timtam
Meter Reader 1st Class
Posts: 61
Likes: 30
|
Post by timtam on Aug 17, 2022 22:21:03 GMT -5
The German guitar research organization GITEC has released the free software "Pickup Wizard" for measuring a pickup's impedance frequency response curve, written by Helmut Keller. The current release is version 1.0 (an earlier beta version had been released several months ago). Full instructions are included in the PDF. Edit July 2023 - New link: The software is written in Matlab script, which means it will run in Matlab or in the free Octave ( octave.org/ ). You don't need to know Matlab code to run it. It has a GUI interface. It would be nice if a future version is also supplied as compiled Matlab code, which would allow anyone to run it with the free Matlab Runtime engine (somewhat simpler than installing Matlab or Octave). The program uses your computer's sound card to generate the test signals and measure the response. The requirements are described as follows: "A sound card (or audio interface) with at least one analog stereo output and one analog stereo input with a nominal input impedance of at least 1 MΩ needs to be connected to the computer before starting Pickup Wizard. The sound card has to support a sample rate of 48 kHz, and a bit depth of 24 bits. Make sure that your operating system does not apply any signal processing to the sound card signals. Also make sure that any direct monitoring and signal processing of your sound card (or interface) is disabled." (p10). The software includes a calibration process for your card's outputs/inputs. The software connects to a pickup via an extremely simple small circuit - the 'measurement adaptor' - that should be easily built (components are one 1 MΩ/0.1% metal-film resistor, up to 4 TS sockets, a 3-way rotary switch, two speaker terminals, and a small enclosure). The PDF also states "If a reduced measurement accuracy is acceptable, and if the pickup is mounted on a passive electric guitar, the pickup may stay in the unmodified guitar during the impedance measurement." (p4) The engineering basis of the software's calculations is described in detail in the PDF. Other than verifying that the software indeed loads in my copy of Matlab (Windows), I have not had a chance to try it out, and may not be able to anytime soon. Maybe someone here wants to give it a go and report back. Most development and testing by GITEC has been in Windows (Matlab and Octave). Some testing under Linux and Mac has been done and more is underway (Matlab and Octave run under all three OS's). |
|
|
|
Post by antigua on Aug 18, 2022 13:37:25 GMT -5
This is very interesting. I'll have to give it a try.
One thing about interfacing with an 1/8 audio jack is that I suspect some people would be tempted to cannibalize the jack and cable from a throw-away pair of earbuds to do this, but those cables will tend to have a high capacitance, so it might have to be noted somewhere not to simply do that.
|
|
|
|
Post by stratotarts on Aug 21, 2022 16:26:30 GMT -5
I downloaded it and had a read of the manual. I haven't tried it out yet. The main thing I'm wondering, is it an "integrator killer"? I was getting ready to do a major design re-make of the 5.9 integrator. This worries me. I don't have any serious "irons in the fire" apart from that, but it seems pointless to invest major time and money into a redesign if a device like this becomes the instrument of choice for measurements. So I'm thinking, if it really works, I might have to walk away from supplying integrators entirely. The whole point of the redesign was to reduce the assembly labour, which I now see was grossly under profitable. But if the demand is going to fall off, I can't really assume that risk. I guess it was just a matter of time.
|
|
|
|
Post by antigua on Aug 21, 2022 19:00:30 GMT -5
I downloaded it and had a read of the manual. I haven't tried it out yet. The main thing I'm wondering, is it an "integrator killer"? I was getting ready to do a major design re-make of the 5.9 integrator. This worries me. I don't have any serious "irons in the fire" apart from that, but it seems pointless to invest major time and money into a redesign if a device like this becomes the instrument of choice for measurements. So I'm thinking, if it really works, I might have to walk away from supplying integrators entirely. The whole point of the redesign was to reduce the assembly labour, which I now see was grossly under profitable. But if the demand is going to fall off, I can't really assume that risk. I guess it was just a matter of time. I don't see the integrator and this software as being the same sort of thing, because the integrator is hardware and works with any hardware based setup, where as the Gitec is software that exploits a soundcard, soundcards which aren't designed for this purpose, and I think it makes some modelling assumptions, but it being summer break I haven't had time to get into it yet, so there is a lot about it that I don't understand yet. I've tried Matlab experiments in the past, and I continue to be much more comfortable with hardware, because there tends to be less dependence on any one link in the chain, so the setup can be adapted to various testing situations, and that includes preferring the Velleman dedicated hardware to the Rightmark software and an audio card. Hardware that it's built for the purpose just always seems to be more reliable than hardware that is pulling double duty. The Velleman USB bode plotter also makes it easy to measure the amplitudes and export JPGs of the plots, multiple plots overlapping, with labels, and I've seen first hand that a couple other pickup testers have exploited those conveniences in their pickup tests. Part of the reason I didn't pursue Matlab bode plotting was because it would have been difficult to get all of the benefits that the Velleman setup provides out of the box. The integrator is really critical to the this setup though, if you were to ever stop offering them, something software based might be the only way to go. |
|
|
|
Post by stratotarts on Aug 21, 2022 20:14:30 GMT -5
Yes, they aren't exactly the same "animal", the Wizard and the Integrator. I plan to build one so I have both. Member "ms" did some similar work with a simple resistor network circuit and posted it here quite a while ago. I haven't made any final decisions about it. I have become short of time lately, and the redesign is a major one, although the circuit itself won't change very much. So it seems daunting, but as the year goes on I might feel more optimistic about it. Could also just be some post Covid blues...
|
|
|
|
Post by ms on Aug 22, 2022 7:09:20 GMT -5
The introduction to the write up of the pickup wizard sounds like a description of my device, but there are some important differences in the measurement geometry and thus the required passive network. The PW uses a large value resistor connected from the test signal to the pickup; the other end of the pickup goes to ground. The voltage across this resistor is used to measure current. Measuring this voltage requires differencing the voltages measured by the two channels of the recording interface. This tends to increase the relative size of errors. Also the high value of the resistor needed to keep the difference large at all frequencies means that full accuracy requires a correction.
