Open Source Controller - 144v 500 amps

I'd like to get peoples opinion of this open source controller

http://www.instructables.com/id/Homemade-100-HP-Motor-Controller-for-an-Electric-C/

I'd also be interested in links to other open source motor controller projects that people like

Linear paralleled FET arrays do not share current well.

For a brushed controller, you can simply use single IGBT solutions, a decent low-side FET driver, and a $15 brushed ebike controller for a brain, but with the shunt input resistance dropped 50-100x (or whatever)

Brushed controllers in most any power level can be made by someone with no skills, no design, no understanding, and work like a champ.

High power brushless controllers on the other hand... Wow! It's like stepping up the difficulty 10-levels. You need to have some pretty advanced skills to end up with something useful.

Looks like this is the main page for this project.
http://ecomodder.com/wiki/index.php/Open_ReVolt

Good news: There seems to be an Open Source code base for using the Atmega in a variety of controller types. Wiki advertises both DC and AC versions.
View attachment Controller.jpg

Didn't look through it carefully. Wiki was either incomplete or hard to navigate.

B.t.w. ES readers might recall that a former member in somewhere in Ontario was building a BLDC controller, with I think also an Atmega processor. Unfortunately the thread went from technical to abrasive and details on how to build it and program was not posted. I couldn't find the link to his thread though. If I reacll right he had pics that looked somewhat similar to the control board layout of this project. (Thought he didn't have a separate power board, but the FETs on the same board)

I think the fellow in Canada may have been using a PIC micro. Or perhaps it was a different one than I'm remembering.

Plenty of app notes for doing this with PIC, AVR or ARM micros. Of course 500 amp boards are a bit more difficult to make work.

Unless I'm mistaken, that's just a brushed motor controller.

Jeremy

liveforphysics said:

Linear paralleled FET arrays do not share current well.

For a brushed controller, you can simply use single IGBT solutions, a decent low-side FET driver, and a $15 brushed ebike controller for a brain, but with the shunt input resistance dropped 50-100x (or whatever)

Brushed controllers in most any power level can be made by someone with no skills, no design, no understanding, and work like a champ.

High power brushless controllers on the other hand... Wow! It's like stepping up the difficulty 10-levels. You need to have some pretty advanced skills to end up with something useful.

Click to expand...

MOSFETs do parallel well as they have positive resistance slope so if one heats up too much the load is taken up by the others. I agree that for high amps IGBTs are a nice way to go. The main problem is switching transients due to the inductive part of the system. The down side of this is that when IGBTs or MOSFETs go they form a dead short so you go like a bat out of hell. With a brushless motor a dead shorted power device will usually mean a dead stop! Take your choice , I'd prefer the dead stop! Whatever system you use the ideal is a pulse by pulse current controlled system. The controlling system by using the current into the motor in the feedback loop will behave much like the accel pedal of a gas car..

themotorman said:

MOSFETs do parallel well as they have positive resistance slope so if one heats up too much the load is taken up by the others. I agree that for high amps IGBTs are a nice way to go. The main problem is switching transients due to the inductive part of the system. The down side of this is that when IGBTs or MOSFETs go they form a dead short so you go like a bat out of hell. With a brushless motor a dead shorted power device will usually mean a dead stop! Take your choice , I'd prefer the dead stop! Whatever system you use the ideal is a pulse by pulse current controlled system. The controlling system by using the current into the motor in the feedback loop will behave much like the accel pedal of a gas car..

Click to expand...

Paralleling large numbers of FETs and then trying to switch them fast, for a high commutation rate controller, for example, is a bit fraught. You're right about the DC conditions, but this isn't the issue. High voltage (c.100V) FETs have a high gate charge, the really, really big problem is driving enough current evenly into the gates of a parallel array at speeds commensurate with maintaining low switching losses. Most multi FET controllers are pretty poor at doing this well, particularly at the budget end of the market.

The very best illustration of this is the user experience from the multi-FET controllers commonly used on big ebikes. As the number of paralleled FETs increases, the current capability of the controller doesn't increase in proportion. This is hardly surprising when you think that a typical 100V 36 FET BLDC controller will have gate drivers that are trying to shift around 1200nC of charge around evenly between 6 gates at a time - a tough challenge and one that needs RF-type circuit board layout around the gates, with equal length transmission lines matched to carefully sized gate resistors to minimise gate ringing. The Chinese multi-FET controllers don't attempt to drive the FETs fast and cleanly, they just slug the switching times right down to get rid of the problem and accept the high FET switching loss, which is the reason for the law of diminishing returns when it comes to increasing the number of FETs versus the power the controller will handle.

