Mark Waldrep and More Silliness...

[Miska]

Either you're terrible at expressing yourself, or you're deliberately cryptic to the point of being intentionally misleading.

 

Probably both, but I cannot know how experienced you are on electronics side. I'm more like an electronics guy by training who ended up doing software.

[Jud]

It has to, or you'd get aliasing. It works exactly the same as any other sample rate conversion.

 

Now you are getting to the point...

 

There is no single translation, as long as you have to bitstream you have option of taking any of the translations.

 

So would it be correct to conclude from these statements that one might do much the same when converting, for example, DSD128 to DSD64, but the result of the sub-sampling would be a DSD rather than a PCM representation of the signal?

[mansr]

Probably both, but I cannot know how experienced you are on electronics side. I'm more like an electronics guy by training who ended up doing software.

I'm experienced enough to know that a digital pin can only take on one value at a time.

[jabbr]

I'm experienced enough to know that a digital pin can only take on one value at a time.

 

Well there is no law that requires DACs to have a single input pin per channel, you generally also have a clock. or take the I2S protocol which transmits multibit data on a single line. If you look at the DSC1 design, the conversion from single bit to multibit is entirely specified (shift register ladder).

[Miska]

So would it be correct to conclude from these statements that one might do much the same when converting, for example, DSD128 to DSD64, but the result of the sub-sampling would be a DSD rather than a PCM representation of the signal?

 

For DSD128 to DSD64 allows using methods that cannot be used for PCM rates, because the conversion ratio is small and the sampling rate distance to audio signals is so high and there is no need to gain lot of bits for expressing dynamic range. Also the ultrasonic noise level vs frequency is low compared to what will be gained at lower sampling rate so there's plenty of headroom. So the process, instead of reducing noise by filtering is adding relatively more noise... (practically shifting the noise down by an octave)

[mansr]

Well there is no law that requires DACs to have a single input pin per channel, you generally also have a clock. or take the I2S protocol which transmits multibit data on a single line.

 

There is no law, but actual DSD DAC chips have one data pin per channel and one clock pin. The I2S interface only handles PCM data at lower rates. Thus if one had a multi-bit DSD file (which don't exist), there would be no way of feeding it into a typical DAC.

 

If you look at the DSC1 design, the conversion from single bit to multibit is entirely specified (shift register ladder).

 

That's a moving-average low-pass filter combined with D/A stage, not a 1-bit to multi-bit conversion. It's a neat way to convert 1-bit to analogue, but it has nothing to do with a multi-bit digital representation.

[Jud]

And how would those four sequential values be represented in the PCM bitstream?

 

That depends on the sample encoding you choose. Why does it even matter? We're talking about arithmetic values here.

 

Should I take it from this that the filtering (always) precedes the sub-sampling?

 

It has to, or you'd get aliasing. It works exactly the same as any other sample rate conversion.

 

So low-pass filtering, followed by sub-sampling using whatever encoding is chosen. If one wants RedBook as the result of the conversion, one chooses binary words of 16 bits in length at a 44.1K sample rate as the encoding. Thus saying "a standard low-pass filter turns DSD into PCM" (http://www.computeraudiophile.com/f8-general-forum/mark-waldrep-and-more-silliness-28952/#post554362) seems to leave out half the story. One might even call it "cryptic to the point of being intentionally misleading" (http://www.computeraudiophile.com/f8-general-forum/mark-waldrep-and-more-silliness-28952/index2.html#post554440).

[jabbr]

There is no law, but actual DSD DAC chips have one data pin per channel and one clock pin. The I2S interface only handles PCM data at lower rates. Thus if one had a multi-bit DSD file (which don't exist), there would be no way of feeding it into a typical DAC.

...

 

That's a moving-average low-pass filter combined with D/A stage, not a 1-bit to multi-bit conversion. It's a neat way to convert 1-bit to analogue, but it has nothing to do with a multi-bit digital representation.

 

Perhaps I just don't understand ... could you point me to the specification of multi-bit DSD so I can compare that to the 33 level output of the DSC1 shift register stage? I would also like to see exactly what multipin format the multibit DAC chips would accept.

[mansr]

Perhaps I just don't understand ... could you point me to the specification of multi-bit DSD so I can compare that to the 33 level output of the DSC1 shift register stage? I would also like to see exactly what multipin format the multibit DAC chips would accept.

To my knowledge, no such spec exists.

[firedog]

Your reasoning appears to go something like this:

 

1. 1-bit DSD is produced by an SDM process.