I prefer a geometry using a small value resistor with one end connected to ground. (The measurement is that of the pickup plus the value of this resistor, which is accurately known and can be subtracted from the measured impedance) One channel of the RI measures this voltage. High frequencies are not an issue (2K versus 1M). Also, the measured impedance is proportional the ratio of the two measurements, further reducing high frequency errors, limited by differences in the high frequency responses of the two channels of the RI rather than the cutoff frequencies themselves.
Another difference is the PW sweeps a sine wave, but my device avoids sweeping by using a waveform containing all frequencies.
|
|
|
|
Post by aquin43 on Aug 22, 2022 10:06:53 GMT -5
The introduction to the write up of the pickup wizard sounds like a description of my device, but there are some important differences in the measurement geometry and thus the required passive network. ...I prefer a geometry using a small value resistor with one end connected to ground. (The measurement is that of the pickup plus the value of this resistor, which is accurately known and can be subtracted from the measured impedance) One channel of the RI measures this voltage. High frequencies are not an issue (2K versus 1M). Also, the measured impedance is proportional the ratio of the two measurements, further reducing high frequency errors, limited by differences in the high frequency responses of the two channels of the RI rather than the cutoff frequencies themselves. Another difference is the PW sweeps a sine wave, but my device avoids sweeping by using a waveform containing all frequencies. I suspect that the PW arrangement was adopted because it allows the measurement of pickups with one end grounded. Your method, which I have now adopted myself, is more like the conventional impedance measurement methods, is more robust and measures the pickup open circuit with well under 1pF effective load.
The section of the PW program that models the lossy inductor is worthy of note and could be applied to the impedance measured by either method.
The frequency response determination from the model which I have referred to in another thread seems to me to be of limited validity and ignores the fact that the magnetic signal from the string may follow a path that is not fully reflected in the pickup's impedance. There will always be a need for a Bode plot to complete the model.
PS: The PW software will run under Octave on Linux with no problems.
|
|
|
|
Post by aquin43 on Aug 28, 2022 15:00:21 GMT -5
Having no need for another pickup impedance measuring system, since I already have one similar to that described by ms, I was still intrigued by the modelling part of the program. I have extracted the modelling section into a stand alone program that can read the real/imaginary data files from my test set and determine the component values for approximating the lossy inductor as well as the pickup capacitance and dc resistance. I have found that this way of modelling the coil works extremely well for most pickups but that it can fail badly for some mini-humbucker types that have close coupling between the coils. The model is obviously not really valid for such a pickup but, even so, quite a good match is possible by reducing the frequency range sufficiently. Fortunately for most of the pickups of this type that I have tried a match is possible up to around 10kHz.
There also seem to be many different combinations of L and R values that can match a given pickup within a fraction of a dB, depending on the starting parameters.
|
|
|
|
Post by aquin43 on Sept 4, 2022 3:43:28 GMT -5
I downloaded it and had a read of the manual. I haven't tried it out yet. The main thing I'm wondering, is it an "integrator killer"? I was getting ready to do a major design re-make of the 5.9 integrator. This worries me. I don't have any serious "irons in the fire" apart from that, but it seems pointless to invest major time and money into a redesign if a device like this becomes the instrument of choice for measurements. So I'm thinking, if it really works, I might have to walk away from supplying integrators entirely. The whole point of the redesign was to reduce the assembly labour, which I now see was grossly under profitable. But if the demand is going to fall off, I can't really assume that risk. I guess it was just a matter of time. I have found that the frequency response calculated from the impedance as in the "Pickup Wizard" is only part of the story and can differ by several dB from the response measured with the exciter. So, if the frequency response is what you want, the exciter with either your integrator or digital processing is the way to go.
|
|
|
|
Post by stratotarts on Sept 14, 2022 21:19:29 GMT -5
I could support the GITEC software by adding the specified hardware to the new integrator. As there was some discussion of different approaches, is there a way that the impedance measurement circuit could be jumper selectable between the GITEC and the circuit mentioned by member 'ms' on Aug. 22 above? I will have to review the circuits. In that case, what I have in mind is generally a panel with separate connectors for integrator and impedance measurement.
|
|
|
|
Post by ms on Sept 15, 2022 6:33:55 GMT -5
I could support the GITEC software by adding the specified hardware to the new integrator. As there was some discussion of different approaches, is there a way that the impedance measurement circuit could be jumper selectable between the GITEC and the circuit mentioned by member 'ms' on Aug. 22 above? I will have to review the circuits. In that case, what I have in mind is generally a panel with separate connectors for integrator and impedance measurement. What you describe in your last sentence is similar to what I have, but I do not use an integrator. A few years ago I used a very high impedance source follower, but I abandoned that. For my purposes, once you have the impedance, the additional information added by an exciter coil is not obscured by the input capacitance of the sampler, which can lower the resonant frequency significantly, but does not obscure the reduction of signal level by metal in the pickup in the frequency range of highest interest. Of course, not everyone would agree with me on this. I do not think it would be hard to make hardware that supports either of the two impedance set ups. You need different software as well; aquin43 has the additional required software for Octavia, I believe. I think that the combination of the hardware and software changes for implementing both methods would make this an unnecessarily complicated project. The method that I suggest makes better use of capabilities of the recording interface, and the software to achieve a simple but good impedance measurement is simpler than the GITEC software (which implements a frequency response correction, according to the documentation), also. |
|
