Minimizing Capacitance with CNC Winding Patterns

[markm]

Pretty much all my other posts have been in relation to side issues of my main project, which is pretty much summed up with the subject line for this post.
As pretty much all the members of this forum are aware, capacitive coupling comes in part from adjacent parallel coils, but more so between layers rather than individual adjoining coils in a given layer.
Anyway it occurred to me that this could be dealt with by using special winding patterns that would avoid or minimize inter-layer parallel segments, and I didn't see that anyone had really tried this. I also realized that "scatter" winding with conventional machines offered neither significant variability in wire angle nor consistency of coil placement, and was thus entirely inadequate for addressing capacitance issues. The key to doing this was creating a winding machine that would allow eccentrically would coils. After a long journey of creating a CNC winding machine and writing the software to run everything, and carefully testing the various pattern functions, I've finally been able to create a few dozen pickups, collect data,
verify assumptions, and test some theories.

Here's a quick tour of my twisted puppy maker:


Stratotart's Integrator project has been a great help with making measurements and I'm gathering all my data in a simple database to facilitate analysis.
Initial results have been exciting with some winding patterns reducing capacitance nearly 50% over a regular helical wind pattern (w/identical bobbin/turns/magnets/wire), and peak frequency response shifted up about 2/3 of an octave in some cases.

Using just identical 52mm Strat bobbins for analysis has been a good control. Most of my test coils have nearly identical wind counts but I have a few high/low counts for a little variety.
The other day I graphed output level vs inductance on identically turned coils with different patterns, and it was all over the map (poor linear fit). This is probably mainly due to the various coil shapes created by multiple layers of my cnc winding patterns. 

Next I took a look at the peak unloaded frequency vs inductance, this to was pretty scattered. I knew that that was probably due to the varying capacitance, so I plotted the frequency vs LC (inductance*capacitance). This turned out to be a very linear data set. Clearly the LC term rules the pickup frequency response equation.



Pickup #1 shown on these graphs had nearly 10,000 windings with a regular helical pattern and really helps the data set for linear fits.

The final step was to look at the loaded peak frequency vs LC:



Pickup #4 and #9 both jump out of line in these graph but were completely middle of the pack on the unloaded graph!
#4 was a very loosely wound coil with extremely low inductance - basically a defective wind due to poor tensioning and not in interesting case, but...
#9 was a perfectly wound test pattern with good inductance and about 30% reduction in capacitance.  Why did this pattern yield a higher loaded peak (but not unloaded) compared to all the others? (more tests underway!)

Anyway that's the basics of my project and a look at some of the analysis I'm doing.  My eventual goal is to science the heck out of this and then try to make superlative pickups. Would love to make some money, but doing it for the love anyway.

I'll try to discuss more details about winding patterns and other issues in future posts on this thread.

[antigua]

I've found that coil geometry is highly important in the overall capacitance, and that there is likely a minimum capacitance that can be achieved based on the geometry of the coil.  I've measure Strat pickups as low as 80pF, and as high as 200pF, with most being between 85pF and 125pF. One possible factor which I haven't been able to investigate is the build thickness of the magnet wire Magnet wire with a thicker coat might result in lower capacitance due to the added space between the turns. It's also possible that different coatings have a substantially different dielectric value. 

As for geometry, if you look at Jazzmaster pickups, which are very wide and flat, they tend to have a low capacitance guitarnuts2.proboards.com/thread/8311/fender-vintage-jazzmaster-analysis-review  and www.offsetguitars.com/forums/viewtopic.php?t=107829 , with capacitance as low as 30pF, as high as 46pF, weell below that of the lowest measured Strat pickups. Conversely, if the coil is predominantly tall, the capacitance tends to be a lot higher. If you look at this thread guitarnuts2.proboards.com/thread/7728/measured-electrical-popular-telecaster-pickups and focus on the Telecaster neck pickup capacitance, it tends to vary from 150pF all the way up to 300pF. Telecaster neck pickup tend to be wound tightly and use 43AWG in order to ensure that the metal cover will fit over the bobbin, so it possible that the extreme capacitance comes from measures intended to make the coil as slim as possible. 

The take away from that is that the physical distance between the starting and ending layer of the coil has a huge impact on the capacitance. I think that even if you manage an exotic scatter pattern by way of CNC winding, you wont do much better than 80pF. There is a solution, but it would be tricky to pull off; and that's a partitioned coil. In other words, you wind lots of little flat coils, and then you stack them on top of one another. You might be able to approximate this with CNC winding, by concentrating the bulk of the coil to one side at the start, and the other side and the end. I believe that, even with perfect partitioning, you would still not do much better than 60pF, which is to say, somewhere between a regular tall Strat coil and the flat coil of a Jazzmaster pickup.

[markm]

There is certainly an almost irreducible coupling effect at the ends of the bobbin where the layers are all very nearly stacked parallel, fortunately the big wide middle is fair ground for reduction strategies.

So far I've actually gotten as low as 52pF with 8200 wraps of #43 poly, not bad eh?

