Plaguey capacitors on PC mainboard.

[John Rumm]

Rubycon are generally thought to be a very good brand (assuming they are not fakes!) - having said that - any brand can (will) fail in time.

[fred]

In article , Etaoin Shrdlu writes

You don't hang about :-)

I hope it continues to hold up for you, an extended period of both memtest86 and prime95 is probably in order and it probably wouldn't do any harm to run these at a slightly elevated temperature, say in a cupboard or ventilation limited space to push the envelope a bit.

Using a pointy bit for desoldering is a useful tip, thanks for the reminder.

[Etaoin Shrdlu]

It's on a Dell board, and I wouldn't expect them to use any old rubbish. But yes, like anything, even the best can fail sometimes.

[Etaoin Shrdlu]

My sister's taken our little boy out for the evening - I have to make the most of the opportunity. Been rushing around doing all kind of things.

[Johny B Good]

I've fixed many an unstable PC by re-capping the MoBo.

When someone brings in a PC, complaining about its lack of stability (propensity to randomly reset or lock up), my first action is to take a swift gander at the CPU VRM caps surrounding the CPU socket to look for the blatant signs of dried out 'blown' caps (tops domed up and the anti explosion stress relief 'scribe lines' cracked open with or without brown staining from leakage of the electrolyte).

I don't think I've ever seen such a lack of capacitor failure evidence on an unstable PC. If anything, I've seen plenty of such failing caps on perfectly stable PCs that heve been brought in to have other issues dealt with. It's surprising just how many caps can fail in the VRM circuitry before obvious symptoms start kicking in.

It's usually the CPU VRM caps that fail ahead of the other smaller ones dotted around the board elsewhere since they might well be having to deal with ripple currents of several tens of amperes (basically what the CPU core demands from the 1voltish 60 to 80 amp VRM supply rail).

The other smaller caps only have to contend with the much lower average duty cycle ripple imposed by the current draw of the load at the end of a relatively long circuit trace feeding memory slots with oddball voltages that don't appear directly on any of the extremely Lo-Z collander ground plane supply rails.

I can only recall just the one occasion when replacing one of these smaller caps (too small to justify the stress relief 'scribe marks') near to the dimm slots on an Intel MoBo proved to be an effective cure to a boot up issue.

Normally, you pick your suspects by MK 1 eyeball on the basis that they don't have the normal appearance of flatness (or, indeed, a suggestion of concaveness). In this case, I chose my 'suspect' on the basis that it merely looked flat as opposed to its neighbours appearance of just the slightest hint of concaveness (and that was with the use of a jeweler's loupe!). I think I cured it on that first 'hit', although I may have replaced a few other suspects nearby just for good measure - I'm not sure since it was a few years ago now.

[alan_m]

It worth checking closely as various korean/chinese sourced capacitors often have branding very similar to that of the well respected manufactures.

I have have seen capacitors branded "Rubjcon" and "Rulycon" using the same font style/colour as Rubycon but with the font size much smaller to make reading the name somewhat difficult without a magnifying glass.

[tony sayer]

I've done a few in the past but its a right PITA doing them;(...

[Johny B Good]

The chances of success with this strategy are quite high so it's well worth dealing with the "Obvious Offenders" when your supply of ammunition is getting low (I've seen plenty of stable MoBos where half of their VRM caps have 'blown' their tops so there does seem to be a high level of redundency/margin for failed caps).

That's one possible cause but half the time it's down to bad dimm slot contacts rather than the ram itself - usually reseating the dimms effects a fairly permanent cure. Although it's far less common, failed decoupling caps around the dimm slots can also cause 'bad ram syndrome'. It's rare because they have a much easier electrical life and are, generally, a little more remote from the major source of heat so run cooler to the benefit of extended life.

Which is an obvious 1st line of approach to diagnosis. I'd be surprised if just replacing the 'obvious suspects' in the VRM area of the board doesn't provide an immediate cure.

The trick to success is to use a 25 to 35 W antex with a chisel blade tip (double that for the feebler 'bit inserted _into_ the heating element lump' type of soldering iron) and don't be afraid to melt a fair sized blob of multicore tin/lead solder over both capacitor holes so you can wriggle it out of the board.

A lump of beeswax to re-flux the excess solder will help it adhere back onto the soldering iron bit after you've sucked the holes dry using a de-soldering sucking tool[1]. There's never the luxury of sufficient clearance to waste time crushing the blown cap to gain better access to the 'stumps' to let you fiddle about with needle nose pliars to allow each one to be removed one at a time (there certainly won't be any room to use these stumps to tack solder the replacement onto without relying upon unfeasibly long component leads).

Once you've discounted all such thoughts in regard to those pragmatic bodges that can be valid on more sparsely populated boards, you can concentrate on getting the job done properly without having to worry about adding excess lead inductance to compromise the original design rules used in the manufacture of the board nor any worries as to the security of the solder joints.

[1] The trick of re-inserting the replacement capcitor leads into the solder filled holes whilst you use a soldering iron to re-flow the solder is extremely difficult to achieve when you're working 'blind' as is almost innevitably the case with motherboard recapping work in the rather congested VRM area. You stand a much better chance of success if you do take the trouble to clear the old solder out of the holes so you can wriggle the new cap into place.

Sweating the new cap into the solder filled holes is fine and dandy when dealing with the other caps standing alone in their own bit of motherboard real estate where you can readily observe where the wire ends are going but it's just impractical in the one location on the board where you can guarantee a replacement of an obviously blown cap or three is going to effect a cure.

[Johny B Good]

Yup, it's worth bearing in mind that "Shit Happens" and that these faults can sometimes be due to the fragile high tech silicon going bad (and even then, it might simply be a bad iinternal lead connection in the packaging of the chip rather than the chip itself but that's just a matter of academic interest when you're down to this level of faultfinding).

I have to say though that I thought I enjoyed a higher success rate with this sort of repair, perhaps somewhere in the region of 80 to 90 percent (at least in the case of Motherboard VRM re-capping - I experienced a far lower success rate trying the same repair technique on ATX PSUs so much so that I've long since given up wasting time on such repair attempts).

[Stephen]

Be aware that there are counterfeit Rubycons around. some of them are really poor counterfeits with names such as Rubicon or Rudycon....

You can even get chinese doll capacitors in the field. This is where a smaller capacitor is put inside the can of a bigger but empty capacitor and hence passed off aso something of a different spec.

see

[Etaoin Shrdlu]

Thanks for taking the time to explain this to me, JBG. This is probably the best and most active NG I subscribe to, and I knew I'd get the information (and more) that I needed here.

The PC in question has been running happily overnight, so I'm feeling quite pleased. I replaced some memory caps and some VRM caps, so I'll probably never know which was the real culprit. But it seems to work well now, so I'm not going to complain.

[John Rumm]

Mmm nice ;-)

I had seen that with batteries before, but not capacitors.