Paul and Sabrina's Cheap 3 Phase Inverter (AC Controller) with Field Oriented Control

[e*clipse]

Thanks for the Speedhut link! I wish they had said a little more about the gauges n' stuff.

Speedhut has some really neat products - GPS speedometers, electronic speedometers, etc. The guages are stepper motor gauges, and should work with any input. They also have custom gauge faces for certain cars, including my Eclipse! I'm sure for a little bit extra, I get a gauge cluster that would look like it was always an EV...

Hmmm... RPM >> AMPS?? How about a voltmeter that reads 650V?? This looks like fun!

- E*clipse

[NeilBlanchard]

I think that pure (distilled) water is NON-conductive.

[e*clipse]

I started writing this, and it started sounding like some lecture. I really don't mean it as such, and I hope folks can wade through my writing it for the information...

There are things that work in a lab, and then there's the real world. For some reason, stupid little details can cause something to not work, when it theory it should. For example, pure water. How long does water stay pure, before corrosion from some metal connection contaminates the theory?

Probably the main problem with water is that it also freezes really easily. Are you willing to risk your motor controller? My dad risked his solar water heater to his preventive maintenance in Idaho; I doubt things would work out much better for a pure water cooled motor controller in Boston.

I just got back from a trip to Idaho, with -10F temperatures. I drove my Biodiesel powered truck and one of the big lessons I learned is that some thermal overkill - low and high side - is a good thing. Trying to start a diesel that's not plugged into a block heater and has gelled biodiesel (using 50% dino-diesel) in the fuel system is not fun. At those temperatures 15W-40 engine oil looks like honey.

I think if anyone is going to use a water-based coolant for electronics, it would be very wise to make sure that water leaks absolutely cannot get to the circuits. I have a dead PC that is pretty good testimony to what tiny little condensation-sized dribbles can do. If you think that this can only kill low pwer electronics, my dad's washing machine is testimony against that. Fortunately the circuit board was pretty old school, and we were able to trace the blown relay that controlled the motor's field windings and thus, the spin cycle. A $3.50 relay for a $700 washing machine. Simply sealing the wires, like a standard automotive harness would have prevented the problem.

I do plan to use water based cooling in my motor controller. For this reason, and because of the amount of water a real world vehicle is subjected to, I plan to seal the motor controller completely, like an OE device. Even though pure water is probably the best coolant from a conductivity/thermal mass perspective and I live in California, I will still sacrifice the ideal performance to a level of thermal safety.

-E*clipse

[e*clipse]

I've been thinking more about the gauges thing, and I think there is a solution that is almost 95% worked out already!

I have one of those little OBDII code reader/gauges that allow you to pretty much display anything you want. It works great for my little 1st generation Insight, which is not a very common car. It also works fine for my dad's Toyota 4-runner.

This gauge is great for tuning, but the display is difficult to glance at and see a measurement. For this reason, I wouldn't use it for something like a speedometer or current sensor, even though it can display these values.

However, SpeedHut has a wide range of guages that simply plug into the CAN bus, like the OBDII gauge. In other words, it's easy to provide a real, analog gauge for anything that's streamed on the CAN bus.

That's where we're 95% there - Paul has made a CAN bus stream of data for the motor controller. If that data were OBDII compatible, there would be a world of gauges/displays available:
For example, the tachometer requires:

• ECU/ECM that supports CAN-bus (SAE - J/1979), For more information about CAN-BUS click here.
• Easy Installation, connects directly into the vehicle's 16-pin OBDII Data Link Connector.
• 12 Volt DC Power

Here's general OBD-II info; EV-specific stuff could easily be added here:
http://en.wikipedia.org/wiki/OBD-II_PIDs
http://www.speedhut.com/about-freedomgauges.html#canbus

I don't exactly know the format, but it seems a couple format changes and all of this could be plug 'n play! If a gauge doesn't exist already, it seems this format and a custom faceplate is all that would be needed for EV-specific gauges.

- E*clipse

[P-hack]

any analog gauge/meter can be pwm'ed very easily, if you update the pwm cycle reasonably fast (i.e. 30 times a second) it will still look analog for the most part.

http://web.archive.org/web/200701270...05abstract.pdf

you just need to figure out pwm values for full and half scale on the bench and sort it out from there.

