Prosessing power of directstream dac fpga


#1

Just reading discussion thread about chord Dave dac in head.fi. Dave’s developer Rob Watts quite openly describes his design. Long thread but interesting reading…

Question: What is the model of Xilinx spartan-6 fpga used in direct stream dac.


#2

A much lower cost version: the LX16 - You don’t loose timing with lower oversampling ratios - the timing is preserved by virtue of sampling a bandwidth limited signal - you get back exactly the same timing of the signal if you have reasonable sample widths. Oversampling is only necessary to sample rate convert or to get some room for gentler filters. I oversample CD and other low rate PCM to 176.4 or 192k and then use a IIR (much more like analog filters than FIR filters) to go to the 640FS sample rate. Using DSD processing and output saves a lot of work and a lot of processing and outboard hardware compared to the multibit sigma delta he describes. Not that I don’t appreciate taking a design to the limit like they do.


#3

I’d buy the DS DAC alone for your communicative sovereignty and smartness, which as far as I supose is technically valid, too smiley-music005_gif


#4

Thanks Ted for your answer. I also like my DS dac. Thru firmware updates we all have seen continuous improvement of DS dac.

Theoretically speaking what would be the next steps to improve it from firmware point of view

how about and HW point of view (mark 2 105_gif)


#5

I’ve posted on the “Life after Torreys” thread (http://www.psaudio.com/forum/directstream-all-about-it/life-after-torreys/) but in short I’ve got three definite things to try, all three should be improvements, tho until I do them I won’t know if and how much…

One is more jitter/noise reduction for all input processing by moving some buffers around.

Another is a filter refinement that should improve low rate PCM a little.

And a little more speculatively: I don’t know if the Spartan 6 will support a high enough multiply rate, but if it does I might be able to lower the ultrasonic noise somewhat.

Since I’m still making noticeable improvements in sound quality with software and there aren’t any significant hardware problems it’s too soon to think about a DS Mk II. But since we’re talking about FPGAs: If I were designing the DS now I’d probably use an Artix 7 FPGA instead of the Spartan series (better optimized for cost/performance with about 1.5 to 3 times the computer power / cell and more cells, multipliers, etc. per $ cost…).


#6

That’s funny - we recently transitioned to the A50Ts at work for some of our designs.


#7

Ha…I see you’ve got the Diapasons. The first high end speakers I heard. It was in our music class in school over 30 years ago. They sounded great in the whole room with a symphony.


#8
Ted Smith said I've posted on the "Life after Torreys" thread (http://www.psaudio.com/forum/directstream-all-about-it/life-after-torreys/) but in short I've got three definite things to try, all three should be improvements, tho until I do them I won't know if and how much...

One is more jitter/noise reduction for all input processing by moving some buffers around.

Another is a filter refinement that should improve low rate PCM a little.

And a little more speculatively: I don’t know if the Spartan 6 will support a high enough multiply rate, but if it does I might be able to lower the ultrasonic noise somewhat.

Since I’m still making noticeable improvements in sound quality with software and there aren’t any significant hardware problems it’s too soon to think about a DS Mk II. But since we’re talking about FPGAs: If I were designing the DS now I’d probably use an Artix 7 FPGA instead of the Spartan series (better optimized for cost/performance with about 1.5 to 3 times the computer power / cell and more cells, multipliers, etc. per $ cost…).


When is the next firmware expected?

#9

Not soon for the FPGA - we haven’t planed for a particular date yet. (I can’t speak to any other patches or releases of, say just the bridge code…)