Is DSD superior or just the audio file du jour?

[Ken Newton]

Too many years knocking around this hobby has turned me from skeptic to cynic. Many of the people who seem to think DSD is the greatest thing since toilet paper on a roll are not people whose opinions I trust. But there are a few here whom I do. What say you?

Tim

DSD is a form of native 1-bit quantized sigma-delta modulation. There are a number of other forms of sigma-delta, which can vary in the number of native quantizer bits, sample rate and noise-shaping strength. In fact, nearly all D/A chips in current production are some form of hybrid multibit with sigma-delta modulation. While I'm not advocating DSD I can readily think of two objective technical advantages for it. One advanage is that 1-bit quantization is inherently linear, as two points can only define a stright line. The second advantage may be less recognized, which is that 1-bit converters don't exhibit the code dependent slew-rate distortion or the code dependent glitching exhibited by multibit converters. Of course, DSD has it's disadvantages as well, such as a very high amount of wideband noise and, if I correctly recall, greater jitter sensitivity.

One of things I find most interesting about DSD versus PCM is that the binary digital signal stream for DSD appears analog. By which I mean, it appears to change in proportion to the analog signal it represents when viewed on a scope. It resembles an FM signal, except that it's a pulse density type of modulation rather than a frequency modulation - however, the modulation of both appears signal proportional meaning, analog like. The binary digital signal stream of PCM on the other hand is a code modulation. It does not resemble the analog signal it represents. That said, I don't know that has any significance for the subjective sound. I just find it interesting to note.

 

[Vincent Kars]

Indeed the output of a DA converter in case of PCM is a stepped one.
You need a reconstruction filter to make it analog.
In case of DSD it is analog (continuous) but you do need a low pass filter to get rid of the quantization noise.

Anyway another definition of DSD:

 

[NorthSky]

Very good ↑ definition.

 

[Don Hills]

... In case of DSD it is analog (continuous) ...

Nope. It's a series of steps. Much bigger and more frequent steps than PCM steps too, so containing much more energy that has to be contained by filtering.

 

[Vincent Kars]

Hi Don
Care to explain it in more detail?

 

[Fitzcaraldo215]

Nope. It's a series of steps. Much bigger and more frequent steps than PCM steps too, so containing much more energy that has to be contained by filtering.

Don - I do not get that. But, I don't get Vincent's argument either. In neither case is the digital representation of discrete samples continuous, an essential property of digital audio. Yes, the processing necessary in either case to turn it into continuous analog differs substantially. DSD might be easier; PCM harder. But, if you look at the digital data stream in either case, it's composed of discontinuous bits. The analog output from the DAC has none. Both have employed some sort of filter to be able to get to that.

The argument that DSD is "closer to analog", therefore better is something I find meaningless.

 

[JoeWhip]

Personally, I want to get closer to reality, not "analog". I have great sounding PCM as well as DSD. The most important thing is the quality of the original engineering. I am a bit tired of the religious devotion to one side or the other on audio forums as to PCM, DSD and the like.

 

[Don Hills]

Hi Don
Care to explain it in more detail?

Let's see if I can do this without maths...
Start with two theoretical DACs (PCM and DSD) and measure them before any filtering. Also assume that they both can output the same maximum (peak) voltage, +- 1 volt. If you put an oscilloscope on the PCM DAC, you see the classic "stair steps". If you put it on the DSD DAC, you see an irregular pulse train. For the PCM DAC, the step heights (changes) from one sample to the next will vary from 0 to full output, changing at the sample rate. For the DSD DAC, each step will be maximum height, alternating between + and - 1 volt, and changing at up to 64 (128, 256 etc) times the "sample rate".

 

[dallasjustice]

It seems like upsampling to higher and higher MHz is becoming more popular. I think there's some merit to it.

To me, the real question is not whether one format is better than the other. The question is which format performs better with a particular DAC.

I believe most modern DACs are SDM DACs. Most DACs will upsample incoming PCM to a higher rate and then use various filters which eventually end up running in the MHz range. I believe the ESS chip runs at 40 MHz.

The DAC's upsampling filters may not be as good as filters which could be applied in a much more powerful server. DACs have limited computing power whereas PCs have much more capacity to use better filters.

I read somewhere a good explanation from Jussi about why folks are upsampling to DSD 256 and even DSD 512. It makes sense to me.

Whether it can be proven with DBT that upsampling produces a better result is unknown to me.

 

[NorthSky]

• http://www.mojo-audio.com/blog/dsd-vs-pcm-myth-vs-truth/

 

[fas42]

To me, the real question is not whether one format is better than the other. The question is which format performs better with a particular DAC.

