Paul & Sabrina's cheap DIY 144v motor controller

[bennelson]

Wait a minute!

So the one I have is MISSING safety features!!?!?

I know I've been a guinea pig on this project, but this is rediculous!

[MPaulHolmes]

LOL. Well, it has all the safety features of the regular old 500 amp version. I just added a few extra things to the newest one like disabling the PWM if the microcontroller latches up. I tested that circuit, by the way, and it works perfectly!

As far as touchy throttle goes, there are a number of things that we could do to fix that. I'm thinking that part of the problem is trying to fit 0 to 1175 amps in the same range as 0-500 amps. On the Zilla, Otmar doesn't use a linear throttle if I recall correctly. We could first try slower ramp up, and if that's no good, then we could make it something like this:

We could transition from a throttle that feels proportional to 500 amp of max current near zero throttle, and a smooth transition to a throttle that's proportional to 1000 amps at max throttle.

I may need to mail you a little programmer for it, since it doesn't have a bootloader. boohoo. They are pretty cheap though.

EDIT: Something like this:

currentReference = 500*(x*x + x), where x is the throttle position, and is [0,1]. For x close to zero, x*x is very small, so it behaves like 500*x. For x near 1 (full throttle), x*x + x is close to 2, so it behaves like 1000*x.

I think Otmar uses a cubic, so maybe something like:
500*(x*x*x + x). That would be like 500 amps for longer, and a more severe transition near the end of the throttle.

The top line is what the throttle is right now, next one down is 500*(x*x+x), the next one down is 500*(x*x*x + x), and the next one down is the throttle for the regular old 500amp controller. Notice how the 2 bend to mimic the 500amp throttle on the low end, and catch up to the current 1000amp throttle.



If those friggen algebra sudents tomorrow tell me that there's no use for parabolas, I'm going to whup them and show them this.

[steiner]

Paul,

I am always curious why people don't use the pedal for velocity control instead of torque control. I realize torque control is much more similar to current automobiles. However, I have an off-road electric vehicle that has a Kelly controller (not what I would recommend) and it has the ability to use either torque mode or velocity mode. The velocity mode is kind of like driving around in cruise control. I simply push down on the pedal relative to how fast I want to go and the controller adjusts the current to deal with the changes in load. It didn't take long at all to get use to this and I really like it now. Wouldn't this get rid of the touchy throttle?

Rick

[jackbauer]

I think i'll take one of the new versions. I'm concocting an evil plan to build a 1ka sepex controller.

[MPaulHolmes]

OK, so the new controller blew up. It was my fault though. The failure was from excessive current, I believe. There were 2 issues I think:

1. I re-enable pwm in the software automatically after <= 1mS regardless of if current has gotten under control. Before shipping it, I almost disabled that feature, but you live and you learn. Normally, the overcurrent protection should never come on. But I didn't have my hardware overcurrent trip point and max software trip points set just right. I think current climbed higher and higher. The hardware overcurrent disable just slowed the climb I think.

2. Ben had an extremely sensitive throttle for some reason. In the garage testing I didn't really observe this, but it was only at 60v and less than around 120 amps. This could be due to noise on the sensor at higher power maybe. I believe the M- bar was basically an FM radio transmitter. With a low side controller like this one was, soldering the mosfets to the M- bar I think caused the oscillations on the oscilloscope that I was seeing. And it was sensing current on that bar.


The bar added some capacitance I guess to the M- signal, which was a square wave. On a high side controller, the mosfets backs are against B+, which is nice and calm. So, I'm almost done now with the final touches on new boards for a high side controller. I needed to order revisions anyway, so this should work really well I think. There will be a mounting point for M-, M+, B-, and B+! Current will be sensed on M-, which will not have much varying voltage, since M- and B- will be a single bar. M- comes in one way, you sense the current and leave as B-. It's weird, but the current coming in and the current leaving will be different. haha.

Hey! I need to change my signature.

[jyanof]

Originally Posted by MPaulHolmes

OK, so the new controller blew up.

Bummer!

So, I'm almost done now with the final touches on new boards for a high side controller.

I forgot what DCDC you were using on this new controller, but i just thought I'd remind you that those 6W Cincon DCDC's won't work in a high side configuration.

There will be a mounting point for M-, M+, B-, and B+! Current will be sensed on M-, which will not have much varying voltage, since M- and B- will be a single bar. M- comes in one way, you sense the current and leave as B-. It's weird, but the current coming in and the current leaving will be different. haha.

so, in the new configuration, M+ (and the source pins of the mosfets) will now have a high dv/dt, right? Would you worry about any stray capacitance there?

just curious - could you leave it as a low side switch, but do the 'single bar' thing for M+ and B+ with your current sensor there? I bet you could hack this together even with the hardware and PCBs you have. This would put your current sensor on a 'solid' bus bar and M- will still have a high dv/dt, but maybe the problem goes away if the current sensor isn't on it. oh, and gate drive stays easy.

