Emulating a fancy LCR meter

[Joe Gwinn]

I think the cheapest way to emulate the fancy LCR meter is with a computer and a good sound card. I've been thinking about this since the last time it came up, and a sound card (or pair of cards) that can simultaneously generate and listen could work.

Basically, one uses the left and right input channels to sense pickup input and output signals, and computes the relative phase and amplitude. As the drive frequency varies, phase and amplitude will also vary, and from this one can compute the inductance and AC resistance as a function of frequency.

The details would need to be worked out, and I would imagine that not all sound cards are equally suitable. There is also the matter of writing the code that drives the soundcard or cards and does the necessary computations.

The first question is if anyone has already done this.

[belwar]

I've got several extra good computers and would put this to the test if someone wants to give me detailed instructions. I dont have very high end sound cards, but I do have Audigy 2 XFi's.

[dai h.]

not sure if this is germane (but thought it might be) :

build a unique PC sound card impedance bridge

http://www.arrl.org/qex/2005/Steber.pdf

another one but it's not on the net anymore apparently as well the program if memory serves:

http://web.archive.org/web/20020810023529/http://www.publi.ro/ady/

just some links I remembered seeing before when googling about LCR meters and such

[Joe Gwinn]

not sure if this is germane (but thought it might be) :

build a unique PC sound card impedance bridge

http://www.arrl.org/qex/2005/Steber.pdfThis seems right on point. I'll have to read the article. I built a simple I-V interface to do much the same, but for signal generator and oscilloscope, not computer. It looks like Steber has an I-V interface there too.


For the audience, an explanation of "I-V Interface" is in order:

This is usually implemented as three opamps.

The first opamp buffers the test signal, driving one end of the component under test.

The second opamp is wired as a high-impedance voltage follower, takes its input from the same end of the component under test as the first opamp drives, providing a buffered voltage signal to the scope (or soundcard) input.

The third opamp is wired as a transimpedance amplifier, which converts current into voltage. The other end of the component feeds the input of the transimpedance amplifier, providing to the scope (or soundcard) input a voltage proportional to the current through the component.

Opamps two and three therefore respectively yield signals proportional to the voltage across and to the current through the component. Impedance is the voltage divided by the current.

The twist is that with reactive components like inductors and capacitors, current and voltage are not in phase. It is this phase deviation versus frequency that allows us to compute inductance and capacitance values.

another one but it's not on the net anymore apparently as well the program if memory serves:

http://web.archive.org/web/20020810023529/http://www.publi.ro/ady/The link didn't work for me.

just some links I remembered seeing before when googling about LCR meters and suchIf you find some other references, please post them.

[dai h.]

http://web.archive.org/web/20020810023529/http://www.publi.ro/ady/

The link didn't work for me.



it's from the web archive ("wayback machine"). Can you see this page (some unknown language--Romanian?? seems to sound similar to Italian) : ?

For Enhancer section please click here Pagina este intr-o constructzie eterna
Mai dar ce bine c-ai aterizat in pagina mea...:)
Gaseshti pe aici trei programele shi... inca ceva...

Shi uite ca m-am gandit sa ma dau shi eu pe internet... Nu de alta dar nu aveam ce face mai bun... :). Ce gaseshti pe aici ? Trei programe pe care le-am facut pt. ca m-a pasionat domeniul audio... Calitatea lor va spune cat de mult am fost pasionat...:))) Pt. cine este curios despre ce poate intampla in "cutiutza" mea, expun aici unele din gandurile mele... shi sa nu uit, mai este o poezie care-mi place in mod deosebit...

Ce altceva mai fac eu ?
Dupa ce am facut o versiune "publicabila" a programului MultiMeter, revin la o pasiune mai veche... care s-a trezit de putzina vreme. AI. Pana imi gasesc ceva mai pamantesc cu care sa-mi omor neuronii, o sa ma ocup de intelighentzia artificiala... Ureaza-mi spor la treaba shi scrie-mi ce parere ai de pagina mea... O ultima chestie : in masura in care mai fac cate un programel sau o sa am (shi alt-) ceva de spus, o sa upgradez shi aceasta pagina...

Numarul de vizitatori
incepand cu 27 Aprilie 2000


up on top where it says "Multimeter", it leads to a page that says this:

MultiMeter...

After a 2-week hard work, (with large breaks in between :), this program has started to work... quite well. What does it do? It measures pieces… like any other multimeter… Why have I done it? Well, I believe this is a much cheaper version of a multimeter, which measures capacities, inductances and resistances. It should become useful to any amateur electrician (or to a poor one), who has a computer (486 DX4, 16Mb minimum requirements) and Windows 9x.

Click to Download In order to make the program work, you will need Windows 9x, a Full duplex sound card with already installed drivers, 2 jack stereo couplings with 1-2 meters wires and a resistance. Download this program (100Ko).
Conections diagram This is the way you must make the connections. Sticking some wires and a resistance shouldn’t be too difficult. You will choose one channel from the speaker exit of the sound card (or you will join both channels using resistances of minimum 8 Ohm and further on the channels will be connected between the resistances), you will make all the other marked connections (to LineIn) and... GO!
The program should have very accurate information about the value of the resistance. The measurement scale is depends on the level of the noise. If you measure values that are comparable to the noise (which you will find at the open calibration, that means without the measurement piece), then the program won’t display the value. The noise and the continuous current, typical of any sound card, are successfully reduced using some recursive filters. Probably in a future version (if anyone should be interested in this program) I will also insert a virtual sample that will help me to reduce even more the disturbing influences. It will consume more resources of the CPU, but "precision before everything :)" At this time, the program works at an amateur level.
Let’s get down to the facts. How do you use it? Supposing that the wires are already sticked (I hope that you’ve already understood how to do that)... After you start the program, you will make the Short calibration (the wires where the piece should be found short-circuit among themselves). You will wait a while, until the Err indicator reaches a minimum value (and possibly a stable one, depending on the sound card). With the wires still short-circuited, you will switch the program to the Open calibration, then you will unbind the wires. You will also wait for a value that is minimum and as stable as possible. Now switch the program to its measurement phase and you can measure different pieces.
The volumes are highly important! The quality of the measurement depends very much on the distortions of the signal. A simple version that regulates the volumes is the following: The exit volume is regulated at approximately one half (no more than 3 quarters even if you have a very good sound card). With your program already going, independent of the work mode, you will regulate the LineIn entrance volume until the indicator at he recording control (of the system) is yellow at most (this also means approximately at a half)... This way you should get minimum errors in most cases.
Does anyone understand what I’m talking about?

About how this program works (although I doubt that anyone should be interested)... How does the program measure the pieces?
Resistances : the difference between the intensity of the signal between the channels.
Inductances/capacitors : the difference between the intensity of the signal between the channels and, as a plus, the phase difference between them.
What is the difference between the 2 types of measurements? The first uses only one frequency (the one in the left) and is based only on the phase difference. The second one uses both frequencies and the result is provided by a system of 2 equations with 2 unknowns ... gotcha, ha? LOL The 2 frequencies shouldn’t be too close (under 10%) or too different (over 200%). If they are identical (I haven’t even tried it, believe me?), the second method can’t work. The second method has one more advantage: it can make the difference between inductances and capacities, and this is why it’s appropriate in these cases. As for the resistances, it doesn’t work too well.

