Wow! Busy afternoon.Kepler said:OK, implemented AutoTorqPAS. It basically worked but still had the same dropout problem .
...
I flashed back in Prelim3 firmware and did a basic RPM PAS config. The RPM PAS worked perfectly without any dropout issues at all. Next I tried AutoTorquePAS with prelim3 firmware. This worked really well with no dropouts and operated just like AutoPAS on the latest version. Still needs some fine tuning but is looking good at the moment.
Interestingly, with the latest firmware, I couldn't get speedPAS to work at all. I could see the cadence bars going up and down but the motor would never activate. At the time, I didn't think too much about it as I figured i had missed a setting and was concentrating on AutoPAS anyway which did work.
Anyway, I have run out of options with Prelim5 and magnetic ring PAS. for now I will stick with B23 on the bike using a magnetic ring PAS and Prelim5 on my other bike with Torque PAS.
Thanks for all the tests - particularly the rollback tests to v3.0P3 which seem pretty definitive. Good stuff.
Too bad the AutoTorqPAS trick didn't get around the issue, but it's good to see that it had a successful road trip . You may have noticed that this is just a special case of the 'Tips and Tricks' gimmick in the Guide where the PAS Assist pot is attached to the Trq input so you can use a 3-pos switch at the same time for throttle control. Here we just left off the pot and rely on the pull-up.
Last thoughts: The only related thing that came up in testing and hasn't been posted was an issue with ThrI->MinInput being set too close to the actual Throttle minimum (dead zone too small; offset less than 0.1v). This is similar to the issue posted here a bit ago about an auto-cruise problem. With the offset too small, a little noise would cause the CA to think the throttle was applied briefly, knocking you out of auto-cruise. I had a similar problem in PAS testing where I couldn't get a PAS mode to start because the CA thought the throttle was engaged. Oddly, everything worked fine in the previous release but a subtle change in v3.0P5 caused this 'throttle detection' issue. I really doubt that this is in play in your case, but scraping the bottom of the barrel for ideas, it's the last config issue I can think of that sounds even remotely applicable.