# Do I need isolation for op-amp stages in a current pump circuit like this?

U3A (OPA703) is used to shift a unipolar DAC signal down to a bipolar signal and feed it to U1 (OPA544). The OPA544 takes this bipolar signal and produce a pretty high current of about 3 Amps at the OUT pin.

In my opinion, U3A is the control stage and U1 is the power stage. I think these two kind of op-amp stages should be isolated, so I created isolated power supplies for them with different GROUND (DGND and AGND).

Does this affect the circuit's operation as a whole?

I used the current source design in LM675 datasheet: www.ti.com/lit/ds/symlink/lm675.pdf

• I answered, but only to explain why you shouldn't do what you proposed. What you actually should do (connect the grounds, partially isolate the grounds, provide isolation in the signal path...) depends on what problem you are trying to solve, which isn't revealed in the question. Jun 23, 2016 at 3:13
• What I'm afraid of is the noise introduced by the AGND because the load connected to OUT pin is grounded to AGND. Jun 23, 2016 at 4:15
• 1. Provide a low-impedance path from the load's return point back to U1 (and the ground connections of C21, C22, C30). 2. Keep U3 well away from this path. You could muck around with star grounds and guard traces and that stuff, but really those are mainly just ways to get you to do 1. and 2. Jun 23, 2016 at 16:01

## 2 Answers

No, you should not isolate the grounds of the two op-amps and then directly connect the output of one to the input of the other.

Imagine in your circuit if DGND were 50 V above (or below) AGND. Then the input to U1 would be well outside the common mode operating range. Realistically, the connection of the circuit would prevent the two grounds by drifting so much. But it would only stop them by allowing excess current to flow, resulting in unexpected circuit behavior.

Your circuit, as shown, will not work. U1 appears to be connected as a malformed Howland current pump. As shown, the output will swing from rail to rail. A Howland pump configuration looks like

simulate this circuit – Schematic created using CircuitLab

and it should be clear that, for 30 volt supplies and the resistors you've shown, especially R4, there is no way you'll get more than about 3 mA out.

So show us what you really have.