New Wilmott box, o'scope Bode, Schaller humbucker

[nuke]

Ok, a little reality check. I just got my Ken Willmott integrator board (pre-built) and coil all put together into a box today. Did a quick checkout and calibration and fired up my Rigol MSO5000 scope with built in Bode plotting function. 

Chime in on whether the results appear plausible. 

I think I'm getting reasonable operation. The scope has a built-in Bode plot function that uses two scope channels and one of the scope's built-in signal generators.  The plot includes phase information as well as amplitude. 

Phase in this case is signal input to the exciter coil, compared to signal output of the integrator. The orientation of the coil matters. In one direction the "0" and asymptote are as in the picture. Flip the coil and the phase 0 and asymptote flip places. But it is interesting to note the phase lag between the exciter and the output with respect to frequency. Not certain what value it might have, just interesting and the waveforms do reflect the phase shift. 

The test subject is a 1980's era Schaller humbucker (labeled "bridge" and bridge spaced poles). 

Params at measured at 100hz on the LCR meter at 70F ambient:

Ls 4.652 H  
Q 0.356
Lp 41.36 H
Cs 542.9 nf
Cp 61.45 nf
Rs 8.206 k
Rp 9.251 k
DCR 8.160

The Rigol plots both phase and amplitude. Green is phase, magenta amplitude. 

First plot is "unloaded" on the Integrator box, showing peak at 10.47Khz:



Second plot in the "load" switch setting on the Integrator, showing peak at 3.019Khz:


 



Shot of the bench top setup of the pickup, exciter coil, the integrator and the scope. 

[aquin43]

Measuring an ordinary pickup, the phase lead at 100Hz has to be an artifact of the measurement setup. The integrator is probably a single pole low pass filter with a low cutoff frequency and the gain will be limited at dc for stability. There may also be a roll off in the feed to the scope if you are using ac coupling. It is easy to make a flat amplitude response but much harder to make a flat phase response.

You could calibrate out the phase error by measuring an air cored coil of as few turns as will give a reasonable reading instead of a pickup and using that as a baseline.

[stratotarts]

The integrator can not reliably measure phase. It can only reliably measure amplitude which is what it is designed for. But it's not mainly because of any part of the implementation of the integrator circuit. It's because the test coil to pickup coupling is not controlled for phase. You found that out when you moved the arrangement and got different readings. There is some amplitude and phase shift at the high and low ends of the integrator response but this has been overcompensated as far as possible. The low end response was set at 18Hz, well below the required test range 100Hz. At that frequency the circuit phase shift is more noticeable than the amplitude variation which is a fraction of a dB. At the high end, the phase accuracy is limited by the bandwidth of the op amp. The JFET amp in use is quite fast, so errors should be very minimal below 20kHz.

If you want to seriously consider phase, you must perform two terminal network analysis, directly at the leads of the pickup. Because that kind of measurement gives you an isolated coupling to the pickup, the actual phase/amplitude values can be processed further mathematically and yield valid results. I once modified an integrator board to do that, but as usual got sidetracked with other work. Member here "ms" and others use this kind of approach and can generate Bode plots from it. Those kinds of plot follow closely what you would measure with a test coil and integrator, except that the cover losses will differ, since the geometry of the pickup coil, and the pickup coil test coil combination are different.

In a reactive circuit, phase and amplitude are intimately related, but it does not mean that you necessarily need to measure both, it definitely depends on the actual behaviour that you are trying to capture. The guitar output chain and human ear (in the limited context of a mono instrument played through an amp) are mostly insensitive to phase, the phase affects the amplitude along the way but it is the amplitude that is relevant at the output. Again, if you do use phase measurements, they have to take into account the phase behaviour of all your test equipment as well as the device under test.

It isn't the case that phase in the guitar circuit is not relevant. It could factor into calculations about the circuit. But you can not learn much from putting a "phase stethoscope" on the resulting signal and hope to learn what is going on inside the "body".

[aquin43]

Looking at the integrator schematic, I see four low frequency roll offs. The ones at the input and output are probably low enough not to count but the integrator itself has two single poles at 8.84Hz, each of which will contribute 5 degrees of phase lead at 100Hz. The gain error, though, will be under 1% (< 0.1dB).

[nuke]

Yeah, I agree. The integrator itself introduces phase shift. There's no filtering without phase shift of course.

I just wondered if the integrator's effect on phase were subtracted out, would the phase change perhaps reveal something about the pickup's LCR properties. Phase was mostly just a curiosity, since the Rigol's Bode function does it and you can't turn it off. (and many times, you'd want it for other work). I hadn't sat down and calculated the poles of the filters, just sort of eyeballed the circuit and it was reasonable. Also, faster and simpler to buy it from Ken than build it myself. 

Get back to the amplitude response, do my results appear reasonable compared with others and the raw LCR numbers? I'd like to be on the same measurement base as other folks. I'm using the scope instead of software/audio interface/PC combo. At first blush, I think I'm in the ballpark. 

[stratotarts]

From a quick look, I can say that the amplitude plots you posted look okay. However, the resolution and division settings you chose for the scope display, are not very fine. So if you want more certainty about the measurements, you should increase the resolution and adjust the plot division/scale to show more of the plot, and show it in more detail.

To re-iterate, I know that the integrator has big phase shifts in the subsonic range. It's not practical to build a simple op amp based integrator without a low end cutoff because the DC offsets in the feedback loop are subject to an extremely high loop gain in that case, which produces unacceptable DC offsets at the op amp input. The current design therefore uses capacitors to separate the AC from the DC gain, which is made equal to one for the absolute minimum input offset. Both the series input RC and the parallel feedback RC network are matched in frequency, to the low frequency cutoff value of about 18Hz that I mentioned.

