Measuring a pickup's capacitance with an LCR meter

[antigua]

As you might be aware, it's generally thought that an LCR meter won't accurately report a pickup's intrinsic capacitance, since unlike a real capacitor, a guitar pickup has electrical continuity. A member on another pickup forum music-electronics-forum.com/t46404-post492764/#post492764 named Helmoltz, who is from Germany and apparently has worked with Helmuth Lemme, informed me that you can measure a pickup's capacitance with an LCR meter than is capable of testing at 100kHz. Neither the Extech 380193 nor the the Global Specialties LCR-58 I have on hand went that high, but the DER EE DE-5000 does, and it only costs about $100, so I ordered one and gave this a try. 

...and it works! 

I measured a Seymour Duncan SSL-1 neck pickup, and got the exact same capacitance value 103pF , reported here a few months ago guitarnuts2.proboards.com/thread/7745/seymour-duncan-ssl-analysis-review

Here's a pic of the measurement:





Same story with a Seymour Duncan SH-1N or "'59 neck", measured 98pF here guitarnuts2.proboards.com/thread/8034/seymour-duncan-neck-analysis-review , showing 99pF with the DE-5000




100kHz was the only test frequency that worked. For the SSL-1, at 10kHz, it shows 4.86pF in PAR, 39pF in SET, both of which are way off, and lower frequncies are increasingly off, showing capacitance values in the nano-farad range. At 100kHz, switching between SER and PAL shows the same value to within 1pF, for bot the SSL-1 and the SH-1N, so while I assume there is a good reason to use PAL mode, SER appears to be very close as well.

Note in the picture that I have the lead wires far apart from each other, and not touching, as when they're brought close together, you instantly see the capacitance rise by a few picofarads.

I'm astonished the two measurements between the DE-5000 and the bode plot calculation methods are in such agreement, since the bode plot method involves such impressions as guesstimating the precise frequency of resonance based on the digitized bode plot, and taking of 10pF capacitance to account for the probe capacitance. 

Since it appears the DE-5000 reliably measures capacitance as well as inductance (only SER mode is suitable for inductance, and the lowest frequency setting of 100Hz is most accurate), and that these value agree with those derived from impedance plots, it's therefore possible to calculate the resonant peak of the pickup using nothing more than the DE-5000.

The resonant peak itself is rarely useful, since in an electric guitar, a high degree of additional capacitance will bring the true resonant frequency down to a much lower value. For the surveying of pickups I've done, I also include resonant peaks with a 470pF capacitor across the pickup, which is intended to represent a guitar cable, and give a more realistic representation of what the pickup will do in situ. I will try putting a 470pF cap across the pickup, then I'll measure the capacitance again, and see how closely the capacitances sum, as well as determine how closely the calculated resonant peak "with load" comes to the peak measured with an oscilloscope. If it turns out the DE-5000 can effectively acquire all these data points, then the only big advantage that would remain for bode plot measurements is in determining how much resonant damping occurs due to eddy currents.

I would think that having a quick and easy way to determine the capacitance would be a boon to pickups winders, since it's claimed that laying the wire in different ways results in different capacitances, this would offer an instant means of analysis, requiring no tedious oscilloscope measurements and calculations. 

When the sun comes back up, I will do some more testing with this. 

[JohnH]

Very interesting, and its evidence in favour of LRC models for pickups.

With respect to impedances, if you take a nominal single coil with 2.5H and 100pF, at 100kHz, the impedance of the inductive part is very close to 100x that of the capacitive part. 1.6M vs 16k. So at that frequency, if the meter is expecting a pure capacitance, its unlikely to be confused by reality by more than a % or so.

That SSL is a fairly pure simple pickup, and effective load to add a bit of damping to it is still many times the capacitive impedance, if using 100kHz. It would be intetesting to try the same correlation on a more complex unit such as a covered PAF, where eddy losses are more dominant.

