mpe wrote on 2021-04-20, 21:24:

(still want to hear opinions about the cap)

I replace a lot a lot of caps and do all kinds of a repair work on a regular basis. I have gone through multiple TVs, tape decks, computers, monitors, you name it.

Please trust me when I say that you absolutely cannot tell if a cap is good or bad from a picture or by looking at it. Most bad caps will look just like a good cap, but they will have dried out and gone bad.

The only way to tell is to remove them from circuit (at least one leg) and test them.

The small component tester on ebay that is super cheap is a good tool to find bad caps. They cost between $5-10; mine was $7 shipped.

The next problem is removing the caps. You can either use solder braid (harder but cheap) or a desoldering gun (expensive-ish at $250, but super easy)

Such capacitors with inconsistent packaging were not uncommon in the 1980s but since the 1990s reputable manufacturers managed to package them with better tolerances. Don't think they are caused by bulging

I believe it's necessary to remove the capacitor from the circuit if you want to measure its ESR, and at that point you could just as well replace it with a new one. I would not trust a cap branded as "OIL" anyway, 🤣.

gdjacobs wrote on 2021-04-20, 20:27:

The bottom line is that visual inspection is only a reliable way of finding caps which have definitely failed. Caps that look good won't necessarily function properly. You also can't reliably test caps in circuit, except for those that have failed short which happens more often with ceramic caps than wet electrolytics.

Couldn't agree more, in-circuit measurements are inaccurate at best. So you either start removing and measuring elcos one by one or trust the readings you get in-circuit. If I had to choose a method (strictly talking about XT-S3-era boards) then I'd choose to replace them all. The usual 10/33/47/470 uF elcos are extremely cheap.
Onto the topic of the correct tools: yes, megatron-uk is spot-on and people need a hot air station if they want to start fiddling with PC components. Doing it with an iron is possible but it's going to be ugly and slow and the risk of damaging other components is high. It's like changing a tyre with a monkey wrench.

It's a slippery slope though. If I replace all elcos on a 25 years old board then I should replace them on the expansion cards as well. But why stop there? PSUs are also filled with elcos, even floppy and optical drives have them. A standard 486 has like 40 caps altogether that need replacing. Quite a lot even with using the proper tools. Sometimes the best method is not doing anything at all. If it ain't broke, don't fix it.
Except for old PSUs. Those should be fully recapped even if they work fine.

Just ordered this device:

The device advertises in-circuit capacitance/ECR testing (given by using 100 kHz pulses and low voltage). Will see how it goes! In worst case I can always use it to test desoldered parts.

I am only concerned that I read somewhere that some motherboards deliberately used high ESR caps as the resistance was a part of the electronic design.

I do own a vacuum desoldering gun - so no problem here. It works very well. Much better than solder wick/manual pump I used before. However, even with that one I am not always confident to desolder caps very close to delicate traces, plastic parts like sockets, slots, ... Simply if there is a reliable non-invasive way of testing caps I would much prefer that.

Don't assume that it will give you a valid in-circuit result. Generally though, you are more likely to get false negatives rather than false positives with one of those, so it's not a waste to at least try it first.

It's absolutely enough to only desolder 1 side of the component; you don't have to fully remove it.

Personally, I always test out of circuit because troubleshooting can take so long that I don't want to risk running down the wrong path for one component. You will end up spending more time (in the balance) on a single line of inquiry than you will in desoldering/resoldering components. Unless you just get lucky. Sometimes that happens, but I don't like gamble my time (and sanity) on getting lucky.

mpe wrote on 2021-04-21, 07:24:

The device advertises in-circuit capacitance/ECR testing (given by using 100 kHz pulses and low voltage). Will see how it goes! In worst case I can always use it to test desoldered parts.

This is doubtful, at least in a general sense. Most of these apply a very low current across a discharged capacitor to measure instantaneous voltage rise (prior to significant charge buildup on the cap). In theory, the involved voltages and currents are so low that semiconductors won't activate and sections of the circuit will remain electrically isolated. The device can still be confused by passive components (capacitors, inductors, resistors) in electrical proximity or by leaking semiconductors not fully isolating a section of circuit.

An even more sophisticated device could potentially detect and measure impedance deviations via reflectometry. Usually this is only done on wires and RF circuits which have fairly distinct characteristic impedance, but it's theoretically applicable in more general conductors over a wider frequency range. Measuring impedance over a wide frequency range would also allow capacitive reactance to be isolated from resistance and inductive reactance, but the sampling performance required means it's likely only possible in a laboratory setting for the moment.

On the plus side, this device does operate at 100khz which is the typical point of self resonance for electrolytic caps (where inductive and capacitive reactance is equal in magnitude). That's the correct measurement frequency for lytics, although it's not necessarily correct for plate caps, ceramics, or other products intended for higher frequency application.

Miphee wrote on 2021-04-21, 06:11:

So you either start removing and measuring elcos one by one or trust the readings you get in-circuit. If I had to choose a method (strictly talking about XT-S3-era boards) then I'd choose to replace them all. The usual 10/33/47/470 uF elcos are extremely cheap.

The analogy isn't exact, but the two perspectives I alluded to earlier in the thread are similar to the debate over predictive vs preventative maintenance. Determining which approach works best is usually very situational and will probably be different for a person starting out in repair vs someone who has experience.

7F20 wrote on 2021-04-21, 15:29:

It's absolutely enough to only desolder 1 side of the component; you don't have to fully remove it.

For this kind of job, it's fine. For completeness I'll just point out that your world gets a bit more complex when dealing with RF circuits.

Removing caps is just a matter of adding fresh solder and heating one leg at a time, walking the cap out of the board.
Adding solder helps improve the heat transfer.
A temperature controlled iron is highly recommended. Back when I used a cheap mains powered iron I had a lot more trouble with leaving burn marks on the PCB.

Got the ESR tester. I tested all 10µF caps on the above motherboard in-circuit.

It is a bit of tricky to test with short leads and aligator clips. I guess I need to get very shor quality probes for this.

I got readings between 0.6-3.6 Ω. That seems to be still good value according to the table (the threshold for 10µF/25V is 5.3). But nevertless, much higher than a brand new caps.

I desoldered those with high values >3Ω and measured them out of the circuit. The values were a bit (0.8Ω or so) lower than when in-circuit. And capacitance was also a bit lower than nominal, but not far off. So the device seems to be working. However, I only know for sure once I find any actual bad cap 😀

mpe wrote on 2021-04-25, 08:22:

Got the ESR tester. I tested all 10µF caps on the above motherboard in-circuit. […]
Show full quote

Got the ESR tester. I tested all 10µF caps on the above motherboard in-circuit.

It is a bit of tricky to test with short leads and aligator clips. I guess I need to get very shor quality probes for this.

I got readings between 0.6-3.6 Ω. That seems to be still good value according to the table (the threshold for 10µF/25V is 5.3). But nevertless, much higher than a brand new caps.

I desoldered those with high values >3Ω and measured them out of the circuit. The values were a bit (0.8Ω or so) lower than when in-circuit. And capacitance was also a bit lower than nominal, but not far off. So the device seems to be working. However, I only know for sure once I find any actual bad cap 😀

You can't know if they're in spec without identifying the manufacturer and series. ESR can vary tremendously depending on how the cap is constructed.