely different architecture from PCM to a delta-sigma modulation scheme such as DSD? Am I right?

Tony

No, it's a new concept entirely.

PCM places the amplitude on a + and - plane at every sample, DSD moves the amplitude either up or down at every sample (hence the single bit). PCM uses it's bit depth to determine how many "points of reference" it has for the + and - plane.


See more info here..

http://www.superaudio-cd.com/technol...plain_english/

http://www.sonydadc.com/products.disk.super.go

http://www.daisy-laser.com/technolog...techsacd11.htm

http://www.scena.org/lsm/sm7-3/dvd-en.html

http://en.wikipedia.org/wiki/Direct_Stream_Digital

http://www.merging.com/2002/html/sacd.htm

http://answers.google.com/answers/threadview?id=5

http://www.sony.net/SonyInfo/News/Press/199904/99-042/

http://www.sweetwater.com/expert-center/glossary/t--DSD


There used to be this nifty picture on Sony's site that explained a lot about it...but I can't seem to find it now.
__________________
_________________

"What is a crossfire hurricane & why wasn't I born in one?"

Randy Wright
#8
3rd December 2005
Old 3rd December 2005
  #8
Lives for gear
 
Joined: Oct 2003
Posts: 521

genericperson is offline
It's different enough to be called different.

I think DSD sounds noticeably better than PCM. The limitations are in processing with effects, multi-track mixing, etc. That seems to be evolving and working its way into more users hands, just like the early days of digital PCM traveled from the elite to the middle class over time.

I'd give it 7 years, and it will be a new tracking and mixing standard for working professionals.
Quote
1
#9
3rd December 2005
Old 3rd December 2005
  #9
Lives for gear
 
nobtwiddler's Avatar
 
Joined: May 2005
Location: Millbrook, NY
Posts: 2,301

nobtwiddler is offline
Genex on it's way here.

I also agree that DSD sounds better!
A few limitations with this format a t the moment, editing, for one.
But I'll give the format a real test next week when I take delivery of my new Genex GX9000 DSD machine.
It's been a long wait, I was waiting for the remote to be built...but now it's ready and I should have it in a few days.

Paul
www.millbrooksoundstudios.com
blissmusic
Thread Starter
#10
3rd December 2005
Old 3rd December 2005
  #10
Gear interested
 
Joined: Sep 2005
Location: New York City
Posts: 24

Thread Starter
blissmusic is offline
Quote:
Originally Posted by djui5
No, it's a new concept entirely.

PCM places the amplitude on a + and - plane at every sample, DSD moves the amplitude either up or down at every sample (hence the single bit). PCM uses it's bit depth to determine how many "points of reference" it has for the + and - plane.

There used to be this nifty picture on Sony's site that explained a lot about it...but I can't seem to find it now.
What do you mean by + and - ???? I'm very confused.

Is this the pic you were talking about?



is there a differnce between what is represented in this pic and pulse WIDTH modulation? what does the width represent in the 1 position in pulse width modulation?

I am assuming of course that DSD is VERY different from pulse width mod., and of course I ask to satisfy my curiosity.

Thanks!

Tony
#11
3rd December 2005
Old 3rd December 2005
  #11
Gear maniac
 
Mrs. Fairman's Avatar
 
Joined: Apr 2004
Location: Germany
Posts: 185

Mrs. Fairman is offline
Quote:
Originally Posted by blissmusic
Is there any advantage to DSD technology over PCM based systems?

What are the pro's and cons particularly in terms of sound quality in each?
Microphone - microphone amplifiers - (spitten) 1 XLR to 2 XLR -

A. XLR 1 in the best PCM Lavry gold AD
B. XLR 2 into the best (admitted) DSD AD

A. play by Lavry gold DA
B. play by DSD DA

A/B ears comparison test. I believe Lavry gold am better.
__________________
D.T.Jambor
www.truetrackrec.de
#12
3rd December 2005
Old 3rd December 2005
  #12
Lives for gear
 
Joined: May 2004
Location: Denver, CO
Posts: 820

2pulse is online now
Conversion artifacts are the achilles heel of PCM. PCM in its purest form has the capacity to sound better than DSD, but unless you're playing back on ridiculously priced audiophile-grade equipment, DSD will almost certainly sound better.

