BK Precision LCR 879B meter readings wrong, or user error

[Baqar79]

Hello there,

I'm definitely at a beginner level with my electronics (I studied it for around 3 years, but with a decade passing without using this knowledge I've pretty much forgotten all but the very basics).  I did electronics though as a hobby 20 years ago with a friend and we ended up with all bits of odds and ends (being 10 or so at the time, nothing is junk ).

My problem is this LCR meter that I just received a day ago (From tequipment.net; I imported it into NZ) is not giving me the expected readings for capacitors vs my Fluke 287 and I can't really figure out quite what I'm doing wrong.

I callibrated it using a copper wire short (open and closed) for each of the ranges, and relative all the results (using the included alligator clips) first before connecting the capacitors...and the results seem good at first, but the results seem to go downhill as I turn up the frequency.  The results of some unused capacitors (only purchased several months ago) are below.  The Fluke 287 (despite it's bulk) is my favourite DMM for it's accuracy with a straight up 1% + 5 digits for capacitors until you exceed 20,000uF; the callibration certificate included suggested it was dead on with the capacitor measurement (though they only used a 5nF to gauge the accuracy).

BK precision 879B readings vs Fluke 287 & BK Precision 2709B:

MKT capactors

(1nF marked)
100Hz   (2.5% + 2 digits)   1.0422nF (parallel) vs Fluke 287 @ 1.04nF and BK2709B @ 1.045nF
120Hz   (2.5% + 2 digits)   1.0419nF (parallel)
1kHz   (0.7% + 2 digits)   1.0370nF (parallel)
10kHz   (0.5% + 2 digits)   1.0260nF (parallel)

If you compare the 100Hz range capacitance with the 10kHz range capacitance it is a difference of just over 1.5% (1.0260nF/1.0422nF = 0.984) which is within spec for the given value.  If the fluke is within spec the value given for capacitance is between 0.98nF & 1.10nF so this looks ok.  On a side note the 2709B seems almost identical to the value of the fluke despite having a 3.0% + 30 digit accuracy at this range.

(1.5nF marked)    
100Hz    (2.5% + 2 digits)   1.5541nF (parallel) vs Fluke 287 @ 1.56nF and BK2709B @ 1.561nF
120Hz    (2.5% + 2 digits)   1.5533nF (parallel)
1kHz    (0.7% + 2 digits)   1.5451nF (parallel)
10kHz   (0.5% + 2 digits)   1.5292nF (parallel)

Within spec, though despite the greater accuracy of 10kHz & 1kHz there is a big change between them, note that the values are decreasing, this seems to happen a lot in these results.

(3.9nF marked)
100Hz   (2.5% + 2 digits)   4.183nF (parallel) vs Fluke 287 @ 4.19nF and BK2709B @ 4.22nF
120Hz   (2.5% + 2 digits)   4.182nF (parallel)
1kHz   (0.7% + 2 digits)   4.165nF (parallel)
10kHz   (0.5% + 2 digits)   4.121nF (parallel) 

Same as above...the results of 100Hz to 10kHz varies by 1.5% with the biggest jump happening from 1kHz to 10kHz

(56nF marked)   
100Hz    (0.5% + 2 digits)   55.67nF (parallel) vs Fluke 287 @ 55.8nF and BK2709B @ 56.01nF
120Hz   (0.5% + 2 digits)   55.66nF (parallel)
1kHz   (0.5% + 2 digits)   55.40nF (parallel)
10kHz   (0.5% + 2 digits)   54.82nF (series) 

Strange, still exactly 1.5% difference between the 100Hz and 10kHz range; yet according to the accuracy charts in the manual the should be within 0.5% of the value...so something is wrong here (probably someone more knowledgeable about LCR meters might know why)

(82nF marked)
100Hz   (0.5% + 2 digits)   79.80nF (parallel) vs Fluke 287 @ 79.9nF and BK2709B @ 79.7nF
120Hz   (0.5% + 2 digits)   79.78nF (parallel)
1kHz   (0.5% + 2 digits)   79.37nF (parallel)
10kHz   (0.5% + 2 digits)   78.15nF (series)

The change in value is now past 2% between 100Hz and 10kHz; even assuming that the 0.5% is on either side of the real value, that wouldn't exceed much more then 1% (even after adding the extra digits)

