So here's a boring geeky post about Switched-mode power supplies.
Amazingly I'd never actually bothered to find out what one is until yesterday. To me a power supply has always been a step-down transformer followed by a full-wave rectifier and a large smoothing capacitor; maybe also a large inductance to filter out the remaining ripple. You can calculate the output voltage from the input voltage and the turns ratio of the transformer, for an unregulated supply (which will drop slightly as load is applied), or make the output of that stage a bit higher than you actually require and put in a voltage regulator IC (commonly 780X for X volts output). I even made a power supply along these lines, mounted it on a block of wood with terminal blocks for the outputs and covered it in a tin made out of the cut-up sides of an old can of oil. It was referred to colloquially as "the bomb" and I think it had three different DC outputs, one of which was variable from 1–21V using a knob on the top (and a variable voltage regulator IC).
So when I took apart a non-working domestic appliance yesterday I was confused.
The device in question was a Freeview receiver and hard-disk recorder which had failed in exactly the way described in comment 4 of a thread on digitalspy. Seeing that my multimeter was wildly fluctuating when asked to measure what ought to have been a DC power rail I surmised that a large AC component was leaking through and that therefore a capacitor somewhere had failed. It was then that I found the DS forum thread, and having read that, I noticed a capacitor with a bulging top. Sadly, replacing that had no effect at all, and all the remaining capacitors looked physically OK and behaved — so far as my multimeter could tell — like capacitors (in fact, so did the one I had taken out). Which is how I came to be taking the power supply circuit board out so I could ascertain a bit more about how it was supposed to work.
The power supply circuit board was in two sections: a smaller mains side and a larger low-voltage side (with quite a lot of different components on it), and there was one IC which bridged the two sides (an opto-coupler, it turns out). What was surprising was that the mains input was itself being rectified and stored in a high-voltage capacitor (which I will treat with a bit more respect next time I poke around after just unplugging the thing!).
So how it works is this: after passing through a low-pass filter the mains input is rectified and charges a capacitor to 339 volts DC. This supplies an IC on the mains side (the "switch" in "switched-mode power supply") which interrupts it at high frequency and feeds it through a high-frequency step-down transformer (which is much smaller than an equivalent mains-frequency one) and this is rectified and smoothed to form the DC output. The opto-coupler feeds back a signal to the switching IC, which varies the duty-cycle — and thus the eventual DC output voltage — until it is at the correct level. So this type of power supply is self-regulating and there is no need for a 780X, which wastes a lot of heat. (But this one does have a 78R12 mounted on it, so go figure.)
So I will take a punt and replace the high-voltage capacitor. It seems to me it's the most critical one in the system (and it fixed it for someone else in the thread).
Having put that device back together without fixing it (but minus one capacitor), I cracked open a VCR that failed some time back, when I said "it's probably to do with the power supply and might be easy to fix". This also turns out to have a switch-mode power supply in it, and this time there is no doubt what is wrong: electrolyte had leaked out of the high voltage capacitor and become crusty underneath it, and one of its legs was no longer firmly attached.
So I will replace that high-voltage capacitor too.
Replacing blown capacitors — isn't that what TV repair men of the early 1980s always used to do? I thought this was a thing of the past — but apparently not.