The differences between enamel/formvar/poly coatings are in the 1/10000 of an inch range. With all the other factors at play, I think this is fairly insignificant and mostly part of the vintage snake oil mystique, but I could be wrong. Right now I just have #43 poly and #42 formvar so I can't make good comparisons based on coating.

One of the things my program is able to do is control how tightly the coils are wound on a given layer (by multiples of the wire diameter). Experiments with this parameter so far have shown that increasing this spacing has minimal effect on capacitance. Tests with patterns that reduce the inter-layer capacitance (making the intra-layer capacitance, which should be most effected by this parameter, more dominate % of total capacitance) seem to support the notion that the enameling is of minor consequence...but more data is needed.

Embarrassing revelation: #9 in my data set previously posted was a data entry error. Actual peak loaded freq was about 500hz lower than entered, and that brought it right back in line on the loaded LC graph.
The perils of publishing data early! I'm winding a coil right now to replace the #4 data point with a properly wound coil since I realized that was a hole in my data set. 

Here's (hopefully) an animated gif that shows one of the basic patterns.



A standard helical layer is first wound, then half a layer is wound starting at the edge on one end, and going to the middle on the other. This eccentric layer proceeds till what was the second end is at the far edge of the bobbin instead of the middle. At this point the bobbin simply rotates 180 and repeats the same pattern, thus creating a cross pattern as shown.
You can imagine successive layers of this basic two layer pattern.
This particular pattern maximizes the angle between coils for adjacent layers, and thus reduces capacitive coupling.
As you may notice the central area where the eccentric coils overlap creates a raised diamond pattern. This also ends up creating space in the coil as other layers pass over this raised middle section. Over the course of many layers this creates an round arc shape to the outer coil layers. A similar thing happens with highly randomized layers.
An interesting aspect of this for future study is the effect this has on microphonics.

This picture shows the shape create by a layered double cross pattern like the one shown above:

The double cross pattern is very useful for compensation when combined with other patterns that create a central bulge.

Ultimately it should be possible to make pickups with tailored output level, microphonics, and frequency response with less linkage between these parameters.

[markm]

...Well the .gif doesn't want to animate on the forum page, but you can click on the rendered coil drawing to see the animation that shows the pattern.

[antigua]

This is pretty cool. I don't know that anyone has ever created guitar pickup coils with such carefully controlled layering patterns, and if they have, I doubt they've studied their creation so closely. 

I see you're putting a lot of emphasis on the capacitance values, and I think that comes from the fact that when we talk about scatter winding, it's all about the capacitance, but there is also going to be a variance in inductance as the turns achieve lower coupling coefficients as the scatter patterns become more wild. The graphs are good, but some tabular data that shows capacitance, inductance, peak frequency and scatter profile, might be easier to draw conclusions from. 

Supposing that L and C drop almost proportionately as a coil begins to look less like a coil, and more like a lot of little coils mashed together along several axis, this inspires questions. Why not just wind a proper coil with fewer turns of wire, and reduce L and C that way? If a person particularly enjoys a scatter wound pickup, would they not just as soon enjoy a pickup that merely features fewer turns of wire?

As I mentioned in the other thread, you might suppose that output would remain high despite L and C being reduced, but since those scattered turns of wire are off axis with the magnetized guitar string, my suspicion is that they will produce less output than turns that are co-axial with the core. I think if a pickup designed really wants to increase the output, or even maintain the output while reducing L and/or C, the game is to make the coil smaller and set it closer to the top of the pickup. Arguably the Lace Sensor already achieves this goal, and the ways in which a Lace Sensor is dissimilar to vintage style pickups is due to the lack of powerful AlNiCo pole pieces, and in their place, much weaker rubberized ferrite, like a refrigerator magnet.

Another pickup that achieves this, flat/wide form factor in a Strat footprint is the Bill Lawrence Micro Coil www.strat-talk.com/threads/bill-lawrence-microcoils-analysis.502500/ which if you refer to the measurements I took, had capacitances around 70pF, inductances of only 1.4H, and a DC resistance of 7k, suggesting that the turn count is probably in the area of 5000 turns, with 44 AWG, or maybe 4000 turns with 45 AWG, but this bode plot below shows that despite having around half the turns of a regular Strat pickup, it has essentially the same output voltage as a typical Strat pickup, because with a typical Strat pickup, the turns of wire are near the bottom of the coil serve to increase the L, C and R, but contribute very little to the actual output voltage, due to their distance from the guitar string, there the magnetic variability is at it's strongest. 




As an aside about the Micro Coils with the low inductance, my understanding is that, while these have AlNiCo pole pieces, the originals had steel pole pieces, and that Bill Lawrence's company created this AlNiCo version after Bill left the building, keeping the turn count the same, and just replacing the steel with AlNiCo pole pieces. If the AlNiCo pole pieces had been steel in these pickup, I expect the inductance would climb from 1.4H up to around 2.1H, given the trend of steel pole pieces increasing inductance by about 50%, which would place them on the same L range as a typical Strat pickup. The reduce magnetic reluctance of steel would likely increase the output, even beyond what is observed with the AlNiCo Micro Coil, and exceeding the vintage style Strat pickup.

