dsd DSD Offshoot Discussion From MQA Topic

[Paul R]

1 hour ago, mansr said:

What makes you believe this? Do you have a good reference? Audio salesmen are not good references. Alternatively, you may explain your position, derived using maths from some starting point we both agree is valid.

Sure - try AES Convention Paper 5396, it is an overview at a basic level , and old enough to have been suitably refuted if there was much of anything incorrect in it.  There are a bunch papers that get significantly deeper, but this one is a good overview. 

 

 

 

1 hour ago, mansr said:

 

The entire book is the reference. If there were any relation to PDM, you'd think the authors would mention it at least once. They do not.

 

I call BS on this. If you read the book, you would be able to pinpoint a reference that defends your point, or if it exists, refutes mine. I am afraid I don't believe you are enough of an authority on the subject to be able to waste my time that way. 

 

Jussi probably is that much of an authority. He pointed to a process control document that is actually applicable, but does not, so far as I saw when reading it, contain a direct comparison of what DSD  - a 1 bit per word bitstream format, is when compared to multi-bit data representation like PCM. 

 

 

 

 

[Paul R]

3 minutes ago, mansr said:

My guess is it's a sales tactic. It enables the usual follow-on claim that it is somehow or somewhat analogue in nature and thus escapes the evils of digital.

 

I will requite exactly what I said. 

 

DSD is the result of modulating a free running clock signal with an analog signal so that the output is a bit stream whose density is relatively proportional to the input analog signal. 

 

The bitstream represents the data is]n a PDM format, the density or frequency of the pulses is what is used to represent the analog signal. Not the value of each pulse, as in PCM.  If you don't understand that, then nobody can help you. But I sure hope your work on MQA is nowhere near as sloppy as your thinking here. 

 

 

[fas42]

For interest's sake I tried putting a WAV track through a PCM -> DSD64 -> PCM conversion path ... hmmm, quite major differences at only 60dB or so down; so, are the 'distortions' intrinsic to that conversion process, or is the software - I used XiSRC - not good enough? Worth trying some variations to that processing chain ...

[Samuel T Cogley]

15 minutes ago, Paul R said:

I call BS on this. If you read the book, you would be able to pinpoint a reference that defends your point, or if it exists, refutes mine. I am afraid I don't believe you are enough of an authority on the subject to be able to waste my time that way.

 

Hi Paul

 

Isn't the answer here for you to read the book and then bring what you learned from that to the forum?  It seems like you're looking for some shortcut to the knowledge.  Why not go straight to the source?  Will it take some time?  No doubt.  But there really isn't any other way for you to gain the knowledge you lack here.  If you're looking for donations to finance the book purchase, I'll gladly chip in $5.  You'll learn something you don't currently know and you'll (hopefully) bring that knowledge to the forum.  Win/Win, no?

[mansr]

9 minutes ago, Paul R said:

I call BS on this. If you read the book, you would be able to pinpoint a reference that defends your point, or if it exists, refutes mine. I am afraid I don't believe you are enough of an authority on the subject to be able to waste my time that way. 

The trouble here is that you are asking me to prove that DSD isn't something or other. Why should the authors of the book bother to mention that DSD isn't PDM when it also isn't PWM, PSK, or a thousand other things? The fact that a 700-page book on the subject not even once (I searched the PDF) mentions either "PDM" or "pulse density" is, to me, a pretty strong indication that there is no meaningful relationship to be found. I also find this obvious for mathematical reasons, but you seem to require the word of an "authority."

 

12 minutes ago, Paul R said:

DSD is the result of modulating a free running clock signal with an analog signal so that the output is a bit stream whose density is relatively proportional to the input analog signal.

No, that is not how a DSD signal is created.

 

14 minutes ago, Paul R said:

The bitstream represents the data in a PDM format, the density or frequency of the pulses is what is used to represent the analog signal. Not the value of each pulse, as in PCM.  If you don't understand that, then nobody can help you.

Still wrong. DSD is a series of equidistant pulses, each with a value of +1 or -1. It has nothing to do with PDM. Some graphical representations of a DSD signal kind of give that impression, which is unfortunate since it isn't an accurate way of thinking about it.

[danadam]

2 hours ago, Paul R said:

Great. How? I just wind up with SMGL entities if I try typing it, rather than copying it. 

On linux that would be entering "u" + unicode codepoint in hex while holding down Ctrl+Shift. As for windows, I don't know, but google shows some results for windows enter unicode, e.g.: How to enter Unicode characters in Microsoft Windows

 

The unicode codepoints in hex are:

• 03a3 for Σ
• 0394 for Δ

[Shadders]

Hi,

Is it possible that one person is discussing how a bit stream signal is created, and another person is discussing how that bit stream is transformed into a sound signal (analogue) ???

