modern switching power supply faults

Tony Duell ard at p850ug1.demon.co.uk
Sun Apr 13 11:51:44 CDT 2008


> 
> 
> I have two Dell switchers out of P4 based 1U servers.
> 
> The first one has a failure in the startup circuit... pressing the power 
> button will
> lead to a flash of the power light... and that's it.
> 
> The second one failed spectacularly....
> 
> There is a wire wound power resistor (5W15RJ) near the AC input... it 
> cooked...

Di U deduyce from that marking that it's a 15 Ohm resistor?

Is it possibly part of a soft-start circuit, to be shorted out by a triac 
or similar once the supply has got going. If so, then if that fails and 
the resistor is left in circuit, it will overheat.

There have been supplies where there's a startup resistor -- to power the 
control circuitry until the rest of the supply gets going -- which would 
overheat if said supply took too long to get goiog. The resistor simply 
couldn't power the circuitry for long on its own. But I would expect such 
a resistor to be a lot more than 15 ohms

> and I do mean cooked.  It came out in pieces (with almost no effort), 
> melted the
> nearby capacitor and a nearby relay.  It also did a pretty good job of 
> puckering/burning
> the PCB (but not bad enough it could not be used).

What is that relay connected to/used for? It doesn't short out the 
reistor by any chance as part of a soft-start circuit, does it?

> 
> I suspected maybe shorted primary full wave rectifier... but I used the 
> DVM on
> diode test and got what I'd expect.
> 
> Oddly, despite there being a fuse on the primary side... it's still 
> good... it wasn't
> taken out.
> 
> That resistor got hot hot hot.  Could that resistor have failed on it's 
> own ?
> I suspect something further down the circuit must have a problem.
> 
> I'd be happy if I could make one power supply out of the two.  These are not
> simple supplies however.  While much of the switching transistors, 
> regulators,
> bridge rectifiers, capacitors and the usual fair in a switcher are 
> pretty readily
> identified, there is a 6" x 1" "brain" board (I'll call it that as it 
> has adjustment
> pots, the ps fan circuit, quite a # of surface mount components on both 
> side of
> it, and it interfaces with the main PCB with 34 connections.
> 

Any recognisable chips on that board?

FWIW, there have been some very complicated SMPUs in classic machines. 
The HP9845B, a complicated machine in all respsects, has a PSU on a 
3-dimentional sculpture of PCBs, with 4 chopper transistors, 2 main 
chopper transformes, and IIRC 17 ICs on the control PCB.

-tony




More information about the cctech mailing list