Even if the steel poled Micro coil realizes a few dBV increase in output, that might not been too impressive, because taking it a step further, consider a PAF pickup. Not only does if feature highly permeable steel pole pieces, but it also has shallow bobbins, like a Micro Coil... and there are two of them side by side. So if you consider that both the PAF and the Strat pickup have about 8,000 turns of wire, the PAF is placing a much larger portion of those 8,000 turns closer to the magnetized guitar strings. All of that together can be summed up with one word, efficiency. Fender single coils are just not efficient, and so a focus on improving the output, or anything else about them, through wind pattern or scatter, is akin to rearranging the deck chairs on a sinking luxury liner.

About the only thing the Strat pickups have going for them is that the AlNiCo pole pieces produce a stronger magnetic field than steel poled pickups, but that is of limited utility, because too much magnetic pull upon the guitar strings begins to have consequences on how the string move, and guitarists end up lowing the pickup away from the strings, defeating the purpose of the stronger magnetic field. 

[pablogilberto]

Hello fellows!


This is cool



Thank you!

[pablogilberto]

Great work!

Just few questions...

Sept 13, 2019 6:13:58 GMT -5 markm said:

Initial results have been exciting with some winding patterns reducing capacitance nearly 50% over a regular helical wind pattern (w/identical bobbin/turns/magnets/wire), and peak frequency response shifted up about 2/3 of an octave in some cases.

What do you mean by this,  shifted up about 2/3 of an octave? 

Using just identical 52mm Strat bobbins for analysis has been a good control. Most of my test coils have nearly identical wind counts but I have a few high/low counts for a little variety.

How do you ensure that the bobbins are identical? I'm interested to do a similar experiment.
Any tips?

Next I took a look at the peak unloaded frequency vs inductance, this to was pretty scattered. I knew that that was probably due to the varying capacitance, so I plotted the frequency vs LC (inductance*capacitance). This turned out to be a very linear data set. Clearly the LC term rules the pickup frequency response equation.

I agree that the LC term rules.

I'm wondering why Strat models in the market can maintain an Inductance with a 5% variation. But upon checking the capacitance, they tend to differ. I usually get 80, 100, 120, 140 sometimes 200pF (usually  on cheap pickups).

With you experiment using CNC, do you get the same amount of capacitance assuming you don't change wire, identical bobbin and same winding pattern?


another question, what is the relationship of the LC parameters when it comes to wire tension?

[pablogilberto]

Sept 15, 2019 3:36:04 GMT -5 antigua said:

Thank you for this. Learning a lot.

I see you're putting a lot of emphasis on the capacitance values, and I think that comes from the fact that when we talk about scatter winding, it's all about the capacitance, but there is also going to be a variance in inductance as the turns achieve lower coupling coefficients as the scatter patterns become more wild.

I want to understand better the relationship between the RLC values and the winding pattern, wire tension and geometry of coils.
Do you have a good reference for this?

Why not just wind a proper coil with fewer turns of wire, and reduce L and C that way? If a person particularly enjoys a scatter wound pickup, would they not just as soon enjoy a pickup that merely features fewer turns of wire?

I'm thinking about the same thing. What if we just use fewer turns of wire? Lower L and C.
What will be the tonal consequence? Will the output level (dB) drop?

About the only thing the Strat pickups have going for them is that the AlNiCo pole pieces produce a stronger magnetic field than steel poled pickups, but that is of limited utility, because too much magnetic pull upon the guitar strings begins to have consequences on how the string move, and guitarists end up lowing the pickup away from the strings, defeating the purpose of the stronger magnetic field. 

I see some pickups using Neodymium magnets because they are strong. And I'm also thinking that too much pull will reduce string movement.
Have you experimented on this? 

[antigua]

Jun 22, 2020 10:10:23 GMT -5 pablogilberto said:

Sept 15, 2019 3:36:04 GMT -5 antigua said:

Thank you for this. Learning a lot.
I want to understand better the relationship between the RLC values and the winding pattern, wire tension and geometry of coils.
Do you have a good reference for this?




I'm thinking about the same thing. What if we just use fewer turns of wire? Lower L and C.
What will be the tonal consequence? Will the output level (dB) drop?




I see some pickups using Neodymium magnets because they are strong. And I'm also thinking that too much pull will reduce string movement.
Have you experimented on this? 

1) The effects of coil geometry on inductance are easy to find, for example www.allaboutcircuits.com/textbook/direct-current/chpt-15/factors-affecting-inductance/ , but I don't know of a good reference that discusses geometrical  consequences on parasitic capacitance, but I know from empirical data gathering that wide coils have much less capacitance than tall ones, and I think the reason is somewhat obvious: the distance across the circuit is a lot greater with a coil's star and finish are physically far apart. The resistance is easy to speak to: smaller and longer copper wire means more resistance.

2) Fewer turns of wire that are laid more efficiently would mean less resistance and a higher Q factor. The capacitance and inductance would be about the same as if more wire had been laid inefficiently (scattered). 

3) with existing pickup design, there's no utility in neodymium magnets, yet that's how they're often used. There is a lot of potential to use them in very small/flat pickup designs, for mounting on acoustic guitars, for example,