Bad caps and stable system (or vice versa)?

[Tom41]

I suppose you're wondering... why is it that sometimes a single bad cap can cause reboots and other symptoms, and other times a system is still stable when all the caps are bad? The answer to this lies in the PSU.

The random crashing and rebooting is a result of the CPU and/or other components on the board getting the wrong voltage, or too much 'ripple current'. The caps and MOSFETs on the motherboard are designed to further regulate the supply from the PSU so that the CPU gets as stable a voltage as possible.

If all the caps on the motherboard vent, but the PSU is still outputting a very stable supply, the blown caps will still be able to regulate the voltage (though the tolerance will decrease). However, if the voltage and ripple remain within tolerance, the system remains stable.

On the other hand, if a cap in the PSU blows, the output voltage and ripple will be severely affected. Even good caps on the motherboard will not be able to keep the voltage and ripple within tolerance, and your computer will start showing bad cap symptoms. In the worst cases, the severe ripple and over-voltage from a bad PSU can cause your motherboard caps to deteriorate and eventually bulge.

If your computer starts showing bad cap symptoms and the caps on the board look fine, check the PSU straight away! The longer you leave it with a bad PSU, the more likely the motherboard caps are to blow.

If you have anything to correct, or want to add information, please feel free to do so in this topic.

[linuxguru]

Slight correction: Blown caps on the VRM O/P will cascade and eventually cause a VRM failure and/or blown CPU, regardless of the quality of the PSU. That's why it's imperative to recap a motherboard immediately after just a single VRM cap has been observed to vent (actually, even before anything vents). A good PSU will not change the outcome in these cases, but a bad PSU can cause it to fail even quicker.

[starfury1]

I actually posed this question in another thread and said
"could a quality supply mask the effects of a failing VRM output"

At first I was thinking to a greater extent yes
but when you look at the design of the VRM, not as much as you might think.

The design of the vrm is to delivery huge currents at very short notice (with minimal opposition to flow) add to this it has to keep the voltage very tightly regulated so its quite a feat and a very demanding design criteria
(there is a "droop" voltage, voltage wise designed into to it. So the voltage doesn't overshoot)

The input side is separated by the regulator circuit from the output side so they are some what independent of each other.

So my thought is to a lesser extent yes.

As linuxguru said....once on fails the domino effect will kick in and the next weakest link most probably the one with the highest ESR will go then so on at increase rate of speed, probably exponentially.

I guess the VRM design and the demands of the CPU (plus tolerances of it) would factor in too....I am still amazed given the design criteria how well some CPU's keep chugging along with all the VRM output caps sitting there with blown tops.....

So guess what the design is and how it will behave with a given CPU may be more the case of what will happen as it fails

Granted, a crap supply which cant keep the input side running smoothly is probably going to cause much more erratic behavior and stress on the output side of the VRM then a quality one which keeps a more pure constant DC.

so yeah a quality supply will compensate a bit for the failing VRM output but not to the extent I thought it might

Anyway thats my thoughts on it but I don't design or even do in depth circuit analysis on them.... so this is all purely speculative from my side.

as above comments welcome

Cheers

[tmiha71]

I suppose you're wondering... why is it that sometimes a single bad cap can cause reboots and other symptoms, and other times a system is still stable when all the caps are bad? The answer to this lies in the PSU.

Wrong.
We can argue how much it is wrong, but at the end of the day - it's wrong.
I do not say that PSU does not play role in this, but not so big role (maybe can patch something sometimes).

The answer is: depends how GOOD CPU-vrm was designed...
Few notes to support this thesis:
- unknown is - REAL CPU demand
- unknown is SAFETY MARGIN (redundancy done on mobo)...
- requested support for wide line of CPU products

Funny thing (from data-sheets) for CPU-vrm design - "ripple current" is defined at free-will (seen few times, approximately 3App). Should be specified by CPU-manufacturers.

