I remember some discussion a little back about possibly developing a DAC.

Any further news?  I am going to be buying a new 1 and figured that I would check here since I have been very happy with my Masterpiece.

Thanks -

Eric
Quote 0 0
Deleting Attachment...
Anonymous
It seemingly died on the vine.
Quote 0 0
RayP
A while back Bruce ran a thread about developing a tube DAC - that is a DAC that uses tubes as part of the conversion process rather than just a conventional DAC chip with a tube stage bolted onto the output, but it didn't progress as far as I can tell.

Anyway, in that context I thought I would share the project that I've just finished, "a DAC that has all critical analogue and mixed-signal functions realized in valve technology." to quote the designer. The DAC is based on an article originally features in 'Linear Audio' (https://linearaudio.net/article-detail/2284) and realised by its designer (https://linearaudio.net/authors/373). The design is fundamentally a Delta-Sigma convertor so requires an SDM (essentially DSD) input and the original featured an integrated stage that converted PCM to SDM.

I have had an ongoing interest in DSD playback for some time and have built several DSD decoder projects (as well as several conventional DACs, such as the Twisted pear Buffalo 3SE and Soekris DAM1121 R2R DAC) and I already use HQ Player software, which is the best solution for delivering DSD to a decoder. I contacted the designer and asked him about modifying the original to be just a DSD decoder, dispensing with the PCM aspects. He was extremely helpful (and has remained so through the project) and delivered a modified design, including revised PCB fabrication (gerber) files. A couple of us here in the UK shared the cost of getting a small batch of PCBs fabricated (not trivial as the PCB is quite large and relatively sophisticated - it is handling digital data and high voltage - and is four layers with a custom layer stack as specified by the designer).

So, a build project started and you can see the result below. All the components, apart from the PCB, were easy to get hold of, mostly from mainstream suppliers, with just a few specialist suppliers (such as for the reconstruction filter inductors and the output transformers). I've not used 'boutique' capacitors or resistors in the build, the nearest thing might be the LCR polystyrene caps in the filters and Arcotronics Polypropylene output caps. Processing of the datastream from HQ Player is performed via a beaglebone black equipped with an isolator/reclocker.

This past weekend the project came to fruition and I've been enjoying listening to it via my very  revealing headphone amp (DIY Audio Noir HPA and BeyerDynamic DT880 headphones).
This DAC is amazing and simply the best I've owned. It has a totally silent background (as far as my ears can tell). The music out of it has a fantastic richness but without any stodginess, every voice/instrument is clearly discernible and, to use an old cliche, I'm hearing things I hadn't picked up on before. There is much texture and detail in the music - I have never heard Eva Cassidy sound so good, for example. I still have the fun of exploring the HQPlayer settings further to see if I can optimise the playback to my tastes even more. Input to the DAC is DSD256; I tried DSD512 too and it works except that the computer I'm currently using isn't powerful enough to cope with the HQplayer processing demands so there are dropouts. I do have a more powerful, audio-dedicated, workstation available so I'll give DSD512 another try at some point. This project is a keeper so I'll design a chassis lid to top it off.

IMO this DAC deserves to be built by more people and it will be good, in due course, to expose it to some of my audio chums to see what they think as my opinion isn't necessarily objective of course! What might be considered to be downsides of the project are that it's not cheap, it needs to be fed a high quality SDM (DSD) datastream (no PCM) and it's ridiculously inefficient to power all those valves to just produce a line-level analogue signal but it is also, IMO, ridiculously good and so easy to listen to.

Quote 0 0
Achim
Great effort, Ray - congratulations!
Quote 0 0
tubehifi
Wow!  That is really cool!  I'm not a digital guy so I don't gave the theory to develop something like that.  Only 6 tubes?  The design I was going to pursue would have been based on a R2R switching network, but it needed a mountain of tubes, so I gave up on it.

I would be very interested in building one. This thing has commercial value.  Perhaps something could be worked out and a kit could be developed that is acceptable to all parties.  I'm pretty good at knocking costs down and can see from the picture that there is a lot of money tied up in power supplies and support circuits.         
Quote 0 0
RayP
Bruce, I have no commercial stake in this so anything in that respect would obviously have to be pursued with the designer. Suggest you check out the Linear Audio material I linked to.

