Failed MCZ rubycons

[UraBahn]

Re: Failed MCZ rubycons

How about Elna RJF then? Those seem to be in the ballpark of MBZ judging by a cursory glance at the data sheet.

[kc8adu]

Re: Failed MCZ rubycons

several of us everyday recappers have noted failures of mbz in ibm sff boxes.since the lower the esr the more they suffer from heat i am not surprised they gave up.
found some mbz a while back that hadnt hatched but were alongside ones that had.looked fine but esr was several ohms!again in ibm sff boxes with socket 478 and 775 hotplates inside.

[mockingbird]

Re: Failed MCZ rubycons

I have to say, I don't like MCZ not bloating... I don't mind that they have only a few years lifespan - that's to be expected, but I'd rather caps did bloat so I can tell that they need to be replaced. Are there any reports of MBZ having the same issue? I think Nichicon HN might be more reliable than MCZ...

Elna RJH, Sanyo/Suncon(?) WG/WX and post-2005 Nichicon HM/HN should be good candidates? Also maybe worth considering polymer caps too.

Elna RJH is not suitable at all for this kind of application.
Sanyo has no electrolytic equivalent to MCZ. WG is an equivalent to MBZ/HM
Nichicon HN is a good replacement... They will no longer be in production shortly.

[EIBM]

Re: Failed MCZ rubycons

Are these machines running 24x7x52? And are MCZs rated for 2,000 hours at maximum rated ripple current and 105C? If so, it may be that at the actual ripple current and actual ambient temperature, these 6- and 7-year-old parts have worn out. Running 24x7x52 is about 9,600 hours, IIRC, which means that in 6 years, those parts will have been operating for nearly 58,000 hours. Granted, a 2,000 hour part may last longer than that at maximum rated ripple current and temperature, but these parts have got a lot of hours on them.

You won't like this suggestion, but if those machines are expected to last several years longer, you might consider replacing all the lytic caps. One of my previous employers produced computers that were expected to last quite a few years. Two types of parts were considered scheduled maintenance items, every 50,000 hours I believe it was: fans and power supply output caps.

Six years for a rubycon? The rubycon capacitors in my 5150's PSU are thirty years old and they're fine. I quite like those "baby blue" ones. And, I have working marcons and sanyos which are over 34 years old.

Granted I understand what you mean, products today are manufactured to be rated for a specific lifetime. But the question remains why MCZs are getting this notoriety for failing when other ones aren't (KZG excluded, they are awful caps to begin with)...

Some of these thinkcentres were subject to 24/7 use (but please note, NONE of the MBZ went, only KZG/KZJ and the MCZ!). My Intellistation however, as noted by the "healthy" KZGs, is taken care of exceptionally well.

Also these computers do *not* have faulty and/or bad PSUs. They are not the cause of the issue. I opened them up and was quite impressed by the components.
Typically a bad PSU will pop more than just a specific series of capacitor!
ALSO worthy to note, the older northwood thinkcentres... *all* of the MBZ caps are fine, but the KZGs have long since bursted. And they use both manufacturers for the same ratings interchangeably. Ergo, if the PSUs were supplying a ripple to specific capacitors, it should affect both the MBZ and KZG, not just the KZG. Which I can also conclude, the MCZ have had some sort of bad formula.

EDIT: The fans are Nidec Beta V, so they'll last longer than the junk in typical computers.
Granted, my pentium 1 had an abused nidec, so I replaced it with a noctua. Also redid the CPU with arctic silver thermal paste. That's always good to change after 15 years.

[UraBahn]

Re: Failed MCZ rubycons

Actually I would recommend replacing all the MCZ even if they appeared fine. (My T0604-datecoded MCZ, which looked fine on the surface, bubbled when I cut it...) Brand new caps would be relatively cheap insurance. Same with the KZG too.

Elna RJH, Sanyo/Suncon(?) WG/WX and post-2005 Nichicon HM/HN should be good candidates? Also maybe worth considering polymer caps too.

I don't think EIBM runs his boards 24x7, and neither does he run them in overly hot or cramped conditions (quite the opposite actually).

[PeteS in CA]

Re: Failed MCZ rubycons

UB, I grew up north of the Flier's Club and south of Steven's Bridge. I attended WHS at both the College Street campus and the West Street campus. To me it's the "new" campus, but, well, you might know when it was built.

[PeteS in CA]

Re: Failed MCZ rubycons

Are these machines running 24x7x52? And are MCZs rated for 2,000 hours at maximum rated ripple current and 105C? If so, it may be that at the actual ripple current and actual ambient temperature, these 6- and 7-year-old parts have worn out. Running 24x7x52 is about 9,600 hours, IIRC, which means that in 6 years, those parts will have been operating for nearly 58,000 hours. Granted, a 2,000 hour part may last longer than that at maximum rated ripple current and temperature, but these parts have got a lot of hours on them.

You won't like this suggestion, but if those machines are expected to last several years longer, you might consider replacing all the lytic caps. One of my previous employers produced computers that were expected to last quite a few years. Two types of parts were considered scheduled maintenance items, every 50,000 hours I believe it was: fans and power supply output caps.

[UraBahn]

Re: Failed MCZ rubycons

If you don't mind me asking, a good batch does of course mean post-2005 date codes (since I imagine everything from 2001-2004 must be bad ^^; ), right? But I also wonder if MBZ and MCZ failures stem from PSU problems. Also, I meant KZG/KZJ, not KZE. ^^; Even if the PSU had good capacitors, something else could be wrong, though it's hard to say what.

