You can upsample as I said from the streamer... the question is if one likes or not a low pass filter at the output of the dac analog chain...I do not like it....my ears did not like it. I am not against oversampling...
There are no op-amps in my valve DAC, maybe you confused it with one of my other designs. There are ferrite core inductors, using large potcores with an air gap and a magnetically soft ferrite to ensure their distortion is negligible.
The original valve DAC requires a 27 MHz crystal rather than a clock oscillator, as the oscillator is part of the design. The DSD-only variant does require an external clock oscillator, but that has to run at the DSD bit rate or a multiple of it, that is, at the usual audio clock frequencies.
Last edited:
By the way, there is no law of physics forbidding you to use air core inductors instead of potcores. It just makes the placement of the inductors with respect to mains transformers and each other more critical, because of their open magnetic circuit.
Regarding Andrea's clock oscillators: some of his and his unknown companion's designs use a crystal at a subharmonic of the intended frequency and a clock multiplier. If you are going to make a DAC based on sigma-delta modulation, you have to be a bit careful with those, see posts 2830 to 2850 of the well tempered master clock thread, https://www.diyaudio.com/community/...jitter-crystal-oscillator.261651/post-6372426
I must be the only one who prefers a hardware oversampling digital filter to software upsampling. No low pass filter for me either.
If your best DAC is the Denafrips that would indicate you appreciate the R2R sound. That suggests for a DIY DAC you are better going down a R2R route. The best available chip and design is for AD1862. If you are fanatical you could make your own discrete ladder but that has high risks.
Miro's AD1862 allows many DIY options. Through reiterations of topologies. with comparisons to your Denafrips to beat, you have a very good chance of getting exactly what you are wanting.
Thank you for the reply, Marcel.
Agreed on the opamps. My bad. Regarding the other dac designs, for me personally I usually seem to like the sound better if I can keep the number of IC opamps to a bare minimum.
Understand reasons for using soft, gapped ferrites. Also understand there are limited options for producing steep analog output filters for audio dacs. Isn't that one reason Chord runs their high end dacs at 170MHz, and with a 17th order modulator? To keep analog output filter requirements simple? Of course there are always engineering tradeoffs, no perfect solutions.
Regarding use of a 27MHz crystal, just wondering what the oscillator close-in phase noise looks like, and to what degree it may affect dac sound it a way some people observe with other dacs. Don't know how sensitive valve dac is to that region of phase noise?
It would be so much easier to feel confident about it if I could hear the dac myself. Next best might be if someone could directly compare valve dac sound with a high end Chord and or a Holo May (with and without HQ Player conversion of 16/44 to DSD1024 for the Holo).
Last edited:
I haven't a clue how good or how bad the close-in phase noise of the crystal oscillator of my valve DAC is. I haven't got the equipment to measure it either.
In principle, the effect of close-in phase noise is the same for all audio DACs: you get a random phase modulation of the audio signal causing skirts around the spectral peaks of the signal. Those are usually small compared to the skirts you get from undesired amplitude modulation due to voltage reference noise.
In principle, the effect of close-in phase noise is the same for all audio DACs: you get a random phase modulation of the audio signal causing skirts around the spectral peaks of the signal. Those are usually small compared to the skirts you get from undesired amplitude modulation due to voltage reference noise.
Okay. Agree voltage regulation can matter a lot too, especially for AVCC type loads. Also, but to a lesser extent it matters for all dac circuitry. Seems like its a matter of, say, fix one thing (e.g. improve AVCC voltage regulation) then the next lower level thing (e.g. close-in phase noise) becomes more audible.
The AKM factory should be producing again soon (second half of this year), according to them. Hopefully everything goes well.
My recommendation for a DAC project would be Marcel's Valve DAC, it is very special. I have the DSD-only version.
Yes, a thread and an article.
Article:
https://linearaudio.net/sites/linearaudio.net/files/03 Didden LA V13 mvdg.pdf
The article describes only the original valve DAC, not the raw DSD version. The raw DSD version is a simplified version without the digital signal processing and the crystal oscillator.
Thread: https://www.diyaudio.com/community/threads/valve-dac-from-linear-audio-volume-13.308860/post-5104207
I've tried to put the most relevant information and links to the most important posts in the first post of the thread, but I made the links in an incorrect way, so most of them don't work. The post numbers are correct as far as I know. I can't update the first post anymore because the new forum software doesn't allow such long posts.