[antigua]

Apr 16, 2018 2:36:15 GMT -5 JohnH said:

With respect to impedances, if you take a nominal single coil with 2.5H and 100pF, at 100kHz, the impedance of the inductive part is very close to 100x that of the capacitive part. 1.6M vs 16k. So at that frequency, if the meter is expecting a pure capacitance, its unlikely to be confused by reality by more than a % or so.

Why wouldn't 10kHz suffice since the pickup is capacitive at that frequency? I'm surprised that passing the resonant peak isn't enough, that it has to go waaaay past it in order to achieve accuracy.

Apr 16, 2018 2:36:15 GMT -5 JohnH said:

That SSL is a fairly pure simple pickup, and effective load to add a bit of damping to it is still many times the capacitive impedance, if using 100kHz. It would be intetesting to try the same correlation on a more complex unit such as a covered PAF, where eddy losses are more dominant.

I did test a '59 humbucker as well as the SSL-1, there is a pic above, and it showed the same accurate capacitance measurement as well, for both SER and PAL modes. Since the main difference between these pickups is equivalent parallel resistance via eddy currents, it appears not to be detrimental to the LCR meter's ability to determine capacitance at 100kHz.

[JohnH]

At 10khz, the L and C impedances are about equal magnitude, but 180 degrees out of phase. ie, around resonance. The meter will see something very different to just the capacitance.

Good news about the humbucker readings. That 100khz capacitor impedance is still a lot lower than that of any effective equivalent damping resistor. And in this case, the relative phase difference is 90 degrees. Pythagoras helps to make the combination almost just like the cap alone. Eg, if the cap is only 10x more conductive than a damping resistance, the net result is sqrt(10^2 + 1^2) = 10.05, just a 0.5% change.

If you consider the 6-part pickup model, then the three damping components are a resistor and an inductor with series resistor. At 100kHz these will also all be insignificant compared to the main capacitance.

[antigua]

We've hit a snag with measuring pickup capacitance @100khz with the DE-5000, some pickups are showing different capacitances than what was measured based on resonant peak and inductance. Even though the SSL-1 measured the same, a Fender Fat 50 is showing very different. The calculated C based on peak freq. was around 120pF, but the DE-5000 shows closer to 60pF; very different. 

The reason appears to be apparent with a high frequency bode plot. I made a direct impedance plot, driving the coil directly, and you can see that, for some strange reason, the Fat 50 has a secondary resonance at 98.2kHz, almost right on top of the test frequency. 




The SSL-1 has a secondary resonance also, but it's at 153mHz far away from the 100mHz test frequency. 



So if the pickup is behaving capacitively, there should be a -6dB oct slope, but these resonances in the area of the test frequency are obviously not that. The pickup becomes inductive again, thereby causing the LCR meter's calculation to be incorrect. The major practical problem is that it's impossible to know if a pickup has these high frequency resonances without first plotting the impedance like this. 

[JohnH]

Super interesting, and I'm as always, impressed with your diligent scientific methods.

What is happening is that when you hit a resonance, the resonant circuit jumps to a high impedance (this is what LC parallel circuits do at resonance). The meter, which has hit this frequency with its test sugnal, thinks it is just a capacitor with a high impedance, and so it interprets it as a low capacitance.

How about you run two measurements, one as you have, and the other with an added known extra capacitance in parallel of say about 50pF. This will move the extra high-end resonances by maybe 20%. If neither of these hit the 100kHz, you should be able to get two readings differing by the extra 50pF. But if one of them does hit 100kHz, then when you interpret the pickup capacitance from the two reading, use the higher value which will be 'off resonance'.

You could try with several caps to be sure.

[antigua]

Apr 21, 2018 16:51:53 GMT -5 JohnH said:

Super interesting, and I'm as always, impressed with your diligent scientific methods.

What is happening is that when you hit a resonance, the resonant circuit jumps to a high impedance (this is what LC parallel circuits do at resonance). The meter, which has hit this frequency with its test sugnal, thinks it is just a capacitor with a high impedance, and so it interprets it as a low capacitance.