As I’ve also written in the program, if you like and use it, please send me an email, so that I can see what you think of it (is it too expensive? :). Enjoy using it! Oh! I’ve almost forgotten! Don’t try the program on the speakers! You may not like the sounds it produces!

How does it work on my computer ? Well, on my sound card (YMF 724) I measure in a continuous scale, from 0.22 uF to over 1000 uF, with the frequencies set at 100/1000 and 20 Ohm serial resistance. If you want to measure small condensers (like 0.01 uF), try on bigger serial resistance (but not bigger than 1000 Ohm) and larger frequencies (but not larger than 1000 Hz). The resistances in a scale from 1 Ohm to some 10kOhm, again with the serial resistance of 20 Ohm. About the coils... not quite good news... the program works only on paper. What does that mean? I don’t have any standard inductances in order to check if it works accurately. However, by the way I designed the program, it should work. Maybe one of the users will let me know how the program works with the inductances.

Well, I’ve been quite in a mood for talking, ha? :)

I will take no responsability for any damage or loss of data resulting from the use or misuse of this program.
You are using it at your own risk.


(I've attached the image from the article also)

[Joe Gwinn]

it's from the web archive ("wayback machine"). Can you see this page (some unknown language--Romanian?? seems to sound similar to Italian)

[snip]

(I've attached the image from the article also)It does seem to be Romanian, from the URL.

The link now works for me.

http://web.archive.org/web/20020811015349/http://www.publi.ro/ady/pics/mm_diagram.gif

The series resistor converts current through the component into voltage applied to the Right input channel. The Left channel gets the voltage across the component (and resistor).

This is a primitive I-V interface. The problem is that to get enough signal the series resistor must be large, but this interferes with seeing the voltage across the component uncontaminated by voltage across the series resistor.

The opamp approach solves this problem by using a transimpedance amplifier in place of the series resistor.

[Mike Sulzer]

For the audience, an explanation of "I-V Interface" is in order:

This is usually implemented as three opamps.


A couple of comments:

1. Why so many op amps? The voltage at the output of the first op amp is what you want to measure. Even if the input to the sound card were to load it a bit (unlikely), it is still "the voltage", and V/I is the impedance. The same thing applies to the first op amp. The sound card only needs to be buffered if the load of the Z would make it non-linear.

2. Maybe I-V is not the best way to go. The region of greatest interest for a pickup is near the resonant peak, second greatest, below the peak down to maybe 80 Hz. The peak is the frequency where the output voltage of the TIA is smallest (lowest current through Z), and so most subject to relative error. The relative error is important because we would be dividing by this small number. The other approach, driving the Z with a current source derived from the soundcard, and buffering the voltage across the Z with a follower gives the highest voltage at the peak, and the division by the current is pretty much just scaling by a complex constant. This method would have more trouble at the high and low ends, but the low end is just resistive, easily checked with a meter, and the top end is generally irrelevant. You have to make sure the current source does not saturate near the peak, but it is not hard to monitor the level, or to put an overload sensor on it.

[Enzo]

Not my area of expertise for sure, but we used to have some software from JBL, and I am sure others make similar products, that analyze sound providing a three dimensional graphic. Freq, amplitude, phase - or something like that. The result is one of those plots that looks like a terrain map of a rolling countryside. I want to say it had a name like SOund Master, but it could have been anything. This was several years ago, but I am sure audio analysis software is still around. It was intended for setting up PA systems in large spaces. Due to the three dimensional display, you could see relationships between variables that might not be so evident when considering them individually.

Would something like that help here. Is my description remotely clear?

[Joe Gwinn]

A couple of comments:

1. Why so many op amps? The voltage at the output of the first op amp is what you want to measure. Even if the input to the sound card were to load it a bit (unlikely), it is still "the voltage", and V/I is the impedance. The same thing applies to the first op amp. The sound card only needs to be buffered if the load of the Z would make it non-linear.The first opamp can be replaced with the soundcard, if it's up to it. But opamps are cheap, and will render the details of the soundcard output irrelevant, as well as buffering the transmission line from soundcard to component. One usually drives the component through a current-limiting resistor to protect both driver and transimpedance amplifier. (Not to mention sprinkling some ESD protection diodes around the component terminals.)

2. Maybe I-V is not the best way to go. The region of greatest interest for a pickup is near the resonant peak, second greatest, below the peak down to maybe 80 Hz. The peak is the frequency where the output voltage of the TIA is smallest (lowest current through Z), and so most subject to relative error. The relative error is important because we would be dividing by this small number. The other approach, driving the Z with a current source derived from the soundcard, and buffering the voltage across the Z with a follower gives the highest voltage at the peak, and the division by the current is pretty much just scaling by a complex constant. This method would have more trouble at the high and low ends, but the low end is just resistive, easily checked with a meter, and the top end is generally irrelevant. You have to make sure the current source does not saturate near the peak, but it is not hard to monitor the level, or to put an overload sensor on it.I-V is how the $15K instruments work. I read lots of patents researching this.

I imagine that it's a lot harder to implement a current source than a voltage source at high frequencies. Stray capacitance is going to be a real problem with current sources, which are by definition high impedance.

The Q of a pickup at resonance isn't that high, so while you are correct that the current is minimum at resonance, it isn't all that small either.

[Mike Sulzer]

I-V is how the $15K instruments work. I read lots of patents researching this.


Yes, a general purpose instrument would use I-V because it is best for most impedances.

What do we really want to know about a pickup? I think it is the frequency and width of the resonance when loaded by the C of the cable and the R of the input of the amplifier. I think that this is most likely to be an indicator of the pickup's tone. One should know the inductance as well, measured at some intermediate (not too high) frequency, low enough to not be affected by the C of the pickup. A pickup measuring machine should measure the device in the proper environment, and the parameters of this environment should be selectable.

By the way, I did some modeling of the results of the coil that I measured with/without slugs (results on the other discussion). For the case with the slugs, the high frequencies and low frequencies are best fit with inductances about 7% different. The case with no slugs does not require using two different inductors. This is a pretty small effect, but it does look real.

[Mike Sulzer]

Here is something to remember about computer-based measurements: You have available both tremendous computing capability and very flexible methods for adapting it to the situation at hand. Thus, rather than have a complicated analog system which the computer samples, one can use a simple analog system, but one that can be well-characterized. For example, instead of a current source (complicated), one can use a large-value resistor, and then correct for its finite value in the computer code. The resistor should not be much smaller than the largest value expected for the unknown impedance, or accuracy can be lost. Also, the TIA could be replaced with a small resistor (as in the diagram above), and the computer code could correct for its effect. The resistor should not be much less than the resistance of a pickup in the case of interest here.

For example, in the modeling that I mentioned in my previous post, the 900K resistor in series with the output of the generator is an essential part of the model. It would affect the accuracy of the measurement near the peak somewhat, but its effect can be taken out in the calculations.

[Joe Gwinn]

Yes, a general purpose instrument would use I-V because it is best for most impedances. Yes. Even twenty years ago it was a big deal to homebrew such an interface, but now all it requires is some jellybean parts.

