SD SH-1N ‘59 model quick test

[stratotarts]

Device Circuit under Test : SD SH-1N ‘59 model
Test Date : Nov 27, 2016
DC resistance (Ohms) : 7,390
Self Resonant Frequency (Hz) : 6420
Self Resonant Peak (dB) : 6.4
Loaded Resonant Frequency (Hz) : 2980
Loaded Peak (dB) : 2.4
Inductance Test Resonant Frequency (Hz) : 1152
Inductance (H) : 4.44
Calculated Intrinsic Capacitance (pF) : 128
Loaded Parallel Q Calculated from Peak : 1.32
Loaded Peak Loss (dB) : 2.8

Gauss: slugs 450 / screws 380

Attachments:

[antigua]

I measured about 75pF for similar braided lead wire guitarnuts2.proboards.com/thread/7725/capacitive-coupling-various-guitar-parts

I've been foregoing calculated capacitance form my humbucker analysis, since without disconnecting the lead wire, it would represent a large portion of the C, and these different humbuckers are offered with different type of lead wire, and guitarists will often cut them shorter when they install the pickup. Add to that the fact that the loaded peak in dB, between 1.5 to 2.5dB, is so close to a hard knee, even in the case of a nickel silver cover being present, that I think 95% of the value is in the loaded peak frequency to see how clear a humbucker will be, and the inductance, to see how loud it will be for a given distance from the strings. It seems kind of trite to summarize PAF clones by those two vectors alone, so I'm keeping an open mind with regard to other factors that might set them apart.

Can you test the inductance at a lower frequency? I actually asked Manfred Zollner his opinion about the 120Jz vs 1kHz issue, and he confirmed that lower is best, ideally 125Hz in order to avoid AC mains 50Hz/60Hz and their harmonics. My Extech LCR meter only does 120Hz or 1kHz, unfortunately.

I can verify that at 1kHz, the inductance does read lower than it really is, for if I test with a Strat pickup, the value barely moves at all. Test with a PAF replica, and the measured inductance drops with frequency. I just got done testing my collection of TV Jones pickups, and it looks like all their measurements are at 1kHz, thus under-reporting the actual inductance. The values I recorded were at 120Hz. 

[JohnH]

I think that the inductances measured by a meter, or derived by idealisation as a simple LC circuit, are only 'accurate' for the purer forms with sharp peaks such as uncovered alnico singles. The humbuckers, and those with covers of more metal in them cant follow a simple model. Hence in those cases an inductance measurement, although still useful, is not so much of a fundamental property but is a measurement of whatever the meter is really measuring in what is truely a more complex case. Id like to know exactly how the inductance meters work in terms of an equivalent circuit, and model that in SPICE to get another bench mark for my equivalent pickup models.

Stratotarts, I recall you were estimating inductances from peak frequencies using a large swamping cap? Thats something that can also be modelled and compared based on knowing the cap and frequency. And given a well defined unloaded peak, and an unloaded with swwamping cap peak, we can derive not only the inductance (at least approx equivalent) but also the self capacitance. The ratio of (cap + swamping cap)/cap = square of peak frequency ratio. Once the self capacitance is worked out, you can get equivalent inductances at the two peaks, which may be different, or check performance of a more complex model with two inductances.

Ps just an aside on attaching images at GN2. If you link an image stored on an external host, all can see it. But attached files and images can only be seen by GN2 members who are logged in. Only really an issue if you want to reference your GN2 posts on othsr sites.

[stratotarts]

The V5 integrator circuit does not perform well when attempting to measure low frequency peaks such as are produced by using a large value swamping capacitor. Here is the result with a 68nF test cap:



In order to avoid clipping, the cal value for the 68nF sweeps had to be lowered. In spite of that, none of the integrated sweeps were accurate. The measured loaded resonant frequency with the integrator is 170Hz and with that capacitor, the calculated inductance is 11H. The other plot with a more modest capacitance of 4.7nF indicates 4.44H. I am not sure of the reason, but my guess is that the dynamic range is always exceeded.

I bypassed the integrator and took a reading (yellow plot) which yields a more believable frequency of 275Hz. That produces a result of 4.19H for the inductance (4.93H with 68nF, for those calculations I used the 81nF that my meter showed when measuring the test capacitor). The test capacitor in the integrator is measured with the same DMM. So I believe it's safe to conclude that the inductance didn't really change very much between 1150Hz and 275Hz.

It has been established via successful modeling that inductance does vary with frequency. It would take more refined testing to establish the exact variation. A single-value inductance specification would be most useful when measured at the same frequency where it is applied. In most cases, that is the loaded resonant frequency. 1000Hz is closer to that than 125Hz. So I would be more inclined to take the a reading close to the 1000Hz value as the "real" or working value.

Here you can see by comparison of integrated and non-integrated plots that at 1150Hz at least, the integrator is accurate:

[antigua]

If the intrinsic capacitance can be solved for from inductance and peak resonance, which inductance at which frequency would you plug into the equation to get the true intrinsic capacitance?

On the point about measuring L at the loaded peak, at least in this plot..


 
it appears that the reactance dips around 1kHz, and the dB at 120Hz is somewhere between what it is at 1kHz or the loaded resonance. 

[stratotarts]

The V5 integrator is not perfectly flat at 100Hz, so the "droop" there is an integrator artifact, not a property of the pickup. Until now, I didn't think much about it because sub 1dB measurements in that frequency range aren't usually important.

The larger the swamping capacitor, the more accurate the inductance measurement is, because the intrinsic capacitance will be proportionally smaller. But it will be measured at a correspondingly lower frequency. That frequency will be much lower than any frequency at which the results will be used to predict important behaviour. The outcomes of measurements and calculation should determine the choice of target measurement values. For example, when attempting to predict the loaded resonant frequency and Q, given certain external components, the L and C that you want to use is the value of L and C at that target frequency. I think we all believe that the C does not vary much. The L does vary, so the best predictor of L is a value of L that was measured at the closest frequency to the target frequency.

I've been aware that measuring L at different frequencies has been an outstanding task, and this problem makes it more urgent. I think that the impossibility of measuring L and C independently, dictates either assuming a relatively constant L with frequency, or using successive approximation. It would be more reassuring to know that at least one parameter can be measured accurately if other parameters must be computed using its value.

[JohnH]

I reckon to use the swamping cap technique, use a cap that gives you a peak at around 1kHz, which you showed can work consistently. Add no resistor load. Most pickups are capable of getting a sharp peak in this condition, and also give a decent peak in the usual fully unloaded test. That would be all that is needed to estimate the self capacitance and the inductance, if you assume inductance was constant.

We know that reactance varies with frequency in a way that is no consistent with a single fixed cap and inductor value. You can think of this as a variable inductance, or as I do, think of it as two fixed inductive branches with different resistive properties. I tink that is a more useful model because we can indeed model it.

The variations can lead to a small difference in the self capacitance calc, as seen in 6-part models compared to those directly calculated directly from the tests. But ii is a small part of a not too significant number.

Example to follow.

[antigua]

I don't see much change in inductance for Strat pickups between 120Hz and 1kHz, do you expect that it would be different at the loaded peaks in the 3 to 4 kHz range? Given the more reliable readings from Strat pickups, I feel more emboldened to calculate for C there. 

With the humbuckers, the foot print seems to be much more fixed. They essentially have the same cores, a row of slugs, a row of screws, and tiny bobbins with coils that have always been machine wound. So you have resistance, and that tells us how much wire the wrapped onto the bobbins, and save for a few exceptions, the coils are matched and wound with 42 or 43 AWG, so how much room is there to have a deviation in inductance, for a given resistance? Very little, as far as I can tell. So if you have an unusually high resonant peak for a given R, and we assume that L tracks closely with R, that leaves only a low C as a plausible explanation. Since scatter winding on such a tiny bobbin seems silly, I would have to assume the just used a thicker build on the magnet wire. 

As far as the inductance itself is concerned, and not as it relates to a calculated C, I think it's useful for determine the output volume of the pickup relative to other similar pickups, so I'd be curious to know which test frequency is best for an inductance that speaks to the overall volume of the pickup. If the inductance reading vary by up the 500mH, that might not be so bad, as high precision is not required if we're only interested in relative loudness, which also relies on the height of the pickup relative to the strings, so suppose I'd 500mH shy of where I want to be with my neck pickup relative to the bridge, putting it a couple millimeters higher or lower would close the voltage output gap. I would be more concerned if the pickups were like 2.0 henries apart.