 

Regards,

Shadders.

[mansr]

1 minute ago, danadam said:

On linux that would be entering "u" + unicode codepoint in hex while holding down Ctrl+Shift.

Or map a spare key to dead_greek.

[mansr]

2 minutes ago, Shadders said:

Is it possible that one person is discussing how a bit stream signal is created, and another person is discussing how that bit stream is transformed into a sound signal (analogue) ???

No, that doesn't help. Paul is still wrong.

[danadam]

12 minutes ago, mansr said:

DSD is a series of equidistant pulses, each with a value of +1 or -1. It has nothing to do with PDM. Some graphical representations of a DSD signal kind of give that impression, which is unfortunate since it isn't an accurate way of thinking about it.

So nothing new in business 🙂 (vide stair step graphs for showing how high res is supposedly better)

[Ralf11]

8 minutes ago, mansr said:

No, that doesn't help. Paul is still wrong.

 

It is your fault for not explaining a complex topic so an 18 yr old can understand it.  Care to try quantum gravity instead?

[fas42]

The basic process of DSD is trivial - it's the noise shaping where the real meat is, and that's buried in maths normally. Visually, it makes sense why it's needed - with the right "pictures" - something to track down.

 

If one can't get an instinctive understanding of what's happening, then maths does absolutely nothing to clarify the matter.

[Paul R]

19 minutes ago, Samuel T Cogley said:

 

Hi Paul

 

Isn't the answer here for you to read the book and then bring what you learned from that to the forum?  It seems like you're looking for some shortcut to the knowledge.  Why not go straight to the source?  Will it take some time?  No doubt.  But there really isn't any other way for you to gain the knowledge you lack here.  If you're looking for donations to finance the book purchase, I'll gladly chip in $5.  You'll learn something you don't currently know and you'll (hopefully) bring that knowledge to the forum.  Win/Win, no?

 

No. 

 

To start with, I don't lack knowledge in this subject. Not that I can not always learn something new, but in this case, it appears to be someone else trying to prove their technical superiority rather than challenging a technical issue. Moreover, I do not require anyone to agree with me.  

 

The burden here is on me only if the wording I originally used was not precise enough.  I have not seen any real critique of that wording. I think people are just being lazy, jumping to conclusions, and missing what I actually said.

 

DSD is the result of modulating a free running clock signal with an analog signal so that the output is a bit stream whose density is relatively proportional to the input analog signal.

 

That is a fairly precise statement, and I do not see that anyone has successfully challenged it.  True you can DoP the DSD data into a Flac file, but if you try to play it through a PCM DAC that accepts FLAC files, you are in for an awful sounding listening session. Another clue is that DSD is a very different data format is that it has such a high sampling rate. Even the most ardent Shannon fan would have to question a sample rate in the megahertz range for PCM data. 

 

 

[Paul R]

35 minutes ago, mansr said:

No, that doesn't help. Paul is still wrong.

 

Well, you just need to admit you are flat wrong and move on with it. 

35 minutes ago, mansr said:

No, that doesn't help. Paul is still wrong.

 

Actually, you are the one who is wrong here.  When you figure that out and are ready to actually talk facts, look me up. 

 

This is a useless conversation because you goal is not to uncover and discuss facts. I look forward to the day you can tell me how much sample information is intrinsically held in a single DSD sample.  

 

[Samuel T Cogley]

3 minutes ago, Paul R said:

To start with, I don't lack knowledge in this subject. Not that I can not always learn something new, but in this case, it appears to be someone else trying to prove their technical superiority rather than challenging a technical issue. Moreover, I do not require anyone to agree with me. 

 

Ok.  Just trying to understand why you're repeating the same thing over and over again and demanding others prove you wrong.  I was just thinking that bringing in citations from places other than your own brain might help bolster your case.

[Paul R]

3 minutes ago, Samuel T Cogley said:

 

Ok.  Just trying to understand why you're repeating the same thing over and over again and demanding others prove you wrong.  I was just thinking that bringing in citations from places other than your own brain might help bolster your case.

 

Seriously, I did. 

 

I also don't consider tossing out a large book that doesn't clearly define the difference under discussion as a "reference."  I would class that as BS myself.  If it was a reference, then chapter, page, and brief reference might be a better choice for someone to offer up. 

 

 

YMMV. 

 

-Paul 

 

[mansr]

1 hour ago, Paul R said:

Sure - try AES Convention Paper 5396, it is an overview at a basic level , and old enough to have been suitably refuted if there was much of anything incorrect in it.