What do we really want to know about a pickup? I think it is the frequency and width of the resonance when loaded by the C of the cable and the R of the input of the amplifier. I think that this is most likely to be an indicator of the pickup's tone. One should know the inductance as well, measured at some intermediate (not too high) frequency, low enough to not be affected by the C of the pickup. A pickup measuring machine should measure the device in the proper environment, and the parameters of this environment should be selectable.Well, we should design an instrument that can characterize a pickup without coloring the answer with instrument limitations.

By the way, I did some modeling of the results of the coil that I measured with/without slugs (results on the other discussion). For the case with the slugs, the high frequencies and low frequencies are best fit with inductances about 7% different. The case with no slugs does not require using two different inductors. This is a pretty small effect, but it does look real.Modeling? Why not hack together a simple physical experiment, and make actual measurements? This is how one validates a model.

[Mike Sulzer]

Modeling? Why not hack together a simple physical experiment, and make actual measurements? This is how one validates a model.

I did make the measurements; they are available in the other thread. The modeling consisted of finding networks with computed curves that match the measured ones.

[Joe Gwinn]

I did make the measurements; they are available in the other thread. The modeling consisted of finding networks with computed curves that match the measured ones.Oh. From the original quote, it sounded like all theory no data. Data is good.

[Mike Sulzer]

Here is an update on my efforts to use to use a computer with soundcard for measuring pickups:

I tried an I-V circuit. This circuit, as discussed above, gives a low voltage at the higher frequencies. There is too much nose pickup, erratic noise probably from the computer hardware, from about 4KHz and above. (I am using a mac book with its internal soundcard.) There are ways around this, no doubt, but a different circuit topology (V-I) puts the higher voltage output at the higher frequencies. This circuit and some comparisons of measurements and modeling are here:
http://www.naic.edu/~sulzer/coreNoCoreComp.png

The basic circuit uses a resistor in the input path of the op amp, and the pickup is the feedback impedance. The noise source is connected to the input through a low pass filter, and the input R is made smaller to keep the high frequency gain high enough. This increases the accuracy of the measurement at very low frequencies. Sine wave testing shows a noisy waveform at the very low frequencies if you do not do this, and it takes a longer time to get a good measurement.

I use SignalScope to collect the data. The measurement averages 1000 1024 point FFTs. The resulting 513 point spectrum is saved in a disk file. The file is read into a program for computations and graphics. (It would be better to have a single application that would be a "pickupMeter".) The zero frequency point is discarded, and then the spectrum from the output of the op amp (V1) is divided by the one from the input(V2), and multiplied by the input R (9.74K in my case). Also measurement must be corrected for a 2.2% difference in gain between the two channels. The resulting measurement is much less noisy than one might expect. 1000 samples gives about 3% fluctuation (noise), but of course the division takes out most of it because the fluctuation is mostly the result of randomly different signal levels at different frequencies, affecting both V1 and V2 in the same way.

The measurement used a single coil of the type used in a humbucker. The resistance is about 3.88K, so this is a bit under wound by some standards. The cores can be removed and replaced, of course, and the purpose of this measurement is to compare the coil parameters in the two cases. The software also allows modeling the magnitude of the impedance, that is, making curves that attempt to match the measurements. The plot shows both measured and modeled impedances with and without the cores. These cores are standard humbucker slugs, bought from Allparts a few years ago.

550 pf is connected across the coil so that the resonant frequency with cores is near the top end of the response of a guitar speaker.

For the case with cores, two models are shown, one that matches at the peak, one that matches at about 700 Hz. With no cores, one model pretty well matches the whole frequency range up to the peak. There might be some additional effects above the peak not in the model.

Modeling parameters with no cores:
Cp = 620 pf (assumed 70 pf coil cap, added 550pf across coil)
Lp = .635H
Rp = 3.88K (resistance of the coil)
Rq = 100 M (across pickup, very large, not really needed in model without cores)

Modeling parameters with cores:
Cp = 620 pf
Lp = 1.02H (high frequency model), 1.22H (low frequency model)
Rp = 3.88K
Rq = 295K (used to model the losses from eddy currents in the cores)

As the plot shows, the high frequency model works at the peak, while the lower frequency model agrees at about 700 Hz. Some additional modeling (not shown on plot) indicates that the inductance increases roughly another 20% near 100 Hz. So from the bottom up to the resonance, the inductance changes about a factor of 1.4 or 1.5.

So these measurements indicated (as expected) that there are are two important effects on the pickup impedance from using cores:
1. The inductance is increased over the same coil with no cores by more than a factor of two at the lowest frequencies, significantly less at the higher frequencies.
2. The losses from the cores are significant, equivalent to about a 300K resistor across the pickup in this case.

It would be interesting to see what the effect of alnico magnets is in a single coil Fender-type pickup.

[Joe Gwinn]

Here is an update on my efforts to use to use a computer with soundcard for measuring pickups:

I tried an I-V circuit. This circuit, as discussed above, gives a low voltage at the higher frequencies. There is too much noise pickup, erratic noise probably from the computer hardware, from about 4KHz and above. (I am using a mac book with its internal soundcard.) There are ways around this, no doubt, but a different circuit topology (V-I) puts the higher voltage output at the higher frequencies. This circuit and some comparisons of measurements and modeling are here:
http://www.naic.edu/~sulzer/coreNoCoreComp.png
I'm a little unclear as to what was tried and why it failed. Voltage followers and transimpedance amplifiers are used well into the gigahertz region. Could you also post the failed "I-V circuit"?

As for noise pickup, I built my circuit on a piece of vectorboard mounted in a diecast metal box, with BNC inputs and outputs, and binding posts for the pickup under test, and had no problems. The dark side of sensitivity is sensitivity -- full shielding is required.

I also used OP27 opamps, versus generic ~741s, and had no problem making measurements at 20 KHz (limit of interest, not OP27 capability). The 741s didn't work quite well enough, as I recall, but I don't recall exactly what the problem was, although limited open-loop bandwidth is a likely contender.

http://www.analog.com/en/other/militaryaerospace/op27/products/product.html

This effort started in May 2003, when I was first trying figure out how to accurately measure pickups. The use of the Maxwell Impedance Bridge and discovery that the Extech LCR meter was suitable arose from this same effort.

[David Schwab]

Mike, do you have access to either a Firewire or USB audio interface? I never used the audio interface in a Mac laptop, and my G4 doesn't have an audio in jack (?!?), but on older Macs I've had, the built in audio wasn't the cleanest I've heard.

Before I put an M-Audio PCI card in the G4 I used a Griffin iMic USB interface which seemed to work OK.

[Mike Sulzer]

I'm a little unclear as to what was tried and why it failed.

Joe, I am sorry about the unclear writing; not the first time nor the last. The analog circuitry did not fail, it was the system as a whole. As David suggests, the problem is likely with the computer. Although I cannot prove exactly where the problem is occurring, I believe it is cross talk between the digital circuitry and the sound card in the computer. A lesser problem is leakage from the outputs of the sound card back to the inputs at high frequencies.

I am using 1/4 of an LM837; it is a good amp for this purpose. 25 MHz GBP, 200KHz power bandwidth, 10 V/microsec slew rate, low noise for pickup type impedances (bipolar input). I have not even thought about using a 741 in an audio circuit for, well, about 35 years. The slew rate is too low to reproduce 20 KHz at full output levels, the gain is too low, and they sound dreadful.

The I-V circuit that I tried is just like the one on the figure except the pickup goes in the input path and a resistor in the feedback path. No followers are required; the computer output is designed to drive headphones and provides 1 V rms into the load used here.

One could use the I-V circuit by increasing the signal in at higher frequencies. But the circuit I am using just does this in a simpler way.

David, yes, I have both those inputs on the computer. If you are suggesting that all the analog circuitry should be outside the computer, yes, that would be better. But that would violate the main goal of this project, which is to measure pickups with a laptop and as little external circuitry as possible. I would do it only with passive components, if possible, but one op amp, either as a buffer or impedance converter is required. The converter has the best performance.

[David Schwab]

David, yes, I have both those inputs on the computer. If you are suggesting that all the analog circuitry should be outside the computer, yes, that would be better. But that would violate the main goal of this project, which is to measure pickups with a laptop and as little external circuitry as possible. I would do it only with passive components, if possible, but one op amp, either as a buffer or impedance converter is required. The converter has the best performance.

I'd just like to have one that works! This looks like a great idea.

UBS audio interfaces are pretty cheap. The iMic looks like a yo-yo (though the new ones are larger) and it's only serving as a sound card external to the computer.

I know it's hard enough to get a noise free recording just being close to my digital mixer.

[Mike Sulzer]

I'd just like to have one that works! This looks like a great idea.

UBS audio interfaces are pretty cheap. The iMic looks like a yo-yo (though the new ones are larger) and it's only serving as a sound card external to the computer.

I know it's hard enough to get a noise free recording just being close to my digital mixer.

A cheap external USB sound card is not necessarily any more immune to the pickup of digital noise than the internal card. RDigital circuitry is required for the interface, and a poor design might not isolate this well enough.

[Mike Sulzer]

I want to look at the characteristics of some pickups, starting with a humbucker. This pickup was purchased from Allparts some years ago, and is supposed to be something like a PAF replacement. Humbuckers such as this use two coils in series, and so have high inductance, and low capacitance. The resonant frequency is quite high when they are connected to a very high impedance load, but the high frequencies are easily lost in the real world of a volume and tone controls, guitar cable and amp input resistance.

Let's begin by looking at the effect of the cover with the pickup unloaded. The blue line in the plot (http://www.naic.edu/~sulzer/japanHumBuc.png) shows the impedance with no cover; the red line shows the impedance with the cover that came with the pickup. This is a truly dreadful cover. The red line deviates from the blue at less than 2 KHz. Remember, there are two effects from the eddy currents induced in the cover. First, they dissipate energy, and so load the pickup down, equivalent to a resistor connected across the pickup. This effect tends to be important at the higher frequencies where the impedance is high. The other effect is that the currents tend to reduce the inductance. One expects this effect at the higher frequencies as well, but in this case it starts well down in the mid-range.

The green line is the results for a better cover, also purchased from Allparts. It appears to have little effect through the range of a guitar speaker.

The black and light blue lines show the impedances with no cover, but the pickup loaded with some external circuit. The black line shows the effect of 550pf, typical of a guitar cable. The resonance moves down to under 4 KHz with moderate Q.

The additional loading on a humbucker is usually a 500K volume control, a 500K tone control, and a 1 meg input resistance to the amp. When the two controls are on 10, they have a parallel resistance of 250K. (The tone cap is not important with the control on 10; it is like a short.) The amp input brings the impedance down to 200K. So the light blue line is the impedance with no cover, 550pf, and a 200K resistor. Note that the Q is very low, and the impedance begins decreasing at less than 3 KHz.

For some, an explanation of what this impedance means would be useful. The operation of the pickup is described with a voltage source in series with the inductor. The inductor and capacitor form a so-called second order lowpass filter, and the resistance determines the Q, or damping of the filter. (The resistance in series with the L and any equivalent resistance from eddy currents contribute to this resistance.) The low pass filter has a flat response at the low frequencies; it rises near the peak determined by the L, C, and R, and then falls off quickly. The values of these components are determined by the impedances on the plot, and so roughly speaking, one can determine the peak and Q of the filter by reading the values off the plot. The data can be used to calculate these values with good accuracy.

So this pickup, IMO, qualifies a a bit of a dud, especially if the original cover is left on. A good candidate for the dead pickup's society. (I suspect that the base plate is also too conductive, but but I will leave testing and replacing it for another time.)

[Joe Gwinn]

Joe, I am sorry about the unclear writing; not the first time nor the last. The analog circuitry did not fail, it was the system as a whole. As David suggests, the problem is likely with the computer. Although I cannot prove exactly where the problem is occurring, I believe it is cross talk between the digital circuitry and the sound card in the computer. A lesser problem is leakage from the outputs of the sound card back to the inputs at high frequencies.OK. Low impedance drive of signals going to the sound card can help. As can a better sound card.

Sound card. We need to figure out which sound cards are best. I'd like one for the Mac as well. Any worthwhile soundcard ought to be able to achieve 60 dB isolation, and if the analysis software takes advantage of the fact that we know the drive signal frequency, we ought to be able to achieve profound reductions in the effect of stray (uncorrelated) noise.

I am using 1/4 of an LM837; it is a good amp for this purpose. 25 MHz GBP, 200KHz power bandwidth, 10 V/microsec slew rate, low noise for pickup type impedances (bipolar input). I have not even thought about using a 741 in an audio circuit for, well, about 35 years. The slew rate is too low to reproduce 20 KHz at full output levels, the gain is too low, and they sound dreadful.The LM837 ought to work OK. With four amps in the package, there is no reason one cannot have full buffering. I usually build such things into a diecast box with two 9-volt batteries to supply +/- 9 volts as Vcc and Vss.

As for the 741, I'll try to keep it clean in the future...

The I-V circuit that I tried is just like the one on the figure except the pickup goes in the input path and a resistor in the feedback path. No followers are required; the computer output is designed to drive headphones and provides 1 V rms into the load used here.No followers? No transimpedance amps? Therein lies the problem.

One could use the I-V circuit by increasing the signal in at higher frequencies. But the circuit I am using just does this in a simpler way.But it seems to be too simple to work well.

I'll publish my circuit in a few days. It is not complex. But I will have to draw it up.

David, yes, I have both those inputs on the computer. If you are suggesting that all the analog circuitry should be outside the computer, yes, that would be better. But that would violate the main goal of this project, which is to measure pickups with a laptop and as little external circuitry as possible. I would do it only with passive components, if possible, but one op amp, either as a buffer or impedance converter is required. The converter has the best performance.Change that to one IC (with 4 opamps) and we can do far better. My original circuit uses only two of the opamps, plus a third to buffer the soundcard output.

[Mike Sulzer]

No followers? No transimpedance amps? Therein lies the problem.

But it seems to be too simple to work well.


The circuit with the pickup in the feedback loop seems like the right one given the impedance characteristics of a pickup (rising with frequency up to the resonance). This amp has low output impedance. The input voltage sample has low impedance at high frequencies (where it matters most) because it is taken across a capacitor.

But I will try followers in different places in the circuit, one at a time, and see what the differences are.