(470nF marked)
100Hz   (0.5% + 1 digits)   486.1nF (series) vs Fluke 287 @ 487nF and BK2709B @ 488.8nF
120Hz   (0.5% + 1 digits)   486.1nF (series)
1kHz   (0.5% + 1 digits)   484.8nF (series)
10kHz   (0.7% + 2 digits)   479.5nF (series)

Just over 1.3% difference...probably pretty close to spec, but most of that drop is between 1kHz and 10kHz

Electrolytic capacitors

(2.2uF marked)
100Hz   (0.5% + 1 digits)   2.1951uF (series) vs Fluke 287 @ 2.27uF and BK2709B @ 2.264uF
120Hz   (0.5% + 1 digits)   2.1849uF (series)
1kHz   (0.5% + 1 digits)   1.9745uF (series) observed dropping very slowly.  After about 1 minutes was 1.9704uF
10kHz   (0.7% + 2 digits)   1.7418uF (series)

Up to 20% off (what a sale!!!) and if the fluke is within 1% + 5 digit spec, the 2.1951uF exceeds this value...so it is starting to look wrong

(10uF marked)
100Hz   (0.5% + 1 digits)   10.306uF (series) vs Fluke 287 @ 10.56uF and BK2709B @ 10.54uF
120Hz   (0.5% + 1 digits)   10.275uF (series)
1kHz   (0.7% + 2 digits)   9.508uF (series)
10kHz   (3.7% + 3 digits)   8.648uF (series)

Over 16% variation doesn't cover the accuracy discrepancies of this range.  Fluke / BK 2709B measurements are quite different from the LCR meter now; Looks like the BK has a pretty decent capacitor meter when compared against the Fluke 287...they coincide very closely. 

(47uF marked)
100Hz   (0.7% + 2 digits)   46.29uF (series)  vs Fluke 287 @ 47.5uF and BK2709B @ 47.67uF
120Hz   (0.7% + 2 digits)   46.15uF (series) 
1kHz   (2% + 2 digits)   44.46uF (series) 
10kHz   (3.9% + 5 digits)   44.74uF (series)

Quite an improvement now, just 3.3% variation between 100Hz & 10kHz, the DMM meters measured readings are still quite different though from the LCR.

(330uF marked)
100Hz   (0.7% + 2 digits)   303.68uF (series) vs Fluke 287 @ 332uF and BK2709B @ 328.2uF
120Hz   (0.7% + 2 digits)   302.25uF (series) 
1kHz   (2% + 2 digits)   285.65uF (series) Slighly fluctuates
10kHz   not rated but a fluctuating value from 314uF-318uF

Around 6% difference between 100Hz and 1kHz (10kHz not listed in manual), the Fluke & BK 2709B agree to disagree with BK 879B.

(470uF marked)
100Hz   (2% + 2 digits)   453.6uF (series) vs Fluke 287 @ 481uf and BK2709B @ 476.8uF
120Hz   (2% + 2 digits)   453.8uF (series)
1kHz   (3.7% + 3 digits)   443.5uF (series)
10kHz   Not supported but reads at around 680-690uF fluctuating

Assuming the LCR meter is right, it is within spec across the range.  But it is quite different from the values on the digital multimeters.

I've tried checking out the waveform outputs on all the meters (when in capacitance range); it was pretty hard to gauge what was happening with the BK 2709B & Fluke 287, where as the 879B puts out a seemingly perfect sinewave at whatever frequency is selected (thanks for the recommendation on the Rigol 1052E oscilloscope Dave!, never could dream of buying something like that for such a price 10 years ago).

I guess the short question after all of this info is whether I return this meter (...blasted shipping will be expensive though), or whether I learn how to use the meter properly to measure values of capacitance via the advice of some knowledgeable guru on this forum!

[saturation]

Thanks for the time and effort to make such a detailed test and post it here.  Your LCR meter on capacitance seems fine.

What you pay for in LCR meters is the capacity to do inductance and measure ESR, DF, Q etc., if you need them.

A standard for capacitance is measurement at 1 kHz.

Its a fairly long story but its described here:

https://www.eevblog.com/forum/index.php?topic=3440.msg45910#msg45910

[alm]

To compare the results from different meters, I usually calculate the highest and lowest possible 'real' value for each meter based on their accuracy specs (so a value of 1nF with 1%+0.1% accuracy on the 100nF range could be between .89nF and 1.11nF), and see if the ranges overlap.