If it (ripple current) is specified at too low level -> CAPACITORS will supply current (will patch whole vrm circuit, until the needed current flows through inductor ) -> wrong "per se" (failure at the design level) -> and capacitors will be doomed (and with this "type" of VRM, capacitance is important factor, and can not be lowered ) because vrm as circuit can not supply enough current in requested time (capacitors do that). Sensitive to failure of single capacitor.

If it (ripple current) is specified at requested (or higher) level, capacitors do serve as supporting element (to hold output voltage, and do minimal work ), and are REAL PART of circuit, and ALL current (almost) always flows through inductor. This is proper way of designing vrm-s. This type of vrm needs - high accuracy, high speed switching elements, and probably some other things. As a result this type of vrm is less sensitive to capacitance failure (requested current goes through inductor), and in some cases can work with almost all capacitors doomed (pretty degraded)...

<EXPERIMENT>
As experiment I can suggest (execute it on YOUR OWN RESPONSIBILITY), taking old mobo, which posts, and changing period of CPU vrm, increasing it cca 10% (it's needed only to change single resistor according to (in data sheets) presented table/graph), ripple current will increase also ~ 10% -> mobo will work better, and higher working frequency (overclock) should be achieved. And probably longer life of capacitors.
Notice: 10% is not much (nothing can be damaged), BUT, if some component is already at its edge of performance, mobo WILL BLOW, CPU probably, PSU most probably. So mobo MUST BE IN CLOSED CASE. And do not do that with CPU with lowest consumption.
</EXPERIMENT>

Additional note: All this hype with Prescott’s/Pentium D/etc.. is probably due to insufficient ripple current of vrm (too slow vrm), on “older” boards designed for lower/slower demands of power (cooks/eats mobos).

Additional note: Increasing capacitance is good thing per se. But, putting capacitors with higher capacitance in vrm-out, in real life, leads to shorter life span of capacitor (they do work harder).

Additional note: Lowering ESR is good thing per se. But, putting capacitors with only lower ESR in vrm-out, also leads to shorter life span of capacitor (they do work harder). Capacitance should go down too... (to achieve same life span).

Note to Notes: That's what theory says. Practice may be different (due to unknown stuff)...

Question: If mobo fries capacitors -> they do work to hard -> ???
Possible answers:
1) Too big capacitance ?
2) Too low ESR ?
3) Both ?
4) Nothing above ?

[Elysarian]

Originally Posted by tmiha71

Question: If mobo fries capacitors -> they do work to hard -> ???
Possible answers:
1) Too big capacitance ?
2) Too low ESR ?
3) Both ?
4) Nothing above ?

I feel there is one important factor missing from your list of possible reasons there:
heat

I would suggest that "too high" capacitance shouldn't really be a factor (if anything the opposite would be an issue as too small a capacitance would lead to more charge/discharge cycles as the cap tries to keep up with the current demands?

[alphaa10]

Originally Posted by Tom41

I suppose you're wondering... why is it that sometimes a single bad cap can cause reboots and other symptoms, and other times a system is still stable when all the caps are bad? The answer to this lies in the PSU.

(clip)

If your computer starts showing bad cap symptoms and the caps on the board look fine, check the PSU straight away! The longer you leave it with a bad PSU, the more likely the motherboard caps are to blow.

----------

I am extremely grateful for this post-- it exactly addresses my concern about a DELL Optiplex 320 whose mainboard caps began to leak at some point. My original question had been, "How long do I have before the crashes start?" But now I understand the PSU probably has been shouldering the burden of maintaining system stability, so immediate attention to the mainboard caps is the next step.

[mikail13]

My g31m3 works properly but sometimes it don't see usb Hdd and rams were always overheated.
today i desolder 7 OST caps near the cpu.
there is no visible faulty i.e . blown ,eruption
they must be 680 mF but all of are 54 mF.

there are 1000 mF TK caps below the CPU and near the ram slot.
also there is no visible faulty .measure on board and i realise something abnormal for another new board. onboard 0.145 mF..must be 3500 or 5000mF

Desolder and measure them. they have over value. one is 2400 mF other 2800mF.
default is 1000mF.

