Digital designers usually regard configuring an FPGA as a kind of hardware design, but it is in the grey area.
Makes sense.
I'm not sure if Mark's post was about DSD.
If I understand it correctly, they convolved and truncated the impulse responses of three hypothetical 128-tap FIR filters to derive the coefficients of the 322-tap FIR filter that they actually designed. It's meant to be fully integrated, a way to make a decent reconstruction filter without needing capacitors that have too large values to put on a chip.
Well, I am interested in dacs that reproduce digital music at a world-class level. So far, DSD dacs seem to come closer than PCM to the ideal of "perfect sound forever." A thin/slightly-recessed midrange sound seems to be the main complaint I am aware of for DSD as compared to PCM.
Last edited:
The modulator of post #2310, https://www.diyaudio.com/community/threads/return-to-zero-shift-register-firdac.379406/post-7623245 , appears to be meant for PCM input. You can come up with ways to get DSD through a similar modulator, but I can only think of inelegant methods - for example, by decimating the DSD or by running the quantizer at the DSD bit rate and the pulse width modulator at a much higher rate. Hence, I wondered whether post #2310 was meant to be about PCM only.
The dacs I have been using recently are DSD only. Either the FPGA-based "Simple DSD Converter" project here in the forum and or HQ Player can do a pretty good job of converting PCM to DSD256.
My thought with respect to Sigma-Delta PWM would be that a dac should be able play in that mode or in RTZ mode, as needed.
My thought with respect to Sigma-Delta PWM would be that a dac should be able play in that mode or in RTZ mode, as needed.
No. I did not have a problem and you made claims about Marcel's dac that proved to be false. Also you accused Marcel for being emotional about this. You seem overemotional to say the least.
Marcel,
Right now, Andrea's DSD dacs can operate in RTZ mode up to DSD256, or in NRZ mode up to DSD512. Its configurable in FPGA. Its not convenient to do on an ad hoc basis, however. Also Andrea has some DSD can boards that I believe could possibly be operated as two separate DSD dacs with a time delay between them to reproduce one channel. IOW, it could also potentially be configured to work as Thor seems to be suggesting with just a change in the FPGA, and maybe with relay switched routing of outputs to the same output stage channel. At least it might provide a convenient way to prototype/proof-of-concept such a scheme. The main issue with Andrea's designs IMHO is that they are not low-cost to manufacture, in part due to the low quantities.
All the above having been said, my friend still says there is something about the sound of your RTZ dac that he likes more. He says it has a more relaxed sound and feels more spacious behind the speakers. Of course, he also said its a little raw sounding. What I am wondering is whether or not all the things my friend is hearing are due to the same underlying causes. Which is to say, if the raw sound is polished up a little, will the relaxed feel go away? It may fall upon me to investigate to see if we can find an answer.
Right now, Andrea's DSD dacs can operate in RTZ mode up to DSD256, or in NRZ mode up to DSD512. Its configurable in FPGA. Its not convenient to do on an ad hoc basis, however. Also Andrea has some DSD can boards that I believe could possibly be operated as two separate DSD dacs with a time delay between them to reproduce one channel. IOW, it could also potentially be configured to work as Thor seems to be suggesting with just a change in the FPGA, and maybe with relay switched routing of outputs to the same output stage channel. At least it might provide a convenient way to prototype/proof-of-concept such a scheme. The main issue with Andrea's designs IMHO is that they are not low-cost to manufacture, in part due to the low quantities.
All the above having been said, my friend still says there is something about the sound of your RTZ dac that he likes more. He says it has a more relaxed sound and feels more spacious behind the speakers. Of course, he also said its a little raw sounding. What I am wondering is whether or not all the things my friend is hearing are due to the same underlying causes. Which is to say, if the raw sound is polished up a little, will the relaxed feel go away? It may fall upon me to investigate to see if we can find an answer.
Last edited:
I agree. Could you now open your own thread about your Topolino dac so we can have some progress in this thread.
I would suggest that opinions are never true in an absolute sense. Perceptions of truth are useful to form some predictive model to act upon, for better or worse. Opinions are just presentations that become weighted of lesser value than evidence for variant factors, as rarely attributable as having no value if the presentation by someone is presumed honest.
It seems that all your submissions are intended to diminish the weighting of all material presented by Markw4 to zero for what seems some prejudicial reason. My question is why zero?
Regarding the 10 kHz distortion, a post of bohrok2610 on another thread
https://www.diyaudio.com/community/...plification-does-it-exist.410047/post-7622819
reminded me of a subtle difference compared to the original filter: doubled gain, which also means reduced loop gain in the stage that has the higher gain.
https://www.diyaudio.com/community/...plification-does-it-exist.410047/post-7622819
reminded me of a subtle difference compared to the original filter: doubled gain, which also means reduced loop gain in the stage that has the higher gain.
My method involves a uniformly-distributed random rotation of the PWM pattern. Theoretically, that has some linearity advantages as explained in the valve DAC article, but it also increases the transition density and makes the bit pattern less suitable for non-return-to-zero DACs. I wanted to make my valve DAC return-to-zero anyway because I also wanted to try a chaotic single-bit modulator.
So, is my method smarter or stupider? I don't know, but it definitely is different.
If DSD comes closer than PCM and DSD has a "thin/slightly-recessed midrange sound" then what attributes would you assign to PCM as being lesser to the "perfect sound forever"?
