LCR meter tests

[mawyatt]

Assuming this isn't the Hioki contributions (which is valid IMO), then is this the DUT capacitor or the fixture, or both? If this type behavior is showing with different values and type DUT capacitors, the fixture is suspect. My guess it's more with the Kelvin Clips than the mechanical aspects of the fixture, but thats just a guess!! Having the "D" go from negative thru zero to positive seems to hint at the Zero Calibration which could be attributed to the Kelvin Clips??

Directly connecting (soldering) the DUT to the BNC connectors should eliminate any Kelvin or other type clip connection uncertainly. The 3D printed fixture has potential issues tho, being plastic with moisture absorption and temperature deforming effects, as well as aging. None of these are well known to us, but the 4 BNCs provide a stable electrical connection as well as some overall mechanical stability for both the DUT and fixture housing.

The 3D fixtures don't weight much, so little flexing stress is placed on the 4 BNC connectors. We designed the fixtures to be easy to print (no supports) with separate front and rear "plates" and a center rectangle "tube". Brass inserts were used to allow 3mm bolts to hold things together. This arrangement also allows relatively easy means for the Cu tape application. We didn't solder things together considering the butt type electrical contacts at the ends were sufficient as this is just a E shield not a H shield.

Would be interesting to see how well these 3D fixtures hold up with the Soviet Mica capacitors, they seem to do well with the DUT types we have so far

Best

[mawyatt]

It's not trivial, is it? But I don't think you want a snug-fitting pocket, more a pocket lined with foam that can support the capacitor without imposing mechanical stress, then flexible leads to the four-wire terminals.

If you are considering the 3D printed fixtures, then a solid wire to the outer BNCs provides good mechanical support while "floating" wrt to ambient. A couple smaller solid wires to the inner BNCs for the Sense connections. This was how we mounted the large PP 1uF cap, but it had thick heavy leads that were directly soldered to the outer BNCs to hold the cap without any other mechanical connection, and added some thin solid wire soldered close to the DUT case for the Sense leads.

With these Soviet DUTs wouldn't expect using the large terminals for mechanical support would cause issues with a heavy soldered wire, or even a Cu tab, to the BNCs.

Anyway, just some random thoughts.

Best

[EC8010]

I've looked into it carefully. I am also suspicious of the Kelvin clips. I have every so often seen a negative value of D (implying negative resistance). I was testing various capacitors including a batch of three Soviet PTFE all of the same value. The capacitors had large silver-plated stub terminals that were quite corroded, so I polished them clean with metal polish. I measured D of -0.00014, -0.00014, and -0.00015. That suggests something solid, not an iffy connection.

I had also wondered about the S/C calibration and on one attempt, I got rather larger correction values than expected, leading me to clean the Kelvin clips. A little dirt was visible on the cleaning paper. After that, the S/C calibration produced more familiar values.

I'm wondering about a piezo-electric effect and mechanical stress because of the fact that leaving a capacitor in the jig overnight and switching the Hioki on immediately gives a stable and sensible value of D, whereas inserting a silvered mica capacitor seems to require 120s or more for stability; the Hioki now has a kitchen timer nearby. Film capacitors do not seem to show this effect nearly as badly as silvered mica, making me wonder about dielectric absorption.

The measurement jig is destined to get an internal RTD so that temperature can be logged, and a couple of white LEDs inside so that I can more easily see what I'm doing. After that, those Kelvin clips may have to go, but I really don't want to have to poke a soldering iron in there every time I want to measure something. One possibility is to make another jig for soldered connections, but it's quite a lot of work...

Meanwhile, I logged a Vernitron silvered mica capacitor for two hours. The drift in capacitance is not due to temperature, perhaps supporting the mechanical stress hypothesis. The D for the Vernitron silvered micas was nowhere near as good as the Soviets, suggesting poor quality control of incoming material. Never thought I'd praise Soviet quality control! Of course, Vernitron couldn't send naughty employees to the gulag for ten years. Lab work is stopping now because weather forecast says it's going to be hot for the next week...

[mawyatt]

Wonder if wrapping a shielded mechanical enclosure hosting the DUT around these LCR Meter Fixtures might work?
Best

[EC8010]

With care, it ought to be possible to mount it inside a die-cast box, and hinge the lid like a door. The awkward bit is that the fixture would need to be at the bottom of the box, so the box will obcure a lot of the tester. Another thought that occurs is that a suitable biscuit or sweet tin would work just as well, but be a lot lighter. A hinge would soon bend the box, so the lid would have to just come on and off. But it would require earth bonding. Something involving a 4mm metal socket on the lid and a 4mm chassis mounting plug onto the TH fixture would do the trick.

[d-el]

I suspect what I'm seeing is mechanical relaxation. Whatever it is, it takes time to achieve stability.

I see a similar effect.

Precondition:
---------------------------
Agilent E4980AL
FUNC: Cs-D
FREQ: 1kHz
LEVEL: 2V
RANGE: AUTO
BIAS: OFF
MEAS TIME: MED
AVG:32

Ambient temperature: 26°C
Humidity: 34%
---------------------------

Capacity vs time:


I also realized that if connect a capacitor to the generator in advance, like, "warm up the capacitor", the transient time is reduced:

[EC8010]

Interesting; I assume your LCR tester had been fully warmed up beforehand? Not sure where it takes us, but it's another nail in the coffin of stable capacitances.

[d-el]

I warm up my meter for more than 5 hours. The room temperature fluctuated no more than 1 degree Celsius.

I have a SSG-1 10000pF that spent a lot of time in the meter. So if you connect it again, the result normalizes after a few seconds. It seems that there is a transient process when the capacitor has been lying for a very long time without signal.

[mawyatt]

Would be interesting to see how well these 3D fixtures hold up with the Soviet Mica capacitors, they seem to do well with the DUT types we have so far

Best

Finally got around to using a 3D fixture with a Soviet 10nf Mica cap. Printed in PETG.

After a couple days thought about using a Thermistor mounted to the Soviet Cap case to monitor the cap temperature, so created another end plate. Didn't use the 0.2mm nozzle (use 0.4mm) so the print is a little rough but good enough for now.

The 10nF Soviet mica has shown excellent stability, so expect equal or better results with stability with 3D fixture 

Anyway, here's our present progress. Looks like the fixture and cap take about 1hr to stabilize and about 2.1C temperature rise. More like ~3C now,

Best

[Hydron]

Hmm this reminds me I have some test results to post from the various capacitors that people have measured. Excuse is that I've been flat out with work and family stuff (I've only spent about half of the last 6 months at home in the UK). Will try and get around to posting something once the latest batch of jetlag has worn off.