Questions about "original sample rate" and DSD

[AstroBenny]

So I was reading an article about audio formats here: http://www.eclassical.com/pages/24-bit-faq.html

And came cross this section, and I quote:

"What we call "Original sample rate" is the PCM sample rate of the highest bit rate files we receive from the labels. The highest bit rate download we offer will have the same sample rate. We ask the labels to give us the highest meaningful quality in which they have the recording. If they didn't convert rates during the postproduction, this will be the sample rate in which the album was recorded. But we cannot know each label's signal chain for each album. We trust them that they don't deliver upsampled recordings (contact our support if you think you found a "black sheep"). Should they send sample rates lower than the original, they obviously have no other choice (at present). Even there, we can't know whether a higher version maybe exists.

We understand that the above may be considered inaccurate for DSD recordings. Of course, there the "original sample rate" was 2.8224 MHz. And some people will definitely want to have the "better" DSD originals, no PCM conversions. But since we decided to only deal with PCM files at eClassical (up to now), we write what you will get when you buy the album. Unless looking up each album manually, we won't even know whether it has been DSD once - or maybe just some tracks of it?  That information rests with the labels.  Therefore, should it be essential for you to know whether a 88.2 or 96 kHz album originates from a DSD recording, we must ask you to check this with the respective label or read in the enclosed booklet. This is what our term "original" means in detail. We hope you understand that we don't have the possibilities to "legally bindingly" inform you about the production process of every album we sell, since we don't have access to each label's modus operandi for each specific record. "

 

So I have a few questions.

 

Why would the above be considered "inaccurate for DSD recordings? Why would the original sample rate be 2.8224 MHz?

And what does it mean if it has been "DSD once" ?

 

Thanks

[Stagea]

They are just saying that they cannot confirm if the recording was converted from DSD to PCM at any one point before it got to them.

[Mira Yurizaki]

8 minutes ago, AstroBenny said:

Why would the above be considered "inaccurate for DSD recordings? Why would the original sample rate be 2.8224 MHz?

DSD uses 1-bit per sample, as it's a PWM based encoding system, versus the PCM encoding based system most other audio codecs use. That's why the sample rate is that high.

 

There's probably a way to guess how a PWM signal will look, but I don't think you can directly compare PWM to PCM like you can AMD to Intel/NVIDIA.

[Guest Kloaked]

2 minutes ago, M.Yurizaki said:

but I don't think you can directly compare PWM to PCM like you can AMD to Intel/NVIDIA.

In a way you can. You can't look at Nvidia and AMD specs and then expect to know the whole story behind what kind of performance each card will have (depending on who you ask anyways). Like how AMD's stream processor count may outnumber Nvidia's cuda core count (they are still considered "stream processors" as well) but will not perform like you think it would if you just looked at specs on a data sheet. It'll always vary.

 

I'm under the personal belief that DSD isn't all that amazing and that most of your audio quality will come from the hardware you're using (speakers/headphones) then the rest on how well mixed and mastered the audio that you're listening to is - to an extent the bitrate but that's a can of worms I dare not talk about in a public forum

[AstroBenny]

15 minutes ago, Stagea said:

They are just saying that they cannot confirm if the recording was converted from DSD to PCM at any one point before it got to them.

So where does the value of  2.8224 MHz  come from?

[Mira Yurizaki]

1 minute ago, Kloaked said:

I'm under the personal belief that DSD isn't all that amazing and that most of your audio quality will come from the hardware you're using (speakers/headphones) then the rest on how well mixed and mastered the audio that you're listening to is - to an extent the bitrate but that's a can of worms I dare not talk about in a public forum

If I did my math right, DSD is about twice as good as CD quality as far as raw bandwidth is concerned. :3

 

Just now, AstroBenny said:

So where does the value of  2.8224 MHz  come from?

DSD uses a sample rate of 2.8224MHz. That's part of the standard.

[SSL]

35 minutes ago, M.Yurizaki said:

DSD uses 1-bit per sample, as it's a PWM based encoding system, versus the PCM encoding based system most other audio codecs use. That's why the sample rate is that high.

 

There's probably a way to guess how a PWM signal will look, but I don't think you can directly compare PWM to PCM like you can AMD to Intel/NVIDIA.

 

It's PDM, not PWM. Pulse Density Modulation vs Pulse Width Modulation. PCM is Pulse Code Modulation.

 

A PDM stream is 1-bit wide, as you say. The "density" or distribution of on to off bits represents the frequency. Digital to analog conversion of a PDM stream requires approximation of the exact output using a process called delta-sigma modulation, which looks at the preceding and following samples to generate the current value.

 

In fact, PDM and delta-sigma modulation is used in the vast, vast majority of DACs - even though most digital audio is encoded as PCM. The PCM is transcoded to PDM in the DAC before conversion.

[Mira Yurizaki]

6 minutes ago, SSL said:

It's PDM, not PWM. Pulse Density Modulation vs Pulse Width Modulation. PCM is Pulse Code Modulation.

Either or describes the encoding DSD uses: https://en.wikipedia.org/wiki/Pulse-width_modulation

[SSL]

57 minutes ago, AstroBenny said:

Why would the above be considered "inaccurate for DSD recordings? Why would the original sample rate be 2.8224 MHz?

 

It's possible that PCM to DSD conversion would mean that there would be less information in the file than if it has originally mastered in DSD.

 

7 minutes ago, M.Yurizaki said:

Either or describes the encoding DSD uses: https://en.wikipedia.org/wiki/Pulse-width_modulation

 

Pulse density, not pulse duration. They are similar, but not interchangeable. e.g. square is a rectangle, rectangle is not a square.