I’ve been looking at various articles about DSD, and I’m having a hard time understanding why anyone would want or need DSD support. Clearly Sony and Philips wanted it for the (failed) SACD revolution, but why prolong the agony? It really seems like one of these snake-oil power cords.
Sure, vinyl is still with us, and maybe DSD is just inertia. Hard to imagine there’s any nostalgia about it, though. Why not just convert all DSD files to PCM and be done with it?
OMG! Not again! Let the flame throwing begin!
No, no flames please. I’m just wondering if there are any real arguments in favor of the format. It appears from what I can read to be a remnant of a failed medium push, not anything intrinsically interesting in its own right.
Not an expert but a lot of the audiophile vinyls (including MoFi as per the scoop) use analogue master to DSD to lacquer transfer. Some of the pretty decent recent remasters like Floyd also use the same. Seems DSD is the preferred format to digitise analogue masters. There must be some merit to the format over PCM. I’d love to know more.
Me, too! I think it’s probably easy to generate a DSD signal; that may be part of it.
According to my limited historical knowledge, DSD (or SDM, PWM, PDM, 1-bit PCM) was initially an archival format. Back in the day, A/D converters had a 1-bit quantizer, so it made sense to save the raw quantizer output in an archive. The idea - as with any other raw format - was probably to let applications decide for themselves what rate and bit depth to use for mastering. This is probably why DSD is a choice for analog tape archival. Once edited however, there is no good reason to go back to DSD for storing or distribution. It does make sense to use DSD up-sampling for rendering at playback time. That is because, as @Bill_Janssen said, of the simplicity of the D/A stage. All you need is a comparator, so there’s no problem with component tolerances. Same goes for the A/D process. That’s the main advantage of DSD. Many DACs today use internal multi-bit D/A stages, with definite advantages over DSD in terms of dithering and quantization noise.
I highly recommend reading this thread as @jussi_laako knows way more about DSD than anyone else commenting in this thread…as you will see:
Good pointer. Thanks!
Lot of jargon in that thread; I’ll have to read it carefully.
Would it be entirely terrible if I asked you to summarize what the advantage of DSD is? If any?
It adds more things for audiophiles to argue about, and companies can sell more products to process PCM and DSD back and forth, all the while claiming vast and apparent sonic benefits that, apart from never withstanding any blind testing also can not in any way be explained by the very science that was used to actually develop both the PCM and DSD formats and the entire processing pipeline for either.
So basically, it adds some significant entertainment value, and isn’t that what it’s all about?
We usually see things differently, but here i agree.
It comes down to the simplicity of transitioning to analog (or from), as i see it.
Once discrete, in the digital domain, the media can be transcoded to PCM for editing, mixing and manipulation. And it can be upsampled/transcoded back to DSD again before analog conversion, to simplify the process.
In a way you could see it as inserting the time component back into the digital signal before conversion, rather than as a part of the DA-process…
The up-sampling is, in my view, part of the D/A process, whether it’s done internally in the DAC or externally through DSP. I’m not sure what you mean by “time component”, but no timing is lost, whether using PCM or DSD, and there is no reason to render in the same manner as the acquisition.
Yes, i am a bit vague on that part, but your “up-sampling” is really what i consider the “insertion of the time component”. And as you know, a low pass filter is in most aspects all it takes to bring a 1-bit delta/sigma based signal back into the analog domain.
I agree, in the sense it restores the “missing” samples to aid in the D/A process.
It’s a bit more complicated than that. In order to low-pass the analog signal, you need an analog signal in the first place, so you need to translate the bits into analog. That’s the actual D/A process. Once that is done, the low-pass filter is all that’s left to do, regardless of the number of bits. It’s true that converting one bit to analog is simpler than multibit, but even with one bit, there are ways to “smoothen” the sharp transitions of the analog output, and that blurs the distinction between single- and multibit.
May I ask some supplementary questions?
For instance, how many DACs truly support DSD? Do most DACs transcode DSD to high-resolution PCM? I know that Chord Electronics supports DSD, but that’s only as an input*. “Native DSD” implies that a DAC supports DSD, yet I suspect this isn’t the entire picture, and could be misleading.
Finally, how many recordings are truly DSD. Again, I’ve read in the past that most recording, mixing, and mastering is in PCM.
*I understand that only one Chord DAC implements DSD in the FPGA; all others convert to PCM.
That would be stupid implementation.
So do most DSD DAC’s, including the one I published.
So far, I have failed to see such in their measured performance.
Although it is totally unnecessary to convert back and forth, you can stay as SDM instead.
Because there are no commercial native PCM audio ADC’s on the market. And only very few native PCM DAC’s (R2R’s).
Converting to PCM introduces number of problems that are hard to overcome.
PCM is mostly historical remnant from the days when available DSP processing power was low. Like when CD was developed in late 70’s. Certainly not the case anymore. Just compare 70’s mainframes to current mobile phones which have much more processing power than the biggest computers in 70’s.
Chips from TI/BB, Cirrus Logic, AKM and RoHM.
Chord DACs always convert DSD to 705.6k PCM with a rather poor algorithm and then upconvert it again to feed it to their modulator.
Those are all very good questions.
I don’t know how many, but I know there are DAC chips that always convert to PCM internally and others that can be configured to do “direct DSD” (and bypass digital volume control), but the DAC using the chip has to support that.
Note that I call it “direct DSD” instead of “native DSD” to distinguish between the A/D conversion and the digital interface (i.e. “native” vs. DoP).
All mixing and mastering is in PCM (i.e. with high bit depth), since you can’t edit a 1-bit signal.
By “native PCM”, do you mean without delta-sigma + decimation? Does it really matter how it was done if the output is PCM?