You mentioned a few posts back that you had started collecting the components. I surely hope that didn't include the FPGA module yet...
I think my ears are not golden enough for the kind of review you are asking for. I don't even hear the difference between the steep and the apodizing interpolation filters for 44.1 kHz sample rate material on my DAC, while others hear that quite clearly.
In any case, assuming that everything works out well, I expect that the biggest difference between the original and the DSD-only valve DAC will be in the algorithms, so it will depend on the software you use.
If you lived in the Netherlands, we could all go to a Zelfbouwaudio Luisterdag (day where people from a Dutch DIY forum demonstrate to each other what they have built) and ask the opinion of the other audio enthusiasts there.
Phew, that's a relief. Thank you. I can get overly enthused at times but I'm hoping that the current discussion might drive some renwed interest in your Valve DAC, especially if the results are in line with my hopes/expectations.
I've not had any joy finding the BCF manual - that's the trouble when I have a tiy up!
Ray
Last edited:
So it looks as though with Dan and I sharing the cost of a batch of PCBs and with Marcel's support, we will be able to have a go at building the DSD Valve DAC.
If there are any lurkers out there who might be tempted to build one, now would be a good time to show your hand to get the benefit of a cheaper PCB.
If there are any lurkers out there who might be tempted to build one, now would be a good time to show your hand to get the benefit of a cheaper PCB.
I think you will need a bit more filtering than the first-order filtering that the BCF2 provides. You could try something like this:
-Place the 10 uF capacitors on the valve DAC main board
-Change the 8.2 nF capacitors into 10 nF each.
-Connect 390 ohm or 392 ohm from each output to ground.
-Connect the BCF2.
This will give you a cascade of two first-order low-pass filters.
Alternatively, you could make the balanced LC filter I wrote about in an earlier post and connect its output to the BCF2. That will give you better out-of-band noise suppression, flatter magnitude response and more phase shift.
In either case, you will still need the BCF2's input DC blocker capacitors to get the upper valve to bias correctly, and for proper balance you then also need to place the ones for the bottom valve.
I did once, late November 2018. The people there generally seemed to like it; some told me so, and I also heard people behind me say to each other that they liked it, so apparently they were not just being polite.
Thanks Marcel. I was planning on using the filter you posted about previously as I know the first order included on the BCF board isn't really sufficient. There's lots to do before I get anywhere near putting a BCF together though, little things like getting the DAC board assembled and tested!
Last edited:
Great, another notch up on the confidence level.
For interest, I am active on a UK forum and we also have get togethers to share what we've been getting up to. you can see and read about my 300B SE-OTL amps about three quarters of the way down this blog about one of our mmetings.
Simon's New Adventures In Hi Fi: Audio Meet - Owston 10th June 2017
Eight 300Bs for one watt of power? Is it a very unusual design or is the reporter just kidding about the output power?
By the way, the big wooden box in this picture 24-11-2018 : Alphen -> AfterBabbel - Pagina 8 - forum.zelfbouwaudio.nl is my valve DAC. It was used with various combinations of amplifiers and loudspeakers made by other members, here you see it connected to a valve amplifier with 805 valves placed to the left of the DAC.
Know, he is mistaken, it actually gives about 1.5W into 8ohms - but the reporter is right about the sound quality and it was very well received at the event and ppeople still use it as a reference point. Incidentally, I use it with 15ohm loudspeakers so life is a little easier for it...
It is a little unusual; it's single-ended and has no output transformers. It's based on Bruce Rozenblit's design that was published in his book 'Tubes and Circuits' but has been significantly tweaked, primarily with specialised DC filament supplies and different, better sounding, current sink devices.
Thanks for the link to the Dutch forum thread that includes your DAC - it looks very anonymous in the simple wooden box.
Ray
I might be interested.
Price for one board?
As an aside, I've been thinking about trying hqplayer with their dsd conversion, but my current audio pc doesn't have the horsepower, and I've not wanted to spend the money to upgrade to an I7. Will obviously need to figure something out for this project, but I do have some native dsd files.
Randy
Randy
Hi Marcel,
I didn't order the FPGA module but did order the output transformers from Jensen and the Pot cores so I’m hoping that I can still reuse them in a reconstructive filter of some shape or form.
The rest can go in the “useful” draw until another project catches my eye.
There were quite a few new parts that I have identified for the DSD only board so they are now on order with Mouser. Let’s see how this pans out!
I know what you mean with the golden ears. Between myself and my brother who is three years younger, I’m sure half the things he says he hears are psychosomatic.
I travel extensively for work, mostly Asia however I was in Greece last week and the job was very intense so I didn’t manage to make it out and about, may be next time.
Perhaps next time it will be Haarlem...
Dan
I didn't order the FPGA module but did order the output transformers from Jensen and the Pot cores so I’m hoping that I can still reuse them in a reconstructive filter of some shape or form.
The rest can go in the “useful” draw until another project catches my eye.
There were quite a few new parts that I have identified for the DSD only board so they are now on order with Mouser. Let’s see how this pans out!
