DIY discrete dac chip.

I was wondering if anyone have made a discrete konverter chip themselves.

Should be a farly simple task. The circuit will give a lot of opertunities for optimizing the perfomance.

Sorry for my poor english but my english spelling check isn't installed yet.

/Lennart

There a different ways to make the dac.

Of coarse the r2r is one of the possibilities.

I was thinking of another way to make the dac. It will possibly be more precise than a R2R dac, The downside of this method is the price.

If there is some interest in this subject i could try to explain how I would implement the DIY DAC chip.

I'm not 100% certain that the dessign is bulletproof but a discussion in this forum will probably show if it is not..

Best Regards Lennart

why wanting better perfomance than R2R converter ?

Hopefully it sound better

There can be several reasons: the jitter can be lowered, the output can be made in a way that perfectly fits the next input stage..

Re: It will possibly be more precise than a R2R dac.......

Jocko Homo said:

I'm glad that you used the modifier "possibly".

Rather than explain how, I want to know why..........

Jocko

Click to expand...

I thought about your quistion once again.

I think a R2R dac is the easiest way to make a high perfoming DAC.

So maybe this thread should discuss how to make a god R2R dac.

Or do you think that the perfomance reached by the commercial dac's can't be beaten by a DIY dac chip ? If this is the case, there is of coarse no reason to make a DIY dac chip.

What about perfomance, if it is possible to get, let's say true 18bit resolution, will this give a sonic improvement? Or is the effort just wasted ?

If the "true resolution" resolution is just about 16bit, a signed magnitude, aproach can be chosen in order to get a good low level resolution. The downside of signed magnitude is the price/compexity, because two dac’s for each side has to be made.

My idea was to make a current switched DAC.

Basically it consist of 16current sources, the values should be proportional to the value of the bit.
I think a current source can be made with at precision needed to accomplish at least 16 bit of resolution.

A two poled switch should then route the current fitting the input value out of the dac.

I think temperature drift is a major problem, if a high true resolution should be reached. The constant current solution will have a very stable temperature since the current in the resistors, actually deciding the precision, is constant.

Oven controlled operation(temperature controle), can be used for the most critical DIY’ers.

A volume control can also pretty easy be made, without putting any components in series with the output. It can be made as a mixed analogue/digital solution. You could be able to change the volume 6dB analogue, and let the bit shift to a lower priority each time the volume is lowered more than 6dB.

I don’t know if I explained this good enough, to get others to catch the idea. But just ask then I will try to do a better job next time

Thanks for the link.

Could not imagine that a discrete DAC wasn't made before.

The outcome of this project wasn't good as i read it. But there was not put much effort into adjusting the resistors to correct values.

What do you think about the current switch DAC ?

Should i draw some schematic in order to make it easier to understand ?

The DAC should be calibrated, no matter which conversion method is used.

The calibration can maybe be done automaticly, by using voltage controlled resistors, but i think this is aiming to high.

Re: Uh, huh......Just as I thought.....

Jocko Homo said:

Weighted current sources.

Anyway.....why is a DIY DAC going to have lower jitter? The jitter problem lies upstream. (Carlos will tell you it is the filter that causes that, and it has to go. He is partially correct.) Making it less sensitive to jitter is what I believe that you mean.

Back to you...........

Jocko

Click to expand...

I ment that the DIY DAC could have a more precise timing, so that the jitter added by the dac is less.

But i really have no idea of how much intrisic jitter is added by a conventional dac itself. But the settling time is pretty high around 200ns for a PCM1704. I think this indicates that some jitter is added by the decive itself.

High speed devices normaly adds less jitter, isn't that so ?
I realise that this is a simplified view, and there is a lot parameters contributing to the overall jitter perfomance.

Terry_Demol said:

Lgrau,

Before you or anyone else embarks on this journey please
consider the following:

The best implementation of a discrete R2R DAC I know
currently available is the Lavry Engineering DA924 which is 24 bit
96kHz.
Before contributing any more to this thread, everyone here
should go to the Lavry site at:

http://www.lavryengineering.com/index_html.html

and download the manual for DA924

http://www.lavryengineering.com/white_papers/DA924m.pdf

Which will explain in some detail what is involved in getting
the kind of performance you are looking for from this dac.

Also consider that the latest breed of DAC chips can outmeasure
the DA924 with the right implementation (I didn't say sound
better).

Definately not for the beginner

Cheers,

Terry

Click to expand...

Hi Terry.

I really like this DAC from a technical point of view.
Oven controlled, auto calibration, glitch reduction, jitter removal..

All very nice features.

I wasn't aware of thesse kind of exotic converters. Can't really find much information regarding the sound of the Lavry converters, but it should have a better and more analog sound than a dcs converter.

a DIY project will not be as advanced as the lavry converter. But i think a DIY solution can be highly optimzed for the best sound, with smaller effort than put into the lavry converter.

I realise that a DIY solution can't compete with new DAC types like the PCM1792, only looking at the measuments of the DAC.

It would be very nice, with some input of the importance of different parameters..Should the effort be put into high linearty, or is it more important to reduce the jitter.

I know there is a lot of knowledge in this forum, if it could be used to make a sound optimized DIY dac chip, it would be very nice.

A DIY dac will also be the optimum "victim", for tweaks and opgrading, when new technologies and components is put out on the marked..


Best Regards
Lennart

Kuei Yang Wang said:

Konnichiwa,

Well, for such a DIY DAC we should consider some questiosn:

1) Do we require high long term stability in absolute levels?
2) Do we require high stability in absolute clock speeds?
3) Do we require high precision of the relative "AC" levels in the short term (minutes)?

If we decide that the long term stability is of no real sonic value (which I would submit to be the case), we can concentrate strictly on the "analogue" side of things.

Click to expand...

Yes we should only concentrate on getting the short term relative levels correct.

but i think the very critical parts like the resistor network will be easier to controle if the temperature is constant.

Kuei Yang Wang said:

In that case I would suggest that in the simple DAC shown the channel resistances of the individual outputs may be sufficiently large and sufficiently varied between individual outputs and devices to cause issues, plus PSU contamination will easily feed through into the audio signal.


I would suggest to add analogue switches with guranteed limits on variations of "on" resistance and a super well filtered and very low impedance (Op-Amp buffered, RC filtered reference?) low noise reference voltage which is switched by the Analogue switches onto the R2R network.

Sayonara

Click to expand...

Yes this some of the most critical elements in the R2R converter.

In the Current switch'ed(or weighted current sources) dac, the variations on the on resistanse isn't that important because the correvt current will be delivered, no matter wich serie resistance is added. To a certian limit of coarse.

Bernhard said:

No wonder that the performance of this DIY dac was bad, the guy used 5% resistors

Click to expand...

I didn't notice that, must be a realy heavy distorted low level perfomance this dac offers.

Here's a simplified schematic showing how a current switched converter can be made. The schematic only shows the operation of two bit. The digital section is also left out.

the rest of the schematic