For this reason I hope to see sacd gain some ground in consumer audio. You can buy a player for $100 at Best Buy that sounds better than $5k dvda players.
#13
4th December 2005
Old 4th December 2005
  #13
Lives for gear
 
djui5's Avatar
 
Joined: Mar 2004
Location: Phoenix
Posts: 6,661

Send a message via Yahoo to djui5
djui5 is offline
Quote:
Originally Posted by blissmusic
What do you mean by + and - ???? I'm very confused.

Is this the pic you were talking about?


No, thats not the picture. When a converter changes from analog to digital, it takes a picture of the waveform and places it on a positive/negative plane. All waveforms travel up and down from positive to negative amplitude.

I'm attaching a homemade drawing.

With PCM, at each sample point the converters will put the waveform at a calculated place on the plane (seen in the drawing-8 positive, and 8 negative. The middle is the 0 crossing point). How many places it has to put the waveform depends on the bit depth, so with 24 bits, you get 2 to the 24th power...which turns out to a shitload of places to put this waveform in either the positive or negative amplitude. This will draw a pretty good representation of the waveform at 44.1 samples, and an even more accurate one at 88.2/96K.

With DSD, at each sample point the converters place the waveform either higher or lower in amplitude, going up or down. That's why it only requires a single bit, because at each sample point the volume is going either up or down. When you combine this with a 2.8 Mhz sample rate, you get one hell of a picture of the waveform.


This is "generally" how it goes. Make sense?
Attached Thumbnails
DSD vs PCM-pcm-graph.jpg  
Quote
1
#14
4th December 2005
Old 4th December 2005
  #14
Lives for gear
 
Joined: May 2004
Location: Denver, CO
Posts: 820

2pulse is online now
Or think of it like this: 1's indicate a decreasing waveform slope and 0's indicate an increasing waveform slope. 1010101010 would produce a constant slope (a line), whose direction is dictated by what bits have already been processed. If it were an upward excursion, it it can be bent down with a series of bits that included more 1's than 0's. Solid 1's would point the waveform straight down.
#15
4th December 2005
Old 4th December 2005
  #15
Lives for gear
 
djui5's Avatar
 
Joined: Mar 2004
Location: Phoenix
Posts: 6,661

Send a message via Yahoo to djui5
djui5 is offline
Quote:
Originally Posted by 2pulse
Or think of it like this: 1's indicate a decreasing waveform slope and 0's indicate an increasing waveform slope. 1010101010 would produce a constant slope (a line), whose direction is dictated by what bits have already been processed. If it were an upward excursion, it it can be bent down with a series of bits that included more 1's than 0's. Solid 1's would point the waveform straight down.


Thanks, that's a good description.
#16
4th December 2005
Old 4th December 2005
  #16
Lives for gear
 
Curve Dominant's Avatar
 
Joined: Oct 2002
Location: Control Room
Posts: 1,958

Curve Dominant is offline
Quote:
Originally Posted by blissmusic
DSD is simply 1 bit PCM? I thought there was a completely different architecture from PCM to a delta-sigma modulation scheme such as DSD? Am I right?
No, not quite. Not even close, really.

In the early days of digital audio (like, over a decade ago), ALL of our A-D conversion was done with 1-bit delta-sigma modulation at 64x (2.8224MHz) conversion. That info was then decimated to 44.1K, and converted to 16bit (then later, 20bit, and 24bit as the benefits of higher bitrates became obvious in the processing stages).

Right there, you see there is nothing "new" about DSD-stylie conversion. 1-bit delta-sigma modulation at 2.8224MHz has been used from the very beginning of digital audio. And, yes, it was PCM then, and it still is PCM. OK...

Later in the evolution of digital audio, convertor design engineers got hip to the fact that 1-bit conversion cannot be dithered, and that means noise, lots of it in the high freqs, which made all those peeps who claimed early digital audio sounded harsh and nasty - well, they were right. It did sound kinda rough because you cannot dither a 1-bit signal...and so where does all that quantisation noise go?? Well, you have to try to push it up out of the audible band, and they tried but that sh*t got reflected back into the audible band in the form of 3rd-order harmonic distortion. And that distortion cannot be removed once it's recorded, at all by any means...and that distortion GOT there because the conversion was at 1-bit (which cannot be dithered). Following me so far? Good...

