I’m auditioning the DSJ. At about 100 hours of breaking and will soon put it to the test. Am coming from the SGCD.
I’m a Roon/ HQPlayer user and seek a better grasp of how all of this integrates with the DSJ.
I understand, quite sadly, that the Bridge 2 does not provide NAA support, so no HQPlayer support. An odd decision on PS Audio’s part given most audiophiles (if Computer Audiophile is any indication) prefer HQP with Roon.
My question is two fold:
The DSJ upsamples everything to DSD 64 over the Bridge, along with 20 times DSD (not quite sure what the 20 times part means). Does the DSJ do all of this regardless of the software source? Or if it receives an already upsampled DSD signal from my Mac (either Roon over the Bridge 2 or Roon/HQP via USB), does that negate whatever upsampling the DSJ does?
Or does the DSJ do its upsampling thing anyway?
Any thoughts on the merits of Roon via the Bridge 2 vs Roon/HQP via USB?
I like the idea of the Bridge 2 serving as a MicroRendu/ SOTM type device.
I’m a DSJ owner and use Roon/HQPlayer via usb.
I’ve compared Roon>usb and Roon>Bridge II and prefer Roon/HQP>usb. FWIW I use either the Poly-sinc or Poly-Sinc-xtr-2s filter and convert all sources to DSD128 in HQP before sending to the DSJ.
No matter what you send the DSJ it will convert to DSDx20. From what I understand it does this because it uses a single clock and all allowed input rates - PCM up to 352.8 and DSD64/128 - are divisors of this clock. So it won’t make any difference which of the specified rates you send, they will all get the same treatment - in theory anyway. But like I said, with Roon/HQP I usually prefer the sound when upsampled to DSD128 (DoP of course with the DSJ) - however sending straight PCM sounds mighty fine too.
I also use Audirvana with the iZotope filter >usb - with that I typically use no oversampling/conversion, just send the native rate to the DSJ. Sounds great too.
There is an endless amount of tweaking you can do with the filters in these computer music players, in the end it comes down to personal preference.
20 times the nominal DSD rate is 10 x as fast as double the DSD rate (DSD 128). The nominal DSD rate is DSD 64 which is 64 * 44.1k. DSD 128 is double rate DSD or 128 * 44.1k.
20 x the nominal DSD rate is 56.448MHz. The DS upsamples everything to 30 bits at 56.448MHz then converts it to double rate DSD (one bit at 5.6448MHz) for final output. This allows small integer upsampling of 44.1k, 88.2k, 176.4k, 352.8k, 48k, 96, 192k, 384k, etc. (by 1280 x, 640 x, 320 x, 160 x, 1176 x, 588 x, 294 x, 147 x, etc.)
You aren’t saving the DS any work by upsampling earlier and the DS has much more computing power available for upsampling than your CPU, so upsample if you like the sound better (or like experimenting with various upsampling parameters) but there’s no technical reason upsample before the DS.
The terminology is confusing: I’ve tried to always use the phase “20 x the nominal DSD rate” which is 20 x DSD 64 which is 10 x DSD 128 which is DSD 1280. double rate DSD (64) is DSD 128. quad rate DSD is DSD 256… [Sorry The silly thing that does links keeps changing my q u a d into q u a d d…]
The DS upsamples any input it gets to the specific rate of 56.448MHz (at 30 bits of resolution). It doesn’t matter if the input is 32k PCM … DSD 128 it gets upsampled to 56.448MHz. This preserves all of the bits of accuracy of PCM (and DSD) using the 30 bits and all of the high sample rate from DSD (and, of course PCM) since it’s a direct multiple.
Wow. I think that would be a more clear way to explain it in your marketing. I’m an average audiophile, I know a lot of stuff, but I’m not a tech genius by any means. The whole time I read about “20xDSD”, I really had no idea what it practically meant from the PS Audio descriptions. I read every DSJ page on the website, still had no idea when it meant.
The reason is I’m conditioned on a more simple level. I know DSD 64, DSD 128, DSD 256 and DSD 512. Anything explained outside of those parameters, is confusing to me. And I assume confusing to lots of people.
But when you say “30 bits of DSD 1280” - bingo, I totally get it. I understand that is remarkable, and exceeds most DACs in the world. The reason it makes sense to me, is its presented in terminology that’s familiar. Its in the same language as the audio files I purchase.
@tedsmith I probably missed this when the firmware change upped the sampling rate to DSDx20 from DSDx10, but this means the diractstream dacs now support 384KHz pcm?
I note the specs on the PSA webpage have not been updated, but this would make sense, since 384 is now an even divisor of 56448.
Tho it could in principle, as I look at the code there are some places that don’t have enough “arms in a switch statement” - there’s no defined encoding for 384kHz so various places in the code might make different assumptions about what happens with 384kHz sample rates. Perhaps I can make supporting 384kHz an explicit feature of the next release (or find out why it isn’t there )
Thanks for the response Ted.
I don’t specifically require it, I thought support for 384KHz might have resulted from the doubling the sampling rate to DSDx20.
I certainly appreciate your continued development of the Directstream code though. Keep up the great work.
Yes the DS still goes to 30 bits @ 56.448MHz then back to quad rate DSD. The specific details like 30 bits vs 31 or whatever might change from release to release.
The TSS will be similar but with a few more bits here and there, in particular in the SDM (the piece that converts many bits, very fast PCM to DSD.) In particular the TSS will have a volume control that will allow you to “go to 11”, you’ll be able to add up to 24dB with the volume control for quiet source material so there will be more bits to handle that.
All inputs are converted to 30 bits at 56.448MHz, then that’s converted to 30 bits at 11.2896MHz , then the volume is applied which results in 50 bits at 11.289MHz, which is then converted by the SDM to 1 bit at 11.2896MHz.
This is true for all inputs, DSD or PCM.
Multibit SDM is using a sigma delta modulator (noise shaper) to narrow many bits to fewer bits, but still more than one bit. The DS goes to one bit since any more bits need more than a low pass filter to convert to analog which reduces the linearity possible during that conversion.