How about you run two measurements, one as you have, and the other with an added known extra capacitance in parallel of say about 50pF. This will move the extra high-end resonances by maybe 20%. If neither of these hit the 100kHz, you should be able to get two readings differing by the extra 50pF. But if one of them does hit 100kHz, then when you interpret the pickup capacitance from the two reading, use the higher value which will be 'off resonance'.

You could try with several caps to be sure.

I tried the Fat 50 with a 470pF cap in parallel, the anomalous frequency drop slightly from around 98kHz to about 95kHz, and the Q dropped. Higher value caps cause it to disappear completely. 



It testing a few other pickups shows that sometimes there is no anomalous peak, but most often there is at least one, and it's frequency tends to be at or above 100 kHz. Helmholtz form the other forum plotted several Strat pickups also:

 

The cause the of the secondary resonance is a mystery. The most plausible theory brought up, IMO, is an inter-winding short that results in a coil within a coil, which the larger coil inductively couples with. That raises other questions though, such as what causes it to happen, why does there tend to be only on of them and why is located around 100kHz?

It also looks like the anomalies might be more common with high volume machine wound single coils as opposed to handed guided pickups, but a larger pool of pickups has to be tested before there is any certainty about that. 

[antigua]

It appears that this is in fact an unreliable means of measuring capacitance of a guitar pickup, because at a test frequency of 100kHz, the pickup might not be 100% capacitive. The reason why this is is not certain, but in all likelihood has to do with and uneven distributed capacitance as a result of uneven coil winding. Most machine wound coils show minimal anomalies, while hand wound coils tend to show more of them. 

The most scattered coils I managed to make myself was so loose that it would be impossible to place a cover over the top of the pickup, and even if you could, it would make a better microphone than guitar pickup. It produced the plot below:




Two prominent secondary resonances can be seen, one at 86kHz and another at 153kHz. 

My suspicion is that the distributed capacitance becomes especially imbalanced when two far ends of the coil manage to come close together capacitively. For example, if a coil is being wound, and  it forms a depression at a far edge of the bobbin, and then the depression eventually gets filled, the wire that falls into the depression creates a capacitively link between the portion of coil at the top and bottom of the depression. To put this to the test, I'm planning on creating a test pickup with a wildly lopsided coil in order to hopefully induce the effect.

A similar effect is seen with series humbuckers. They have a secondary resonance which is usually at a rather low frequency, and is attributed to the fact that the whole of the coil is broken up over two halves, which disrupts the lumped capacitance of the whole.  In either case, this secondary peak   can be modeled be putting some small capacitance across some portion of the coil:

[markm]

Gentlemen, what an excellent discussion here.
I've given capacitive coupling a lot of thought, and designed my own CNC pickup winding machine capable of producing various winding patterns to minimize capacitance. A number of the issues mentioned here confirm my own theories.
Happy to share more about my project but don't want to be blatantly promotional in my first posting.
I've got a little bit of a background in physics but never got trained up proper with electronics, and haven't had access to quality measurement equipment so as to learn the finer aspects of practical testing and measurement.
Unfortunately I just have a very cheap LCR meter (pickup inductance is too high to measure). While my confidence in the absolute values has gone down a bit after reading some of the comments, I will try to presume that at least the relative values I measure between pickups have some meaning. One of my pickups measured 40% less capacitance than a comparable standard helical wound coil.
I am interested in what type of equipment is recommended for doing LCR measurements as discussed here (I really like Antiqua's graphs), as well as understanding practical frequency response testing methods.

[markm]

Were the graphs done from the DE-5000?

[antigua]

Aug 10, 2019 22:01:24 GMT -5 markm said:

Gentlemen, what an excellent discussion here.
I've given capacitive coupling a lot of thought, and designed my own CNC pickup winding machine capable of producing various winding patterns to minimize capacitance. A number of the issues mentioned here confirm my own theories.
Happy to share more about my project but don't want to be blatantly promotional in my first posting.
I've got a little bit of a background in physics but never got trained up proper with electronics, and haven't had access to quality measurement equipment so as to learn the finer aspects of practical testing and measurement.
Unfortunately I just have a very cheap LCR meter (pickup inductance is too high to measure). While my confidence in the absolute values has gone down a bit after reading some of the comments, I will try to presume that at least the relative values I measure between pickups have some meaning. One of my pickups measured 40% less capacitance than a comparable standard helical wound coil.
I am interested in what type of equipment is recommended for doing LCR measurements as discussed here (I really like Antiqua's graphs), as well as understanding practical frequency response testing methods.

