Regen causing acceleration (sine / async motor) [SOLVED]

[Bena]

As the title says, my car is accelerating when it tries to regen. Happens with sine firmware 5.27R and 5.35R.

I have three suspects:

Inverter configuration is wrong. I followed the tuning guide to set it up initially. https://openinverter.org/wiki/Parameters#Tuning_Guide

Encoder on the motor. No specs, seems to be a magnetic quadrature encoder. Swapping A and B causes the motor to barely spin. It is a small ic potted in transparent resin (cannot see any markings), very close to a magnet attached to the end of the motor shaft. Magnet is polarized "sideways" so the field is spinning. Waveforms looked like a typical quadrature encoder. I tried to count the pulses per revolution spinning the motor by hand, but i would not claim the result as accurate.

Current sensors. LEM HAS-400-S. Original inverter board has some op-amps to convert the output to a 0-3.3 signal with 0 at the midpoint. Readings seem about right. I calibrated them at ~8A and checked at 1A and they are reasonably accurate.
Datasheet: https://www.lem.com/sites/default/files ... -s_v16.pdf


The car works fine if all regen is disabled, in fact i have driven it around the farm for almost a year now. 110km on the trip meter.

I think the problem is that the inverter doesn't actually know what the motor is doing. What parameters affect this? I have guessed the values for nearly all parameters since i have no real specs on anything. I have attached the parameters (regen disabled, did not save during testing)

I did some testing, setting regentravel to 20%, offthrotregen to 50% while keeping brakeregen at 0%. The car is low on power due to worn batteries, smaller regen values barely felt different from no regen. I drove uphill in first gear for this test. You can see the idc value going negative during regen, however car continues to accelerate. See the attached screenshot. idc and ilmax on left scale, fstat on right scale. Idc in the positive direction is confirmed accurate by a dc clamp meter on a battery cable.

Acceleration during regen seems to get more aggressive as motor rpm increases. Under 500 rpm nothing happens and the car just cruises. Under ~800 rpm the car does barely slow down on a flat surface. Slowing to around 500 rpm it will slowly oscillate for a few seconds between trying to slightly accelerate and doing nothing.

Another thing i noticed, reversing up a hill, stopping and letting the car start rolling back down forwards, i can drive forwards with the car still in reverse! Seems like the inverter is confused which way the motor is spinning. Regen does not do anything in this scenario.

Low-end torque is lacking, a lot of throttle is needed to get moving. Further tuning is clearly required, but made difficult by the poor battery performance. Difficult to tell if a parameter change increased power when the batteries are quickly becoming empty. I have ordered a nissan leaf battery pack to do a lithium swap, will probably take a few months to get it done after it arrives.


More specs:
Small chinese electric car from 2011 but registered as L7E fourwheeler (certification is for this vehicle class). It is a conversion of an ICE car. 5-speed manual gearbox included.

Batteries: 8* 12V 120Ah lead-acid (~100v no load, inverter udcmin configured to 72v). 250A main fuse.

Inverter: Shenzhen Greatland Electrics GLC16008A2 (96V, 330A). Original "power stage", current sensing, power supplies. Stock logic board did nothing apart from flashing an led. I desoldered the original DSP and a logic gate ic driving PWM outputs. Jumper-wired the OI mainboard V3 to appropriate locations on the original pcb. All boards in the inverter are double-layer only so it was fairly easy to reverse-engineer.

Motor: Nameplate has no specifications. Dyno graph from certification documents shows 20kW peak power. It is a normal three-phase motor, no permanent magnets.

[johu]

First assumption : you didn't count the correct number of pulses per turn. Look at one of the two signals and count only positive edges. There is no tolerance on this parameter (numimp) it must be spt on

[Bena]

First assumption : you didn't count the correct number of pulses per turn. Look at one of the two signals and count only positive edges. There is no tolerance on this parameter (numimp) it must be spt on

This was the case. I noticed the numimp was actually the default value, maybe i just forgot to set it initially... Encoder outputs 128 pulses per turn. I did not count the pulses manually, instead i looked at the flywheel from an inspection hole and compared the motor position reported by inverter to actual movement. Took three passes to get it right, last calculation gave 128.004 pulses per rotation.

Car would not move with this new value initially, changing motor polepairs to 4 restored movement. I guess the default values just happened to be close enough for it to work in the first place. Regen works now as well.

[johu]

Great