LCR meter opinion

[indman]

You have to check the specs for the LCR meter.

This is fair, but now I see that 3 different instruments have readings at 10 and 100kHz differ by more than 200 times! What testimony should I trust in such a case? How do you reconcile this with such a measurements below?

[rsjsouza]

Yepp.

Today afternoon a test with an wima mkp10 cap, 2.2µF:


1Khz:

U1733C            ST2830         DE5000

2.102µF            2.103µF       2.103µF
D0.001             D0.000         D0.000

10Khz:           

2.105µF            2.106µF       2.104µF
D0.018             D0.000        D0.000

100Khz:

2.351µF            2.389µF     2.121µF
D0.232             D0.015       D0.081

Still nothing to worry, next test perhaps with a 3.3 or 4.7µF - although it makes no sense at 100khz because of the high cap value.

Thanks for doing this comparison, Martin. MKP tends to have a larger tan δ variation in comparison to similar materials (interesting TDK article)



My Reed R5001R reads the following on a few capacitors lingering around. All MKP

		2,5 µF				3,9 µF				6,8 µF
Freq		C	tan δ			C	tan δ			C	tan δ
1 kHz		2,379	0,003			3,535	0,000			6,801	0,001
10 kHz		2,340	0,015			3,509	0,007			6,754	0,006
100 kHz		2,23	0,099			3,47	0,082			6,89	0,065

[Martin72]

I strongly disagree, there is reason for concern, because the difference in the results for U1733 20 times, for DE-5000 - 80 times!

Hi,

What have we done...
We´ve measured cap values beyond their claimed specs.
In a meter-range with mostly undefined specs(in case of the de5000, for the two others I must have a look in their specs.)
This is not the way it should go.
Normally you got a look at the datasheet from the cap-manufacturer, on which frequency the values based.
For example, wima their special snubbercaps, all the values mentioned in the spec were measured with 10khz.
It makes no sense to measure the cap with 100khz and judging about the displayed values.
Apart from this:
I´ve taken some oscilloscope pics from the testfrequencies.
There´s an interesting behaviour on the keysight u1733c and de5000.
Both sending a sinewave to the DUT, which amplitude decreases the higher the testfrequency get.
So far, so good, the ST2830 does it too.
But, while the sinewave of the ST2830 is still an ac signal at 100Khz, the two others their signals are getting above zero-crossing.

MKP tends to have a larger tan δ variation in comparison to similar materials

Indeed, it´s getting more and more interesting for me, therefore I´ll do more cap-measurings and post them here.

[indman]

But, while the sinewave of the ST2830 is still an ac signal at 100Khz, the two others their signals are getting above zero-crossing.

A similar effect is also mentioned here and how do you explain it?
https://www.eevblog.com/forum/testgear/any-one-mod-the-de-5000-for-1mhz-frequency/msg2673546/#msg2673546

[Martin72]

Let me think about for a while..
It could be the reason, why the u1733c and de5000 showing more dissipation than the st2830.
Could, not will.

Here the pics..

[tautech]

Martin, probably a bit unfair to compare devices using 4GSa/s and 1GSa/s DSO's.

[nctnico]

I strongly disagree, there is reason for concern, because the difference in the results for U1733 20 times, for DE-5000 - 80 times!

Hi,

What have we done...
We´ve measured cap values beyond their claimed specs.
In a meter-range with mostly undefined specs(in case of the de5000, for the two others I must have a look in their specs.)
This is not the way it should go.
Normally you got a look at the datasheet from the cap-manufacturer, on which frequency the values based.
For example, wima their special snubbercaps, all the values mentioned in the spec were measured with 10khz.
It makes no sense to measure the cap with 100khz and judging about the displayed values.

That doesn't matter. The readings should still agree.

[Martin72]

@tautech:
Hm?
Shouldn´t make a difference in this case...

@nctnico:

That doesn't matter. The readings should still agree.

Agree to what ?

[nctnico]

@nctnico:

That doesn't matter. The readings should still agree.

Agree to what ?

