SO.... What Capacitance are your Resistors...?

[The Electrician]

If I measure the capacitance of several resistors with an impedance analyzer, I get results like this.  There is an effect known as the "Boella effect" that causes the parasitic capacitance of resistors to vary with frequency. See: http://www.edn.com/electronics-blogs/anablog/4309676/the-Boella-effect and https://www.bing.com/search?q=boella+effect&qs=n&form=QBLH&pq=boella+effect&sc=2-13&sp=-1&sk=&ghc=1&cvid=C4F6F6882C064753A2EF5296043B04A9

Here's the measured capacitance of a 1 watt precision molded 1 megohm resistor vs. frequency.  The bottom of the scale is 1/10 pF:



Here is the measurement on a 1 watt carbon comp resistor:



And here is the measurement on a 1/2 watt metal film resistor:

[The Electrician]

In the previous post, I showed that the parasitic capacitance of a small metal film resistor was much less than 1 pF.

When I measure the capacitance of the carbon comp resistor with its larger parasitic capacitance, I get a value of 7 nF on a Fluke 189.

The measured value for the metal film resistor is 9 nF.  This is interesting; the DMM's measured value of the resistor with smaller parasitic capacitance is larger than the resistor with larger parasitic capacitance.

The waveforms from the DMM when making these measurements vary a lot with different resistors.  Some of this variation is probably due to the differences in parasitic capacitance, and some is due to other factors.

But one thing I do see is large variation in capacitance as measured by the DMM method with respect to apparent small physical variations in the resistors.

These values gotten from the DMM measurement are mostly meaningless.  They certainly don't correlate with the actual parasitic capacitance as measured with an impedance analyzer.

An impedance analyzer can correctly measure the parasitic capacitance of resistors and inductors, but a DMM's capacitance range can't; it's strictly intended to measure the capacitance of devices whose dominant property is capacitance.

[vk6zgo]

Wraper - I think you're getting a bit too obsessed with the 'capacitance' thing a bit too much. Perhaps it's my fault for titling it so precociously lke i did.

Yes,if you had said something like this at the outset:-

"I am getting funny readings if I try to read the parasitic capacitance of some resistors using the "C" range on my TX1 DMM.
I know it isn't really capacitance,as the values are much higher than expected.
My new Chinese metal film resistors read "400nF,whereas my older resistors from other sources read 4nF.

What kind of resistor characteristic would cause :-
(a) An obviously false capacitance reading
(b)One which varies so widely."

We could then have answered "We don't know!" & gone home.

Back in the day,the device of choice to look for strange effects on resistors,would probably have been a "Q" meter,with an LCR bridge as second choice.
Today it may be a VNA,with an LCR meter as second choice.
The last actual LCR meter I used was an '80s version,but it was very capable.

[Jay_Diddy_B]

Hi Group,

Here are some pictures from my experiments:

First I have a Fluke 87 connected to a 0.47uF Film capacitor. The more experienced readers will recognize the Phillips (Mullard) capacitor. The meter reads the right value.



Then I connected a 1M Ohm metal film resistor. The meter is reading 0.147 uF



So I tried the 1M resistor in series with the capacitor and I got 0.098uF:




If I check (C₁ x C₂) / (C₁ + C₂)

I get (450n x 147n) / (450n + 147n) = 110nF

This is close to the 0.098uF measured.

I then tried the parallel combination:



I got 0.59uF


I then tried the same experiments with a Fluke 289:

First the 470nF capacitor:



I tried the 1M resistor but I did not get a stable reading, so there is no photograph.


I then tried the 1M and 470nF in series:



I got a reasonable reading for the capacitor.

HP4274A LCR Meter

The HP 4274A is a very accurate LCR Meter that was probably introduced in the late 1970's but a very capable instrument.

Here is the open circuit reading after zero calibration:



That is zero fF. a fF is a femto Farad 1^-15 F or 1/1000 of a pF

Here is the same meter with the 1M Resistor in the test fixture.



The value is 160 fF or 0.16pF. This is very similar to the 0.11pF predicted by Ian.M in an earlier post. There are reasons why the value is higher. One is the metal end caps on the resistor construction reduce the distance.

Comparing the measurements from the various meters, I would conclude that even some very good DMMs are fooled by a high value parallel resistor into giving erroneous readings. The more specialized LCR meter give the correct answer. The correct answer matches the calculated value of the capacitance from the physical dimensions and the material properties.

I learned from this thread to be careful measuring capacitors with a DMM.

Regards,

Jay_Diddy_B

[The Electrician]

Jay_Diddy_B, compare what I got as shown in the third image of reply #52.  Around .19 to .16 pF over the swept frequency range for the metal film.

