Bad Rubycons...

[Super Nade]

A board arrived yesterday with a solitary bad cap. It was a Dell, S478 board. To my surprise, it was a Rubycon MCZ that had failed.






So I fired up the soldering Iron and replaced it with a 1500uF, 16V UCC- LXZ. I didn't have a 6.3V in hand so I just had to make do with what was available. I'm not sure why the Rubycon failed? Possibilities include random chance, it being under-rated i.e needed a 16V instead of a 6.3V or just a bad batch.

I exclusively use UCC/NCC in all my recap jobs (PSU's, MB's other devices..etc) and they have never failed me till date.





The astute reader would observe the snake like trace on the PCB. I suspect it has something to do with minimizing EMI along that trace. Never seen anything like that before. Have a snake like pattern is not generally good due to the additional stray capacitance it introduces, but I suppose that does not matter here...

[gonzo0815]

Re: Bad Rubycons...

May be this snaky traces are necessary, to keep signaling time in check wit the other lines.
Have seen them often since i owned my first computer.

[gonzo0815]

Re: Bad Rubycons...

Regarding the MCZ i think it is a hot area, due too the linear voltage regulator heating up the cap.
It is well known that high water low endurance caps are not suited very well for such applications.
Most Asrock boards are famous for using linear voltage regulators vor Vmem & Vagp and indeed have this problem.
On my Asrock DS II they alleviated this problem by using three not so low ESR caps at the same place instead of using just one.

[Phenos]

Re: Bad Rubycons...

i've heared they use snake patterns to give the traces equal length with the other traces because the memory is paralel. It has something to do with latency sync i think. same as above post.

[Super Nade]

Re: Bad Rubycons...

Gonzo, why didn't the other three Caps on the opposite side blow up then? Also, wouldn't the RAM stick shield the cap from the heat?

[Per Hansson]

Re: Bad Rubycons...

All memory controllers have "snake" traces

This is because they are parallell interfaces, they are incredibly difficult to layout and the timing is paramount, all 168 traces times as many memory slots you have must have equal lengths else the board will have memory problems

This is the sole reason mainboard layout is very difficult. Most of these designs are only made by Intel or AMD, the third party contractor just use their design (with a few rare exceptions)

On the flip side the serial interfaces being introduced have none of these problems, and most of the time can cram more data through only 1/10 of the ammount of traces!

[Super Nade]

Re: Bad Rubycons...

Good info fellas! Thanks!

Now, the VRM seems a bit far from the Rubycon. Any other ideas on why it failed? I hope the problem is not intrinsic to the board and hopefully the UCC replacement stays alive longer than the poor Rubycon.

[severach]

Re: Bad Rubycons...

That's a GX270 SFF board. Rubycons there fail quite often. Panasonics fail much less often which makes Panasonic outfitted boards more desirable.

If it was under voltage it would have blown immediately.

Every large Rubycon cap around the memory slots, processor, and heat sink are failure prone.

Perhaps you're looking at the wrong VRM. Heat damage always surrounds the 13N03LA pair, Q1067 and Q22 which is visible in your picture. Q1067 barely visible behind the right memory slot is nearly touching your failed cap.

My GX270 board has two short Rubycon 1500's at the other end of the memory slot.

[gonzo0815]

Re: Bad Rubycons...

From your pic`s i see that the MOSFET between the Intel chip and the AGP? port is very close to the bad MCZ cap.
Furthermore it is clearly visible, that a massive coper layer is leading from the MOSFET to the bad cap too.
And there is my assumption, that probably several other massive coper layers are buried within the epoxied.

A linear converter does produce much more heat, then e.g. a MOSFET in a switching buck converter, i think this is obvious.
The MOSFEt is probably providing the 1,5v to the AGP port or it may supply the memory.
I have seen this design on countless low end boards from Asrock, ECS and probably from any other budget board producer.
And at least for the Asrock boards, most are developing problems with caps near those circuits after several years.

Combining the fact, that the MCZ series is a 2000h series @105°C i think the failure is inevitable, if my conclusions are correct and the cap was running hot.

From the pic`s provided, i can`t say much above the other caps, but i think it should not that hard to check the temperature either with your finger or may be with a thermometer may be after running some 3d apps with a beefy (power consumption wise) GFX card.
This could either support my theory or not.

Personally i would not assume, that any other cap, from a comparable series would last much longer then a Rubycon.
The only comparable series from Panasonic would be the FJ ones, which has yet to proof better at such applications, then those already widely used.

[severach]

Re: Bad Rubycons...

Panasonic FL is what Dell is using.

[linuxguru]

Re: Bad Rubycons...

Aqueous electrolytics in high-temperature applications are bound to fail eventually. If the application doesn't call for ultra-low ESR or high ripple current, a non-aqueous cap like the Panasonic FC, Nichicon PW or NCC LXZ is a more durable alternative in hot-running SFF motherboards.

[Computer-Tech]

Re: Bad Rubycons...

Super Nade,
The Capacitor (C259) is the filter cap for the board video chip set.
..
You should also replace C257 and C258 as well. If one of these caps is bad, then the other two are most likely bad (or will fail shortly).
..
By the way you also need to replace C392 and C1194 (on the other side of the Memory/DIM slots). These should be 1,800 uF 6.3V 105 degree Caps with a dimension of 8mm x 20mm and 3.0mm lead spacing.
..
The other high failure caps are the 1,800 uF 6.3V caps located between the CPU and Regulator heatsink. There are 9 each (easy to figure out which ones).
...
There are also three 1,800 uF 16V 105 degree Caps hidden behind the Regulator heat sink. They have dimensions of 10mm wide x 20mm high and a 4 mm lead spacing. These are the Voltage Regulator Supply Filter Caps. These are the ones that give you the dual solid amber indicator lamps during power on self test.
..
The only other problem caps are the three 820 uF 6.3 V caps, and the 680 uF 10V caps. these are also 105 degree caps. They should be replaced if visibly bulging or visibly damaged. If not leave them alone. NOTE: If a board has all of the 1,500 uF and 1,800 uF caps bad and visibly distorted and ruptured - I would replace the 820 uF and 680 uF caps as assurance that the board will not come back.
..
God luck
..
Dave Mason
Computer-Tech@phoenixcomputerlabs.com
...