CS4329 DC offset

[theAnonymous1]

I have a CS4329 based DAC and there is a small amount of DC offset on each channel. I can measure the offset on the input and output of the opamp.

I found this schematic for a different CS4329 DAC board and it looks VERY similar to mine. Could resistor tolerance mismatch be the cause?

http://www.quadesl.com/pdf/cs4329_dac.pdf

An externally hosted image should be here but it was not working when we last tested it.

 

[analog_sa]

Time to get one of those fancy voltmeters that actually provide readings in numbers

Are you are measuring the same offset at the ouput of the dac (between the two outputs)?

 

[georgehifi]

When you say a small amount of DC offset, the OPA2132 does give +- 2-4mV in itself which is quite acceptable, what opamp are you using?

Cheers George

 

[theAnonymous1]

Very sorry guys.

OPA2134PA....

Left input: 1.258v
Right input: 1.267v
Left output: 20mv
Right output: 9mv

CS4329 outputs....

Pin 13: 2.194v - Pin 14: 2.188v = 6mv
Pin 17: 2.206 - Pin 18: 2.205 = 1mv

After closer examination the above circuit is identical to my DAC. The parts values differ a bit though.

 

[Cauhtemoc]

Yes, resistor mismatch appears to be problem.

 

[analog_sa]

Everyone shares the blame: DAC, opamp, mismatched resistors. You should certainly match the resistors as this has an effect on distortion. It is not so obvious what to do with the remaining offset.

 

[theAnonymous1]

Thanks guys, I'll try replacing all the resistors with hand matched values and see what I come up with.

 

[theAnonymous1]

A quick thought...... If I replace the resistors and there is still some offset is there any way to build a trim circuit to null it? A pot across the 5v supply connected to the input of the opamp?

 

[analog_sa]

A quick thought...... If I replace the resistors and there is still some offset is there any way to build a trim circuit to null it? A pot across the 5v supply connected to the input of the opamp?

There will certainly be offset even with perfectly trimmed resistors.
What you propose is possible but not without issues. The trim circuit will also unbalance the differential gain so other resistors will need adjustment. Using the 5v supply is certainly not a good idea though - it will be injecting ps noise.

Cap coupling is almost tempting.

 

[theAnonymous1]

Are you telling me I'm SOL? Is there nothing that can be done?

I went through great pains to get rid of the caps on the amplifiers input, this offset problem is the only thing keeping me from DC coupling (my goal).

I'll do anything it takes!

EDIT: I think I've got it!

When I put a 1Mohm resistor across R17 the offset goes from 20mv to 48mv, if I put it across R15 it goes down to -7mv. Those two resistors measure 14.93k and 15.01k. A significant mismatch no?

Thanks!

 

[analog_sa]

It's a simple choice. You can either match the resistors closely for best common mode rejection and minimum distortion or for minimum output offset but not both. You can compensate all sources of offset in this manner but at a price. Hopefully thermal drift won't be too much of an issue.

A listening test might lead to the best compromise.

Out of curiosity how are you handling volume control?

 

[theAnonymous1]

It's a simple choice. You can either match the resistors closely for best common mode rejection and minimum distortion or for minimum output offset but not both. You can compensate all sources of offset in this manner but at a price. Hopefully thermal drift won't be too much of an issue.

A listening test might lead to the best compromise.

Out of curiosity how are you handling volume control?

Thanks analog_sa, that clears things up.

Maybe my logic is wrong, but the way I see it is if some or most of the offset is from the resistor mismatch and it sounds good now, then mismatching them on purpose to get no offset shouldn't change the sound much, if any.

The volume will be done by a 10k pot in front of a buffer.

 

[theAnonymous1]

OK, I put in some hand matched resistors and now the voltages of the differential outputs match. I now get -3mv and -4mv offset. I'm sure the leftover offset is from the OPA2134. Maybe I will put in some OPA134 on adapters with offset trimmers.

Thanks again all!