Hfe needs to be high enough to prevent excessive loading of the diff amp output. You might want to consider a Darlington.
Keep in mind that power dissipation might be an issue.
Why did you mention Widlar in the title? I know who Bob Widlar is, but I never heard of any connection between him and the Howland current pump.
Oh nvmd..I was looking at
this page explaining different current sources. It calls that circuit"Widlar Op-Amp Current Source for Grounded Load" and it looks just like an improved Howland current source with negative input to the inverting input. so i wasn't sure.
A Mosfet needs a gate voltage that is much higher than its source voltage for it to completely turn on. So it needs a boost-the-voltage circuit to drive its gate to the higher voltage so that it works as a source-follower.
The BUZ21 and most other Mosfets need the gate to be 10V higher than the source for it to completely turn on.
But it might work when the total supply is 10V and the maximum output is only 5V at 0.5A.
Thank you for reminding me that. This is a problem and instead of dodging the lack of high hfe by using a MOSFET, I think i should check out Darlington as Ron suggested.
Also I plan to use both 2.5V and 5V at the +IN of AD8276, and in the case of 2.5V, even a Logic Level MOSFET (Rds specified at Vgs=5V) probably wouldn't work.
I think a Darlington like this 2N6038 rated at 40W will do.
Also I had to loosen the tolerance to 1% on that 5Ω resistor to find one that meets the power requirement.oh well. i can address that in my calculations after I measure the actual resistance.
**broken link removed**
Just a comment from me. I have some experience with the Howland current pump.
Sense voltage across R2, the 1.5 mOhm resistor will be very small, so low input offset error is required from U2, the AD8603. If you substitute this part number, make sure that the substitute has low input offset voltage specified. I don't know if it is still true, but op-amps with bipolar input transistors had superior input offset voltage spec than FET input op-amps.
You might also consider eliminating U2 altogether. The value of Vref inside U1 is 40K. There is no reason why the output of R2, the 1.5 milliOhm resistor, would need buffering in order to drive a 40K load. The circuit will still maintain high precision without help from U2.
Solid. I got rid of the buffer.
++++++++++++++++++++
I have a question related to power Regulation for the said circuit.
**broken link removed**
In the sketch I get 24V from a noisy power supply, the LM2940T is a linear regulator (LDO) that outs 10V, and the three LM4040 are Shunt Voltage Reference rated at 5V 2.5V and 1V(this one not a LM4040).
I'm not so certain about connecting voltage references to the output of a linear regulator (I'm worried about currents and input impedance issues)
Can someone tell me whether it's ok to do this?