EEVblog Electronics Community Forum
Electronics => Metrology => Topic started by: hydrogen maser on December 18, 2016, 01:17:28 am
-
Ok, I know that none of the capacitor testers I have are really "lab quality" but I am trying to figure out what is going on and if one of my capacitor testers is wacky. I am staring to suspect that my DER EE DE-5000 reads low because, well it seems like it reads low. Testers tried: the DE-5000 , Fluke 87 V, cheap GM328 component tester. Of those three it would seem (???) that the DE-5000 should be the most accurate if it is working correctly, but it always reads lower than other testers and lower than I expect (brand new top quality electrolytic caps > 1mf or more). Examples: 5300mf PS caps Nichicon new - read just under 6K on the fluke and cheapo tester and closer to 5K on the DE-5000. That is just an example and I consistently see lower capacity on the DE-5000 vs what I would expect and what other testers show. Also it seems like the DE-5000 is inconsistent - it is always less than I expect but sometimes it shows nothing (bad cap) - turn it off and turn it back on and it shows a capacity near what I expect, but always a little low -10/-20% of stated cap capacity. Technically this could be correct for many caps that have 20% tolerance but it is odd that it always reads on what I would think is the low side. I always run the calibration before I test a batch of caps and test at 120hz for large electrolytics on the DE-5000. Thoughts?
-
Assuming that your DE-5000 is working correctly, it should be the most accurate. It is specced at 2% - 0.30% Depending on capacitance range/value. The Fluke is specced at 1% all around. The DER actually implements a AC sine impedance ratio measurement, while the Fluke works by charging the capacitor, and determing the value that way.
-
Good chance is they're all correct, capacitance of a capacitor changes with frequency.
Try a MKP or other foil cap. these are pretty stable with capacity vs frequency and should read about the same value on all different meters.
-
^ This :-+ ^
You may find the test frequency being applied to the electrolytic capacitors by the multimeter is of a much lower and somewhat variable frequency when compared to the fixed 100 or 120 Hz that the DE-5000 is providing, the specifications for the DER LCR meter appear to cover most of the higher capacity capacitors but these values I suspect are for other capacitor types and larger electrolytics when tested on the DER-5000 can give a misleading interpretation of labelled values because of the applied test frequency, test and compare the values of other types of capacitors and you will probably find they are in much closer agreement. Also just as a guide if the LCR meter is working correctly resistance measurements should be in close proximity to those shown on the multimeters.
-
Yes, I understand that frequency is a factor and caps are specified based on the frequency and I don't know what frequency the Fluke and cheapo are using, but it still does not add up and inconsistent readings on the DE-5000 (non-repeatable outcomes) lead me to believe that something is off in the 5000 and I am looking for a methodology to confirm or deny that theory.
-
When i started i had simply an 380193 Extech LCR meter. But I needed repeatability, if not absolute accuracy and was never sure whether or not it was trustworthy in this regard. The same capacitor would differ by a percent or so, with the Extech. Subsequently, I bought a Eucol UT822C which claims to have 0.1% accuracy. The item that convinced me, however, was to buy a simple LCR calibration box:
http://www.ebay.com/itm/High-Precision-Inductance-resistor-capacitor-LRC-Calibrate-Reference-Box-/111241824712?hash=item19e68789c8 (http://www.ebay.com/itm/High-Precision-Inductance-resistor-capacitor-LRC-Calibrate-Reference-Box-/111241824712?hash=item19e68789c8)
which I used to check the Extech and UT822C. Since then, I have bought a few precision capacitors as further checks. One thing I have seen for myself is that there is a large temperature dependence to be aware of, especially for polystyrene, when measuring to this level of accuracy.
The UT822C measures at five different frequencies. The capacitors I have measured never differ by much at lower than 10KHz. For example, my 0.5uF +-0.5% poly reference capacitor measures 500.7 nF at both 100, 1K and 10 Khz. It measures 503.4 at 100KHz.
(The Extech measures it at 499.3 @1K)
Hope this helps,
Randy
-
Yes, I understand that frequency is a factor and caps are specified based on the frequency and I don't know what frequency the Fluke and cheapo are using, but it still does not add up and inconsistent readings on the DE-5000 (non-repeatable outcomes) lead me to believe that something is off in the 5000 and I am looking for a methodology to confirm or deny that theory.
The one thing I've noticed with the DE-5000(similar suspicions) is that the readings very a bit when plugging components into the slots directly, which has led me to scrubbing component leads, suspicious of the clamping force applied by the meter if yours is the same as mine the component lead clamps feel a little on the week side, just moving the components can send the meter off on a tangent.
