Thanks for responding. Regarding ADCs, I'm referring to
this post and the posts surrounding it. It seems like the only ADC with a native PCM front-end has been discontinued, so they are all at least multi-bit PDM, and many single-bit PDM. (I don't understand the distinction between multibit delta-sigma and PDM, I've always seen the terms used interchangeably.) So it seems to me the purest audio chain is had by taking a 1-bit PDM ADC and directly sending the output to a DSD file, which appears to be what many of the audiophile DSD labels are doing (either using live music or analog tape masters as sources).
And I guess one can argue about the algorithmic complexity of the noise-shaping/feedback loop in the ADC, but when considering playback, DSD surely is substantially more "analog" than PCM.
Here is a quote from Paul
McGowen of PS Audio.
"
Here’s the interesting part of this: if you take a DSD stream and run it through a simple analog lowpass filter to smooth out the on/off transitions, you get music! This is amazing considering that if you do the same with PCM you get only noise...
DSD is a lot closer to analog than PCM ever thought to be."
Of course as you point out, to get the best music the playback circuit isn't quite that simple. But it does produce music. To me, that's very special and even a bit magical, for a file that can be stored as a stream of bits on a hard drive, just like PCM.
And I don't think the DSD to HDCD comparison is fair. HDCD had a strong corporate-pushed beginning, a brief peak and then a quick death. Apparently it was just a variation of PCM, in some proprietary format, nothing really exciting about it. DSD, in the form of SACD, also had a strong corporate-pushed beginning, a brief peak and then appeared to die... but with the birth of high-end computer audio was resurrected, and has been gathering lots of momentum in the few brief years that affordable USB DSD DACs have been around. And as discussed above I think it's so fundamentally different a format that there are plenty of reasons for people to get truly excited over it. We will see if the train keeps gathering steam.
I don't want to continue the DSD discussion endlessly, but I do have to disagree with the "magical" aspect of its playback, and in Paul McGowan's interpretation of it as being "closer to analog." I have the greatest respect for Mr. McGowan, but we definitely differ on how we view DSD.
While DSD can be viewed as audio by looking at it with a low-pass filter, this has nothing to do with magical qualities, and everything to do with the computation-intensive noise shaping done at the encoding phase. A 2.8MHz bitstream (or 5.6, or 11.2) has a wideband signal to noise ratio of 6dB, which is a mathematical fact. This can only be reduced in the audio band by computation that pushes the noise out of band, as explained in Stanley P. Lip****z and John Vanderkooy's AES paper:
Why One-Bit Delta Sigma Conversion is Unsuitable to High Quality Applications
The abstract:
"Single-stage, 1-bit sigma-delta converters are in principle imperfectible. We prove this fact. The reason, simply stated, is that, when properly dithered, they are in constant overload. Prevention of overload allows only partial dithering to be performed. The consequence is that distortion, limit cycles, instability, and noise modulation can never be totally avoided. We demonstrate these effects, and using coherent averaging techniques, are able to display the consequent profusion of nonlinear artefacts which are usually hidden in the noise floor. Recording, editing, storage, or conversion systems using single-stage, 1-bit sigma-delta modulators, are thus inimical to audio of the highest quality. In contrast, multi-bit sigma-delta converters, which output linear PCM code, are in principle infinitely perfectible. (Here, multi-bit refers to at least two bits in the converter.) They can be properly dithered so as to guarantee the absence of all distortion, limit cycles, and noise modulation. The audio industry is misguided if it adopts 1-bit sigma-delta conversion as the basis for any high-quality processing, archiving, or distribution format to replace multi-bit, linear PCM."
This paper specifically addresses 1X DSD. 2X and 4X DSD address some of the shortcomings outlined in the paper, but as you know, 2X and 4X DSD recordings are very, very thin on the ground, and the number of 2X and 4X DSD recordings that remained DSD all the way through processing is markedly smaller. The link regarding three new ADCs that support DSD recording does not address the fact that the output of the vast majority of studio ADCs is PCM, and that most all studio processing is done in PCM, so any DSD recording is likely to be of mixed provenance, unless taken directly from a single-mic recording feed, or taken off of original master analog tapes (which have their own issues with noise.)
It's important to note that this is not a screed to tear down DSD, it is just to try to show there is no magical "direct to the source" format. Similarly to DSD's provenance issues, most modern PCM recordings go through an intermediary multibit delta-sigma stage within the A/D converter before being output as PCM, and (these days) are usually converted to a multibit delta-sigma output as part of the D/A conversion process. Records go through RIAA equalization and are completely dependent on the mechanics of the cutter used to make the master pressing, as well as the quality of the vinyl, the quality of the cartridge, the quality of the phono preamp, etc.
If you're interested in delving more into the grey areas in PCM vs DSD, arguably the best comparison of DSD to PCM at the current state of the art is here, done by Charles Hansen of Ayre:
World’s First Valid Comparison of PCM versus DSD
So what does all this talking get us? In the end, we have different philosophies.
DSD advocates believe that upending the entire recording chain (and somehow maintaining DSD through the chain) with playback through a pure DSD DAC will end up creating audio nirvana. It is important to note that most DACs that play back DSD actually (gasp) convert it to multibit delta-sigma during the D/A conversion process, so they are not pure DSD.
We believe that the widespread use of multibit delta-sigma D/A converters, as well as the use of mathematically intrusive strategies such as asynchronous sample rate conversion and open-form digital filters has limited the capability of the PCM format. As with DSD, only a handful of PCM DACs actually use R2R D/A conversion, so it is not pure PCM.
So, two different philosophies, two different approaches. We'll see if we can get closer to the original with Yggdrasil, and other companies will see if they can get closer to the original using DSD. We'll see what the listeners say when they're both fully realized.
In the end, the market will likely shrug at both, since, by and large, it will be streaming compressed music via monthly subscription, rather than buying 2X DSD or 24/96 PCM.