Need help to minimize the DC-drift on my OPA2209

Xun,

OPA2209 draw a total IQ of 2*2.5mA = 5mA at room temperature which in worst case should result in chip warming up by delta T = P*134C/W (for D package) = 5mA*30V*134C/W = 20C.  Since the maximum drift op OPA2209 is +/-3uV/C (see table below), the largest Vos drift you should see is 2*20C*3uV/C = 120uV (2 times because the drift is not correlated between two op amps).

Since you claim that the total drift is as large as 300uV, this make me think you have a resistive load at the outputs that increase the total power dissipated in the op amps.

Do you have any resistor connected between the outputs?  I assume you do not short the outputs without balancing resistors.  Under quiescent condition, you should not see a larger Vos drift than 120uV - anything larger than that indicates that the output sinks or sources current which further warms up the chip.  If that's the case, the only thing you can do to minimize the drift is to reduce the loading.

An another explanation for the excessive OPA2209 drift you see could be that you use resistors to short the the inputs to the ground.  Even though the OPA2209 has a IB current cancellation circuitry, they do not exactly track over temperature.  Thus, as the IB changes in each op amp by slightly different amount it will causes a different drop across the input resistors and shows up as additional drift at the outputs.  For example, 0.2nA differential IB drift would result in 200uV for 1k input resistors.  If this is the cause of your drift, you need to completely remove or minimize the size of the input resistors.

Xun,

I think you are attempting to do what I had suspected you may be doing - I had brought the issue by asking if you have any resistor connected between the outputs and commented that "I assume you do not short the outputs without balancing resistors. "  But it seems that is exactly what you do when you say: " if I short the output of OPA2209, there is minimum dc-drift from THS4524, that's why I am picking on OPA2209." 

If you short the outputs of OPA2209 to the ground or to each other, you will push the OPA2209 outputs into short-circuit condition where around 65mA is drawn from each op amp (total of 130mA) .  Under such conditions, the chip will attempt to warm up almost instantaneously by: 15V*0.130A*134C/W = 261.3C! and if it was not for the internal thermal protection circuitry inside the OPA2209, you would have just destroyed the part due to excessive junction temperature. 

The "drift" you see is not really a drift by the effect of the thermal protection circuit turning on and off.   At around 170C, the OPA2209 circuit detects excessive heating and in order to prevent the damage to IC turns off the output current - the thermal protection is turn-on and the outputs becomes high impedance.  As the chip slowly cools off, at around 150C, the thermal protection circuit turns-off but the chip quickly is pushed back into thermal shutdown due to persisting short circuit condition.  This can go on and on forever as long as you keep outputs of OPA2209 shorted. I hope this helps.

Sorry for misleading you, but when I said "" if I short the output of OPA2209, there is minimum dc-drift from THS4524, that's why I am picking on OPA2209.", I removed the OPA2209 at that point just to prove THS2454 was not the source of the DC-drift

This is the dc-drift curve