Using single big FETs, or IGBTs if the voltage is over around 120 to 150V, where IGBTs start to look better than FETs, is a much simpler proposition, as it removes at a stroke the difficult multiple gate drive challenge. It is far, far simpler to supply a single gate with a nice, clean and fast gate drive pulse, and the whole controller becomes a very much simpler thing to lay out.

I'm of the view that controller designers have thought too much about the DC issues, like Rdson losses and associated junction to ambient thermal conductivity problems, and have only paid lip service to the need for clean gate drive and proper RF type layout of the drive signal path. One look at the plethora of messy controller designs around, often with fairly grim gate drive layouts, tends to confirm this.

Jeremy

Jeremy Harris said:

I'm of the view that controller designers have thought too much about the DC issues, like Rdson losses and associated junction to ambient thermal conductivity problems, and have only paid lip service to the need for clean gate drive and proper RF type layout of the drive signal path. One look at the plethora of messy controller designs around, often with fairly grim gate drive layouts, tends to confirm this.

Jeremy

Click to expand...

Hear hear! The difference been thinking and doing in a nutshell. Thinking often leads for a focus on the irrelevant problems. Doing will quickly point out the real ones. One hour at the lab bench is worth ten hours hypothesizing. We see it on this forum, as well as in the scientific literature, where most published papers lead to no more than a higher paper count.

(I would have hypothesized just like motorman that paralleling would be relatively simple. In fact I've built linear amps with up to 24 parallel TO3 FETs to make gigantic audio amps. Works like a charm when driving with a continuous signal, but as Jeremy points out, it's a different ball game with a pulsed signal)

Jeremy Harris said:

I'm of the view that controller designers have thought too much about the DC issues, like Rdson losses and associated junction to ambient thermal conductivity problems, and have only paid lip service to the need for clean gate drive and proper RF type layout of the drive signal path. One look at the plethora of messy controller designs around, often with fairly grim gate drive layouts, tends to confirm this.

Jeremy

Click to expand...

Exactly my friend. The RdsOn is where controller designers like this focus, and yet completely fail at the inductive path layout of the devices (linear...)

It reminds me of the cell tester arrays that I made. 24x 300amp FETs in a row, it blew up up in a few seconds at 600amps. 8x 300amps FETs in a row with extensive emphasis on controlling gate ringing, it would run for a few minutes at >600-800amps, yet still would fail from the inductive path being sh*t (linear) and unbalancing the current loading through the FETs at high frequency. 4x 300amp FETs in an RF style circular array, and BINGO! The thing works continuously (at least for minutes at a time), the FETs share current effectively at all freqs, it's never had a problem since, even at much higher current levels.

Layout and gate path/control is critical. This controller failed pretty hard where it counts, and went the throw-parts-at-it approach to try to compensate.

Still though, I love to see anyone building their own controllers, and REALLY like to see folks building open-source controllers, much respect to them!

Somewhere on the web I saw a DIY high current brush motor controller where the guy did as you have and fitted the FETs in a circular array. IIRC it worked pretty well.

Jeremy

Jeremy Harris said:

Somewhere on the web I saw a DIY high current brush motor controller where the guy did as you have and fitted the FETs in a circular array. IIRC it worked pretty well.

Jeremy

Click to expand...

Hmmmmmm? I am too old, too out of date, too stupid and too lazy, but hi-power radar circuits come to mind. The thinking seems to be dragged off track every once in a while because of someone's audio experience.
I have only been looking and listening carefully for one year, but it is very obvious to me that the MM Chinese solution is just cheap, dirty and stupid. As LFP is experiencing, a few minutes of smoke beats hours of math.
Academic circle jerks keep the thinkers employed but seldom produce a usable device.

Non-relevant topic of a lady thinker and doer:
http://blog.ted.com/2011/01/13/the-politics-of-breast-cancer-qa-with-deborah-rhodes/

themotorman said:

liveforphysics said:

Linear paralleled FET arrays do not share current well.

For a brushed controller, you can simply use single IGBT solutions, a decent low-side FET driver, and a $15 brushed ebike controller for a brain, but with the shunt input resistance dropped 50-100x (or whatever)

Brushed controllers in most any power level can be made by someone with no skills, no design, no understanding, and work like a champ.

High power brushless controllers on the other hand... Wow! It's like stepping up the difficulty 10-levels. You need to have some pretty advanced skills to end up with something useful.

