[regal]

Many of us like tubes and want a true tube DAC, so I have been looking at this for a while. For reasons I won’t go into here, V-out DAC’s are best for a tube DAC


I came across the Lampizator site a while back. This guy seems like a straight shooter. His claim is that a CS4397 (basically a Zhalou) and a cathode follower will beat anything on the market (tube or SS.) Well I believe him but there is a lot of room to improve his design. Also it really shows that the old myth that a V-out DAC is just a i-out DAC with an internal opamp is not true. A well designed timeslice D-S DAC like the Wolfson 8471 or the CS4397 does not have an internal opamp performing I to V.


What is wrong with the LAMPUCERA ? A lot really. First off all he is throwing away the differential output of the CS4397. This is a bad idea because it can be used to increase CMRR and increase gain. His power supply design is also very rudimentary.


I found this blog TubeFest where a Zhalou (CS4397) mated with a modified balanced Aikido won the European Triode Festival.

Broskie proposed a much simplier and cost effective design if we don’t need the balanced outputs. It has the CMRR plus all the benefits of an Aikido linestage.



Tests here DAC Final have shown that the Wolfson 8740 sounds a little better than the CS4397, and the WM8741 is even better because there is less post DAC filtering required.



We are at a time where there are so many existing DIY “modules” that it makes little sense to start from scratch creating a new PCB, changing a few components, and calling it a new design. Why not use existing designs and build something that transcends the sum of its parts? Here we potentially have a Triode Festival winner for very little investment:

If there is interest I can put together all the component values for the filter, tube resistors, etc. I definitely plan on building this one.

 

[tomb]

Sounds OK in principle. Like you, I wish we could get a tube DAC and wouldn't even care if the tubes were just a linestage, either. Dsavitsk has already done something similar here, I think:
Modular, Upsampling, Oversampling, Differential DAC

Add a USB input, too, and I'm interested.

 

[regal]

USB is easy to add to this just replace the SPDIF module with TP's USB module or Deode's (USB-I2S

Dsavitsk's concept looks similiar, he is using 2 triodes per channels but looks to be outputing balanced. I prefer to use the aikido to do a balanced to unbalanced transfer because it almost nulls the PSU noise. It is possible however to do this with balanced outputs, cost goes up about $200.

I've yet to hear a PCM1794 DAC I liked, most others prefer the WM and CS too, but I am sure Dsavitsk has cracked the code. Its great to see more tube OS designs showing up.

The awesome thing about the Aikido output stgae are the endless tube rolling possibilities, you just need to be flexible with gain.

 

[dBel84]

This looks like a great concept Regal, I will keep an eye on your progress..dB

 

[TimJo]

I have been thinking about this for a while as well. I had this bookmarked and thought about using the Twisted Pair modules to simplify the build as well. I think your on to something. Hadn't seen the TubeFest article before. I'll take a look.

 

[pabbi1]

Quote:

Originally Posted by tomb /img/forum/go_quote.gif Sounds OK in principle. Like you, I wish we could get a tube DAC and wouldn't even care if the tubes were just a linestage, either. Dsavitsk has already done something similar here, I think: Modular, Upsampling, Oversampling, Differential DAC Add a USB input, too, and I'm interested.

Whatever became of Colin's tube DAC?

 

[tomb]

Quote:

Originally Posted by pabbi1 /img/forum/go_quote.gif Whatever became of Colin's tube DAC?

I think he decided that it's impossible to use tubes for I/V. That's kind of why regal is talking about a tube line stage here. It's sort of a slight-of-hand way to involve tubes, I think - but still desired, IMHO.

Colin is still working on his solid-state DAC, though. It sounds like he has a killer discrete I/V board that he's prototyping.

 

[Uncle Erik]

Very interesting idea, regal. Please keep us updated on your work.

I think I have an Aikido PCB floating around somewhere.

 

[regal]

The tubes here are more than just a line stage. With new DAC's like the Wolfson or CS it is important that the differential signal be used so a Common Mode Ripple Rejection can be employed. The Aikido stage is configured so that the balanced signal from the DAC is transfered to unbalanced and any common noise to the + and - is suppressed. It means a cheap TREAD will be perfectly fine to power the DAC. Also the DC offset present from the DAC is used to bias the tubes.

I agree with Colin on the tube I/V. I have been working with the PCM1704 which is considered one of the top current output R2R DAC's. I have come to the conclusion that these types of DAC's are not very compatible with tubes.
Passive I/V even with a step-up transformer loads the DAC too much.


For now there are a couple of descisions I need to make:

1. CS4397 vs WM8471. I can get a CS4397 board from Ebay for cheap. The WM8471 board ends up $100 more. The CS4397 is well liked in the Zhalou but most reports I have seen favor the Wolfson.

2. Tubes - Octal like the 6SN7 or 9 pin like the 6922, 6H30, 12AU7, 12BH7.

 

[regal]

Here is the Broskies mark-up that shows what is being done. I can work up component values when we decide which tubes would be best.

 

[cgrums]

Hmmm.....this is exactly what my work setup is calling for. I'm definitely going to keep an eye on this.

 

[regal]

With the way the DC offset is used to bias the first tube, the resistor under the cathode resistor determines the load seen by the DAC.

This is the design point because the WM8471 does best with a >10 kOhm load. Russ Wright said the WM8470 did OK down to 1 kOhm. But probably best not to push.

Means we need a medium gain low current tube.

This leaves the 12AY7/6072 as the best tube choice for the input stage. It seems to have a good rep.

It will load the DAC at around 15k Ohms with a 1.75mA bias. Output will be about 3.5Vrms, a little high but I think manageable.

 

[cetoole]

Regal, why not just use a Broskie Cathode Follower? Sure, there are two extra coupling caps, but they are small ones, and it isnt like you can (easily) go DC coupled on this kinda circuit anyways. I really dont like the idea of loading the output of one of these DACs with the load that the + input will see, and they all put out 2vrms anyways. What is the point of a 10:1 voltage divider followed by 30x of gain? I know I wouldnt want a DAC output that hot (my DDE v3 is outputting about 1v, and I am happy with it. Also, the Broskie Cathode Follower will load both outputs much more equally.

 

[Cauhtemoc]

Quote:

Originally Posted by regal I have been working with the PCM1704 which is considered one of the top current output R2R DAC's.

The PCM1704 is the only multibit DAC still in current production, however there are better ones.

Quote:

Originally Posted by regal The CS4397 is well liked in the Zhalou but most reports I have seen favor the Wolfson.

The Zhalou uses a CS4398, not CS4397, but yes, most people will agree that the Wolfson sounds better.

Quote:

Originally Posted by tomb I think he decided that it's impossible to use tubes for I/V.

Impossible is nothing! Well maybe some things, but not a tube I/V.

Quote:

Originally Posted by regal This is the design point because the WM8471 does best with a >10 kOhm load. Russ Wright said the WM8470 did OK down to 1 kOhm. But probably best not to push.

Most voltage output DACs will handle a load of 1k, but it is indeed better not to go below 10k.

Quote:

Originally Posted by regal This leaves the 12AY7/6072 as the best tube choice for the input stage. It seems to have a good rep.

A tip, the current production Electro-Harmonix 6072 is very, very good.

 

[cetoole]

Quote:

Originally Posted by Cauhtemoc /img/forum/go_quote.gif Impossible is nothing! Well maybe some things, but not a tube I/V.

Not impossible definitely, but a good tube transimpedance amp without any iron is kinda tricky.