|
Post by JohnH on Jan 30, 2017 14:28:16 GMT -5
OK thanks, I see. It loads reasonably fast for me, especially bearing in mind the months of work that it is presenting in a second or two.
If it helps, now that it is working, the function-generating code could be stripped right back to something much shorter. The 'X' lines have the full 6 part model in them, and the whole way its written, creating a new variable on each line was only a work-around so it could be directly created and checked in excel.
I still think that some offset in overall volume might be worth building in, probably by adding a db or two at the very end after line starting X26. Otherwise, a CS69 is looking super bright and just as loud as a TX special etc. That's how I present them in GF, with the vintage pickups at 0 db and hotter ones a db or two more overall, and humbuckers about 5 db more.
The eddy function looks like it can gives a good shape though I don't follow it fully. I suspect that a little of what that function represents is really embodied in most pickups, and this could correct the tendency to slightly wider than real peaks in these plots. If you added the slope, and then noted the db drop that this creates at the peak frequency, then added that to the input peak db value, it would sharpen up the curve, cutting a little from the slopes keeping he peak the same. This would not necessarily cause a dip if it is a very mild correction. Can show more if that sounds of any interest.
|
|
|
Post by antigua on Jan 30, 2017 15:34:15 GMT -5
What I have found is that there is some sort of curve with respect the inductance and voltage output. Take for example the Texas Hot set... The inductance of the bridge is 3.9H and the neck and middle are 2.8 and 2.9H. I place the driver coil in the same location for all the pickups, so the difference in dB peaks to the difference in voltage output, and it can be seen that for 1H higher inductance, the bridge put out about 2dB more. I was planning to survey the relationship between voltage output and inductance for pickups types and see if a curve emerged. As for eddy current slope off, I want to do that, but again the math is tricky. A survey of eddy current losses will probably show that for a given dB peak, there is a corresponding slope downwards. Here are some examples: You can see that if the peak is 0dB, or on the same Y plane as the the broad inductive reactance, there tends to be a downward slide and a little bump at resonance. This is especially obvious in the Filter'trons. The extent to which this is true is partly due to the cause of the eddy currents, but there is also a strong correlation to the overall amount of eddy currents. So, rather than call it 0dB, we could slope it off, say 2dB per decade, the add 1 or 2 dB to the resonance, and you will get that nice swoop shape.
|
|
|
Post by JohnH on Jan 30, 2017 16:20:28 GMT -5
When ive looked at tbe data before, my impression was that DCR is a bit more directly related to output than inductance. Eg double the turns doubles the DCR and doubles the output voltage, whereas inductance usually has a turned^2 term in its formulation, and is also more disturbed by ferrous parts.
So, my suggestion for a db offset based on comparison to a nominal reference pickup would be:
Offset db = 20 log10 (DCR ratio x gauss ratio x area ratio) + type offset
Area ratio is 1 for 42 swg and 0.796 for 43 swg.
Type offset is a generic db offset to relate overall pickup types, such as alnico single, ceramic single, PAF, hot ceramic HB, Filtertron etc.
Doing that using DCR puts the Texas Hot B about 2db above tbe neck, which is about where it appears on the plots above. A TX Sp neck ends up 1 db above a CS69, which also seems about right.
|
|
|
Post by antigua on Jan 30, 2017 16:43:53 GMT -5
I'm willing to try that out. Ohm's law says voltage will be V = I * Z. The inductance is an addend of Z, so I think it has to be accounted for in a voltage output equation. The bigger mystery is obviously I, and it sounds like you want to relate Gauss and an approximation of the coil area in order to get I. That would be really cool if that could be made to work. Since this a practical tool, if the end result is close enough, that would be great, but if it takes real variables into account, so much the better.
What do you intend for the Gauss ratio? Halving the flux density doesn't halve the current of course, but I'm not sure what the coefficient would be. I can do some testing, but practical testing with a guitar string is the most tedious kind. I need to create a jig that will let me mount test pickups, then I would record plucks with different Gauss and correlate the Gauss with the voltage.
I also have a pickup winding machine, and in the near future I'm planning to wind specific pickups, like 5000 turns or 41, 42, 43, and 44 AWG just for the sake of seeing how they vary in metrics.
|
|
|
Post by JohnH on Jan 30, 2017 17:10:09 GMT -5
My thought about Z is that the offset would be really to get the output relatuvely correct at low frequency, where inductive reactance is low and dcr is more dominant.