[cts_casemod]

If one drives a early 90's car, its perfectly acceptable to drive analog gauges with PWM. For the newer ones , tought, control is made digitally.
The cluster I use sends a pulse every so often (about 20Hz) and measures some drop in voltage from the float level and temperature sensor. It expects the signal to be zero otherwise. Driving it with PWM won't work as the cluster CPU is expecting a signal within a certain range/voltage. Some have an internal current source.

This is something to have in consideration while designing a universal controller. In my opinion its easier to have a digital controller and just make bar graphs. Speedometer and tachometer are generally digital and can be interfaced easily. I have a digital readout on display to accommodate older cars that may have a mechanical tachometer. This is used to calculate SoC and remaining mileage. Total mileage and SoC are saved to an FRAM by an interrupt at shut down. This takes a fraction of the time needed to write a standard EEPROM.

In general I find it best to keep the cluster and controller separate. The controller can use CAN or whatever to send a limited amount of data, but the display has its own CPU.

As for my controller, I'm very pleased that a version of the ILI9341 controller is now available for a 2.8" display. I recomend using this as a starting point. With a co troller supporting DMA one can get a very decent speed in a much more compact form than using parallel access. I'll be probably moving the whole thing to a cortex CPU soon.

[MPaulHolmes]

I think I found a nice option for an inductor on the input for people who want to boost their DC voltage up. There are a bunch of 20HP cheap AC motors on Ebay. Like $200 range. That is continuous duty, so you could do more for hard accelerations. I'm not sure what the 30 second rating is on a 20HP motor though.

So, I'm hoping that someone who puts 144vDC in their car can just add one of these inductors on the front end right after the battery pack, and then use a 230VAC cheap Ebay motor. Winding it sounds pretty easy too. It's a couple big 165mm (outer diameter) toroids made of "High flux" 26u or 60u (permeability). Then, you just use around 25 wraps of very high strand super flexible welding cable to mitigate the skin effect. 15 feet worth would be enough if you could wind it perfectly. Otherwise, you may have to spring for a few extra feet. It sounds perfect for the DIY person.

[P-hack]

I'm hoping people will actually use the AC motors as motors Have thought about a cheap DC motor as inductor, but laminated steel isn't the best material, but worth some experimenting maybe. A reluctance motor might make a better air-gap bridge between coils than a squirrel cage.

That toroid at 60μ looks to be about 3800 NI @ %30 rolloff, so targeting 152 amps at 83uH effective? Interesting looking at the 125μ high flux, it really holds on to its AL value, for the same turns it looks like 128 amps at %30 rolloff, but the effective inductance at 128 amps bias is 172uH.

I am finding the 125μ in sendust to be more of a liability where high dc current is desired.

Does skin effect vs diameter count if all the wires are touching? do they need to be individually varnished to have a "skin"? And does skin effect only apply to the ripple current?

"Hence current will tend to preferentially flow near the skin of the bundle of wires, just as it does with a single solid conductor of similar overall diameter."

Edit, if I'm doing this right, the high flux at 125μ would deliver 184 amps at the same effective inductance @ ~18 turns.

[MPaulHolmes]

I had written to XFMRS, and they came back with a design suggestion, along with a .xmls file with all their design steps. I asked about the skin effect when I saw how large of diameter cable it was that they were suggesting, and they said the stranding "helps to mitigate" the effect. Wait! But there isn't (much) AC current in the strands! All the AC (ok not all!) current is between the capacitor and the IGBTs. Doesn't that mean the skin effect isn't very applicable?

[cts_casemod]

Originally Posted by MPaulHolmes

I think I found a nice option for an inductor on the input for people who want to boost their DC voltage up. There are a bunch of 20HP cheap AC motors on Ebay. Like $200 range. That is continuous duty, so you could do more for hard accelerations. I'm not sure what the 30 second rating is on a 20HP motor though.

160% torque for 15 seconds and 200% torque for 5 seconds for S1 Rated motors at nominal nameplate speed and V/Hz.

For 30 seconds I would say 130% torque should be safe.

Above 90Hz one has to derate by a certain factor, cant remember how much now.

This might also help:
http://www.sirus.in/images/crompton-ac-motors.pdf