-
I've just run a few tests on a number of higher value electrolytics with the DE-5000 alongside a couple of different multimeters and as Vgkid touched up on earlier that the methods used to determine a capacitors value does vary somewhat depending on the particular meter used, I did however have consistent readings of higher value electrolytic capacitors on the DE-5000 even though the displayed values were lower than those shown on any of the multimeters.
-
got a function gen and scope? you can use those to also check capacitance.
-
Assuming that your DE-5000 is working correctly
It appears that ship has sailed.
Give it another chance with fresh batteries and cleaned contacts.
If it insists, take it apart and check for bad caps or send it to someone who will
-
My experience is that the majority of modern caps are in fact, a bit low. They'll also read a bit higher in both value and dissipation factor when measured under bias, often hitting nominal. The method used probably matters with large value caps, a traditional 120 Hz sine wave bridge may give a different answer than other techniques, plus the amplitude matters. Leakage may affect the reading as well- try putting a resistor across the cap to simulate it.
-
Slightly off topic, I've been considering the purchase of this AC bridge pictured below which has now been listed for a month or so with no takers and at around $80 it looks like a nice unit, I suspect that it probably wont offer up any more than my DE-5000 LCR but I couldn't find any information in regards to somebody actually using one, every time I have mentioned a listing such as this in the past the items seem to disappear pretty quick under my nose or the bidding gets way out of control, any guidance would be most appreciated.
-
That Belco bridge does not look very accurate / easy to use. There is only an amplitude detection. So adjustment for higher loss caps could get slow: you don't directly know whether to adjust capacitance or loss.
I would guess modern direct reading instruments / circuits could give similar or better results.
-
If you were closer, I'd sell you my GR bridge for half that. A lot more dials!
-
Thank you both very much, I have a tendency to fall for gear that simply looks the part as opposed to being practical or ideally suited for the specific application, in my view it's still a good looking gadget nonetheless and those other big monster things with the evil eye scare the hell out of me. :)
I'm going to save the eighty bucks for something else I probably don't need. :-+
-
Good chance is they're all correct, capacitance of a capacitor changes with frequency.
Try a MKP or other foil cap. these are pretty stable with capacity vs frequency and should read about the same value on all different meters.
This. For high capacitance, you may get some conductive polymer caps which will measure almost the same among all of the meters.
Also it seems like the DE-5000 is inconsistent - it is always less than I expect but sometimes it shows nothing (bad cap) - turn it off and turn it back on and it shows a capacity near what I expect, but always a little low -10/-20% of stated cap capacity. Technically this could be correct for many caps that have 20% tolerance but it is odd that it always reads on what I would think is the low side. I always run the calibration before I test a batch of caps and test at 120hz for large electrolytics on the DE-5000. Thoughts?
Did you check at what frequency you are measuring? High frequencies wont work for high capacitance. Also I noticed that it may fail to measure some capacitors close to maximum of the range, like 10 000uf 63V ELNAs I have.
-
If you were closer, I'd sell you my GR bridge for half that. A lot more dials!
What bridge do you own?
-
If the Fluke meter charges the cap for the measurement, this is called quasi-static CV. It has its limitations just as AC impedance measurements have limitations.
The behavior that quasi-static CV exploits is that a capacitor voltage and charge Q = C*V which differentiates to dv/dt = I / C.
The limitation is the ability to accurately resolve dv/dt. If your time slices end up with too few significant digits, the measurement becomes problematic. If the capacitor is small value, it charges too fast to measure time.
Measuring leakage of the cap while measuring C with quasi-static requires careful technique and very accurate current forcing and voltage measuring which most hobbyists will not have access to.
For a very large capacitor dv/dt will be very very slow so the delta V will be difficult to resolve.
For hobbyist usage, I have a TENMA LCR meter. For job related measurements, I have used all of the expensive style gear which I cannot afford for my hobbyist lab.
-
If your main interest doesn't have to cover fractions of a pF to supercaps, it's very easy to build an accurate bridge. You might try that before laying out cash for one. See http://conradhoffman.com/cap_bridge.pdf (http://conradhoffman.com/cap_bridge.pdf)
IMO, the best used bridges are GR, but expect to do some service and calibration. The 1650A or B is popular and shouldn't cost too much. The 1608 is a heavy boat anchor, but very accurate. The 650 is ancient but can get the job done. There are some lesser known models I like, but all have limitations. Be sure they have a range that covers the caps you want to test! Many are limited to 1100 uF. Be sure you have the bench space for any of them.
-
Measuring non-polarized foil capacitors, I have compared my DE-5000 against several classic bridges (all of which have hermetically sealed standard capacitors) and secondary-standard capacitors and it always was within spec. I do worry about testing polarized electrolytic capacitors without DC bias on them, but haven't followed through with my proposed adapter to allow applying battery bias to a DUT with the DE-5000.
-
So I have been trying different caps on the DE-5000 and it looks like electrolytic caps are the only ones that "read low". NPO ceramic, poly dip, polystyrene etc read very near what they are marked, so, hmmm.. Just the way electrolytic caps and the exact way the DE-5000 behaves are? I seem to recall complaints about eletros being terrible and their only saving grace was high capacity....
-
Do realize that elco's do not have very tight tolerances. What values are you measuring?
I can measure a few caps of the same value on my 1658 tomorrow, if you like.
-
It's only a theory and many of my theories are wrong, but I believe cap makers in this day and age can control the value quite well. Since they don't want to waste expensive foil, they purposely build caps at just over the minimum tolerance value. When I buy new caps, I almost never see any that are over the nominal, as a Gaussian distribution about center would suggest. Your meter may be better than you think.
-
It's only a theory and many of my theories are wrong, but I believe cap makers in this day and age can control the value quite well. Since they don't want to waste expensive foil, they purposely build caps at just over the minimum tolerance value. When I buy new caps, I almost never see any that are over the nominal, as a Gaussian distribution about center would suggest. Your meter may be better than you think.
I agree.
As to the accuracy of meters, I also know that not all test the same way and any discrepancies between the 3 that I use are common with 1 normally the odd man out.
Mine:
Fluke 15B
Siglent SDM3055 bench meter (2nF-10k uF)
SMD Smart Tweezers (most used but not always accurate) This is the only one of these 3 that has adjustable cap test frequency.
-
Well, the manufacturers state that the capacitance is dependent on temperature, frequency and voltage. For elcos there is a quite significant drop in capacitance at higher frequency, so perhaps you are measuring at a higher frequency than the manufacturer when determining the nominal value?
http://www.kemet.com/Lists/TechnicalArticles/Attachments/191/Why%2047%20uF%20capacitor%20drops%20to%2037%20uF-%2030%20uF-%20or%20lower.pdf (http://www.kemet.com/Lists/TechnicalArticles/Attachments/191/Why%2047%20uF%20capacitor%20drops%20to%2037%20uF-%2030%20uF-%20or%20lower.pdf)
-
Curiouser and curiouser... tested a new batch of electros that just arrived today (yes, I am constantly replacing electro caps in 20+ year old obscure scientific equipment ebay scores). These read (at 120hz) very near what the stated capacity is - just a little low, but not by more than 2% to 5%. Ok I still have no idea what the deal is, but I have noticed that the higher the stated capacity the larger the discrepancy (on the low side) as a percentage the DE-5000 reads (generally speaking). And yes, technically lots of these caps can have up to 20% tolerance, so are manufactures just cheap bastards and build these larger elecro caps at the low end of the tolerance range because it is cheaper? Even the "good" brands? An EPCOS 10uf 500v cap measures just over 10uf, an exception to the trend, but that is a pretty expensive cap, hmmm....
-
Have you attempted to measure the capacitance of the polarized capacitors with a DC voltage bias applied? The bias circuit will depend on your instrument, be careful not to blow it up. I would think that somewhere between 1.5 V and 9 V (from dry batteries) would suffice.
-
There is also a difference in Cp and Cs. Cs = Cp x (1 + D^2). Large value capacitors can have a large D, which means a significantly larger Cs than Cp.
If you are using Cp mode, your readings are normal.
Cs can be extremely high on bad capacitors, and I would not trust it by itself. I'd also want to know D or Cp or ESR.
-
Slightly off topic, I've been considering the purchase of this AC bridge pictured below which has now been listed for a month or so with no takers and at around $80 it looks like a nice unit, I suspect that it probably wont offer up any more than my DE-5000 LCR but I couldn't find any information in regards to somebody actually using one, every time I have mentioned a listing such as this in the past the items seem to disappear pretty quick under my nose or the bidding gets way out of control, any guidance would be most appreciated.
I used to have one of those back in the early 70's. They were OK at the time when I was using moving coil multimeters, but the accuracy is not great. Just cheap 1% resistors inside with unknown temperature drift - no quality parts at all. The nulls were often a bit vague. I probably wouldn't bother above about $20.
