An alternative way to consider eddy currents in a pickup?

[perfboardpatcher]

I made this discovery while I was playing with the numbers in my spreadsheet I. The spreadsheet was originally used to put two pickups in parallel but it's also capable of doing the 1.5db scoop due to eddy currents. Although it works in terms of computation I'm not certain whether it's justifiable to create a voltage source and inductor out of thin air!

Beside the voltage source for the pickup there is a voltage source for the eddy currents. Both voltage sources have the same magnitude but are out of phase with each other.
I took the data of the DiMarzio Humbucker From Hell as provided by Antigua as example. I used the data/plots for the unloaded pickup.

[JohnH]

I can fully believe that what you show can work. I've also been modelling pickups with the same arrangenent of 6 components, but without the 2nd voltage source. It does a good job of matching the test plots at different loads, once the optimum values are worked out.

[antigua]

This is a great insight. An AC source of opposite phase is theoretically consistent with how eddy currents work. I tended to stick to a single AC source since there is only the moving strings, and the eddy currents are "passive" in that respect, but if you think less about the guitar string and more about having an arbitrary flux change in the vicinity of the coil, then two AC sources is justified.

There might actually be justification for both models at once, because you have both "active" (as defined above) and "passive" (current model) eddy current actions; the active action would correspond to eddys caused by the moving string, where as passive eddys would correspond to those caused by the magnetic field of the coil itself.

This difference can be seen with Telecaster pickups; with the neck, you have the cover in between the coil and string, producing eddys in relation to the moving string, or "active", and those losses are very large. But with the bridge, there is no cover, but there is a base plate. The base plate is far from the strings, but close to the coil, so those eddy losses would be relative to the coil's alternating magnetism, or "passive".

I'll try working with up with LTSPice later and see if I can get the same results there.

[JohnH]

If you want to experiment, you could take one of the 6-part data sets from the current Guitarfreak as a starting point. Then add the second source. Then adjust values to get the best match to loaded and unloaded tests, then see if the matches are improved.

Ive so far avoided a second source, since the relative magnitude of it adds another variable, and so far, all the pickups that Ive processed have got a good match with just one source.

But there could be truth in the two source model, and it could probably also be shown that by analysis, the model can be condensed to an equivalent single source with different component values.

[antigua]

I think the magnitude of the second source would be a summation of eddy current flux, so it would be a coefficient of the primary AC magnitude, that takes into consideration the thickness, conductivity and permeability of the cover, as well as the geometry of the cover. If this is all proven to be true, maybe that coefficient could be determined, and I suspect that it would apply to a broad range of pickups, since so many of them use the same covers.

I suspect that the existing model is adequate in modelling eddy current losses because even though the eddy currents might be said to have distinct causes, the result is mostly the same: frequency dependent impedance. The major difference was that the "passive" eddy currents that interact with the coil caused a reduction of inductance, where as the "active" eddys caused by the moving guitar string did not effect the inductance, and this distinction had an ever so slight impact on the resonant peak.

[antigua]

When I first looked at the diagram I didn't realize it was the existing eddy current model with an added AC source in reverse.

I'm not sure how a reverse AC fits into the picture, but it does seem to enable much more dramatic losses without changing anything else:

Without negative AC:




With added in negactive AC, where AC = -AC * 1.5




I checked to see if these parallel eddy models (without negative AC) can be stacked, and sure enough, it causes a similar compound scoop. 




Maybe this is representative of a scenario where you have multiple sources of eddy loss, as is the case with a humbucker assembly. For example, maybe the first set represents the cover, the second set represents the pole pieces and the third set represents the base plate. 


Another thing to think about; on ST I was modeling a veritone and I realized that the original veritone circuit closely resembles our in-house eddy model:



You can see that there is a cap and a 1.5H inductor in parallel with the pickup, when this control is a part of the circuit. 

And the particular values of a veritone cause it to function as a band stop:




So I believe that is essentially what our eddy current model amounts to; a band stop which is positioned at a frequency and with a bandwidth that approximates the characteristics of eddy losses.

[antigua]

BTW, I have emailed back and forth with Helmuth Lemme a couple times in the past week, so if we want to, we can request his opinion on modelling eddy currents.


My worry is that it ultimately it comes down to requiring a frequency dependent AC source, which stratotarts achieved with a laplace (I don't understand how it works yet. I've put some time into it, though). Some implementations of SPICE allow you to customize the AC sweep, but LTSpice does not. 