The area factor is just for the wire cross section, and the factor 0.796 is the cross section area ratio between 43 and 42 gauge wire. It compensates for an apparently 'hot'pickup with high dcr, not having so many turns as otherwise implied.
The gauss ratio, I just assumed it would be linear. At least, zero gauss leads to zero output! Maybe we can look up the relationship of induced voltage to gauss and rate of change of gauss.
|
|
|
Post by antigua on Jan 30, 2017 17:51:23 GMT -5
With the gauss, what is needed is the difference in flux displacement. The strength falls off inverse square, and since the string never actually touches the pole piece, it never realized the full flux measured. The difference is therefore two distance points along that inverse square curve. Some math happens, and while A3 is half the flux density of A5, but the A3 will not show half the voltage output.
Some FEM modelling would be great for getting some basic value of flux min and max distance, but it's a rather tricky program to use.
I think some sort of heuristic the best solution for the time being.
Also a note about AWG, I couldn't identify the AWG for a handful of pickups, especially where it might have been 43 or 44 AWG. The JB for example, I have no idea. Plus, DiMarzio frequently uses 42 AWG for one coil and 43 for the other, further complicating matters.
|
|
|
Post by antigua on Feb 7, 2017 13:14:51 GMT -5
I've come very close to curve matching eddy currents. It's not the most elegant math, but from a practical standpoint, it delivers the goods, and I'm anxious to share, but I'm still hung up on an important detail: how much louder is a humbucker than a single coil for a given inductance? You have Filter'trons and Strat single coils that both come in around 2.5H, and yet I'm sure the Filter'tron is quite louder - but how much louder?
I wasn't able to use DC resistance as a clue because I don't have reliable AWG wire thickness data in order to offset for finer wire types. There are a handful of pickups where I honestly can't tell which they've used.
|
|
|
Post by JohnH on Feb 7, 2017 14:30:54 GMT -5
I reckon as good a way as any is to grab some guitars with examples of the various pickup types and strum them as consistently as you can into a pc via a neutral buffer. Then you can look at the wave form amplitudes to judge the relative levels. Its not 'accurate' but it cant be completely wrong either. I do this using Audacity software. My rule of thumb benchmark is 2.4k Texas special is about 5db below my 57 classic. Then from there you can generate values within a pickup type from other measured properties.
btw, Ive been wrestling with fitting the curve equation that we've discussed here, which was an adjusted 3 part model, to an equivalent 4 part model. They do match exactly for db output, but with 4 parts you can get also the impedance and phase angle more correct. The difference amounts to a rotation and scale factor of the real and imaginary parts which change with frequency. Very interesting but tricky! From there, I want to go onto the 6 part model, hopefully derived from the data table.
|
|
andyholmes
Rookie Solder Flinger
Posts: 12
Likes: 0
|
Post by andyholmes on Feb 8, 2017 0:05:26 GMT -5
|
|
|
Post by antigua on Feb 8, 2017 0:40:55 GMT -5
I've happened across the UIUC Physics pickup resources, but it's been well over a year since I last reviewed it, and I'll probably see it with a much different eye now, so thanks for reminding me about it.
I believe you can incorporate a pickup model into a pickup LTSpice circuit and have it model correctly. I think the most important thing is modelling the input impedance correctly. Member stratotarts had modeled his integrator circuit, used for testing pickups, with LTSpice before prototypes were made, and IIRC, it correctly modeled the ultimate outcome. I think that was somewhat similar to this scenario, we're just replacing the test device with an effect pedal.