-
Ok, I just can't let this one go even though it should not matter that much. Been looking at older "good brand" caps and they typically show higher capacity at or near the stated capacity on the cap, newer caps show lower. Starting to have some thoughts about what is going on, but need to wait until I get a mixed batch of caps and see what they show. I am starting to suspect two things: recent vintage caps are tested and specified to different standards vs the method the DE 5000 uses and perhaps some of the caps I have are fakes. And no, I am not willing to put a bias across the caps when measuring them because I don't want to blow up my meters.
-
Ok, I just can't let this one go even though it should not matter that much. Been looking at older "good brand" caps and they typically show higher capacity at or near the stated capacity on the cap, newer caps show lower. Starting to have some thoughts about what is going on, but need to wait until I get a mixed batch of caps and see what they show. I am starting to suspect two things: recent vintage caps are tested and specified to different standards vs the method the DE 5000 uses and perhaps some of the caps I have are fakes. And no, I am not willing to put a bias across the caps when measuring them because I don't want to blow up my meters.
A "bias tee" is used to block the DC from your AC measurement side.
-
Your meter is probably fine. I see pretty much the same thing and attribute it to several factors. As mentioned before, I think modern processes are good enough that manufacturers can build caps to the low end of spec, saving money on foil. Next, as caps age it's very common for the value to rise. Everybody thinks opposite, but they tend to go up before they go down. In the old days (how old?) I think they tried to hit the nominal value and you'd get a distribution around that, unlike today. I see a slight rise measuring caps under rated voltage (a few volts is worthless) in both value and dissipation factor, maybe 5% or so. I think the specs are for zero bias, unless the fine print says otherwise. As for Cs and Cp, if the dissipation factors are high enough at 120 Hz that it's a big difference, the cap is ready for the acoustic test- see if it makes a nice thunk when it hits the bottom of the waste basket.
-
I discuss the reason why measurement of aluminum electrolytic capacitors with a DVM gives different results than when measured with an applied sinusoidal excitation voltage: https://www.eevblog.com/forum/testgear/about-capacitance-measurement-with-dvms/msg489690/#msg489690 (https://www.eevblog.com/forum/testgear/about-capacitance-measurement-with-dvms/msg489690/#msg489690)
Also some discussion here: https://www.eevblog.com/forum/beginners/bad-cap-or-bad-measurement/msg487010/#msg487010 (https://www.eevblog.com/forum/beginners/bad-cap-or-bad-measurement/msg487010/#msg487010)
This post has a link to the same Kemet paper referred to by guenthert. The explanation for the anomaly is the etching of the foil; see figure 2 of that paper and also the plots in figure 4.
And starting here: https://www.eevblog.com/forum/testgear/mastech-ms5308-lcr-meter-with-esr-measurement-on-discount-at-the-moment/msg107489/#msg107489 (https://www.eevblog.com/forum/testgear/mastech-ms5308-lcr-meter-with-esr-measurement-on-discount-at-the-moment/msg107489/#msg107489)
-
I discuss the reason why measurement of aluminum electrolytic capacitors with a DVM gives different results than when measured with an applied sinusoidal excitation voltage: https://www.eevblog.com/forum/testgear/about-capacitance-measurement-with-dvms/msg489690/#msg489690 (https://www.eevblog.com/forum/testgear/about-capacitance-measurement-with-dvms/msg489690/#msg489690)
Also some discussion here: https://www.eevblog.com/forum/beginners/bad-cap-or-bad-measurement/msg487010/#msg487010 (https://www.eevblog.com/forum/beginners/bad-cap-or-bad-measurement/msg487010/#msg487010)
This post has a link to the same Kemet paper referred to by guenthert. The explanation for the anomaly is the etching of the foil; see figure 2 of that paper and also the plots in figure 4.
And starting here: https://www.eevblog.com/forum/testgear/mastech-ms5308-lcr-meter-with-esr-measurement-on-discount-at-the-moment/msg107489/#msg107489 (https://www.eevblog.com/forum/testgear/mastech-ms5308-lcr-meter-with-esr-measurement-on-discount-at-the-moment/msg107489/#msg107489)
You did an excellent job of measuring and comparing results of different methodologies of capacitor tests.
Since I do not own any of the gear you used, and many here may not either, it would be helpful if you cited the uncertainties of the measurements you took. If the capacitor measurement differences between the methodologies are smaller than the uncertainty of the measurements, then a definitive conclusion cannot be reached.