I found a free PDF of that paper. It actually agrees with me. Witness (emphasis mine):

"One-bit modulators (SDM's) differ from PCM AD convertors in the sense that the quantizer covers 2 levels only, and that the quantization error is fed back to the input. As a result, the output power of a SDM is always constant; its signal level is either +1 or -1, giving a signal power of 1."

[Paul R]

57 minutes ago, danadam said:

On linux that would be entering "u" + unicode codepoint in hex while holding down Ctrl+Shift. As for windows, I don't know, but google shows some results for windows enter unicode, e.g.: How to enter Unicode characters in Microsoft Windows

 

The unicode codepoints in hex are:

• 03a3 for Σ
• 0394 for Δ

 

Thank you. 

[mansr]

18 minutes ago, Paul R said:

Moreover, I do not require anyone to agree with me.

That's certainly helpful if you're going to insist on being wrong.

[Samuel T Cogley]

5 minutes ago, Paul R said:

 

Seriously, I did. 

 

I also don't consider tossing out a large book that doesn't clearly define the difference under discussion as a "reference."  I would class that as BS myself.  If it was a reference, then chapter, page, and brief reference might be a better choice for someone to offer up.

 

 

Well, I just ordered the book because I want to know.  The abstract for that book discusses more applications for Delta-Sigma Data Converters than just audio, so while I'm expecting it to be quite dry, it should be at least illuminating.

[fas42]

Just to add some more thoughts on the matter, here is a page by AP, who a few people here tend to think are OK, .

 

https://www.ap.com/technical-library/more-about-pdm/

[crenca]

10 minutes ago, mansr said:

I found a free PDF of that paper. It actually agrees with me. Witness (emphasis mine):

"One-bit modulators (SDM's) differ from PCM AD convertors in the sense that the quantizer covers 2 levels only, and that the quantization error is fed back to the input. As a result, the output power of a SDM is always constant; its signal level is either +1 or -1, giving a signal power of 1."

 

Just to clarify (or perhaps muddy the waters further)  that last sentence means that the density (the D in PDM) does not actually change but rather remains constant right?

[Paul R]

10 minutes ago, mansr said:

I found a free PDF of that paper. It actually agrees with me. Witness (emphasis mine):

"One-bit modulators (SDM's) differ from PCM AD convertors in the sense that the quantizer covers 2 levels only, and that the quantization error is fed back to the input. As a result, the output power of a SDM is always constant; its signal level is either +1 or -1, giving a signal power of 1."

 

Well Duh- what is your point?  The first line clearly says that SDMs and PCM ADs are different? Yep, DSD is definitely not the same format as PCM... I mean, how much clearer can that be? If you are just arguing that no digital data can be a true PDM signal, I'll agree with that. The signal is encoded  in PDM format however, which is what I said in the first place...

 

Quote

One-bit 􏰮􏰯 modulators (SDM’s) differ from PCM AD convertors in the sense that the quantizer covers 2 levels only, and that the quanti- zation error is fed back to the input. As a result, the output power of a SDM is always constant; its signal level is either +1 or -1, giving a signal power of 1. As known from PCM theory, dither is necessary to remove artifacts due the non-linear behaviour of such a device. The fact that the quantizer only spans two levels obviously removes the possibil- ity to introduce the standard PCM dither [14]: dither with a triangularly shaped probability distribution function, with a width of 1 LSB, which clearly spans 3 levels. These two observations indicate that dithering a SDM is not identical to dithering a multi-bit quantizer. As dithering is a very important concept in linearizing both PCM and SDM quan- tizers, we will devote a section to this issue. In this paper, we will not address any theoretical issues, such as covered in [9], where it is shown that dithering higher than first-order sigma delta modulators is viable, but approach the subject from a more pragmatic point of view and show the results of various simulations. All simulations have been performed with a 7’th order SDM, designed according to the general rules sketched in [1]. The design is based on a Butterworth design for NT F(z), with a corner frequency of 70 kHz. By adding feedback over 2 integrators, a resonator section is created which allows to shape the noise in the baseband according to the general shape of the human hearing threshold curve. 

 

 

[mansr]

7 minutes ago, crenca said:

Just to clarify (or perhaps muddy the waters further)  that last sentence means that the density (the D in PDM) does not actually change but rather remains constant right?

That's one way of looking at it.

[phosphorein]

22 minutes ago, Paul R said:

 

Well Duh- what is your point?  The first line clearly says that SDMs and PCM ADs are different? Yep, DSD is definitely not the same format as PCM... I mean, how much clearer can that be? If you are just arguing that no digital data can be a true PDM signal, I'll agree with that. The signal is encoded  in PDM format however, which is what I said in the first place...

 

 

 

 

I read this paper just now and I have no idea how you come to this conclusion based on what is presented. Please show us ( and don't repeat the same sentence again ).