To everyone, this comes back to an earlier point about RTZ or NTZ dac's. If we assert that signals prior to the resistors feeding the analog converter are zero impedance and as theoretically perfect digital representations, it seems that the nature of the distortion mechanism comes down to variances in the RMS values. How are RMS errors manifest in these two formats? This is to suggest that even though RTZ might contain spikes, how do these spikes manifest in distortion if they exist identically in the repetition (as notwithstanding creating difficulties later on in the actual I/V conversion)?.
Powerful well defined bass, more real-life HF (such as cymbals), wide and deep soundstage, very detailed and musical presentation with very low distortion, incredible percussive instrument sounds with detailed, say, for example, hand drum head decay modes, etc. -- its pretty much everything, except the warm, pronounced midrange common to R2R dacs, which is say the big chest voice type of sound, warm and full. Of course, when a trumpet with a mute sounds warm and full, unless it was mic'ed with a ribbon or something, then I start to get suspicious about whether the warm and full is entirely real.
Yes, in effect we see here in operation something roughly similar to the classic FIR "oversampling" filter with 100's of tap's shifted into the well, let's call it analogue domain.
We are at this point kind of near where ESS ended up, with 128 unitary weighted Bit switches per channel, in their 8-Channel Chip, so configured as 2-Channel DAC, 512 Bit (Thermometer) per channel and being clocked at 100MHz.
With a suitable logic core, it should be possible to implement almost any possible analogue FIR response and error scrambling by selecting random "bit's".
Somehow we seem to have come almost full circle (kind of) from the TDA1541 (which I am still listening to).
In principle It should be quite feasible to make a DAC IC with a very large number (say 2^16) unitary weight bit switches and current sources, capable at operating at 10's of MHz, for a "multibit" audio DAC with genuine 16 Bit hardware level resolution and beyond 16 Bit with noise shaping and oversampling (which is added by SAA7220 to the TDA1541).
While it takes a bit a little silicone area, 90nm is now over 20 Years old, there should be capacity and prices should be very low. Add in a fairly basic mask ARM Cortex CPU core to do all the processing. We probably have plenty left to implement a studio grade remastering Suite of FX too, at that. Not just EQ, but the whole board.
Really. Call me a luddite, I'll take a TDA1541.
Ok, so there was no rise in HD at higher frequencies, related it seems to something common mode. Ok, as said, you win.
I made statements about the principle, not a specific implementation,
I saw no proof so far.
Pull out ye olde 300MHz multichannel 'scope with a 1:10 probe (spring pin for ground ) and poke it around your PCB. I know, I'm a luddite and sim's are all we need and all that, no need to check if the real device matches the (furthermore partial) simulation, as we all know, in theory there is no difference between theory and practice.
Not Marcel, far from it. He is totally on the level.
Coming to think of it, I had more in mind those who seem to feel a need to defend Marcel against imagined sleights and going to the point of ad hominem attacks (not me, unless I count the above I hasten to add) when arguments are wanting.
Thor
Andrea makes multibit dac too that people say outdoes a 1541, for those who like that sound. He even has a DSD dac that is somewhere in-between. DSD with that multibit sound.
No. There is tendency of the most favored dac in each class, PCM or DSD, where they have certain perceptual attributes. TDA1541 has that warmth more than a DSD dac, but doesn't have the smooth, detailed, low distortion resolution that extends down below 16 or so bits. Probably one thing you might need to know is that the system here is unusual. There are big Sound Lab ESL panels, DS Audio optical phono, etc., that were put together to more or less match the system of one the better high end designers in the business, one who IIRC NP once described as "still having one of the best pairs of ears in the business." This system is said to outperform systems costing in the six figures (not that I paid close to that). The system is designed this way primarily for one purpose, which to be like a test instrument for audio electronics. Now you can believe that or not; either way you would be welcome to visit if you want to see for yourself.
One forum member who visited twice posted about some of what he saw and heard at: https://www.diyaudio.com/community/threads/dac-recommendation.376015/post-7560022
The first thing I noticed was that the electrostats could play loud with little to no distortion. Much more so than my horn speakers, which can go plenty loud. I had heard electrostats before and was always disappointed that I couldn't turn it up without hitting the limit. On these, the limit is high enough that you can enjoy realistic sound levels and stunningly low distortion.
This low distortion from the speakers resulted in a lot of transparency and ability to hear down into the preceding components. And it is those components whose sound we are interested in.
...
And of course all the other stuff, such as natural timbres and textures of instruments and voices was there.
Last edited:
Thanks to Cestrian for the closely matched Vishay/Beyschlag Melf , I've removed the 3k01 Susumo RR and given the Melf a try . I can only go by listening tests to well known tracks but the highs seem to be a little more clear ,not huge but noticing more with cymbals so to me was worth the time and effort
I have a pair of OPA1612's (thanks Cestrian) would it be worth changing the first pair of OPA2210 or second pair of OPA1678 in the active stage to the OPA1612's ?
I have a pair of OPA1612's (thanks Cestrian) would it be worth changing the first pair of OPA2210 or second pair of OPA1678 in the active stage to the OPA1612's ?
Attachments
BTW, I wrote about testing different resistors in Andrea's and Marcel's dacs. I warned about Susumu RG, and just got more flack from usual noise source. I don't think Cestrian believed me because he went ahead and tried it himself. But somehow nobody attacked his credibility for saying the resistors had a sound.