It is understood that post-2005 Nichi HM/HN are OK. Yes, it is possible that something like ripply power could have killed the MCZ, but it's curious how it'd only affect a certain farad range and spare the differently-rated caps.

EIBM, some date codes on those dead MCZ would be appreciated. Perhaps we can isolate which run(s) could be potentially faulty?

[UraBahn]

Re: Failed MCZ rubycons

UB, maybe you answered me before, but I missed it. Are you in Yolo County?

Why yes sir. I live in Woodland to be exact.

[Wester547]

Re: Failed MCZ rubycons

The rest of the caps are nichicon HM(M), but they obviously must have been from a good batch.

If you don't mind me asking, a good batch does of course mean post-2005 date codes (since I imagine everything from 2001-2004 must be bad ^^; ), right? But I also wonder if MBZ and MCZ failures stem from PSU problems. Also, I meant KZG/KZJ, not KZE. ^^; Even if the PSU had good capacitors, something else could be wrong, though it's hard to say what.

[EIBM]

Re: Failed MCZ rubycons

Okay here are the images.

In the first '06 thinkcentre, you'll notice the 1500uF MCZs have blown (it's more difficult to see the burnt electrolyte on the photo). The ones circled in blue show no physical signs of bursting. I've observed a multitude of these same machines with the same MCZs, and ALL of them exhibit exploding 10v @ 1500uF MCZs.


In this '05 thinkcentre, a whole load of the KZGs are slowly dying... being a total of *28*, it'll be a pain to replace.


Finally, in the 2005 intellistation, that one big KZG has shown signs of popping. The other KZGs are fine as I keep this thing cooled *very* well. The rest of the caps are nichicon HM(M), but they obviously must have been from a good batch. Oh, and a handful of oscons.


--> so the conclusion. KZGs seem to pop intermittently, and MCZs seem to be affected by a certain farad range?
Regarding MBZs, all five of those IBM machines contain MBZs and they are PERFECT. I do not have to replace them.
Later *Lenovo* quality thinkcentres may have bad PSUs (I've seen some with bestec): and I did see many Lenovo-made computers with low quality components, including OST and G-Luxon capacitors.

All of these machines in question are pre-2007.

[severach]

Re: Failed MCZ rubycons

I've seen too many MBZ and MCZ fail where other brands don't so I don't consider MBZ and MCZ to be any better than the cheapies. The problem with bad caps isn't that they can't last. Some cheapies last a long time. The problem is that they don't all last. You can't tell from the label whether a KZG, MBZ, or MCZ will last or fail early.

With a Sanyo, Panasonic, Nichicon, and other series from UCC and Rubycon you can tell.

[PeteS in CA]

Re: Failed MCZ rubycons

UB, maybe you answered me before, but I missed it. Are you in Yolo County?

[UraBahn]

Re: Failed MCZ rubycons

While we're on the subject of Rubycon...

Here's an obvious fake, "Rukycon"....
http://www.ebay.ca/itm/1000-Rukycon-...item415c9264b9

[UraBahn]

Re: Failed MCZ rubycons

Failed MBZ, huh...

What PSU was used on that ThinkCentre? Also, was it an IBM or Lenovo one?

[c_hegge]

Re: Failed MCZ rubycons

Topcat had some MBZs which failed in well ventilated thinkcentres recently. '07 datecodes IIRC.

[UraBahn]

Re: Failed MCZ rubycons

All that said, I still believe MCZ are just a touch thermal sensitive compared to MBZ and Rubycon's other series. Again, not nearly to the same extent as KZG/KZJ/TMZ/TMJ are.

Maybe we should record the date codes of failed Ruby MCZs? Perhaps it was a one-off, slightly faulty batch that was produced within a specific time frame. (Slightly faulty in the sense that it was a little over-sensitive to heat/ripple compared to normal.) For my two cents, the suspected ones I cut/removed were T0604.

[EIBM]

Re: Failed MCZ rubycons

As they're 10v rated, I would guess the 1500uF ones are probably on the 5v coming from the power supply, while the 6.3v 1800uF are on the VRM side (chosen perhaps to keep the esr low)

Are you sure it's not a problem of excessive ripple from the power supply mixed with heat from the VRMs? The 1800uF may be close enough to the heatsink to be cooled a bit by the cpu fan.

How about some pictures of the boards?

Yeah I'll get some pictures of the systemboards, as well as outline the caps that didn't fail (which are nestled in the same area).
It has nothing to do with the PSUs. These are IBM FRUs (either made by delta or acbel), so they're engineered to a higher standard.
The caps inside these PSUs are either chemicon or nichicon (I opened them up and checked)... sometimes even rubycon! An awesome acbel I have uses rubycons.

Heat isn't an issue, and if it is, they're way too sensitive (as the caps nearby weren't affected except specific ones: which I outlined above).

I have a thermo-nuclear SFF P4 thinkcentre reaching 120C, and the MBZs (and even the KZGs!) were fine. I'm keeping the cooling stable (now), I just wanted to see how hot it could get. The heatsink did cause some burn marks to nearby surfaces.

[UraBahn]

Re: Failed MCZ rubycons

I have no plans on getting this board working again, I bought it just for the caps. :P

I wouldn't doubt, however, that the inductor coils would get searing hot.

[mariushm]

Re: Failed MCZ rubycons

Nice. After you get the board working you should also measure the temperature of those inductor coils.

I wouldn't be surprised to see those coils hovering at 40-60c

And the caps behind that heatsink... I guess it depends on the case but it could be there's little airflow back there.