If your "AC source" in reality is changing flux density, and in reality, eddy currents cause a frequency dependent net loss of flux density, then the losses are truly external to the circuit, and could only ever be 100% accurately modeled that way.

With the in-house eddy model, does a set of given values used to curve match an unloaded plot and work for the 470pF / 200k loaded plot, or does the eddy model have to be tweaked in order to curve match both both?

[JohnH]

Oct 4, 2016 23:19:33 GMT -5 antigua said:

With the in-house eddy model, does a set of given values used to curve match an unloaded plot and work for the 470pF / 200k loaded plot, or does the eddy model have to be tweaked in order to curve match both both?

The 6 part models that Ive made do quute a good job of matching both loaded and unloaded curves, using all fixed components and a single source of constant voltage. The same values could be put in any Spice sim. I could post some if youd like to try. Mostly they are in GuitarFreak now.

[antigua]

Can you simply program your AC source to decrease 6dB per octave directly?

[perfboardpatcher]

I posted this schematic on Strattalk some time ago:




This was how I measured my guitar pickups before I had the excitation coil & the circuit with the integrator.
(SGEN= signal generator, CRO= oscilloscope)
I still use the same model for the pickup with a parallel resistor (R2).

routine #1:
A: small cap or no cap at all for C2 (330p) - Find the resonance frequency with the maximum voltage output
B: big cap for C2 (10n) - Find the resonance frequency with the maximum voltage output
Determine L and C1 with some fit and measure

routine #2:
For a given cap value for C2,
Find the resonance frequency with the maximum voltage output,
calculate the attenuation at that resonance frequency,
calculate R2


Now to address some issues raised by Antigua:

Antigua to JohnH

If your "AC source" in reality is changing flux density, and in reality, eddy currents cause a frequency dependent net loss of flux density, then the losses are truly external to the circuit, and could only ever be 100% accurately modeled that way.

1. When I go to my spreadsheet for a guitar pickup used as voltage source and enter the values for L, C1 and R2 and I compare the graph with the results from measuring the pickup by excitation coil/integrator, then I find that the actual losses in the pickup are higher than found by strictly going by the electrical properties of the pickup. (The height of the peak when measured by excitation coil/integrator is lower.)
One could say that the pickup isn't modelled properly but I don't see how replacing R2 with other components would solve the problem. It seems that excitation coil & integrator circuit is the only way.


Antigua to JohnH

With the in-house eddy model, does a set of given values used to curve match an unloaded plot and work for the 470pF / 200k loaded plot, or does the eddy model have to be tweaked in order to curve match both both?

2. Speaking from my own experience I can say that when the resonance frequency (C2!) changes R2 changes along. If I go back to my spreadsheet for the DiMarzio HFH and change the values in column F (L_h2 6.00, R_h2 80000, factor 0.0) I wind up having a graph that's almost identical with the graph I posted. Now if I change the capacitance into (557p, was 87p) and resistance ( (360kx200k/(360k+200k)=129k, was 360k) and when comparing the resulting graph with the graph plot provided by Antigua for the loaded pickup I notice that the peak on my spreadsheet is too high. That makes perfect sense, the parallel resistance should be lower than 360k. The load resistance has to be around 120k to achieve an identical looking graph.

[stratotarts]

The values from exciting the pickup directly with a constant current source, and from exciting it with an external magnetic field, differ in the location of the magnetic field. In the case of electrical excitation, the field (and therefore the measured influences) are located around the pickup coil and core. In the case of the exciter coil, the measured influences are based on the field around the exciter coil. This is by design, in order to capture the asymmetric nature of the system. The interposition of a cover is a perfect example of this. It more strongly shapes the response because it intervenes between the excitation field and the pickup coils. It is important to keep in mind that although electromagnetic laws are reversible mathematically, systems composed of multiple components with intervening losses and frequency dependent coupling are generally not.

However, it is true that the results from internal and external excitation correlate strongly because after all, many of the system properties are common.