SO, the convertor designers decided to re-design their convertors as multi-bit on the front-end of the digital stage of the convertor chip. Which meant they no longer had to deal with all that noise which is generated by an un-dithered A-D convertor (such as the ones used in DSD - which is still using the OLD OUTDATED convertor design). Convertor chip manufacturers such as Burr Brown, Crystal and AKM all switched over to this NEW multi-bit format. And digital audio began to sound much better as a result.

Oh, and before I forget: The NEW multi-bit convertors were ALSO delta-sigma modulation designs, just like the OLD 1-bit convertors were. Right? STILL delta-sigma, STILL PCM. Same as it ever was - as far as all THAT goes. Still with me? OK...

So now you know:
1) DSD is not new
2) DSD is PCM
3) DSD is not "different" in any sense, except...

DSD does not re-convert to multi-bit (uless it's "DSD Wide" in which case it's multi-bit and now you're back to the whole PCM thing...ARGH!), or down-sample from the 2.8224MHz sampling rate (and who the hell can hear at that freq - not to mention that freq will literally burn your tweeters - ARGH!).

Not to mention the fact that one cannot process (EQ, compress, edit, etc.) a 1-bit DSD signal (unless you convert it to multi-bit, and why not just do that at the recording stage anyway?), so forget about all that.

Happy hunting!
#17
4th December 2005
Old 4th December 2005
  #17
Lives for gear
 
Curve Dominant's Avatar
 
Joined: Oct 2002
Location: Control Room
Posts: 1,958

Curve Dominant is offline
Quote:
Originally Posted by jbuntz
I dont think DSD classifies as pulse code modulation for the simple reason that the 1 or 0 represents the differential (or delta sigma) and not the individual voltage values for every sample. 1's and 0's as PCM would sound massively distorted and would have a 6 dB per octave rise to the top end.
That second sentence is essentially correct. The resulting distortion is inherent in 1-bit conversion, as 6dB is indeed the headroom of a 1-bit signal.

The first sentence doesn't make as much sense. How is a 1-bit delta-sigma modulated A-D converted signal not PCM?
#18
4th December 2005
Old 4th December 2005
  #18
Lives for gear
 
Curve Dominant's Avatar
 
Joined: Oct 2002
Location: Control Room
Posts: 1,958

Curve Dominant is offline
Quote:
Originally Posted by djui5
No, it's a new concept entirely.

PCM places the amplitude on a + and - plane at every sample, DSD moves the amplitude either up or down at every sample (hence the single bit). PCM uses it's bit depth to determine how many "points of reference" it has for the + and - plane.
Randy,

Basically, what you've said there, is that PCM is 6, and DSD is a half-dozen.

Which kinda-sorta helps us all understand the "difference" between the "two."

The difference being there are two terms which equate to the same effect.

Let's all try to remember what PCM stands for:

Pulse Code Modulation

1) The pulse represents the sample of audio taken
2) the code represents a value given to that sample
3) modulation represents that sample's relative value to samples before and after it in the time sequence

Modulation can manifest in an X-Y axis, or an X-Y-Z axis.

Either way, both are PCM.
#19
4th December 2005
Old 4th December 2005
  #19
Head of Bumping Security (B.S)
 
Joined: Feb 2004
Location: in the hills of Southern California
Posts: 3,223

jdunn is offline
Quote:
Originally Posted by Curve Dominant
Oh, and before I forget: The NEW multi-bit convertors were ALSO delta-sigma modulation designs, just like the OLD 1-bit convertors were. Right? STILL delta-sigma, STILL PCM. Same as it ever was - as far as all THAT goes. Still with me? OK...
Actually some of the very best sounding D/A designs these days do not use the "delta-sigma" technique, but the "cosign-modulator" technique, such as the Burr Brown PCM 1704 and others.

Don't bother trying to challenge me on the technical details of this, as I really don't know.

