Rubycon MCZ and Chemicon KZG mass failures

[mockingbird]

Re: Rubycon MCZ and Chemicon KZG mass failures

Chemicon *KZJ* is perfectly fine.

[Pentium4]

Re: Rubycon MCZ and Chemicon KZG mass failures

Agreed, I have seen lots of KZJ failures in PSU's, motherboards, and video cards. Mostly motherboards though. They seem to be better than KZG at least

[c_hegge]

Re: Rubycon MCZ and Chemicon KZG mass failures

^
Agreed. KZJ are just as bad as KZG. I've seen lots and lots of them fail in cool and open environments. Ruby MBZ/MCZ usually only fail in hot SFF Roasting oven PCs (with the exception of a few bad batches).

Unfortunately polymers cannot be used to replace huge capacitors at 1500uF and beyond (it would be like trying to recap your PSU with polymers)-- the whole circuit would have to be reworked to use smaller uF values in series to make up for the huge 1500uF.

Disagreed. On motherboards, you can just drop the capacitance back to 820 or 1000uF when polymodding. The lower ESR of the polies means that they can filer more ripple at a lower capacitance than what electrolytic caps can. I've done it several times before.

[LLLlllou]

Re: Rubycon MCZ and Chemicon KZG mass failures


What he said.

KZJ bad, KZJ go boom!

KZG usually just vent or dry out with no visible signs.

[Sparkey55]

Re: Rubycon MCZ and Chemicon KZG mass failures

I have witnessed KZJ and KZG bloat and vent tops from just sitting in storage with no voltage applied.

[theokretes]

Re: Rubycon MCZ and Chemicon KZG mass failures

A bit of personal background ... I've worked in P/Ss since 1980, except for some 4 years at a motor control company (which is pretty similar in many respects). Ultra-low impedance caps such as Rubycon MBX/MCZ, Nichicon HM/HN or UCC KZG/KZJ weren't what I worked with - 2000 hours rated life (at max rated ripple current and temperature) wasn't appropriate for products expected to last 10 or 20 years (the motor controllers). But Nichicon HE, Panasonic FM, Rubycon ZL and UCC KZE are of that electrolyte type (not quite as low impedance) and are rated for 5000-10000 hours life. So I compared a part in Rubycon's ZL and RX30 series, 3300uF, 10V (the lowest voltage rating available in RX30):

ZL: 12.5mm x 25mm; 18 m-ohm impedance; 2.77A ripple current; 5000 hours;

RX30: 16mm x 25mm; 34 m-ohm impedance; 1.90A ripple current; 4000 hours.

Here is the corresponding data for Nichicon's PW (7000hours) and PA (5000 hours) series parts:

PW: 12.5mm x 25mm; 30 m-ohm impedance; 1.95A ripple current;

PA: 12.5mm x 25mm; 22 m-ohm impedance; 2.28A ripple current

Other than being rated for 130C rather than 105C, Rubycon's RX30 is more similar to Nichicon's low impedance PW series than Rubycon's almost ultra-low impedance ZL series (water-based electrolyte) or Nichicon's almost ultra-low impedance PA series (non-water-based electrolyte).

All that said, the RX30s might do very well for a long time in your mobo if the mobo design has enough margin in ripple tolerance, in its regulator compensation, and in the ripple current to which the caps are subjected (or if the user isn't stressing that area of the mobo to the max). But if is one of the biggest words in the English language.

Yes, the IBM boards have great ripple tolerance-- intellistations are always competent systems: one nice thing is that the power supplies use huge KY filters, and YXG for the smaller caps (*gasp* a real power supply!) Anything that uses TEAPOs should be promptly discarded. Not acceptable.

But in all honesty, the fact that the KZGs were there means RX30s would be an upgrade to the circuit (a poor capacitor series that is practically already dead won't offer any ripple tolerance)... also they are of considerable higher voltage.
The systemboard was running fine until I took it out again for some other changes, but that's my fault.

I have witnessed KZJ and KZG bloat and vent tops from just sitting in storage with no voltage applied.

Yeah KZJ is bad (I was thinking of KZE).

[Heihachi_73]

Re: Rubycon MCZ and Chemicon KZG mass failures

I found a small form factor IBM P4 on the side of the road the other day, and the mainboard and PSU were both thick with dust.

All caps on the mainboard were were Rubycon (MBZ, CE and YK), with the rest being Chemi-con KZGs and polys.

A total of twelve caps had failed on the board - notably, a handful of KZGs, a few tiny Rubycon CEs (with the small [R] logo) which had between 5 and 12 ohms ESR (only one of these was still good, with 1.44 ohms ESR), and three failed YKs, one of which was open circuit. In the mix were three Rubycon MBZs and eight Chemi-con polys, all of which were still good.

Unfortunately I never took any photos of the IBM board and it went in the bin after the caps and useful bits were pulled - in fact, I don't even remember what model the PC was aside from it being a Pentium 4 from 2003 (going by the PSU it may have been an S50). It has a laptop style floppy drive, with a ribbon cable going directly to the board in place of the regular floppy cable, which the board also had.

I never even bothered to power it up with the dead KZGs in there - the hard disk and CD drives were already removed when I found it as well. The CPU was also passively cooled, just to make it even more hotter in there I gather (just perfect for KZGs and general purpose, 85°C caps).

I haven't opened up or tested the PSU yet but there is a crapload more dust in there (surprisingly, the fan isn't seized). The PSU is a 200W AcBel API2PC23, and might make it to the PSU pictorial thread if I get enough time.

[Wester547]

Re: Rubycon MCZ and Chemicon KZG mass failures

In my cap testing (some 5 years ago, now!) I found the water-based electrolytes seemed to expand more than did conventional electrolytes. Among series with water-based electrolytes, some parts did better in my torture test than others (e.g. HE did better than HD, ZL did better than YXG, etc.). I don't have any explanation to offer, just the empirical observation.

I hope you don't mind that I bring such an old post up but I couldn't help but notice:

In my very abusive testing, HE and ZL are approximately peers, and KZE is close. YXG didn't fare as well. Because I have a hard time getting Rubycon parts (their sales/marketing in the US is execrable, and has been since at least the late 80s), I haven't tested as many of their parts. YXG did do better than PJ, PL, PW, and LXV, however.

Did YXG really do better than PW in your torture testing? I recall you stating about a year ago in the HM/HN discontinuation thread that PW/LXZ/FC did well and YXG did poorly, unless those were separate tests. I only find it hard to believe because PW isn't really aqueous in comparison to YXG (though I recollect that YXG is pretty low on the aqueous spectrum) and the kind of solvent PW uses is much more resistive to heat, and dries up much slower (though I suppose the other additives used in the electrolyte might have something to do with it as well).

Also, those SFF cases don't cool very well, so I'm not surprised there were failed KZGs, MCZs, and 85*C capacitors in the unit Heihachi_73 found, especially with a passively cooled CPU...