With PCM2DSD downstream neither Amanero nor I2SOverUSB is optimal. It would be better to place the isolators and reclocking after PCM2DSD so the isolators and reclocking in I2SOverUSB are misplaced in this case. If both Amanero and I2SOverUSB are using similar power supplies and clocks I doubt there will be much difference when feeding PCM2DSD. Without PCM2DSD (i.e. Amanero/I2SOverUSB outputting DSD) l2SOverUSB may work better than Amanero.
Using multiple separate boards connected with wires and pin headers is understandable in testing and breadboarding but it also is susceptible to noise pickup (e.g. EMI/RFI). For best results it makes sense to have more integrated boards using solid ground planes.
Using multiple separate boards connected with wires and pin headers is understandable in testing and breadboarding but it also is susceptible to noise pickup (e.g. EMI/RFI). For best results it makes sense to have more integrated boards using solid ground planes.
It does make sense. However Marcel's dac is already two boards connected by pin headers, and it does not have integrated power supplies. All the problems with building a one-off dac with multiple boards can be solved. In fact, most of the solutions are shown and or described in this thread. Gentlevoice described a master clocking scheme, and I showed how to reclock and galvanically isolate the I2S bus just before it connected to Marcel's dac. Radiated EMI/RFI coupling was controlled using steel shield walls. A pic can be seen at: https://www.diyaudio.com/community/threads/return-to-zero-shift-register-firdac.379406/post-7430757
That said, what I found is that eventually I got down the point where IMHO the main remaining problems were with the dac itself. The dac wasn't here long enough to see what could be done about those things. The final consensus opinion of the people who listened to the dac here was that it had a nice sense of space and soundstage, but otherwise sounded a little raw. However, that is as compared to best we have heard here so far. Marcel's RTZ dac may be the best dac most people will ever have heard if most or everything known to do in the thread so far were to be implemented. In my case that included ultra-low phase noise clocks, lots of isolated power supplies, multiple steel chassis, etc.
Is it possible to do better? Sure. You can buy the stuff to do it now if you want to and if you can afford it. Otherwise its possible to work on getting the best out of Marcel's open source RTZ dac, for which there are various options at various levels of cost and complexity.
That said, what I found is that eventually I got down the point where IMHO the main remaining problems were with the dac itself. The dac wasn't here long enough to see what could be done about those things. The final consensus opinion of the people who listened to the dac here was that it had a nice sense of space and soundstage, but otherwise sounded a little raw. However, that is as compared to best we have heard here so far. Marcel's RTZ dac may be the best dac most people will ever have heard if most or everything known to do in the thread so far were to be implemented. In my case that included ultra-low phase noise clocks, lots of isolated power supplies, multiple steel chassis, etc.
Is it possible to do better? Sure. You can buy the stuff to do it now if you want to and if you can afford it. Otherwise its possible to work on getting the best out of Marcel's open source RTZ dac, for which there are various options at various levels of cost and complexity.
Last edited:
Yes, all the problems can be solved and much better than shown in your picture.
Thanks Mark, this is a big thread so I'm trying to collect as much as I can together to try out . BTW did the Marcel dac you try have NE5532's in the active filter stage instead of the OPA1678's ?
I have a Mini XMOS XU208 usb digital interface with spdif out and i2S which I used with another dac , maybe worth trying it with a HDMI to i2S receiver into the pcm2DSD . Its run from an external psu
Attachments
It did, but you could very clearly hear the noise floor go up when you touched the coaxial cable that carried the data signal - touch the insulating sleeve around the shield, that is. I decided to go for a four-layer printed circuit board for the final design, to get better and more reproducible shielding.
It also taught me how not to design a valve-based crystal oscillator. I originally wanted to make an oscillator that could be tuned with a reactance valve (which is not a special valve, but a simple circuit) to make it lock to an S/PDIF, Toslink or AES3 source, but never got it to work over the frequency tuning range required for that.
It also taught me how not to design a valve-based crystal oscillator. I originally wanted to make an oscillator that could be tuned with a reactance valve (which is not a special valve, but a simple circuit) to make it lock to an S/PDIF, Toslink or AES3 source, but never got it to work over the frequency tuning range required for that.
@PJotr25 wanted to try the RTZ shift register DAC in combination with the digital part of my valve DAC, so he designed a board with the digital part of the valve DAC and a crystal oscillator, see
https://www.diyaudio.com/community/threads/delta-sigma-modulator-by-marcelvdg.409353/post-7603492
https://www.diyaudio.com/community/threads/delta-sigma-modulator-by-marcelvdg.409353/post-7603492
Hopefully tomorrow I'm going to try Marks idea of 1uf across C18/45 (15pf) and pair of parallel 1meg resistors from pin2 of U8/15 (opa210) to ground
Unfortunately I have no way of measuring it so can only go by ear
I'd also like to try 120R across each of the shift register bypass caps , would the onboard regulator be ok to run this ?
Did anybody else try the Nazar-style shunt regulator mod posted by Marcel #1,865
Unfortunately I have no way of measuring it so can only go by ear
I'd also like to try 120R across each of the shift register bypass caps , would the onboard regulator be ok to run this ?
Did anybody else try the Nazar-style shunt regulator mod posted by Marcel #1,865
I measured the temperature of the pass transistors in free air. With the 120R resistors to ground, the transistors were at about 50 degrees C, which I thought to be within reason.
What's that about? The mod I did to slow down the common mode servo was more less as follows: Basically C18 in the schematic below becomes 1uf. R36,R48 each become 1 meg. To make it reversible, I add pin headers so that I could put 1uf in parallel with the 15pf C18. Then I cut traces and installed pin headers so I could put 1M in series with each of R36 and R48.
Last edited:
Ok thanks Marcel, I'm just looking at things to try it would be nice to keep the thread for your dac active
I agree it would be nice to keep the thread active. Also IMHO there are some useful things that can be learned from modding the dac. It can be made to sound even better using standard commercial components. There may not be published literature on all of it, but that's were we are with this type of discrete resistor dac. A lot of what we don't know is what can we do, given what we know how to make with electronics, so that the end result is optimized for perception of an average human. I say this knowing that there are better sounding DSD RTZ dacs than this one. But those are proprietary boards. There are some proprietary boards can be used for diy projects (and I am about to get one of two new prototypes here from Italy
, in case anyone is interested in that). However, what we need here is the forum for people seeking a lower cost way to get better sound is for more people to try the mods that have been suggested for Marcel's dac and see if there a consensus that certain modifications make the music listening experience feel more accurate/realistic to the human auditory system. IMHO, that's useful knowledge that could be developed right here in this forum if people were willing to contribute a little work and report back results. And people who make their system sound better will be rewarded by having a better sounding system, plus know they are contributing to the group.As a side note, at some point I am probably going to do some proprietary R&D on Marcel's dac. I may be able to share what I find with Marcel, but probably can't share details publicly if there are positive results. What I am saying is that I am serious about Marcel's dac having some potential for improvement. Hint, hint. That's all.
Last edited: