Curtis 1231C Repairs
Ian Hooper, 21 July 2008

Introduction

Having recently completed my MX5 conversion, the first problems I have been encountering are with the motor controllers. At first I had a Kelly KDH14500 installed, and while I had heard they were underpowered, let's just say that turned out to be an understatement - there's no way it would have passed licensing. So a friend kindly lent me a Curtis 1231C they had spare after upgrading to a CafeElectric Zilla controller, which he said I could use to get the vehicle licensed, until I got around to building something better.

In general I would never recommend a 500A controller such as the Curtis 1231C for any direct drive applications (such as my MX5), but it did have enough power for the vehicle to keep up with traffic at least. That is, until it went pop. The first signs of its displeasure were occasional stuttering under acceleration, though I'd heard the Curtis were quirky so didn't give it too much thought. But then, about 5 minutes from home, as I was pulling up to a stop it gave a brief surge then went dead. I first checked over the power wiring thoroughly (initially suspecting that the fault might be mine) but everything looked fine. The Curtis was warm but not excessively - maybe 80 degrees. So I called a friend who brought their dino burner down and gave me a tow the rest of the way.

So a fairly inauspicious first day out for the MX5, but these things are a useful learning experience if nothing else. I've since opened up the Curtis to see what the problem was, here are some pictures..

Curtis Disassembly

The cases are potted, which makes them a real pain to get in to. Potting compound softens when heated so I put the Curtis on the stovetop for a while, until the case was about 100 degrees and the compound was soft enough to peel/tear out.

A view into the case after the potting was removed. You can see the logic board on the left, the capacitors down the middle and the power semiconductors (FETs and diodes) run in two rows on the outer faces of the centre aluminium extrusion.

Even the 8 screws which hold the centre extrusion in place are sealed in potting compound! It must have taken me half an hour to clean the potting compound out of these recesses to get at all the screws. Kind of annoying when manufacturers make products so difficult to service!

With all 8 screws removed, the guts of the controller slide out of the case. You can already see the blackening on the left where something has gone badly wrong...

So here's where the bad stuff happened; the MOSFETs closest to the logic board got fried. Incidentally I've heard this is the most common failure mode of the 1231C, blowing the FETs closest to the logic board. The problem (for anyone interested in the details) is that the 1231C has one FET driver for all 19 FETs, and it is located at the far corner of the case on the logic board. So the PWM signal propagates out from here, reaching the nearer FETs fractionally sooner, which then switch on fractionally earlier than the FETs at the other end of the case, and for a few microseconds they are carrying the whole load. If the motor has high inductance (such as the Warp 11 in the MX5), there is a huge surge of current through the nearest FETs, and it can be far beyond their rating. In fact Curtis have relatively high gate driver resistors (46 ohm) as a kludge to slow down the FET turn-on times and work around this problem with the physical layout. But the drawback with slower turn-on times, of course, is greater power loss during switching transitions and hence more heating. The 1500hz "whine" observed under low throttle with the C revision of the 1231 is another kludge added to help it cope with high inductance motors. The first obvious damage other than the charcoaled FETs is a signal track which has peeled off the board due to the intense heat. It's a little hard to see what else has been damaged at this stage due to all the blackening..

So What's Next?

Many readers will have heard me complain about the Curtis 1231C previously, regarding its primitive design and excessive price, so perhaps blowing one up just proves my point.. ;) But despite this, it is still an expensive bit of gear so it'll be worth me spending some time fixing it. I certainly won't be putting it back in the MX5 again, but it's still useful for people running smaller motors.

One upshot of killing the Curtis has been confirmation of the ZEVA controller's design. It's interesting to see how Curtis have designed the 1231C, but I'm now confident the ZEVA controller will be superior in many respects.

Firstly, use of more modern semiconductors have allowed my design to use fewer, higher power FETs and keep them all within 80mm of the PWM source (instead of about 300mm for the Curtis). I'm also using multiple FET drivers located close to the FETs, with symmetric track lengths from the PWM generator to allow for much faster switching - hence lowering switching losses dramatically.

The 1231C is using 19x IXTH50N20 MOSFETs with a 50A rating and 45mOhm on resistance. I'm using 10x IXFK120N20 MOSFETs, which have 120A rating and 16mOhm on resistance. So even with half as many FETs it'll still have 20% higher current rating and 35% lower on resistance! (i.e less power lost to heat)

I'm also amazed that the 1231C's logic board is entirely analogue, made up of a dozen op amps (comparators) and hundreds of resistors! I guess it serves its purpose, but a microprocessor allows for a far simpler, tidier logic board with far more functionality. Using a micro seems like a no-brainer to me, but I guess the 1231 was designed a long time ago now.

And lastly, this experience has given me a good benchmark for the ZEVA controller: If it lasts more than 2 hours in the MX5, I'll feel confident releasing it to market!

Comments

	j-brown on 2nd Feb 2015 can the IXFK120N20 MOSFETs work in a curtis controller mod-1205-101 voltage 24-36 amps 400 all of the mosfet is burn out and i need to get it fix
		Ian Hooper on 9th Feb 2015 Hi J, apologies for the delay (I was on vacation for a while). You might be able to use IXFK120N20 in a 1205-101 repair, but even if the package type is correct (I'm not familiar with the internals of that model), it's almost certainly not the optimum component! A controller rated at up to 36V is probably using 50-60V rated transistors, which often have a much lower on-resistance (and hence better efficiency) than a 200V component like that IXYS part. IRFP2907 would be significantly better, and there may be other options out there even better. That said, there can be unexpected issues with running different transistors (such as current limiting not working correctly anymore) so it's often safest just to rebuild with the exact same part that blew, or the closest equivalent you can find!