Jesper,
My plan was to start with the onboard clock, but I would also like to try their external oscillator board as a next step to see whether this brings any measurable differences.
The advantage of this specific oscillator board is in being a standard product, whereas Andrea’s board has 22Mhz and needs to be upgraded to get 45Mhz.
But I’m sure Andrea’s board must be very good.
For testing I want to be ready for up to DSD512 without having to modify anything.
Hans
My plan was to start with the onboard clock, but I would also like to try their external oscillator board as a next step to see whether this brings any measurable differences.
The advantage of this specific oscillator board is in being a standard product, whereas Andrea’s board has 22Mhz and needs to be upgraded to get 45Mhz.
But I’m sure Andrea’s board must be very good.
For testing I want to be ready for up to DSD512 without having to modify anything.
Hans
Hooked up Marcel's dac to Andrea's FIFO board via u.fl cables. Spoofed DSD_ON and MUTE signals at dac to enable output relays. Roughly speaking, I would say maybe half the blurriness went away with the better clocks.
Still wasn't satisfied though, so the output filter board was removed and a quick, field-expedient passive filter was fashioned as follows: The dac resistor array outputs passed through the parallel combination of 100uf electrolytic, and a .01uf MKS film. After that was 270pf to ground, then attached an RCA SE cable about 4'-5' long from the caps at the dac to the line amp input. Cable measured 430pf. The input of the line amp is a transformer with roughly 150kHz bandwidth and an electrostatic shield between windings. No additional information on cable or transformer is available.
Also tried a shorter piece of cable, maybe 28" or so, but the sound was too bright so stuck with the longer cable. Again speaking only roughly, FR of audio sounds within reason. Close enough for a quick check anyway. Roughly, maybe 95% of the blurring is now gone. Also, removing the output stage kind of makes it more like a no-dac-dac. For anyone who has experimented with those things and who has one of Marcel's dacs available, it might be worth giving it a try to see what you think. Mostly I was playing DSD256 from Simple DSD Converter connected between the input the the FIFO board and the output of the I2SoverUSB board.
Anyway, its enough to reveal enough-for-now about the potential sound. That is to say, its possible to hear well enough what is left over from the remaining parts of the dac. IMHO the sound very good, subjectively speaking that is. Quite a bit closer to the potential. Soundstage is pretty good too. Some people might be surprised.
Still wasn't satisfied though, so the output filter board was removed and a quick, field-expedient passive filter was fashioned as follows: The dac resistor array outputs passed through the parallel combination of 100uf electrolytic, and a .01uf MKS film. After that was 270pf to ground, then attached an RCA SE cable about 4'-5' long from the caps at the dac to the line amp input. Cable measured 430pf. The input of the line amp is a transformer with roughly 150kHz bandwidth and an electrostatic shield between windings. No additional information on cable or transformer is available.
Also tried a shorter piece of cable, maybe 28" or so, but the sound was too bright so stuck with the longer cable. Again speaking only roughly, FR of audio sounds within reason. Close enough for a quick check anyway. Roughly, maybe 95% of the blurring is now gone. Also, removing the output stage kind of makes it more like a no-dac-dac. For anyone who has experimented with those things and who has one of Marcel's dacs available, it might be worth giving it a try to see what you think. Mostly I was playing DSD256 from Simple DSD Converter connected between the input the the FIFO board and the output of the I2SoverUSB board.
Anyway, its enough to reveal enough-for-now about the potential sound. That is to say, its possible to hear well enough what is left over from the remaining parts of the dac. IMHO the sound very good, subjectively speaking that is. Quite a bit closer to the potential. Soundstage is pretty good too. Some people might be surprised.
Last edited:
Interesting. This is something I could also experiment with since I have two virtually identical RTZ DACs with identical PSUs, PCM2DSD, line tx, etc.
Hi Hans ...
Entirely fine, Hans - thanks for the feedback. Just thought I'd offer you the option.
Cheers, Jesper
Entirely fine, Hans - thanks for the feedback. Just thought I'd offer you the option.
Cheers, Jesper
Hi Mark,
I suppose you connected your 100uF//10nF caps to both shiftregister outputs.
If so, you are using a balanced connection 🤗
And as far as I have read, you didn’t try giving the filter stage a separate supply and transferring the two signal from the balanced outputs over the same SE cable as you are using now, true ?
