Keysight 34461A - Inaccurate SMD Capacitor Measurement?

[mrthanhapple]

Hi everyone,

I'm using a Keysight 34461A bench multimeter and I'm trying to measure some SMD ceramic capacitors (mostly 100nF to 10uF). However, the readings seem quite inaccurate compared to a handheld LCR meter and expected values.

I'm using standard test leads (no special LCR fixture), and I tried using the NULL function to cancel out lead capacitance, but still getting strange results. For example, a 1uF cap measures around 0.4uF.

Has anyone experienced this issue with the 34461A when measuring capacitors, especially SMD ones? Is there a proper technique or limitation I should be aware of?

Thanks in advance!

[coromonadalix]

could be the probes who drift the value ?

to be noted  capacitance tests on many meters are time constant related, and not the same frequencies related as "real LCR meter" tests ...

try shorter leads ?  and test on a clean surface, you have many debris ...

[wraper]

You can't expect measurements of these capacitors made under different conditions to match.  Not to say expected capacitance is not straightforward for these. They can lose like 30% of capacitance just by ageing (reset by heating above Curie point), even without considering voltage bias. And if you measure Z5U, Y5V, they can easily lose 30% of capacitance just by heating them with your fingers.

compared to a handheld LCR meter

I only see 2 different multimeters on the pictures. No LCR meter.

[mrthanhapple]

You can't expect measurements of these capacitors made under different conditions to match.  Not to say expected capacitance is not straightforward for these. They can lose like 30% of capacitance just by ageing (reset by heating above Curie point), even without considering voltage bias. And if you measure Z5U, Y5V, they can easily lose 30% of capacitance just by heating them with your fingers.

compared to a handheld LCR meter

I only see 2 different multimeters on the pictures. No LCR meter.

hi! I tested it with fluke 289, the previous test I borrowed from a friend's LCR and measured it wrong too. The smd capacitor on it has a value of 10uF (20%). When measuring with fluke 289 it gave a result of 7.2uF and with 34461a it gave a result that seemed higher than 12.4uF. I tested many smd capacitors and they were all wrong like that. But when measuring large electrolytic capacitors it was right.

[wraper]

Did you solder it by chance before measuring 12.4uF? In any case, multimeters measure in a very different way than (proper) LCR meters. When measuring parts with flawed characteristics like class II MLCC, this can result in very different readings. LCR meter frequency setting impact reading as well, although it's mostly seen when measuring electrolytic caps. However 12.4uF is suspicious, try measuring DC resistance of this capacitor to verify if it's not faulty.

[mrthanhapple]

hi !
I have tested and measured many capacitors with values ​​from 1uF to 4.7uF, 10uF with an error of 20%. Fluke returned results of 0.7xx uF, 3.7xx uF and 7.4xx uF respectively, and 34461a measured results higher than the nominal value of the capacitor (I calculated correctly, it was 20% higher than the nominal value of the capacitor). I don't know what the problem is or if it's just its nature, but I find Fluke's results more accurate.

[wraper]

What do you get if you measure film capacitors by 34461a?

[mrthanhapple]

What do you get if you measure film capacitors by 34461a?

I measured the 10uF capacitor and it returned 12.xx uF, the 1uF SMD capacitor returned 1.1xxuF. If it's correct, it should be lower, right?

[Someone]

34461a measured results higher than the nominal value of the capacitor (I calculated correctly, it was 20% higher than the nominal value of the capacitor). I don't know what the problem is or if it's just its nature, but I find Fluke's results more accurate.

Not more accurate, but measuring the capacitance in a different use case. As above by wraper, ceramic capacitors have wildly varying capacitance depending on how they're biased and what voltage/current waveforms are used for testing (which are different in the two multimeters for your example).

[wraper]

What do you get if you measure film capacitors by 34461a?

I measured the 10uF capacitor and it returned 12.xx uF, the 1uF SMD capacitor returned 1.1xxuF. If it's correct, it should be lower, right?

Generally, yes. That's why I'm asking about film capacitor measurement which should tell if there is something wrong with the meter.

