Simple DSD SRC for BeagleBone

^ Not exactly.

Valve dac FPGA does upsampling but not exactly ASRC, and certainly not conversion of PCM to DSD. In fact, some of the guys building those dacs are interested in the AK4137 boards of this thread so they can convert PCM to DSD256.
 
Last edited:
On the first page it says about the original valve dac: "Supports DSD64 and DSD128 by first converting them to PCM."

The it says for the raw DSD version: "Designed for raw DSD; it doesn't need any FPGA or SRC4392, because the raw DSD signal is already a sigma-delta modulate"

Goes on to say: "Yet another approach could be to use an AK4137 board for the conversion from PCM to a sigma-delta modulate, but no-one has tried that yet."
 
Last edited:
Pavel ...

Do you know if someone can do the same job as AK4137 with FPGA?

From the designer of the valvedac,

Ray uses the raw DSD version with a computer running HQPlayer, so he can use the raw DSD version and still play PCM. Basically he uses the computer to do what my FPGA board does, although the details of the used algorithms are undoubtedly not exactly the same.

Yet another approach could be to use an AK4137 board for the conversion from PCM to a sigma-delta modulate, but no-one has tried that yet. Compared to the FPGA board and SRC4392 used in the original valve DAC, it should be much cheaper, but it is also much less flexible; with the original valve DAC you can change the Verilog code and build in any filter curve you like, any sigma-delta algorithm you fancy or any amount of headroom for intersample overshoots that you want, provided you have enough knowledge of digital signal processing and Verilog code to do so.

I am currently building the raw DSD version and plan to experiment with the SAA7350 as DSD source.
 
The original valve DAC uses an SRC4392 to asynchronously convert signals to 200 kHz sample rate, the rest of the digital signal processing is done with an FPGA board. The whole digital path is as follows:

DSD64 and DSD128: decimated to 201.6 kHz PCM by the FPGA, converted to 200 kHz PCM by the SRC4392, interpolated to 3 MHz PCM by the FPGA and then converted to a 27 Mbit/s sigma-delta modulate by the FPGA. You have a choice between three different algorithms for this last step. The filter used in the first step is one with a short impulse response.

PCM: prefiltered with or without 2 times interpolation by the FPGA (to work around some limitations of the SRC4392), converted to 200 kHz PCM by the SRC4392, interpolated to 3 MHz PCM by the FPGA and then converted to a 27 Mbit/s sigma-delta modulate by the FPGA. You have a choice between three different algorithms for this last step and three different filter types (steep, apodizing or no prefiltering) for the first step.

DSD is a Sony/Philips trade name for a sigma-delta modulate, so in the DSD case, a sigma-delta modulate gets converted to PCM and then back to a sigma-delta modulate. I chose this solution to be able to pass it through the SRC4392 asynchronous sample rate converter.

The reason for using asynchronous sample rate conversion was that I wanted the DAC to be able to synchronize to an incoming signal, as required for S/PDIF, Toslink and AES3 signals, and couldn't make a large enough tuning range with my crystal oscillator.

I haven't a clue if it sounds better, worse or exactly the same as an AK4137.
 
Last edited:
Don't know if anyone else here tried it, but I compared CT7302 DSD modulator to AK4137 and I can't tell you bad the 7302 sounded. I tried working with their factory contact, but they didn't seem to understand or care that it was unusably bad.

Regarding the sound of software DSD modulator algorithms, one can easily try some of the various modulators in the trial version of HQ Player. They all sound different.

Thus, I personally doubt that two different modulator designs are likely to sound the same.

Also, DSD only starts to get really good up at DSD256 (or higher). For most DACs a DSD256 modulator must clock out at 11.2896Mhz, or less commonly at 12.288MHz. Don't know about the Tube dac modulator?
 
Last edited:
Don't know if anyone else here tried it, but I compared CT7302 DSD modulator to AK4137 and I can't tell you bad the 7302 sounded. I tried working with their factory contact, but they didn't seem to understand or care that it was unusably bad.

Regarding the sound of software DSD modulator algorithms, one can easily try some of the various modulators in the trial version of HQ Player. They all sound different.
Please post the measurements so that we would have easier time understanding what you say.