Don't know.
However, do know the one big thing about Marcel's dac already. Its a trick other people in the forum have used before, and it goes against theory. Suggest to try listening to only the positive phase of each analog output in unbalanced mode. How does it compare to the sound of balanced? Which do you suppose is closer to the vinyl?
It sounds bright and distorted, but why? IME its because now you are clearly hearing all the stuff that needs fixing. Fix it all and you will likely have something very good.
However, do know the one big thing about Marcel's dac already. Its a trick other people in the forum have used before, and it goes against theory. Suggest to try listening to only the positive phase of each analog output in unbalanced mode. How does it compare to the sound of balanced? Which do you suppose is closer to the vinyl?
It sounds bright and distorted, but why? IME its because now you are clearly hearing all the stuff that needs fixing. Fix it all and you will likely have something very good.
Yours has an ATSAM3U1C, the other has ATSAM3U2C. Only flash memory size differs.
https://www.microchip.com/en-us/product/ATSAM3U2C vs https://www.microchip.com/en-us/product/ATSAM3U1C
Here is a pic of the way the dac is hooked up now:
Power supply is set to +-15v (with knobs taped to make sure it doesn't get bumped). Regulators are LM317, LM337, and LM7805. All share a common ground which is then grounded to the AC power line ground. USB board is USB powered. Its all sitting out in the open. This will represent about the worst case setup. How does it sound this way? Not very good, really. Sounds are blurred together; a lot of detail is missing—veiled over, one might say. However some people might not even notice something is wrong if that's the kind of sound they are used to.
Next go round will be to start fixing some of the obvious issues with the setup.
Power supply is set to +-15v (with knobs taped to make sure it doesn't get bumped). Regulators are LM317, LM337, and LM7805. All share a common ground which is then grounded to the AC power line ground. USB board is USB powered. Its all sitting out in the open. This will represent about the worst case setup. How does it sound this way? Not very good, really. Sounds are blurred together; a lot of detail is missing—veiled over, one might say. However some people might not even notice something is wrong if that's the kind of sound they are used to.
Next go round will be to start fixing some of the obvious issues with the setup.
You are right, I remembered it wrong. The bitclock is sandwiched between two sigma-delta modulates on the headers, but not in the ribbon cable. Nevertheless, my flatcable had better shielding, as the pin-out of the connector I used was
SDinL ground
SDinR ground
ground ground
bckin ground
see schematic DAC3_5.pdf. It was replaced with three U.FL connectors in later versions.
I expect the effect of sigma-delta modulate to clock crosstalk to be too subtle to see anything with a scope.
What DSD rate were you using and what software is on your computer to play the .dsf files.
Hans
Just moved Amanero to clean, isolated +5v, instead of USB power. Dramatic improvement in SQ. Most of the veiling is gone. More low level detail and accuracy.
Playback at the moment is still with SACD rips (DSD64).
EDIT: Now that the USB board power is fixed, other problems are becoming less masked. There is a bright, sort of metallic edge on the sound. Part of that is probably from the +-15v ground being shared with the +5v digital ground going back to the AC grounded power supply. More isolation of power and grounds may provide further improvement. We'll see.
EDIT 2: Still listening in single-ended output mode, BTW.
Playback at the moment is still with SACD rips (DSD64).
EDIT: Now that the USB board power is fixed, other problems are becoming less masked. There is a bright, sort of metallic edge on the sound. Part of that is probably from the +-15v ground being shared with the +5v digital ground going back to the AC grounded power supply. More isolation of power and grounds may provide further improvement. We'll see.
EDIT 2: Still listening in single-ended output mode, BTW.
Last edited:
Curious result, I listened to it single-ended and I heard no brightness or distortion, in fact, in spite of being on a breadboard, the sound quality compared very favourably to my Valve DAC, which uses balanced outputs, turned single-ended via Lundahl transformers - the valve DAC is a tad richer but I felt the RTZ revealed a little more detail and both are amazingly good IMO. These observations were reinforced by fellow UK-based audio DIYers when I took Marcel's original protoype to an audio meet, at which I also demonstrated a Valve DAC - the result is that two of those DIYers now own RTZ DACs. That said, as I reported earlier, I don't own an Amanero and wouldn't contemplate buying another one as the JLSounds board is superior IMO.
