rlogiacco:
I did put some load on each and every line: 10 Ohm resistors on 3.3V and 5V, 20 Ohm resistor on 12V and a couple more on -12V and -5V (don't remember exactly the values).
Gathering that you actually used ten Ohm and twenty Ohm resistors, that sounds just fine. At least one of those supplies requires a reasonable minimum current. The negative supplies do not need to be loaded as they may not actually supply power to anything in modern computers. (-12V for RS-232 ports, if present, some audio boards; -5V - nothing!)
rlogiacco:
I've just noticed something weird you might be able to help me interpret: with no mains plug and with the switch either on or off I measure low impedance between all the power lines and ground. I noticed that while trying to measure the two negative resistance values with my multimeter as it reported some odd values I wasn't recognising. Here is the listing:
5V 95.7 Ohm
3.3V 427 Ohm
12V 0.3 Ohm
-12V 2.2k Ohm
-5V 267 Ohm
Vsb 95 Ohm
I don't know if that's normal, but to me it seems the 12V line is shorted.
That sounds almost certain. The trick is of course, that if your meter applies a voltage in the direction that make the rectifier conduct, you will read the voltage drop of the rectifier rather than a resistance as such. On most multimeters, the voltage applied to measure resistance will be less than a silicon rectifier threshold, but these PSUs generally use Schottky diodes with a very low voltage drop, so may read all sorts of things. The "diode test" range on the meter offers a somewhat higher current to charge capacitors and directly read a diode voltage drop.
In the other direction, applying positive voltage to the positive outputs and negative to the negative, you will see a varying voltage as the capacitors charge, and then it will settle at quite high value, at least in the kilohms range except for whatever resistors are sensing the voltages to control the regulator. This resistor will generally be on one line only.
rlogiacco:
May I ask you a little bit more detail, may be a link or something? The PSU is already fused, I was trying to add an additional protection on top of the internal one, trying to save the internal fuse from my fat fingers :-D.
On a PC power supply, the negative goes to the mains ground. If a fuse in the ground line as you describe blows, there can still be a return path through the mains ground and other equipment connected to your circuits, or through another one of the supplies and whatever heavy current blew the fuse, may be directed through something else. Need I point out how that could be a problem?
The mains fuse inside the PSU is only to protect it from internal faults. A short on any of the outputs should shut down the switchmode system- as it appears to be doing here.
So what to do? Unsolder and remove the rectifier from the 12V line. Test it. Look for any connection to the regulator circuitry - if it is (and it most likely is) either 5V or 3.3V that is regulated, then the PSU will run without it.
Ok, that leaves me with some hope this thing is not blown. Any suggestion about something else to check?