I do know that DSD does basically use a PCM converter, but in a way such that it eliminates several of the filters. This is likely one of the reasons that many people prefer DSD to PCM.
#20
4th December 2005
Old 4th December 2005
  #20
Lives for gear
 
Curve Dominant's Avatar
 
Joined: Oct 2002
Location: Control Room
Posts: 1,958

Curve Dominant is offline
Quote:
Originally Posted by jdunn
Actually some of the very best sounding D/A designs these days do not use the "delta-sigma" technique, but the "cosign-modulator" technique, such as the Burr Brown PCM 1704 and others.
I believe the subject of DSD and PCM referred to recording, in which case we're talking about A-D.
#21
4th December 2005
Old 4th December 2005
  #21
Head of Bumping Security (B.S)
 
Joined: Feb 2004
Location: in the hills of Southern California
Posts: 3,223

jdunn is offline
Quote:
Originally Posted by blissmusic
Is there any advantage to DSD technology over PCM based systems?

What are the pro's and cons particularly in terms of sound quality in each?
Here's the original post. It doesn't seem that specific to me. D/A is part of the system. You must be able to reproduce the waveform, not just record it.

No big deal. I was just pointing out that delta-sigma is not the only technique of conversion currently in use. You are correct though, in that I'm unaware of any A/D that uses cosign-modulation.
#22
4th December 2005
Old 4th December 2005
  #22
Lives for gear
 
Curve Dominant's Avatar
 
Joined: Oct 2002
Location: Control Room
Posts: 1,958

Curve Dominant is offline
Quote:
Originally Posted by jdunn
I do know that DSD does basically use a PCM converter, but in a way such that it eliminates several of the filters. This is likely one of the reasons that many people prefer DSD to PCM.
There are filters used in DSD conversion. Have no doubt about that.

They are not the SAME filters used in multi-bit conversion. So Sony is therefore allowed to say "DSD does not use the filters that are used in PCM." And they do say that. That's very clever of them, don't you think?

I'll give the DSD/SACD marketing folks this: They made a very admirable attempt to market a doomed format. A lot of audio engineers (some of them very reputable) drank the DSD Koolaid.

But as far as "many people prefer DSD to PCM"?? Besides a few audio forum posters, I don't know anyone who knows much if anything about DSD, let alone prefers it. The consumer market came down in favor of MP3. I personally demo'd DSD in an acoustically sealed room with $3K speakers, and didn't think it sounded very good. It sounded like 1-bit conversion, basically.
#23
4th December 2005
Old 4th December 2005
  #23
Lives for gear
 
Curve Dominant's Avatar
 
Joined: Oct 2002
Location: Control Room
Posts: 1,958

Curve Dominant is offline
Quote:
Originally Posted by jdunn
D/A is part of the system. You must be able to reproduce the waveform, not just record it.
I'm not sure how that makes any sense in regards to this discussion.

DSD/SACD systems in fact do not require D-A conversion. Since you seem to know something about D-A conversion, I'm sure you can figure out why that is.
#24
4th December 2005
Old 4th December 2005
  #24
Gear addict
 
Joined: Aug 2002
Location: Cambridge, Massachusetts, USA
Posts: 364

Send a message via AIM to jajjguy
jajjguy is offline
Curve, I really don't follow your explanation of the mechanics of DSD. (Or did you not explain that?) Are you saying that djui and 2pulse have incorrectly described the difference, or are you just adding more background and nuance to the description?

Quote:
I personally demo'd DSD in an acoustically sealed room with $3K speakers, and didn't think it sounded very good.
That's the part of what you said that I understand. Thanks.

Quote:
It sounded like 1-bit conversion, basically.
Whoops, lost me again. Are you saying it sounded worse than an old Speak'n'Spell, which I believe was 8 bit? Or the godawful recordings that my digital camera makes, which are 12 bit?
__________________
cellist, recordist, acoustics geek
Dos
#25
4th December 2005
Old 4th December 2005
  #25
Dos
Gear nut
 
Dos's Avatar
 
Joined: Apr 2005
Posts: 98

Dos is offline
Quote:
Originally Posted by Curve Dominant
or down-sample from the 2.8224MHz sampling rate (and who the hell can hear at that freq - not to mention that freq will literally burn your tweeters - ARGH!).
It doesn't actually produce this freq!

PCM is not DSD, your posts on this topic have been erratic and misinformative.

Going back to the original post...

The Pros are, DSD is much higher quality, whether you can hear it or not.
That's pretty much the only pro!! (not a bad one tho)

The cons are, the fastest G5 can only handle something like (correct me if i'm wrong) 10 tracks of DSD audio!! and thats with no processing.
You'll also need to change all your PCM based software. 2 pretty big cons if you sak me.

