matching transistors & diodes

[Martin Czech]

There was a thread about matching transistors & diodes a couple
of weeks ago. We all know the "moog manual" method, constant Ie and
measure Ube. I've found somewhere a note that says : match at least at
two points.  That makes sense if the Ic is modulated over a wide range
(like in controlled gain stages). 

A switch for two reference currents is all that is needed to do this
with the above mentioned circuit. E.g. the test could be at 500uA and
1uA or so.

At 1uA you get something like the ideal diode characteristics and at
500uA the nonideal resitive component comes out. With only one measurement
point it could well be that two transistors seem to match, because ideal
diode characteristics and resistive components together are equal, this
could cause errors above or below the fixed Ic.

Now we've got two numbers, but we need a scalar to compare the transistors.
How can we come to such a scalar : arithmetic or geometric mean value
of the differences ?

Any ideas?

m.c.