We have been sending the PRCL length to PRM M3 and M2.  Even with the modified driver we are using for M2, there is no frequency where we can get more actuation with M2 than either M1 or M3 has.  There have been a few problems related to this not great offlaoding.  Locklosses due to using up the M2 range, glitches when the drives were near 2^16 (alog 18983 ) and locklosses where a 20 second oscillation used up the M3 drive (alog 19464 ).  

We are now sending PRCL to both M3 and M1.  A screen shot of the measured crossover (currently around 0.5 Hz) is attached.  We are roughly compensating the suspension resonances at 1.4 and 2.8 Hz, just to prevent mulitple crossovers.  We are curently using a pole at 0.01 Hz, a susComp filter that compensates for the suspension resonances, 27 and 60Hz notches and an elliptic lowpass at 70 Hz.  It is probably possible to push the crossover up above the suspension resonances, as my original suspension compensation was designed to do, this didn't work, it might be that I accidentally used an undamped suspension model to design it.  

This is implemented in the ISC_DRMI guardian now, and is fine.  It works for PRX, but the gaurdian has not been updated yet so the old offloading will come on until that is done.  We can do the same for SRM soon.