It's clear that DMM capacitance ranges are only suitable for measuring actual capacitors.  I posted elsewhere about DMM capacitance measurements.  Let me find the post and I'll edit in a link.

Link at: https://www.eevblog.com/forum/testgear/about-capacitance-measurement-with-dvms/msg489690/#msg489690

I asked for other people to try some of the measurements I described, but everybody's too busy with other stuff!   

[jaunty]

interesting - all the replies i've gotten since first posting this - sorry I've been distracted with other work stuff since first posting about this topic ... i've looked into a few more results with 'other' (vishay, welwyn wirewound) resistors since then and found i get a null capacitance result ... though of course that's all the result from a DMM - but that fact is not uninteresting at any rate

HOWEVER

i find it interesting the replies i'm getting on the topic ... from the perspective that capacitance is generated by a set of relationships between materials that have a capacitor like structure (think of an RCA jack maybe - or perhaps even a silicon junction on a transistor etc) i don't find it SOOOO surprising in principle at least that some electronic products CAN have capacitance or capacitance like properties ... i just wanted to ask to see if this was a 'known' issue or simply the result of the measuring or measurement method etc ... seems the topic is still open ... might make a great topic for a research paper.

[T3sl4co1l]

Yeah a DMM isn't going to tell you that, they require low leakage to register anything as a capacitance.  That's a consequence of how they measure it, an RC oscillator at low frequency and low current, basically, usually.

More generally, any impedance that varies against frequency is inductive or capacitive, to some extent.  Even if you only know |Z|, you can recover the RLC equivalent because the network must be causal (no negative inductance, resistance or capacitance), and (to be perfectly technical), |Z| is the magnitude of a continuous analytic complex function Z(f) for which the Kramers-Kronig relations apply.

The only real way to test a component, is to sample that function at enough points -- measure the impedance or resistance or reactance at enough frequencies -- to derive the network to the desired accuracy.

If you know already that a given component should be simple -- a carbon film resistor, or a ceramic capacitor, for example -- it's probably okay to sample a few points, and extrapolate the 2nd order (series/parallel RLC) model from that.

Also given that the component is simple, like these examples, you can guess that ESL should be on the order of the physical length, and EPC on the order of the physical width.  If the resistance (ESR/EPR) is large with respect to the aspect ratio times the impedance of free space (377 ohms), we expect C to dominate.  If smaller, then L.

Tim

[mzzj]

If I stick multimeter probes to wet soil and try to measure the circumference of earth based on that sometimes I get up to 1 kilogram difference in the  measured distance between earth and sun.

In other words: Like about 100 people have tried to explain, You have wrong tool for the job. GIGO

[MK14]

sorry I've been distracted with other work stuff since first posting about this topic

Around 2.5 years worth of distractions!

[MK14]

The following article on the parasitic capacitance of resistors.
http://www.resistorguide.com/resistor-capacitance/
Seems to be saying that

Foil resistors, on the other hand, have superior characteristics for high-frequency use, with the capacitance usually less than 0.05 pF which makes them cope with frequencies up to 100 MHz.

Read more http://www.resistorguide.com/resistor-capacitance/

So, if you tried to test that resistor type, you would be trying to detect 0.05pF, probably using meter probes with lots of pF, and possibly with your hands holding it, adding even more pF. (Not to mention, the resistors resistance, will completely mess up the reading, as well, typically).

Analogy:
You are putting a 0.001g postage stamp (ok, it probably is more like 1g, really) onto your Stones and Pounds, weighing scale. Standing on the scales, hoping to see the 0.001g postage stamp weight, on top of your, 5 to 50+ stone weight.

The combination of the resistors resistance, and trying to accurately measure just <0.05pF, is a difficult thing to measure accurately.

As a side note, will it be 2020, or 2021, when you reply ?

[T3sl4co1l]

A very basic introduction, but falls well short of a complete description.

Curiously, the diagram shows ESL, which is promptly forgotten in the following equations, and the ratio of R to L or C is not at all noted in the text (scanning it, anyway).

Such generalizations are not helpful.

Again, capacitance dominates at high values.  It is hard to find a wirewound resistor that's bad as low as 50kHz, but perhaps a high value (~100k?) precision part would be.  Those aren't very common nowadays.

Capacitance absolutely does not dominate at low values.  At least, I will be impressed -- and very interested -- if you can present to me a 0.01 ohm resistor that is capacitive-dominant at a few MHz and costs the same as any other.

Tim

[MK14]

I wonder how the resistor manufacturer's, measure the tiny capacitance, of a tiny surface mount resistor, of a type which has very low stray/parasitic capacitance ?
Especially over a very wide range of resistor values, such as 0.1 Ohms up to 10 Megaohms.

