500 ppm LCR-meter: measuring high-value resistors (10..100 M

[porcelijn]

This LCR-meter works a treat for "regular" resistor values, but shows unstable results for values over 10 Mega ohms. These hi-R measurements should be done at low frequency (100 Hz) to minimise the typical lowpass filter effect of hi-R resistors. No need to say that these resistors are measured within a properly earthed, full metal enclosure, with maximal samples averaged.

But that's not enough. The LCR-meter allows each individual sample a very short timeslot, thus minimising time/phase faults while alternating between I and U measurements. These samples are probably not fully synchronised with the 100 Hz sinussignal (and this signal itself is definitely not synchronised with the omnipresent AC mains noise).

To compensate for this a much longer timeslot would be needed. And wouldn't it be nice to add an extra "zero" choice to the 3 sinus frequencies? Then with all sinus switched off (and DUT still in situ) the meter could measure the actual noise level, taking this into account when the sinussignal is restored. This and an extended timeslot could probably be realised in a next software version.

Or am I off topic here? Maybe measuring resistors proper should better be done with a DC source and DC mV meter instead of a sinussignal-based LCR-meter?

[snoopy_94]

These hi-R measurements should be done at low frequency (100 Hz) to minimise the typical lowpass filter effect of hi-R resistors.

This is not quite correct! Indeed the principle of measurement to obtain both real and imaginary components. In the case of high value resistance, the resistance (Rp) and the parallel capacitance (Cp).
For example a 10 MΩ resistor, the measured value (DC) with a Fluke 8840A at 10.221 MΩ, LCR-Meter measure (Max precision unchecked and 4 for Averaging, in the Preferences panel, use of the TH26001A Test Fixture):
-10.22 to 10.24 MΩ / / 0 to 0.3pF at 100 Hz
-10.22 MΩ / / 0.21pF at 1 kHz
-10.23 MΩ / / 0.21pF at 10 kHz

With a 100 MΩ 1% resistor:
-100.8 to 101.4 MΩ / / -0.05 to 0.07pF at 100 Hz
-100.0 to 100.3 MΩ / / 0.04pF at 1 kHz
-106.0 to 106.3 MΩ / / 0.05pF at 10 kHz

I think the best is to use 1kHz.

These samples are probably not fully synchronised with the 100 Hz sinussignal (and this signal itself is definitely not synchronised with the omnipresent AC mains noise).

The samples are synchronized with the test frequency, but the test frequency is not synchronized with the AC mains.

To compensate for this a much longer timeslot would be needed. And wouldn't it be nice to add an extra "zero" choice to the 3 sinus frequencies? Then with all sinus switched off (and DUT still in situ) the meter could measure the actual noise level, taking this into account when the sinussignal is restored.

Now the signal is sampled for 10 periods and 100 samples per period. Then the average of x times the "Averaging" value is performed (x = 2, 9, 14 for 100/120Hz, 1kHz and 10kHz). So it should be 10 x 100 x 2 or 2 kB for measurement (2 bytes per measurement - ADC is 16 bit). It is difficult to have more. I have to look if noise can be measured.

Maybe measuring resistors proper should better be done with a DC source and DC mV meter instead of a sinussignal-based LCR-meter?

I did this long ago for the measurement of resistance between 50 and 1000 MΩ!

Jean-Jacques

[porcelijn]

Jean-Jacques, your accurate measurements are amazing!
Mine were disappointingly different. With the same settings and 100M 5% resistance I got these results:

93.04 to 94.18 Mohm / 0.65 to 0.73 pF at 100 Hz
122.4 to 123.3 Mohm / 0.38 pF at 1 kHz
216.9 to 217.3 Mohm / 0.04 pF at 10 kHz

These results are unusable of course, and they were obtained with virtualy the same equipment you used. What went wrong in my testing? I didn't dare posting this reply until I had managed to improve my results somewhat.

Phase recalibrating (and doing that two, three times..) gave a minor improvement. Using another 100 Meg array also produced slightly better results. (I don't have one single 100 M resistor, so I use 10x 10M or 5x 20M).

But the real breakthrough came when I realised that in my effort to minimise noise I had actually spoiled the 4-point measurement system. I had started my tests with a self-constructed pair of Kelvin leads: two pairs of thin, 60 cm long shielded cable, each pair ending in the tip of a multimeter-testpin. Nice and handy for low and medium impedance measurements, but prone to pick up noise at high R-values.

So I then decided to connect these hi-R resistors directly to the LCR-board's bnc-connectors: the 2 in the middle (= the current sense leads). These can on board by J21 and J22 be bridged to both outer bnc-connectors (= the voltage sensors), so outside no noise-picking extra leads are needed, it seems.

But here I was terribly wrong! In AC-measuring hi-R values even very small capacitors can heavily distort results. And the stray capacity of the 4 bnc-connectors should not be neglected here. Bypassing the voltage-sense connectors disturbs the balance between the 4 Kelvin leads, leading to wrong results.

So today I remounted my "multimeter" Kelvin leads, put heavy metal shielding all around the testing setup, and measured 100 Meg again. Results were far better now:

95.18 to 99.74 Mohm / 0.21 to 0.72 pF at 100 Hz
97.31 to 98.06 Mohm / 0.13 pF at 1 kHz
100.2 to 99.96 Mohm / 0.16 pF at 10 kHz

Jean-Jacques, that's still not up to your admirable reference results, but 5% accuracy is a workable start. At 100 Hz the apparent jitter is clearly mains noise, so shorter Kelvin leads could improve things. Or better still, a TH26001A Test Fixture, as you used. But a homemade equivalent shouldn't be too hard to construct. I'll try that for a start.

[thijsbeckers]

This is gold, you guys! Those are some pretty extremes...

[porcelijn]

After doing the Elektor-advised hardware modifications I get strange results when testing hi-ohm resistors (80 megohm and over) at 10 kHz. Correct values are shown at 1 kHz and 100 Hz, but at 10 kHz the value almost doubles. This suggests unwanted noise or oscillation at 10 kHz. I hoped to improve things by soldering the new 10 ohm resistor not over R45 but directly over the cut-out gap at pin 5 of U5, but this made no difference. Does anybody share my experience?