Make PCM sound like DSD
I finally read it. Lynn, nice. I hope to have some useful comments. Notice though that while I actually use your own (said) base for all, I put it it in the 180 degree opposite context. Could be nice ...
Let me first virtually quote from others (like Mani has done it more in the beginning of this thread) that when DSD is compared to PCM, and then in the perceivedbly most direct way (chains) for both the DSD and PCM situation - and which would be self recorded (Tascam) DSD played back through a Mytek vs self recorded (PMII) through the NOS1, DSD sounds saltless. We could say "more analog" but in the end we maybe must redifine how analog is sounding, knowing that the dynamics in there can't be the very best, plus knowing that theoretically digital performs better there. Much more to say with pros and cons for both, but the grasp of it is hopefully clear. So, this PCM wins "hands down" and it is in the dynamic character of the sound, which at the same time is smooth and silky. So, notice the explicit recognition of being dynamic and grainy, which is not difficult to achieve. Have it silke and dynamic, is. And before we forget, this is actually comparing Redbook to DSD, or IOW lowres to hires. This is important knowledge for the below. Now from your article :
The classical PCM spectra has a very different effect on slew-prone analog electronics. The extremely fast rise-times – in the nanoseconds – slews the analog electronics at the rising and falling edge of every sample, and the rise time of the switch array in the converter is pretty much the same whether the converter is operating at 44.1kHz or a much higher frequency like 705.6kHz. The slew events are so short that it doesn't appear on FFT-based distortion measurements, since the FFT measurement is averaged over a second or longer; obviously, a few nanoseconds occupy only a very tiny portion of a second.
Emphasis is mine.
What you assume here, is a high
frequency, which at least with Redbook can not be the case (ehm, that much); it is just not in there. Now :
What *does* happen, and in fact all the time, is the very same fast rise time, but from
transients. A transient is not a frequency (unless we observe it in the electrical domain) but it is just a one-shot fast rise (never drop) which occurs at attacks (think the rim of a drum for example).
No real world attack rises inifinitely fast (but let's not forget synths), but, because of the too low sample rate it will look like close to that in digital. So, say that this rim shot evolves so fast to its (sort of) sustain level, that from one sample to the other it goes up from 0V to 1V. This really happens 1000s of times (with the 1V as the real example were it for 2VRMS output of a DAC) and thus it has to be dealt with. Has to be ?
Yes, because a transient is a transient and nothing tells that it can't be dealt with. Ok, the slew rate does. Aha. But anyway, because it is not a frequency, it just can be done by any analog device which can follow it (you may recall my dirac pulses plot).
since the sample duration has been affected by the slewing, which is not the same as linear low-pass filtering.
Here you actually touched my subject above. Well, this is how low pass filtering will flatten transients, unjustiffied.
Now your PCM=DSD solution ("BUD");
Although I didn't do the math for merits, it is clear that the DSD principle allows for transients which are derived from the sample speed, the maximum allowed voltage and thus the shortest time possible to reach our 1V from above. Of course it is clear that many samples are needed to get from 0V to 1V and it is only that the e.g. 2.8MHz sample rate can do this fairly fast (notice that at least for SACD it is not allowed to have more than (IIRC) 14 subsequent DC rises so that also slows down). Still PCM does this infinitely fast. And now the fun :
Where you actually slowed down (smoothened) the rise time (hey, and fall time because you talk about frequencies) so it will sound more analogue (actually : more DSD), I can use the very same principe of DSD to point out that it can't deal with transients as good as PCM can.
So ... you knew this, but we both used a different context and I see the infinite possible rise time of PCM as a virtue while you see the slow rise time of DSD as a virtue.
Maybe a year or so back I talked about a similar subject and how the higher frequencies may not be important in the first place, but transients always are, and there it was the first time I could find a reason why PCM-hires sounds worse than Redbook. Please keep in mind, this is "our" NOS1 usage and ever and always knowing that the thing operates electrically 100% the same for either format (because all is done in software preceeding it). So, read your own article Lynn, and see how it can happen indeed that PCM-Hires needs your DSD approach. Why ? because *then* you talk about frequencies and possible messy slewing.
That's it. Shoot me if needed.
Peter