Yeah, sorry, I've been held up a bit as I ordered a couple of cases from Aliexpress only to be messed around, so I've changed tack now and got a couple of nice Hifi2000 cases on the way to me now from Italy 🙂
Currently my dddac looks like this 🙄
But I'm back on the case now and will be trying out a couple of wiring options in the next week or so and hopefully seeing if things like snubbers are helpful in this application.
Yeah, sorry, I've been held up a bit as I ordered a couple of cases from Aliexpress only to be messed around, so I've changed tack now and got a couple of nice Hifi2000 cases on the way to me now from Italy 🙂
Currently my dddac looks like this 🙄
But I'm back on the case now and will be trying out a couple of wiring options in the next week or so and hopefully seeing if things like snubbers are helpful in this application.
No worries mate I knew you were probably busy. I have found Andae and Hifi2000 absolutely fantastic to deal with. I have bought 8 cases from them and found them exceptional.
I'll hold off on my PSU build until I know your findings.
Thanks for your help !!!
Could it be reflections on the I2S/BBCK line that are messing up the dac sync? That is what the 100R are for pre the input to the dac I assume, perhaps there are some changes afoot there in some way, as Dar as I understand digital traces (and that isn't much) they are terminated at the end of the run not at each dac. I may be well off base here just remembering something in the recesses from some thing Steve Nugent wrote.
Just a thought.
Laters
Drew.
Just a thought.
Laters
Drew.
OK, let me jump in here. Nice detective work from Hermann by the way, this makes our common hobby so interesting. Always some new perspectives 🙂
What we see here is that the dac chip is outputting its analog level (a current) within a few clock cycles. Indeed this has nothing to do with jitter, delay lines or what so ever. We should remind ourselves that the 1794 is a complex 1 bit converter with some 6 bit tricks in it. The datasheet is so vague (I guess on purpose) that we do not know what really goes on, except that it is a mash converter. What I expect is that what you see here is the effect from the dithering which happens inside the chip. We also should not forget that these clock cycles are far away from the 44.1 kHz sample frequency and that this experiment is done with square waves of 22.1 kHz. You will not find these in CD recordings of course.
Now it might very well be, that the resulting step case of 3 clock cycles helps a bit smoothing the digital output. In fact what you see is a kind of Low Pass Filter on a one bit signal.
If you would run a normal sine wave through this test it will be much harder to see the effect of course.
Nevertheless, I think Hermann contributed very nicely to understanding why more decks sound better than one single deck. The HF filtering and the averaging out out of the mash converter effect does the trick it seems 😎
On the point if controlling this would help. As said above, you will create a kind of FIR filter with no feedback, so a straight Low Pass Filter. And that is exactly what we saw in Hermann's scope pictures. Not sure if this is really needed. you can also put a capacitor on the Output as is already an option on the DDDAC Blue board...
What we see here is that the dac chip is outputting its analog level (a current) within a few clock cycles. Indeed this has nothing to do with jitter, delay lines or what so ever. We should remind ourselves that the 1794 is a complex 1 bit converter with some 6 bit tricks in it. The datasheet is so vague (I guess on purpose) that we do not know what really goes on, except that it is a mash converter. What I expect is that what you see here is the effect from the dithering which happens inside the chip. We also should not forget that these clock cycles are far away from the 44.1 kHz sample frequency and that this experiment is done with square waves of 22.1 kHz. You will not find these in CD recordings of course.
Now it might very well be, that the resulting step case of 3 clock cycles helps a bit smoothing the digital output. In fact what you see is a kind of Low Pass Filter on a one bit signal.
If you would run a normal sine wave through this test it will be much harder to see the effect of course.
Nevertheless, I think Hermann contributed very nicely to understanding why more decks sound better than one single deck. The HF filtering and the averaging out out of the mash converter effect does the trick it seems 😎
On the point if controlling this would help. As said above, you will create a kind of FIR filter with no feedback, so a straight Low Pass Filter. And that is exactly what we saw in Hermann's scope pictures. Not sure if this is really needed. you can also put a capacitor on the Output as is already an option on the DDDAC Blue board...
oh... this delaying of one bit signals and than adding together is exactly what you do with DSD signals to bring out the analog value. Only you use a much larger number (compare here nr of DAC decks). Depending if hardware or software from like 8 /32 to several hundreds
This got lost a bit between the other posts...