[perfboardpatcher]

Oct 7, 2016 9:49:52 GMT -5 stratotarts said:

The values from exciting the pickup directly with a constant current source, and from exciting it with an external magnetic field, differ in the location of the magnetic field. In the case of electrical excitation, the field (and therefore the measured influences) are located around the pickup coil and core. In the case of the exciter coil, the measured influences are based on the field around the exciter coil. This is by design, in order to capture the asymmetric nature of the system. The interposition of a cover is a perfect example of this. It more strongly shapes the response because it intervenes between the excitation field and the pickup coils. It is important to keep in mind that although electromagnetic laws are reversible mathematically, systems composed of multiple components with intervening losses and frequency dependent coupling are generally not.

However, it is true that the results from internal and external excitation correlate strongly because after all, many of the system properties are common.

I've read your comment a few times, aren't you essentially like Antigua and me allude to that there are more than one voltage sources - "from internal and external excitation" - at work?

I would like to present the following problem:

Consider a configuration set up as in 78932
Pickup under test: some strat sc pickup with alnico rod magnets
We're performing the test, we measure -> test results #1

Now test #2:
First we degauss the magnets
We're performing the test once again, we measure -> test results #2

Do you believe like I do that test results #1 and test results #2 should be identical?

[antigua]

After learning a little bit more about eddy currents recently, I have to take back my speculation that the two different eddy models, one with and without -AC were both valid, I was trying to keep a open mind, but I feel confident that these are the relevant facts of the matter:

a) you can only have one AC source per physical coil. Eddy currents cause a net loss of flux with increasing frequency, so you might have a lot of tributaries of positive and negative flux fields, but they sum together as a net change of flux through the coil. What has to happen is that the AC source needs to supply frequency dependent amplitude in order to model that sum of flux.

b) the eddy current model we have settled upon is actually a complicated RLC filter that features a dual behavior, where you have one capacitor and two inductors resonating semi independently, emphasizing one over another depending upon the frequency, and that just happens to mimic the effects of eddy current losses.

I've been trying to learn more about what this filter is, and it's all very complicated, but it appears to be like a Chebyshev third order filter:







I'm not certain that this is what it is, I'm only noting the similarity.

If you stack these resonant tank circuits, you can see multiple stop bands, or "ripples":

[JohnH]

Oct 8, 2016 13:11:34 GMT -5 perfboardpatcher said:

Oct 7, 2016 9:49:52 GMT -5 stratotarts said:

The values from exciting the pickup directly with a constant current source, and from exciting it with an external magnetic field, differ in the location of the magnetic field. In the case of electrical excitation, the field (and therefore the measured influences) are located around the pickup coil and core. In the case of the exciter coil, the measured influences are based on the field around the exciter coil. This is by design, in order to capture the asymmetric nature of the system. The interposition of a cover is a perfect example of this. It more strongly shapes the response because it intervenes between the excitation field and the pickup coils. It is important to keep in mind that although electromagnetic laws are reversible mathematically, systems composed of multiple components with intervening losses and frequency dependent coupling are generally not.

However, it is true that the results from internal and external excitation correlate strongly because after all, many of the system properties are common.

I've read your comment a few times, aren't you essentially like Antigua and me allude to that there are more than one voltage sources - "from internal and external excitation" - at work?

I would like to present the following problem:

Consider a configuration set up as in 78932
Pickup under test: some strat sc pickup with alnico rod magnets
We're performing the test, we measure -> test results #1

Now test #2:
First we degauss the magnets
We're performing the test once again, we measure -> test results #2

Do you believe like I do that test results #1 and test results #2 should be identical?

I would agree with you. Whether the poles are magnetised or not, should not affect the coil inductance (though the fact that the poles are there at all does so), and also, the flux field they create if magnetised is fixed, and so induces no signals. Its only when some moving strings get magnetised that a moving flux is created dependent on the pole magnets. Al the above should apply equally to Strat tart's and Antigua's tests too, with the driver coils.

[JohnH]

Oct 8, 2016 15:44:16 GMT -5 antigua said:

After learning a little bit more about eddy currents recently, I have to take back my speculation that the two different eddy models, one with and without -AC were both valid, I was trying to keep a open mind, but I feel confident that these are the relevant facts of the matter:

a) you can only have one AC source per physical coil. Eddy currents cause a net loss of flux with increasing frequency, so you might have a lot of tributaries of positive and negative flux fields, but they sum together as a net change of flux through the coil. What has to happen is that the AC source needs to supply frequency dependent amplitude in order to model that sum of flux.

b) the eddy current model we have settled upon is actually a complicated RLC filter that features a dual behavior, where you have one capacitor and two inductors resonating semi independently, emphasizing one over another depending upon the frequency, and that just happens to mimic the effects of eddy current losses.