When I demo'ed the RTZ DAC at the UK meet I used Native DSD source files, mostly at DSD256.
I can sample at 5GHz using an active probe with spring ground and capture a one-shot of quite a bit of digital data. If there are glitches, runt pulses, etc., there is a good change to catch them. Also, its possible to experiment with the zero ohm I2S input resistors on the dac board. If there is crosstalk sufficient to cause what, a bit error, then something should probably show up. OTOH if some increase in data pulse jitter due to cross-talk is a concern, then that could be another matter. Or, maybe I'm missing the point of what the mechanism of concern might be?
Could be. I was listening to it after it had only been running a little while. It was going through some changes in sound, which IME is not abnormal for a new device and or with a USB board that hasn't been powered on for a few years. Also depends on the playback system. The Neurochrome HP-2 and Audeze LCD-X don't make for a very forgiving combination.
As far as amazingly good, as we move along here and keep making improvements to the setup we'll see how amazing it gets. The reason I started out with a poor setup was to chronicle how changes in setup affect the sound. When everything is done as well as possible then the dac will sound its best. I could have jumped directly to that point, or close to it, but there are still skeptics who think benefits of setup/support-systems are all imagined because the FFTs don't look all that different. What do you do about those people to get them to understand its not just imaginary?
Last edited:
If it's bad enough to cause bit errors, I'm sure you can measure it with your oscilloscope. I was thinking about crosstalk causing clock jitter rather than bit errors.
We are probably going to get some jitter simply because of Amanero clocks and because I2S bus is buffered in a CPLD. Later on I will try a different approach to board interfacing and jitter reduction. At that point I often don't use any kind of ribbon cable. That said, I don't always like u.fl cables in parallel (although there are ways to fix that too). We'll see how things turn out as we go, is all I can say for now.
BTW, what's your opinion on SE output mode listening? Is that what you use?
BTW, what's your opinion on SE output mode listening? Is that what you use?
In my testing I used a galvanically isolated 5 Volt.
However when measuring I used the balanced outputs all the time because noise and distortion were so much better than with a SE output and that doesn’t just mean cancelation of rather harmless even harmonics but also CM disruptions.
That’s why I checked the gain of both positive and negative path to be exactly the same by adding both outputs and tuning the sum to zero volt.
Hans
However when measuring I used the balanced outputs all the time because noise and distortion were so much better than with a SE output and that doesn’t just mean cancelation of rather harmless even harmonics but also CM disruptions.
That’s why I checked the gain of both positive and negative path to be exactly the same by adding both outputs and tuning the sum to zero volt.
Hans
I haven't listened to it much. It was certainly meant to be usable in single-ended mode, although the performance is better in differential mode (no extra noise from the common-mode loop, better cancellation of common-mode artefacts of the core of the DAC, cancellation of even-order distortion of the filter).
Thing is the ear/brain system is not exactly an FFT analyzer. And music is not exactly PSS. If there is disagreement, and if the end use is for music enjoyment by humans, then why carry on the likely-fiction that reproduction systems are sufficiently LTI for phase-free PSS FFTs to define what is closer to the end goal?
IMHO its a general problem in the technical world of audio. It wasn't specifically a reply to you.
However, if SE sounds more real to humans, and if balanced measures better, then maybe we aren't measuring/interpreting something that matters more/most to humans in the time domain view?
However, if SE sounds more real to humans, and if balanced measures better, then maybe we aren't measuring/interpreting something that matters more/most to humans in the time domain view?
Last edited:
Because of the lack of practical alternatives, I guess. Sighted listening tests are unreliable and blind ones are difficult to organize properly, and neither of them can be easily compared to simulation results, so judging performance by measurements is often the most practical option, even if they are a bit too simplistic.