INA118 instrumentation amplifier PSRR

Hi Rick,

I just duplicated this test for 3 INA118's in the lab here. I configured them for a gain of 100 and swept the power supply from +/- 10V to +/-15V with the inputs shorted together and attached to a 1.5V power supply. Of the three units, the worst unit's output voltage changed by 1.43mV over this range. This is an output referred offset shift of 143 uV/V or an input referred offset shift of 1.43uV/V . The typical spec is 1 + 10/G or 1.1uV/V in this case which is close to the measured value of the worst unit. 

Looking at the max spec for the parts, in a gain of 100, you could conceivably see a change of 6 mV over this power supply range for the PB/UB version of the part, or 11mV for the P/U version. However it would be unlikely to have this happen on 3 out of 3 boards. 

It might be helpful to see the schematic of your circuit to see how it differs from my test circuit if you are seeing offset shifts much greater than this. 

What is the difference between offset RTI vs. power supply and the power supply rejection ratio?  I think I'm not understanding something here.   As for your test circuit, it sounds like you have it down.  I think the only thing I forgot to mention is that the source impedance of the signal input is about 250 ohms.  It's a 500-500 ohm voltage divider from a 3V supply that feeds both inputs.  This is for test only, as the actual application uses a wheatstone bridge.  Not sure if that matters for this issue.  Thanks.

Hi Rick,

It's just different terms for the same measurement PSRR is defined as change in input offset divided by change in power supply voltage. As an example from above, the output offset of the part changed 1.43mV over the swept range. Dividing by the gain the input offset changed by 14.3uV. The change in the power supply voltage was 10V, to calculate the PSRR in dB, we get 20 * log (14.3uV / 10V) = -116.89 dB. 

The curves in datasheets are generally taken on a few number of units and don't guarantee the performance of the device. Specifications in the table with a max spec are tested to ensure that the product meets that spec before it is shipped. Therefore, typical curves can show different values from the spec table. 

However, I'm still not sure that we have reached the root cause of your issue. In your original post you stated the offset was changing by "tens of millivolts". Do you have a more exact value? Are you changing both supply voltages simultaneously? If you are changing one supply voltage only, then this will introduce a common-mode error as well. Also, if you are only measuring the output with a volt meter it would be good to examine the output with an oscilloscope to ensure that there are no oscillations that are being averaged into the measurement. 

I ran a simulation yesterday initially to determine if source impedance could introduce this error and Tina-TI did not show an effect. I will test this on the bench to confirm the reality of this later today. 

Hi John,

Well, I guess I had a bit more to learn about IA circuits.  I think I didn't understand PSRR because I wasn't looking at it with respect to the input.  I also have egg on my face about the circuit, as, after going back and looking at the schematic, there is a second gain stage with a gain of 11 in the circuit before the output measurement.  I was able to modify the circuit such that the first gain stage, the IA, is now powered from an on-board +/-3V regulated supply.  The second stage op amp is still powered from the unregulated supply, but the drift is now within acceptable limits.  Thanks for all your help on this.  I will have to read up on this a bit more... there may be more specs I think I understand but actually don't.

Hi Rick, 

Glad I could help. I'll admit that I had been scratching my head the last couple of days trying to figure out where 10x greater offset change that you were seeing could come from!

The topic of PSRR comes up quite a bit on here so I  may write a short post about it on our blog, The Precision Hub. 

Good luck with your project!