Dashboard Clock goes off?

[sven337]

Just replacing the transistors doesn't seem to work (because they'll die again soon), so let's figure out what they're for exactly and take a higher level look: what is being implemented here?
The way they're connected, the six of them are basically in parallel and drive the cathode (the horizontal filaments) of the VFD. Their job is to get a steady 300mA through these filaments (full brightness, figure empirically found with a bench power supply).

In fact, you can remove the 6 transistors, and connect a constant-current power supply set to 300mA between the + and - (leftmost and rightmost) pins of the VFD, while it's plugged into the board - it will work normally.

I couldn't fully work out how the transistors were driven (=what their base is connected to), but it all goes back to the mainboard and a, on the power board, a pin that seems to be used to dim the VFD. I don't remember it dimming when turning on the headlamps but maybe that's actually what happens. I'm about to lose that feature anyway.

So, instead of trying to fix this broken cheap power supply with such a terrible efficiency that the transistors dissipate multiple watts (about 500mW each, which is not nothing for SOT-89), let's just replace it with something that works better. My theory for why the repair isn't working is that this design relies on the board helping to dissipate some heat, and when it's as badly burnt as mine it doesn't dissipate as well. Of course as soon as 1 transistor dies the rest will follow due to the parallel design. Crappy power supply design that saved maybe 2 bucks.

My plan is to find a buck module with a 2.5V output that I can use as a constant-current device, or if I can, add a series resistor to limit the current to about 300mA.
Then I'll rip out the 6 transistors and connect the buck module where I need it.