DC Trimming an op amp output

[saint gillis]

We have several ways to remove DC after an op-amp output, among them : AC coupling with a capacitor, Servo DC circuits, ..

I guess we could also bias one of the inputs, something like that :


Is that method reliable? Will it drift with temperature? Will it be stable over a long time?

 

[CJ]

they gots some pretty good opamps nowadays as far as offset, how many micro volts is too much?

maybe use two resistors for R2, 2.35 M 1%

and a 20 turn pot,

 

[Newmarket]

Everything has some temperature dependence. What tempco on the resistors, particularly the trimmer ?

 

[saint gillis]

maybe use two resistors for R2, 2.35 M 1%

and a 20 turn pot,

Why 2 resistors instead of one?

Everything has some temperature dependence. What tempco on the resistors, particularly the trimmer ?

I don't know, the question is purely theoretical, I was also minding if the op amp itself would drift

 

[Newmarket]

Why 2 resistors instead of one?


I don't know, the question is purely theoretical, I was also minding if the op amp itself would drift

Resistor drift likely to dominate. Even using eg 10ppm precision resistors

 

[ccaudle]

Another change which may help stability is to connect V+ to resistor to pot to resistor to V-. You will not ever need to set the pot all the way to V+ or V- if you are just trimming DC offset, so restrict the pot to being used for very small adjustments around the chosen set-point (which is presumably 0V). Choose very low temp coefficient resistors, and they should dominate. Obviously choose a low temp coefficient pot as well.

I was also minding if the op amp itself would drift

Input bias current changes with temperature. An op-amp specified for low offset should describe the unit-to-unit range of input bias current as well as the temperature behavior of bias current.

You would likely find description of offset behavior and how far it is possible to minimize in texts describing instrumentation and measurement designs. That is obviously something of concern for precision voltmeters, for example.

 

[KA-Electonics.com]

The typical bias current drift of a 5532 is 5 nA/°C. No maximum spec is given in the ON Semiconductor datasheet.

With perfect trim components there is still a 110 µV/°C (typical) error term from the 5532 with a 22KΩ bias resistor.

 

[Hubbub]

For an AC coupled input the most certain way of achieving stability is a design that does not rely on resistor tempco or supply stability. Stabilizing the output with a CAZ (commutating auto zero) type amplifier servo scheme is the most predictable way. The output too be stabilized feeds a real slow integrator with at most a few uV input offset. The integrator output then feeds back to the controlled amplifier input so as to null out the controlled amplifier's offset(s). Now resistor and capacitor tolerances affect the corner frequency, and maybe feedback loop gain, but not the offset stabilization voltage.

 

[JohnRoberts]

We have several ways to remove DC after an op-amp output, among them : AC coupling with a capacitor, Servo DC circuits, ..

I guess we could also bias one of the inputs, something like that :
View attachment 119254

Is that method reliable? Will it drift with temperature? Will it be stable over a long time?

As Wayne already replied that (old soldier) op amp will be the weakest link for DC stability in that schematic among several (like PS rails).

There are better modern op amps, but what is the actual problem you are trying to solve?

JR

 

[Newmarket]

As JR asks - what is the problem to be solved wrt DC offset ? For audio we can basically use capacitive coupling and sort it like that. Or servo as appropriate whilst acknowledging and accounting for the time constant.
Or simply use very low offset devices and live with that.
For DC instrumentation issues it becomes more challenging. But I believe modern chopper op amps are now very good compared to some years past when switching noise breakthrough could be a real issue. fwiw When this was a thing when requiring very high DC precision in analytical instrumentation we used OPA177 to replace the choppers. The internal "customers" for this found it fine in the application.

 

[saint gillis]

Originally, I was looking for ways to trim the differential offset of a THAT 1646 (for reasons that would be long to explain here), and I thought of trimming the op amp driving the 1646 to achieve the goal.
Then I was just minding if this way of biasing an op amp was stable enough, just for the sake of learning.

 

[JohnRoberts]

Originally, I was looking for ways to trim the differential offset of a THAT 1646 (for reasons that would be long to explain here), and I thought of trimming the op amp driving the 1646 to achieve the goal.
Then I was just minding if this way of biasing an op amp was stable enough, just for the sake of learning.

Long term stability is always an issue with such trims.

Years ago I considered the possibility of incorporating a cheap MCU driving a few DPOTs to periodically trim various aspects of analog circuits and them save those trims in flash memory until revisited again. This could behave similar to a DC servo but without the inherent HPF.

This involved more complexity than benefit and I abandoned the thought experiment (I was going to tweak more than just DC).

JR

 

[Flatpicker]

FWIW I've been working on a few mic pre designs with DOAs feeding the 1646 going directly out to ADC inputs of my interface with no issues of DC that I can tell. I have bipolar 10uF's between the output and sense lines of the 1646 as suggested by the datasheet.

Now as for the connection between the op amp and the 1646, that's a different story and various amounts of DC are exhibited depending on the op amp used (and you know I had to try them all, lol!). I did a lot of listening test between using coupling caps (Nichicon 'audio' ele's) and a servo using the OPA177 here and could not distinguish a difference in sound whatsoever. Again, this is just MHO and YMMV, but today's ele caps are pretty doggone good!

 

[jokeramik]

Due 100% DC-feddback you don´t need this.
Best regards!

 

[gyraf]

What about a single-node DC servo?

 

[drtechno]

Why 2 resistors instead of one?


I don't know, the question is purely theoretical, I was also minding if the op amp itself would drift

All drift, its when its not acceptable. NE5532 is strapped like that in many balanced line driver circuits. But I wouldn't use it as a buffer in an oscilloscope. So it depends on the application. DC coupling is the best.
I wouldn't use R2 and the potentiometer.
If you have dc offset, then adjust Rin (22K) but most likely any offset would be more of a product of the +15 and -15 Voltage not being the same in the opposite directions.