What is causing mismatch in this current mirror?

[solis365]

This is the circuit in question. The operation is thus: The opamp drives the emitters of Q1 and Q2. The bases of these devices are grounded, so the output of the opamp is sometimes negative. The opamp works to establish Vref at its inverting input. Thus, the voltage on Rin will be VCC - Vref, and the current I_in will be (VCC-VREF) / Rin

Since the opamp is controlling the V_be of both transistors, they should both sink approximately the same current; of course the Early Effect must be considered, which would cause mismatch if R_out was different from R_in.

I have prototyped the circuit on a breadboard using an LM741 opamp and an LS313 Matched Pair audio BJT transistors. The datasheet isnt that great but they are similar in performance to the SSM-2210 matched pair, which has a much better datasheet.

For certain combinations of Rin and Rout, the matching error becomes huge. But there is almost no discernable trend!

For all of the graphs, the vertical axis is matching proportion, calculated as ABS[(Iout - Iin)/Iin]

It doesnt just happen at low current, or high resistor mismatch, or anything that I can tell. does anyone have any ideas where the mismatch is coming from??


Note - Amp Out is the voltage at the output of the amplifier, or V_E of the BJTs
Note 2 - Vin and Vout are the voltages ACROSS the resistors, not referenced from ground.

 

[MrAl]

Hello there,

Without studying this too deep yet, i see you have a possible source of high imbalance sensitivity to one parameter. That parameter is Vbe for the two transistors. The circuit is driving the two be's with a low impedance source, and with only a tiny difference in Vbe you might see a large difference in collector current. There's no wiggle room for even tiny differences in the Vbe's.
Try doing a sensitivity analysis for the two Vbe's and see if it starts to match your data.
The fix might be some small value resistors in the bases or emitters, if that's the problem. This might cause an offset, but at least it will be more predictable.

 

[solis365]

Ill try sticking some series voltage sources into the base of each transistor and seeing how the current varies as that changes. However I don't see how the Vbes can be different at all, they are shorted together. I feel like adding base resistors would only exacerbate the problem, as there is no way that the resistors will be matched as well as the BJTs are.

could you explain a little more about how the Vbes would differ?

 

[KeepItSimpleStupid]

A 741 OP amp?


Vos and Ibias will be OP amp errors.

I don't know what power supplies your using for the 741 either nor the current your trying to mirror (500 nA?)

 

[solis365]

the 741 is on +/- 10V rails. Vos shouldnt affect the matching between Iin and Iout, just the absolute matching between Vref and Vin.

I am mirroring a range of currents, as shown in the graphs. ~500nA is one of those, but the error is not ALWAYS high at low currents... in fact, sometimes the error is very low at those low currents. My question is why isnt there a discernable trend in the matching error? it doesnt seem to be correlated to anything in the circuit.

 

[RCinFLA]

Cannot tell from the info you provided.

Assuming the devices are infact matched devices and do not have a temperature gradient between them the causes of mismatches would be too little Vce and devices too large for small currents. Beta peak Ic should be no more then a couple few hundred times above operating Ic for good matching.

I assume you have sufficient negative supply for op amp and the system is not oscillating.

 

[solis365]

I assume you have sufficient negative supply for op amp and the system is not oscillating.

System is not oscillating, but this made me think. Just looked up the output swing of the 741... with +/- 20V supply it can only swing +/- 15, with a 15V supply it can only swing 10V... I have a 10V supply so it looks like I may be hitting the rails for certain cases. some of my output voltages are 6.2V, which sounds like it might be right at the limit of the output swing.

I should have just started with a better amplifier... I was afraid to toast the $20 AD549 precision amp that I have set aside for the job.

 

[ronsimpson]

Matched transistors are pretty good but I never connect two transistors together b-b and E-E. Add two emitter resistors. Re=Rin/10. This will cause the op-amp's output to go more negative. I would aim for -5 volts at max current.

 

[crutschow]

The op amp output always is very near -0.7V, independent of transistor current, since that is the typical Vbe for the transistors. So I see no problem with the op amp if it is operated from ±10V supplies.

The cause of your mismatch may be differential heating between the two transistors. Any slight mismatch in voltage across the transistors. Are Rin and Rout identical?

To minimize this mismatch you need a dual-transistor with both transistors to be on the same chip.

 

[ronsimpson]

The transistor is a dual transistor. From the first post "LS313 Matched Pair audio BJT transistors".

The Vbe is matched to 0.4mV typ. 1mV max.

Over extreme temperatures the match is 1mV typ and 5mV max.

The Hfe matches to 5%.

 

[MikeMl]

Answer: input bias current of the 741, which can be as much as 500nA.