If for some reason the simulations don't look right, and you're willing to share, we can look at your LTSpice circuit and see if there are any apparent problems with the pickup model. I don't know so much about the active circuitry, but others in present company do.
|
|
|
Post by JohnH on Feb 8, 2017 2:58:31 GMT -5
Hi Andy. Welcome to GN2 I believe the models of pickups proposed by Lemme would work for you in modelling stompboxes. I have done the same. You can also do a bit better and use more components to model the pickup. Take a look at this: guitarnuts2.proboards.com/thread/3627/guitarfreak-guitar-frequency-response-calculatorThis spreadsheet uses 6 components to model each pickup, based on Lemmes model with extra parts to match antiguas tests. You should be able to pull some values out to put into your Spice model. Even if you dont use the spreadsheet itself (which is based only on passive wiring), theres enough info in the screenshots on the link to cover a few typical humbuckers or singles.
|
|
|
Post by antigua on Feb 8, 2017 10:47:39 GMT -5
Someone mentioned that there is a version, or a fork, of LTSpice that would output simulated audio based on a circuit model, like an "audio LTSpice". Does anyone remember what that was called? I'd like to give that a try now.
|
|
andyholmes
Rookie Solder Flinger
Posts: 12
Likes: 0
|
Post by andyholmes on Feb 8, 2017 13:44:00 GMT -5
Actually, I think LTSpice does that natively now. If I recall you can input to a voltage source with .wavefile=filename.wav and output with .wave=filename.wav, but I've never tried it myself. I remember trying input with ngspice but had problems controlling the amplitude, probably LTSpice is better.
On a related note, I've noticed when running a sine wave through the pickups the output of the pickups is reduced considerably, I can account for it but I was curious if you had any measurements of the approximate output level of pickups or if it's possible to extrapolate from the other variables?
|
|
|
Post by antigua on Feb 8, 2017 14:34:44 GMT -5
I'll have to give that audio feature a try. It would be especially useful for demonstrating the tone of a guitar with various peak resonances, and veratone mods that have peculiar band stop curves. If circuit modelling can be incorporated with audio I/O, it's only a matter of time at this point until someone can create a full pickup modeling suite based on all the information and technology we have on hand. I know that Line 6 has done something similar with their Variax, but the technology employed there is opaque to us. On a related note, I've noticed when running a sine wave through the pickups the output of the pickups is reduced considerably, I can account for it but I was curious if you had any measurements of the approximate output level of pickups or if it's possible to extrapolate from the other variables? It sounds like you're treating the pickup as a series inductor (which it is) with an AC sine wave source voltage, so you're getting DC resistance plus some degree of reactance, which is all frequency dependent. What frequency is the sine wave? Here's a question for anyone that can answer: with DC current, the only resistance is the "R" copper wire resistance. With AC, the resistance is copper wire "R" plus reactance "L" and "C" , so at very high frequencies, is the total effective resistance of the pickup actually lower than the DC resistance, due to the overtaking capacitance? I think the answer to this is yes, but I'm not 100% sure.
|
|
andyholmes
Rookie Solder Flinger
Posts: 12
Likes: 0
|
Post by andyholmes on Feb 8, 2017 15:27:44 GMT -5
Yes, I use John's GuitarFreak model basically how it's presented in the spreadsheet, which analogous to how Helmuth shows his model applied, so basically my subcircuit is 'T' shaped. I've been using 640Hz ever since a friend pointed out it's a more octal (octaval?) middle-ground on a guitar.
|
|
|
Post by JohnH on Feb 8, 2017 15:29:58 GMT -5
Here's a question for anyone that can answer: with DC current, the only resistance is the "R" copper wire resistance. With AC, the resistance is copper wire "R" plus reactance "L" and "C" , so at very high frequencies, is the total effective resistance of the pickup actually lower than the DC resistance, due to the overtaking capacitance? I think the answer to this is yes, but I'm not 100% sure. In principle Id say yes just based on the various RLC models that we have discussed. But it could be only above a very high frequency. Eg a 100pF cap only gets below say 7000 ohm impedance at frequencies above around 200kHz. But C is only a derived equivalent based on matching aspects of the pickups behaviour at much lower frequencies. So a better andwer may be that we dont know but that the answer is probably no in the audio range.
|
|
andyholmes
Rookie Solder Flinger
Posts: 12
Likes: 0
|
Post by andyholmes on Feb 9, 2017 10:23:03 GMT -5
I'm plodding along in SPICE trying to apply what's being discussed, so tell me if I'm way off base here.
