Yet another ESR thread...

[torch]

I have been promising myself I would build an ESR meter. Soon. Really. As soon as I get a round tuit. In the meantime, I have been using my Hantek 3x25 to generate a 100Mhz waveform into a 50 ohm terminating resistor into the scope, teed off to a pair of test leads. It's bulky, but it works.

Yesterday, She Who Must Be Obeyed was complaining that her computer was real slow booting. A couple more tries and it wouldn't boot -- no "Power Good" signal coming back to the PS. A little bit of knowledge is a dangerous thing, so of course I declaimed "Ah ha!" I bet it's the caps. I can fix that. A half-hour later the computer was reduced to a pile of cards and wires and I was staring at a bulging cap and a split cap near the CPU. Being almost as bright as I think I am, I realized the other caps may also be on their way out, fired up the jury-rigged ESR described above and began to poke and prod. Fortunately, all the caps tested OK, so I desoldered the bad ones, installed the replacements and started trying to figure out where all the wires came from.

Now if you re-read the last sentence with greater care than I took performing it, you will realize I said "ALL the caps tested OK". During the reassembly process it dawned on me that two of the caps were physically bad, so why did they test OK?

I started playing with the bad caps. The capacitance value was really not that far off from the rating (rated 3300uf, measured ~4500). The ohmeter climbed to infinity. But the top was split. The machine now fired right up normally, so obviously changing the caps was the Right Thing To Do. Connecting the caps across my benchtop variable PS showed the reason: leakage current. Despite the ohmeter's assurances, as I cranked up the DC voltage, current started to flow -- about 1/2 amp at ~5 volts (caps rated 6.3v).

I have seen occasional comments that the ESR is not the be-all, end-all proof of a capacitor's worth, and this clearly illustrated the issue. Further playing around with the frequencies and various new caps showed a large cap will have a very low ESR even with a 1/2 hz wave, especially at low voltages, so I'm not certain how much damage it would take to show up when testing a lower value cap.

I wonder if, instead of looking for the lowest possible ESR when an AC source is applied, it would be better to apply a DC bias to the AC signal (ie: add an offset to the Hantek's output) and look for a range. IE: below a certain value indicates DC leakage current? Any thoughts on a suitable circuit from those that actually know what they are talking about?

[robrenz]

I don't know what I am talking about but I will give my 2 cents anyway.  Just put your meter in current mode in series with your power supply with suitable current and voltage limits and test for leakage.

[Alana]

I saved few caps from last bad cap repair and i will try to test them for leakage. No function gen so i wont be able to test for ESR... yet.

[kripton2035]

this DIY esr meter measures esr, capacity AND leakage...
http://kripton2035.free.fr/digital%20esr/esr-sprut-de.html
but not sure if it can measure in-circuit?

[PA0PBZ]

I have been using my Hantek 3x25 to generate a 100Mhz waveform

I hope that is a typo and you mean 100KHz?

Anyway, I think leakage testing can not be done in circuit, as you pointed out the current was flowing at about 5 volts, which would open the semiconductors on the board. I'm not saying that you should not do a leakage test, but it takes away the comfort of not having to remove the capacitors when doing a normal ESR measurement.

[cwalex]

I think it is fairly standard to replace all caps of the same type as those with the bulges since the caps are cheap and the labour is expensive. At least that is what I do, I don't bother checking ESR, I just replace them all.

[FenderBender]

Well sometimes if there isn't a real visual indication or there's a different line of caps that are on the board but don't show signs...but might actually be dying.

You're right, but it's nice to have.

[torch]

I hope that is a typo and you mean 100KHz?

Anyway, I think leakage testing can not be done in circuit, as you pointed out the current was flowing at about 5 volts, which would open the semiconductors on the board. I'm not saying that you should not do a leakage test, but it takes away the comfort of not having to remove the capacitors when doing a normal ESR measurement.

Ooops. Yeah, I meant 100 kilohertz, not megahertz.

There was current flowing long before I hit 5v -- that's just the maximum I cranked the PS up to. I thought it might be possible to detect leakage a lot sooner than that, possibly at a volt or so. However, after looking through that German ESR page, I realize that might not work for higher-voltage capacitors than the 6.3v ones I was working with.

[T4P]

When it comes to modern badcaps it's very likely they might not be the same series but others will go soon ... Unless you had a 100$ EPS v2.91 PSU which tends to use nippon chemi-con as the Input/APFC cap and Teapos for the output section
which is a dumb idea