MOV degredation monitor? how about fuse condition? (with RF)

[coppercone2]

Is there a way to check MOV integrity some how? And some how extrapolate the data to a curve that shows some kind of information about what you can expect from them?

Is there a way to do it with specialized equipment, like leakage in a megger? I have seen MOV holders that are built like fuse holders, so you can replace your MOV as a periodic maintenance procedure if you are interested in high reliability, like proactively replacing fuses on a maintenance schedule. I have never seen a fuse tester before though, only replacement to fall within some kind of SOG.

Is there a way to do it in circuit while powered? Leakage current monitoring? Some kind of high frequency measurement (like using a periodic RF signal to test their integrity).. I would like to boil it down to some kind of simple analog circuit if possible.

For instance you can do this with peltier modules (kinda), measure their impedance, to see the deviation from normal, and a useful pass/fail system can be developed this way. Usually done at like 100Hz. I guess you can monitor the IV characteristic too if you prefer.

Same goes for fuses, I can't imagine what kind of information you would get from them though, maybe a raise in resistance? Can some kind of change of alloy stress/crystal phase change be detected using RF/Microwave and sensitive equipment for electrical measurement? I assume you might be able to check the fuse structure change on a x-ray crystalography machine but that's really crazy. It's really hard for me to imagine a DC impedance test to provide useful information ,but I thought maybe some kind of microwave test might show something (really far out I know), maybe something related to skin effect due to different alloy composition? I assume it work hardens and a different crystalline structure of whatever the fuse is made of will show up due to fast cooling on the surface. Maybe this manifests itself as a nonlinear effect at some terribly high frequency?

I also thought some kind of porosity test can be conducted on the fuse wire to maybe see microcracks or something on the surface related to many cycles but again thats impractical and crazy unless its some expensive fuse.

I mean data that can generate a useful reliability curve.

Maybe tempco gets weird when they are degraded?

This provides some insight
http://www.digikey.fi/en/pdf/l/littelfuse/lf-varistortesting

But I am interested to know if there are better methods. You can't do the voltage drop method in circuit because it requires current limiting, unless you switch in a resistor/DCsource and disable protection for the duration of the test. Since they are DC devices a AC/RF test could be used with the appropriate filter networks, even if they are energized to high voltage.

Do you think with a MOV you can measure a high frequency voltage drop by feeding it a high voltage high frequency signal that's  current limited, and  block the supply by putting a inductor or ferrite on the MOV to Line/DC (connect the MOV through a LPF)? Where the test signal frequency is >> then the line voltage so you can properly direct it with filter impedance?

It would effect the performance of the MOV but you can use a different circuit element to handle HF HV stuff maybe that is not widdled down by the LF, like connect another MOV through a beefy HPF after the MOV that has to deal with the 60Hz. And then monitor that mov with a low frequency that does not go through the HPF. The reason for this would be because line impedance is unstable usually, so you would need to either stabilize line impedance or use filters to set up impedance so you are not measuring a rando asshat impedance leakage.

Don't know if you can beat the response of a GDTMOV combo with filtered MOV though. Maybe. Also on lower voltage MOV.

[IconicPCB]

I used to run a bare board test facility, testing bare boards prior to population.

One of the problems we saw was the fact that a tested board would fail during assembly process.
Sometimes it would be a problem associated with hole copper barrel sometimes a micro crack in the copper track.

It was understood that a functional test was needed to augment the usual connectivity and isolation testing of the PCB.
This was in response to..."yes there is copper but how good is it..."
The answer to this question was a test stimulus consisting of a DC +RF signal. DC would be used to provide localised heating at the potential fatigue point while the RF was used to identify the presence of such a fatigue point.

Copper will act as a non linear conductor at the fatigue point and creating a hot spot will increase the non linearity.
RF will experience distortion when passing through such a spot and can be readily picked up and measured with a spectrum  analyser.

Harmonic content of measured signal being a measure  of likelihood of future fault in the copper track.

I am sure a similar regime of excitation could be used to quantify MOV failure/reliability.

[coppercone2]

Sounds like it would work for fuses too.

I am kind of thinking, a slow AC signal should still cause spot heating similar to a DC signal right?

Any idea at what wavelenghts relative to device or track geometry you start getting some kind of spot heating changes?

Do you just model it using the first order model, and calculate how much energy goes through the parasitic capacitance and how much goes through the ideal resistance? Then kind of treat it like a resonance where there are similar energies going through both paths and that means the heating profile will be significantly different?

How do you deal with the parasitic inductance? Is this line of thinking even relevant? Conceptually I have no idea what is going on, in the microwave range in a waveguide I can kinda imagine the voltages along the waveguide but I don't think thats relevant

[IconicPCB]

Parasitic elements are not an issue since the heating is supplied by DC.

So a rated current based on track geometry is employed to create localised heating ( may be some experimentation will be required) this is augmeted with an injected RF signal and harmonics are indicative of non linear junctions in the copper track in this case the MOV ( which of itself is non linear ).

[coppercone2]

No, I mean if heating with DC is not an option because you want to test in circuit, while its operating, but you have option of using still kinda low frequencies in comparison to the high frequencies that would notice the nonlinear junction, but frequencies you can filter out from the circuit power supply (so raise the impedance at that frequency between the MOV and the Power rail to isolate your RF heat source (low frequency but isolatable and leaving the MOV useful) from the AC power source)

How will heating at a low frequency that is way above DC but way under typical nonlinear junction frequencies (I assume this would be at least in the VHF range). I also assume the spot heating time constants would be in the miliseconds or so, so you can turn off your RF heater and quickly turn on your higher RF measurement frequency, in the microseconds, to prevent measurment error from saturation in filters, RF front ends or what have you.

I am thinking this like a bug detector, I don't see those run at under 900MHz. I figure at least 100MHz directly coupled ? Say you wanna heat it with like 100KHz, so you can connect the MOV through something that is high impedance @ 100KHz but low impedance under that. These are just random frequency numbers I pulled out of my ass.

Then use another MOV through a high pass filter, this mov is not likely to get much energy going through it unless there are real fast surges, so its less important to monitor.

It seems that if you do  this method there will end up being some kind of unmonitorable MOV eventually, but it will degrade the system less if you disconnect that one to measure it.

Even if this is not practical I am still interested in the frequency response of things with regards to spot heating behavior deviation from DC because it is conceptually interesting.

If you use multiple MOV you might as well disconnect them one at a time for measurement if they are in parallel, the only problem would be that a control system could fuck up and leave your circuit vulnerable (so long you monitor the system while the MOV is disconnected to make sure one of them did not get hit and take a big surge by itself), so there is still merit to some kind of frequency staged MOV thing since it has a intrinsic systemic failure avoidance thing going on, and doubling up systems is not nearly as interesting, because that transfers over to hardening a control system rather then exploring weird physics effects and fancy analog circuits. 

Also there may be other circuit elements better suited for RF transients then MOV's which can be put upstream.

[T3sl4co1l]

Yes.

AFAIK, leakage increases with age/number of surges.

The internally fused kinds, of course, will show abnormally low leakage on failure.

Tim