[antigua]

Apparently the reason this '59 has a wood spacer is because it's the "vintage correct" '59, with long legs and braided wire. The short legged, four conductor '59 has the plastic spacer. Apparently SD's attitude is: get it all right, or get it all wrong. 

[reTrEaD]

long legs

What does that mean? How far the ends of the baseplate extend downward where the two mounting screws are?

[antigua]

Exactly

[stratotarts]

I wanted to nail down the inductance measurements of the SH1n at different frequencies. So I bypassed the integrator, and ran some plots with different swamping capacitor values. I measured all the test caps with a meter, and then subtracted the aggregate capacitance due to cable and pickup internal capacitance (calculated from a known estimate of L and an unloaded measurement of F).



It does appear that the measured inductance does not vary much when measured at different frequencies, with different chosen swamping cap values. For the benefit of non-members who can't enlarge the charts, the result is:

Frequency   Inductance
---------    -----------
310            4.2
420            4.1
715            4.0
930            4.2

Previously, I was not applying recursive calculation to my measurement with the integrator. When I do that, I compute 4.3H measured at 1150Hz (using the 4.7nF load which actually meters out to 4.3nF). So all in all, I feel that the integrator measurement must be quite close. For some unknown reason, the swamped readings don't work properly with the integrator at the lower frequencies. If I want wide range inductance measurements, I will probably have to build a permanant test fixture to repeat what I've done here in a convenient fashion.

The problem with understanding how the Extech meter fits into this, is not knowing exactly how it makes measurements at the settings of 100 and 1000 Hz. Although I didn't go below 310Hz, from what I saw, the Extech should read almost the same value for this pickup at the two settings. I think we are going to have to do some comparisons that will qualify the Extech readings, in order to come up with reliable inductance measurements.

[JohnH]

Id like to think about how those lower frequency results compare to the standard loaded and unloaded ones.

On the first post, the self capacitance is estimated at 128pF, ie about 0.13nF. I see the last table at lower frequencies added 0.31nF to the swamping caps. is that intended or is it a typo?

If you take the original unloaded readings and the 930hz results, the frequency ratio is 6240/930 = 7.237

If Inductance was constant (not saying it is though) across that whole range (but at an unknown value), then the self capacitance would be in proportion to the square of this ratio eg (assuming initial estimate 0.13nF):

7.237^2 = 45.0

6.73nF/45.0 = 149pF, ie quite believable, allowing for some lead cap too.

I think that is a good way to start, and then when I am modelling (just with the loaded and unloaded plots), I adjust my various components and keep cycling back through a calc like that to home in on the best set of values, which vary a little from there.

[stratotarts]

The 0.31nF is not a typo, it includes the test cable capacitance as well as the pickup capacitance. In this test, the new unloaded resonant frequency was 4425 Hz. Using the ratio method as you did, you have 6.73nF/(4425/930)^2 = 297pF which comes close to 0.31nF.

I start with a close approximation of the inductance, which can be obtained from any of the swamped measurements. In those cases, the capacitance is a small fraction, so the error is not great. Using that and the unloaded resonant frequency, I obtain an estimate of the intrinsic capacitance. Since the inductance is already known with fair accuracy, the intrinsic capacitance is therefore known with fair accuracy. I add this quantity to the capacitance of the swamping cap, to accurately reflect the load when subsequent calculations are made with the different swamping caps. Any error in the original estimate of inductance is divided by the ratio of swamped to intrinsic capacitance.

This procedure is open loop, not requiring any values that have been calculated, to replace earlier ones. Perhaps there is an algebraic method of doing it in one step, but I am not seeing it right now (my math abilities are often strained, so I generally avoid the effort unless there is a big payback). I added a column to my spreadsheet to accommodate this improvement.

I think that testing with 100nF was interesting because the error is inherently small due to swamping, and the frequency is lower, which some may prefer. The question of whether the inductance remains the same at higher frequencies remains unproven, we have not been able to verify it because there is a chicken-and-egg dependence between intrinsic capacitance and inductance at such frequencies. At the moment, my approach is to assume that inductance remains relatively constant vs. frequency. If the 6 part model predicts something different, then we do have a bit of mystery.

Without that assumption, we would not have a way to know whether we see a high inductance and a low capacitance, or a low inductance and a high capacitance at high frequencies. That is because the inductance can not be reliably isolated at high frequencies, and capacitance can not be reliably isolated at low frequencies.

Perhaps you could find out more with a network analyzer, but those are expensive.