I personally think that DSD upsampling will not give you much audible benefit, but there are plenty of people who do. It
is possible to use software to convert PCM to DSD and upsample it in real time, the only software I am aware of that can do this is
HQPlayer. If you do use HQPlayer, you need to be aware of the following caveats:
- HQPlayer is expensive! USD$326.
- Performing 6-8 channels of conversion + convolution is EXTREMELY CPU intensive. It is not uncommon for people to use dedicated PC's for HQPlayer, and these are typically fire breathing watercooled monsters with high end graphics cards (HQP takes advantage of nVidia CUDA cores) typical of gaming setups.
If you decide to use HQP, you can use either the embedded version (where it lives on its own dedicated PC) or the player version (where you use a third party playback program like Roon on the same computer). Again, there are people who believe that separating the convolution/upsampling function from playback into two PC's results in better sound. I am not one of them.
Don't forget that the thermal cost of using HQP is significant and you will need a strategy to deal with that. If you decide to air cool your PC, your CPU fans will be running all the time and it will raise your room's noise floor unless you move your PC to another room and run a long network cable to send output from the PC into your Merging via Ravenna. Or you could consider watercooling, which does reduce noise, but increases the complexity of your build.
Now, I know for a fact that my Merging NADAC has a lower SINAD when fed a DSD signal (-118dB) compared to 44.1kHz PCM (-110dB). I have measured it with an E1DA. However, upsampling PCM from 44.1kHz to 192kHz also brings the SINAD down to about -116dB. The CPU cost for upsampling PCM 44.1kHz --> PCM 192kHz is minimal, my CPU typically sits at 8% usage when I am doing this. OTOH, upsampling PCM 44.1kHz --> DSD128 results in CPU usage of 95%, and I sometimes get audible signal dropout. The benefit is a 2dB improvement in SINAD. I can tell you that I can not hear a difference between PCM 44.1kHz and PCM 192kHz, and I definitely can not hear a difference with DSD. The reason I upsample to 192kHz is because it is a "free upgrade" in SINAD and it only costs me a tiny bit extra CPU usage and raises my electricity bill by a tiny amount.
I spoke earlier in this thread about our own biases, and this is one of mine. IN MY OPINION, DSD upsampling is not needed. If it came with a low CPU cost, I would do it and grab the extra -2dB improvement in SINAD even though I know perfectly well it is not audible. What you are really doing is expanding a lot of effort into pushing an inaudible number into an even more inaudible number. There is plenty of low hanging fruit to pick before you even do that. I can tell you for a fact that creating better convolution filters through improvement in my understanding of room acoustics, measurement methods, knowing what to correct/what not to correct, and better usage of my software (Acourate) results in massive objective and subjective improvements to the sound.
However, there are people who want "the ultimate" and build two PC's, upsample to DSD, electrically isolate all network connections by using optical, use atomic clocks, audiophile grade switches, and the like ...
here is an example of such a system. I applaud his effort, determination, and attention to detail. In fact I have to say that I admire him, even if I disagree with half the things he is doing. You can decide if you want to go down that path, but I doubt that many people on ASR would think that going to such lengths is fruitful.
