Weird controller behaviour

Hi all

My project (electric go kart) is almost finished but I've hit a problem that I can't seem to figure out. Any ideas would be greatly appreciated.

My controller doesn't respond to the throttle application. The only way to get it to spin the motor is through connecting the "self-learning" wires together. Upon connecting them it acts strangely and I don't know what to make of it. The motor will run at a medium speed in one direction. If I give the throttle a twist, it will then stop and then run in the opposite direction. It will continue to run until I twist the throttle, whereupon it will stop and reverse again. If I hold the throttle, it stops and then spins in the opposite direction twice as fast until I let it go. And then it the same story again...
Any one had the same issue?

Controller - Sunwin SH60U45GV18T-WZ2
Motor - Heinemann PMS120-WB294

It sounds lime the self learning procedure that the controller is running through. The throttle during this time may be an instruction to reverse the rotation.
You need to find out what the condition of the procedure is... Some controllers you just leave for a minute or two but others require a specific completion process to store the learnt info.
The throttle not responding until the controller has learnt the motor sequence is normal as it doesn't know what to do.
Google could help, but someone here will know the Sunwin process for sure.

SOLVED!

Basically, my controller is 60v from factory and my battery is about 46v, so the low voltage protection kicked in and that's why my throttle was unresponsive. I think self-learning function ignores the LVP which is why it still spun up the wheels but of course I couldn't control it beyond just that. Still don't know exactly why it behaved the way it did in self-learning mode but it doesn't really matter now. Main thing it works ok, now that I've opened up the controller and changed it to 48v (desoldered the relevant tab). Did a small shunt mod while I was at it too

Well done.
AFAIK the way the self learning works is the controller puts an ordered firing sequence through the phases then checks the hall response. If it isn't suitable it tries a different phase combo then checks again. Then it runs through the sequence through the halls and sorts the best combo and rotation direction. The motor often does some weird lurching while this is happening and then shortly after runs smoothly in whatever direction it lands on. Would explain your observations.
Don't quote me on this though...i could be spouting utter bs