Me again.
I would like to quickly point out that, since my study is on-going, my layman ways of understanding and absorbing these things is also ongoing.
As far as DSD or 1 bit audio, the little iFi iDSD Micro Black Label addendum I penned as it pertains to multi-bit DSD and conversion via ESS chipsets and DAC's ----if written today would probably be worded a bit differently, and be a little clearer this time around. But, the end result/opinion is the same.
ESS and others...Too much DSP, too much filtering, too many conversions and reconversions on a format that only needs ONE low pass filter at the output stage.
As far as ESS is concerned, what they REALLY do with DSD, no one but those guys know. The 'for inside guys only' datasheet and info chart shares little more than squat. But what we DO know from some golden nuggets* shared by some of their team, ESS feels that filtering DSD digitally, digital volume control, ASRC, and another trip down their Hyperstream Delta Sigma Converter and analog filter output stage leads to a better sounding DSD in the end vs the 'old' way subscribed to by Burr Brown and a few others.
NOW, the COOL thing is there is ONE chip maker whose patents are readily available for perusal that tells us what is going on in their '1-bit domain, non-decimation, no PCM intermediary' technology. CIRRUS. LINK BELOW..
https://www.google.com/patents/US6784816
LOTS of INTERESTING stuff. I imagine ESS does MANY similar things. Actually, when the Cirrus paper speaks of IIR filter implementation versus FIR, I thought of ESS immediately.
Again, I am just a layman here. Musician is my trade. But I CAN in those patent explanations follow a significant chunk of what is going on with their DSD processing. One of their methods actually reminds me of Sony's DSD-Wide, which takes the 1-bit DSD stream, FIR filter's the 1 bit stream, (I forget how many delays and taps) and noise shapes down the resultant multibit samples to an 8-bit 2.8 mhz bitstream. Of course, the impulse/time response itself may be no better (or maybe worse) than just decimating to 352khz DXD ala Pyramix. HMMMMM.... but I digress...
LOTS of GREAT info in that google link about the modern state of DSD processing. (seems quite a bit ahead of the old Sony 8 bit DSD-Wide that many a DSD lover listens to on a regular basis... ahem.. ahem... I digress, I digress yet again)
The bottom line, many of these chip/DAC makers feel that implementing a 'good enough' analog filter for a DSD bitstream is too difficult to do. The end result if done that way is just simply a conversion filter that ends up sending down too much ultrasonic noise that compromises the downstream equipment. Or if you design that filter with aggressive cut and rolloff to compensate, you totally screw up the DSD advantage a different way. Damned if you do, damned if you don't.
So these chipsets result to various DSP's, doing everything possible to avoid conversion to low sample rate PCM, surely for marketing reasons but surely also their shared belief that the time resolution of DSD should be maintained as much as possible therefore the filters should be better than regular PCM decimation filters. And they believe that their noise removal via internal DSP, upsampling, filtering, re-modulation in multi-bit delta sigma, combined with their 'better' more easily implemented analog output stages makes for better DSD playback than those DAC's that simply ue ONE ANALOG FILTER at the converter output.....
ala iFi and their Burr Brown chip.
To ME, in spite of any inferiority that allows for more ultrasonic noise downstream, I just have to believe the simple approach of a short moving average FIR filter applied directly to the previously untouched 1-bit bitstream will sound better.
My own non-scientific anecdotal listening evidence says the same as well.
In spite of that right now I am listening to all my digital files, which includes a couple TERRABYTES of DSD, on a very very nice sounding ESS chipset DAC. And yes, all that DSP processed DSD sounds very very very good.
BUT... I am holding out for the iFi iDSD Pro, which I expect to BLOW OUT OF THE PROVERBIAL WATER these DSP based DSD comers. '
iFi, PLEASE say it is coming our way soon! I am practically begging you to take my money in exchange for one
back to happy listin'
*( one of those Golden Nuggets corrected my previous ideas about where the ESS volume control was in regard to DSD. It, like PCM, is at the very beginning, right after the USB I2S. Check out the linked paper about Cirrus DSD processing, which is what I am fairly convinced is how ESS does it that early in the signal chain. With DSD being just a two number signal, 1 and -1, a constant scaling multiplier is applied)