Next version: adjustable frequency range, and adjustable sensitivity Help us out Soeren
That would take the circuit from simple to something quite involved and I guess Webbot was aiming for letting even the wettest B1FF have a chance.
Frequency range selection would best be done in DSP, where you can make extremely steep filters (assuming a fast enough clock speed). An HP and an LP, both done as Chebychews should go a long way (for robotics, a passband ripple of 3dB or even higher shouldn't be an issue).
A similar analog filter (HP and LP) of a much lower order (i.e. not very steep) would take a load of opamps and would be a nightmare for an amateur (you will need horrendously precise resistors and caps with a low tc).
The adjustable sensitivity is easy, in Webbots circuit, just change the 100k resistor - lower resistance means more negative feedback (and hence lower gain).
Personally, I would have thrown an extra transistor into the circuit, added decoupled emitter resistors (to stabilize the transistors and stay in control of their input impedance) and most important, decoupled the bias voltage by one or two R/C circuits - to keep electrical noise from modulating the bias supply which will be seen as audio on the mikes output pin and then be amplified like any real sound.
The most usefull I could do here, would probably be a dual layout with a relatively balanced and noise free gain stage and with the right ECM capsules (there's a large variety in the range of max. allowable SPL from one make to another), I guess it could be set to a high gain and just go into a variable potential divider - perhaps with a "digital potentiometer" controlled from the µcontroller.Disclaimer: Just from the top of my head right before bed time.
If anyone is serious about making sound based circuits, I suggest looking at either dedicated DSPs , dsPICs or similar.