2. An SDM can also produce multi-bit output.

3. Therefore any output from an SDM is DSD.

4. All DACs include an SDM stage.

5. Therefore all DACs convert to DSD.

6. Therefore PCM playback cannot outperform 1-bit DSD.

 

If you can't see the flawed leaps of logic in this, there's probably no point trying to educate you.

 

What you refuse to understand is that in the Waldrep article he was trying to debunk a quote about DSD and how most DACs and ADCs convert "everything" to DSD. What both you and he fail to understand - or intentionally misunderstand - is that "DSD" in that quote basically is audiophile vernacular for PDM/SDM, so it refers to both 1 bit and multi-bit DSD. When you understand that, it it clear that the referred to quote is correct and Waldrep and others are grasping at some sort of literalist interpretation to make a point that is irrelevant to the quoted material. The original quoted passage came from a post at this site. Do you somehow think every post here is written with the rigour of a peer reviewed article?

 

I would say there's "no point in trying to educate" you, as you refuse to understand the simple and obvious meaning of the referred to post-when you read it in the context in which it was written. The same point has been made numerous times on this forum by Jud and others - and is basically accepted as true. Please come back and tell me that most modern DACs don't convert PCM to hi-res PCM and then use a form of sigma delta conversion - be it one bit or multi-bit - before filtering and producing analog. I want to hear you claim that.

 

That was the point and claim of the passage Waldrep is bending over backwards to refute. Of course, like you, he is refuting a point that wasn't made. Sometimes in order to understand reading material you also need to understand the context in which it was written. Literalism isn't always the best method for understanding.

 

Logical enough for you?

[Miska]

Thus if one had a multi-bit DSD file (which don't exist), there would be no way of feeding it into a typical DAC.

 

Don't tell that to my DACs.

 

That's a moving-average low-pass filter combined with D/A stage, not a 1-bit to multi-bit conversion. It's a neat way to convert 1-bit to analogue, but it has nothing to do with a multi-bit digital representation.

 

It can take in N-bit SDM. There is completely no need what so ever to think it would be limited to "1-bit". It is actually not much different from for example ESS Sabre or dCS DACs. DCS1 has 33 output levels, Sabre has 65 and dCS has 25.

 

As long as you talk about "1-bit" you seem to equate multi-bit SDM representation to multi-bit PCM representation. There are many fundamental differences.

[mansr]

Don't tell that to my DACs.

 

 

 

It can take in N-bit SDM. There is completely no need what so ever to think it would be limited to "1-bit". It is actually not much different from for example ESS Sabre or dCS DACs. DCS1 has 33 output levels, Sabre has 65 and dCS has 25.

 

As long as you talk about "1-bit" you seem to equate multi-bit SDM representation to multi-bit PCM representation. There are many fundamental differences.

 

Your DAC isn't typical. As for the rest, their internal machinations are not relevant when they lack input pins/protocols for anything but 1-bit DSD or PCM. Not that it matters. I'm still waiting for someone to show me even one multi-bit DSD file available for purchase (or otherwise).

 

I'm not equating multi-bit with PCM, but I am assuming it must mean something where each sample point is represented by more than one bit, whatever those bits mean. If this is not what you mean, could you be so kind to explain what I'm missing? Please?

[徐中銳]

Incidentally, if Sony's your argument, I recall the following being somewhat silly, from sonylearntv, 11 minutes :

« DSD (Direct Stream Digital) is the ideal format for delivering High-Resolution Audio. Find out why and how it works in this episode. »

[Paul R]

Your DAC isn't typical. As for the rest, their internal machinations are not relevant when they lack input pins/protocols for anything but 1-bit DSD or PCM. Not that it matters. I'm still waiting for someone to show me even one multi-bit DSD file available for purchase (or otherwise).

 

I'm not equating multi-bit with PCM, but I am assuming it must mean something where each sample point is represented by more than one bit, whatever those bits mean. If this is not what you mean, could you be so kind to explain what I'm missing? Please?

 

Correct me if I am wrong, but guys - a multi-bit DSD file would just be 1-Bit DSD packaged more conveniently - no? As in something like DSD over PCM (DoP)?

 

If you convert it to PCM, such as to DXD, then it isn't DSD or even PDM any longer - it is PCM. Like the difference between FM and AM radio. Same information, totally different format.

[mansr]

Correct me if I am wrong, but guys - a multi-bit DSD file would just be 1-Bit DSD packaged more conveniently - no? As in something like DSD over PCM (DoP)?

 

How do you suppose DSF and DFF store the bits? The correct answer is 8 per byte.

[Paul R]

How do you suppose DSF and DFF store the bits? The correct answer is 8 per byte.

