Magic Pie 2 with internal 12 FET controller

[Phil O Dendron]

Hi,

got inspired by Bikemad's Dual controller thread, so credits go to him 

Had six mosfets laying around that I bought when the pie's battery wiring went up in smoke.
In theory in this application the fets could easily be doubled.

So I cut a heatsink for them. This is the most annoying part when you do not have the right tools for the job.


Wired the fets into the pie, using the shunt that is already on the board.


There's a little mess inside the pie now. Lots of zipties. 


Backside (fixed the screws with hot glue before reassembling, not shown in the pic)


Bridged the shunt about two third to get a wider range when tuning with the cruise controller software.
Turned the current limits down to 20amp continous and 25 amp peak.

First "run" with rear wheel off the ground, no load, everything was fine.
First real start was WHOOOOOOA!!!, but blows the 35amp car fuse immediately.

Had to build a "twin fuse pack" because I could not get higher rated fuses in that style.


Turning down the "acceleration" setting to 95% may have been helpful, too.

Now the pie pushes a lot harder than before.
Hope to get the front suspension forks this week, 'cause the setup of the 'goose is very "front happy" now 


Took it for a test ride. Big big smile 
The good thing is the pie seems to keep cooler than before.
Maybe the fets have a better operation point now.

Maybe bad: the first time my batteries got warm (not hot).

I'm not sure if this setup will stand the double power as the wire length and the cooling of the fets is not equal,
so maybe they will not share the load fifty-fifty.

Will keep you up to date as tuning is in progress.


cheers, phil.

[Morgen 3Eman]

Hi Phil,

Your FET paralleling is a much better idea than Alan's controller paralleling, as the FETs for a given phase are being turned on and off by the same signal source.  After reading Alan's adventure story, I'm surprised that there wasn't more damage done to the controllers, because of small timing variations between controllers.   It wouldn't take too many nanoseconds of having the upper FET turned on in one controller and the lower FET not  turned off on the other to do some serious over current stress. 

Having said that, I have to tell you that Alan is one seriously brave dude!  You have to like a guy who goes out and finds the answers to his questions....

TTFN,
Dennis

[Phil O Dendron]

Hi Dennis,

Hi Phil,

Your FET paralleling is a much better idea than Alan's controller paralleling...

Having said that, I have to tell you that Alan is one seriously brave dude!  You have to like a guy who goes out and finds the answers to his questions....

Thanks 
It was Alan's idea, too. But I think all the external controllers with 12, 18 or 24 fets use the same principle.

I'd even considered putting another six (or 12) FETs on a separate heatsink plate, linking the power and the three phase wires directly, but with each of the gate pins being connected to the gate pins on the existing FETs. I think a shunt would still be needed on the ground connection to prevent too much current from going through the new FETs.
But I think a second controller would be the easiest option.


Alan

Turned up the settings for the current limits a bit more, the pie has now a peak power of 3800 watts.
The pie is still quite cool, but my battery gets hotter and hotter.

cheers, phil.

[e-lmer]

You may want to look into a circuit breaker.
I got one from Cooper.
http://www.cooperindustries.com/content/public/en/bussmann/transportation/products/circuit_protection/circuit_breakers/series_187_marineratedcircuitbreaker.html
But this kind seems much nicer than mine.

It does look a little expensive:
http://www.ebay.com/itm/Bussman-DC-Circuit-Breaker-MRCB-50-amp-187050F-Manual-/280640402464

[Morgen 3Eman]

Hi Phil,

If all the FETs are driven by the same signal source, paralleling the output FETs  can allow the passage of more current, but if the control signals are based on the original shunt signal, there will not be an an actual increase in current to the windings.  The basis of the error signal to drive the FETs is the voltage developed across the shunt, so you have to decrease the shunt resistance to increase the control signal pulse duration that determines the winding current.  At some point, the conductors in the current path will need to be increased in cross section so they don't just melt under the load, or more likely set the circuit board on fire.   (They smell terrible.)

By the way, replacing the FETs with higher rated FETs can achieve the same goal as paralleling the FETs.

TTFN,
Dennis

[Morgen 3Eman]

One more thing, If you decide to parallel some transistors, you should be aware of an affect called "current hogging".  It is mostly a problem in bipolars (PNP/NPN),  where one transistor in a set a paralleled devices starts carrying more current than the others, which causes the junction temp to go up, so it carries even more current.   The other devices carry less current, and thus run cooler, and you can see where this goes.  I have even found the problem in multi-emitter transistors.


I haven't noticed the problem with power FETs, but I would be very leary of using different device part number FETs in a parallel system. 

TTFN,
Dennis

[Morgen 3Eman]

Well that was clumsy:)

[Phil O Dendron]

You may want to look into a circuit breaker.
I got one from Cooper.
http://www.cooperindustries.com/content/public/en/bussmann/transportation/products/circuit_protection/circuit_breakers/series_187_marineratedcircuitbreaker.html
But this kind seems much nicer than mine.

It does look a little expensive:
http://www.ebay.com/itm/Bussman-DC-Circuit-Breaker-MRCB-50-amp-187050F-Manual-/280640402464

thanks e-lmer, will try to get some of these.

Hi Phil,

If all the FETs are driven by the same signal source, paralleling the output FETs  can allow the passage of more current, but if the control signals are based on the original shunt signal, there will not be an an actual increase in current to the windings.  The basis of the error signal to drive the FETs is the voltage developed across the shunt, so you have to decrease the shunt resistance to increase the control signal pulse duration that determines the winding current.  At some point, the conductors in the current path will need to be increased in cross section so they don't just melt under the load, or more likely set the circuit board on fire.   (They smell terrible.)

By the way, replacing the FETs with higher rated FETs can achieve the same goal as paralleling the FETs.

TTFN,
Dennis

Dennis, thanks for the hints.
The shunt is modified, should have about 50% of resistance of the original.
The pcb now has solid copper wire where high currents flow.

One more thing, If you decide to parallel some transistors, you should be aware of an affect called "current hogging".  It is mostly a problem in bipolars (PNP/NPN),  where one transistor in a set a paralleled devices starts carrying more current than the others, which causes the junction temp to go up, so it carries even more current.   The other devices carry less current, and thus run cooler, and you can see where this goes.  I have even found the problem in multi-emitter transistors.


I haven't noticed the problem with power FETs, but I would be very leary of using different device part number FETs in a parallel system. 

TTFN,
Dennis

I was not aware of "current hogging", never heard of that before.

I thought in an application that uses the fets as switching devices the circuit should be designed to deliver
a more-than-enough signal to the gates to make the fets switch fast and "safe".
When the voltage applied to the gate is so big, small differences within the two parallel fets (of the same type, of course)
should not matter.

I'm not completely sure if I understand your last line, maybe I have to study the theory of Mr. Leary's circuits 

cheers, phil.

[Kris]

unlike junction transistors, power mosfets have a positive temperature coefficent. So fets in parallel should automatically current share.

[Phil O Dendron]

unlike junction transistors, power mosfets have a positive temperature coefficent. So fets in parallel should automatically current share.

thanks, kris!
so the cooler mosfet of the pair should take more current because of its lower resistance (if the wiring was equal)?
that`s REALLY good news for me