It will be a while before people convert to DSD but as we all know 'a while' in the digital world could be pretty soon.
Quote
1
#26
4th December 2005
Old 4th December 2005
  #26
Lives for gear
 
djui5's Avatar
 
Joined: Mar 2004
Location: Phoenix
Posts: 6,661

Send a message via Yahoo to djui5
djui5 is offline
Quote:
Originally Posted by Curve Dominant
down-sample from the 2.8224MHz sampling rate (and who the hell can hear at that freq - not to mention that freq will literally burn your tweeters - ARGH!).

What the hell are you talking about?

Earth to Curve:


SAMPLE RATE IS IRRIVELENT TO THE FREQS BEING REPRODUCED, with the expection of the Nyquest theory which states that the sample rate should be twice the highest freq being reproduced/recorded. How do you figure that DSD reproduces freq's of 2.8Mhz? Do your guitars squeel like that?



Holy crap, I can't believe you said that.

Also, saying DSD is the same as PCM is like saying a Pinto is the same as a Ferrari because they both have wheels and an engine, and go forward.
#27
4th December 2005
Old 4th December 2005
  #27
Lives for gear
 
Curve Dominant's Avatar
 
Joined: Oct 2002
Location: Control Room
Posts: 1,958

Curve Dominant is offline
Quote:
Originally Posted by djui5
What the hell are you talking about?

Earth to Curve:


SAMPLE RATE HAS NOTHING TO DO WITH THE FREQS BEING REPRODUCED.
That's right. Because they are FILTERED.

Remember Randy? There was a point in the discussion where FILTERS came up?

Earth to Randy? Hello?

Let's everybody at least try to act like we're all engineers, and actually read the thread before we post.

Thanks!
#28
4th December 2005
Old 4th December 2005
  #28
Lives for gear
 
Curve Dominant's Avatar
 
Joined: Oct 2002
Location: Control Room
Posts: 1,958

Curve Dominant is offline
Gee, Randy, you sure do edit your posts quickly.

IRRIVELENT - hmm, I'm going to jump to the conclusiion that you meant to type the word irrelevant.

And I believe it's Nyquist, not Nyquest.
#29
4th December 2005
Old 4th December 2005
  #29
Lives for gear
 
djui5's Avatar
 
Joined: Mar 2004
Location: Phoenix
Posts: 6,661

Send a message via Yahoo to djui5
djui5 is offline
Quote:
Originally Posted by Curve Dominant
Gee, Randy, you sure do edit your posts quickly.

IRRIVELENT - hmm, I'm going to jump to the conclusiion that you meant to type the word irrelevant.

And I believe it's Nyquist, not Nyquest.


Yeah, my spelling sucks, I know.


And I did edit my post. I was trying to say that sample rate has nothing to do with the freq's being reproduced, and I didn't want you saying something smart like "well what about the nyquist theory?", so I was trying to word things carefully. I know how you like to take things out of context.


What do filters have to do with it anyway? Are you trying to tell me now that DSD does reproduce freq's as high as 2.8Mhz?

Do you understand what sample rate really means Curve?
#30
4th December 2005
Old 4th December 2005
  #30
Lives for gear
 
Curve Dominant's Avatar
 
Joined: Oct 2002
Location: Control Room
Posts: 1,958

Curve Dominant is offline
Quote:
Originally Posted by djui5
And where do you propose those high freq's come from Curve? Are you trying to tell me that because you have a sampling rate of 2.8Mhz, that is creates freq's that high?
I'm sorry, Randy, you're going to have to edit or elaborate on these questions so that they make sense, and then I'll see if I can respond.
New Reply Submit Thread to Facebook Facebook  Submit Thread to Twitter Twitter  Submit Thread to LinkedIn LinkedIn  Submit Thread to Google+ Google+ 
 
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Similar Threads
Thread
Thread Starter / Forum
Replies
soloan / Remote Possibilities in Acoustic Music & Location Recording
63
88fingerz / So much gear, so little time!
13
Inky Goddess / Remote Possibilities in Acoustic Music & Location Recording
52
nOiz / So much gear, so little time!
1

Forum Jump

SEO by vBSEO ©2011, Crawlability, Inc.