The large difference in the larger caps may be (partly) explained by the fact that electrolytics tend to be quite far from ideal capacitors. They often have a significant ESR, which may confuse some meters. Try putting a few ohms of resistance in series (expected for low-value electrolytics) and see how this changes the readings.

[Baqar79]

Thanks for the information, it is a bit beyond me at the moment so I'll revisit this later once I have a better idea how these LCR meters work and the reason for these results.

cheers!

[saturation]

Typical Impedance of a capacitor vs frequency:




Impedance Z, is the sum of system resistance + reactance



The reactance of a capacitor is frequency dependent:



If R is constant say as the ESR, as well as C in your test capacitor, then:

Z then is proportionate to  -1 / [2 pi f]
 
Giving an ideal curve, graph #1 at the top.  Thus, in the ideal, it doesn't matter what frequency you use to measure the capacitance, because if it follows the ideal model, any frequency can be used to calculate C based on the value of Z.

However, real capacitors vary by construction and value when actually tested with an LCR meter capable of bode plots.  So the measured value will differ, and since 2 pi f is a constant, the cause of this variance is the other things in the actual construction of the capacitor that are called parasitics; these are unintended inductance, resistance and capacitance that exists within the test capacitor:



Which then produces more complicated 'real' world models of a capacitor based on the bode plot response, here its MCAP:



or in some instances simpler:



versus the ideal model:

[Mechatrommer]

not to forgot. temperature changes capacitance too. nice explanation and graphs there saturation. however i'm not sure if this is a typo...

Z ~ -1 / 2 pi f

yes there are many model for capacitors that try to mimic real world capacitors, but i wonder what they based on in developing such models, and which model is the best! that can approximate the real capacitor?

[saturation]

ooops, yes mecha its should have been  



But the forum doesn't support printing that symbol. 

You are absolutely right, ordinary capacitors are very sensitive to temperature.

The model is based on the bode plot, the parasitic elements are added to explain the curve you see.  It could be consistent for all capacitors made by a specific brand for a specific size, and are related to the components and method of manufacture.  

For example, when you see the light blue curve flatten out early, its very indicative of ESR of about 1 ohm.  So the impedance becomes purely resistive at 5 kHz and a simple model to explain Z is a cap with series resistance.

The complicated model you see is made with one of the high end Agilent LCR meters.

Thus, getting a handheld LCR meter to measure LCR is really not that much more accurate than the one in a good DMM, except many DMM don't measure L.  What DMM don't do are the other L C parameters like DF, Q and ESR.  As you see in the other graph, ESR becomes evident in 5-10 kHz for some types of capacitors, and not others.  If its the only dominant item at that frequency, the straight line result is fine so sampling a frequency will suffice and your interpolation will be accurate.  But if other parasitics are there, is could 'dance' at another frequency, you don't know, that's why you buy a tester and you need to test a cap to the resonant frequency.

If you look at the ideal model curve the resonant frequency of typical caps are 1 MHz and up, and this is the 'end' of the capacitance test, as the curve is now dominated by parasitic inductance, and thus Z rises with frequency.  Yet only the most expensive LCR begin to sample at this frequency.  Further, most design work is now done far over 1 MHz, so testing a cap with a handheld is fraught with errors.

I think spending money on more accurate handheld LCR meters is not very cost effective.  If you want to characterize a cap well and aren't a pro with a good LCR meter, you can do far better with just your scope and the Hantek 3x25, and manually make a bode plot by making a small RC filter and put the scope across C and calculate capacitance from there.

If you're going to measure it, it will be cumbersome, but it can be far more accurate at lower cost and not mislead you like with a handheld.

not to forgot. temperature changes capacitance too. nice explanation and graphs there saturation. however i'm not sure if this is a typo...

Z ~ -1 / 2 pi f

yes there are many model for capacitors that try to mimic real world capacitors, but i wonder what they based on in developing such models, and which model is the best! that can approximate the real capacitor?

