SPICE Simulation Challenge

[Qiaozhi]

As for steel frequency response, seek a paper "FREQUENCY-DEPENDENCE OF RELATIVE PERMEABILITY IN STEEL" by N. Bowler. That paper alone is enough to sink a lot of PI nonsense related to Fe.

Found it.

[Teleno]

As ever you like it. Good luck obtaining correct phase response.

As for steel frequency response, seek a paper "FREQUENCY-DEPENDENCE OF RELATIVE PERMEABILITY IN STEEL" by N. Bowler. That paper alone is enough to sink a lot of PI nonsense related to Fe.

Care to explain in plain language the practical implications in metal detection/discrimination for those of us unable to understand the paper?

Thanks.

[dbanner]

Ok. What I understand from reading the posts is that a target whether ferrous or non ferrous, placed in the field causes an imbalance in the coupling, either by way of absorption or redistribution of magnetic field. Fine.
An IB Rx loop with perfect null will show rise in amplitude to both ferrous/non ferrous.
Rx loop which is misbalanced will show a reduction in amplitude dependent on which side misbalanced and whether the target is ferrous or non ferrous.
Now let us say that Rx loop is misbalanced such that a ferrous target causes reduction in amplitude, then if the ferrous target were sufficiently large, would it cause an initial reduction in amplitude then as the Rx loop moves through the perfect null, then an increase in amplitude? A sufficiently large ferrous target would affect the coupling to put Rx loop back to perfect null then beyond, so that it is now misbalanced to the other side resulting in increasing amplitude? If this is the case, then is this why some detectors cannot discriminate big iron?

[dbanner]

Maybe limiting the decreasing amplitude to zero would help discrimination of big iron targets?

[Qiaozhi]

Now let us say that Rx loop is misbalanced such that a ferrous target causes reduction in amplitude, then if the ferrous target were sufficiently large, would it cause an initial reduction in amplitude then as the Rx loop moves through the perfect null, then an increase in amplitude? A sufficiently large ferrous target would affect the coupling to put Rx loop back to perfect null then beyond, so that it is now misbalanced to the other side resulting in increasing amplitude?

That's exactly what does happen, and is why a large ferrous target will cause a beep when it's close to the coil.

If this is the case, then is this why some detectors cannot discriminate big iron?

As an example, a Garrett GTAx1000 will beep and indicate a target at the top end of the VDI scale. However, it is possible to notch out the top segment on the display and consequently ignore these "false" targets. Unfortunately this causes some reduction in sensitivity, so it's not recommended.
Which particular detectors are you referring to?

[dbanner]

Golden sabre, in fact all tesoro detectors of similar circuits. Or any coil so balanced( slightly off to one side).
My humble tgsl sometimes beep with big iron target, but not with every sweep, seems intermittent.
With tgsl, false beep on big iron is sure indication of a correctly misbalanced coil.

[dbanner]

How does one know the proper coefficients to apply in simulation?
I imagine that degree to which the Rx is decoupled when a target is presented would vary with the type and size of material and also the distance . I don't think conductivity of metals is linear across the non ferrous spectrum. Rather complex problem.
I'll have to go back to school.

[Qiaozhi]

Golden sabre, in fact all tesoro detectors of similar circuits. Or any coil so balanced( slightly off to one side).
My humble tgsl sometimes beep with big iron target, but not with every sweep, seems intermittent.
With tgsl, false beep on big iron is sure indication of a correctly misbalanced coil.

Just tested a Laser (Tesoro) Hawkeye, and it doesn't beep with an enormous rusty bolt close to the coil, just a bit of clicking. But a Tesoro Bandido does beep with the same bolt.

The coil for the Hawkeye has a 4-pin plug, and the Bandido uses the 5-pin type. These two coil types have vastly different inductances, and may be balanced differently as well. It's possible that the coil is mis-nulled on the other side of the balance point, in which case all ferrous targets will cause an increase in amplitude, and consequently cannot flip the RX signal. Whereas all non-ferrous targets will cause a decrease, and most are of a size that cannot flip the RX signal and cause it to be rejected.

In fact, I've just tested the Hawkeye with a huge biscuit tin lid close to the coil, and it gives a double beep. This is a sure sign that the RX signal has been flipped over.

[Qiaozhi]

How does one know the proper coefficients to apply in simulation?
I imagine that degree to which the Rx is decoupled when a target is presented would vary with the type and size of material and also the distance . I don't think conductivity of metals is linear across the non ferrous spectrum. Rather complex problem.
I'll have to go back to school.

The conductivity of non-ferrous metals is not linear across the spectrum. The targets we are attempting to simulate are just representative (ideal) samples.

[Davor]

The whole idea is to fit, within reason, the frequency response of a target, or a target family. Because amplitude may walk from nanovolts to tens of milivolts, only the frequency response is meaningful.

