I definitely was not trying to insult Marcel. On the contrary, IMO claiming that his design loses detail when operated as intended is more of an insult especially since no evidence was presented.
Marcel, bohrok2610 has certainly dissed your dac by saying it costs more and sounds worse than dacs apparently designed by better engineers good enough to produce superior and flawless products at lower cost. In fact, he hasn't even heard your actual dac, only the similar one he made which he is claims is superior to your design. IOW, bohrok2610 plainly implies FFTs prove the actual fact that he is a better dac designer than you are, because his FFTs look prettier. I call BS on that. I know his dac can measure better without sounding better. That's because he optimizes for spectral analysis FFT looks, not for SQ. If we provide good support circuitry for your design, I am quite sure it will sound far better than his, pretty looking FFTs or not.
Also, I don't mean to give short shrift to FFTs themselves and the various ways their data can be analyzed, including the use of phase information. The common problem is with the over-reliance on spectral analysis looks by subjective human engineers.
Also, I don't mean to give short shrift to FFTs themselves and the various ways their data can be analyzed, including the use of phase information. The common problem is with the over-reliance on spectral analysis looks by subjective human engineers.
Last edited:
How do you know what I do? Especially since I do not optimize for spectral analysis but for SQ.
If you want to listen to my designs you can try this:
https://www.diyaudio.com/community/threads/fixed-gain-field-recorder.373352/post-7862343
So I'm not allowed to express my opinion? Prefering something else is not dissing.
I don't believe you. You would have stopped using ferrites and nonlinear caps in your clock circuits if you bothered to listen. What is true is that you can't see what's wrong with ferrites and nonlinear caps in a spectral view. Yet you claim not to be optimizing for FFT looks?
Out of interest Bohrok whats different with your design that could be making it measure better ? Sorry if it's already mentioned but this is a big thread . You used a different output stage opa1632 ? Do you use a different regulation to Marcels for the shift registers ? Maybe it could it be the layout ? I think you mentioned you added re clocking ?
I definitely noticed this. I just I guess reacted poorly to what I saw as another episode of attacking Mark. Spectrum does different things to different people 😉
I haven't seen you being attacked one bit. This is a good, informative thread and I think Mark's contributions with listening testing are really valuable - so I encourage them and discourage what I sometimes see as non-helpful comments, attitude et al.
Apologies all, having a rough time and reading into some stuff a little more negatively as a result.
Edit: argh! I can't edit out my previous post (again - though I think the window has been lengthened) so just to say head's back on shoulders and sorry @bohrok2610 for the harsh attitude of my own. I'll step aside unless I see where I can contribute or ask pertinent questions.
Thanks again everyone.
It is probably a combination of things: reclocking of DCLK, power supply, USB-I2S, layout, component choices, not to mention the ADC used in measurements. But the basic design is Marcel's. I haven't made any major circuit changes other than to the output stage. I have several output stages as "hats" but OPA1632 version is the same you are using.
@bohrok2610 and @Markw4 , you have both expressed your opinions on measurements, informal listening tests, the DAC of this thread and related issues quite clearly, can we leave it at that for now?
For the record, the DAC of this thread was never meant to be the world's greatest audio DAC. It was merely an experiment to test a hypothesis of JohnW and produce a useful DAC in the process, with everything open source so people can adapt it whatever way they like. There is no reason to assume that its design or layout is optimal in any sense, and certainly no reason to assume that it cannot be outperformed by far more complicated circuits as used on DAC chips.
For the record, the DAC of this thread was never meant to be the world's greatest audio DAC. It was merely an experiment to test a hypothesis of JohnW and produce a useful DAC in the process, with everything open source so people can adapt it whatever way they like. There is no reason to assume that its design or layout is optimal in any sense, and certainly no reason to assume that it cannot be outperformed by far more complicated circuits as used on DAC chips.
Got to admit I was first sceptical regards re-clocking , I didn't think it would make much difference but gave I it a go. The good thing is I'd be able to disconnect it and revert back if not convinced . I can say now the reclocker won't be going anywhere , the parts etc used may change over time but for me I like it better .
The same with the output stage , the tweaks to try are easily reversible. Things either sound the same , better or worse I've never been too proud to admit if I can't hear something . What I can say is that i'm up for trying things that I think may find to be an improvement.
The same with the output stage , the tweaks to try are easily reversible. Things either sound the same , better or worse I've never been too proud to admit if I can't hear something . What I can say is that i'm up for trying things that I think may find to be an improvement.
Sure. And maybe the same about S-cut crystal clocks. You just wait, its gonna blow you mind how much better its gets. But first you need have a good clock board. And maybe you need to put some physical distance between that and the USB/PCM2DSD boards.
The transformer output stage is also way better than Marcel's output stage even with slowed differential summing.
We have done the experiments and listening tests, and I have people with a very good track record at that. They think we have an opportunity to design a justifiably very expensive world class dac based around the designs of Marcel and PJotr25 (and requiring Acko level clocks). I would have to do most of the actual work to commercialize it though, which I would rather not do. The benefit would be that Marcel, PJotr25 and I would receive royalties, and Acko would make some money on clocks; and I know people would be eager to try them out worldwide, because there is already a worldwide distribution channel. Again, I don't want to do it. I would rather give the IP away so that the commercial window would pass by. Problem now with what I would rather do is crazy hard, its trying to get people to just try it so that there are no longer trade secrets.
In other news, I have some Chinese transformers coming which I will soon be testing. I'm not expecting too much, but maybe I will be surprised. I'll let you know.
Last edited: