AW: LOG100 current mirror

[Haible Juergen]

Not many replies to this topic so far ...
Interesting circuit - selected diodes should work as well as the expensive
LM394 (?) - the transfer courve of the transistors / diodes performs a
compression
of the operating voltage for the opamp mirror - this should help avoiding
offset
problems from them opamp, and keep the voltage drop low.

but:

Have you noticed that the input voltage of this circuit follows the output ?
That is, if you want to source the output current into a grounded load,
the reference voltage of the input will change ? (not a problem if your
input is a true current sink with high impedance)
But even worse, try to control a 3080 with this circuit. 3080 pin 5
is almost at negative supply voltage, and so is the input reference
node of the mirror !
To avoid this, one could fix the output voltage of the mirror to some (high)
level. First shot would be a pnp in common base configuration
(base at 10V DC, feed in current mirror output at emitter, and control
OTA from collector), but then there is the 1/beta error which you'd
also have in a simple 2-pnp-current mirror.
Next shot, use a p-channel JFET. Should work, but not a cheap
solution either.

Haven't done further investigations yet - but the circuit is surely
interesting
(with the extension of a pnp transistor) when you can live with 1/beta error

and only a npn pair is available.

Comments ?

JH.