Having failed to find anything wrong with the capacitors C38 (large electrolytic), and C2/C46 (small blue ceramic capacitors) in the charger I'm repairing, I've decided to take the advice of a colleague and attempt to separate the gate control signals from the drain/source output signals. Last time we were testing, he detected faint crackling sounds, almost like high voltage arcing over. I thought that could surely not be the case, since arcing would create a fearsome bang. But he suggested that perhaps something was leaking slightly, causing one of the MOSFETs to start conducting when it should not. This is not a problem until the CPU directs the PWM chip to start switching all the MOSFETs; when its companion on the same half-bridge switches on, there is "shoot through" (two transistors conducting from bus+ to bus-), and then the fearsome bang came on cue.
Throughout the charger, there is generally at least 5.08 mm (0.20") pin to pin between any components with high voltage. The MOSFETs are a glaring exception; these are TO-220 devices (so many of them are) with 2.54 mm (0.10") pin to pin spacing. There are thick tracks to the source and drains of the MOSFETs, to reduce inductance and resistive losses. This makes the solder connections bulge somewhat, reducing the gap between for example between drain and gate. It is possible for all sorts of material to bridge that gap - resin from soldering, bits of insulating material (black, white, and yellow gunk), and so on. Dust and moisture can creep across this bridge, allowing an arc to form. The drain jumps from bus- to bus+ in a short time (a microsecond or two I believe) when the upper MOSFET turns on, so even if there isn't much conductance, there can be capacitance coupling the drain to the gate of the bottom MOSFET. All this is encouraging shoot-through.
His suggestion was to cut an air gap between the pins if possible, or stagger the leads a little (so they form a triangle rather than a straight line), or both. It turns out that the former was more possible than I thought, since there is the large cutout in the PCB for the heatsink, but there isn't clearance for the repositioned drain lead.
He also mentioned that I needed to clean off as much of the burned epoxy (now from two sets of blown MOSFETs), as the burned epoxy would presumably contain a lot of carbon, which is of course conductive (in that form).
The above was taken before I extended some of the slots a bit further. You can see an area at the left where I started to make room for a staggered drain lead, but decided that it would just make the clearance issues worse.
I was also not vigilant last test for scratches in the solder mask:
The yellow circles show areas where the solder resist has been scratched, revealing bare copper. It happens that these are gate leads (marked with a red "G"), near a B+ area (also scratched!) and a drain/source/output track (marked D/S). These also connect to CON32S, the 32-pin connector between the main board and the control (daughter) board. So again, gate signals are routed very near drain/source signals, and I had been removing black gunk from components to read their values and so on. So I'll be using some silicone or the like to protect these areas from creepage before the next test. The scratched solder resist will of course also end up under silicone.
[ Edit: In the above photo, you can just make out the values for C2 and C46: 2.2 nF (marked as 222K) and 220 pF (marked as 221J). These happen to be 1 kV parts; some chargers have 2 kV parts. I think they just use whichever is available. ]
So now I need to replace the MOSFETs again, hopefully for the last time. These holes have no thermal relief, there are heavy copper wires right up close to the MOSFET leads, and the PCB is extra thick to take the weight of all the magnetics (1.8 mm (0.0709", 9/128") instead of the usual 1.6 mm (1/16"). All these factors make it very difficult to remove the solder from the holes, using either a hobbyist solder sucker or solder braid. I don't have access to a proper desoldering station; maybe one day.
I've had success in the past using pins. Pins are usually made of steel with a thin coating that doesn't stick to solder all that well. But the usual household pin is about 0.6 mm diameter. The MOSFET leads are about 0.7 mm across. I tried quilting pins (0.7 mm diameter), hat pins (also 0.7 mm diameter, but lovely and big), and various sizes of safety pins. The safety pin was tolerable, but the coating was adhering to the solder too much, so I had a lot of trouble getting the pin out one it poked through.
In the end, I used ordinary sized paper clips; these are about 0.8 or 0.85 mm diameter. They are cheap enough (and I'm pretty cheap
) to simply cut and discard:
Here you can see the clip has poked through the hole and rotates freely, but a lot of solder has adhered to the clip, making it very hard to remove. So now I merely cut the paper clip with side cutters, and the clip having almost no solder away from the end comes out easily.
In the left of the above photo you can also see R13, which is 1.5 mm x 38 mm of tinned copper wire. This was mentioned recently by KennyBobby; thanks, Ken! Normally this text is hidden under a piece of heat conductive insulating material, and in some cases also under a piece of foam that holds it gently away from the chassis of the charger. That's a 2 kW charger; the layout is slightly different for 1.5 kW models, but the resistor appears to be identical.
[ Edit: middle photo had some tracks mislabeled. ]
Learning how to patch and repair PIP-4048 inverter-chargers and Elcon chargers.