1.) Looks like there is a strong subharmonic (-5dBc at the high end) - can this be filtered out, or are there any silicon rev changes that will improve this?
2.) VCO autocal
:It looks like the default autocal time is about 30 milliseconds. This is a very long time for us. Something like 0.3 milliseconds would be preferred, but I understand if there are hardware limitations making this impossible. I played around with the R9 settings and it looked like it can lock a lot faster. Do you have any preliminary data for a wider loop bandwidth (20-100 kHz), or any preliminary data on register 9, which controls the VCO auto-tune timing?
Alternately, is there a way to read / write the registers affected by auto-cal (e.g. one slow sweep while we read the ideal VCO settings at each frequency and store them, then simply write them back with autocal disabled for subsequent sweeps)?