Scholar or technical articles about DSD ? Specially looking for DSD 128 and above specifically

[Zareli]

Hi community:

 

I've been intrigued by Hi-Res DSD formats lately. I already know DSD's main claim is the extended dynamic range, being around 120db (although somewhere I read on non-music sound tests it could reach 180db potentially). Also everyone seems to agree that PCM 24/48 is technically superior than DSD64 and more efficient (raw uses less than DSD, losslessly comrpessed uses less than DST). I want to know how do higher rate DSD compares to PCM but no one seems to be making that comparison. I read the whole experiment of the Mendelssohn Session and heard the tracks. I heard little if any difference (I can't play native DSD256 anyway) but I was more interested in the discussion than the raw result.

 

So conceptually this is what I "learned":

Higher rate DSD allows noise shaping into a higher frequency realm, which allows a higher cutoff, which means extended frequency response. But what about dynamic range ? From my basic knowledge about Sigma-Delta converters, the oversampling (over nyquist frecuency) is what ultimately allows to reduce noise floors and increase dynamic range. I'm inclined to say that DSD128 then would have a higher dynamic range as well as a higher frequency range.

 

If that is so,

 

Then what would be the closest "PCM mathematical equivalent" to DSD128 and DSD256 ? (I know there's no such thing but let's say 20bit/44.1khz or 20bit/88.2khz can more or less transparently represent DSD64 and vice versa).

 

How much dynamic range increase you get by doubling the DSD sampling rate ?

[Ralf11]

I would be looking for listening tests published in journals with a sound methodology - and focusing on impulse sounds, like drums, snares, etc.

 

That is where I'd expect greater bit density to make sense...

 

Closer to your line of thought, I'd also look for info on different filtering techniques "more gentle slopes" etc.

[Zareli]

I would be looking for listening tests published in journals with a sound methodology - and focusing on impulse sounds, like drums, snares, etc.

 

That is where I'd expect greater bit density to make sense...

 

Closer to your line of thought, I'd also look for info on different filtering techniques "more gentle slopes" etc.

Although listening tests are helpful sometimes, I believe anything above 48khz/24bit pcm is beyond what humans can detect or differentiate in AB listening tests. About DSD, I belive DSD64 still can get more dynamic resolution but I don't know if DSD128 can provide that extra resolution from a theoric point of view and anyway it's difficult to provide the conditions for a listening test with EFFECTIVE 120db of dynamic range and above (peaks above 120db that can be harmful), so ultimately one can assume that humans can probably detect the benefits of having more than 120 db of dynamic range, but the conditions are practically impossible (a room with <30 db floor noise and very specific and expensive lab components) and it would hurt the listeners anyway.

 

After that considerations it is safe to assume that DSD128 really doesn't offer practical benefits over DSD64 just like 96/24 pcm doesn't offer practical benefits over 48/20 pcm for the end listener and listening tests would be pretty much "useless" to answer the original questions. And if any difference can be heard at all it is probably an artifact rather than a real difference between the formats (you're hearing the downsample processing artifacts instead of the true difference between DSD256 and DSD64, similar to what was hypotethised in the Horus experiment.

[tailspn]

Then what would be the closest "PCM mathematical equivalent" to DSD128 and DSD256 ? (I know there's no such thing but let's say 20bit/44.1khz or 20bit/88.2khz can more or less transparently represent DSD64 and vice versa).

 

I can't speak to your scholarly quest for information, for I'm not qualified. But I can speak to the recording side of the equation; the music you purchase to enjoy through your listening system, that may influence your understanding.

 

The problem with a DSD/PCM equivalency is they are a different species. PCM is a amplitude digital value based system, describing a continuous analog signal as a series of discrete valued binary coded samples. Not unlike frames in a motion picture film strip.

 

DSD, or any like Sigma-Delta modulator derived bit stream, contains no digital coded values, just an analog density of clock bits describing a continuous analog signal through a continuously varying density of bits. Collectively, they express no value; just their density expressing a percentage of modulation. That percentage describes the current signal level as a percentage of the full dynamic range. It's further defined that 50% modulation equals 0dB. Depending on the voltage extremes of the demodulator (DAC), that can be any defined voltage value.

 

The fact that DSD and PCM are convertible one to another through a lossy process has people questioning each's quality, seeking some kind of definitive objective "this one is better than that one" answer.

 

Examining the complete recording through playback system, one objective statement that can be made is that all current, and for more than the past 10+ years, analog to digital converters used by the music recording industry are front ended with Sigma-Delta modulators, producing one or multi-bit Pulse Density Modulated bit streams. To obtain PCM word samples, the bit-stream(s) are decimate filtered and converted to PCM in a lossy process. The fewer those conversion processes occur in the production cycle, if at all prior to the making of the delivery media, all things considered, the higher the resulting true sound quality.

 

I a nutshell, while PCM is presently a necessary digital format for post processing Sigma-Delta modulated bit streams, it is always the product of the original Pulse Density Modulated bit stream. If the objective is to make the original sound different, called post processing, them PCM is currently the only path.

[Jud]

Although listening tests are helpful sometimes, I believe anything above 48khz/24bit pcm is beyond what humans can detect or differentiate in AB listening tests. About DSD, I belive DSD64 still can get more dynamic resolution but I don't know if DSD128 can provide that extra resolution from a theoric point of view and anyway it's difficult to provide the conditions for a listening test with EFFECTIVE 120db of dynamic range and above (peaks above 120db that can be harmful), so ultimately one can assume that humans can probably detect the benefits of having more than 120 db of dynamic range, but the conditions are practically impossible (a room with <30 db floor noise and very specific and expensive lab components) and it would hurt the listeners anyway.

 

After that considerations it is safe to assume that DSD128 really doesn't offer practical benefits over DSD64 just like 96/24 pcm doesn't offer practical benefits over 48/20 pcm for the end listener and listening tests would be pretty much "useless" to answer the original questions. And if any difference can be heard at all it is probably an artifact rather than a real difference between the formats (you're hearing the downsample processing artifacts instead of the true difference between DSD256 and DSD64, similar to what was hypotethised in the Horus experiment.

 

Then why are you asking. If you already know what you think, why is it someone else's job to persuade you of anything different?

 

 

Sent from my iPhone using Computer Audiophile

[Zareli]

Then why are you asking. If you already know what you think, why is it someone else's job to persuade you of anything different?

 

 

Sent from my iPhone using Computer Audiophile

I just have an opinion about listening tests in the matter and I didn't expect someone else to persuade me regarding that aspect. I just asked for recommendations about articles on theoretic and/or instrumental measurements regarding higher DSD rates improvement of DSD64 limitations. I'm sorry if I expressed myself in a misleading way.