emetter follower temperature drift problem.

does anyone knows how to ajust the common emitter amplifier temperature drift for my audio circuit? i attached the sch here. the two output voltage drift differently. which is more bad. if they drift the same, it can be ok. but the drift differently.
affected by the temperature. can anyone help solving it?

Perhaps you can run a simple experiment 1st. Let the circuit stabilize thermally , next heat up carefully the NPN alone and watch the drift, next heat up the PNP, and watch the drift, finally the opamp, do the same. It would be preferably to have matched NPN/PNP monolithically.

In any event I suspect some of the characteristics with the NPN. Anyway if you could run the experiment then share the info. To het it up, for one cpu could simply put your fore finger on it with some pressure.

Once you figure out which components are responsible for the drift (using Earl's method, above) then perhaps put those components into an "oven" -- i.e. a small container that has just enough space to contain the components, a resistor (used as the heating element) and a thermistor (used in a feedback loop with the resistor to maintain the temperature at near constant). Choose a temperature that is hot enough to stabilize the drift in all environments this circuit will be subjected to, yet cool enough to minimize thermal stress on the components.

If it turns out to be the two transistors, you could also try connecting them thermally (smear a little heatsink compound on the flat faces of each transistor and then press the faces together and wrap them with something or otherwise bind them together). That way, their temperatures are more likely to track. Even better would be to stack a couple of SMD transistors together (one NPN the other PNP) using, once again, a little heatsink compound between the transistors.

BUT, you are dealing with microvolt variations, so good luck ;)

Another consideration, are you sure it's not your meter?

Is this based on simulation, lab results, or both?

If I understand your notation properly, you are seeing a drop of 120uV on the high side versus 30uV on the low side as temperature is increased (how much?) with DC signal inputs.

Given that each opamp is in a voltage follower configuration, I believe the difference is opamp related. Several parameters play into your 'room temperature' results including the offset voltage, the bias currents, the non-symmetrical drive capability of the opamp, the open loop gain, etc. Over temperature, all those parameters will move.

Looking at the datasheet for the OP295, the output current versus temperature curves show for a single supply configuration, the ability to source current drops off faster with increased temperature than the ability to sink current, which would affect the high side opamp more. The curves show the reverse capability for a dual supply configuration. That may be something you can test.

The opamp can also drive closer to it's negative rail than its positve rail and in this configuration the high side opamp has to drive Vbe above the output voltage (closer to the suppy rail) compared to the low side which only has to drive Vbe below the output voltage (closer to ground). The drive capability decreases with increased temperature.

All that said though the PNP seems to have less drift. There must be a reason. The op amps are in the same package and on the same die so they are fairly well matched (should be). So looking at the individual drifts should point in the right direction.

Wait to see what comes out of that.

Hi Earl,

Agreed the opamps should be reasonably well matched and I made the assumption that if this is a real circuit, that no significant temperature gradients exist across the circuit (op amp in particular).

Even assuming well matched opamps, there exists the difference in respective function that each half of the opamp package is performing. One half is sourcing current, driving toward its positive rail, and the other is sinking current driving toward its negative rail (ground in this case). The opamp performance is non symmetrical with repect to its rails and some of those non-symetries seem to diverge with increasing temperature, based on the datasheet curves. I suspect that is where the drift originates. Note even at the initial temperature there is a difference in input-output errors for each opamp.

The transistors are mostly out of the voltage control function, the output errors are reflected back to the opamp inputs and the opamp is doing its best to drive its (non symmetrical) outputs to minimize those errors. I once worked on a circuit where the opamp ran out of gain, even at DC, due to a diode in the feedback path (like here).

As Steve pointed out, measurement error could certainly be a factor. Knowing if this is simulation versus real would help rule out some possibilites.

"The transistors are mostly out of the voltage control function, the output
errors are reflected back to the opamp inputs and the opamp is doing its
best to drive its (non symmetrical) outputs to minimize those errors. I once
worked on a circuit where the opamp ran out of gain, even at DC, due to a
diode in the feedback path (like here)."

Yes the amplifier states a gain of 1000, so one can not expect the inputs to be the same. In any event it is interesting that the difference is more with a lower Vbe on the NPN device while the current gain is more when the transistor is hotter. In any event that's the part I don't quite understand. The asymmetric aspect of sinking vs sourcing is interesting.

The part that is strange is that the NPN circuit is worse in regulation which I didn't quite expect. Measurement error, could be. Didn't ask what kind of measurement instrument was used or if multiple meters could be used for verification.

In any event the amplifier gain could be increased by cascading stages and that should reduce the overall error. Also I would suggest building two of these circuits to see if they both do the same.

Something isn't making sense in terms of what one would predict.

The temperature coefficient of the op amp Vos dominates the circuit tempco. It is 1uV/deg C typical and 5V/deg C maximum.

The op amp spec gain is 1,000,000 to 10,000,000.