30 Nov
2005
30 Nov
'05
5:32 p.m.
Random failure of the weekend was a UC3844 IC in a switchmode power supply -
with a very loud bang, the entire top of the chip blew off, but there's no
*obvious* sign of any other failure within the supply.
This was one of two such supplies in a Sun RAID disk controller; the two do
current-sharing under normal conditions - but the unit will run on one supply
if the other fails, which is exactly what happened here. Leads me to
*tentatively* say that the RAID unit itself is 100% and the failure was
totally in the PSU.
So, given that assumption for now, anyone know of a likely fault in a typical
SMPSU that might cause such a catastrophic chip failure but no other obvious
damage? Seems like a strange one to me, but I assume that something else in
the supply has failed and the chip just couldn't handle it. Whilst knowing a
bit of SMPSU theory (and practice), I'm not familiar with the UC3844...
Of course maybe it was just time for that chip to die (it was at startup, and
this in a server which had probably only been power cycled a handful of times
in its life), but it sure was noisy about it.
Shame Sun don't publish schematics!
cheers
Jules
1 Dec
1 Dec
2:48 p.m.
Random failure of the weekend was a UC3844 IC in a switchmode power supply -
with a very loud bang, the entire top of the chip blew off, but there's no
*obvious* sign of any other failure within the supply.
This deos not mean there are no other component failurs. This sort of
thing is very common actially.
A typical SMSPU circuit contains a section something like this
HT+
---
|
)
) Chopper transformer primary
COntrol IC )
(Part) )
|
------------+ base resistor /
| (~10 Ohms)_ |/
Out |----/\/\/------------| Chopper transistor (NPN)
| |\
| >
| |
I Sense |------------------------+
| |
| /
Gnd | \ Sense resistor (~0.5 Ohm)
-------------+ /
| \
| |
+---------------------------------+
|
---------
HT common
There are likely to be more diodes, snubber networks, etc, but this is a
start. HT+ and HT common are the 2 ends of the mains smoothing capacitor
and have about 400V between them
What happens is that the chopper trnasistor goes short-circuit all ways
round. The HT line is now effectivel shorted out via the transformer
primary (very low DC resistance), the shorted transistor, and the sense
resistor. The latter normally burns out, it is likely to be a 'safety'
flameproof resistor so you may not see any signs of failure. If you are
unlucky, some of the PCB tacks fail as well. The chopper control IC now
sees about 400V between its ground and its output/current sense pins. It
generally expires esplosively.
I would be very suprised if an IC failed that violently without there
being other component failures.
This was one of two such supplies in a Sun RAID disk controller; the two do
current-sharing under normal conditions - but the unit will run on one supply
if the other fails, which is exactly what happened here. Leads me to
*tentatively* say that the RAID unit itself is 100% and the failure was
totally in the PSU.
Very likely. Most SMPUS failures result in no output, not over-high
outputs (although watch out for 2-stage supplies if the regulator chopper
fails!).
Shame Sun don't publish schematics!
It wouldn't do you a lot of good if they did, the supply is almost
certainly bought-in and there wouldn't be a schematic in the service
manual anyway. But it can't be that hard to draw out, can it?
-tony
4 Dec
4 Dec
4:39 p.m.
Tony Duell wrote:
I must have been lucky... all the ones I've had to fix before have had failed
in other ways...
It's not *that* simple, because this is such a compact unit. There are a
couple of vertically-mounted boards soldered on the main PCB. One contains the
dead UC3844 and holds the circuitry that sits between the high voltage input
side and low voltage output. The other board seems to be purely on the HV DC
side and has a few surface-mount resistors on it, a small transistor, and a 16
pin IC labelled UC385x (there's a splodge of dye or something on it so I can't
read the last number - doubtless it'll be easy to find via google when I come
to look it up anyway).
Then there are three large L-shaped heatsinks on the board which wrap over the
top of the board components, and then another large slab heatsink is screwed
to those.
The slab heatsink comes off easily enough, but method of PSU construction is
to bolt heatsinked components to the L-shaped heatsinks *before* soldering
them in place. There's no access to undo the transistor / diode heatsink
clamps without desoldering half the board. Component level repair was
obviously not on Sun's list :)
I can trace stuff with a continuity meter, common sense, and a visual on the
PCB traces on the board underside (where luckily 95% of them are) - but a
visual conformation of the PCB's topside is out without some major
dismantling... (which of course I'm going to have to do if the chopper
transistor's dead anyway in order to change it...)
cheers
Jules
Random failure of the weekend was a UC3844 IC in a switchmode power supply -
with a very loud bang, the entire top of the chip blew off, but there's no
*obvious* sign of any other failure within the supply.
