Based on others experience having reported good results with ceramic bearings reducing the '76 sec' error I decided to give it a try and ordered Kit 8421 from VXB. At $39 each and using 3 bearings, it was a $120 experiment. I expected good results after reading some of the details on how these are made and their low rolling friction characteristics but was surprised how bad they performed. Below is only a small sample of the several tests conducted over a few nights to try to make these expensive bearings perform at least as well as their steel counterparts.
Note: PemPro does allows you to fine tune the PEC profile by adjusting drift and delays. I've included drift fitting screen shots to show 'none' and the best fit for comparisons.
Ceramic bearings tested with McLennan gearbox and Ruland coupler.
(Clicking on the images will open a larger version to better see the details)
Worm Block with 2 steel bearings reinstalled in the coupler side block and 1 ceramic bearing on the thrust side.
This is a test of using only one ceramic bearing on the thrust end (opposite the coupler).
The two steel bearings were reinstalled in the block on the coupler end.
Measurements with increased bearing preload.
The 2 mounting blocks were moved inwards to preload the bearings more. This reduced the '76' error a lot though PE was still high.
Summary
This was a big disappointment as I had better expectations. I can't recommend these bearings based on the test results shown but to be fair, these bearings had no ABEC identifications so at the least they were ABEC-1. Maybe a higher ABEC rating of 5-7 would improve it more as I suspect it will but those bearings are around $80 each and I don't want to risk another expensive mistake.
What was interesting is two different bearing types and still a small amount of 76 sec error (not occurring at the 80 second fundamental mark) is still present. What do the two bearing types have in common? Both have 8 balls in the race and designed for radial loads. See the section on '76 sec error investigation' for more info.