I've been trying to learn more about what this filter is, and it's all very complicated, but it appears to be like a Chebyshev third order filter:







I'm not certain that this is what it is, I'm only noting the similarity.

If you stack these resonant tank circuits, you can see multiple stop bands, or "ripples":

Its an interesting exploration. I find that multiple peaks occur only if multiple caps are present (ie, all the bits of capacitance are not all in parallel and so cant be condensed into one). But I cant prove whether or not that is fundamental though, ie, why the various inductors don't create different separate peaks with the one shared cap.

At the end of the day, they are all just models, and they are good models if they are useful or give helpful predictions of the true situation which is way more complicated, in this case with many overlapping capacitive couplings and inductive and eddy effects between winding coils and other parts.

[antigua]

Oct 8, 2016 18:46:14 GMT -5 JohnH said:

At the end of the day, they are all just models, and they are good models if they are useful or give helpful predictions of the true situation which is way more complicated, in this case with many overlapping capacitive couplings and inductive and eddy effects between winding coils and other parts.

I've not looked at the internal working of the GuitarFreak, can you just modify the current by frequency mathematically? 

I think there is terminology to describe this difference, but I can't recall what it is, but you have models that represent physical reality, and then you have models that disregard the physical reality, because the physical reality is either unknown or excessively complicated, but still manage to produce useful approximations. 

[JohnH]

I've not looked at the internal working of the GuitarFreak, can you just modify the current by frequency mathematically? 

As it is, the signal source is effectively a zero impedance constant voltage source at all frequencies. It steps through the spectrum like that, calculating one semitone at a time, similar to a Spice model. But its easy to add any frequency-dependent functions on top of it, and I do that to catch other effects such as pickup placement and string harmonics. An overall frequency vs amplitude function would be easy, since its all just math from first principles, it can do what ever its set up to do. But I don't think there's really a need, since as it is, with everything switched on and starting from a constant source, it can get a good representation of an actual recorded signal.

[stratotarts]

Oct 8, 2016 13:11:34 GMT -5 perfboardpatcher said:

Oct 7, 2016 9:49:52 GMT -5 stratotarts said:

The values from exciting the pickup directly with a constant current source, and from exciting it with an external magnetic field, differ in the location of the magnetic field. In the case of electrical excitation, the field (and therefore the measured influences) are located around the pickup coil and core. In the case of the exciter coil, the measured influences are based on the field around the exciter coil. This is by design, in order to capture the asymmetric nature of the system. The interposition of a cover is a perfect example of this. It more strongly shapes the response because it intervenes between the excitation field and the pickup coils. It is important to keep in mind that although electromagnetic laws are reversible mathematically, systems composed of multiple components with intervening losses and frequency dependent coupling are generally not.

However, it is true that the results from internal and external excitation correlate strongly because after all, many of the system properties are common.

I've read your comment a few times, aren't you essentially like Antigua and me allude to that there are more than one voltage sources - "from internal and external excitation" - at work?

I would like to present the following problem:

Consider a configuration set up as in 78932
Pickup under test: some strat sc pickup with alnico rod magnets
We're performing the test, we measure -> test results #1

Now test #2:
First we degauss the magnets
We're performing the test once again, we measure -> test results #2

Do you believe like I do that test results #1 and test results #2 should be identical?

I am referring to the two kinds of test, one is direct excitation of the pickup through a high value resistor. You, in 78932 above, Errede, Carson, Lawson all have used this method. The other is by external magnetic excitation as you report that you now do, and Lemme, Moore, myself and now Antigua also use.

In the modeling area, I feel that Carson came the closest because of its simplicity. His model has only a fixed capacitor, and a frequency variable inductor and resistor. I found that making the inductance constant made almost no difference (probably because it acts at a low frequency where the inductive reactance is also low). So even a simple LCR model with variable resistance (it obeys an exponential law with relation to frequency), models reality very closely. I did not pursue the implementation of it because I have been focusing on refining test equipment. In these discussions about modeling I mostly sit back and read because it's gotten ahead of my full understanding. If I understand correctly, we are wondering what a good model would be. In my view, that would be one that balances simplicity with predictive power. How well it corresponds to the actual physical mechanisms inside the pickup is not as important. I believe that those mechanisms are much too complex to model directly without a quantum leap in computational methods (for example multiple discrete element simulation such as is used to emulate string vibration).

