Dear Paul,
Any opportunity and possibility for Direct Stream DAC to be supported DSD 256 via firmware upgrading in future?
Thx
Raman
Ted has answered this question a few times, in some detail. Short version: not really, no.
The long version includes the fact that the internal processing in the DirectStream DAC happens at 10xDSD rate which is not an even multiple of DSD256 (4xDSD), which greatly complicates the processing. All other supported input rates are integer divisors of the 28.224MHz upsampling frequency.
1 Like
I hope Ted and Paul reconsider their position one day.
I am starting to read some observations that DSD 512 is another league… that’s why T+A Dac8 that does it is getting more popular.
Is there anything available recorded at octaDSD? Who is selling it?
My understanding is the current DSD architecture is unable to accommodate this rate, not that Ted and Paul have taken a position one way or the other. If the rate becomes popular I’m sure PS Audio will support it.
1 Like
I really don’t see how we could go beyond 2xDSD in the current implementation.
The way that the DS DAC works is by sampling ANY incoming signal at 10xDSD - which allows the DS DAC to ignore the incoming signal’s clock and is one of the ways jitter is minimised.
The FPGA then looks at the sampled data and then correctly decides if it is 44.1, 48, 88.2, 96, 176.4, 192, DSD 2.8, DSD 3.1 or DSD 5.6 and processes it correctly.
Note that 10x DSD is a common multiple of those signals. This way the sampled data will be able to capture ALL the information in those sample rates above. But it won’t work with DSD6.2 because thats DSD10x is not a common multiple of that.
For it to process 4xDSD or higher, you’d need to have the incoming signal sampled at 20xDSD (if my maths is right) and I don’t believe the processor or clock is rated for that.
Elk said Is there anything available recorded at octaDSD? Who is selling it?My understanding is the current DSD architecture is unable to accommodate this rate, not that Ted and Paul have taken a position one way or the other. If the rate becomes popular I’m sure PS Audio will support it.
Possibly no recordings yet but over at CA folks are overwhelmed with the HQP upsamling to DSD512 with T+A DAC8 DSD which is discrete 1 bit chipless DSD implementation. For some reason or the other, anybody who is auditioning or experimenting with T+A DAC and HQP upsampling to DSD512 is becoming a convert and agrees that the sound quality is anything they have heard, sometimes even beating a good vinyl rig.
http://www.computeraudiophile.com/f10-music-servers/sound-galleries-music-server-25805/
Note: these are very long threads, so it might take some patience to filter things out. It just happened that I have been following these for a while now.
You are not alone following these threads 
However, I am very reluctant to buy a PC accommodating HQPlayer (or the dedicated and pricy SGM server for that matter).
Frode (aka Distinctive at the CA forum)
Could DSD256 input just be multiplied by 2.5 to generate the 10x DSD rate for processing? The XMOS USB input receiver can handle DSD256 I believe.
Dsd 256 is a bust. I have a bunch of tracks and while there is something there is minor at best and this is with large speakers with headphones and expensive ones I could say I heard it at all. Iknow I am old and maybe goin deaf but dsd 256 for me and I think in the near future is a bust.
While on this has anyone tried the water marked tracks I posted.
Could DSD256 input just be multiplied by 2.5 to generate the 10x DSD rate for processing?You can't have half of a bit.
EldRick saidCould DSD256 input just be multiplied by 2.5 to generate the 10x DSD rate for processing?You can’t have half of a bit.
I don’t mean the bits, but the sampling frequency. Ted converts everything to 10x the DSD64 rate before processing, which is 2.5x the DSD256 rate. So for those inputs on the Directstream that can handle DSD256, could the multiplier not be 2.5 to get the 10x rate? (I realize it would be then downsampled to DSD128 for processing in the Directstream. But who knows, maybe in some cases it would sound better than the same file in DSD128.)
Still saying we handle DSD 256 (or higher) and sending DSD 128 to the hardware is like claiming you have a 192k capable DAC and sending 96k to the DAC chip… We can probably do higher sample rates with enough work, but at what cost to the sound quality of the rates we handle now? Higher sample rates will definitely add more jitter, noise and THD to the outputs. Whether this sounds better to people or not isn’t obvious, but it would definitely sound different. In any case I’d not expect it the near term (or even medium term.)
