Dry contact resistance testing

[cweagle]

(Hi all, longtime listener first time caller)

There is a class of milli/micro ohmmeters that uses <20mV signals to measure resistance, in order to not puncture any oxide layers that might be present.  I've found that Keithley makes some of course, there's a recent Agilent unit (4338B), and the Instek GOM-805.  I wouldn't mind spending the $1000 that the Agilent seems to run, if it were also a >=5.5 DMM.  The Instek is about that much but... I dunno, Instek.

Any recommendations for other models to look for?  Or a "reasonable" Keithley?  I don't need micro ohm resolution, but 0.1 milli would be nice.

[Vgkid]

Look into the ballantine 3205b , various valhalla models , LOM-510 , simpson 444

[mairo]

Hioki also has various units available:
https://www.hioki.com/en/products/list/?category=11

[ch_scr]

Also be aware that an SMU like the Keithley 2400 can be set up to do that (and so much more).
The K2400 supports 4-wire mode and ticks the 6 1/2 digit DMM box (it ticks the precision power supply and load box as well).
If you need to do exclusively dry contact testing, maybe get a more specialized unit. If you need a unit capable of that on your "general" workbench, I would highly recommend the K2400.
The necessary configuration (U/I limit, current reversal with offset compensation, 4 wire, show resulting resistance) can be made on the unit without need of a computer or programming.
The resulting configuration can be saved as a preset to be recalled as necessary.
Of note is the full control over the measurement parameters (some DMM do the reversed current with a lower amplitude or no current at all [just measuring the offset voltage]).
Full current in both directions is preferred to have best signal to noise ratio in the measurement.

[Gyro]

You could always take the approach of shunting down the measurement probes with a resistor until the test voltage is <20mV and then treating the resistor and contacts as a parallel resistor pair. It would clearly tell you immediately whether the dry contacts are open and it wouldn't be too difficult to establish the contact resistance with reasonable accuracy.

Not ideal, but it would avoid the need for too specialised equipment.

 

[bdunham7]

Do you need it to be accurate?  Dry-contact sensing is an arcane feature that only shows up on a few instruments, as you've observed.  It doesn't make sense to buy one of those instruments just to get that feature if you can easily construct a small tester From a few resistors and an op-amp. 

[cweagle]

You're right, the 2400 is indeed a very nice meter!  About $2-3000 too nice, unfortunately.

[cweagle]

Do you need it to be accurate?  Dry-contact sensing is an arcane feature that only shows up on a few instruments, as you've observed.  It doesn't make sense to buy one of those instruments just to get that feature if you can easily construct a small tester From a few resistors and an op-amp.

What I'm actually doing is looking for barely-broken conductors in cables and FFCs: they have a little bit of oxide in the break so they still work for signaling (kind of) but have definitely failed.  This sort of break happens when the flex life of a conductor is exceeded; it's nothing you'd see from 'normal' mechanical damage.  We use bespoke continuously-measuring current meter fixturing to look for these while the flexing is happening, but I'm suspicious that they're starting earlier than currently observed, and instantly healing due to oxide punch-through.

Sometimes a TDR pulse can find them, but they're subtle.  When you open the cable you find a little spot of cupric oxide where the conductors fall apart; it's quite difficult to see with even a good x-ray machine.

Looking at the prices & availability, I might just get a fast 6.5 meter and make up the drive circuitry from a bank of op-amps.  The AC thing is a good idea though; I could mod/demod for better SNR.

[ch_scr]

The problem is the limited current dry contact testing allows, which means at low resistances the voltage you are measuring is so low thermoelectric voltages become quite relevant. When test current is reversed, the voltage across the resistance reverses as well - but the thermoelectric voltages are still same polarity. Thus the difference between the two measurements can be substracted out and the real measurement result remains. Maybe get a second hand K2400 from a reputable dealer and get it calibrated / adjusted?

[cweagle]

Ah-ha, that's why the AC is used.  I remembered an old Jim Williams chopped milliohm meter app note where he did mod/demod for noise rejection.  I hadn't considered the thermal EMFs.  I suppose that if I had bothered to get the Keithley book out (to see what their product line was like 15 years ago) I'd have learned that.

Well this isn't especially good news though.

[bdunham7]

What I'm actually doing is looking for barely-broken conductors in cables and FFCs: they have a little bit of oxide in the break so they still work for signaling (kind of) but have definitely failed.  This sort of break happens when the flex life of a conductor is exceeded; it's nothing you'd see from 'normal' mechanical damage.  We use bespoke continuously-measuring current meter fixturing to look for these while the flexing is happening, but I'm suspicious that they're starting earlier than currently observed, and instantly healing due to oxide punch-through.thing is a good idea though; I could mod/demod for better SNR.

You might consider using a small signal and op-amp circuit as an amplifier rather than the simple comparator I was thinking of, but use a ramp signal as your output so you can observe the nonlinearity that happens when the oxide 'punch-through' happens.  That way you don't have to try and predict the voltage level that the nonlinearity happens--you'll be starting from zero.

I've run into something similiar a long time ago with phono-preamps.  Sometimes one channel starts out weak (but not necessarily dead) and then a loud groove or a tap on the top of the record player starts it working properly.  That is usually a bad connection, but sometimes it can be a broken lead wire or what I found a few times, a bad solder joint in the preamp itself.  It seems that the high frequencies make it through but the low ones don't--not too hard to understand.  Perhaps you could also consider using a dual-frequency test signal, say 10Hz and 1MHz.

I'm not sure the actual 'dry-contact' feature that you find in the few products that have it is exactly what you are looking for.

[cweagle]

It's not clear to us right now what happens -after- the oxide punches through - clearly at high enough energy levels (voltages) you get more or less reasonable conductors again, but there's probably some sort of in-between that has information too.  Now I'm thinking about low-amplitude square waves that slowly get bigger, with inferred resistance calculated with each pulse, and look for the change 'point' (or whatever).

Seems like I should be able to do that with a waveform generator with a big attenuator, and a triggered data acq.  Interesting.

There are also a number of other tactics in play here of course, including some novel TDR work, non-linear junction detection (copper oxide made usable RF detectors before germanium was a thing, after all), and some inductive network analyzer experiments (sensitive RF detection keeps the voltages very low).