I had a discussion about clock subharmonics and DAC performance on the well-tempered master clock thread and also ran a few simulations, see post 2850 of that thread, https://www.diyaudio.com/forums/dig...itter-crystal-oscillator-285.html#post6373471 As expected, the effect of the -52.75 dBc spectral peak at 3/2 times the desired frequency was rather dramatic: you get an FM modulate with 0 Hz carrier added to the audio at a level of about 66 dB below the maximum Scarlet Book level, so whenever you listen to a soft sound, you get a frequency-swept sine wave through it.
Andrea Mori pointed out that he didn't design his clock generator with sigma-delta DACs in mind (just like I didn't design the valve DAC with frequency doublers in mind), and that the subharmonic performance can be improved as much as you like by improving a filter between the frequency doubler and the slicer (or squarer, as he calls it) that converts the signal to a square wave. When you manage to reduce subharmonics to -100 dBc or so, the FM modulate is well below the noise floor.
I would still prefer to avoid the whole issue by not using an oscillator that runs at a subharmonic of the desired clock frequency at all, even if that makes the close-in phase noise a bit worse, but if anyone would like to try it, it should work well with the extra filtering.
Andrea Mori pointed out that he didn't design his clock generator with sigma-delta DACs in mind (just like I didn't design the valve DAC with frequency doublers in mind), and that the subharmonic performance can be improved as much as you like by improving a filter between the frequency doubler and the slicer (or squarer, as he calls it) that converts the signal to a square wave. When you manage to reduce subharmonics to -100 dBc or so, the FM modulate is well below the noise floor.
I would still prefer to avoid the whole issue by not using an oscillator that runs at a subharmonic of the desired clock frequency at all, even if that makes the close-in phase noise a bit worse, but if anyone would like to try it, it should work well with the extra filtering.
Last edited:
Maybe of interest?
https://www.diyaudio.com/forums/dig...-simple-dsd-src-beaglebone-4.html#post6378889
Ray
https://www.diyaudio.com/forums/dig...-simple-dsd-src-beaglebone-4.html#post6378889
Ray
That depends on what you want to do with it. Are you mainly interested in PCM or in DSD? What do you want to connect to the DAC?
Except for what I wrote about them in appendix A of https://linearaudio.net/sites/linearaudio.net/files/03 Didden LA V13 mvdg.pdf , there is nothing I can post on a public forum without getting sued for violating someone's copyright. The details are in references 38 and 47:
R. H. Wilkinson, "High-fidelity FIR filters based on central-difference operators", IEE Proceedings on Circuits, Devices and Systems, vol. 141, no. 2, April 1994, pages 111...120
R. H. Wilkinson, "High-fidelity finite-impulse-response filters with optimal stopbands", IEE Proceedings-G, Circuits, Devices and Systems, vol. 138, no. 2, April 1991, pages 264...272
I don't use DSD, so PCM would be great. I use a Windows PC playing flac on MusicBee over USB, sometimes optical. Really I just love DIY projects (I've done a Beta 22, Gamma 3, and TU-8800) and this seems really awesome and unique. I would use it as an alternative to the Gamma 3 for fun.
If Pavel has no objections, I am considering running a small quantity group buy for assembled boards for his simple DSD SRC. If it's of interest to you please post on his thread;
https://www.diyaudio.com/forums/dig...-simple-dsd-src-beaglebone-5.html#post6380243
Then you are right. The original valve DAC supports PCM and has four inputs that can be used as S/PDIF, AES3 or optical inputs, depending on what circuit you connect to them. With version 2.1 of the FPGA configuration file it also has an I2S input, so you could connect a USB to I2S interface to it.
Mind you, it is far more expensive and complicated than the raw-DSD version.
Marcel's PCB is 4-layer and uses a custom layer stack as a fundamental part of its design. When Dan and I decided to build the DSD version the best price I could find was for a minimum order qty of 5 from PCBway and that order cost us over £400 IIRC, that's the wrong side of $500 I guess. I think I'm right in saying that the PCB design for the original version is bigger than that for the DSD version too.
Incidentally, I, and I'm pretty sure most of the others who plan to build the DSD version, mostly play PCM files; in simple terms the sample rate conversion elements of Marcel's original version have been moved 'upstream', in my case to PC based software (HQPlayer) or, perhaps, something like the SRC board that ppy has developed - see my post a few previous to this one.