One would believe, because one’s belief that DSD sound quality versus PCM sq is purely subjective, leading to one’s library consisting of very few examples of DSD, that the absence of DSD filtering would be of absolutely no concern… but…
I remember reading from a Mojo kb article that PCM is native to R-2R DACs and DSD is native to delta-sigma DACs, and that conversions are necessary if the bitstream isn’t native. Am I to assume this is still the case, and is it during the conversion that filtering is applied (=> concern)? Or, is the filter applied to the analog result, and filtering is independent of conversion (=> no concern)?
I did audition the Cayin RU6 R-2R dongle DAC before ending up with the L&P dongle, but I do remember liking the RU6’s NOS filter best, whereas the NOS filter on the L&P seemed to be a different animal altogether. Would the absence of any DSD filters be the same as NOS?
well, some of us, no names mentioned lol, have a couple terabytes of DSD material.. mostly native in some way.
There are several ways to convert DSD... here are the most common.
1. Native (or, as close to native as we can get)
The bitstream stays at 1 bit, untouched. No DSP, Bypassing all volume control, no pre-filtering, no second delta sigma modulating stage. The actual digital to analog conversion is done by discrete parts. These discrete parts are a shift register out of which multiple 1-bit streams come forth offset by 1 sample each, some kind of analog dynamic matching/scrambling, and bit switches, usually collected into multiple groups to create 'analog coefficients" Yeah, its a digital FIR filter made with all analog discrete parts. This is what Burr-Brown DAC's like the DSD1793 at iFi have. T+A makes a discrete version of their own. Holospring and Denafrips all use something similar, and Miska (his handle at AudiophileStyle) has a open source DAC anyone can build that is DSD only using this technique.
Why is it so good? The DSD bitstream is only touched once, and is ingeniously a combination of the necessary FIR filter and the digital to analog converter in one piece.
It might look something like this.... (note that this is extremely similar to the output stage of a multi bit Delta Sigma DAC as well. However, this is the way to bypass all the DSP involved in a multi bit Delta Sigma DAC and in the end, have basically a 1 bit DSD (Delta Sigma) stream converted directly to analog via a single end of stage FIR filter)
1 Bit DSD enters Shift Register( we will make it super short, like T+A which I believe only 4 bits/clock cycle of delay) means it looks like this.........
2. CONVERSION OF DSD VIA DSP TO INTERMEDIATE MULTI-BIT STATE, RE-MODULATION BY MULTI-BIT DELTA SIGMA NOISESHAPER, CONVERSION TO ANALOG IN SIMILAR MANNER TO THE ABOVE.
THIS KIND of Conversion you will find in the ESS chipsets, although they give us very little REAL details on how they do it, and honestly, some of it just doesn't add up and needs a more open explanation. After all my reading we are lead to believe that a gain stage is applied directly to the one bit signal to create a volume control and a multi-bit intermediary. Well, it CAN'T be that simple. We are dealing with a 1 bit 2 level binary here. If you are going to simply multiple each 1 or 0 by a 32 bit gainstage number, well, that just won't work. You can't just convert 1 bit 2.8mhz DSD magically into 32 bit 2.8mhz 'super DSD' via a gain stage. MORE must be involved, such as using something like the above example 1.... a short moving average filter which yes, you CAN then multiply by a gain stage and change volume. (BTW that is EXACTLY how Sonoma 8bit DSD-WIDE is made from 1 bit DSD.) But rather than harp on ESS lack of transparency, we can talk about AKM, or Cirrus, or any other that is a bit more transparent about their process.
AKM offers TWO modes. It offers a direct mode exactly like I explained above in Example One. But the DAC manufacture has to choose the added complexity of implementing this mode along with the more convenient processed mode, which allows for gain matching and volume control. The Topping E70V does not use the direct mode. It uses its processing mode. So, then how does it work?
Of course this is very simplified. It also 'reads between the lines' of the AKM Datasheets, and so often those are not exactly accurate and are used to sell chips moreso than actually provide detailed information about what actually is going on.
I hope many of you find this at least somewhat edifying.... now, I am back to some final measurements on the Topping DAC's.