[bennelson]

Yep, I've been doing some secret beta-testing the last fews days of the 1000A DC Open ReVolt controller.

I'm getting MUCH better at blowing things up. I must have toasted this one in HALF the time as the 500 Amp controller. I'm getting more efficient!

Again, I take my life in my hands, testing out EXPERIMENTAL electronics hardware in real-world conditions. I find the flaws so you don't have to!

I think GM and Toyota should do this - just let lots of people onto the streets with early versions of their cars, and see what happens.....

Anyways, yes, I blew it up. It was rather undramatic. There was no "failing on" motor run away like there was when I blew the 500 Amp controller. In fact, we think that the controller failure was due to a miscalibrated "safety feature". (In event of something unsafe, your controller will blow itself up to stop you from driving!)

Pre-charge and main contactor control worked great. I really like that as a feature built in right on the control board. It even still worked AFTER the power section had all its magic smoke removed!

Here's a couple photos of the controller after the incident.

http://300mpg.org/wp-content/uploads...1/DSC_1015.jpg
http://300mpg.org/wp-content/uploads...1/DSC_1014.jpg

On the second photo, if you look on the left side, between the two vertical aluminum spacers, you can see a diode that's totally ker-plow. Because all of the MOSFETs are sandwiched tight below, I really couldn't get a good photo of any of those. Judging from the nice even soot pattern, I would say that I smoked far more than just one of the MOSFETs. Are we taking bets on how many I blew?

If you would like to see some nice photos of the (non-blown-up) 1000 Amp controller, I have those posted at: http://gallery.me.com/benhdvideoguy#103390

That also includes some photos of the 1000 Amp and the 500 Amp controllers right next to each other. It's sort of fun to visually compare the two. The 1000A is longer and has all the capacitors in a very tight staggered configuration.

[MPaulHolmes]

It's just that I was warned about soldering the mosfets to the M- bar. The source legs are going to be going crazy up and down, but the new driver has 6 isolated supplies, each of which won't care about what the neighbor supplies are doing, one supply babysitting 2 mosfets. They were suggested to me by Fran specifically for high side drive. It's actually 6 homemade dc-dcs.

I think your idea about including an M+/B+ bar is a really good one, but I'm just worried that part of the mosfets failure and horribly dirty gate turn-on was because of the noise on the M- bar. I think with the gate to source having to stay in -30v to 30v, the mosfets will have a longer life if the gate turn on doesn't look like it's going from -5v to being clamped at 27v from the P6KE20A-T TVS diode at 200 MHz.

This is a guess, but I think the bad part about the capacitance on M- was noise the gate was picking up from it. the source capacitance might not be the same problem because the mosfets' gate was really really close to the radiating M- bus bar, since the mosfet back was soldered directly to it. So we are talking 1mm or however thick the isolation is between the gate inside the TO-247 package to the copper back of the package. I didn't see noise like that on the 500amp version, and I'm guessing that it's because the M- bar was "far away" from the inside of the mosfet case. There were instead solder joints and PCB tracks, and then some silver conductive epoxy, and finally the M- bus bar.

So, the only component to suffer being soldered directly to a wildly changing bar would be the diodes, but they don't have any control signals for them, and I think they are a little more robust because of it. I of course don't really know what I'm talking about, but I really really wanted to clean up that horrible gate turn on.

[jackbauer]

Anyone who has worked with power electronics and not blown something up hasn't actually worked with power electronics.

[DJBecker]

Originally Posted by MPaulHolmes

1. I re-enable pwm in the software automatically after <= 1mS regardless of if current has gotten under control.

Our controller has software-only overcurrent control, which means we are much more careful about it. (Plus we had two batches of counterfeit IR MOSFETs which caused us to rewrite the limit code a few times.)

We check just before setting the PWM register, and skip the next PWM cycle (set the width to 0) whenever the current is above the limit.
This is a separate, explicit check that is not part the normal control loop. That way changing the control loop constants won't accidentally defeat the current limit.

We hit the current limit frequently, and can tell immediately by the hissing noise from the motor that results.

We also clamp the MOSFET to the live bus bar. We haven't seen any gate drive issues from that configuration, even with one DC-DC converter and a single gate driver driving all of the MOSFETs. It may be because we run a twisted pair from the gate driver board to each MOSFET. The actual connection to the MOSFET pins is made with a tiny (0.5"x0.325") circuit board that has a 10R gate resistor and 1R0 source resistor.

That reminds me, you might consider adding a gate drive LED for debugging.
Our gate connection circuit board has pads for a TVS (SMBJ12A), and a LED+resistor (both 0805 size). The LED is only installed on the end MOSFET board. Even pointed the wrong way, the reflected glow lets us know that the gate driver is working during tests. In retrospect, we should have put LEDs on the gate driver board itself, but they had already been made.