The component values to achieve the 18Hz cutoff are near the limit of what you could use and not introduce biasing issues with the op amp. It might be possible to extend that lower by using adjustable input offsets or some other modifications to the circuit, but first I would like to know the reason. Because, the phase by itself doesn't provide much information besides confirming general electronic theory (it shows a typical phase shift for this kind of circuit). More than that, would require some sort of wide band calibration with some "perfect" pickup in order to null out the phase effects of the test coil, generator and integrator as a unified signal path.

Amplitude testing isn't required below even 100Hz because the outcome is always so predictable due to the diminution of eddy currents and capacitive reactance. So, I used that as a design target and settled for a minor (~0.2dB) shelf at 100Hz. I expect that users will be at the same time aware of it, and also of its unimportance to the measurements.

If there is a specific reason for making special phase testing while also using an excitation coil, then I might be interested in thinking about, and presenting solutions. Otherwise, I am sorry but it seems like a waste of time. The integrator is not intended for phase measurements, and I never claimed that it is. Having said that, there is also the aspect (or question?) that assuming that you did want to capture the phase, why it would be important at the lower frequencies (just as, the reason for measuring amplitude at lower frequencies would need an explanation or experimental justification)?

If you are really, really interested in that kind of phase response and you want to measure it with the integrator, you could restrict the frequency range to something reasonable and perform a wide band calibration with a nearly perfect, almost theoretically correct DUT such as an air core inductor. It's possible, it might be interesting, but as you can guess, I'm skeptical of the value unless the underlying theory or idea is explained to me. Since you just pose it as a question, "would it reveal anything?", well the way I would treat that myself is to look at the existing plots in the light of that question.

However, in general, phase has no "stand alone" relevance in the context of the pickup circuit, outside of the relationship with impedance. That can be much more easily and reliably measured using the two terminal tests that I mentioned.

[nuke]

Although I've owned this scope for some time and use it plenty, I've never used the Bode function before. 

It appears there's no means to adjust the scale. I've hunted for menus or settings and there aren't any. Nor is there anyway to turn off the "phase" graph. Not much of an explanation of how Rigol implemented the function. Definitely not a piece of vintage HP test gear with great manuals. 

At least, I can export the Bode values as a table and import them into Excel or what not. 

[stratotarts]

Have you tried turning off the phase plot? Edit - oh, you did. NM. What happens when you Change the Amp/Freq settings under the Param Set menu as outlined on page 19-3 of the manual?

[nuke]

The param set menu page just programs the waveform generator with the steps to sweep. Frequency/points-per-decade/amplitude and there's some other options for piece-wise amplitude. But it doesn't affect the scale of the plot in anyway. 

I need to goof around with it some more and figure out how it works. It doesn't really say much about how it generates the plot itself, nor can I find any means to override it. Seemingly, it sets the scale at the first step that it measures and that's it for the rest of the plot. 

[stratotarts]

There are screen shot images in that section of the manual, that show considerably less range than the +60 to -60dB range in your plot. So there must be some way.

[nuke]

The manual is just not in sync with the firmware. I have later firmware on the scope. Looks like there's a new update on the Rigol site, but the release notes file is corrupted... So gonna hold off on that one. 

The manual says the range doubles when switching the source generator impedance from 50-ohm to high-z, but that does absolutely nothing on the firmware I have on the scope. It also says some other things in other functions that are not exactly correct, so Rigol haven't updated the manual content to match the firmware revisions. 

In any case, I did work with it a bit and figured out what it actually does. The program sweeps the generator in steps, based on the parameters setup in the table, voltage, start and stop frequency, steps per decade in log sweep mode. It uses two channels of the scope to generate the plot, "in" and "out".  While it steps through the generator to take measurements, it autoscales the scope channels and sweep. It simply subtracts the voltage readings of channel 2 and channel 1, applies the db formula and plots it. The phase comparison is also based on channel to channel math that's done internally.  

I'm using a scope probe for channel 1 at the terminals of the exciter coil where the generator is attached.  Channel 2 is a straight BNC to the output of the integrator. 

Did a little experimenting with different settings to figure out what it is doing, and it is performing a proper Bode plot based on the difference between the two scope channels. It just doesn't offer any means to scale the display. It's just fixed in the display window, kind of a dumb thing, but it is what it is. I played with the scope probe at 1x and 10x at the generator/coil connection, as well as the gain setting on the box, and it is generating correct and reasonable numbers.  The shape of the curve remains the same, it just offsets linearly up or down, so I'm happy.  Using Excel, it is pretty easy to normalize the plot to base line it on the 0db point for comparison purposes. I think somewhere around 250hz might be a good spot to zero the offset. 

It is easy to export the data in CSV format from the scope and move it to my Macbook. Excel is happy to make scatter plots out from the .csv files and bam, bob's your uncle. 

I just collected data on a couple of interesting pickups that aren't in Antigua's database. I set up 100 points per decade on the sweep, and got some nice smooth plots. 

Would like to get my methodology relatively in line with what the rest of folks are doing so we can make meaningful comparisons. 

[nuke]

Here's a quick and dirty excel graph of Gibson P100 Bridge pickup, showing unloaded and loaded curves.



Measured data:

Ls @ 100hz: 5.37H
Cp @ 100khz: 226.4pf
Q @ 1khz: 2.55
DCR: 9.41Khz

Plot of loaded and unloaded curves:






Photo of pickup:

[stevewf]

Mar 24, 2024 18:38:54 GMT -5 nuke said:

Here's a quick and dirty excel graph of Gibson P100 Bridge pickup, showing unloaded and loaded curves.

Oooh! Congrats on getting the rig running.

I'd really love to compare that with the profile of a Gibson P90 Bridge pickup!

[nuke]

Graph with the data normalized closer to the 0db line by subtracting 18 from everything.