You can only really measure the RXQUAL through a drive test (thats the most normal way at least).
As said before there are two lots of RXQUAL to measure. If you’re in dedicated mode and your network uses DTX then all measurements should be RXQUALS. (RXQUALF will still give a measurement but will almost always show a poorer result so should be ignored). The same applies to RXLEV measurements.
Normal measurements for RXQUAL are 0-5 are good call quality, 6-7 is poor quality, and should trigger handover selection if possible, if not the speech becomes unintelligible and dropped calls may occur.
Also as mentioned, the BER is the method for the measurements. The RXQUAL is just a scale given to ranges of the BER to simplify measurements.
FER is a a measurement of how the codec is decoding the frames. As the radio conditions worsen, more errors are introduced and the ability of the TRX to rebuild the frame obviously reduces. Hence the FER rises. Normally, unlike the BER/RXQUAL the FER tends to shift quite quickly.
All 3 should be looked at in conjunction with RXLEV to make a reasonable assumption about the conditions of the call at the time.