Paul & Sabrina's cheap DIY 144v motor controller

[DJBecker]

Originally Posted by MPaulHolmes

One thing I like about the dspic is that you can do 4 A/D conversions simultaneously. Throttle, current, temperature, voltage in my case. The specs are that it is basically a regular old control board, but now surface mount, dspic30f4011, with the driver section removed, and voltage monitoring included.

Do you see this as a big advantage?

I was initially disappointed in the slow speed of the AVR A/D converter, but it turned out to be only a minor problem when I went to write the actual code.

The current sensor is the only thing that needs to be constantly monitored. And doing that on alternate cycles seems to be good enough, leaving plenty of A/D time for everything else.

Almost all of the other sensors only need an occasional sample. The software limits the throttle ramp, so 10 samples per second would be fine. We are using plastic package temperature sensors (LM335 and MCP9701) where 1 Hz sampling is more than good enough. Sequentially scanning all 16 A/D channels on the "odd" cycle is easily fast enough for all of these.

The next step up would be sampling the motor current at the center of both drive and freewheel phases. But would that really tell us anything new? It would get rid of some sample noise and bias, but is unlikely to add more device protection or a better driver experience.

I initially had the battery current on my fast sample requirement list, but there isn't much that you can do with that in the fast control loop. You can only verify that the battery current is approximately proportional to the PWM percentage, and that is best done in the slow monitoring/communication loop.

Now if we had a really fast A/D converter, with programmable offset and a DMA transfer, we could make o-scope waveform data available to the dash display. But that's a big step up. And an AVR-class processor not only doesn't have the hardware, speed or computational horsepower, it doesn't have the memory.

[MPaulHolmes]

It's not an advantage at all really. Well, it's a huge advantage for the AC controller field oriented control code, where you need the 3 currents at the same time, but I just like it because the code is so simple in the DC case.

[DJBecker]

Originally Posted by MPaulHolmes

It's not an advantage at all really. Well, it's a huge advantage for the AC controller field oriented control code, where you need the 3 currents at the same time, but I just like it because the code is so simple in the DC case.

Surely you don't begrudge the five lines of code it takes to scan the other channels ;-)

You need the code anyway, if you want to report values from the other A/D channels.

For the new driver board, would it be possible to split it into two half-sized boards? With pads/holes for a chaining connector at each end? Keep the components tightly clustered in one spot so that it's easy to cut the board to as few as three or four devices. Right away that would double the order quantity, and perhaps make it broadly appealing for people driving arrays of MOSFETs for other purposes.

Also, you can get a more balanced gate drive by running individual traces to the gates rather than chaining them. That seems to be more important than making equal-length traces by doing back-tracking or serpentine paths.

[badfishracing]

Paul,

Long long lurker, first time poster. I'm trying to do a IGBT controller, my own flavor of microprocessor, basically your same driver. I was thinking of using two driver ICs, even going to the TO220 style. It doesn't look like you're tying the two drivers outputs together? Is there any risk of timing differences between the two drivers? Or is it best to just let each do it's own thing?

Darin Gilbert

[QUOTE=MPaulHolmes;215959]Here's the driver board I'm planning on using. 2 drivers should be good for up to 20 mosfets, but I'm just going to try 14 for my prototype. That's about 71 or 72 amps per device for 1000 amps. I'm just hoping that it could do that for like 10 seconds. Enough to get up to speed in stop and go traffic, and do burnouts.

[MPaulHolmes]

Hi Badfishracing! My understanding is that ideally, you should have a separate isolated supply for each driver, but that in practice, it works to just have 2 running in parallel, with outputs not tied together. I know Ian Hooper in Australia tried it with 6 drivers for 12 mosfets if I remember right, and it worked fine.

[DJBecker]

I haven't seen any design guidelines that suggest a power supply for each driver. Certainly you want good decoupling caps, following the usual guidelines: a small ceramic cap to damp the high frequency noise, and a bulk cap with low ESR that is more than 10x the combined gate charge to supply the instantaneous current.

If you connect the gate driver outputs together, you will get some amount of shoot-through. Even with perfect matching and putting them on opposite ends of the board, the power use will go up, perhaps way up.

The only reason I could see to connect the outputs of two gate drivers in parallel is to drive a single really big devices. The answer used to be to use a little gate driver feeding a totem-pole stack of big bipolar transistors. But today you can just buy a bigger gate driver. I've never used one, but my jaw dropped when I read the datasheet on this part:

IXDN430YI 30A 0.4ohm 35V TO-263-5 $4.40

[MPaulHolmes]

Here's the idea for isolated supplies per driver. I came across it when they were talking about 4 isolated drivers vs. 6 isolated drivers in a 3 phase inverter. For very high current applications, the B- bus bar's inductance starts to be important. So, during switching, some of the mosfets would turn on at different times, because of the potential difference from one end of B- to the other. But if you have 2 different isolated supplies that are both really connected to B-, when the ends of B- have a potential difference, it doesn't matter. Take that to the nth degree with 1 isolated supply per mosfet, and you have the "ideal" situation, except that it's hopelessly expensive and complex. haha

[DJBecker]

Originally Posted by MPaulHolmes

For very high current applications, the B- bus bar's inductance starts to be important. So, during switching, some of the mosfets would turn on at different times, because of the potential difference from one end of B- to the other.

You avoid that situation with MOSFETs by having a gate driver voltage significantly above the threshold, and using per-gate resistors. The gate signal starts looking like a current source, and is less sensitive to minor variations at the source pin. If the MOSFETs are similar, they all switch about the same time because they are synchronized by the Miller knee.

I can see the problem you are talking about being much more of a issue with bipolar transistors, where even a minor difference in B-E voltage makes a big difference in collector current. You can't match/equalize base current with a series resistor because that kills your gain. The usual solution is a low-ohm resistor at the emitter. (I say "low ohm" rather than "small", because that fractional-ohm resistor usually generates a lot of heat.)

[foor]

[QUOTE=MPaulHolmes;215665]I'm almost done with a 1000amp power section, control section, and driver board for the mosfet version. It's non-SR, since I just found some much improved freewheel diodes that are 150amps per device instead of the 60amp ones I've been using. QUOTE]

Paul,
your development of the 1000amp non-SR design is the most exciting news I have heard since back on post #600 where you got the original version running in your car. I believe this addresses the ONLY real short-coming with your 500amp version, at least from the open-source, KISS approach you have taken.
Will you be releasing the design details on this new version: schematics, layout, BOM, etc? Possibly another kit for your store?
-Wes

[MPaulHolmes]

yes. I'll post all of that. To truly keep it simple, maybe I should just keep the voltage monitoring, change the dspic back to atmega, dump the CAN, and remove the driver section from the control board, and make it all thru-hole. Mr. Bigh's boards are designed for 15v input voltage instead of 12v. HEY!

Check this out:
EC5A-12S15 Cincon DC/DC Converters & Regulators

I think it's even non-isolated. That would be perfect. And 10 watts continuous would be plenty for driving those igbts I think.