24 Aug
2014
24 Aug
'14
2:19 a.m.
On 2014-Aug-23, at 11:30 PM, Josh Dersch wrote:
On 8/23/2014 7:12 PM, Chuck Guzis wrote:
Sure, the 7805 could fail after warming up and either go open or short internally. To test
it you could put a load drawing current equivalent to the computer (say 0.75 to 1A) on it
and see if you get the same failure after some minutes.
A scenario that may be at work is a short develops on the +5 line after something warms up
- an IC or a capacitor (are there any tantalums on there?). When the short occurs the
current-limiting in the 7805 regulator kicks in, so the shorted device doesn't see
enough current to go poof and fail completely. You could poke around to see if anything is
getting unusually warm.
A way to distinguish between a short vs the reg going open would be to watch what happens
to the voltage on the input side of the reg when the failure occurs. If it's going
open, the V should goes up due to the removal of the load. If it's a short the V
should go down due to the increased current draw.
I think Dwight had a technique for tracking power rail shorts with a 4-contact ohmmeter.
If you have 800$ to spare you could get one of these and trace current on the PCB:
http://www.aimtti.com/go/iprober/
The gonzo technique is remove the 7805, connect the logic Vcc line to a 20A +5V supply and
find out what component does go poof. Unfortunately it may be a PCB trace burning up
before the bad component does. (All liability disavowed).
Josh,
I'd first try lifting one side of CR22 and see if your +5 returns.
If not, you can remove IC 63 (probably what, a 7805?) and see you can read anything at
the high side of C67. You should be able to--unless you've got a bad diode.
If the CR22 maneuver doesn't result in any improvement, then you've probably got
a fault somewhere else on the board. Make a quick resistance check while you have IC 63
out with the power off, checking the resistance between the +5 lead and ground.
That should get you started on where to look.
--Chuck
Thanks! Lifting CR22 (which took awhile to find, it's actually clear on the other
side of the PCB from the rest of the supply) made no difference in behavior.
Removing IC 63 (which is a 7805, nice guess) and reading voltages on the positive lead of
C67 yields about 15VDC. The resistance between the +5V lead of the IC63 socket and ground
is about 3.5K ohms, so it doesn't look like there's a dead short here. Would the
7805 be a likely suspect here? Any easy way to test it?