I wondered if "total parallel inductance" should be balanced the way you do with R1||R2||R3 = DCR, so I gave that a shot but it seemed to reduce the dB peak without helping the frequency peak. So I was reading about parallel inductance and Wikipedia says "If the inductors are situated in each other's magnetic fields, this approach is invalid due to mutual inductance" which made me wonder if humbuckers, or even multiple single-coil setups, should be measured for mutual inductance, and I should use the SPICE model for coupled/mutual inductors? Am I right in assuming this plays the dominant role in the larger 900-1000Hz frequency response dip typical in humbuckers?
|
|
|
Post by antigua on Feb 9, 2017 12:19:39 GMT -5
Are you talking about parallel inductance in the six part model? If you do the simple Lemme three part, the inductance is only series.
|
|
andyholmes
Rookie Solder Flinger
Posts: 12
Likes: 0
|
Post by andyholmes on Feb 9, 2017 18:53:26 GMT -5
I've tried the Lemme model, and it generally works but entirely lacks the frequency response dip of humbuckers in nearing the 1kHz mark. It seems negligible at guitar output levels but I find it becomes quite audible in high-gain circuits. The 6 part model works quite well, but given how powerful SPICE is I'd rather have a model that takes the metrics/measurements of the pickup as-is, rather than curve matching. I came across a page about mutual inductance in humbuckers and I've had some limited success incorporating it into Lemme's model, I might try some other approaches tomorrow. .SUBCKT pu_test 1 2 3 L1 1 6 {ind/2} L2 6 4 {ind/2} R2 6 3 47k K1 L1 L2 0.02 R1 4 2 {res} C1 2 3 {cap} .ENDS Attachments:
|
|
|
Post by JohnH on Feb 9, 2017 21:32:39 GMT -5
I see you are working around 57 Classic properties. I've had a few goes at this, here is another set of 6-part properties to try, for a covered pickup R1 (k) 8.77E+00 L1 (H) 5.33 C1 (nF) 0.12 Rd(k) 1.80E+02 Rl(k) 208.9 Ld(H) 3.30E+01 Working them out is tricky, but I'm gradually improving and getting closer to real measurements. In this and other recent models (later than in the current GF), both the R and the effective inductance are matched to meter readings based on the combination of the parts, as well as aligning with frequency response from loaded and unloaded tests. This is a screen shot of the results from this model compared to physical tests: The dashed lines are traced from tests, the solid lines are calculated. Loading is 10pF and 10M for the unloaded and 480pF with 200k with the loaded test. You could run your model with loaded and unloaded conditions and see how it compares. I think getting a model to match tests at different loads as well as produce credible impedances is what is needed to have any confidence in it, and, if this is achieved, then it is a good model of reality for the purpose of testing its effects on following circuitry. Im interested in what you show, but I don't follow what the SPICE code means, could you describe what is happening in that code?
|
|
andyholmes
Rookie Solder Flinger
Posts: 12
Likes: 0
|
Post by andyholmes on Feb 10, 2017 4:21:28 GMT -5
Thanks for the updated 6-part properties, I can see you have your model pretty nailed down and its my choice for comparing my SPICE models to measured results. I've taken your advice and started to run tests loaded and unloaded, certainly more of a challenge but also enlightening. Here is a slightly revised model: .SUBCKT pu_test 1 2 3 <-- Define "pu_test" as a discrete component with three pins, 1 (IN) 2 (OUT) 3 (GND). An external AC voltage source (~1Vpp) is placed positive to pin 1, negative to pin 3. L1 1 6 {ind/2} <--------- When used like "pu_test 1 2 3 ind=4.831" {ind} is passed as a variable L2 6 4 {ind/2} This way I can perform operations like splitting the inductance as in Lemme's model Rd 6 3 100k <------------ Most components have a value and 2 or more node attachments, see how the resistor connects to the node where L1 & L2 meet K1 L1 L2 0.6 <----------- This is the mutual inductance coefficient between L1 & L2 R1 4 2 {res} <----------- Series DC resistance C1 2 3 {cap} <----------- This cap runs between pin 2 (OUT) and pin 3 (GND), since these nodes are defined above Rload 2 3 2000k <-------- Unreasonably large load resistor, the first letter is specific to the device the rest can be chosen; they are not accessible from "outside" though. .ENDS I hope those comments are helpful. This may be somewhat interesting to you: ltwiki.org/index.php5?title=B_sources_(complete_reference)I'm made some small progress, but you can see I'm using a giant load resistor (Rl) and the mutual inductance seems far too high to be likely. You mention in your PDF how the capacitance is derived from a ratio between the loaded and unloaded frequencies. When I used your value of 120pF the results were much more accurate, could you describe this further? Is it possibly to derive mathematically? I had previously confirmed the measured capacitance of 205pF via the equation 1/((2*pi*f)^2 * L) where 'f' is the self-resonant frequency (Hz) and 'L' the inductance (H). Lastly here is my current 6part model and example netlist: .SUBCKT pu_6part 1 2 3 * Ensure R1||R2||R3 == DC Resistance .param calc_res={1/(1/res - 1/resd - 1/resl)}
L1 1 4 {ind} R1 4 2 {calc_res} Ld 2 5 {indd} Rd 5 3 {resd} Rl 2 3 {resl} C1 2 3 {cap} .ENDS
V1 1 0 AC 1.13 Xpickup 1 2 0 pu_6part ind=5.33 cap=120p res=8.04k indd=33 resd=180k resl=208.9k Cload 2 0 {load_cap} Rload 2 0 {load_res}
.step param load list 1 2 .param load_res table(load,1,10Meg,2,200k) .param load_cap table(load,1,10p,2,470p)
.ac oct 100 100 20k
Attachments:
|
|
|
Post by JohnH on Feb 10, 2017 5:10:36 GMT -5
It looks like you are making interesting progress, which ill read again.
Just on cap values and frequencies. The usual formula relating L and C to the Frequency of peak output is only true in a pure unloaded LC circuit. Once there is load or damping, it drifts away from accuracy and so gives misleading results.
With regard to a numerical solution, I managed to crunch an LRC model to take the deviation into account, when making a curve formula for antiguas table in this thread. But im struggling to fully solve more complex models. So the 6 part models are all developed interatively. I adjust one component to correct one output parameter, and the others all go out of whack but gradually converge. This is done by adjusting each component with resoect to the parameter that it most greatly affects.
In tbe case of C, it has a primary affect on the unloaded frequency and the ratio of loaded to unloaded frequency. The latter is what i was using. Suppose the tested loaded and unloaded peak frequencies were a factor of 2 apart, but with the current estimate of C, the ratio is only 1.5. It means that tbe ratio of C plus 470pF to C needs to increase, ie C needs to be smaller. Its approximately proportionsl to the square root, so if C was the last estimate, the next estimate is C adjusted so that C/(C+470pF) increases by (2/1.5)^2
Due to the damping etc, it doesnt nail it exactly but its a better new estimate for the next iteration.
|
|
|
Post by antigua on Feb 11, 2017 0:08:21 GMT -5
I've set out sampling amplitudes of pickup classes, and at first I tried the "break fine wire away from the string" method with some 41 AWG. It sounds very consistent, but when you look at the dB wave form, it's not very consistent. A difference of 2dB is pretty hard to hear, but it's obvious to the eyes when you see on the screen.
What I settled on is pulling the string away from the fret board above the 12th fret as far as I can get it, then releasing and letting it slam into the fret board. The collision seems to have a normalizing effect. The results are reasonable consistent, but still far from perfect.
|
|
andyholmes
Rookie Solder Flinger
Posts: 12
Likes: 0
|
Post by andyholmes on Feb 11, 2017 17:54:29 GMT -5
John, my simulations so far indicate that the parasitic capacitance is pretty important here, so I was lingering quite a bit on this. I noticed that antigua had posted a graph here that showed the self resonant frequency (SRF) of the '57 Classic at 4.78kHz, as your graph does, which makes the the calculated capacitance 229pF as opposed to 5.06kHz/205pF listed on the interactive table. I couldn't help wondering if the 120pF you use in your model wasn't suspiciously close to half 229pF so I ran side by side sims with 120pF and 115pF and it seems to edge closer to the measured result except the post-SRF slope is slightly worse. The dB peak at higher capacitance loads is lower with both but 6150pF is a heck of a long patch cord. With the coils on this pickup being so comparable, I wonder if this is due to the coils being in series, mutual inductance or just coincidence, any thoughts? Below I have one graph roughly on antigua's scale, and one roughly on yours for comparison. Antigua, are the amplitude measurements you take in dB or V? I'm somewhat unclear on the relationship between dB, dBV and Vpp. I usually work with Vpp either in SPICE or on my scope and my crude tests of my bridge humbucker is around 1Vpp. To be honest every time I thought about testing output I imagined some bizarre contraption of pully weights and a pick on a lever or something.
