All of that was written by Dustin Foreman of ESS. He’s a friend and neighbor of mine and the lead engineer at the ESS Sabre division. He’s been over at my house several times listening to my system and sharing trade secrets. Probably the smartest guy I know.
1 Like
Thanks @MiveraAudio
That’s brilliant! I’m not sure I got the idea right, but understand DSD64 limitation is noise start to creep out after 22.05kHz, upsampling to DSD256 and above means noise shaped start to creep out much at higher i.e, 88.2kHz. Between 22.05kHz to 88.2kHz, if there are some music contents, it will not obscure by noise!
One may not need to go higher DSD sampling rate like DSD128 and above, file size is going to be a major problem. I find DSD64 is more manageable like 88.2kHz/24 FLAC.
Well with most 7th order software modulators and DSD 256, with my Purestream DAC the noisefloor is -120dB till 100khz, and peaks @-70dB up @ 180khz. From there it starts dropping. However in order to achieve this performance from DSD256, we needed to design to DAC to handle only DSD256. This is why we have a DAC that’s damn near the level of the $90000 MSB Select DAC 2 for only $3000.
However the Purestream has been discontinued. I find it shameful to only be 2nd best 
Hi,
Am I reading it correctly that you are saying Celeron J3455 is capable of upsampling to dsd256?
1 Like
During the brief testing I did in our particular setup with Parallelize Sigma Delta Modulator enabled, yes.
For reference:
https://www.cpubenchmark.net/cpu_lookup.php?cpu=Intel+Celeron+J3455+%40+1.50GHz
I would suggest whenever you upsample to turn this on…especially for DSD
What strikes me is all the technical analysis, but what does it sound like? In my experience the upsampling done by Roon is a worthwile addition. For the filters I prefer the “precise, minimum phase” on my DAC. This is all PCM stuff, but how does conversion from PCM files tot DSD sound like?
It sounds good if: a) Your DAC can deal with it natively and b) you like what it does. My overall impression is DSD can sound more relaxed but some hear that as losing some impact and urgency in the music.
3 Likes
I upsample everything on my main system to DSD256, so I like the way it sounds, but I chose my current DAC with the idea that I’d be doing this. I think that how much perceived improvement you get from doing this both depends on personal taste and is highly dependent on what DAC you’re using. In my experience, well-designed ESS-based DACs do well with upsampling to high-rate DSD, as do DACs that use proprietary designs that are optimized for DSD playback (e.g., T+A DAC 8 DSD, Holo Audio Spring DAC).
2 Likes
Thank you David and Henry for the clear response. I do own a DAC capable of native DSD, I’m planning on buying some DSD content. I do believe in a good sounding DSD music, but it should be DSD from the recording / production onwards like 2L and Pantone do. Otherwise it strikes me as strange to have a recording in PCM converted to DSD as a better sounding solution, but that’s me… Look at how PS Audio handles all music content as DSD/1 bit in there converters.
For now I’m just going to enjoy my music 
Oh, for the record, my DAC is a NOS design…
Right — filtering and upsampling to high-rate DSD in software is along the same lines as what PS Audio does in firmware/hardware in its DirectStream DACs. But without getting into the technical justification, as long as your DAC supports it, I’d urge you to give it a try and see what you think — it’s easy enough to do in Roon. If you don’t like what you hear, just go back to what you were doing previously.
Separately, if you’re on the lookout for good DSD recordings, check out Channel Classics.
1 Like
Will do that, the proof of the pudding is in the eating
Thanx for the tip!
NOS designed DACs tends to perform better when using Roon’s up sampling algorithm; this is because the internal over-sampling digital filters can be fully disengaged.
don’t be fooled by the NOS marketing jargoon: true NOS DACs do not have any internal oversampling filter
upsampling in software, with such DACs, lets you more easily shape your own filter (as you don’t have to, first, bypass the internal ones) 
1 Like
If one can bypass the internal over sampling digital filter doesn’t it perform like a NOS DAC? There’s no fooling here, its just switchable. Whether it is in a discrete or in IC form, if there are internal over sampling digital filters, all these can be fully bypassed.
Designers can have an option to use their own hardware or software over sampling/up sampling techniques. Now the DAC behaves strictly by converting digital to analog signals😃
The hole point is that you take out the oversampling calculations out of the DAC chip. But for me it just sounds better as a whole package: the complete DAC.
Not just the oversampling but also the digital filters which greatly influence the sound signature. Having a software programmed digital filters give user more flexibility.
Hi guys, I bought my old man an iFi iDAC2 to use with the Roon membership I bought him, so he can play around with DSP, both PCM and DSD up-sampling. The iDAC2 has bit-perfect mode for both PCM and DSD.
I’ve read plenty about about the pro’s and con’s of DSD up-sampling, like the higher you go, the further you push quantization noise into ultrasonic frequencies. But also phase noise can become a problem, which is where you need ultra low phase noise clocks.
But can someone explain the pro’s and con’s of PCM up-sampling, with a NOS type DAC? Are there any downsides with going higher and higher with PCM up-sampling rates, provided the DAC supports it?
I’ll tag a couple of experts @brian @jussi_laako
Many thanks in advance
Yes, you are right about clock, most DACs clocked max out around 22.579MHz (44.1kHz x 512) and 24.576MHz (48kHz x 512). Say if you up-sample to DSD512, you have to use 22.579MHz crystal clock with no division. Operating a clock without division has a higher chance of jitter and phase noise. The only way to reduce the higher sampling phase noise of the clock is to select a higher crystal clock frequencies, at least 2x then divide down. They other way is to select ultra low noise femto crystal clock oscillator.
1 Like
Yep plenty of stuff about DSD up-sampling.
But as above, what are pro’s and con’s of PCM up-sampling, with a NOS type DAC? Are there any downsides with going higher and higher with PCM up-sampling rates, provided the DAC supports it?
And what does going higher and higher in PCM sample rates achieve? With DSD it pushes quantization noises up and up but what about PCM up-sampling. Say up-sampling to PCM192 to PCM384, what is achieved with a NOS type DAC. And the potential downsides?
Of course ultimately it comes down to our ears and what sounds better/best. But putting that aside for one moment and just discussing the technical aspects.