It doesn't take a long time in audio electronicsland to need some way to convert voltages from linear to log with a high degree of accuracy (wow that sounds geeky?) If you work with 1V/octave audio VCO's for instance you will need this.
OK the first thing you need is to match transistors' Hbe to create an accurate long tailed pair; If the pair isn't matched, your 1V/octave conversion will probably be butt. Beyond VCOs we see this matched long tail pairs in high performance discreet preamps....you know that as well?
Anyway, how to match? We're in the Interweb age, and MFOS has pages on it; the legendary Ian Fritz has a PDF on it, from which I copy my test circuit; this youtube dude has a vid about it; I think he copies Ian's circuit and methodology? Who i am to cast stones there?
Or: you can use transistors arrays like the 3046 still if you look around, but 3046's only match to +/- 5mV Hbe, and we want more like +/- 1mV. Finally, you can buy matched pairs and log conversion-ready arrays from vendors like small-bear (I freaking love small-bear) but any of the THAT corp. etc. ones are going to set you back in terms of dough. many of us have a whole bunch of 2N3904's just waiting to be matched....
I went a slightly different route: I scared up some .01% 40K resistors from Electronics Goldmine (one of my favorite surplus places) and used 4 of those in circuit.
The breadboard tester complete with cheap ZIF sockets:
The schematic (hand drawn, PDF) for this here.
Most schematics I see for this show 100K as the resistors in circuit. At 40K .01% I have found my match readings (using a 5.5 digit DVM) to be even more sensitive than if I selected say 200K .01%. So I am going to match within +/1 1mV in my rig and then plug it into some log converters. Stay tuned to see how well the converter scales.