2 Likes
Minimalist saidDSD is 1 bit per sample, so bit == sample. You cannot have half a sample and cannot have half a bit. As I remember, the calculations in the DirectStream DAC are in the integer domain and the sample rate conversion is synchronous rather than asynchronous. You cannot have a fractional sample or bit... period.EldRick saidI don't mean the bits, but the sampling frequency. Ted converts everything to 10x the DSD64 rate before processing, which is 2.5x the DSD256 rate. So for those inputs on the Directstream that can handle DSD256, could the multiplier not be 2.5 to get the 10x rate?Could DSD256 input just be multiplied by 2.5 to generate the 10x DSD rate for processing?You can't have half of a bit.
Ted could recode the DAC to upsample to 20x DSD rate, but that would be a major update to the DAC “design” and has significant probability of degrading the performance for all lower rates, just to accommodate a format that is in its early days and has not proven itself to be clearly superior and of a significant percentage of sales against the currently common formats. Perhaps DSD256 will prove itself, become a commonly distributed format and force Ted’s hand in accommodating it in the DirectStream. Only time will tell.
J.P.
Ted Smith said Still saying we handle DSD 256 (or higher) and sending DSD 128 to the hardware is like claiming you have a 192k capable DAC and sending 96k to the DAC chip... We can probably do higher sample rates with enough work, but at what cost to the sound quality of the rates we handle now? Higher sample rates will definitely add more jitter, noise and THD to the outputs. Whether this sounds better to people or not isn't obvious, but it would definitely sound different. In any case I'd not expect it the near term (or even medium term.)
I’m happy as it is. I’ve got around 2000 albums (no junk stuff, just what I need) of which 1/3rd to half are hires in the meantime, of which maybe 200 are single rate DSD. Of just two of them a higher DSD rate exists at all, not knowing if they really were recorded in that rate.
Bridge replay in single rate sounds better than USB double rate anyway in my case (without intensive tweaking of the USB connection).
So it’s most important to me, that what I really have sounds best.
Nevertheless I’m quite sure, higher DSD rates might get usual at least for selected labels during the next years, but I don’t count the amount of relevant music, to be expected within a few years, as essential. And storage demand is huge.
Whenever it will really make sence, I count on PS Audio to offer a good trade in program 
Dev saidElk said Is there anything available recorded at octaDSD? Who is selling it?My understanding is the current DSD architecture is unable to accommodate this rate, not that Ted and Paul have taken a position one way or the other. If the rate becomes popular I’m sure PS Audio will support it.
Possibly no recordings yet but over at CA folks are overwhelmed with the HQP upsamling to DSD512 with T+A DAC8 DSD which is discrete 1 bit chipless DSD implementation. For some reason or the other, anybody who is auditioning or experimenting with T+A DAC and HQP upsampling to DSD512 is becoming a convert and agrees that the sound quality is anything they have heard, sometimes even beating a good vinyl rig.
http://www.computeraudiophile.com/f10-music-servers/sound-galleries-music-server-25805/
Note: these are very long threads, so it might take some patience to filter things out. It just happened that I have been following these for a while now.
Indeed hqplayer Upsampling to dsd512 with t+a dac 8 dsd sounds very nice. Detailed and solid, more meat on the bones. Very very much analog like. I would say no digital artifacts at all. Ds dac has more Air and micro detail. More 3D soundstage. More analytical in good way. Both are excellent but with different sonic characteristics.
Setup: NAS+roon server+hqplayer on ASUS rog laptop win10+upsampled to dsd512 with t+a, no USB tweaks.
If time allows I would like to try hqplayer+ T+A with high quality NAA device to get rid of possible RFI/EMI issues.
