My caps LOOK ok, but my board doesn't work!

[Topcat]

This board is definite proof positive that your caps can LOOK ok in appearance, but are absolutely unusable or dead. The customer contacted me, stating that his caps looked perfect, but the system froze one day, and never POST'd again. The board is an ECS D6VAA, a common one for the plague.

When I got the board in for repair, he was right... heh... It wouldn't POST. Indeed, his board packed with G-Luxon caps looking as good as they did the day they were manufactured. Not a single one had even the slightest bit of swelling.





Before recapping, I checked the VRM's for shorts, just to make sure I wouldn't be pissing in the wind and wasting caps trying to rebuild it. VRM's were fine, so I recapped it with some nice pretty new Rubies (MBZ Series), and she's operating perfectly now!

Lesson boys and girls: They don't always have to appear bad to be bad!! I get asked this a lot: "They look ok, whats wrong with my board?" Even if they look ok, take note of the brand. We all know G-Luxons are turds out of the factory...

[Neo2_000]

That's where an ESR meter comes in handy, revealing suspect caps where there is no physical evidence.

I would be interested in learning more about testing VRM's. I assume you are referring to the 3 legged chips that usually have the middle leg cut off. I have seen these referred to as VRM's and mosfets among other names.

What are these things really, what function do they actually perform?
What are common problems you may test them for and how do you test for them?
Do you have to remove them from the circuit board to effectively test them?

I recall reading one post where there was no joy after replacing the caps on one board, so the tech also replaced these chips and the board came good.

Thanks,
Neo.

[AK0R]

VRMs (voltage regulator modules) are a specific class of MOSFETs; MOSFETS are a specific type of FETs. FETs have several advantages over bipolar transistors. For a basic description on the workings of a voltage regulator, search on "7805 data" on Google or Yahoo (7805 is a basic 5-volt regulator).

VRMs take one voltage (such as +5 VDC) and produce a different, lower voltage that is needed by various components (microprocessor, memory, etc.). Each VRM can only produce one output at a time, so two different VRMs are needed if the CPU and memory operate at different voltages. The output of the VRM can be varied slightly, depending on how external components are connected and controlled, so some motherboards are able to have settable voltages for memory. CPU voltage is usually, but not always, taken care of automatically. True variable power supplies, where the voltage is variable over a large range, aren't used in PCs. The middle leg is usually cut off, because it is connected to the case, which is soldered to the board. The case is often, but not always, connected to ground. To test one, you need to know what the input and output should be, then measure them with respect to a known ground. If the VRM produces no or low output with proper input, either it's bad, or an external controlling component is bad. If the input is bad, look for upstream damage. Usually, it's best to test a VRM in-circuit (with CPU, memory, etc. removed!) so that the output can be checked at the immediate output of the VRM, as well as at the point of connection of the powered device.

[MD Willington]

I'm looking for a good right up on testing VRM's since I get asked how to test them every so oftem, in all honesty, I have not tested a VRM in a long time so I'm kind of foggy oon the whole procedure.

MD