Between eachother. All LCR meters measure the capacitance at a given frequency. So in whatever way the DUT is affected by the frequency, they should read the same value (within their accuracy specification).

[Martin72]

They do, see the values at 1khz and 10khz.
At 100 khz they differ for some reasons, but they all are in their specs.

[TimFox]

@nctnico:

That doesn't matter. The readings should still agree.

Agree to what ?

Between eachother. All LCR meters measure the capacitance at a given frequency. So in whatever way the DUT is affected by the frequency, they should read the same value (within their accuracy specification).

My point is that the change in measured capacitance is affected by the fixturing (including clips and lead length) and short-circuit calibration, which can differ slightly between LCR meters.

[Martin72]

Yes, therefore I will bring the three on the nearest level when doing the next measures.
Actually, the keysight got short 2-wire alligator clips.
DE5000 got short 4-wire alligator clips and ST2830 4-wire kelvin clips.
But all got these "direct" terminals (ST2830 a fixture for this), next time I´ll use these ones, when the cap allow it.

[TimFox]

Pay special attention to the connections for "short" calibration, to ensure that they are very equivalent to each other to get a repeatable parasitic inductance (outside of the internal inductance of the capacitor).
It might be better to use a very short piece of medium-heavy copper wire between the clips, rather than connect the clips to each other directly.

[Martin72]

Will have it in mind.

Interesting thing:

chinese bench lcr

[mawyatt]

Pay special attention to the connections for "short" calibration, to ensure that they are very equivalent to each other to get a repeatable parasitic inductance (outside of the internal inductance of the capacitor).
It might be better to use a very short piece of medium-heavy copper wire between the clips, rather than connect the clips to each other directly.

Agree, this is the inductive equivalent of removing the few femto farads from the test fixture. A small piece (20mm x 10mm x 3mm) of solid copper sheet works well as a good inductive & resistive shunt.

Best,

[Martin72]

Today I took several caps to work, starting to measure them, hope of finishing before the weekend starts.
1µ, 2.2µ, 3.3µ, 4.7µ and 8.2µF.
First I took the 8.2µ . ....You can forget it at 100khz..
Every LCR are showing nonsense at this frequency but that doesn´t surprise me much, I´ve expected it already.
It´s simply not the frequency for µFs.
Nevertheless, the values will be recorded in the table.
The 16µf the keysight shows, the 39µf the ST2830 shows and the 9.1µF the DE5000 shows (Winner!  )

Martin

[ogden]

First I took the 8.2µ . ....You can forget it at 100khz..
Every LCR are showing nonsense at this frequency but that doesn´t surprise me much, I´ve expected it already.

No wonder! Reactance of 8.2uF capacitor at 100 KHz is 0.1941 Ohms. Lowest impedance of most LCR meters is 100 Ohms meaning by measuring such a big capacitor at such a high frequency you are throwing resolution of your instrument out of the window. Only reason to measure such capacitor at 100KHz is to show that you don't know what you are doing

[Martin72]

I wasn't going to say it so directly, but now you have.
Thank you...

[ogden]

Right. Check specs of decent meter like U1733C and you will see that measurement accuracy for capacitors larger than 20pF is better at 1KHz rather than 10KHz or 100Khz. Inductor measurements are what you need 100KHz LCR meter for.

[edit] Decent primer of impedance measurement: https://www.keysight.com/zz/en/assets/7018-06840/application-notes/5950-3000.pdf

[Martin72]

Point is, I know that, many others too.
But some doesn´t and then they´re getting upset.
Meter is crap, faulty, have issues, not recommendable, and so on...
You doesn´t measure µF caps with more than 10khz, point.
When cap manufacturers don´t so, why should you.
Then the D-Factor thing...
On keysight and DER LCR, the D-factor will displayed with maximum 3 digits after the comma.
It´s  a kind of nonsense for most of the foil caps.
So mostly you´ll see only 0.000 in the display, when testing with the best suitable frequency.
The D-factor is a calculated parameter, based on the measure.
You can´t trust measurements with 100khz what µFs concerns - So you musn´t get upset, when the calculated D-factor rises up 80x in comparison to lower testfrequencies.
Best example is the test with the 8.2µF on the keysight.
It shows a dissipation of 4.26.....Before, with 10khz, it was 0.001...
So it should be clear to everyone, what that means.