As for a steel response paper, it should be noted that there are 2 plateaus, the upper one at lower frequencies, and a lower at a higher frequencies. The PI technology samples are band limited at the lower side by fmin~1/(2t2), where t2 is the end of main sample, and for the short samples you are limited to higher frequency range. However, EF sample is frequency limited much lower. Fe has strong response at low frequency, and therefore the EF will be raised above the proper late sample without Fe. For that reason you have EF sinking the main sample below a no-target response (for some steel targets, but not all), and for some Fe targets appearing to go negative against the non-Fe. Also the schemes for discriminating Fe depend on the slope between plateaus to decide presence of Fe. But when you go through a few different kinds of steel, you'll notice that the slope is in fact mostly in a frequency response range of a PI, but not necessarily so. Hence Fe discrimination, as attempted so far, is a hit or miss.
What can be done? Re-think the role of EF in PI detectors used for Fe-infested terrains. Alternating polarity is a way to drop it completely.

[Teleno]

As for a steel response paper, it should be noted that there are 2 plateaus, the upper one at lower frequencies, and a lower at a higher frequencies. The PI technology samples are band limited at the lower side by fmin~1/(2t2), where t2 is the end of main sample, and for the short samples you are limited to higher frequency range. However, EF sample is frequency limited much lower. Fe has strong response at low frequency, and therefore the EF will be raised above the proper late sample without Fe. For that reason you have EF sinking the main sample below a no-target response (for some steel targets, but not all), and for some Fe targets appearing to go negative against the non-Fe. Also the schemes for discriminating Fe depend on the slope between plateaus to decide presence of Fe. But when you go through a few different kinds of steel, you'll notice that the slope is in fact mostly in a frequency response range of a PI, but not necessarily so. Hence Fe discrimination, as attempted so far, is a hit or miss.
What can be done? Re-think the role of EF in PI detectors used for Fe-infested terrains. Alternating polarity is a way to drop it completely.

Thank you for the explanation. If I uderstood correctly the Fe decay is not exponential (log linear) but more extended in time and can still be significantly present at the end of the Rx period or even above the level of the main sample?

[KingJL]

... Alternating polarity is a way to drop it completely.

Perfectly on target!!!

[Davor]

Thank you for the explanation. If I uderstood correctly the Fe decay is not exponential (log linear) but more extended in time and can still be significantly present at the end of the Rx period or even above the level of the main sample?

You may observe it as 2 kinds of contributions superimposed. But perhaps the correct approach would be to devise a simplified mathematical expression that would be simple enough to tackle with the known mechanisms. By casually browsing through patents, the manufacturer-who-must-not-be-named sees hyperbolas everywhere. How well it approximates Fe is debatable, but if a piece is small enough, just about any function will do for piecewise approximation. So why not a hyperbola?!

[Qiaozhi]

I may have reached an impasse with this endeavour.

As a sanity check I monitored the RX pre-amp output of Raptor versus the TX signal and compared the signal change for various targets. For the concentric coil used, all targets caused a phase-shift left. Non-ferrous (eddy) targets produced an increase in amplitude, and ferrous (magnetic) targets caused a decrease in amplitude. When Carl mentioned that the magnetic response is limited to 0 to , and eddy responses incur an automatic shift to give a range of 90 to , this is referring to:
which I can clearly measure on the scope by sampling at zero-crossing and quadrature points of the RX waveform wrt the no-target response.

The bottom line is that we have a problem. The simple LR model is ok for simulating non-ferrous (eddy) targets as you can see the phase-shift left and the increase in amplitude, but the ferrous (magnetic) targets (although they also phase-shift left) need to show a decrease in amplitude. Even if the LTspice simulator supported the Chan core model with mutual coupling between inductors (as Pspice does with the Jiles-Atherton model) the results would still be incorrect. To produce a decrease in amplitude when compared to the no-target response, the signal from the target would need to be partially in anti-phase with the no-target RX signal in order to cause the reduction, which clearly violates what we see on the scope. The problem is that the amplitude decrease is caused by a distortion in the magnetic field pattern whereby the applied field is absorbed by the ferrous target, and this mechanism is not modeled by the LTspice simulation. Teleno managed to fudge around the problem to some extent by assuming that the target affects the loop inductance, and hence by messing with the coil inductances you could give the impression that it was working. However, the purpose of this investigation was to create some target models that could be used in both VLF and PI simulations to investigate the possibility of ferrous detection with a PI using a mono coil. Unfortunately, simply fudging the results for a VLF does not achieve this goal.

I also tried using a collection of controlled sources to allow the Chan model to be included in a mutually coupled coil simulation, but with limited success, as the absorption phenomenon is not being modeled that way either. So, I'm stuck...

It looks like (in this instance) that real circuitry and a scope are the only alternative to poking around with LTspice.

[dbanner]

I problem is that the amplitude decrease is caused by a distortion in the magnetic field pattern whereby the applied field is absorbed by the ferrous target, and this mechanism is not modeled by the LTspice simulation..

Well, that sounds like the end of the road. Close but no cigar.
Someone will have to write code to upgrade spice software.
Or you could just subtract double the increase at the Rx output.