This deos not mean there are no other component failurs. This sort of
thing is very common actially. I would be very suprised if an IC failed that
violently without there
being other component failures.
Yes, so would I to be honest.
Shame Sun
don't publish schematics!
It wouldn't do you a lot of good if they did, the supply is almost
certainly bought-in and there wouldn't be a schematic in the service
manual anyway. But it can't be that hard to draw out, can it?
5 Dec
5 Dec
2:37 p.m.
Tony Duell wrote:
I must have been lucky... all the ones I've had to fix before have had failed
in other ways...
You have been very lucky. I have never had a chopper transistor fail
open-circuit.
SMPSU problems tend to come in 4 categories for me :
Startup resistor open-circuit
Chopper transistor shorted, many other components burnt out as a result
High ESR capacitors
Others.
Random failure of the weekend was a UC3844 IC in a switchmode power supply -
with a very loud bang, the entire top of the chip blew off, but there's no
*obvious* sign of any other failure within the supply.
This deos not mean there are no other component failurs. This sort of
thing is very common actially. Shame Sun don't publish schematics!
It wouldn't do you a lot of good if they did, the supply is almost
certainly bought-in and there wouldn't be a schematic in the service
manual anyway. But it can't be that hard to draw out, can it? dead UC3844 and holds the circuitry that sits between
the high voltage input
side and low voltage output. The other board seems to be purely on the HV DC
side and has a few surface-mount resistors on it, a small transistor, and a 16
pin IC labelled UC385x (there's a splodge of dye or something on it so I can't
read the last number - doubtless it'll be easy to find via google when I come
to look it up anyway).
Then there are three large L-shaped heatsinks on the board which wrap over the
top of the board components, and then another large slab heatsink is screwed
to those.
The slab heatsink comes off easily enough, but method of PSU construction is
to bolt heatsinked components to the L-shaped heatsinks *before* soldering
them in place. There's no access to undo the transistor / diode heatsink
clamps without desoldering half the board. Component level repair was
Can't you desolder all the transistor and diode connectuons and pull the
heatsink out? That's what I would do (and how it was presumably
assembled).
obviously not on Sun's list :)
Alas not...
I can trace stuff with a continuity meter, common sense, and a visual on the
I find the continuity test a lot better than my eye at following a track
(particularly on a board with many find tracks covered with soldermask).
But watch out for anything that will appear as a short on your meter --
inductors, transformer windings, low value resistors (often used for
current sensing in SMPSUs), etc. Unoslder these (at least at one end)
before you start or you will go insane.
-tony
8 Dec
8 Dec
8:20 a.m.
Tony Duell wrote:
Until now :)
Tony Duell
wrote:
I must have been lucky... all the ones I've had
to fix before have had failed
in other ways...
You have been very lucky. Random
failure of the weekend was a UC3844 IC in a switchmode power supply -
with a very loud bang, the entire top of the chip blew off, but there's no
*obvious* sign of any other failure within the supply.
This deos not mean there
are no other component failurs. This sort of
thing is very common actially. I have never had a chopper transistor fail
open-circuit.
casualties so far are both choppers (IRF740's), the UC3844, and a couple of
resistors (they were hidden by the heatsink before; rather obviously cremated
now I've removed it!)
The UC3844's current sense line got a blast of mains down it which is what
killed it...
I'll test other components in the area though in case they've failed with no
visible effects.
cheers
Jules
12:08 p.m.
I have never
had a chopper transistor fail open-circuit.
casualties so far are both choppers (IRF740's), the UC3844, and a couple of
resistors (they were hidden by the heatsink before; rather obviously cremated
now I've removed it!)
The UC3844's current sense line got a blast of mains down it which is what
killed it...
I'll test other components in the area though in case they've failed with no
visible effects.
6954
days inactive
6961
days old
5 comments
2 participants
participants (2)
-
ard＠p850ug1.demon.co.uk -
julesrichardsonuk＠yahoo.co.uk