[soundmasterg]

I want to look at the characteristics of some pickups, starting with a humbucker. This pickup was purchased from Allparts some years ago, and is supposed to be something like a PAF replacement. Humbuckers such as this use two coils in series, and so have high inductance, and low capacitance. The resonant frequency is quite high when they are connected to a very high impedance load, but the high frequencies are easily lost in the real world of a volume and tone controls, guitar cable and amp input resistance.

Let's begin by looking at the effect of the cover with the pickup unloaded. The blue line in the plot (http://www.naic.edu/~sulzer/japanHumBuc.png) shows the impedance with no cover; the red line shows the impedance with the cover that came with the pickup. This is a truly dreadful cover. The red line deviates from the blue at less than 2 KHz. Remember, there are two effects from the eddy currents induced in the cover. First, they dissipate energy, and so load the pickup down, equivalent to a resistor connected across the pickup. This effect tends to be important at the higher frequencies where the impedance is high. The other effect is that the currents tend to reduce the inductance. One expects this effect at the higher frequencies as well, but in this case it starts well down in the mid-range.

The green line is the results for a better cover, also purchased from Allparts. It appears to have little effect through the range of a guitar speaker.

The black and light blue lines show the impedances with no cover, but the pickup loaded with some external circuit. The black line shows the effect of 550pf, typical of a guitar cable. The resonance moves down to under 4 KHz with moderate Q.

The additional loading on a humbucker is usually a 500K volume control, a 500K tone control, and a 1 meg input resistance to the amp. When the two controls are on 10, they have a parallel resistance of 250K. (The tone cap is not important with the control on 10; it is like a short.) The amp input brings the impedance down to 200K. So the light blue line is the impedance with no cover, 550pf, and a 200K resistor. Note that the Q is very low, and the impedance begins decreasing at less than 3 KHz.

For some, an explanation of what this impedance means would be useful. The operation of the pickup is described with a voltage source in series with the inductor. The inductor and capacitor form a so-called second order lowpass filter, and the resistance determines the Q, or damping of the filter. (The resistance in series with the L and any equivalent resistance from eddy currents contribute to this resistance.) The low pass filter has a flat response at the low frequencies; it rises near the peak determined by the L, C, and R, and then falls off quickly. The values of these components are determined by the impedances on the plot, and so roughly speaking, one can determine the peak and Q of the filter by reading the values off the plot. The data can be used to calculate these values with good accuracy.

So this pickup, IMO, qualifies a a bit of a dud, especially if the original cover is left on. A good candidate for the dead pickup's society. (I suspect that the base plate is also too conductive, but but I will leave testing and replacing it for another time.)

That was very interesting!

So was the original cover that came with the pickup made out of brass? I've noticed most of the pickups with a brass cover tend to sound like crap, with high end harshness, and the one with nickel silver covers usually sound better.

Greg

[Mike Sulzer]

That was very interesting!

So was the original cover that came with the pickup made out of brass? I've noticed most of the pickups with a brass cover tend to sound like crap, with high end harshness, and the one with nickel silver covers usually sound better.

Greg

I do not know, but when I get home sometime tonight, I will scrape off some of the plating and take a look.

Mike

[Mike Sulzer]

Joe,

I put a follower between the voltage input, that is, the top of the capacitor (V2), and the output to the sound card. This does indeed make a difference. Typically a bit more than 1% of the impedance value, but more like 2% near the peak (13 -14 kHz) and decreasing to near zero at about 20KHz. So the resistor in the input path of the op amp and the sampler see different voltages. I think the explanation for this is that the the sampling process occurs over a short time interval and loads the input significantly just during this time. This means that one should have an identical follower on the output of the main op amp so that the voltage ratio is as accurate as possible. This makes three op amps. So I might as well use the fourth op amp to buffer the input noise signal and use its feedback loop for the frequency gain adjustment. One should give this amp some gain (about two or three) so that the input resistor in the main op amp can be increased, decreasing its gain and so further increasing its accuracy. All this might be a bit more than necessary for looking at pickups, but as you said, one might as well use the whole chip, especially once one sees that there are problems with using just one amp. (I had hoped to eventually have a really low power circuit using batteries that would last a very long time.)

[Joe Gwinn]

Joe,

I put a follower between the voltage input, that is, the top of the capacitor (V2), and the output to the sound card. This does indeed make a difference. Typically a bit more than 1% of the impedance value, but more like 2% near the peak (13 -14 kHz) and decreasing to near zero at about 20KHz. So the resistor in the input path of the op amp and the sampler see different voltages. I think the explanation for this is that the the sampling process occurs over a short time interval and loads the input significantly just during this time. I would worry about transmission-line effects and cable capacitance first, before worrying about sampling offsets. Specifically, the output impedance of the V or I buffer amps needs to match the impedance of the coax line to the soundcard, and the soundcard impedance must match (or be made to match) the coax line as well. At the opamp end, the traditional solution is a series resistor of value equal to the transmission line impedance. At the soundcard end, one may require a T-adapter with terminator on one port.

Also, the lower the drive impedance, the less the effect of capacitance, especially at the higher frequencies.

This means that one should have an identical follower on the output of the main op amp so that the voltage ratio is as accurate as possible. It's sufficient that all ratios are accurately known. I usually measure all components to 1% before assembly into a circuit, and then use the actual measured values in the analysis software.

This makes three op amps. So I might as well use the fourth op amp to buffer the input noise signal and use its feedback loop for the frequency gain adjustment. One should give this amp some gain (about two or three) so that the input resistor in the main op amp can be increased, decreasing its gain and so further increasing its accuracy. All this might be a bit more than necessary for looking at pickups, but as you said, one might as well use the whole chip, especially once one sees that there are problems with using just one amp. (I had hoped to eventually have a really low power circuit using batteries that would last a very long time.)I must say that putting an uncontrolled (from unit to unit) and complex thing like a pickup in the feedback loop bothers me. There has to be a reason all those $15K LCR instruments don't do it that way. They cannot be scrimping on opamps for sure.

But you are getting plausible curves. I would try the circuit out on some components of known properties, like a RC network, to see how accurately these known properties are measured.

One practical disadvantage of the feedback-path approach is there must be a DC path through the component under test, so things like the bifilar-wound pickup that couples through the capacitance between the windings could not be handled. One could bridge the component under test with a 1 Mohm resistor, but this loads the component, leading to a discussion about the significance of this loading.

[Mike Sulzer]

I would worry about transmission-line effects and cable capacitance first, before worrying about sampling offsets.

I do not think that one need worry about transmission line effects when using a cable a few feet long at 20 kHz and below. Cable capacitance, maybe. But 250 pf at 20 KHz is a bit over 3e4 ohms. It does not seem likely that this would bother an amp that can operate into 600 ohms. I will try a very short cable tonight and see if that makes any difference.

[If matching did matter it could be handled at the load end only. A match there kills any reflection. If matching were required at both ends, rf transmitters could not necessarily be made efficient, but they can.]


It's sufficient that all ratios are accurately known....

Again, I am not being clear in my writing. The use of identical followers would be to make the sampling issues identical for the two signals, and so the problem would drop out when the ratio is computed.


I must say that putting an uncontrolled (from unit to unit) and complex thing like a pickup in the feedback loop bothers me. There has to be a reason all those $15K LCR instruments don't do it that way.