-
My $20 ebay AVR Transistor Tester and my JYE $8 capacitor tester (both based on AVR devices) both seem just as accurate as my multimeters. The incredibly useful AVR transistor tester also shows ESR.
-
FWIW, my above comments are from traditional measurements at 120 Hz on a sine wave bridge. High frequency measurements of physically large electrolytic caps are, IMO, fraught with problems. Even if you measure as close to the cap body as possible, the effective lead length, or loop, is still too long to say much about performance. If you want good performance at HF, you can't beat SMT. HF measurements probably make sense with smaller radials, but the minute you install them in the typical circuit board with real traces, all bets are off.
-
Although I have Mastech MS-5308, I believe its almost identical with your DE-5000, there was a discussion regarding references for hobbyist grade LCR meters like ours, we've verified and it seems like your DE-5000 is quite accurate.
An example from this thread References for LCR meters (https://www.eevblog.com/forum/projects/reference-for-lcr-or-esr-meters/), here is my MS-5308 results compared to "many-many" dollars priced >:D Wayne-Kerr 6440B Component Analyzer.
(https://www.eevblog.com/forum/projects/reference-for-lcr-or-esr-meters/?action=dlattach;attach=82692;image)
-
FWIW, my above comments are from traditional measurements at 120 Hz on a sine wave bridge. High frequency measurements of physically large electrolytic caps are, IMO, fraught with problems. Even if you measure as close to the cap body as possible, the effective lead length, or loop, is still too long to say much about performance. If you want good performance at HF, you can't beat SMT. HF measurements probably make sense with smaller radials, but the minute you install them in the typical circuit board with real traces, all bets are off.
Exactly. And, small value capacitors are problematic to measure at low frequencies.
The very large capacitors are used to smooth 120 Hz full-wave rectified AC. Why would you want to now what it is doing at MHz? If you want to extinguish RF, you need tiny capacitors very close to the source of the RF.
-
I discuss the reason why measurement of aluminum electrolytic capacitors with a DVM gives different results than when measured with an applied sinusoidal excitation voltage: https://www.eevblog.com/forum/testgear/about-capacitance-measurement-with-dvms/msg489690/#msg489690 (https://www.eevblog.com/forum/testgear/about-capacitance-measurement-with-dvms/msg489690/#msg489690)
Also some discussion here: https://www.eevblog.com/forum/beginners/bad-cap-or-bad-measurement/msg487010/#msg487010 (https://www.eevblog.com/forum/beginners/bad-cap-or-bad-measurement/msg487010/#msg487010)
This post has a link to the same Kemet paper referred to by guenthert. The explanation for the anomaly is the etching of the foil; see figure 2 of that paper and also the plots in figure 4.
And starting here: https://www.eevblog.com/forum/testgear/mastech-ms5308-lcr-meter-with-esr-measurement-on-discount-at-the-moment/msg107489/#msg107489 (https://www.eevblog.com/forum/testgear/mastech-ms5308-lcr-meter-with-esr-measurement-on-discount-at-the-moment/msg107489/#msg107489)
You did an excellent job of measuring and comparing results of different methodologies of capacitor tests.
Since I do not own any of the gear you used, and many here may not either, it would be helpful if you cited the uncertainties of the measurements you took. If the capacitor measurement differences between the methodologies are smaller than the uncertainty of the measurements, then a definitive conclusion cannot be reached.
The conclusion is that the difference in measured values among the various meters and metering methods with respect to aluminum electrolytics as discussed in this thread, is due to differences in the etching depth of the aluminum foil. On page 3 of the Kemet paper they say that etching gains as high as 300 are possible.
Since there's no way to know the nature of the foil etching for a given electrolytic without contacting the manufacturer, the variance of measurements among various meters and various electrolytics will be impossible to predict.
But we can conclude that these variances don't necessarily mean that there is anything wrong with the meters.
To compare among meters use a film capacitor, not an aluminum electrolytic.
-
FWIW, my above comments are from traditional measurements at 120 Hz on a sine wave bridge. High frequency measurements of physically large electrolytic caps are, IMO, fraught with problems. Even if you measure as close to the cap body as possible, the effective lead length, or loop, is still too long to say much about performance. If you want good performance at HF, you can't beat SMT. HF measurements probably make sense with smaller radials, but the minute you install them in the typical circuit board with real traces, all bets are off.
Exactly. And, small value capacitors are problematic to measure at low frequencies.
The very large capacitors are used to smooth 120 Hz full-wave rectified AC. Why would you want to now what it is doing at MHz? If you want to extinguish RF, you need tiny capacitors very close to the source of the RF.