I know what you mean with the golden ears. Between myself and my brother who is three years younger, I’m sure half the things he says he hears are psychosomatic.
I travel extensively for work, mostly Asia however I was in Greece last week and the job was very intense so I didn’t manage to make it out and about, may be next time.
Perhaps next time it will be Haarlem...
Dan
I'm sure they can be used for the reconstruction filter. If you only use the higher DSD rates and don't mind getting a few tenths of a dB of treble roll-off at 20 kHz, you can use the reconstruction filter as published in the original article, or the alternative mentioned in the pdf file on the linear audio website (the Gaussian-to-x-dB filter). If you also want to use the lower DSD rates and/or want the filter to be as flat as possible up to 20 kHz, the capacitors and the number of turns of wire on the potcores may have to be changed, but you can still use the same potcores and transformers.
I also tend to be skeptical about what people claim to hear, but there certainly are people who can hear far better than I can. Mooly and PMA amazed me with what they could hear under double-blind conditions in this thread: High-order dither listening test Unfortunately I made a stupid methodological error, so the test was inconclusive after all.
Hi Randy, good to see your interest.
It would be great to have you come on board for the board purchase, it just makes the maths harder! I'll PM you (and include Dan) so we can discuss possible arrangements.
Just to be clear, we're looking to order the DSD-only version of Marcel's Valve DAC board. It is just the main decoder section with the DAC cores and V Ref. so you would need the upstream DSD source elements and also build your own reconstruction filter section. Please make sure you've seen the pdf document Marcel mae available a few posts back.
If you come in, would you want a solder paste stencil or are you happy to and solder without one?
For my HQPlayer server I purchased a powerful workstation, reconditioned ex-business, and made a big saving - worth looking into. Also, don't forget that suitable Nvidia based graphics card can be used to supplement the CPU processor - check out CUDA offloading on the Signalyst side and the links there from - I did that with a middling spec video card and it made a significant difference. With the BBB HQPlayer NAA the HQPlayer server is located at the other end of the house to the listening room. Happy to try and aswer any questions on it.
Ray
Last edited:
For reference, Marcel's suggested reconstruction filter for the DSD-only project is post #102
https://www.diyaudio.com/forums/dig...-linear-audio-volume-13-a-11.html#post5857250
Marcel, would anything need to change (apart from obviously have filter components on both pos and neg sides) for anyone wanting to buld a balance reconstruction filter. I assume that, should you choose to use the output transformers too they would go straight onto the back of the reconstruction filter as with your original Valve DAC?
As the reconstruction filter boards are very simple I may try to sketch out a schematic/PCB to continue with my KiCad learning. A batch of PCBs would be very cheap compared with the main board.
Ray
https://www.diyaudio.com/forums/dig...-linear-audio-volume-13-a-11.html#post5857250
Marcel, would anything need to change (apart from obviously have filter components on both pos and neg sides) for anyone wanting to buld a balance reconstruction filter. I assume that, should you choose to use the output transformers too they would go straight onto the back of the reconstruction filter as with your original Valve DAC?
As the reconstruction filter boards are very simple I may try to sketch out a schematic/PCB to continue with my KiCad learning. A batch of PCBs would be very cheap compared with the main board.
Ray
See post #105, https://www.diyaudio.com/forums/dig...-linear-audio-volume-13-a-11.html#post5857695 The required 1.8 mH (theoretically 1.77368 mH) inductors can be made with 84 turns of wire on a 250 nH/turn^2 potcore, like the ones Dan ordered - but of course there are only four of them in a stereo balanced filter, rather than twelve.
The problem I have with the reconstruction filters is that I don't know exactly what the noise shaping looks like, nor what rate of change the amplifier can handle without generating significant subslewing TIM, so I can't predict what filtering you really need. For the original valve DAC I took a conservative approach and used a filter with a slightly higher order than the noise shaper that cut off immediately after the noise shaper's band of interest, which led to a sixth-order filter. Most people on the noDAC thread seem to be using even less than third order and getting away with it, so that's why I guess third order Butterworth will suffice.
The transformer should be connected to the filter output and it requires an impedance correction network, like in the original design. I'll try to draw up some schematics this evening.
The problem I have with the reconstruction filters is that I don't know exactly what the noise shaping looks like, nor what rate of change the amplifier can handle without generating significant subslewing TIM, so I can't predict what filtering you really need. For the original valve DAC I took a conservative approach and used a filter with a slightly higher order than the noise shaper that cut off immediately after the noise shaper's band of interest, which led to a sixth-order filter. Most people on the noDAC thread seem to be using even less than third order and getting away with it, so that's why I guess third order Butterworth will suffice.
The transformer should be connected to the filter output and it requires an impedance correction network, like in the original design. I'll try to draw up some schematics this evening.
Last edited:
Thanks Marcel - I don't know how I missed that other than because it was late and I was very tired when I was posting last night!
One could use the output transformers for balanced to single-ended conversion?
Ray
Last edited: