Failed repair attempt on a BSH dishwasher 9000 683 387 power board (TNY264GN) (#P49)
Here’s something unusual for today: A failed repair attempt. Well, the patient is technically alive again, but very very sick and needs replacement. But he’s currently in a working state as a proof-of-concept 
This PCB is from a work colleague, it’s not from my dishwasher. The “MELECS 9000 683 387” branded power board seems to be a major failure point in some BSH appliances (Bosch, Siemens, Gaggenau, Neff, Constructa, maybe even Thermador, Balay, Pitsos and Profilo if these sub-brands of BSH sell dishwashers, I didn’t check). Failure is easy to spot – the device is out of power, nothing gargles when powering on, no functionality is available from the control panel.
The failure on the PCB is also easy to spot – the TNY264GN chip has disintegrated, your fraction of remaining IC may of course vary (cruel chopping of the legs for desoldering not done by me, I just got the remains packaged together with a bunch of new TNYs):
This destruction is typically not localized – usually, a fusible 100R resistor (3-5W?) in the power path also dies in the process. In our case, that didn’t happen, instead a PCB track clearly designed to evaporate in overload situation did it exactly that. If you see a repair kit that includes TNY264GN and a big ass 100R resistor – that’s why.
It is also possible to damage more components in the process, but some people in various forums got lucky and a replacement of those two things did the job. Since our PCB did suffer more damage and will be discarded, I cannot describe the process to the full repair, but offer some observations I made along the way.
First, here’s two aligned pictures of top and bottom, including another component that needed replacement. This is most likely not directly related, but the two + two X2 caps did have severely different remaining capacities and I only had a potted one with wires sticking out in my parts drawer. If you need to find a connection from top to bottom, this might be easier than flipping around your board a bunch of times:
(yes, I know, many large components do obstruct the view onto the tracks, I’ll add a photo if I get to keep the PCB and can remove all the large bits)
Back to the X2, it’s a Epcos/TDK part labelled B32922 X2 MKP/SH 40/105/56/B 1414 (250V~) 1283 (305V~), so 0.22µF, 305VAC, -40°C to 105°C rated polypropylene X2 cap with 15mm pin spacing. This must only be replaced with another X2 capacitor that fails in a safe manner (No old Rifa caps, *cough* #P14). The one close to the failed track was alright, the one closer to the AC connector was down to like 10nF and clearly worn out from overvoltage events from the mains. For cheapo power supplies, this is a failure mode, as already demonstrated in the repair of the Calore heater controller (#P30).
The electrolytic cap should be checked as well, since in the event of TNY failures like mine, there is a sawtooth pattern on the secondary power supply side that goes up to about 28 to 30VDC, which is way above the rating of this 470µF/16V Chemicon cap. If this has happened – please replace, even if a quick check says the part is still in spec, like this one way. Replacing it with a 470µF/40V type also made the sawtooth stop at around 16V, which does mean the original cap was either defective regardless, or very high ESR…so also defective.
Speaking of the sawtooth pattern: This is the voltage on the + side of this very capacitor. If you can see this, and this is slow enough to be visible with a regular multimeter, especially on ones that have a faster bargraph display: The repair is not over, this is not normal behaviour. For a fixed power IC, this should be a constant voltage from 8 to 15V (?), I don’t know the exact value. The circuit from here, supplied over a 78M05, does work from 8V, at 10V the relays really contact hard, and I’ve read about 13.5V being a good number. On my unit with a hard fix on the TNY side, this goes up to as much as 18V, which isn’t really plausible given the 16V rating of the associated smoothing cap shown above.