The issue with capacitors you mentioned would be one reason. The possibility that someone would come up with an impedance that would make it oscillate is another. I think the high resistance and capacitance of a pickup coil imply that there are no stability issues.

[Mike Sulzer]

That was very interesting!

So was the original cover that came with the pickup made out of brass? I've noticed most of the pickups with a brass cover tend to sound like crap, with high end harshness, and the one with nickel silver covers usually sound better.

Greg

Yes, the original cover is brass, as is the base plate.

[Mike Sulzer]

The results of comparing a short cable (less than one foot) to a longer one (about five feet) from the output of the op amp to the sound card input (V1): There is a small difference (less than 1%) above 14 KHz. If the designed is changed so that both outputs use similar op amps, this effect should tend to cancel out.

[Joe Gwinn]

I do not think that one need worry about transmission line effects when using a cable a few feet long at 20 kHz and below. Cable capacitance, maybe. But 250 pf at 20 KHz is a bit over 3e4 ohms. It does not seem likely that this would bother an amp that can operate into 600 ohms. I will try a very short cable tonight and see if that makes any difference.By transmission-line effects I include the effect of a reactive load on the driver. The operate into 600 ohms part isn't really the issue, and the opamp output impedance will be far lower. As for the 30,000 ohm reactance at 20 KHz, we probably will want more headroom. More below.

[If matching did matter it could be handled at the load end only. A match there kills any reflection. If matching were required at both ends, rf transmitters could not necessarily be made efficient, but they can.] If one will match only one end, it's better that it be the driver end, so the opamp output is isolated from the reactance of the line.

The vast majority of RF transmission lines are matched at both ends. One reason being that matching makes cable capacitance irrelevant.

The issue with [in-line] capacitors you mentioned would be one reason. The possibility that someone would come up with an impedance that would make it oscillate is another. I think the high resistance and capacitance of a pickup coil imply that there are no stability issues.Given that we want to be able to track the pickup through and beyond resonance, all possible kinds of (non-negative) impedance are going to be explored. And not all pickups have high resistance. For instance an alumitone.

[Joe Gwinn]

The results of comparing a short cable (less than one foot) to a longer one (about five feet) from the output of the op amp to the sound card input (V1): There is a small difference (less than 1%) above 14 KHz. If the designed is changed so that both outputs use similar op amps, this effect should tend to cancel out.Both attenuation and phase shift would need to be matched over the entire frequency band of interest, or computations of the complex impedance of the device under test will be affected. Phase is usually the harder one, especially at higher frequencies, especially if the device is complicated (meaning requires many circuit elements to model).

[Joe Gwinn]

Here is my I-V buffer circuit from October 2003. For driving long pieces of coax, I would probably add series 50-ohm resistors to the outputs of A1 and A2. In my case, the load was the inputs of a two-channel scope. I think I set the inputs to 50 ohms, but don't recall. In any event, I would clean that part of the design up, as there were annoying phase shifts at 20 KHz.

[Mike Sulzer]

By transmission-line effects I include the effect of a reactive load on the driver. The operate into 600 ohms part isn't really the issue, and the opamp output impedance will be far lower. As for the 30,000 ohm reactance at 20 KHz, we probably will want more headroom. More below.

Yes, the 600 ohm drive capability has little to do with the output impedance. But the ability to drive power does have an influence on how little the device will be affected by a reactive load of a given magnitude. The reactive part of the load is relatively high here, and probably not a big concern.

If one will match only one end, it's better that it be the driver end, so the opamp output is isolated from the reactance of the line.

I agree, that is, if matching is needed at all.

The vast majority of RF transmission lines are matched at both ends. One reason being that matching makes cable capacitance irrelevant.

If you have control over only one end of an rf system and power is not an issue, you certainly will match it to make sure your equipment snuffs reflections. And if power is a concern, you might want to go to a system using a hybrid. But cable capacitance is not an issue in a system which is end terminated only. As long as the load matches the cable, that is the impedance that the source sees.

Given that we want to be able to track the pickup through and beyond resonance, all possible kinds of (non-negative) impedance are going to be explored. And not all pickups have high resistance. For instance an alumitone.
If you want to look at an alumnatone without its transformer, you will need a circuit designed for very low impedances.

[Mike Sulzer]

Both attenuation and phase shift would need to be matched over the entire frequency band of interest, or computations of the complex impedance of the device under test will be affected. Phase is usually the harder one, especially at higher frequencies, especially if the device is complicated (meaning requires many circuit elements to model).

But small variations in amplitude or phase should not matter if the drivers for the two outputs and the cables are the same. The variations should drop out when the ratio is computed.

[Mike Sulzer]

Here is my I-V buffer circuit from October 2003. For driving long pieces of coax, I would probably add series 50-ohm resistors to the outputs of A1 and A2. In my case, the load was the inputs of a two-channel scope. I think I set the inputs to 50 ohms, but don't recall. In any event, I would clean that part of the design up, as there were annoying phase shifts at 20 KHz.

Good clean design, Joe. I think the 50 ohm resistors are a good idea. This is on my list of things to try. Why would you use an Rs as high as 1 Meg?

[Joe Gwinn]

Good clean design, Joe. I think the 50 ohm resistors are a good idea. This is on my list of things to try. Why would you use an Rs as high as 1 Meg?Thanks. I stole it from Agilent et al.

Use a series resistance equal to the coax impedance, whatever that may be. Audio cable is not necessarily 50 ohms. It may be necessary to measure the cable impedance.

It may be useful to add a voltage follower stage between the output of the transimpedance amplifier A2 and the coax, with the series resistor between this new follower and the coax.

As for Rs, I was playing with it, to see the effect. High values of Rs limited bandwidth due to stray capacitance unless the device impedance was quite low, which then made noise more of a problem. While in theory Rs=0 ohms would be optimum, again it's best to isolate practical voltage sources from the vagaries of the device under test, and also to protect the voltage source against shorts. And to prevent overcurrent into the transimpedance amplifier input. But the exact value of Rs isn't critical, and I ended up with 2.2 Kohms.

[Joe Gwinn]

Yes, the 600 ohm drive capability has little to do with the output impedance. But the ability to drive power does have an influence on how little the device will be affected by a reactive load of a given magnitude. The reactive part of the load is relatively high here, and probably not a big concern.There are two issues here. First, if the load impedance is too low, the amplifier will struggle. Second, assuming that the amplifier isn't struggling, a reactive load can provoke oscillation. It is this second issue I was alluding to. Either capacitance or inductance load on an opamp output can provoke oscillation. A series resistor pretty much abolishes the effect.

If you have control over only one end of an rf system and power is not an issue, you certainly will match it to make sure your equipment snuffs reflections. And if power is a concern, you might want to go to a system using a hybrid. But cable capacitance is not an issue in a system which is end terminated only. As long as the load matches the cable, that is the impedance that the source sees.I guess by "if power is a concern" you mean DC power, as from a battery in a cell phone or the like. Cell phones are physically small and have control of all aspects of the line and what it is connected to, so many tricks are available.

In my day job, I (and a thousand of my closest friends) build megawatt radars. Believe me, everything is matched, both to ensure phase stability, and because at those power levels even a small mismatch can reflect enough power into someplace it doesn't belong to release smoke.