|
|
|
Post by JohnH on Feb 11, 2017 19:17:04 GMT -5
The approx. 1/2 factor you noticed is just a coincidence. The reason that the simple F=1/(2Pi.(LC)^0.5) formula leads to around 200pF and the 6 part model has much less is because this is a case where the damping, which is embodied in the 6 part model but not in the simple formula, leads to the resonant frequency being reduced with a given L and C. The test results show max measured values, and obviously include the real damping. If there is no damping, resonance happens where the impedance of the L and C are equal and opposite, which gets you to that simple formula. With damping, the greatest output often happens where zC is more than -zL, ie, a lower frequency. The smaller capacitor in the 6 part model is what you need here, and its reactance is greater in magnitude than that of L at the frequency of max output. If we are working out outputs from these component networks using hand of spreadsheet maths from first principles, we need to work in impedances and voltage magnitudes, and ratios of voltages. Once the maths are done, the results are best presented in db. This is what SPICE sims do and it is also what is presented in antiguas test numbers and charts. For these purposes, db =20log (voltage ratio), where log is base 10. So, x10 voltage ratio between say input and output is 20db. The reason we use 20 and not 10 is that it gives a representation of the amount of power that could be produced downstream if this was amplified. For example, if you had x10 voltage you would also get x10 current so x100 power, 2 orders of magnitude, 20 tenths of a Bell where one Bell is x10. Although we are not dealing with power in the pickup calcs, this convention is still used. Here are some diagrams that illustrate how to visualize all this: These are plotted using complex numbers. resistances are across the page, inductive reactions are up and capacitavive reactions are down. We work with impedances, then relate the relative values of voltages to the ratios of impedances, the same way as with a pure resistor network. In addition to deriving relative magnitudes, the phase angle between input and output can also be worked out, though it is not needed so often. Part 1 shows a general condition of an RLC resonant circuit, where R is relatively small, and it is shown at a frequency above resonance, so impedance of C is less than of L. We follow a series chain of R, L and C (ie, as they go from input, in series with each other, to ground), and input is applied across the whole of that chain and output is just across the capacitor. Part 2 is at resonance, where output is max. If R is small, it is where zL = -zC and you can see on the sketch how this results in the greatest ratio between Vin and Vout, and it also illustrates how we get more output voltage than input voltage if damping is low. Part 2B - this is a variation where R becomes significantly high, and it is one way to add damping. You can visualise here how in this case, the max output to input ratio may be when zC is greater than -zL, ie at a lower frequency than previously calculated. This condition of an increased R in a three-part model was what I used to get the curves for antiguas charts. It is the easiest way to picture how adding damping may reduce the frequency where output is max, and it creates credible curves though absolute impedance and absolute phase are not correct, hence not good enough for adding further components in SPICE models etc (see below). Part 3 adds an extra R2 to provide damping instead of increasing R1. It is moving towards the 6 part model and now we have to figure reactance's in parallel. I've put a formula there for R2 and C in parallel. Its not so obvious as 2B but it also implies capacitive reactance is greater than inductive at max output, and so resonant frequency is reduced. The curves produced by this model can be identical to 2B above, with the same relative phase angle between input and output, but absolute phase is different and the total impedance magnitude is different and also varies with frequency. So this four-partmodel is moving towards being a good representation of a pickup-, while 2B is more just a useful visualization tool. Is that any use as an explanation?
|
|
andyholmes
Rookie Solder Flinger
Posts: 12
Likes: 0
|
Post by andyholmes on Feb 11, 2017 21:12:54 GMT -5
Yes, that's very useful overview since this is my first real foray into RLC circuits. I'm going to break out wxMaxima and keep plodding along, thanks.