The reason I measured the capacitance of large (and small) aluminum electrolytics over a wide frequency range was to show that their values can vary substantially vs. frequency when they are manufactured with etched foil. Since the measurement frequency of DVMs is not selectable, and since the apparent measurement frequency of DVMs when measuring high value electrolytics can be quite low, we can expect that an instrument using a sine wave at the conventional frequency of 120 Hz may measure a lower value than a DVM, for large electrolytics.
This is what the OP asked about, and showing measurements over a wide frequency range shows the effect of etching the foil.
The fact that the upper limit of the frequency range of the analyzer was 1 MHz was in no way intended to suggest using the capacitor at 1 MHz.
The finding that the OP's DE-5000 sometimes measures aluminum electrolytics low, has a good reason as I have shown, and shouldn't lead the user to mistrust the DE-5000. Just don't expect the measurement with sine wave excitation to be the same as the value given by a DVM.
The standard method used by manufacturers for measuring aluminum electrolytics has for years been to use a frequency of 120 Hz and an excitation voltage of .5 volts. But the value of electrolytics varies with temperature, age, manufacturing tolerances, etc., so we can't expect well defined results even using the industry standard method.
-
And, small value capacitors are problematic to measure at low frequencies.
Hmm, how small is the value and how low is the frequency? Let's say 1pF and near DC, with time constants about 1second ;) . With the right equipment it is not that problematic, unless the capacitor's own time constant is too small. 1pA current would charge a 1pF capacitor by 1V in 1second. With an electrometer you can either discharge a capacitor and measure the charge, or charge it with a known charge/current and measure the voltage.
On the subject - the capacitance of an electrolytic capacitor is not a very well defined figure, due to large levels of dielectric absorption. A capacitor with dielectric absorption is modelled as a number of series RC networks in parallel to the main capacitance with it's own ESR, so the behaviour of that combination considerably depends on the signal frequency or timing of a charge/discharge.
Cheers
Alex
-
On the subject - the capacitance of an electrolytic capacitor is not a very well defined figure, due to large levels of dielectric absorption. A capacitor with dielectric absorption is modelled as a number of series RC networks in parallel to the main capacitance with it's own ESR, so the behaviour of that combination considerably depends on the signal frequency or timing of a charge/discharge.
Cheers
Alex
This is fully discussed on page 4 et. seq. of the Kemet paper: http://www.kemet.com/Lists/TechnicalArticles/Attachments/191/Why%2047%20uF%20capacitor%20drops%20to%2037%20uF-%2030%20uF-%20or%20lower.pdf (http://www.kemet.com/Lists/TechnicalArticles/Attachments/191/Why%2047%20uF%20capacitor%20drops%20to%2037%20uF-%2030%20uF-%20or%20lower.pdf)
-
And, small value capacitors are problematic to measure at low frequencies.
Hmm, how small is the value and how low is the frequency? Let's say 1pF and near DC, with time constants about 1second ;) . With the right equipment it is not that problematic, unless the capacitor's own time constant is too small. 1pA current would charge a 1pF capacitor by 1V in 1second. With an electrometer you can either discharge a capacitor and measure the charge, or charge it with a known charge/current and measure the voltage.
On the subject - the capacitance of an electrolytic capacitor is not a very well defined figure, due to large levels of dielectric absorption. A capacitor with dielectric absorption is modelled as a number of series RC networks in parallel to the main capacitance with it's own ESR, so the behaviour of that combination considerably depends on the signal frequency or timing of a charge/discharge.
Cheers
Alex
Creating 1pA with better than 1% uncertainty is expensive. Not a practical way to measure a 10-cent 1pf capacitor.
Low freq, from my experience, is 1kHz and below. To measure 1pF at 1kHz and below is problematic.
Many applications for capacitance measurement will specify real/imaginary impedance measurement.
-
Small value caps can be measured at low frequencies, though not 1 Hz, with something like a GR 1615. Though it isn't very difficult, to get any accuracy it has to be a fully shielded measurement in a dedicated box, not some loose part using naked test leads! Considering what small value caps are used for, I'm not sure why one would try it. That situation is the high impedance end of the problem.
The low impedance problem is larger caps at high frequencies. Without a 4-wire measurement, dissipation factor (or esr) will err on the high side.
I collect bridges and in spite of having a fair number of them, none are particularly good at the frequency extremes, especially with larger values. A modern instrument like the Keysight E4980A will do a much better job from 20 Hz to 2 MHz than any of the old stuff, but at a fairly steep cost of $16k, or more with options. A swept analyzer is probably more useful, but less accurate. Horses for courses.