I don't think of modeling itself as adding to the understanding of the internal physics. It seems to me, more about being able to analyze data and perform simulations. So it seems arbitrary to me, whether components of the model actually exist in the physical pickup. Two different voltage sources, or 12 Henry inductors, neither really exists in the pickup, right?

Lastly, I think results #1 and #2 should be identical.

[antigua]

I think if it's possible to model eddys via the AC source directly, it should be done that way. Now that I have a slight grasp on why the parallel inductor and resistor work the way they do, I don't think they should be used in the model, because they are hypothetical components in an model that is otherwise real. I thought they might somehow be modelling actual eddy currents, but I know that's not the case now. 

I might use it in LT spice since it doesn't allow for a dynamic AC source, but I would a label beside the eddy model that describes it as "hypothetical eddy current losses". 

[JohnH]

Oct 8, 2016 22:21:15 GMT -5 antigua said:

I think if it's possible to model eddys via the AC source directly, it should be done that way. Now that I have a slight grasp on why the parallel inductor and resistor work the way they do, I don't think they should be used in the model, because they are hypothetical components in an model that is otherwise real. I thought they might somehow be modelling actual eddy currents, but I know that's not the case now. 

I might use it in LT spice since it doesn't allow for a dynamic AC source, but I would a label beside the eddy model that describes it as "hypothetical eddy current losses". 

Are you thinking to model the eddy current and damping effects, rather than with the 6 part model, instead  by a frequency dependent source amplitude together with a 3 part model? I may have misunderstood, in which case I apologise. 

Agood model needs to work with one set of parameters to cover all cases within its range. So if a frequency dependent amplitude was to be a part of it, it should be the same function to capture both the loaded and unloaded results, and other cases too.  The fixed 6 part models do that quite well.

But consider what would be required if the amplitude was varying as the main way to capture damping, with a simple 3 part model.

Here are some plots relating to the Seth Lover neck pickup, the blue lines are both from the same 6 part model derived from the tests (really, I should show the test results themselves, but these are very close and ok to show a point). The higher curves are from 3 part models, with inductance and DCR as measured, and capacitance derived to match unloaded frequency.



So the job of a variable signal source with the 3 part model would be to bring the upper curves down to the levels of the 6-part = approx. test results. If you compare respective curves, not only is the required corrections varying with frequency but it also varies between the two load conditions. So there is not a single set of frequency-dependent adjustments that solves the whole model for both conditions. The biggest changes are needed at the resonant frequency for that particular load, so such a model would very difficult to derive and implement IMO.

[perfboardpatcher]

Thanks a lot for the comments, guys! I've read it all but I myself am still in the process of piecing something together.

[antigua]

Oct 9, 2016 3:55:42 GMT -5 JohnH said:

Oct 8, 2016 22:21:15 GMT -5 antigua said:

I think if it's possible to model eddys via the AC source directly, it should be done that way. Now that I have a slight grasp on why the parallel inductor and resistor work the way they do, I don't think they should be used in the model, because they are hypothetical components in an model that is otherwise real. I thought they might somehow be modelling actual eddy currents, but I know that's not the case now. 

I might use it in LT spice since it doesn't allow for a dynamic AC source, but I would a label beside the eddy model that describes it as "hypothetical eddy current losses". 

Are you thinking to model the eddy current and damping effects, rather than with the 6 part model, instead  by a frequency dependent source amplitude together with a 3 part model? I may have misunderstood, in which case I apologise. 

Agood model needs to work with one set of parameters to cover all cases within its range. So if a frequency dependent amplitude was to be a part of it, it should be the same function to capture both the loaded and unloaded results, and other cases too.  The fixed 6 part models do that quite well.

But consider what would be required if the amplitude was varying as the main way to capture damping, with a simple 3 part model.

Here are some plots relating to the Seth Lover neck pickup, the blue lines are both from the same 6 part model derived from the tests (really, I should show the test results themselves, but these are very close and ok to show a point). The higher curves are from 3 part models, with inductance and DCR as measured, and capacitance derived to match unloaded frequency.



So the job of a variable signal source with the 3 part model would be to bring the upper curves down to the levels of the 6-part = approx. test results. If you compare respective curves, not only is the required corrections varying with frequency but it also varies between the two load conditions. So there is not a single set of frequency-dependent adjustments that solves the whole model for both conditions. The biggest changes are needed at the resonant frequency for that particular load, so such a model would very difficult to derive and implement IMO.

