Paul & Sabrina's cheap DIY 144v motor controller

[CamLight]

Originally Posted by chapper

From CamLight (#3035)
Ahh...so the FETs would be mounted on a bare (copper?) bus bar that would act as a heat spreader and also conduct the current out? And then the bus bar would be mounted to an electrically insulated heat sink to get rid of the heat?

Precisely! Sketch is on the way... probably tomorrow.

Looking forward to it.
Love the idea...combines the sink and bus bar very nicely. My only concern would be the very high thermal resistance between the FETs and the bus bar (without compound). But, with compound you wouldn't be able to conduct all that current. All depends on the amount of heat you need to remove though.

What case type were you considering for the FETs? If it was TO-220 you can have a bus bar on top of the tab and still mount the rear of the FET to a heat sink.

You'd probably need a LOT of TO-220's though.

[MPaulHolmes]

Testing started!

Results so far:
With 0.5" separating bus bar and current sensor, 166 amps actual through the bus bar caused the voltage to change from 2.469v to 2.900v on the current sensor. For this current sensor, if the output voltage is 4.6v, then that indicates the "maximum current". So in my case, on the first attempt, my current sensor has a range of -808 to 808 amps.

I think I'll give it a bit more space, and see what happens. Not bad for a first space guess though!

[squiggles]

Originally Posted by MPaulHolmes

OK, now for the heat...

At 1000 amp, 100% duty, it would generate about 75 watts per mosfet (I guess a little more, due to switching losses). However, 100% duty and 1000amp would be very uncommon, since the 1000amp would usually only happen at very low rpm, which would mean very low duty.

My understanding was that the switching frequency and the switching time of the device affected the heat dissipation more than the duty. Theory being that it is the time spent in the linear region that is the big issue.

[MPaulHolmes]

I'm sort of worried about that. The gate charge is more (maybe 50% more) than the irfq mosfets that I"m using now too. I may have to lower the gate resistor. My driver still has some head room, though to put out more current.

[jackbauer]

A few people have asked me what , if any , mods I made to the control board to drive igbts. The answer is none! I did however use the circuit below in place of the normal gate resistor. The purpose is to help the igbt turn off quicker and overcome the "miller knee". If i were doing it again I would change the dc dc converter to one with a 15v output as this is how igbts like to be driven. That said the 12v version seems to work fine on mine. Very low switching losses. Just friday I pumped 250 battery amps through the controller on a hill climb while on the way to see Ireland's answer to Chip Foose The heatsink went from 6c to 8c!

[williamson]

Has anyone researched a CPU cooler? It uses heat-pipes to transfer the heat. The pipes contain a liquid, most likely alcohol, that evaporates and absorbs a lot of heat. The fan then condenses it back. I would guess this must be operated with a designated side down, so the liquid flows. They are sold cheap on ebay. Such as # 160407436678. Of course, it's small. BTW, in case anyone wants to know trivia, the miller capacitance is the multiplication of the real gate to drain capacitance, by the drain voltage rise. It slows the turn-off.

[chapper]

Here are two variations on the same idea.

The upper drawing is a section through a controller much like Paul’s “Revolt” except that it now contains a structural/heat transfer member and an air channel above and around the MosFETs.

The lower drawing (The Off-Road Controller) shows the same configuration, but it’s held together with anodized aluminum clamps for true electrical isolation and equalized surface clamping forces. It avoids screw threads abrading their tapped holes and potentially contacting different voltages. (In the upper drawing, note the 3 screws holding the buss bars to the heatsink. What happens if…?)

Silver plate the buss bars and PCB with “Cool-Amp” Conducto-Lube Co.
Between the MosFETs, diodes and buss bars use an electrically and thermally conductive compound like “Conducto-Lube”. Conducto-Lube Co.
Between the circuit board copper and the buss bars, use an electrically conductive compound like “Conducto-Lube”
Between the buss bars, and the aluminum heatsink, use a thermally conductive compound like: Aavid “Ultra Stick”
All anodizing is Type III

[jyanof]

Originally Posted by CamLight

How many watts of heat do you estimate each FET will have to dissipate?

LOL, sorry about all the design questions. I'm just fascinated with high-power thermal-transfer design and it always leads to lots of questions.

I dig the thermal stuff too, and did a bunch of calculations and testing previously on the 500A Revolt prototype in my car.

See my posts on these pages:
http://ecomodder.com/forum/showthrea...tml#post153760
http://ecomodder.com/forum/showthrea...ng-9325-2.html

http://ecomodder.com/forum/showthrea...-6404-214.html

Some basic heat generation formulas for mosfets were used that approximates both conduction and switching losses (just google 'mosfet heat dissipation' or something).

At cruising speeds, ~200 motor amps, the total calculated heat generation was roughly 300W. Roughly 2/3 of that were from the 10 mosfets, so that's only 20W per mosfet. When I went through the calculations, I think I was predicting a junction temp somewhere around 80C at these steady state conditions in the hot AZ weather. The heat generation is roughly proportional to motor amps (duty cycle has an effect too), so conceivably 400A would have twice the heat generation. Accelerations would also create excursions into higher temperatures, though they often occur after coasting or sitting still for several tens of seconds which allow enough time for the controller to cool down a little bit.

Paul's SR version will use the same number of mosfets, but I believe he is lowering the gate resistance to decrease the turn on/off times. This should reduce switching losses significantly and reduce the total heat generation. Additionally, any output above 500A *should* only last for several seconds. A dynamic analysis would be more appropriate to estimate how much time at full power it would take to hit the thermal cutback threshold. Getting rid of the thermal pad would surely benefit this dynamic thermal performance. Even still, I read a post that the Zilla 1k has a hot-day rating of 1000A for 10 seconds - the SR version may have something similar.

All this being said, I agree that it's all about cooling. The soliton1 claims 1000A continuous with proper water cooling. My setup is a basic finned heatsink with 3 wimpy computer fans. I think there is a lot of room for improvement if you were to take the time and effort. Someone mentioned heat pipes - I've been thinking about that as well as it should be fairly easy and effective, but what I have now seems to work!

[frodus]

Can the serial output be configured to "spit out" the controller measurements for current, voltage....etc?

Interested in being compatible with a serial based iphone interface? We want to support other controllers and products with our software.

Options right now are serial and ethernet, but canbus is forthcoming.

[adamj12b]

Originally Posted by frodus

Can the serial output be configured to "spit out" the controller measurements for current, voltage....etc?

Interested in being compatible with a serial based iphone interface? We want to support other controllers and products with our software.

Options right now are serial and ethernet, but canbus is forthcoming.

When the controller is started up, you send the command rtd-period #, where # is an interval setting in milliseconds.

The output string is a fixed format and it looks like :

TR=000 CR=000 CF=000 PW=000 HS=000 RT=0000 FB=00 BA=000 AH=000.0

Where you have:

TR = Controllers calculated throttle position. 0-511
CR = Controllers current reference based on throttle and other factors. 0-511
CF = Current feedback from LEM current sensor 0-511
PW = PWM duty cycle of controller 0-511
HS = Heat-sink Temp in raw ADC counts 0-1023
RT = Raw Throttle counts from the ADC 0-1023
FB = Fault Bits. Hex output with each fault and status code. Normal 00
BA = Calculated Battery amps 0-511
AH = Calculated Amp Hours consumed 0-999.9

The default output is through the serial port. 19200,8,N,1.

If you need some more info, feel free to email me directly.

-Adam