-
And, small value capacitors are problematic to measure at low frequencies.
Hmm, how small is the value and how low is the frequency? Let's say 1pF and near DC, with time constants about 1second ;) . With the right equipment it is not that problematic, unless the capacitor's own time constant is too small. 1pA current would charge a 1pF capacitor by 1V in 1second. With an electrometer you can either discharge a capacitor and measure the charge, or charge it with a known charge/current and measure the voltage.
On the subject - the capacitance of an electrolytic capacitor is not a very well defined figure, due to large levels of dielectric absorption. A capacitor with dielectric absorption is modelled as a number of series RC networks in parallel to the main capacitance with it's own ESR, so the behaviour of that combination considerably depends on the signal frequency or timing of a charge/discharge.
Cheers
Alex
Creating 1pA with better than 1% uncertainty is expensive. Not a practical way to measure a 10-cent 1pf capacitor.
Low freq, from my experience, is 1kHz and below. To measure 1pF at 1kHz and below is problematic.
Many applications for capacitance measurement will specify real/imaginary impedance measurement.
Just did a test on a 1.5pF ceramic cap with my old precision capacitor bridge (+/- 0.01% accuracy):
100Hz: 1.5612 pF (the main problem here was eliminating the 50Hz mains hum harmonics in the measurement)
1kHz: 1.5578pF
10kHz: 1.5555pF
As long as a capacitor is mounted in a shielded container so you can do a proper 3 terminal capacitance measurement, there is no problem making and measuring 1pF capacitances to very high accuracy.
If you were talking surface mount in production, you could probably sensibly talk about 5% tolerance. Through hole - maybe 10% is achievable. If you look at a capacitance mounted in a shielded box so that environmental effects are eliminated, you can measure 1pF to much better accuracy then 1 part in a million without difficulty if you have the right equipment. Usually, the most accurate measurements are done at 1kHz.
-
Now I've got to try a 1 pF measurement at 20 Hz!
I got a shipment of caps today for some rebuilds. I ordered 2700 uF snap caps to replace some 2300 uF parts. Sure enough and true to form, they came in at 2400 uF. All the small values, 220 uF and less, were nearly dead on, but all the larger snap caps I've ordered for a few years have come in low.
My traditional cap bridge will read higher dissipation on large caps than my Digibridge (which isn't actually a bridge). Not a huge error, but more than I'd like. They agree perfectly for lower values, say 100 uF, and I'm not sure how to verify the loss calibration for a large cap to determine which one is wrong.
One tripping point to be careful of when making higher frequency measurements- bridges will often give or let you set the measuring voltage. It's easy to choose a value that's possible for low frequencies, but would require amps of current at high frequencies. At high frequencies the amplitude simply goes down. Then one draws a graph, not realizing the test conditions have changed as the frequency changes. Even the very expensive Keysight unit isn't very good about warning when AGC is lost. It does, but you have to be looking for it.
-
Now I've got to try a 1 pF measurement at 20 Hz!
It is easier then you think :)
Imagine the capacitor suspended in a box or tube. The top end doesn't matter to much - that is the driving end where you will apply the signal between the capacitor lead and the case. At the other end the capacitor leas solders to a feedthrough that goes to a low noise opamp inverting input with w 10M resistor to the output. The non-inverting input goes to the case. Ideally you follow this with a bandpass filter tuned to the applied signal frequency.
At 20 Hz, then for every rms volt you apply to the input end, you will get 1.256mV out of the opamp with a 1pF capacitor in the box. Because the opamp is maintaining the input at zero volts, the input capacitance of the opamp and the feedthough capacitance has no effect on the measurement. At 1pF, you can probably treat the capacitor is ideal and calculate the value from the voltage.
So even at 20 Hz, and 1V input, you can measure capacitors down to fempto farads.
If you increase the voltage to 100V rms , you can measure capacitance down to the low attofarads at 20Hz. At 1pF, it is easy to make capacitors with 500V + voltage rating.
Imagine if you were measuring at the amazingly high frequency of 1kHz! You will find tuneable bandpass filters make all the difference because it will cut out most the opamp noise and input signal harmonics. It is important to measure the 20Hz level and nothing else.
Richard
-
My traditional cap bridge will read higher dissipation on large caps than my Digibridge (which isn't actually a bridge). Not a huge error, but more than I'd like. They agree perfectly for lower values, say 100 uF, and I'm not sure how to verify the loss calibration for a large cap to determine which one is wrong.