Are you saying that simply lowering the AC source with increasing frequency won't cause the Q factor to also drop, as is observed, and that a benefit of the six part model is that it does lower the Q enough to match the actual measurement?

[JohnH]

Yes indeed, that's pretty much how I see it.

I think also, that it is reasonable to accept that the true complexities of damping etc can be considered as a partly resistive and partly reactive load, and that the extra 3 parts are a reasonable way to approximate rhem, calibrated by the test results.

[perfboardpatcher]

Oct 8, 2016 22:10:11 GMT -5 stratotarts said:

I don't think of modeling itself as adding to the understanding of the internal physics. It seems to me, more about being able to analyze data and perform simulations. So it seems arbitrary to me, whether components of the model actually exist in the physical pickup. Two different voltage sources, or 12 Henry inductors, neither really exists in the pickup, right?

Lastly, I think results #1 and #2 should be identical.

As a matter of fact the 12 Henry exists in the pickup! You will find out in the next post of mine.
But adding a Shebyshev or notch filters to the pickup model as Antigua suggested on the other hand isn't easily justified, I agree with you!

So we all agreed results #1 and #2 should be identical.

[perfboardpatcher]

I was thinking of a way to combine 2 ways of measuring
I do not favour the pickup model that I describe here in particular. I also prefer a simple pickup model. The signal chain from pickup to guitar loudspeaker is long and all along the signal chain a lot happens in terms of eq. It makes sense to focus only on the eq effects caused by the pickup that are radical and difficult to be reversed: the 2nd order high cut and resonance peak.

It's just an experiment to see if the following "theory" will get me anywhere.

1. Measure the guitar pickup as impedance or how to model a guitar pickup (as voltage source) by measuring its impedance curve.



When measuring a guitar pickup like with as in 78932 a model consisting of L, R, C and Rp seems able to cover it all.
The impedance for the combination of L, R and Rp is Z1h = (XL + R) || Rp .
In my experience the model only works for one particular resonance frequency at a time. To model the same guitar pickup but set up with a different resonance frequency the measuring procedure has to be redone with the pickup set up with that new resonance frequency (new value for Cld). Together with the new resonance frequency comes a new value for Rp.



The results for L, R, C and Rp can be entered in Spice sim/spreadsheet that models the guitar pickup as voltage source which results in output in the form of a graph or a set of datapoints.

Now the idea is to replace the single branch of L+R and Rp with a combination of 3 separate branches of L+R in parallel: #1 "guitar" deals with the guitar signal, #2 "damping" and #3 "scoop" deal with the losses.
The impedance for these multiple branches is Zmu = (XL1 + R1) || (XL2 + R2) || (XL3 + R3) = XLmu + Rmu
Obviously the impedance curves of Zmu and Z1h should match as close as possible.
R2 has the same value as Rp, L2 has a low value. Or possibly, alternatively, L2 and R2 together could provide damping independent from the resonance frequency.



2. Measure the guitar pickup as sensor

The procedure to measure the guitar pickup by means of an excitation has been explained by Antigua and Stratotarts .

The results of measurements of the pickup with excitation coil and integrator circuit can be compared with the simulated graph from (1) .
To make up for the discrepancies the following schematic of the pickup as voltage source applies.
I've added voltage sources VS2 and VS3 - what else? - which are proportional to VS1. (As an example: VS2 = -0.2 * VS1)



Furthermore I want to mention that in JohnH' 6-model there's one L+R branch that does all the damping and scoop. The parallel resistor only seems to limit the resistance of the unloaded pickup at resonance frequency. I presume this model is tuned to work for capacitance values from 0 (unloaded) up to 470pf but I'd expect it to go off the rails at much lower resonance frequencies, say, 1000 herz.

[JohnH]

Oct 10, 2016 13:24:26 GMT -5 perfboardpatcher said:

I presume this model is tuned to work for capacitance values from 0 (unloaded) up to 470pf but I'd expect it to go off the rails at much lower resonance frequencies, say, 1000 herz.

I think you forgot to complete that statement with..."because.....", followed by a rational theory of why you think that might be. But yes, any model will deviate from reality to some extent.

