Years ago I tried to read up on all the ML patents and white papers I could find. Not sure I ever understood it completely and the time elapsed since hasn't helped. Basically all the FBS detectors from Minelab transmit square waves at two different frequencies. One frequency is varied slightly during a noise cancel. (Fun, hold the coil up close to your earphones while you are wearing them, and hit the noise cancel button. The coil will induce a current in you headphone speaker coils and you can hear the frequency shift. You have to be a geek like me to consider this fun.)
It's looking now that Minelab's patents on multi-frequency technology other SMF detectors out there work in the same way on the transmit side.
The two transmit frequencies can be used to generate many frequencies on the receive side. What I believe I remember is that if a detecting frequency resulting from a noise cancel ends up on or close to a 50 or 60 Hz power line harmonic they just ignore that frequency because it is very likely to be very noisy.
I've seen some videos where people think that certain noise cancel frequencies can make a difference on what the detector can find. I'm pretty well convinced this is BS, and described why somewhere on this forum years ago.