Martin

[indman]

The DE-5000 has problems measuring high quality film capacitors at 100kHz and no one has convinced me otherwise. Best of luck with the measurements!

[Martin72]

When you don´t use the right frequency for testing, it´s not the problem of the lcr.
Here a couple of datasheets snippets from high quality foil capacitors, their dissipation factors(see attachments below).
They all have one thing in common, 100khz only for caps max 100nF(except 383 series from vishay).
I can understand your concerning, it shows something at 100khz you can´t trust.
But that is not important, it´s not a issue, a showstopper, a problem.
Because you don´t test mikro-farads with 100khz, sounds simple, is simple.
When the DE5000 would have this behaviour on the right frequency, then it will be a problem, a isssue, a no go thing.
But it don´t have it.
And it´s getting "better":
DE5000 and the unlikely more expenisve U1733C from keysight have one thing in common, the D-factor display is nearly useless for foil-caps because of the poor display resolution(lowest value 1x10-³ = 0.001).
The ST2830 fits better.

Martin

[nctnico]

Sorry but you are not convincing me by showing capacitor specifications. It is not like you can't use these capacitors above the frequencies used to measure / specify them.

Loss factor is frequency dependant so it will show different values at different frequencies. How much different depends on the dielectric losses which can vary wildly with frequency depending on the dielectric material used:
https://www.sciencedirect.com/topics/earth-and-planetary-sciences/dielectric-loss
https://www.intechopen.com/chapters/39574

In a real world scenario I would like to know how a several uF capacitor behaves at higher frequencies (heck, your LCR meter likely has a several uF capacitor at the input to seperate the DC bias from the signal and the designer very likely specified / measured them to assess fitness for the purpose). An LCR meter typically covers the lower frequencies where a network analyser typically doesn't work very well.


Bottom line: saying LCR meter can't measure capacitors at higher frequencies doesn't make sense. Your results scream measurement / equipment error to me. Are the Kelvin clips actually Kelvin clips? How about shielding?

For example: this 10uf 0402 capacitor is specified up to several GHz: https://www.murata.com/en-eu/products/productdetail?partno=GRM155D70G106ME18%23


I did some testing with my own LCR meter (GW Instek LCR-821 to be precise; an older higher end model) and I measured some polyester/MKT capacitors: 1.5uf, 1.8uf and 2.2uf from my parts bin (unfortunately the largest I have around is 2.2uf). They all read to within a few % of error from 100Hz to 200kHz using Kelvin clips. With every frequency change I performed an offset calibration (with a piece of busbar for the short test). As expected the dissipation factor changes with frequency. The 2.2uf goes from 0.0011 (@100Hz) to 0.043 (@ 200kHz).

[Martin72]

Bottom line: saying LCR meter can't measure capacitors at higher frequencies doesn't make sense. It screams measurement error to me.

I don´t tell that. 

It´s important to do measures according to the specs for getting more or less constant reproduceable results you or others can work with.
Specs what the selected cap concerns and what the measure device concerns.
Measuring values with a frequency higher/lower than the frequency on which the values based according to the datasheet means simply, that what you measure then is not specified.
Same when using a frequency range of the meter.
When you want to do measures with a frequency that covers all ranges of the meter, you must choose a frequency that will do so, according to the specs of the meter.
When you want to measure "higher" capacities with e.g 100khz for what reasons ever, you must use a meter which can do that.
It´s not a fault of the meter you used before.
When you test a cap beyond it´s specs with a meter beyond it´s specs, what do you expect to get ?

[Vgkid]

My LCR meter has a bunch of calculations to calculate the measured accuracy.
Most high end LCR / cap meters should use a accuracy formula...
the easy to use calculator for my unit:
https://www.ietlabs.com/notes/digibridge_accuracy_calculator