[eccerr0r]

I have a Abit KT7-??? (IIRC) that has virtually all of its caps leaking - all the big caps are bulging and have brown spots on them. Yet this board is still perfectly stable, more stable than some other boards I've owned - at least for now.

I don't think I'll recap this board however...

[Wester547]

Originally Posted by mikail13

My g31m3 works properly but sometimes it don't see usb Hdd and rams were always overheated.
today i desolder 7 OST caps near the cpu.
there is no visible faulty i.e . blown ,eruption
they must be 680 mF but all of are 54 mF.

OST RLA capacitors which have gone open. VERY VERY bad capacitors (resleeved TK ATWBs anyone? The datasheets match and I noticed the vent designs are very similar) and easily OST's worst series. Worse than KZJ. I can't recall seeing a board with these capacitors on which there wasn't a failed OST RLA, blown or not. Not to say any OST is good though.

there are 1000 mF TK caps below the CPU and near the ram slot.
also there is no visible faulty .measure on board and i realise something abnormal for another new board. onboard 0.145 mF..must be 3500 or 5000mF

Desolder and measure them. they have over value. one is 2400 mF other 2800mF.
default is 1000mF.

Sounds like the oxide film on the anode foil has totally eroded with the aluminum essentially unprotected, victim to the irate might of unstable electrolyte with high concentrations of H2O.

I don't think I'll recap this board however...

If it hasn't stopped working yet, it will eventually, and will probably kill several of the MOSFETs in the VRM output. Their duty cycle increases dramatically as the ESR of the capacitors goes through the roof, causing them (the FETs) to fail prematurely. Seldom is there the occurrence of "bad" capacitors not "bad" - the scenario in which capacitors that have bloated still measure in spec, but even then, as they spill electrolyte onto the board, the highly acidic and conductive electrolyte tends to corrode and eat traces for one thing...

To answer the thread more directly, another issue is open capacitors vs. shorted capacitors. Capacitors which develop an internal short (weak points in the dielectric causing an arc of sorts) will certainly "trip" critical circuits, preventing the board from operating. With open capacitors, it's a slower process, as covered above. However, if there are several other "good" capacitors in parallel (preferably polymers), then it might not effect functionality that much, but replacement is still apprised as they can still cause the output voltages to go out of whack (another way that bad caps kill MOSFETs). Bad capacitors can also cause inductors to overheat and fail, but by themselves (the failed capacitors, if open circuit) may not be dissipating as much heat as you'd think because the means in which to do so are nil (the electrolyte is essentially the conductor of heat and no electrolyte or dried electrolyte means no way in which to conduct current and therein heat).

And even if the board does work, its speed will certainly suffer as ICs will become less responsive when so much noise is present in the output. It is also true that some boards are designed better than others and will remain more stable than others with failed electrolytics. And I think a PSU with a very clean and stable output would help the VRM input capacitors more than anything.

[mikail13]

here is another story of msi g31m3..

it works properly.

i measure caps near the rem slot onboard
which are linked parallel.

all of them 0.145 mF.. TK

They must be app. 5000mF..

i've changed Four of Five 2 months ago. all of TK.

because of one is looks good.
but apparently something wrong with this cap.

desolder and measure it.

it is 3500 mF. Must be 1000 mF.

After the change all caps value becomes 4900-5000 mF on board.

Why do overated caps reduce the value? why are they looks good?

is they harm to system?

[eccerr0r]

Originally Posted by Wester547

If it hasn't stopped working yet, it will eventually, and will probably kill several of the MOSFETs in the VRM output

That is, if it's powered up I don't know if I'll ever use this KT7 and 900MHz Athlon board that's probably on the order of 1/20th the speed of my i7...