... or system. I did some experiments a while back on a desktop PC, just using the low cost DAC on the motherboard - resampled to ludicrously high rates, talking here of, say, 768k PCM, etc. Yes, there was an impact, irrespective of how the player and OS software was handling the data - and that was that the sound improved with each step up in "speed" - absurd file sizes of course, over 1G, etc, so completely impractical ... but it "proved" a point, for that system at least ...

 

[Thomas savage]

Did not prove anything as purely relies on your subjective opinion.

I have up sampled things and found no such reliable conclusion. Indeed my subjective assessment led me to conclude it's likely you can go too far depending on original source.

 

[Sal1950]

... or system. I did some experiments a while back on a desktop PC, just using the low cost DAC on the motherboard - resampled to ludicrously high rates, talking here of, say, 768k PCM, etc. Yes, there was an impact, irrespective of how the player and OS software was handling the data - and that was that the sound improved with each step up in "speed" - absurd file sizes of course, over 1G, etc, so completely impractical ... but it "proved" a point, for that system at least ...

"Proved" as in there are documented results from bias controled blind listening tests?

Edit, Dang Thomas, you beat me to the draw.

 

[Thomas savage]

"Proved" as in there are documented results from bias controled blind listening tests?

Edit, Dang Thomas, you beat me to the draw.

Sure thing cowboy, that's a youthful itchy trigger finger for ya.

( that's youthful in the context of this forum)

 

[Sal1950]

Personally, I want to get closer to reality, not "analog". I have great sounding PCM as well as DSD. The most important thing is the quality of the original engineering. I am a bit tired of the religious devotion to one side or the other on audio forums as to PCM, DSD and the like.

Ah yes Joe, a pet peeve of mine.
Analog is not a ultimate sound quality goal, the Edison cylinder is the most pure analog. Sing into the horn with a blank wax cylinder then play it back. That is not the ultimate in reproduction though.

 

[Blumlein 88]

Personally, I want to get closer to reality, not "analog". I have great sounding PCM as well as DSD. The most important thing is the quality of the original engineering. I am a bit tired of the religious devotion to one side or the other on audio forums as to PCM, DSD and the like.

Dammit, if you aren't for religiously fanatic debate you have no business posting on a forum. Get with the times man.


Just kidding.

Doing a bit of recording, I am unable to find what PCM loses vs the mic feed. Haven't recorded in DSD, and have little interest in it.

I mean really , what would it take to get everyone to agree a format finally has exceeded human ears and is aurally perfect? At which point it becomes a non-issue.

 

[Sal1950]

I mean really , what would it take to get everyone to agree a format finally has exceeded human ears and is aurally perfect? At which point it becomes a non-issue.

Never
The believers argue over the sound of power cords for Gods sake. LOL

 

[Blumlein 88]

It seems like upsampling to higher and higher MHz is becoming more popular. I think there's some merit to it.

To me, the real question is not whether one format is better than the other. The question is which format performs better with a particular DAC.

I believe most modern DACs are SDM DACs. Most DACs will upsample incoming PCM to a higher rate and then use various filters which eventually end up running in the MHz range. I believe the ESS chip runs at 40 MHz.

The DAC's upsampling filters may not be as good as filters which could be applied in a much more powerful server. DACs have limited computing power whereas PCs have much more capacity to use better filters.

I read somewhere a good explanation from Jussi about why folks are upsampling to DSD 256 and even DSD 512. It makes sense to me.

Whether it can be proven with DBT that upsampling produces a better result is unknown to me.

I don't find myself able to hear differences in affordable digital. Is DSD at some high rate technically better? If so, it is more than enough. So how can it matter? Regular old PCM seems fully transparent at very affordable levels.

So if PCM is undetectable blind, and neither is DSD, why need we bother? 48/24 seems all we need. 96/24 if you simply for philosophical principle wish to leave out nothing the mics pick up. Few are the mics with response 96 khz will miss. I even have an Earthworks mic which would push that presumption a bit. Further, let us suppose 48/24 misses a bit. How much is missed? It is a damnably small amount. Not one which would make or break a recording. Lots of hyperbole in audiophile minds over what are mostly imaginary and at best infinitesimal real differences.

 

[Thomas savage]

Good sensible assertions..

 

[JoeWhip]

Dammit, if you aren't for religiously fanatic debate you have no business posting on a forum. Get with the times man.


Just kidding.

Doing a bit of recording, I am unable to find what PCM loses vs the mic feed. Haven't recorded in DSD, and have little interest in it.

I mean really , what would it take to get everyone to agree a format finally has exceeded human ears and is aurally perfect? At which point it becomes a non-issue.

The religious fervor is one of the reasons I don't post much anywhere anymore. It is insane. As for mic feeds, I have been in several recording sessions now and the 24/96 playback sounded identical to the live mic feed. Good enough for me.