PSU recapping - what's the general consensus on when to do it?

[Pickelhaube808]

My general rules towards when I should put in the time and effort to recap...

- Not working (obviously)

- Flakey functionality that is likely related to power problems (a bad power supply can affect everything on a board)

- Contains known bad capacitors (like ones from the cap. plague)

- The components or complete system are old and rare/expensive enough to justify it even if nothing is currently wrong

 

[modem7]

At one place where I worked (in the 80's I think), time was a premium but not so money. If a SMPS failed, we would spend, say, 10 to 15 minutes on it at most, looking for open/shorted PN junctions, etc. If we could not fix the SMPS within that time period, the SMPS would get shelved, and the faulty parent equipment get a replacement SMPS. At some point, a company was found that would fix 'any' SMPS for a fixed cost. If they could not repair the SMPS, there would be no cost. This arrangement worked out well, and for us, quite economic. We started to take a look at the soldering in the fixed SMPS' to see what was being replaced (yes, we looked at the soldering before the unit was sent out). The redone soldering indicated that the company was simply doing a replacement of all of the small to mid value aluminium electrolytics around the switch mode controller chip. The large input filter caps were not touched, neither the filter caps on the outputs.

I can imagine some non-technical person doing the cap replacement (i.e. labour not requiring technical knowledge), and then the SMPS went to someone who had the knowledge on how to test it.

Occasionally, we would see additional repair work, e.g. optocoupler swapped out.

 

[SiriusHardware]

We started to take a look at the soldering in the fixed SMPS' to see what was being replaced (yes, we looked at the soldering before the unit was sent out).

That's quite interesting to hear you say - I used to work for a company in the UK which undertook similar 'repair anything' work, but we used to go one step further and try to ensure that the soldering of the replacement parts was done exactly the same way as it had been for the original parts, if the solder was sparsely applied, we did that, if the solder had originally been heaped on, we did that, if the capacitor legs had originally been bent over to keep the part in the board, we did that. In short, we went to considerable trouble to exactly reproduce the original style of soldering.

If you clean a solder job up really thoroughly with the right sort of cleaner, it gives the new joints the same 'old' dull looking finish as all of the other original joints, making it difficult to tell what has actually been resoldered. The reason we went to all this trouble was to make it more difficult for anyone to 'steal' our hard-won knowledge of what were essentially common recurring faults in equipment. We knew them, but having found them the hard way in the first example seen, we didn't want to give that knowledge away to any casual observer.

Occasionally this did cause us problems when a customer would take a look to see what we had done, and it looked as though we had replaced absolutely nothing except of course that their item was now working - this would then lead to the suspicion that we had just found a loose connection or something and charged them for more than we had done.

 

[modem7]

That's quite interesting to hear you say ...

The company we were using obviously didn't care.
Maybe they thought there was no point since an alternate method (albeit requiring more effort) of detection was to record components via things like make, model, date, and packaging.

 

[SiriusHardware]

Of course you are right, if anyone was determined to know what we had changed they could have recorded the details, brands, colours, values, markings of every single component in the unit before submitting it for repair - we just aimed to make it more difficult for anyone looking for obviously disturbed or resoldered components at the after-repair stage.

Just to give you an example, there was a particular type of motor drive unit which we used to get in some numbers from a company which obviously used a lot of them. Fairly early on, we realised that essentially the same two electrolytics were failing every time so they were a good earner for us, a predictable easy fix, but would also have been an easy fix for the company we were fixing them for if they had any technically competent people like you who were able to 'read' what we had replaced. If they had been able to determine what we were doing to fix them, then that particular revenue stream might have dried up for us.

Known 'Stock' faults like this have always been the bread-and-butter of any repair technician / repair company whether they work in the domestic or industrial field, they compensate for other jobs which are long and difficult and may have to be repaired at a loss, the first time at least.

 

[modem7]

Known 'Stock' faults like this have always been the bread-and-butter of any repair technician / repair company whether they work in the domestic or industrial field, they compensate for other jobs which are long and difficult and may have to be repaired at a loss, the first time at least.

VCR's were very complex. I remember doing a 'VCR theory of operation' type course (even though it would be rare that I encountered one). Still, it could be a lot of effort to track down the cause of certain faults. A technician where I worked discovered a VCR/TV repair technician selling a hard-copy database of field-learned fixes for many make-models of VCR. We all chipped in to buy the document.

 

[MattisLind]

I read an article on the web a few years back (which I obviously cannot find now) which had some basic rules which I adhere to.

Basically I remember the checklist as follows when dealing with some piece of equipment to restore :

* Replace if leaking, bulging, smelling.
* Replace capacitors of same type like those which were leaking, bulging, smelling.
* Replace capacitors that measure out of tolerance, both capacitance and loss-wise
* Keep the rest

So for example in a Tektronix 611 from the early seventies which I just recently worked with I only replaced one capacitor. It was a tantalum capacitor that was corroded. I checked the other capacitors and they all measured very good. Low loss and very close in capacitance.

Another project is a DEC VR241 screen made by Hitachi. I checked the capacitors. One had a slight bulging top. Replaced all of the same type (even though they measured fine out of circuit). Found one 2200 uF cap which was only 1700 uF. The rest was within spec and low loss and no leaks neither smells could be detected.

The slightly annoying thing is that the replacement cap for the 2200 uF was measuring slightly below 2000 uF, so still within 10%. It is not certain that a new capacitor is measuring better than an old. DEC often used epoxy sealed capacitors in their designs. I seldom find those to be bad.

On the other hand, in a Macintosh SE/30 supply I worked with a couple of years back most of the output filter caps were leaking so they got replaced and so did the capacitors on the motherboard since they also leaked.

 

[pkhoury]

If it's average consumer-grade stuff, yes, I'd get out the meter and check the electrolytics. But I'm not going to junk a bunch of oil capacitors because they're "too old". And there's a whole bunch of fairly recent stuff that's been destroyed by "shotgun replacement" of caps. Any time that soldering iron comes out, one must take into account the possibility of damage.

It's the "shotgun" advocates that I have the biggest problem with.

I don't really recall mentioning oil capacitors, though I did replace some non-functional PCB-laden caps with modern Chinese/Grainger caps (they are for a wig wag, an antique railroad signal). I'm also not trying to shotgun replace everything, hence the purpose of this post in the first place.

 

[mark0x01]

Some caps just bulge without even being installed. I had a bag of "cheap" 105C low ESR ones I left on a sunny window sill and they bulged out within a summer.
I also have "new old stock" PSU's from the late 80's that have done the same thing on some caps while just sitting in a box.
Drying out isn't always visible, so testing is the only sure way of weeding out the potential issues.
I usually reform large linear psu caps and then do a timed voltage drop check without a load to check for leakage, and then with a resistor to check it's value.