The best way to probe this voltage is likely this via here (+) against one of the many stitching vias around the 78M05 (-) directly next to it. The rather extensive GND network is also marked in the full view above with a black – (I was too lazy to re-export the image without…)
One more hint for probing: Since this is an AC power board, running this as intended while probing about is dangerous and potentially lethal. However, the TNY264 has a specified minimum drain voltage of 50V, which I can confirm from my tests. If you can manage to generate slightly above 50VDC from a lab power supply, either directly (my HP6644A does 61V), by combining several channels, or by offsetting it with let’s say 24V from a regular power brick (that’ll do for standard 30V lab power supplies), it is possible to run the unit from low-ish DC, making it much more safe to probe (the coil still produces spikes exceeding 250V!). Just make sure to use the correct pins on the AC connector – BSH has of course skimped and only uses a two-diode rectifier, meaning the center pin of the AC connector is used as regular GND, and the lower one, closer to the TNY chip, needs to be the + voltage. The top one can be ignored, unless you want to invert all voltages.
More on the TNY264GN: Pin 1 (BP) needs to be at 5.8V (mine is 5.89V vs. GND) all the time. If that isn’t the case, check the mandatory 100nF MLCC close by, it might be damaged.
Pin 4 (EN/UV) can be forced to GND via a 4k7 to 3k3 resistor, which disables pausing on overload or other strange scenarios, which was the case here. With the forced enable, the chip will just keep on pulsing, allowing to get the 5V and derived 3V3 rails up for further testing. This is absolutely NOT recommended as a fix since this will likely overheat and damage other components nearby (the transistor and the zener diode for example), but will allow for testing of the rest of the dishwasher to decide if a PCB replacement is worth it, or if other components might be damaged as well.
This is a thermal view of the board powered via AC and with all the fixes and force-on patch in place – if your board temperatures exceed this, check the hot spots.
Hottest component is the patched THT diode – a working unit should stay wayyy below this, as secondary voltages will be much lower and the TNY pauses when the desired voltage is reached. Here, the diode eats away quite some power, which does not happen in regular operation.
Also careful with the two 1103 = 110k resistors here on the low voltage side: These show up in the thermal image for a reason, these divide full rectified AC voltage down for whatever diagnostic purpose. Meaning the top one is at 400V, and the lower one is at about half that, before feeding into a small resistor that offers some 0-3V readout voltage (or something in that ballpark). Note further that, because BSH skimped again, this board has no real fuse – yes, it does have the 100R fusible resistor and a MOV, but there is no PTC or regular fuse anywhere. If you short something probing here, the 100R resistor limits your fault current to 2.3A (500 watts!) and holds that until the magic smoke has escaped, which could take a moment or two for that thing.
Other than that – well, the EPCOS labelled coil 50 11 009921 does have a resistance of roughly 0.1R and 4R for the two taps; the center tap on the one side does not exist, it is purely for mounting purposes, so it doesn’t get in backwards. I think these values should be correct and similar on working units. The unlabelled coil right next to it is just for EMC purposes (surprising given BSH’s penny-pinching) and plays no critical role in generating any voltage. It should have sub-1R resistance both ways, but checking of any common-mode filtering parameters is not necessary for debugging.
My unit came alive when touching the drain pin with a multimeter probe – the other side not connected to anything. This must have been some capacitive effect, as I could not reproduce this with either a fixed resistor or a capacitor to GND. It also did not occur with oscilloscope probes. I noticed this when checking the voltage on the drain pin, and suddenly relays started clicking. My money would be on a related fault (cracked via, micro open on some track), but in the end I did not find it and I have to give the board, wiring and plastic mounting parts back. When doing this, the primary voltage went up by several dozen volts, the secondary sawtooth voltage also got a little higher, and the 78M05 was able to keep the 5V just long enough in regulation so that the microcontroller downstream was able to make the first relay click. Not a fix, but likely worth investigating.
One last thing: The HFE8 12-1H-L2(610) relay is a bistable one (I don’t know why, but maybe the rest of the dishwasher allows for switching these?), meaning it does have a set and a reset coil. If the board is completely dead, you could check if this one is a) active and b) working. If it is not, the board power is completely cut from here on. Haven’t read about relay failure in any of the forums, but this could be a failure mode as well, leaving the entire board inoperable despite zero downstream faults. When applying 12V to set/reset this component, check the datasheet for polarity.