That would IMO have been a fair compare to your passive transformer solution.
I’m just mentioning this because with the Fifo and the passive reconstruction filter you are making very big steps now, where smaller steps in between could possibly have given a similar change in subjective perception.
Hans
P.s. connecting the transformer to the 2 active filter outputs could also be a step in between.
No, connection to dac looked like this:
Positive phase and ground only for right and left channels. Electrolytic and parallel .01uf film cap combination, were in series with the dac outputs to provide DC blocking.
Regarding a separate supply for the opamps, it would probably help some but IME it likely would have been sort of a half-way measure in this case. Wanted to cut to the chase, as it were.
EDIT: In case its not clear, this doesn't sound like a nodac-dac. There is no remaining noise problem after the transformers, at least no worse than vinyl. IOW, whatever noise remains is not a problem I can't live with, and the sound is really quite musical. Let's wait until some other people try it and see how well they can do with it.
Last edited:
The capacitor and cable are 700 pF together. They are effectively in parallel with an 8.2 nF NP0 capacitor on the DAC board, so the total capacitance is 8.9 nF plus the winding capacitance of the transformer. 3020 ohm/8 (assuming 10 ohm shift register output resistance) and 8.9 nF gives a corner frequency of 47.3711 kHz and a treble loss of 0.7124 dB at 20 kHz, excluding the effect of the transformer.
Yeah, I can hear its slightly dull. But its better than the little HF grunge that would be there otherwise. IOW its adjusted to sound good given the way the rest of the dac is at this point.
In the end, I would still like to see an all discrete active filter stage.
Mark, no personal experience but from a few assists in recording the local orchestra, as well as a good friend who plays and deals in the better guitars, I was surprised to hear that tuning an instrument by measurement is easy, but leaving it at that, usually is oke if you only want to play one key/string. Of course we don`t and when chords etc are played, the instrument sounds off. Kind of your story about using steady state signals when testing an audio device.
Why didn’t you try direct from HQPlayer instead? Thought you have the trial version that could go up to 30mins
I play, tune instruments, intonate them, etc. Its an involved subject that could get way too off topic here.
Don't know that its exactly analogous to using steady state test signals and discarding phase information as is typically done in PSS spectral measurements.
Maybe its more that all analogies are superficial and break down at some point if picked apart. Its that two different things are never exactly the same. The purpose of an analogy is not to say two different things are the same, or parallel, down to a deep level. Analogies are only for introducing someone to a new idea by sort of comparing it to something they are already familiar with.
I could have done that, but I like the sound of Simple DSD Converter version 3 firmware better than I like the HQ Player algorithms I can think of that can run on a less than stellar PC. Since I use an old laptop most of the time for playing test music, I can't run the most computationally expensive HQ Player algorithms.
Actually, I think Simple DSD Converter version 3 firmware is quite good. IMHO its good enough for commercial use (if you work out some agreement with them). Just to be clear, I am only involved at the hobby level, its entirely their IP. The only small flaw that bothers me a little can be ignored without problem. Its that the midrange isn't quite as open as it might be, but its only a slight effect. Remarkable the guys did such such a good job using a modest FPGA. BTW, they haven't given up yet. Have some more listening sessions I want to do for them, just have had to take a break from it for awhile with some of the other stuff going on.
Last edited:
Ok, thanks, very interesting!
What about the effects of just FIFO in the chain. Looks like Marcel’s DAC is still doing all the processing like before (no FIFO) and just getting input feed with greater accuracy?
What about the effects of just FIFO in the chain. Looks like Marcel’s DAC is still doing all the processing like before (no FIFO) and just getting input feed with greater accuracy?
Basically. The BCLK signal has lower phase noise, which may have some effect on dac sound.
It is a Marcel as has said, to the effect that a dac needs a time reference and a voltage reference. BLCK is the time reference.
understand better now, thank you.
I was more interested in finding out if time domain isolation is also beneficial even for a synchronous system. As we have seen, physical isolation of signals is like a must have now for both measurement and listening
I was more interested in finding out if time domain isolation is also beneficial even for a synchronous system. As we have seen, physical isolation of signals is like a must have now for both measurement and listening