On 7/23/07, Josh Dersch <derschjo at msu.edu> wrote:
Thanks, everyone, for your helpful responses!
Happy to help.
I'll definitely test the
power supply out with a dummy-load, just need to run out and get some
parts when I get a chance.
You can use a lot of things for a dummy load - the reason that
headlamps have been suggested is that a) they are resistive and
shouldn't make the combination PSU+load circuit too complex, and b)
they are much less expensive than other things in case something goes
wrong. I have used old hard drives in the past as "dummy loads", with
no ill after-effects, fortunately. Sand-block resistors in the 5W
range can also be used, but unless you happen to have an assortment
lying around, light bulbs are probably easier to aquire at an auto
parts or general merchandise store.
Have I missed any reasons?
It does appear to run fine with no load
whatsoever, and the voltages it's putting out look to be correct under
those circumstances.
There must be some sort of minimum load present in the PSU (not an
uncommon design for most switchers, though I have seen switchers made
in the 1980s that _do_ require an external load of several watts to
kick on).
What's the best method for testing for bad
capacitors?
Aluminum electrolytics frequently (but not universally) exhibit
visible bulges or leaks, but they can still be defective and look
perfectly fine. Ceramic and tantalum caps, being rigid and brittle,
don't tend to look bad at all (but defective tants have been known to
explode).
The only way I know to be certain is to take the cap out of the
circuit and check for shorts with a VOM and/or capacitance with a
capacitance meter (some not-cheap Fluke meters, for example, can do
both at a twist of the function dial).
After you run your PSU on a dummy load, if it checks out, I'd check
your motherboard for shorts of power rails to ground. If it's the +5V
line, there will be lots of caps that are suspect. If it's one of the
other voltages, there will almost certainly be fewer candidates. The
trick to any component-level debugging is to divide and conquer. Try
to set up measurements that will help you eliminate parts of the
circuit or that will help you zero in on a suspicious area.
It's possible that you could have a chip shorted to ground, but unless
your board happens to be 100% (or nearly so) socketed, it'll be tough
to check them all.
You might also check the board for signs of repair. I've seen hack
jobs caused by clumsy fingers that had nothing to do with dead
components. If all the joints and pads have a uniform grayness to
them, then you probably don't have that as an issue, but if you see
shiny solder joints on one or more parts, take some time to snoop
around that area to reassure yourself that you don't have a bridged
connection somewhere.
Given how old an Apple III is, though, it wouldn't be a shock (no pun
intended) to find out you had a defective cap or two. Checking for a
supply-line short to ground is a quick test that should help either
narrow down or eliminate swaths of the board quickly.
-ethan