Paul & Sabrina's cheap DIY 144v motor controller

[MPaulHolmes]

I'm finished right now with a power section, and I just barely started the control board layout. I'm going to try some stuff on the breadboard a bit more this time, to get a really really good handle on the noise. Also, a few professional motor controller people are helping me to work on the layout and code.

I'm sure there will be a well working controller by July! There will be improvements along the way, but the improvements will mostly be to the control section. However, I'd like to try Synchronous Rectification with a different chip from Atmel, the AT90PWM3, which has the same core as the ATMega8 and ATMega16. ya!

[MazdaMatt]

Hey, i've been off work sick for a couple days and just got my home internet up again. I happened to be cruising YouTube and saw the burnout vid... immediately jumped over here to see the progress! Awesome work! Now's when the fun begins... remember the 80/20 rule. It is now 80% done, which means you're through 20% of the dev time. Time to get the "bugs" (shudders, current mis-reads, etc) worked out.

Paul, I was just thinking about the concept of reading the current at the same point on the PWM each time. Presumably, the best time to measure is just slightly after the rising edge plus switching time? Or at least some time well into the "on" portion? Otherwise you'd be consistantly measuring something below max. As I mentioned about 300 pages back, there are microcontrollers that are created JUST for this sort of application that generate an automatic ADC reading after a set delay from the PWM rising edge. After you make it to your show with a "smoothed out" program, maybe it is time to step up your game again and dive into a bigger and better microcontroller.

[MazdaMatt]

Hey Paul. I found this article describing three "generations" of synchronous rectification techniques. Likely you'll want to start out with the first generation - calculate the expected turn-off time of your drive mosfets, give it a little space, then turn on the rectifying mosfets. It has some inherant flaws (anything that simple has to have flaws), but it will be a major improvement over diodes. Seems to me that leaving the diodes in place is not a bad idea anyway.

Give it a read-over just so you're aware of the other techniques in case you want to jump on them. Not sure if your ATM8 is going to handle the advanced techniques.

Predictive Control Maximizes Synchronous Rectifier Efficiency

[MPaulHolmes]

Originally Posted by MazdaMatt

Paul, I was just thinking about the concept of reading the current at the same point on the PWM each time.

I ordered a quartz crystal for the micro-controller, so it will have a 16.000 MHz clock now instead of 8. Also, I got some good advice today about how to generate an interrupt in the center of each PWM "on time" of the ATMega8/16/32, so I'm going to try that out. Then I'll read the current (a more stable value now I hope!) and update the throttle during each and every single pwm wave. I'm thinking it will make things pretty stable.

[dcb]

Originally Posted by MazdaMatt

(anything that simple has to have flaws)

I find most flaws in over-engineered things.

Before frolicking down the optimization path, one really needs to have a solid benchmark suite, to know if all the effort really made any difference.

[dcb]

Originally Posted by MPaulHolmes

I need a good source for 4 ounce copper pcb.

Well, the drains are bolted and bussed to the heat sink, you could bend the source leads up(down) and clamp those together in line with a couple lengths of bus bar. That should take care of the high current bits and allow the rest to use "regular" pcb.

So if this is the current schematic, you would run a large cable from the source bus to battery ground and another large cable from the heatsink to the motor (assuming the heatsink busses the mosfet drains and not individually insulated)

[MazdaMatt]

Originally Posted by dcb

I find most flaws in over-engineered things.

Before frolicking down the optimization path, one really needs to have a solid benchmark suite, to know if all the effort really made any difference.

Hehe... I meant in theory. Something so simple as a hammer is sure to have fewer flaws than an electromagnetic nail shooter. But you sure could optimize the hell out of that shooter.

Most certainly the thing needs to be smoothed out before optimizing in most cases. It seems that Paul is looking at a pretty big gain from relatively easy picking fruit by going from diodes to synch rec - it is just wise to look at the further options available and scale back to "bang for the buck" level of optimization. Heck, the efficiency gain from the finer points of hundreds of hours of synch rec development could probably be matched by just using better wires.

[Roger Heuckeroth]

I believe there may be some pitfalls in synchronous rectification that will drive you crazy trying to work them out. I would stick with the diodes subscibing to the KISS principal.

The diode is always there ready to freewheel the inductive current. With synchronous rectification what happens at current limit? If your PWM signal gets cut off how do you keep the synchronous mosfets switched? I'm sure there are ways to make synchronous rectification to work reliably, but its a lot of work to gain a bit more efficeincy.

Also, if both sets of mosfets short out, doesn't that make a direct short right across the traction battery pack. That will probably only last a few miliseconds before the main fuse blows or the power stage vaporizes.

My company makes air polution control equipment. If something happens during the warranty period, it almost automatically costs me $1000+ (travel, labor + material). With that type of incentive I have learned to make things as robust as possible. This is a bit different in this case in that worst case is a burnt out power board. Still not something that you want to have happen.

Just my two cents.

[MazdaMatt]

That's why you gotta build it right

You can leave the diodes in there AND use synch rec mosfets

[MPaulHolmes]

I hadn't thought of that!