|
|
|
Post by antigua on Feb 20, 2017 15:34:34 GMT -5
I've been messing with LTSpice trying to wrap my brain about how simply separating a coil into two halves can case the overall parasitic capacitance drop become half what it was, but if the modeling is to be believed, that's how it plays out. So I measure 110pF for the Antiquity neck. If I subtract 70pF for the cable, that leaves 40pF remaining. Two 80pF coils in series theoretically give you 40pF total. 80pF per each 3.8k 42 AWG coil is not unrealistic for a machine wound coil with a higher tension and uniform layering, so that at least gives me confidence that the numbers are not astronomically wrong, but I agree with your point that the inductance is going to be lower at the resonance than what is measures at 120Hz, though that would mean the capacitance would therefore have to be revised upwards even further to offset the lower inductance. One way to look at the series capacitance thing is to first say that series impedances always add. This is true for Rs, Ls and Cs. Then, since the impedance of a capacitor varies with the inverse of the capacitance, while the impedances add, the value of the total capacitance must go down.This is still messing with my mind. Parasitic capacitance in a coil is a bad thing, people want to get rid of it, and we see that in a series humbucker the capacitance is very low because you're combining capacitance in series. So if you want a single coil with low capacitance, why not just stack concentric coil in series in order to get that same series drop of capacitance? I know if doesn't work that way, but I don't know why it does not. It appears that the lack of proximity between the coils plays a role; that if you take the two halves of a series humbucker and stack them together on the same bobbin, then you have yourself a Strat pickup (or something remarkably close to Telecaster bridge pickup), and suddenly you have way more capacitance. But it wasn't the coils that changed, it is merely how close the two halves are to one another. I do not see a solid explanation for why this is, unless modelling coil capacitance as a parallel lump sum is grossly inaccurate.
|
|
|
Post by JohnH on Feb 20, 2017 16:21:11 GMT -5
The bold part above is clear enough, reactance of a capacitor is inversely proportional to capacitance.
I think the other conundrums come from the limitation of the simple RLC 'flat earth' model of a pickup. Im calling it that because like thinking of the earth as flat, it is useful for certain things but breaks down if pushed too far. Its not so much a wrong model but it is a crude model.
With regard to capacitance, a string of two RLC models in series would double R and L and halve C. A real test of two coils or two pickups would deviate. At least R would follow the simple expectation but L and particularly C would not. C is just a derived value dependent on the model we are using to represent reality, not on reality itself.
The real combined effect of capacitive effects within a pickup would be a combination of capacitances not just from one winding to the next, but from each winding to every other winding, plus some directly from each winding to the leads and ground. A massively complex network/contimuum. Some of these will tend to always partly add up whether in series or parallel, particilarly those that go directly or closely to ground, and some will work more as expected adding RLC models in series.
|
|
|
Post by antigua on Feb 20, 2017 16:41:33 GMT -5
With regard to capacitance, a string of two RLC models in series would double R and L and halve C. A real test of two coils or two pickups would deviate. At least R would follow the simple expectation but L and particularly C would not. C is just a derived value dependent on the model we are using to represent reality, not on reality itself. The question still stands though: is a single coil not analogous to two (or more) RLC models in series? The point at which you decide the coil stop and start up again in arbitrary. And yet, as you add more windings, C increases, it doesn't decrease. The real combined effect of capacitive effects within a pickup would be a combination of capacitances not just from one winding to the next, but from each winding to every other winding, plus some directly from each winding to the leads and ground. A massively complex network/contimuum. Some of these will tend to always partly add up whether in series or parallel, particilarly those that go directly or closely to ground, and some will work more as expected adding RLC models in series. Here's a very big question I need to find or receive an answer to: do the innermost windings also couple with the outermost windings, or do the intervening wingings in between act as a shield? Another way to phrase the question, if the entire coil capacitively related to itself, or are windings only capacitively related to their nearest neighbors?
|
|
|
Post by antigua on Feb 22, 2017 0:46:45 GMT -5
I've been reading about coil capacitance, and I came across an interesting revelation here coil32.net/theory/self-capacitance.htmlWhat this is saying is that the windings only capacitively couple with their neighbors who are perpendicular to the axis. So if you have a pickup sitting on a table, windings capacitively couple with windings to their left or two their right, but not above and below. The interesting thing about that is that it means a single layer coil has about the same capacitance as straight wire, because such a coil only has neighboring wire above and below. Of course a pickup is multilayer, but the point of it all is that the winding capacitance is limited to one plane throughout the coil. A practical application is that if you want to manufacture low C coils, some sort of layer spacers, or spacing agent, might do the job, such as a layer of lacquer or wax. You wind, then dip, then wind, then dip, etc.
|
|