Paul & Sabrina's cheap DIY 144v motor controller

[DJBecker]

Would removing the CAN make the hardware much simpler? A MCP2515 controller and transceiver are inexpensive, and make the controller far more expandable.

I would also bump the priority of a tachometer input. I use a ICP pin to trigger an input capture register freeze on a free-running 62.5KHz 16 bit timer, but a GPIO pin triggering an interrupt or just read in the 16KHz timer loop would give enough resolution to be useful.

Once you have CAN and a tachometer input, it's only a little more software to implement redline limits and fixed-RPM operation commanded over CAN. That gives you both cruise control and PTO/idle-speed option when paired with a instrument cluster ECU that has switch inputs. The instrument ECU can deduce the gear from speed vs RPM calculations, making per-gear redlines and shift control possible.

I also recently added a remote throttle CAN message. It was mostly for testing, but would also allow the body/instrument ECU to handle the pedal input. Combined with an isolated CAN transceiver ($3-$4 for a TI ISO1050 transceiver instead of $1 for a non-isolated, albeit the isolated part is only available in an 8SOIC package), that would make a traction voltage referenced motor controller a little safer to implement. (Although there isn't much of a problem leaving an engine compartment TPS referenced to traction voltage.)

Both command-operation messages (fixed-RPM and throttle messages) have a one second timeout as a fail-safe mechanism. The controller needs to send frequent updates or the cruise control / throttle drops to zero. The set-redline command setting persists, although a watchdog reset could unexpectedly clear it.

The point is that having CAN bus opens up opportunity for many advanced features. They can often be implemented without even changing the motor controller firmware. Putting a cap on motor current or speed when in reverse, motor speed matching for shifts, tweaking the current limits and P-I parameters for dropping traction voltage and S-O-C, and dozens of other features.

[mrbigh]

Originally Posted by 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.

>
I would keep the DSPIC on a socket (SMD if viable)and keep the CAN on board with voltage monitoring and pre-charging relay output; thru-hole assembly and remove the driver's section (keep it external for two different output versions, MOSFET and IGBT) to make the PCB more versatile.
The parts for the IGBT driver modules are in, I will assemble one for a picture shooting session.

[bga]

Heavy copper tracks.

A good way to beef up the wide copper tracks on a power PCB is to solder heavy gauge solder wick or copper strands along the track. The track has to be mostly bare copper, free of resist, to do this.

[DJBecker]

Originally Posted by bga

Heavy copper tracks.

A good way to beef up the wide copper tracks on a power PCB is to solder heavy gauge solder wick or copper strands along the track. The track has to be mostly bare copper, free of resist, to do this.

Most commercial boards have a solder mask coating. This is very difficult to mechanically remove without damaging the copper layer. Using solvent is considered an even worse option -- it damages the board and wicks under the foil, breaking down the adhesive all along the trace.

Remember that adding metal lowers the resistance, but does not increase the surface area thus doing nothing to improve the cooling. Doubling the thickness results in less than a 1.4x increase in current carrying capability. And since solder has about 10x the resistance of copper, you should only count the copper added when estimating the increased cross-section.

[sawickm]

Originally Posted by 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.

Hi Paul,

Can't wait to see your "Rev.E" Controller !!!

It would be nice to keep it compatible with Adam's RTD Explore software, that would mean continue using the ATmega. The CANbus option would also be a good feature to have on your new controller.

When you say that you are removing the driver section from the controller does that also mean the driver power supply too? Most of the industrial IGBT driver control PCB's contain the driver/driver power supply as one unit .igbt-driver.com/

That means the "Rev.E" ATmega Controller would just use a simple regulator power supply for the ATmega, control, saftey, and interface logic.

-Mark

[MPaulHolmes]

For me it was easier to leave the DC-DC on the control board, since there was other stuff up there like the low voltage disable circuit for the 12v supply. There would have been a lot of things that came down. Usually the professionals use drivers that have all the built in protections, but all mine are individual components. The only signal I really really need is the output of the optocoupler, which is a tiny amount of current, so it can be longer without the di/dt penalty of the long gate drive wires, which are carrying I have no idea, but like over 10 thousand times the current each. Also, the voltage monitor circuit uses that isolated 12v supply, so it would have to come on down as well. PRetty soon, it's a big dang party down there!

CAN is in addition to the serial communications so it has both right now. I can also do the serial communications to mimic the output of the atmega no problem. So RTD would still be OK. But the thru-hole version of the dspic30f4011 is huge, and the surface mount is TQFP44 with 0.5mm pitch, which isn't my favorite thing to hand solder. Also, that's the only one that has both serial and can at the same time. Well the 4012 does too I think, but the serial pins are used by something else important that I can't remember right now on the 4012.

[DJBecker]

If you stick with an ATMega AVR, that probably means using an external MCP2515.

Which isn't a bad thing. It shares most of its SPI connection with the ISP header, so you only need one or two dedicated pins. And you might be able to use the inputs and outputs of the 2515 as slower GPIO.

[MPaulHolmes]

I'm done with a basic thru hole version for the 1000amp power board and driver board with the voltage monitoring. It's basically just the 2C control board, with voltage monitoring, no driver section and no CAN. It also has larger isolation rings, and thermal relief. I priced 10 surface mount boards to be placed automatically, and it was over $1000. hahaha. That sealed the deal for thru hole. I also went with the same chip as before, since it's still the same thing. The dsPIC version will have the CAN, and like 50 other options, like a little pcb mounted relay and precharge resistor black thing that drops in resistance when it heats. So, just a turn of the key will precharge and then close the contactor.

I have a few other pins on the atmega. jack, what did you want with those? Just a hole?

It's just got 2 holes for running wires from the output of the optocoupler to the driver section. If you are using an IGBT driver board, I think you just need to use that above DC-DC instead of the other cincon. I need to doublecheck that though. it's 2 am and I can't think straight. haha.

DJBecker! You are doing some SR experiments?! ya! I still really want to do that too.

[Nevyn]

Soldering TQFP isn't so bad, once you know how to cheat! I'm not saying it's easy, it's just not demoniacally hard.

SMD Soldering Guide by Infidigm

Instead of using a sucker, use some "Solder-Wick" or copper braid for desoldering. It leaves behind just enough to make contact, but slurps up everything extra!

Takes a little bit to get use to this method, but you don't go blind doing it!

[twotires]

Collin Cunningham in one of his video tutorials on soldering SMD (search for his name and/or Make Magazine and/or Jameco Electronics on Youtube) suggests pinching off a teeny-tiny little dot of plasti-tack and using that to hold the piece in place before beginning soldering. I've never tried soldering SMD, but I thought that was a neat trick.