outputting CV (DC voltage) with a soundcard

[mescalinum]

I've been thinking to this subject since some time.
Nowadays we have access to cards with 16bit DACs for very cheap, and 20bit, 24bit also very accessible. (moreover I actually have 14 unused DACs, 24bit and 20bit, from my soundcards).

Unfortunately most of the soundcards available, even pro soundcards, are not DC coupled, so it's not possible to directly output a CV.

Tried on my Fostex VC-8, and it resulted in something like a 30Hz hi-pass filter on the DACs.
Tried on my Echoaudio Gina24, this behave a little differently: with an op-amp buffer it could keep the voltage, but a little drift (say 5%) would happen, so not properly reliable, i.e. nevermind controlling a VCO.

What are the options?

I opened my Fostex VC-8 and found two electrolytic caps for each output, but shorting them didn't work (moreover, VC-8 uses delta-sigma converters; would be ever possible to have DC come out of them?)

So I started to think to a more general solution, and "demodulators" came to mind.
Either 16bit or 24bit (read: cheap or pro), we should be able to produce a tone (0 DC) of constant frequency (say 8kHz) and variable amplitude.
Rectifying it (diode bridge, superdiode, ...) then extracting amplitude (I tried with a 2nd order butterworth filter) we should be able to get a constant DC voltage out of it (loosing one bit cause of the rectifier, still [ideally] 15bit / 19bit / 23bit would remain!).

I tried, both on LTspice and on breadboard, measured the ripple, and didn't seem big (unfortunately I didn't take note, but will do).
If anyone's interested I can post schematic.

What do you think?
Do you have ideas for improving it?
Reasons for not to do it?
Alternative solutions?

I made also some tests with a frequency to voltage converter, but the ripple seemed to be much higher at lower output voltages. (butterworth filter exhibits opposite behavior (!)).

P.S. this is my first post here, so, hello to you all!

[varice]

to electro-music, mescalinum!

Here is a topic that you may find interesting:

http://electro-music.com/forum/topic-12266.html

It addresses some of the issues of your post.

In my opinion, it would be best use a soundcard that does not have a DC blocking cap before the output, or one that can be modified to bypass the cap.

[mescalinum]

so here's the schematics and output of a simulation.
I'm going to test this on breadboard and will report measured ripple.

---

demod_sch.png
Description:	one of the designs I've been working on. first stage is an active rectifier (peak amplifier), second stace is an active butterworth filter, low pass, 2nd order.
Filesize:	19.73 KB
Viewed:	307 Time(s)
This image has been reduced to fit the page. Click on it to enlarge.

---

---

demod_out.png
Description:	spice output. green track is virtual modulation signal, generated from a sequencer [0..1], enters the circuit as a AM signal 10 Vpp, carrier 10 kHz. red track is the output. if 0.2ms of timing constant is too much you can lower C2 and C3, but you will be
Filesize:	34.16 KB
Viewed:	257 Time(s)
This image has been reduced to fit the page. Click on it to enlarge.

---

[robotfunk]

perhaps this is of interest:
http://muffwiggler.com/forum/topic-13698-0-0.html

[Tim Servo]

There are DC capable sound cards, although they tend to be on the expensive side. There is also software meant for producing control voltages using those sound cards. Otherwise, if you're going to go with the envelope follower method, you might want to look at this design by Harry Bissell. It uses multiple stages to get fast response and low ripple:

http://www.edn.com/article/490267-Envelope_follower_combines_fast_response_low_ripple.php

worth a look, anyway.


Tim (slow response, LOTS of ripple) Servo

[Paco]

and something like this?

http://www.expert-sleepers.co.uk/siwaacencoder.html

[forbin]

I think that the biggest problem that you will have with most soundcards producing DC is the Codec or Digital-to-Analog Converter. Pretty much all the soundcards that I have seen recently use Sigma-Delta converters. The way they work is basically a one bit converter running at something like 256 times the sample frequency (so usually they are clocked well into the MHz). Because part of the design is an integrator which tends to accumulate a DC bias they nearly all (well all the ones that I have seen anyway) have a High Pass Filter that rolls off down in the sub 10Hz region to overcome accumulated DC error. This also helps with pops and other nasty DC offsets getting into the analog chain downstream. Problem will be getting a DC signal through... One option might be to use a TDA1543 (goggle it for info) which is an old-school resistive ladder design. There is a lot of discussion as to whether these really are 16 bit though and generally is thought that they are really only 12 bit internally. The capacitors on the output are probably only removing a dc bias on a single rail design?

[mescalinum]

[forbin]

Hi mescalinum

Your right about the DC output. It was the Input of the codecs that I was looking at. I was trying to use a dsPIC to make a VCO and wanted a high resolution DC accurate converter. I am yet to find one... even the AKM's!

[mescalinum]

[forbin]

No all the Codec's that I have looked at (not just AKM) have a HPF on the analog in. It is to remove DC offsets in the input but seems to be inherent in the design of a lot of Sigma Delta converters. I want a fast, accurate BUT also DC accurate converter to do phase/amplitude/frequency modulation in my design. You can get dc accurate sigma deltas but they are pretty slow... I will keep looking... The PIC ones are OK but you need really quite high accuracy and they are only really good to 12 bit and even that is a bit iffy really. I tend to find that really the top couple of bits on most converters are noise at the best of times... Ideally I would like 16 bit 40kHz dc accurate converters...

[skrasms]

[rumpofsteelskin]

A more convenient way might be to do this might be to output the CV in binary on a serial port (you could use a midi port) and use a look-up table on a pic to convert back into CV