 

Well, if you happen to be on a machine that uses 8 bit bytes, and byte addressing, sure.

 

But the data isn't changed - it is still 1 bit DSD regardless of how you store it. (shrug)

 

-Paul

[Miska]

Correct me if I am wrong, but guys - a multi-bit DSD file would just be 1-Bit DSD packaged more conveniently - no? As in something like DSD over PCM (DoP)?

 

I've been using the most convenient way, utilizing unsigned 8/16/32-bit sample WAV at native SDM rate and defining custom media type. But of course one could use DSF or DSDIFF for the purpose too.

[mansr]

Well, if you happen to be on a machine that uses 8 bit bytes, and byte addressing, sure.

 

But the data isn't changed - it is still 1 bit DSD regardless of how you store it. (shrug)

 

Yes, so this multi-bit thing everybody insists I'm not understanding has to be something else.

[YashN]

I'm not equating multi-bit with PCM

 

Are you sure?

 

but I am assuming it must mean something where each sample point is represented by more than one bit

 

???

[tailspn]

Correct me if I am wrong, but guys - a multi-bit DSD file would just be 1-Bit DSD packaged more conveniently - no? As in something like DSD over PCM (DoP)?

 

If you convert it to PCM, such as to DXD, then it isn't DSD or even PDM any longer - it is PCM. Like the difference between FM and AM radio. Same information, totally different format.

 

Multi-bit is/are parallel 1-bit streams at the same bit rate, who have a mathematical relationship to one another. That allows the "x" bit wide "word" to be processed in a digital computer. Other than a several dB improvement in noise margin, there's no practical advantage of multi-bit (streams) over a single 1-bit stream to a consumer. There's a large advantage to someone producing recordings who does not want to suffer the losses of PCM conversion in order to post process a DSD recording. With multi-bit (wide) streams the digital processing math can be applied to the original bit rate data without filtering and decimation. This is how the Sony Sonoma DSD-Wide (8 bits wide/8 streams) system operated.

 

There are a number of companies actively working on multi-bit Digital Audio Workstation software presently, which within a year should allow an alternative to PCM/DXD post processing of DSD recorded content.

 

There's no analogy to DoP, which is a serial bit packaging scheme into PCM word frames for transmission purposes. You're correct that DXD is pure 352.8KHz 24 or 32 bit PCM.

[YashN]

What you refuse to understand is that in the Waldrep article he was trying to debunk a quote about DSD and how most DACs and ADCs convert "everything" to DSD. What both you and he fail to understand - or intentionally misunderstand - is that "DSD" in that quote basically is audiophile vernacular for PDM/SDM, so it refers to both 1 bit and multi-bit DSD. When you understand that, it it clear that the referred to quote is correct and Waldrep and others are grasping at some sort of literalist interpretation to make a point that is irrelevant to the quoted material.

 

That's my understanding as well.

[tailspn]

How do you suppose DSF and DFF store the bits? The correct answer is 8 per byte.

 

DSD is of course organized in bytes, to facilitate storage and transmission. But the important segmentation is into 13.3ms 1-bit words, known as CD frames. It is on the frame boundaries that edits take place in order to render them clickless.

[mansr]

Multi-bit is/are parallel 1-bit streams at the same bit rate, who have a mathematical relationship to one another. That allows the "x" bit wide "word" to be processed in a digital computer. Other than a several dB improvement in noise margin, there's no practical advantage of multi-bit (streams) over a single 1-bit stream to a consumer. There's a large advantage to someone producing recordings who does not want to suffer the losses of PCM conversion in order to post process a DSD recording. With multi-bit (wide) streams the digital processing math can be applied to the original bit rate data without filtering and decimation. This is how the Sony Sonoma DSD-Wide (8 bits wide/8 streams) system operated.

 

Since when is a several dB improvement in noise margin not a good thing?

 

I'm also still waiting for someone to explain how these multiple bit streams can be fed simultaneously into a single DAC pin, or failing that, concede that the claim that "any DAC" can play such streams is false.

[mansr]

Incidentally' date=' [u']if Sony's your argument[/u], I recall the following being somewhat silly, from sonylearntv, 11 minutes :

« DSD (Direct Stream Digital) is the ideal format for delivering High-Resolution Audio. Find out why and how it works in this episode. »

That was hilarious.

[Jud]

I'm interested in intermediate DAC formats. For example, is the "5-bit" internal product of a TI chip 5 1-bit streams, a combination of 5-bit PCM and a 1-bit bitstream, or something else?

 

Also, if a DSD128 file is input to a typical (non-ESS) DAC, what if any internal processing is done prior to final conversion to analog?