[wraper]

I had a few 2.2uF 6.3V 0603 MLCC on my desk. Measured 1.5uF @1kHz by ST5C LCR tweezers, 1.7uF @1kHz by DER-5000 LCR, 1.65uF by 34461A, 2uF by U1272A, 1.7uF by U1282A, 1.7uF by BM037, 1.8uF by DT-2843R.

[mrthanhapple]

I had a few 2.2uF 6.3V 0603 MLCC on my desk. Measured 1.5uF @1kHz by ST5C LCR tweezers, 1.7uF @1kHz by DER-5000 LCR, 1.65uF by 34461A, 2uF by U1272A, 1.7uF by U1282A, 1.7uF by BM037, 1.8uF by DT-2843R.

i think it should be lower than the nominal value, but my 34461a seems to have a higher value. and i have a lot of smd capacitors on the iphone board. i don't know if it's because i upgraded the FW to the latest version (A.03.03-03.15-03.03-0052-04-030). the original FW was A.02...

Can I recalibrate myself?

[wraper]

First of all measure a good stable capacitor as I said. Secondly, if it gone like 30% off, it's unlikely it needs a simple calibration adjustment unless calibration data corruption happened.

[Dr. Frank]

as others already said, high CV types of MLCC have a big variety on their nominal capacitance value, especially when you have low grade ceramics, or a no name manufacturer.
High CV means either high voltage and/or high capacitance value for a given size.. look out for "edge" value type, i.e. the highest values in a series.
Every MLCC above 1µV mostly suffers from degradation (decrease) of capacitance versus temperature, time and especially versus applied DC voltage, on the order of many 10%. The manufacturers, especially the bad ones, then tend to deliver much higher initial capacitance values, like +30%.

If you have a 22µF / 6V3 capacitor, and you measures its capacitance at 5V working voltage (typically behind a LDO), you will notice that its capacitance might degrade by -50%, and additionally by e.g. -20% over time, when this operational voltage near its nominal voltage is applied continuosly. This might lead to instabilities of the LDO, because it really needs these 22µF.
If the application is run at automotive grade temperatures, like +85°C, additional degradation sets in. Most of these effects are reversible.

All these effects might lead to the different measurements, depending on the method used, i.e. if the measurement is done on a small DC level, or at high DC levels.
The DE5000 uses practically zero DCV and an AC method, whereas DMMs often use a DCV charging method.

To check your 34461A, you might compare film capacitors, where you don't have these effects, but it's difficult to get them at higher than 4µ7.

Frank

[mawyatt]

High density ceramic capacitors utilized high K dielectrics which inherently have a high voltage, temperature coefficients as well as poor aging, and are generally a poor choice to compare capacitance results.

We've found tantalum capacitors when higher capacitance values are desired, perform better than high K dielectric ceramics wrt to voltage, temperature dependance, and aging.

We just compared a 22uF Tantalum with a U1223A handheld, KS34465A, SDM3065X DMMs and DE-5000 LCR Meter (@100Hz). The results were 22.5, 21.83, 22.14 and 21.85uF respectively. The Zoyi ZT-MD1 tweezers report a value of 22.01uF @ 100Hz.

For even higher value capacitors some of the better quality electrolytics with low ESR might be considered.

Suspect the 34461 and 34465 utilized the same capacitance measurement scheme.

Best

[mrthanhapple]

Hello everyone! Today I tested it on a 1uF 10% tantalum capacitor. It seems to match the fluke 289 and the correct nominal value. I used a scope to measure its frequency and its pulse seems a bit different from the fluke's, so I think the 34461a can only measure capacitors under certain conditions. As for the smd capacitor, it probably doesn't meet the conditions, so the measurement results are not reasonable.
thank you very much

[ivo]

If the 65/61As have continued the same capacitance measurement technique from the 34410/11 (seems like it from your scope trace), then they use a constant current to generate a charging ramp, and use the fast ADC digitizer to quickly get enough points on the ramp to fit a time-constant curve and calculate the expected capacitance to generate such. So it's sort of a DC technique of measuring capacitance.

It seems they rather regard it more as a "toy" feature for spot checking, than any measurement to be relied upon, unlike the DMM's main functions.

Of course for different capacitors, this technique of measurement is liable to give you different results than say sine-wave vector / phase measurement of a proper LCR at different frequencies and different voltage amplitudes.

