- DSC02111-800x800.jpg (108.74 KiB) Viewed 768 times
- DSC02110-800x800.jpg (128.37 KiB) Viewed 768 times
Parts of the system will always have to be analog as there is no other way to determine amplitude and position of things.
that's an easy one: ice bears are in the north, if you see penguins you're at the south pole!!
I mean it would be instead of hall sensored start. And hall sensors divide the electrical rotation into only 6 pieces, so it is quite coarse. The one from the video was setup up for dividing an electrical rotation into 512 pieces, so should be much smoother. But am not there yetizeman wrote: ↑May 08, 2018 3:14 amwould your sensorless start improvments help with "regular" sensored start in some way as well?
even though the controller/software now (since last autumn's update with hall data integration in sensorless mode) works better than ever, there is still some improvment possible in the (sensored) start situation.
start under heavy load can be very rough and stuttering, and power application is still very "soft". soft so far is great for all day riding, but can be hindersome in cases where you need instant power like wheelies for hopping over a fallen tree or something.
Interesting, what is the function of the additional analog processing between the current sense and the MCU?Lebowski wrote: ↑Apr 14, 2018 12:32 pmBusy with some extremely interesting stuff.
I've started programming on a different chip, one of the 33EV types. Goal is to have sensorless start with full torque at standstill. Relative (sensorless) position at standstill is quite easy, but to get absolute phase information requires a bit more doing. The picture above shows the different versions of analog signal processing that I am considering for processing the current sensor signals. The left one has 4 opamps, the right one 8 at the moment but should be 10... but have already scrapped that one in my mind and am drawing up schematic for version 3....
The new processor plugges onto the powerPCB. The analog bits plug into the bottom right header of the experimental processor board.
You could do this really easily with delta sigma based current sense, you could sample the same data sets with different digital filters to simultaneously produce a low latency 10 bit output and a slower high resolution 16bit or 18bit output. I don't think dspic has the needed hardware for that though unfortunately.
That would be ideal but I can't think of how to do that without resorting to sensors.Lebowski wrote: ↑May 15, 2018 11:14 amMy goal is to have something which can sense the correct phase even when the motor is (mechanically) locked, so that the motor can provide torque in the correct direction against the mechanical lock. So no cheating by having a necessary small bit of movement to try and figure out which way we're going....
The math is horrendous though, got pages and pages full of sine and cosine etc
I imagine the challenge is current sense range, you need enough to run the motor at full power but somehow have enough resolution to detect very small changes when trying to determine rotor position while unpowered.Njay wrote: ↑May 16, 2018 12:30 pmPhase inductance should be different depending on rotor position. If that's enough info, and "easily" measurable, to know the exact rotor position, it's another story. Maybe it's also possible to use a phase to induce current into another phase and from that get position information.
I actually want to sense changes in phase inductances (to deduce position) while powered, i do not want something where i have to alternate between powering and sensing. The video a few posts back was sensing and powering at the same time.lizardmech wrote: ↑May 16, 2018 1:16 pmI imagine the challenge is current sense range, you need enough to run the motor at full power but somehow have enough resolution to detect very small changes when trying to determine rotor position while unpowered.Njay wrote: ↑May 16, 2018 12:30 pmPhase inductance should be different depending on rotor position. If that's enough info, and "easily" measurable, to know the exact rotor position, it's another story. Maybe it's also possible to use a phase to induce current into another phase and from that get position information.