At Project Day yesterday, Weber noted that the above topology is somewhat clever; with the interleaved nature of the SCC, there is always a smooth current drawn from the input (when one half is fluxing its inductor (top transistor of the pair on), the other half can be free wheeling (bottom transistor, really just a low loss diode, turned on). Presumably this reduces the current that has to be supplied by the capacitor at the input (this capacitor would absorb solar panel energy when none of the top capacitors are conducting, and supply it later). For those interested in the technical details, here is an article on multi phase buck converters
. It looks like ripple current reduction is the main benefit, but there seem to be others.
I spent most of project day installing the parts recently purchased for dockarl's PIP and the spare board he bought. It looks like his original board was one of the first PIPs manufactured, made in 2013. The replacement board is from 2014. It seemed to make sense to work on the 2014 board first. Seven devices went into that board: 2 MOSFETs from the buck stage (and thanks to Scott's manual, we now realise that these aren't in parallel*), 4 IGBTs from the bus end of the DC-DC converter, and the switcher for the bus soft start circuit. This latter one we replaced, even though the originals seemed OK, because the original fault code was 09, "Bus soft start failed". (It was cheap, contrasting with the > $10 each plus tax for the IGBTs). We now believe that when nearly anything blows up from along the power train, the microcontroller will pick this up as a bus soft start error. Weber pointed out that you can't soft start if you have no power, after all.
[ Edit: * Actually, they are in parallel, the manual's block diagram seems to be highly idealised. The MOSFETs or IGBTs for the buck stage are actually in the negative side of the bus, for example. ]
[ Edit: * It turns out that later models have only the one IGBT/MOSFET in the buck stage. These have a TO-220 diode in place of the TO-247 package.]
One of the buck MOSFET/IGBTs (they could be either) is right behind the biggest film capacitor I think I've ever seen. It measures 50 x 40 x 30 mm. When trying to put this MOSFET in, I thought this is nuts, I may as well pull out the capacitor and make things easier for myself.
It's doubly crowded in, by the fins of the heatsink on top and this capacitor in front. Others have the smaller (but still large) blue capacitors in front as well, but fortunately not this one.
Well, it has "leads" that are the same size as the contacts of automotive fuses; they're almost like quick connect terminals. So it was hell getting it out, and even more hell getting the solder out so I could put it back later. Fortunately, there are many plated through holes around the slots where these "leads" pass through the printed circuit board, so I didn't have to worry too much about losing plating between top and bottom side of the board. I ended up using a drill as a crude milling tool, carefully taking my time so that the thin drill didn't break. It did make the MOSFET easier to replace, but for the spare board, I decided not to remove the big black capacitor from that board
I set up a current limited power supply, with 24 V of batteries in series, so that hopefully nothing too violent would happen if anything went wrong. I switched on, and... nothing. No light from the LCD display, no fans whirring, no power LED. Huh. I'm sure it was working at least to that level last time, so now with all these repairs it's worse. But wait, last time we were running off the mains, not the battery. Ok, set up the mains wiring, switched on and... nothing.
Huh? But last time... Weber pointed out that the mains wire was attached to the other board, the 2013 one, so we must have tested on that board, not this one. But before I did the whole replacement exercise on the other board, I thought I'd try and see what's wrong with this one, if only to figure out what parts to buy. After all, with Scott's manual for the Axpert inverters, the one fairly complete schematic is for the power supply. Well, for the battery power supply; it turns out that there are two similar power supplies, one that runs off the battery, and one that runs from the AC input. They both use a UC3845 "current mode PWM controller" IC, but get this: one is a through-hole version, and one is surface mount! I can only imagine that one came first, say the battery one, and then it was decided that the AC input one would be added later, and they only had room to put it surface mount underneath. I guess they didn't want to mess with the existing through hole one; don't mess with something that works.
All the semiconductors seemed to check out OK, except for the UC3845 chip. The resistances didn't seem close to what the repair manual suggested that they should be. When I checked the other board, the resistances were much closer to what was specified. Resistance checks like these can be very unreliable, since they depend so much on the multimeter used, but I felt confident that the other board would work.
So I mentally noted to buy this chip, and set out to make the repairs to the other (2013) board. It turned out that it wasn't such a hard thing to replace the buck MOSFETs with the big black capacitor still in place, when I managed to squeeze a socket in there and rotated it by a combination of finger power and pliers. So this time, I connected up the power supply and battery, switched on, and... nothing. Waah!
But last time... Ah, again, that was on the mains. Could it be that three of the four power supplies were broken, and the one I hadn't tried was still good? Only one way to find out. I connected the mains wiring, switched on, and... Light! Beautific, glorious light was poring forth from the liquid crystal screen! Joy!
Now to connect a battery before I had to head home; it was getting late by this stage. But at that point, I spotted a medium sized white capacitor. Oops!
I had removed a white capacitor that was blocking the IGBTs (it has normal pigtail leads, so it's easy enough to remove and replace), and had forgotten to replace it. (Part of it is visible at the right end of the photo.) Sigh. So I'd have to wait for Weber to replace the capacitor and try again next morning.
That was this morning, and it seems to charge the battery just fine from the AC input. As I understand it, that involves the whole power train (full bridge at the AC output, buck converter, and two full bridges at the DC-DC converter on either side of the high frequency transformer). Actually, it may not involve the first two, I just don't know how the battery charger function actually works. But it's a pretty good sign. Weber didn't attempt to fire up the inverter part as an inverter, reasoning that something bad may happen because the power supply (from the battery) is not working.
I have the other board here now, and will attempt to get it working. Hopefully, the same repair will get the other board working as well. So dockarl, we think we're getting close to repairing the two boards.
[ Edit: by the heatsink -> by the fins of the heatsink; white capacitor was blocking IGBTs; visible at right end of photo. ]
[ Edit: buck converter is not in reverse when AC charging.]