In my opinion hqplayer is very nice software and with it’s high quality filters it provides excellent results. Only downside is that it is very CPU hungry. E.g. My ASUS ROG laptop with i7 CPUs and NVIDIA 980M GPU (hqplayer can utilize CUDA cores), CPU utilization was around 50% when up sampling to dsd512 (poly-sinc-2n+asdm7)
Hi, I am reviving this mid-2016 thread to ask if anything has changed since then which would make DSD 256 (aka 4xDSD and DSD 11.2 MHz) more feasible in the DS?
Two years ago, I would have scoffed at 4xDSD as numbers racket overkill. But lately I have become interested in the recordings offered by the 2xHD audiophile label (no affiliation), which transfers analog recordings to digital at 4xDSD and sells digital downloads in the native 4xDSD format as well as various down-sampled formats, including 1xDSD and 2xDSD. (I purchased a 2xDSD version of this 1968 Bill Evans Trio concert recording recently and it is spectacular, both musically and sonically).
2xHD describes its mastering chain as follows:
The mastering chain consists of a selection of hi-end vacuum tube equipment which includes a modified vacuum tube NAGRA HD dac (DSD) to make it possible to obtain a real analog sound from a digital master. The 2xHD FUSION mastering system is powered by a super capacitor power supply, with a new technology that lowers the digital noise in which is found in the lowest level, something that was not possible before. The resulting signal is then transformed into high resolution formats by recording it in DSD11.2kHz using Merging Technologies’ Horus A to D converter, which is also powered by a super capacitor power supply. All analog and digital cables are the most state of the art available today.
http://www.2xhd.com/technology.html
It would be interesting to experience 4xDSD on the DS and possibly compare 2xDSD and 4xDSD versions of the same recording. I recognize that not being able to play 4xDSD is a classic “First-World problem,” and if I never hear 4xDSD that would be okay. I am just wondering whether the advances the DS has undergone since 2016 could make 4xDSD a possibility. Thanks.
We are closer to 4X, essentially all of the FPGA code runs at 4x (a few input routines aren’t tested at 4x, but all of the audio processing is at 4x.)
On the other hand since PS Audio allows customers to use whatever source hardware and software they want, getting all this hardware and software working for 4x is a problem. If PS Audio required certain programs, specific drivers, only PS Audio I2S hardware, etc. we could get everything to talk together much sooner.
S/PDIF and AES/EBU top out at 192kHz which is only enough for double rate DSD…
Many I2S sources are produced by dividing down a 100MHz clock (especially those that convert from USB), Raw DSD at 4x is a 45.1584MHz bit rate and 100MHz is too low to generate the I2S waveforms reliably (and DoP would be 50% more bits.) (The DS’s I2S inputs aren’t tested at 4x but they should be fine.)
USB Audio 2.0 (which Microsoft just starting to support) requires 8k USB blocks per second, but those blocks have a maximum size which won’t hold the number of bytes needed for the 4x data at that rate. There is a newer USB Audio (3.0) standard that extends this block size, but all drivers will have to be updated, and worse, probably firmware in any XMOS chips (or other USB processing chips) that are in any hardware on the path to the DS (including the XMOS in the DS for that matter.)
So getting 4x to the DS is the real problem now.
That doesn’t mean that PS Audio doesn’t care about 4x or isn’t working on 4x, but it’s an uphill battle.
1 Like
Thanks for the detailed response Ted! Can’t say I completely understand it all, but when the DS first came out your position on greater than 2x DSD required hardware changes to achieve. Then a few months back you made some comments that, at least to me, inferred that greater than 2x DSD could possibly be done via software. Now with your response above it’s a little ambiguous, but I’ll take a leap and say we’re back to hardware updates (or non user performed software updates) to get there. Is that correct? Would also presume if it is a hardware change, no matter what level, it would be a new model DS.
Perhaps it could be implemented in the new Octave server(s), particularly the standalone unit (rather than Bridge III, or whatever we are now calling that, as it might have more limitations)? That all-in-one unit could store the files on internal (or external) hard drive and transfer to the DS over I2S.
OMG if I had more files on DSD256, I had no idea how to store them. I already have more than 8TB now…this would easily double having only up to 1/4th in DSD256.
But anyway one thought:
assumed Bridge III could handle DSD128…would its USB input be able to route through DSD256 or would it have the same limitation?