There has been a few cases where such a 6-part model that has been derived from the two tests has then been tested with a different load and found to give a good predictions. But it would be a good instead to do more, say with a higher cap representing a longer cable or even a tone control turned right down. Id be happy to put one up as a prediction, and then let it be tested in reality.

[antigua]

Oct 10, 2016 13:24:26 GMT -5 perfboardpatcher said:

The results for L, R, C and Rp can be entered in Spice sim/spreadsheet that models the guitar pickup as voltage source which results in output in the form of a graph or a set of datapoints.

I appreciate that you're giving thought to this, because it seems that you understand some aspects of RLC filters better than I do. I'd take it upon myself to model this in LTSpice and see what happens, but I still have about twenty more sets of pickups to analyze for raw data, and whatever else turns up. Have you tried modelling this to see if it works as you'd expect?

[perfboardpatcher]

Oct 10, 2016 14:24:58 GMT -5 JohnH said:

Oct 10, 2016 13:24:26 GMT -5 perfboardpatcher said:

I presume this model is tuned to work for capacitance values from 0 (unloaded) up to 470pf but I'd expect it to go off the rails at much lower resonance frequencies, say, 1000 herz.

I think you forgot to complete that statement with..."because.....", followed by a rational theory of why you think that might be. But yes, any model will deviate from reality to some extent.

There has been a few cases where such a 6-part model that has been derived from the two tests has then been tested with a different load and found to give a good predictions. But it would be a good instead to do more, say with a higher cap representing a longer cable or even a tone control turned right down. Id be happy to put one up as a prediction, and then let it be tested in reality.

Because plots for unloaded and 200k||470pf were all you got from Antigua. Yeah, if you could post a Guitarfreak plot for one of Antigua's pickups loaded with 10M||2.2nf or 10M||10nf - whatever cap value Antigua has - that would be nice.

[perfboardpatcher]

Oct 10, 2016 15:14:37 GMT -5 antigua said:

Oct 10, 2016 13:24:26 GMT -5 perfboardpatcher said:

The results for L, R, C and Rp can be entered in Spice sim/spreadsheet that models the guitar pickup as voltage source which results in output in the form of a graph or a set of datapoints.

I appreciate that you're giving thought to this, because it seems that you understand some aspects of RLC filters better than I do. I'd take it upon myself to model this in LTSpice and see what happens, but I still have about twenty more sets of pickups to analyze for raw data, and whatever else turns up. Have you tried modelling this to see if it works as you'd expect?

Nope, not yet. But what was showed in the first post looks like a good starting point.

[JohnH]

OK, well it would depend on whether doing this is of interest to Antigua, since we rely on him for all the excellent test data.

My suggestion would be to try this with one simple alnico single and one PAF. The singles tend to show very low damping and model very closely with the way I'm doing it, the PAF's get quite close but they are trickier to capture because peaks are less defined and there are more damping effects happening.

Which pickups they are could be selected from those that have already been tested, and are still available for a further bench test (ie not put back into guitars), Antigua to chose.

Here's the steps, in order to try to make it scientific:

1. Pick two pickups (Antigua)
2. Find an extra cap in the range around say 5nF, doesn't matter what, but then measure it with the LCR meter (Antigua).
3. JohnH makes 6 part models (if not already done) of the two pickups from the basic loaded and unloaded tests already posted, and posts these to show how they match or deviate from the 10M/10pF and 480pF/200k tests.
4. JohnH runs the models with 10M + 10pF + extra cap and also 200k + 480pF + extra cap and posts the results
5. Antigua retests the pickups, loaded as the original tests plus the extra cap in parallel.

Just putting that out there to see if that is of interest, timeframe: 'in due course'.


Results may be surprising, and I would be surprised if there is not a good match. The reasons I have this view are that:

1. With two load points matching the data quite well, this orientates the data so that changes in load within the tested range, and beyond it, are at least moving the results in the right directions away from an accurately known fixed point which is the loaded test.
2. Im sure the 6 part models are at least better than basic 3 part models. When comparing these, the greatest differences are with lower loading. With greater loading, (resistive and/or capacitive), the differences are less dramatic and the 6 part models can be seen to plotting a credible few db lower than the 3 part models.
3. On a couple of cases, we have test results for greater loads such as from 2x250k pots. The 6 part models derived from lower loading results model these quite well
4. There's nothing in the maths that would lead to very large deviations in results at different loads, from models that are well aligned at the two test loads. There will be some deviation however.