It's tricky to make contact with it, without (typically adding) to the stray capacitance.

Especially if it is a smaller resistor manufacturer, who can't afford to buy expensive ($12,000), test equipment. E.g. A small plant in India, making resistors. But maybe they can or should afford the right equipment.

Another possible snag, is ignoring the stray inductance, which will also attempt to hinder accurate measurement of the resistors stray capacitance.

But I can believe there are tried and tested methods and/or old journals with the answers.

[ArthurDent]

Here is a simple way to check this ‘resistor is really a capacitor theory’. You can use a resistive divider on D.C. to get a ratio voltage and this is common in metrology. You could also use a resistor divider on A.C. like in a scope probe but you might have to adjust compensation depending on the use and frequency. You can use a capacitive divider only on A.C. and this is pretty common as well.

If you use 2 equal resistors as in figure 1, the voltage, V1, at the output will be V/2. If you use a signal generator and use A.C. with equal capacitors as in figure 2 then the output voltage, V1, will also be V/2.

If you ‘think’ you have a resistor that actually has a large capacitance then if you choose a real capacitor of the same value for the lower capacitor in the divider circuit in figure 3, then if the resistor that you think is a capacitor is placed in the top half of the divider, the output should still be, V1, of V/2.

Try it with a generator at various frequencies from 1Khz to 1Mhz and let me know what you get for V1.

[MK14]

One issue that comes to mind. Is that the resistors stray capacitance, is probably rather sensitive to frequency (and probably other factors, such as voltage).
So the method you described, may limit the choice of frequencies.

E.g. low frequencies, may be unsuitable.

I guess the resistors datasheets could do with a typical stray capacitance value vs frequency (and voltage if it gives much change, which it is known to with some capacitor types, anyway).

[ArthurDent]

MY14 - "Is that the resistors stray capacitance, is probably rather sensitive to frequency ..."

I realize that and just mentioned using 1Khz to 1Mhz for checking to eliminate both high and low frequencies. This was kind of a wild ass guess and I thought this range might work for testing. However, I think that the output voltage from figure 3 will probably be too low to measure because the ratio of capacitance will be so large.

[MK14]

MY14 - "Is that the resistors stray capacitance, is probably rather sensitive to frequency ..."

I realize that and just mentioned using 1Khz to 1Mhz for checking to eliminate both high and low frequencies. This was kind of a wild ass guess and I thought this range might work for testing. However, I think that the output voltage from figure 3 will probably be too low to measure because the ratio of capacitance will be so large.

I bet Marvin the Robot, brain the size of a planet, would know how to measure it.

0<.05pF expected stray capacitance from some resistor types, mentioned above. Would present a significant challenge.

Actually, no. This thread seems to have the answers.

See a post, many posts above this one.
It doesn't seem to like me trying to link to it directly.
Please goto post "Reply #55"


It shows a HP LCR meter, being used to measure down to 1/1000ths of a pF.

[ArthurDent]

"I bet Marvin the Robot, brain the size of a planet, would know how to measure it."

He's too depressed to solve so menial a problem. 

[T3sl4co1l]

One issue that comes to mind. Is that the resistors stray capacitance, is probably rather sensitive to frequency (and probably other factors, such as voltage).

If you use a series equivalent (R + C), you'll have a f^2 dependency.

The parallel equivalent is the more accurate first order model.

The impedance is measured on a fixture with a VNA.  Pad capacitance, connecting traces and all that are de-embedded, leaving only the contributions due to the component itself.

A 49.9 ohm 1206 chip resistor will fall over at >3GHz.  Yes, a high bandwidth is needed to test some components.

Tim

[MK14]

If you use a series equivalent (R + C), you'll have a f^2 dependency.

The parallel equivalent is the more accurate first order model.

The impedance is measured on a fixture with a VNA.  Pad capacitance, connecting traces and all that are de-embedded, leaving only the contributions due to the component itself.

A 49.9 ohm 1206 chip resistor will fall over at >3GHz.  Yes, a high bandwidth is needed to test some components.

Tim

Thanks.
Yes, needing >3GHz, is a very surprising fact. Intuitively, one would have expected, more like 10KHz (as in simple capacitance meters, such as some DMMs), to do the job of component testing and measurement.  But that is because intuition, can get it badly wrong at times.

In this case intuition is likely to be getting it wrong, because of NOT taking into account the tiny signals the very low stray capacitance gives (stray sub pF capacitance, does not do much at 10KHz, in most circuits). Compared to the big dominance of other factors, until they are removed/minimised using the techniques described in your post.