If you want to look at an alumnatone without its transformer, you will need a circuit designed for very low impedances.Actually, the current circuit could do it, but with smaller resistor values. But I would keep the transformer. There are other pickup coils to consider, such as those in active pickups.


The point is to make our I-V Buffer as general as possible, so we only need to build one model. I think if one adds up the costs in time and money, the metal box, connectors, battery hardware, and circuit board are 90% of it, and the circuitry is almost free. The money is always in the boring stuff.

[Mike Sulzer]

There are two issues here. a reactive load can provoke oscillation. It is this second issue I was alluding to.

Yes, that is the issue I was thinking of.

I guess by "if power is a concern" you mean DC power...

No, I meant loss of rf power or efficiency. If you need to worry about power coming back down the transmission line from the antenna in a high power system, you want to divert it to a waster load rather than use a series resistor. You cannot always match the antenna to the line under all conditions, so this is a real concern.

In my day job, I (and a thousand of my closest friends) build megawatt radars. Believe me, everything is matched, both to ensure phase stability, and because at those power levels even a small mismatch can reflect enough power into someplace it doesn't belong to release smoke.

In my day job, I use a megawatt radar to do science. I am well aware of loud noises followed by smoke.


Actually, the current circuit could do it, but with smaller resistor values. But I would keep the transformer. There are other pickup coils to consider, such as those in active pickups.


And lower inductance coils used when you put a coil on each string. In either the classic I-V or the circuit I am using, one could use two or more resistors for different impedance ranges.

[Mike Sulzer]

The impedance plot of a tele bridge type pickup is available here: http://www.naic.edu/~sulzer/tele2.png.

This is a 7K pickup wound with #43 wire using bobbin material and magnets from StewMac. The interesting thing about a tele bridge pickup is the effects of the back plate and bridge. The plot shows three curves of the unloaded pickup (No R or C across the pickup) and three curves with 500pf (to simulate a cable) and 110K (to simulate the loading with volume and tone on 10). The three curves for each set are for 1. no plate or bridge, 2. plate but no bridge, and 3. plate and bridge. The effect with no R and C is large, because the frequency of the resonance is high. (The effect of eddy currents generally increases with frequency.) Even loaded, the effect is quite noticeable.

The unloaded resonant frequency for this pickup is lower than that of the japanese humbucker (posted last week). However, the loaded resonant frequency of the tele is quite a bit higher than that of the humbucker. The humbucker, with two cols in series, has high inductance and low capacitance, and so is affected more by the cable loading than the single coil tele.

[Mike Sulzer]

This post discusses the impedance curves of a strat type pickup. I put 5000 turns of #43 wire on a Guitar Parts USA plastic bobbin. This type of bobbin allows he cores to be removed, so one can compare the magnets to an air core and also use different types of material.

The two plots discussed in this post are here:
http://www.naic.edu/~sulzer/GPUSA2.png
http://www.naic.edu/~sulzer/GPUSA2Lows.png
The first shows the whole frequency range of the measurements; the second shows just the lower part.



Four of the five lines on the plot show both measurements and model results, that is, impedance curves resulting from calculations. (The fifth shows the impedance with the alnico magnets and the pickup loaded with cable and volume and tone.) The model consists of a C, an L, and two resistors, one in series (Rs) with the L, and the other (Rq) in parallel with the pickup. The pickup capacitance and the series resistance are the same for all the core types:
___C____L___Rs___Rq__core type
146e-12 .76 4800 5.5e6 aircore
146e-12 .99 4800 .59e6 alnico
146e-12 1.24 4800 .46e6 slugs 1 per hole
146e-12 1.4 4800 .33e6 slugs 2 per hole

The inductance varies with the core type because the permeability of the material is different. An air core gives the lowest inductance, but also the highest Rq, indicating the lowest loss. (The loss would be from eddy currents flowing in a metal core.) Why is Rq not infinity? Two possible reasons: 1. the pickup capacitance might be somewhat lossy; it is the result of electrostatic coupling between the turns of wire, and there is a significant total resistance. 2. The measuring instrument might not be accurate for high Q/high frequencies, due to the finite high frequency gain of the op amp.

Notice that in addition to matching at the peak, the model for the air core case matches very well at low frequencies. This is expected when there are no eddy currents as in a metal core. The good match between measurement and model indicates that the pickupMeter is working well.

The measurement and model using the alnico magnets, as intended in the design of this pickup, shows significantly higher inductance, due to the permeability of the material, and much higher losses (lower value of Rq). But both the permeability and the conductivity of the alnico are lower than the material used in humbucker slugs. A humbucker slug is both shorter and thinner than than an alnico magnet, so one in each hole is less magnetic material, but two (with one sticking out quite a bit) is probably too much, so a good comparison might be in between.

Since the inductance has been matched at the peak, it is too low at the lower frequencies since the effect eddy currents in decreasing the inductance is more pronounced at high frequencies. On the low frequency plot, the orange dots goes with the yellow curve, and the blue dots go with the green curve.

[David Schwab]

In my day job, I use a megawatt radar to do science. I am well aware of loud noises followed by smoke.

LOL! :D

And lower inductance coils used when you put a coil on each string.

Like one of these suckers:

http://www.sgd-lutherie.com/images/WAL_pickup_inside.jpg

[Mike Sulzer]

LOL! :D



Like one of these suckers:



Exactly! I am working on a six coil pickup system for guitar with separate preamp and distortion for each string in the guitar. Will discuss it on this forum when ready. (It will take some time.)

[David Schwab]

Exactly! I am working on a six coil pickup system for guitar with separate preamp and distortion for each string in the guitar. Will discuss it on this forum when ready. (It will take some time.)

That's a Wal bass pickup. I'm going to be making my own version of it soon. I have all the parts made and just have to wind it.

Hex fuzz is one of the best things ever. I used to have an ARP Avatar guitar synth, and it had that feature. Even with the simplest gnarly sounding fuzz, you get this very clear tone when you play chords due to the lack of intermodulation.

[bbsailor]

Exactly! I am working on a six coil pickup system for guitar with separate preamp and distortion for each string in the guitar. Will discuss it on this forum when ready. (It will take some time.)

Mike,

You will need to use some pretty thin wire to get enough turns on each magnet core to get an output in the 100mv to 200mv range. You can get around this by using AWG 32 to AWG 36 wire connected to a transformer to boost, by a high transformer turns ratio (TR), back to a high impedance. Miniature output transformers (inexpensive) 3.2 ohms or 8 ohms to 20K or 50K would work.

You can also use 1/8" neo magnets to obtain a little more winding area on the individual string bobbins. Used as a bridge pickup, the 1/8" magnets would work where the string movement is less. Use a 3/8" teak plug cutter, available at marine stores to make your own 3/8" plastic or fiber washer that can be directly glued at the ends of each neo magnet to make the individual string bobbins. With a bobbin about .5" tall, you can probably get 4 to 8 ohms of wire on each coil. Burns, in his pickup patent, only uses 1 ohm DC per individual coil but needs a transformer with a 1:50 TR. With the 1/8" magnet core, you will be in the range of 200 to 400 turns and would need a TR of around 1:20 or higher to get the output in the typical pickup range. Using a 3/16" magnet only gives you 3/32" of space to wind over the magnet core. If you angle the pickup or alternate the strings in two rows (coil 0.75" OD each string to fit inside a humbucker cover), you can obtain a some more space to fit the coils between the strings.

