Voltage Mirror?

This is similar to cascaded PNP-then-NPN emitter followers, in that Vbe cancels to first order. That arrangement has a lot higher input impedance, since the PNP has gain. The problem, as you note, is matching differences in Vbe caused by heating of the output transistor.

Analog Devices makes matched transistor arrays. Most other multiple transistors are poorly thermally-matched monolithics or, worse, multiple isolated die in one piece of epoxy.

Thermal matching can be terrible in these:

ftp://jjlarkin.lmi.net/UPA800_80mW_one-side.jpg

This pair has, vaguely, 100K/W differential theta. So 100 mW dissipation in the output transistor would create a positive offset error of roughly 25 millivolts.

The bummer is, all sorts of fun circuits are blown away by things like cheap opamps.

John

It's an emitter follower with a V_BE in series with it. Used all the time for low accuracy stuff. For higher accuracy, use an op amp--it would need a bit of a bandaid to work at 100V, but nothing too awful.

Cheers

Phil Hobbs

The Vbe match of the two transistors is VERY bad; the emitter current in those two transistors isn't the same, after all. At best, this will buffer a Zener (voltage source) with a limited input current, and has somewhat better thermal behavior than a simple emitter follower.

Take a long look at a uA741 input circuit: the Vbe matching is assured by current-matched loads.

Bad? At most a few volts off since we are programming the voltage and not the current(hence it is working on a logarithmic scale). As long as the supply is stable it is not a big deal as I am using feedback to program it. Drift might be an issue to some degree but it doesn't have to be perfect in my case. I'm not looking for mV accuracy.

It shouldn't be too big of a deal as any current changes will have a logarthimic effect on the voltage. I'll be using feedback to control the voltage so the voltages do not have to follow exactly(or even close but just linearly).

Any cheap opamps that work at in the hundreds of volts range rather than

You do need to put a bandaid on them. It's easiest when the input range is within the op amp supplies, but you can use floating op amps to fix up the offset voltage of other kinds of buffers. Whether it's worth it depends on your application.

Cheers

Phil Hobbs

ge

Add in a (-12V) supply and for 0-100V in/out, use:

(warning, bad ASCII art here)

Input + | 10R | +----------+----10R---------+---(output) | | | \ +-R---(-12V) | | +-|- \ A | | >--R--+ +--+--| +----|+ / A / | | | | / --+-- 10R |\ R | | | | (GND) +----+--(+100V) (-12V)

(the "A" characters are intended to indicate NPN emitters)

It's basically a follower, but with input and output voltages attenuated, and with voltage-to-current converter on the op amp output pin, and a level translator transistor plus an emitter follower.