Some other DMMs and cheaper LCRs might use a square wave technique (easier to generate) and this might also cause varying results, and also depends on the frequency as well of course. Overall because capacitance is more-so an AC property, for any capacitor there's no single "correct" DC value it can have, like a resistor could.

[Kleinstein]

The ramp technique still has an effective frequency or time scale for the capacitance. A problem with the capacitance mode one many DMMs is that the frequency can depend on the measured value and it is often not given in the data sheet.

A lossy capacitor kind of has to have a frequency dependent capacity. For electrolytics this can make up some 10% for a factor 10 in the frequency. X7R can be similar, possibly even worse. In addition there is an effect of a DC bias. Not sure how much the capacitor could reach - it may well be in the 5 V rage. Chances are the ohms part can only work with one sign.

[Someone]

A lossy capacitor kind of has to have a frequency dependent capacity. For electrolytics this can make up some 10% for a factor 10 in the frequency. X7R can be similar, possibly even worse.

Not a significant effect at the frequencies that are used by these capacitance measurements (see below for an X6S example). Whereas the variation from AC alone is similar in magnitude to what the OP saw yet in the opposite direction!

[Dr. Frank]

Hello everyone! Today I tested it on a 1uF 10% tantalum capacitor. It seems to match the fluke 289 and the correct nominal value. I used a scope to measure its frequency and its pulse seems a bit different from the fluke's, so I think the 34461a can only measure capacitors under certain conditions. As for the smd capacitor, it probably doesn't meet the conditions, so the measurement results are not reasonable.
thank you very much

Sorry, you still do not get the point.

It's NOT the 34461A which measures "not reasonable", but the type / technology of the capacitor itself greatly affects the result.

The 34461A very probably delivers fully correct readings for the capacitance of your MLCCs, but at elevated DC voltages, where these capacitors show a degradation of its nominal capacitance value.

To correctly characterize your MLCC, you'd  need a capacitance meter which measures with an AC method at low level, with the possibility to make DC biased measurements.

Then you would see the same DC-bias effect, i.e. a lower capacitance value, at about the DC level which the 34461/465A are using.

Frank

[mrthanhapple]

Hello everyone! Today I tested it on a 1uF 10% tantalum capacitor. It seems to match the fluke 289 and the correct nominal value. I used a scope to measure its frequency and its pulse seems a bit different from the fluke's, so I think the 34461a can only measure capacitors under certain conditions. As for the smd capacitor, it probably doesn't meet the conditions, so the measurement results are not reasonable.
thank you very much

Sorry, you still do not get the point.

It's NOT the 34461A which measures "not reasonable", but the type / technology of the capacitor itself greatly affects the result.

The 34461A very probably delivers fully correct readings for the capacitance of your MLCCs, but at elevated DC voltages, where these capacitors show a degradation of its nominal capacitance value.

To correctly characterize your MLCC, you'd  need a capacitance meter which measures with an AC method at low level, with the possibility to make DC biased measurements.

Then you would see the same DC-bias effect, i.e. a lower capacitance value, at about the DC level which the 34461/465A are using.

Frank

Thank you for your detailed explanation. Perhaps my English comprehension is limited. I hope you understand.

[mawyatt]

Here's a link to something the OP might find interesting.

https://www.eevblog.com/forum/projects/ceramic-capacitor-behavior/

Note post #10 which shows how High K dielectric ceramic capacitors behave with DC bias. These plots were created with a lab grade bench LCR meter and custom DC Bias Adapter.

Anyway, as Dr Frank mentioned your KS34461A is probably fine, but the highly variable and unrepeatable high K ceramic capacitors are not good candidates for various capacitor comparisons with different instruments which employ different techniques and test levels & frequencies.

Best

[TimFox]

As mentioned above, DMMs typically use charging the unknown device in some kind of timing or relaxation oscillator, where the voltage across the device varies during the measurement and the loss (ESR) is ignored.
The capacitance calculation assumes a linear capacitor (Q = CV) with no voltage co-efficient or parasitic loss.
True LCR meters and bridges use sinusoidal waveforms to measure capacitance and loss at a well-defined frequency.