Click to expand...

MOSFETs do parallel well as they have positive resistance slope so if one heats up too much the load is taken up by the others. I agree that for high amps IGBTs are a nice way to go. The main problem is switching transients due to the inductive part of the system. The down side of this is that when IGBTs or MOSFETs go they form a dead short so you go like a bat out of hell. With a brushless motor a dead shorted power device will usually mean a dead stop! Take your choice , I'd prefer the dead stop! Whatever system you use the ideal is a pulse by pulse current controlled system. The controlling system by using the current into the motor in the feedback loop will behave much like the accel pedal of a gas car..

Click to expand...

I blew hundreds of MOSFETs (IRF840 assembly line leftovers) trying to get to the bottom of why paralleling them caused failure once the power level of the circuit crossed a fairly particular level. I finally thought, what the hay, I'll remove the separate gate resistors and parallel the MOSFETs pin for pin even though that went against the available information prescribing separate gate resistors.

I was quite surprised to find that it completely solved the problem. I just use a single gate resistor for a set of paralleled MOSFETs now. I haven't experienced a failure yet after many hours of using them that way on my personal SMPS projects. I recommend keeping the paralleled MOSFETs as physically close as possible, however.

My present charger/push-pull booster/controller uses a single gate resistor per pair of 75 ampere MOSFETs for each leg in the push-pull portion, but I have also used other MOSFETs paralleled without separate gate resistors in totem pole (half bridge) configurations without any problem, and at power supply voltages of at least 160VDC.

On the topic of current mode control in general, indeed, current feedback is great for the smoothness and feel it provides. That is what the booster/controller portion of my present gadget uses.

ferrites are your friend... 8)

grindz145 said:

ferrites are your friend... 8)

Click to expand...

It'd be a real bear carrying around 50 or 60 hertz transformers in an onboard charger.

Solcar said:

I blew hundreds of MOSFETs (IRF840 assembly line leftovers) trying to get to the bottom of why paralleling them caused failure once the power level of the circuit crossed a fairly particular level. I finally thought, what the hay, I'll remove the separate gate resistors and parallel the MOSFETs pin for pin even though that went against the available information prescribing separate gate resistors.

I was quite surprised to find that it completely solved the problem. I just use a single gate resistor for a set of paralleled MOSFETs now. I haven't experienced a failure yet after many hours of using them that way on my personal SMPS projects. I recommend keeping the paralleled MOSFETs as physically close as possible, however.

Click to expand...

Is it possibly so that with separate gate resistors variance in the gate capacitance (perhaps extra big if the FETs you used were production rejects), cause the FET with the lowest capacitance to turn on well before the others, thus having it carry the whole current spike?

With paralleled gate resistors and a physically close layout all gates would track each other.

That ancient FET has more than enough internal resistance in it's gate. The reason for individual gate resistors is because otherwise the other FETs are kept from switching until the gate charge finishes saturating on the FET with the lowest gate voltage plateau. In this case though, the FET itself had more than enough resistance to correct for whatever gate plateau differences there may have been.

Is it possibly so that with separate gate resistors variance in the gate capacitance (perhaps extra big if the FETs you used were production rejects), cause the FET with the lowest capacitance to turn on well before the others, thus having it carry the whole current spike?

With paralleled gate resistors and a physically close layout all gates would track each other.

Click to expand...

There seemed to be something other than that going on since the actual power the circuit was carrying at destruction was plenty low so that when the circuit was operated on just two single MOSFETs in the half bridge, the failure didn't happen. I also have no indication that the failure was due to overheating. All indication I had in using the MOSFETs in various projects before that was that they were fine in quality. They really looked like original IR MOSFETs.

liveforphysics said:

That ancient FET has more than enough internal resistance in it's gate. The reason for individual gate resistors is because otherwise the other FETs are kept from switching until the gate charge finishes saturating on the FET with the lowest gate voltage plateau. In this case though, the FET itself had more than enough resistance to correct for whatever gate plateau differences there may have been.

Click to expand...

I can't think that any of the 20 or so different varieties of MOSFETs I have in stock are late technology, that is, that are not ancient. I have paralleled SO-8 types like IRF7492 (I don't know what the actual part was.) by simply stacking them on top of each other. It always works for me when I parallel any of those 20 or so types (each paralleled set always the same individual type), but maybe it's the age of their design structure that lets it work. Anyone who tries it with more modern types without separate gate resistors, I'd be interested to know how it does. It'd be great to have a comparison between success with separate gate resistors vs. a single gate resistor.