There are many sources of commercially made solenoid coils larger than 3/8" diameter. The pickings are slim to non existant looking for a commercially made stock solenoid 3/8" OD with an ID of 0.125".

Hint: Go to Wal-Mart and pick up 2 four packs of plastic Singer Class 15 sewing bobbins PN/30023. They can use .25" magnets X .5" tall and 6 will fit inside a humbucking cover in two rows. You should get about 6000 turns of AWG 42 on each one. The wire winding area is .0375" high by 0.125" wide.

Keep us posted about your progress.

Joseph Rogowski

[Mike Sulzer]

Joseph,

Thank you for the suggestions. The pickup is done, but I am still working on the electronics. This pickupMeter is a lot of fun, so I have been working on it instead.

The design has gone in the opposite direction. Each core is two .4 " long ferrite beads end to end. The diameter is similar to the usual alnico magnet, and the relative permeability of the material is 5000. A small neo mag disk (1/8" dia 1/32" thick) is on top of each pole. That is all you need with the ferrite behind it. 4000 turns #43; could have fit more on the ~.6" long spools. Output is about 300 mv p-to-p, which gives very good SNR using a quiet bi-polar amp (LM837). Each coil is loaded with a resistor and capacitor to bring down the resonant frequency and Q to reasonable values. A seventh coil serves as a hum field sensor; it is preamped and subtracted from the others. There is about 20 db isolation between string outputs from the coils, which is increased to about 30 db with a resistor summing network after the preamps.

I will start a discussion on this sometime after after Thanksgiving (have a lot of work that week). Thanks for your interest; I look forward to discussing it with you and all others interested.

Mike

[David Schwab]

Mike, that sounds really interesting. I'm looking forward to reading about it.

Here's the parts my prototype Wal clone. They aren't assembled, just sitting against the keeper bars and magnets. I'm going to wind them next week.

It's not going to have a poly-phonic output, but it could.

[Mike Sulzer]

Here is the schematic of the version of this pickup meter, in use for a couple of weeks: http://www.naic.edu/~sulzer/pumSchem.png.

The noise generator feeds a low pass filter; this improves the accuracy of the device at low frequencies where the main amplifier has low gain due to the low impedance of a pickup at low frequencies. When the follower on the main amp is omitted, the signal level at a peak above 12KHz drops by about 1 or 2%. So there is some loading effect, probably due to cable capacitance. The follower on the low pass amp assures that the output impedances of the two channels are very nearly equal.

Attempts last weekend to make an I-V design, using a high pass filter on the noise source to help overcome leakage in the sound card, failed. It is still not good enough. Even with the high pass filter, the fall off of a peak above 12KHz is obscured by the leakage.

It is important to measure the accuracy of the device, more on that in the next week or so.

[Joe Gwinn]

Here is the schematic of the version of this pickup meter, in use for a couple of weeks: http://www.naic.edu/~sulzer/pumSchem.png.

The noise generator feeds a low pass filter; this improves the accuracy of the device at low frequencies where the main amplifier has low gain due to the low impedance of a pickup at low frequencies. When the follower on the main amp is omitted, the signal level at a peak above 12KHz drops by about 1 or 2%. So there is some loading effect, probably due to cable capacitance. The follower on the low pass amp assures that the output impedances of the two channels are very nearly equal.If cable capacitance is the problem, one solution is to have matched impedances at both ends of the cable. While this cuts the the available voltage in half, it will flatten things out, and so long as noise sources are shielded out, amplification will save the day. The buffer amplifiers can easily be arranged to provide some gain.

Attempts last weekend to make an I-V design, using a high pass filter on the noise source to help overcome leakage in the sound card, failed. It is still not good enough. Even with the high pass filter, the fall off of a peak above 12KHz is obscured by the leakage.Leakage from where to where? And of what magnitude? If from output to input, one solution is dual sound cards. But most cards have very good isolation between sections, because the ear is very good at detecting such things. And because the marketing dept needs something to talk about.

What make and model of sound card are you using? I'd like to read up on it.


More generally, it will be widely useful if we can identify which sound cards are suitable, just as we did with LCR meters. Actually, what will be really useful is to identify makes and models that are not suitable, and why they fail to be suitable.

[Mike Sulzer]

Leakage from where to where? And of what magnitude? .

Joe, thanks for keeping me honest here. The problem was that on my test card the film bypass caps were not connected. (The electrolytics were.) Since I have been using the same amps for both circuits and for checking leakage, everything was affected. The i-v works much better now. I will try to get some valid comparisons tonight.

It looks as though the i-v does require a resistor in series with the signal source as you suggested for certain cases. This is for when one connects a cap across the pickup to act as the cable cap.

[Mike Sulzer]

Here is the IV circuit (http://www.naic.edu/~sulzer/ivSchem.png) now in use as a pickupMeter. Why 100K in series with the noise generator? A resistor is needed to limit the gain of the iv amp when the impedance under test has a very low value, such as a capacitance at high frequencies. But selecting the right value for this resistor can also help alleviate two problems: 1. A wide variation of the level of the output of the iv amp; 2. the need to adjust the level of the noise source with different test impedances.

The pickupMeter will be used with impedances that typically can vary from somewhat less than 10K to about 1M. With 100K in the feedback loop of the iv amp, its gain varies from .1 to 10, that is, 100 to 1. We can use the resistor in series with the noise source to reduce the variation at the amp output, transferring some of it to the buffer amp that drives the line to channel 2 with voltage across to the test impedance.

This plot (http://www.naic.edu/~sulzer/iv01rs.png) illustrates a test of the performance of the circuit using three different values of resistors. The values plotted are those measured by the pickupMeter as a function of frequency divided by the dc value measured by my DMM. The 100K resistor in the iv circuit was measured with the same meter, and it is this reference that is used in the software. (This is not a good absolute reference, but is adequate for testing.) Ideally all the values on the plot should be one. Note that the scale goes from .995 to 1.003; so the errors are small.

Obtaining good performance over the whole frequency range requires careful wiring to keep the noise source away from the summing junction of the iv amp and a single point ground system. It also requires good power supply bypassing (10 microf electrolytic, .047 microf polypropylene, and .047 microf ceramic in parallel from each power supply terminal to the ground). The problems at low frequencies with the 1.005M resistor are the result of hum pickup. This would not be a problem with a normal pickup that has low impedance at low frequencies.

This plot (http://www.naic.edu/~sulzer/iv01c1000pf.png) shows the measurement of a nominal 1000pf capacitor. The measured impedance is converted to a value of capacitance versus frequency. Hum affects the measurement at low frequencies, but I think there is another source of inaccuracy as well. The measured value is consistent at high frequencies but increases at low frequencies. It is up 1% of the high frequency value at somewhat under 2 KHz.

Finally, here (http://www.naic.edu/~sulzer/iv01jpnhb.png) is a plot of the impedance of the same japanese humbucker shown in an earlier earlier (with the cover removed). The peak has a somewhat higher impedance, showing the higher accuracy of this circuit at high frequencies and high impedance.