[Mechatrommer]

The model is based on the bode plot, the parasitic elements are added to explain the curve you see.  It could be consistent for all capacitors made by a specific brand for a specific size, and are related to the components and method of manufacture.  

ok. based on bode plot but from samples they are testing. maybe a good brand name and type. so maybe if we are testing different capacitors type and cheapo brand thats not comply to the model built into the meter, then maybe the parameters read out will be off? or maybe all "regular" capacitors on earth will resemble the characteristic of the "ideal" graph Z vs Hz or closer to "real world" graph Ohm vs Hz? making every measurement, usable if not accurate.

[Mechatrommer]

and not the least. from the way i see it, single freq analysis (like the OP posted), is not capable of measuring ESR and capacitance simultanously, let alone the other parameters

What you pay for in LCR meters is the capacity to do inductance and measure ESR, DF, Q etc., if you need them.

btw, i found this usefull (besides many others) for simple impedance measurement www.kennethkuhn.com/electronics/impedance_measurement.pdf which is a superset of LCR measurement.

[saturation]

Yes, mecha you are right.  Cheapo or brand, the curves you see in the data sheet can be off, so if you really need LCR accuracy, you measure it yourself to be sure of what you have.  If most capacitors are ideal, then a cheapo LCR meter will suffice, and since these are cheap it won't be as big an expense as getting the intermediate ones between $500-$2000.

But why bother when you can do all that with just the Rigol and the Hantek 3x25 and a calculator and do what a $30,000 LCR meter can, if you have the time to do it.  The calculations are all algebra.

That's why your work with the Hantek is invaluable, you can do a lot with a sine wave source to 100 MHz.

The model is based on the bode plot, the parasitic elements are added to explain the curve you see.  It could be consistent for all capacitors made by a specific brand for a specific size, and are related to the components and method of manufacture.  

ok. based on bode plot but from samples they are testing. maybe a good brand name and type. so maybe if we are testing different capacitors type and cheapo brand thats not comply to the model built into the meter, then maybe the parameters read out will be off? or maybe all "regular" capacitors on earth will resemble the characteristic of the "ideal" graph Z vs Hz or closer to "real world" graph Ohm vs Hz? making every measurement, usable if not accurate.

[Mechatrommer]

That's why your work with the Hantek is invaluable, you can do a lot with a sine wave source to 100 MHz.

yes, thats the mission actually. but first i need to get it right in the head first... knowledge.

[mobbarley]

Adding a newbie LCR meter question to this thread -

If I am making small air-core PCB coils (uH) for operation in the ~10 MHz region, what are my chances of a 1kHz / 10kHz LCR meter giving me a reasonable inductance reading?

[amspire]

It will probably be OK for a single layer air cored inductor, but you can verify by checking for resonance with a capacitor at 10MHz.

So say you made a 1uH inductor. 2*pi*F = 1/(L*C)^1/2

So for a 10MHz resonance you need a 253pF capacitor. If you make it up from a few NPO ceramic caps that you measure at 10Khz, the results should be accurate. The NPO caps behave pretty well up to around 100MHz.

Anyway for a single layer air wound inductor, the approximation:

L = (r²*n²)/(9*r + 10*l)  uH   r = radius and l = length in inches. n - no of turns

is accurate to 1% for l > 0.8*r

[mobbarley]

Thanks! Got the tuning part but as the coils are an unusual shape i'd like to actually measure them - Up till now i've been making a best guess and attempting to tune with a library of capacitors but without a good starting point it is very tedious.

[IanB]

I guess the short question after all of this info is whether I return this meter (...blasted shipping will be expensive though), or whether I learn how to use the meter properly to measure values of capacitance via the advice of some knowledgeable guru on this forum!

It is easy to assume that that every measurement you make has a single "correct" value. But in the real world, this is not the case. In the real world, the value you measure depends on how you measure it. If you are comparing your measurement to some reported value, you must make your measurement in exactly the same way as the original value was measured and reported.

In the case of real capacitors, the measured capacitance depends on the frequency. So if you change the frequency you will get a different measurement. Some capacitors are more sensitive to this than others, but electrolytic capacitors for example are very frequency sensitive. It is well understood that big electrolytics are increasingly ineffective and high frequencies, so as you increase the frequency the capacitance falls. You will find in practical circuits that a small ceramic capacitor is often put in parallel with a big electrolytic for this reason.

So in summary, if you are expecting your capacitors to have a single, "true" capacitance measurement, you are expecting something unrealistic. A perfect, ideal capacitor might behave that way, but real capacitors do not. Circuit designers have to take that into account.