Actually there are seven tubes, two trios form the delta-sigma cores to the decoder and the seventh is to provide a reference voltage. As I said above, Delta Sigma (which a lot of modern DAC chips do under the hood BTW) means you need to provide SDM/DSD so if you have typical PCM music files (like CDs) they would have to be upsampled to DSD. IMO the best way to do that is via HQPlayer but if you were prepared to compromise you might be OK with an upasmpling IC, like the AK4137, or use the ASRC module in the designer's original project - those approaches mean you wouldn't need any specialised support infrastructure such as my build has.

Regarding that, my build appears more complex than it actually is because there's quite a bit related to my use of HQPlayer and streaming the data over the network to a Beaglebone renderer (conceptually similar to the RPI-based music player) and that accounts for a lot of the complexity of  power supplies you perceive. I've actually substituted a HV regulated B+ supply for the CRCRC supply that is on the PCB too, and I use a regulated filament supply for the tubes but it'll work fine with AC filaments - to be honest, I don't thnk you could strip much, if anything, out of it if you built it to the designer's schematic - remember you need to provide HV as well as LV supplies for the digital elements as you can't get away from using silicon for things like the clock management (the silicon is smd, mostly on the bottom of the PCB). I can post some more pictures if it helps?
Quote 0 0
Wolfgang
Something like this could work as alternative to the HQP approach:

https://www.audiophonics.fr/en/diy-interfaces/digital-upsampling-interface-src-ak4137-i2s-32bit-384khz-dsd256-with-screen-p-13770.html
Quote 0 0
tubehifi
Ray,
I just ordered Volume 13 which has the article.  Amazon distributes it.  I assume they sell boards for it.  At the very least, this should make for a great project to share even if there is no commercial pathway.  We shall see where it all goes.
Quote 0 0
RayP
tubehifi wrote:
I assume they sell boards for it.


Nope. If you re-read my first post I made reference to having to order a small batch (5) of PCBs, the cost of which was split between myself and another UK-based DIYer. They were the most expensive part of the project as they're 4-layer with a custom layer stack as specified by the designer. I've obviously used one of my PCBs and the other is reserved but I can make enquiries about the other boards if you would like?

There's a build thread with plenty of my pictures on DIY Audio; https://www.diyaudio.com/forums/digital-line-level/308860-valve-dac-linear-audio-volume-13-a.html
Quote 0 0
RayP


Yes, you could but I don't think it'll be as good as using HQPlayer. One thing to be wary of is that Audiophonics resell a lot of cheap Chinese stuff, which is often not very good quality and sub-optimal implementations.

This might be a better bet; https://www.diyaudio.com/forums/digital-line-level/358147-simple-dsd-src-beaglebone.html
Quote 0 0
RayP
tubehifi wrote:
I just ordered Volume 13 which has the article.


Also bear in mind that my build isn't exactly the same as the design in the article because it has dispensed with the ASRC elements however, it'll give you a good handle on the design principles.
Quote 0 0
RayP
Apart from the numerous through hole parts visible in the picture above, here are a couple of pictures of sections of the bottom of the PCB, the smd parts are involved with managing the data and clock signals. You can see why a PCB is necessary for this project;



Quote 0 0
tubehifi
Does the author expect the hobbyists to develop their own multi layered boards for this?   That is too much to ask.
Quote 0 0
RayP
tubehifi wrote:
Does the author expect the hobbyists to develop their own multi layered boards for this?   That is too much to ask.


No Bruce he doesn't. He has already used his professional expertise (see the short auhor bio. on the Linear Audio site) to do the hard work on designing the PCBs, which are relatively complex given the nature of the project. To quote from my first post on this topic, the designer "...was extremely helpful (and has remained so through the project) and delivered a modified design, including revised PCB fabrication (gerber) files...". All the development work on the PCBs for either version is done (as, IIRC, is the code for the ASRC element of the original version). If you want to build this you just need to pay to get PCBs fabricated from the files he has developed and made available - I used PCBWay. Having the PCBs fabricated wasn't cheap (several hundred £ for a minimum order of five PCBs) and ideally a group buy is the way to proceed as the unit cost would fall quickly.

You need to remember that the designer's motivations are quite different from your own and he seemingly has no interest in any commercial aspects; to quote from the Linear Audio webpage, "...he considers it a nice intellectual challenge to see how far he could get with building a DAC that has all critical analogue and mixed-signal functions realized in valve technology" and, having succeeded, has been generous enough to share his design with the audio enthusiast community via Linear Audio and continues to be generous in his suppport on DIY Audio.
Quote 0 0
Achim
Ray, did you ever have a chance to listen to an implementation of Jussi Laako's DSC1 or DSC2? In comparison to the tube solution?
Quote 0 0
Deleting Attachment...