Article:
https://linearaudio.net/sites/linearaudio.net/files/03 Didden LA V13 mvdg.pdf
The article describes only the original valve DAC, not the raw DSD version. The raw DSD version is a simplified version without the digital signal processing and the crystal oscillator.
Thread: https://www.diyaudio.com/community/threads/valve-dac-from-linear-audio-volume-13.308860/post-5104207
I've tried to put the most relevant information and links to the most important posts in the first post of the thread, but I made the links in an incorrect way, so most of them don't work. The post numbers are correct as far as I know. I can't update the first post anymore because the new forum software doesn't allow such long posts.
Last edited:
@MarcelvdG - Very impressive project and great write-up. I think this is a bit more complicated than I want to tackle at this point, but it's definitely interesting.
Let's assume for a moment that everything in the digital playback chain is not a factor in your DIY (or other) DAC implementation. How significant - in terms of real world listening - would the Andrea's clocks be in improving the sound over the Crystek clocks? Recommended at the time. Would there be an audible and noticeable difference? Or, would the difference be at such a nuanced level, as to be imperceptible to all but the very highest of high-end systems.
A few years back, you helped school me in understanding where improvements could/should be made to get the best possible sound. Primarily, the improvements in PSUs. To that end, I adopted Ian's DAC modules (I know you have some issues with them) for the first time, and included better PSUs and other implementations improvements. At that time, I did upgrade the standard clocks to the NDK SDA and replaced the standard ones that came with the FiFoPi and DAC modules.
When complete, the results were more impressive in terms of audio quality then I would have ever thought possible. It sounded better than any of my other RPi based DAC HATs I had been playing with. So, it earned a permanent spot on a shelf of my primary system. For the most part, the fatique I came to expect from other DACs was simply gone. I could listen for hours without every sensing the harsh characteristics of digital playback over I2S. Except for the one-off track that I believe was simply badly produced. I was very happy with the results of the quality of the music. I have since upgraded to some commercial products (a good pre, and active speakers), and now I'm embarking on a new version of that first DAC, broadening the implementation to include other of Ian's modules to complete a full-featured streamer.
For this next version I'm looking to upgrade the clocks again. In particular, I been reading a lot about the Accusilicon AS318 clock. It's challenging because there are a ton of thoughts and comments regarding the Accusilicon - both negative and positive. So, what's the final take-away about the Accusilicon clocks? Would there be any other clocks that would provide an 'audible' improvement without breaking the bank? 🙂
A lot depends on how clocks are powered, bypassed, buffered, laid out on a PCB, and how clock signals are distributed. Crystek 957 can perform anywhere from very well to rather bad, depending.
IME for many or most diy dacs, clocks are not usually the biggest problem that could use some improving. Again IME, if everything else is done well then then clock improvements can easily be audible.
All that having been said, if using FIFO_Pi with Crystek clocks I would suggest to start with diyiggy's recommendation for bypass caps: Rubycon .22uf, 805 size SMD film caps. Powering the clock modules from an external low noise power supply is something that was also reported to help (I would probably prefer TPS7A470x over LT304x, along with the use of a dedicated power transformer winding). Also if using FIFO_Pi version 3 its possible to leave both clocks enabled at all times for best stability.
Regarding Accusilicon AS318B, most people seem to prefer them over other clocks for use with FIFO_Pi. However, they may cost more too.
IME for many or most diy dacs, clocks are not usually the biggest problem that could use some improving. Again IME, if everything else is done well then then clock improvements can easily be audible.
All that having been said, if using FIFO_Pi with Crystek clocks I would suggest to start with diyiggy's recommendation for bypass caps: Rubycon .22uf, 805 size SMD film caps. Powering the clock modules from an external low noise power supply is something that was also reported to help (I would probably prefer TPS7A470x over LT304x, along with the use of a dedicated power transformer winding). Also if using FIFO_Pi version 3 its possible to leave both clocks enabled at all times for best stability.
Regarding Accusilicon AS318B, most people seem to prefer them over other clocks for use with FIFO_Pi. However, they may cost more too.
Last edited: