VAX 4000-500 PSU Overload?

[Maciej W. Rozycki]

On Sun, 13 Dec 2015, Robert Jarratt wrote:

> >  No leaking ChemiCon SXF-series capacitors inside?  They seem to suffer
> > from some kind of design or manufacturing defect and consequently the
> > electrolyte gets through the seal eventually even in parts never used.  A
> > number of DEC PSUs have these capacitors, including the H7874 PSU which
> > BA430/BA440 VAX 4000 cabinets use.
> 
> 
> Just opened it up and there are indeed leaking caps, the brown ones. There
> is also a strange deposit on a couple of screws (picture here:
> http://1drv.ms/1micVN7), but the screws are nowhere near the leaking
> capacitors. Any thoughts on that?

 Obviously zinc plating of the screw has come in reaction with something.

 It doesn't seem from the photo electrolyte could have reached there, 
however I'd double check that, in particular inspect the surface of the 
PCB throughout looking for any difference in appearance.  This stuff can 
reach far, e.g. in one of my failed H7821 PSUs which was unfortunately 
stored at an angle to its proper operational position (or otherwise the 
upside down position of problematic capacitors have prevented the seal 
from failing), electrolyte travelled at least 15cm/6" away from the failed 
capacitor.  And this screw is located in a bottom corner of the PSU when 
in its operational position so gravity will drive leaked electrolyte 
towards it even if capillary action does not.  Other than that I have no 
idea.

 Lead or tin from the soldering alloy aren't as reactive as zinc is and 
therefore they don't show signs of corrosion so quickly as zinc does, 
although as I observed in the failed PSUs I have they eventually do too.  
As does copper from any exposed traces and can the solder mask or the 
epoxy substrate of the PCB itself too.

> >  Unfortunately replacing the parts is not easy as heatsinks block access
> to the
> > soldering pads and you need to desolder main rectifiers first to gain
> access.
> > You need to clean any electrolyte spills too as they will cause corrosion
> and
> > shorts.
> 
> Yes, I have had one of these PSUs apart before and know how horrible it is
> to get at these parts. I will however go ahead and replace these caps.

 Indeed, and good luck!

 NB the two brown capacitors seen in the middle top of your photo are 
among ones that leaked in my PSU; an SXF marking is barely recognisable on 
the edge of the lower one in your photo.  After considerable effort 
earlier this year I succeeded with removing the intermediate heatsink 
blocking access to these caps, by desoldering the four rectifier 
dual-diodes holding the heatsink in place -- at the cost of losing a leg 
from one of the smaller ones (a Motorola MBR3045PT) and some damage to the 
PCB.

 Fortunately MBR3045PT parts are still available and I was able to get a 
replacement, and the damage to the PCB is I believe not critical.  What's 
important the larger rectifier dual-diodes (84CNQ045; no clear indication 
of the manufacturer) have survived intact, as these seem to have become 
unobtainium now.  Apparently the last die foundry capable of making these 
parts has discontinued them earlier this year due to lack of customer 
interest, so the only source remaining might be part recovery from 
otherwise broken equipment.

 Overall I think I'll need better tools to be able to desolder such stuff 
in a more repeatable and less destructive way.  The thick legs of the 
84CNQ045 parts combined with the large volume and consequently thermal 
capacity of the intermediate heatsink seem to be able to take heat away 
virtually instantaneously.  I'll appreciate your advice on choosing a good 
soldering/desoldering station, capable of handling such high-current (and
consequently highly heat-conducting) parts.

 For the curious and possibly to provide some information on the parts 
involved I've documented progress with the disassembly of this module at: 
<ftp://ftp.linux-mips.org/pub/linux/mips/people/macro/private/h7874/> 
(large photos!).  Removing the offending caps revealed C322 and C323 
designation underneath; these are 330µF/25V parts.  I decided not to move 
forward with installing replacements and reassembling the module, or 
proceeding with disassembling the other module without upgrading my 
soldering/desoldering tools first though.

> >  I observed similar symptoms with the failed PSU as you do: it started up
> > briefly, enough for diagnostic output to start coming from the system
> > through the console port, and then within a couple of seconds the PSU shut
> > down.
> 
> I don't have an easy way to apply test loads to all the outputs at the same
> time. Is it possible to remove the final output boards for each "half" of
> the output and test one at a time without causing the PSU to shut down?

 No idea.  I've only recently started finding my way with this system; I'm 
mostly a software person.  Admittedly quite a low-level one, but still on 
the software side.  And it's been only earlier this year I discovered the 
issue with the dreaded SXF caps, even though I've been using various DEC 
hardware/PSUs for over 15 years now.

  Maciej