The graphs are created with a Velleman PCSU200 (often in tandem with stratotart's integrator, but the graphs in this thread are just directly with the pickup).  DE-5000 is a bit more useful. Both pieces of hardware cost around $100 - $120, I'd buy both, but I'd get the DE-5000 first. In Ls mode at 100Hz, you can get an accurate inductance measure, and in Cp mode at 100kHz, you can often get a good capacitance measure, especially with neat machine wound coils, and it also has a DC resistance mode as well.


The main benefit of the Vellemen plot is that you can get a precise capacitance by calculating if from the peak frequency and the inductance (which is easiest to get with an LCR meter like the DE-5000), and you can determine the amount of eddy current losses caused be a pickup's conductive parts by observing how call the peak amplitude is compared to other pickups, but neither the peak amplitude nor the capacitance is as important as the inductance, which is easiest to acquire with the DE-5000. 

If you are testing winding patters, I would create bode plots in order to look for differences between the coils you're making. The simple LCR measurements of the DE-5000 are most useful the pickup is of a known type, like a generic Strat or PAF pickup. If you are doing something exotic, then the plot detail can reveal interesting things that you would want to be aware of, such as secondary resonances, as seen in the "homemade #6" plot above.

[bbgo]

Hey tell me about your model. I'm trying to model 2 single coils in series in Spice and I'm not sure how you got the two sources working.

[antigua]

Mar 13, 2023 21:00:48 GMT -5 bbgo said:

Hey tell me about your model. I'm trying to model 2 single coils in series in Spice and I'm not sure how you got the two sources working.

It's unintuitive, but the AC sources have an AC amplitude of 0.2, although it doesn't matter what the value is here, and then you click the running man icon at the top to start a simulation, select the "AC Analysis" tab



and then input these values, which make sense for audio analysis and give decent resolution of the response plot, and click "OK" and it will output "ac lin 500 10 20000" into the schematic somewhere. When you click the running man icon a second time, it will process the plot, you have to click on a node in the schematic to see what the response is at that point. Clicking on the resistor that signifies the amplifier input makes the most sense in this case. 

[dylanhunt]

Thanks Antigua!  I am always excited to see posts from you, here or elsewhere. [And I am just realizing that I've read the original post before and that it is just the thread with the hypothesis about why <100kHz cap results and bode plot results might not be reliable (and about graphing methods) that is from the last few days.  Well, the following might still be relevant to that hypothesis.]

Quoting: "I'm astonished the two measurements between the DE-5000 and the bode plot calculation methods are in such agreement, since the bode plot method involves such impressions as guesstimating the precise frequency of resonance based on the digitized bode plot, and taking of 10pF capacitance to account for the probe capacitance."

Somebody might have already said this, in more technical terms, but my understanding is that the LCR meter is not actually measuring capacitance (some do, I imagine: digital ones don't?).  Rather, it is measuring something and then is able to do a host of calculations based on the sensed values.  This wonderful Aussie blogger (David E. Jones on EEVBlog.com) makes that point at the timestamp of the video below (and elaborates throughout the video).  Earlier in the video, Dave says that the LCR meter measures AC voltage and current going (from the meter's AC source) through a circuit at two phase angles (0- and 90-degrees) and then calculates series resistance and series reactance, and then extrapolates the values of interest, such as capacitance:

[ms]

Think of the LCR meter as measuring the ratio of voltage to current at some frequency f, both amplitude and phase.  If you have told it that it is measuring a capacitor, it assumes that you are right and then uses the values it measures to compute the capacitance and a loss factor. Same for inductance, and so on.  So the capacitance measurement of a pickup must be made at a high enough frequency so that the inductance (in parallel) does not affect the measurement very much.