Separate to the development of the 3-phase controller, handling the power requirement was also required.
The intention is to use simple electronics and deal with the issues of high power electrics separately.
To that end a test-bed vehicle was made by converting a Daihatsu HiJet van.
This was selected due to the access under the floor to the power train.
The van floor was completely removed as was the existing piston-engine power train up to the prop shaft.
Then an ex-milk float motor was added and connected directly to the prop shaft.
Also a wooden bench was fitted in the van over the motor for experimentation purposes.
This circuit is just for reversing the series wound milk-float motor using contactors:
Blue is MCU PIN36 output, Green is gate, Red is IGBT1c, 1uS/div 20kHz near 0% mark.
An alternative to using a zener to cap the gate voltage is to provide a LV supply for the driver circuit.
The problem is that about 2v is lost on the gate is the battery discharges to a low voltage, so the gate may not switch well,
but it means that the gate should switch on cleaner against the LV supply using C2 as ballast for the spikes.
Also 11R series gate resistance was added to attempt to dampen the charge/discharge ringing.
The circuit seems fast switching (transitions are much less than 1uS) but the problem is noise.
The frequency of that supply oscillation is ~6.4MHz, period ~156nS.
The on delay from the IGBT spec is 150nS so could be that.
This is the current seen by the controller during the test:
Shades go: current sample (darkest), moving average over 4 samples, then 16, then 64 (white)
The IGBT is allowed to exceed it's constant current rating of 100A for brief spikes, but is reigned in quickly.
The average load seems to be held at around 35A so the device could probably be allowed a higher limit, but in this test the controller limits well.
Now testing on the vehicle with the wheels on the ground again.
Basically the current limit is not high enough for it to move much, but that's the point.
Yellow is control input, green: motor output
and the top plot is current (raw-grey, 4 point moving average, 16, 64-white)
Obviously this is limiting the power output as you can see the green plot is being capped.
The motor was held stationary so the max power output was expected to be the same.
Possibly the current sensor is too simple as the sensing is sampled at a low frequency (in the 10Hz to 100Hz range) and may miss the peaks.
Maybe develop a circuit with an amplifier and capacitors to store the peaks for a 100mS so the sampling is more reliable.
More Sensors For Less
For testing purposes all the thermal sensors will need to be observed individually.
In the final controller they can probably just be bridged as only overload is needed and should be a rare occurrence.
Thermal sensors don't spike or change very quickly.
So the addition of an external multiplexer allows this.
The controller just times the pulses from the hall switch to get the speed as the gear ratio is fixed at 40:9 in the diff and each wheel rotation is 1.67m.
Just a quick test to assess the usefulness:
Left scale: rotations per 100mS, right: MPH, the red line is the control.
Incorporating another run into the graphing:
Top to bottom: current (A), voltage (V), power (W), Speed (MPH), and control input (left scale), motor PWM output (right scale)
So the gearing of the vehicle will allow up to 45mph so 30mph should be attainable.
Just a short road test to show it moving under it's own power:
The first test (above) was using a hand held control, but a foot operated control is needed.
A simple cable operated one does the job for now:
Took it down the street, 3-point turn and came back.
Top to bottom: current (A); voltage (V); power (W); left:Speed (MPH), right(yellow):economy (MPGe); and left(blue):control, right(red):motor
The motionless sections of the graph are truncated.
The speed is not very accurate currently, but a software revision will fix that.
During the run the controller handled a peak current of 414A and many times held the motor at 200A (over 10kW) without problems.
The outbound run was 12 seconds and accelerated to 15mph, the return was 8 seconds and got up to 13mph.
Also since both the power and speed are measured in the controller it's possible to calculate the efficiency.
Even convert it to MPG equivalent, which in an electric vehicle moving at around 15mph is around 500MPGe
The van accelerated to 17mph before stopping with the brakes.