2QD Controller Works (Stopped By A Solder Bridge!)

Finally, I have finished the 2QD controller, and tested it ok. It only took me...two months (almost) since receiving the PCBs. ::rolls eyes::

I got it started the day I got them, but naturally I did not yet have quite all the parts collected (thought I did, but I'd missed some transistors and capacitors). Next few days saw me finding the rest of the parts and assembling it fully, but it did not work, which meant that one or more parts were bad (not surprising, since they were from recycled junk), or I had installed something wrong.

Then I got sidetracked, because of too many things taking up time, including (so far futile) job searches and whatnot since the one I have now keeps cutting my hours to the point I can't pay rent and utilities, much less eat. I went back to it as often as I could, but since part of what I had to do was figure out exactly how it was *supposed* to be working, and it is a more complex design than one would first expect for it's size and component count, due to safety features and feedback from one section into another, it ended up taking until tonight to fix it.

I intended to but never had time to build the 2QD in a SPICE program; I started it but the learning curve for the program is pretty steep! I thought it would help me understand why I was getting what appeared to be all the correct waveforms but no output, and no actual modulation of the pulse-width for the throttle input variation, beyond the first perhaps 1% or so.

I physically tested every component in and out of circuit, and got all the right readings. Even pulled all the transistors and tested them in a cheap old DMM that has a transistor tester, which displays the beta of the TUT. All tested ok, with all the 3904s having similar betas, and the 3906s as well. I'd also used some power transistors at one time for the MOSFET driver transistors, but decided to simply replace them with the small-signal 390Xs when I pulled them all out, since they'll still work, just not turn the MOSFETs on as fast. Later, I'll replace them with the power transistors when I put it on the bike. So no component appeared to be defective....

I have a friend here in Phoenix that had given me some very old but quite usable test equipment including the Tektronix 531A tube-based scope I used to troubleshoot it, and I asked him if he might be able to take a new-eyes look at the problem I was having. He said he would, and then I went back to working on it, since it was my project for today, and I thought I could at least narrow down some more things.

Of course, now that I'd asked for help, I found the problem later that night. ;-)

I had a hairline (and I mean VERY thin) whisker bridge from the gate of the bottom MOSFET to ground, right at the point where the trace from there ends at that end of it's gate resistor, to the ground plane of battery negative.

I have checked and double and triple checked for that kind of thing and never saw it. :(

I found it only because I decided to jumper from the top of the 9V zener to the MOSFET gate just to see if the MOSFET would even turn on. It didn't, so I disconnected it and turned the board over to desolder the obviously bad MOSFET :-) and the light glinted just right and I saw the bridge!

I didn't take a pic of the bridge, because I could not get a camera angle in which it would show up. But it was right about where you see the scope ground lead clipped to the edge of the board, between the battery negative lead (black) and the MOSFET visible on the heatsink.

The MOSFET was not bad; just that they don't work well if you ground the gate. ;)

At least I know the feeback loops for the self-protection circuitry all work! :) Since the MOSFET wasn't turning on, there was no voltage drop on motor negative, and so the feedback from there told the comparator to not increase the pulse width, since it was not going to do any good and might potentially fry something if it kept trying.

Now it tests ok and works with all the motors I have here to test with, including under stall loads on one of the powerchair motors. I don't have a meter that goes high enough, so I don't know what the current draw is, but it's more than the max 20 amps it reads for the few seconds I dare to test for. (the $2 Centech meter is only rated 10A, and isn't fused!)

Most testing was done as you see below:

A powerchair motor with gearbox, the cheap old DMM (given to me by Wilgar), and the Tektronix 531A scope (given to me by Mark) that is older than I am (but still works fine!). Power provided by the pair of batteries I'd normally have on the bike, the MK-Powered 12V31Ah SLAs. For the tests, I used very thin wires, in case something went wrong and the circuit breaker didn't pop, so I could easily yank them loose. Never had to, but still....

The single pair of small MOSFETs I have on it right now don't even get warm, even with the maybe 4 or 5 square inches of heatsink, in the hour or more I ran the heaviest motors on it, with light loads on them. Barely got warm for the short time I could test stall currents for. The bigger 4pole powerchair motor gets pretty warm just sitting there running, with no load on it, after an hour or so. I'm wondering if something is wrong inside causing too much friction or something with the brushes/commutator, perhaps. I have to take it apart and check. But the controller doesn't warm up.

Now I just have to put it into the case, wire the big MOSFETs on the external heatsink up to it instead of the on-board ones, and take it for a test ride on the bike. :)

This is a video of the testing of the controller with one of the powerchair motors (it has a commutator segment lifted, I think, because it clicks everytime it passes a brush--different from the 4-pole motor's apparent heating issue).

The loud fan is in the Tek531A, to keep the vacuum tubes it uses cool. Even so, you can clearly hear the motor problem as it speeds up.