Hardware Insights Forum

How often do capacitor failures in PCs go unnoticed by the user??? No doubt it's a regular occurrence, but how common is it relative to the equipment becoming non-functional???

If you let bad caps go long enough, they will always cause a problem, so this is more of a question of how often you catch them in their early stages before they cause the device to fail. For me, that's quite often, maybe 75% or more of the time.

It depends on how the capacitors fail. Problematic ripple voltage can often go unnoticed for months on end by the user and if the capacitors are spec'd well enough on the motherboard the filtering there can make up for it just for a bit. However, if a critical rail experiences an overvoltage as a byproduct of SMPS output capacitor failures, it will render the power supply unusable if it even had overvoltage protection, and if not that it will fry the motherboard to which it is connected. So yeah, bad capacitors are definitely a problem, but they are not the only potential for the failure of a power supply, motherboard, etc... a shoddy design especially can result in many other failures.

Wester547 wrote:It depends on how the capacitors fail.

Yes, but the poll is in general (with the most frequent failures i.e. motherboard VRMs and PSU output filters given the highest weight).

Problematic ripple voltage can often go unnoticed for months on end by the user as if the capacitors are spec'd well enough on the motherboard the filtering there can make up for it just for a bit.

Yes, that's one thing that can happen. Another possibility is that if you have two identical caps on a PSU rail, one on each side of a pi filter, the one before the coil is put under more stress, and when it fails, the cap after the coil takes the load. It eventually fails too, given enough usage, but could indeed keep a PSU running for some time after the first cap fails with, if not in-spec ripple, at least a level that the connected hardware could function with.

Another possibility involves mixing electrolytics and polys (or potentially ceramic capacitors if they're in close enough proximity) on a VRM output. As the electrolytics fail, the polys take over.

I'm not even sure why well-known manufacturers used the two-transistor +5VSB circuit for as long as they did, as technically superior switching ICs have existed since at least 1996 (TOPSwitch).

LongRunner wrote:I'm not even sure why well-known manufacturers used the two-transistor +5VSB circuit for as long as they did, as technically superior switching ICs have existed since at least 1996 (TOPSwitch).

Yes. Delta/Newton used the two-transistor +5VSB circuit in their PSUs all the way up to 2005. Most of those OEM Dell units that are Newton do have two-transistor circuits and even worse have the critical capacitor - often it's a 47-100uF 25V or 50V LTEC TK or TH, LZG, Taicon VT, or PJ. If you're fortunate it's something like Nichicon PJ or Rubycon YXF. The original Xbox power supplies, either a Foxlink branded Delta or plain Delta, also used a two transistor +3.3VSB circuit, possibly with the critical capacitor either being a Teapo SEK or something else... even worse, often, the +5VSB/+3.3VSB transistor was not heatsinked, though it would often be a MOSFET that has the metal tab at least. Yet I've never seen discoloration on the PCB after years of +5VSB/+3.3VSB use in those units so Delta must have done something right or better than the other manufactures who defer to two-transistor circuits. I guess using a MOSFET that has an extremely low ON resistance could help but even still...

Also, Delta seem to be very good at getting capacitors like Teapo and LTEC to last a very long time, even in the +5VSB circuit assuming there are no scorching diodes or resistors present on the PCB(s), so that's probably why you didn't hear about a mass of failure rates a la Bestec.

Hipro did not... as of 2001 they moved to either a Fairchild Powerswitch or Topswitch to generate +5VSB, which was nice. My respect goes up for Hipro for that.

Another possibility involves mixing electrolytics and polys (or potentially ceramic capacitors if they're in close enough proximity) on a VRM output. As the electrolytics fail, the polys take over.

Capacitors aren't really predictable with regard to the manner in which they fail, though. Electrolytics, I mean, at least the really cheap ones - sometimes they can fail so catastrophically that they will not only preclude the board from POSTing but they will often blow MOSFETs and possibly even cause discoloration to the motherboard. It's never a good idea to put any trust to failing capacitors.

Wester547 wrote:...Foxlink branded Delta...

I thought Foxlink was a separate company. That they both had problems with fractured solder joints on the mains input could have been a coincidence. The Foxlink units did not have adequate reinforcement on the mains connector (two flimsy plastic tabs and that was it). Not exactly sure what happened with the Delta models, but inadequate soldering of the mains pins seems to be within the realm of possibility.

M$'s response to that problem was to replace the power cords with either special arc-fault interrupter equipped units (for the "affected" consoles) or otherwise a seemingly conventional version that supposedly has tighter contacts (which however would be the opposite of safety improvement as far as breaking the solder is concerned). Which is plainly ridiculous because the cord in question is a standard item, and if it was truly the cord at fault, you could fix the problem by getting a replacement from just about anywhere.

Capacitors aren't really predictable with regard to the manner in which they fail, though. Electrolytics, I mean, at least the really cheap ones - sometimes they can fail so catastrophically that they will not only preclude the board from POSTing but they will often blow MOSFETs and possibly even cause discoloration to the motherboard.

Yes, but we're talking about the usual boring failures - high ESR and low capacitance.

LongRunner wrote:

Wester547 wrote:...Foxlink branded Delta...

I thought Foxlink was a separate company.

To my knowledge, Foxlink either contract Deer or Delta/Newton to build their power supplies... clearly, the power supplies in the original Xboxes weren't Deers. They often had similiarities to Deltas (the Foxlinks), such as the type of input coils used (exactly the same coils), the type of standby circuit, and the underside had a soldering job very very similar to that of Delta's. Also, the Foxlink model numbers end with a PS (FTPS) just like Delta and Newton's units do (NPS/GPS/DPS/DPSN). I get that Lite-on also sort of shares that similarity but still.

EDIT: Okay, maybe Foxlink do build their own power supplies, the internals of other Foxlink power supplies look too drastically different from Delta units...

Wester547 wrote:To my knowledge, Foxlink either contract Deer or Delta/Newton to build their power supplies... clearly, the power supplies in the original Xboxes weren't Deers. They often had similiarities to Deltas (the Foxlinks), such as the type of input coils used (exactly the same coils), the type of standby circuit, and the underside had a soldering job very very similar to that of Delta's.

What do the transformers look like??? Do they have the Delta-style no-tape-around-the-core construction???

I'm sure even the infamous cheap-and-nasty PSU companies are capable of making quality PSUs. They're just trying to profit from people who don't know what a quality PSU is.

LongRunner wrote:

Wester547 wrote:To my knowledge, Foxlink either contract Deer or Delta/Newton to build their power supplies... clearly, the power supplies in the original Xboxes weren't Deers. They often had similiarities to Deltas (the Foxlinks), such as the type of input coils used (exactly the same coils), the type of standby circuit, and the underside had a soldering job very very similar to that of Delta's.

What do the transformers look like??? Do they have the Delta-style no-tape-around-the-core construction???

They look almost entirely different save by the "130" numbering I usually see partaking Delta transformer labels. I don't think Foxlink power supplies are made by Delta anymore, though, looking at the internals of their other power supplies yields too vastly different a design as the previous edit mentioned.

I think a computer is much more likely to keep working with bad caps when the bad caps are on the motherboard, but not the VRM. If it's still getting low ripple from the PSU they can sometimes work for quite a while. Of course they should be replaced, but people don't know or seem to care when their capacitors go bad unless it ceases use of the computer completely. Although we had one in the shop with a StupidPower, 90% of the Fuhjyyu caps were out of spec (Even non bulging ones) but the computer was working surprisingly normal. It had about 50/50 of Rubycon and nichicon on the motherboard, but they must have been getting stressed pretty hard.