Try this test. Prepare a low value test resistor with the ability to connect to your instruments without introducing extra error due to the connection means.
I suspect the large caps you're referring to have ESRs in the neighborhood of 10s of milliohms. Perhaps one of your instruments isn't accurately measuring resistances in the range of 10s of milliohms. That could be checked by using a pure resistance.
I soldered some short pieces of 8 gauge stranded wire to an Ohmite 1%, .02 ohm resistor like this:
(https://www.eevblog.com/forum/metrology/what-capacitor-tester-do-you-believe/?action=dlattach;attach=280359)
Then I measured it with AC and DC excitation on the Hioki Impedance analyzer with this result:
(https://www.eevblog.com/forum/metrology/what-capacitor-tester-do-you-believe/?action=dlattach;attach=280361)
I also measured it on a Wayne-Kerr component analyzer with a result of 19.98 milliohms.
By using stranded wire you could flatten the strands so the flat ends could be inserted into the 4 terminal blade connectors on your Digibridge.
-
Possibly there's a way to fool the bridges, but I think they need the reactance of the capacitor to report the dissipation factor. Unlike the way a modern unit works, they won't give the esr independent of the value. A modern unit typically gets voltage, current and phase, then a uP calculates everything from that. The Digibridge will certainly give ac resistance down to a very low value, and I've checked this, but all I can do for the traditional bridges is confirm the range and dissipation resistor values, and that the range switch isn't contributing to errors. I doubt I could lash up a bench test that was much better than the bridges, though I could greatly increase the reference capacitor (we have bigger films now than back then) and maybe gain a little bit.
[rant mode=ON] Only recently, OK, maybe a couple decades, have people switched over to talking about capacitor esr. For switching supplies and such, it's the parameter of interest, and that's OK. For most other things, dissipation factor has been the preferred expression of loss. It just bugs me because nowadays if I talk about dissipation factor, nobody has a clue. I know, it's all just math. [rant mode=OFF]
-
Possibly there's a way to fool the bridges, but I think they need the reactance of the capacitor to report the dissipation factor. Unlike the way a modern unit works, they won't give the esr independent of the value. A modern unit typically gets voltage, current and phase, then a uP calculates everything from that. The Digibridge will certainly give ac resistance down to a very low value, and I've checked this, but all I can do for the traditional bridges is confirm the range and dissipation resistor values, and that the range switch isn't contributing to errors. I doubt I could lash up a bench test that was much better than the bridges, though I could greatly increase the reference capacitor (we have bigger films now than back then) and maybe gain a little bit.
I thought that the bridge could measure a pure resistance, but that's a result of my having used modern impedance analyzers so much that I begin to think the old bridges had the same flexibility. Have you asked Henry Hall about this problem? Perhaps you could make a measurement of a particular large electrolytic on several modern instruments (at work) and get a feel for what's the right value for the ESR. Having confidence in the correct value, you could then determine who is right, traditional bridge or Digibridge.
[rant mode=ON] Only recently, OK, maybe a couple decades, have people switched over to talking about capacitor esr. For switching supplies and such, it's the parameter of interest, and that's OK. For most other things, dissipation factor has been the preferred expression of loss. It just bugs me because nowadays if I talk about dissipation factor, nobody has a clue. I know, it's all just math. [rant mode=OFF]
I think this is partly due to the increasing use of switchers as you say, but there is another reason. Since the realization that ESR could be measured at 100 kHz by assuming the reactance of a moderate to large sized electrolytic is less than the ESR and then applying a pulse of current and measuring the voltage drop, low cost so-called ESR meters have become common. These meters "measure" ESR directly, hence the popularity of ESR as the parameter to measure. Unfortunately, the limitations of these devices are often not understood.
I have thought that a "D" meter that could measure dissipation factor directly would be a nifty item if a design could be implemented similar in cost to the "ESR" meters.
-
Never thought about a D-meter. Since D is just the tangent of 90 degrees minus the phase angle of the cap, and you can build a decent phase meter out of a few gates, it might not be too hard. OTOH, since the actual circuitry of almost anything is trivial compared to the amount of work to build it and put it in a chassis, and given the DDS chips and such we now have, I've wondered if one could do a DIY version of the high end Keysight or Chroma (formerly QuadTech) wide range LCR meters.
-
I have a big, clunky HP 4262A and an Agilent 1733C I believe, that I bought here. They both read the same to .1 of anything, which surprises me. You really have to be careful though, with the leads when reading .1pF level accuracy because just the slightest bend will throw them off.