If that is the case, I would doubt if there is really I2S signal coming from the spdif board? (as you say, USB board works....)
So the only way to solve the riddle is to take a scope and check for signals...
You need to start somewhere and this is the first step...
I have long thought that I could sometimes hear a slight change in the character of the soundstage produced by my own single chip PCM1794A based experimental DAC whenever I even briefly pause and restart disk playback. It is a subtle effect, so I thought perhaps it was only my imagination, yet the impression persisted. Dusty128's findings now lend some measured proof that the behavior of the PCM1794A does indeed change should the input data stream be halted and restarted. Although I've no solid idea as to why it should do that, I suspect it has something to do with the operation of the internal sigma-delta modulators, of which there are seperate modulators for the left and right channels.
The stair-stepped output waveform of the multi-deck DAC is apparently unintentionally applying a little bit of linear interpolation, realized in the analog domain. Linear interpolation realized in the anlog domain is a technique that's currently utilized in at least one highly regarded commercial DAC - The Trinity DAC: Trinity DAC - Page 52
In short, linear interpolation does not produce the impulse response ringing of convential brickwall filters. That is because linear interpolation does not feture a brickwall frequency response, and so, is poor at suppressing the signal's undesired ultrasonic image products. The 4 deck unit is, in effect, applying oversampled linear interpolation to the otherwise NOS operation. However, this is only a quasi-linear interpolation, as the stepped levels are not always uniform or what's known as monotonic, and the stepped transition levels are not evenly spread across the full sample period.
Last edited:
I would suggest it a form of overdubbing or similar, if you are summing the multiple analogue outputs of the DAC's and there is a brief time delay between the samples...
You will know as soon as you start measuring the signals. It is the only way. You seems to have done everything right, so you have to start excluding possible errors now....
Indeed! At the moment I'm running a bog standard two deck unit, which now with matching 100r resistors on both decks sounds good. My single decker with super regs and separated analogue /digital psus is waiting for its next upgrades being the buffer circuit and Cinemag output txs. It already sounds absolutely stunning: silky smooth, detailed and very dynamic! I could happily live with the two decker but I'm damn glad I've developed the single decker super Dac!
Tentlabs Shunts
My 8 and 3.3v tentlabs shunts have arrived today. I am aware some components like caps etc should be removed on the DACS module. Can someone please share the list of component removals\changes with me? Thanks
My 8 and 3.3v tentlabs shunts have arrived today. I am aware some components like caps etc should be removed on the DACS module. Can someone please share the list of component removals\changes with me? Thanks
Take a look at Supersurfers post here:
http://www.diyaudio.com/forums/digi...pcm1794-waveio-usb-input-129.html#post4191839
You can pretty much tell from his photos what he has removed.
As far as I can see the delay varies according to the bit rate...
You are still summing analogue channels that are offset in time, so it is going to alter the sound similar to dubbing, due to the varying start up of the different DACs, at the moment this is random as there is no direct control.
Thanks Ian and I will but I also want to be 100% sure what people have done to avoid issues.
I do not see how it can be better on its own, but as the delays will probably smear the jitter out and there will be added i.e higher combined output current with multiDAC into a transformer primary, it may recover sufficient SQ to be more pleasing to the ears. The averaging process with higher numbers of DACs in stacks may also add to the improvements still claimed.
my theory is that it's not smearing the jitter, but filtering the aliasing, which is supposedly 1 of the only real downsides to a NOS dac.
John Swenson has done a fair bit of work on this in the past, there's a fairly concise explanation of his thoughts and findings here: Computer Audio Asylum: RE: Question about DSP and Filtering by John Swenson
Hi James,
There are more factors in the audio chain that contribute to dirty high frequencies, like; dome tweeters, resonating filter capacitors in your loudspeaker filters and amplifiers, badly designed amplifiers, wire cry (yes it exists), switching power supplies, and of course jitter.
It is a priviledge to own good tube amps, these (the audio transformers in the chain) are able to do some analog filtering 🙄
So for me NOS has no down side 😀
Yes indeed!
No technical specifics from me this evening, but I've re-worked a few things prior to fitting into my new cases this evening and from a couple of hours test listen to make sure these bits are working OK I'm in love with it all over again 🙂
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
I'll do a proper explanation of a few bits very soon, but for now just wanted to say how thankful I am to experience such delights 🙂