SIMULATIONS

[mikebg]

All depends on sample delay or frequency

Skin effect dominates on copper plated coin with iron core. And the magnetisation of the iron core is there too.
Do you all see the difficulty in implementing a working discrimination, when the orientation or the ground mineralization screws your theoretical models?
Aziz

"Skin effect" is a term of frequency domain. When you use it, you should point at what point of frequency spectrum it dominates. The same is valid for time domain: At what sample delay magnetization of the iron core prevails?
In frequency domain all is clear. When we have combination of conductivity and permeability, there is a resonance frequency (point Q) at which X=0. If you have coins 1, 2 or 5 Eurocents, you can measure the resonance frequency.

[Ferric Toes]

Could we compare the non conductive ironstone with a ferrite?
Could we find a ferrite that has the right hardness to be similar enough?

Tinkerer

Yes you can. "Soft ferrites" as used in TV line scan or switchmode psu exhibit high permiability/susceptibility but little or no viscosity; much like a sample of California black sand that I have. Ferrites as used in MW and LW radio antennas exhibit viscosity. It is all to do with the grain size. Fine grain has viscosity, coarse grain hasn't.

Soon I wil be able to do log/lin, log/log, and linear plots to see what is going on. Just have to get a small batch of my Magnetic Viscosity Meters finished. The latest linear curves I have done were using the MVM sensor and front end
amplifier.

Just bought a beer with a steel cap and looked at the linear scope plot. Bit surprised at what I saw. Report later.

Eric.

[Ferric Toes]

Skin effect dominates on copper plated coin with iron core. And the magnetisation of the iron core is there too.
Do you all see the difficulty in implementing a working discrimination, when the orientation or the ground mineralization screws your theoretical models?
Aziz

Plated 2p is 93% mild steel core and 7% copper plating. Weight of coin 7.12gms. Copper contribution is about 0.5gm spread over the whole surface. Skin effect will be there but invisible to PI as we know it. Bet that all the response we see is the mild steel core.

[Davor]

Guess you'll be a bit surprised again
At VLF these behave funny, provided you have a two tone machine, and in PI they may behave even more funny at fast samples.

I often laugh when I come across such coins with my detector - the sound of junk - you can't miss it.

[Tinkerer]

Plated 2p is 93% mild steel core and 7% copper plating. Weight of coin 7.12gms. Copper contribution is about 0.5gm spread over the whole surface. Skin effect will be there but invisible to PI as we know it. Bet that all the response we see is the mild steel core.

Skin effect will be there but invisible to PI as we know it. Bet that all the response we see is the mild steel core.[/QUOTE]
This is one of the problems of traditional PI. We can not see what happens during the first few us after switch OFF. This information information remains hidden in the Flyback and Pre-amp saturation.

Is there a way to see what happens during the first few us? Yes, by using an IB configuration we can largely eliminate the Flyback in the RX coil.
We can also make simulations that show us that time period, but, unless we have a real circuit to compare the simulation with, we do not really know if the simulation results are right.

I have designed a circuit that allows us to see the total cycle time. The simulations are much the same as the wave forms of the real circuit. There are some distortions of the waveform, due to the filtering, but they can easily be recognized as such.

For this purpose I make sure that the pre-amp does not saturate. This means relatively low gain. I make up for that by increasing the TX power, such that the target response can be seen, when using a real detector coil.

Of course, now I have the problem that nobody recognizes the wave form they see, because they have never seen it before.
To get out of this dilemma, I thought to show 2 different simulations side by side, Traditional PI and IB-PI. I am still stuck with that. How do I make a traditional PI simulation to come out right, if I have no means to see a real corresponding circuit?

Attached is the screenshot of the simulation as well as the asc file in zip format.

There are 2 circuits and 3 different targets. Changing the inductance in the targets, changes the TC of the target. Ex, 10pH=TC of 10us.

The problem is with the K coupling between the various coils. For the individual sim's, the K has to be changed, I have not managed yet to run them at the same time.

Maybe somebody can help?

Tinkerer

[Tinkerer]

Here we have the traditional PI simulation.
3 non magnetic targets, TC's 10us, intense blue, 100us, red, 500us, light blue.
With a TX of 50us, we see how the 10us target TX ON eddy currents have reached their peak and are diminishing, while the eddy currents in the longer TC targets are still raising.
The light green trace is the TX coil current. The raise is near linear, because the TX time is very much shorter than the coil TC. At switch OFF, the current drops fast, then raises again, to drop to the 0 level, due to the critical damping resistor.
What causes the transient during the decay curve? It is the stray capacitance, set at 200p, to represent the inter-wire and wire to shield capacitance.

The TX coil also serves as RX coil with the target coupling at 0.0001

The target response traces are separate from the TX trace, but this we can not see on a real circuit. What we see on a real circuit is the sum of the TX wave form and the target wave form and all the other stuff, like environmental noise, ground etc. that the RX coil sees.

On the simulation we can sum the traces by adding <I(L1)+I(L3)> etc. to give us a trace that is more like the trace we see on the scope from the real circuit.

For complex target responses, like magnetic and conductive ground etc. I sum R and X responses. Summing several targets with different TC's and different R and X, seems to be able to reproduce just about any target response wave shape.

This is as far as I got with the simulation. Please, help me here. The better the simulations, the better we can all understand the real circuits.

An interesting add, would be a noise source.

[Aziz]

Oh well, I have completely eliminated the magnetic viscosity effects.
But the skin effect is even present at the decaying eddy currents.

Eric, I'm looking forward to your interesting report.

Cheers,
Aziz

[Ferric Toes]

OK, that makes sense, as I've just realised that you're putting the target inside a solenoid coil. This is not the same situation as you would get using a typical detector coil with a metal target below it. In the latter case the vertical flux lines would tend to cancel due to superposition, resulting in a stronger signal for the horizontal target.

Not quite sure what you mean here. The coil I am using is a vertical shielded solenoid of such a length that when I put a 10ml sample holder inside on the platform, it is in a uniform vertical field. This is the situation for soil and rock measurements and OK for a vertical coin, where in the case of the plated 2p, I get the magnetic response. With the coin horizontal as in the picture, the flux/coin situation should equate to the object being under the centre of a standard detector coil.

Eric.

[ODM]

I'm curious about this measurement setup and its coil setup. Is there a separate receiving coil or flux sensor, and what measurements does it yield? There is no closed magnetic loop like in most magnetic characterizing setups.

Some earlier threads are mentioning the MVM, too. I presume it's no longer an active product. What type of setup is/was it?

[PiTec]

Hi,

basically, the response for iron targets is similar to that of non-iron ones, except that depending on shape, orientation and alloy some reactive response has to be added in case of an IB setup.

The problems with crown corks have been discussed already – and I think this actually makes a useful simulation with LTspice impossible. As proposed, you may limit the simulation to 3 target positions (plus 90° rotation this would make 6), but to really simulate the influence of target position and rotation you would need a suitable software, like the 3D coil simulation software from Aziz, but with a life target response plugin With LTspice, you would have to use experimental values for target TCs and coupling coefficients for each target position and rotation angle.

A very basic simulation that shows the principle could be like this:

• IB setup so the reactive response can be seen as well (double D with two identical RX coils RX1 and RX2 and the TX coil around both). The response of a conventional PI is of course also visible after the flyback pulse.
• Target TC can be varied to simulate in which angle the TX field strikes the target. For simplicity only one single TC, although real iron targets may have several TCs plus other effects that may influence the response, like skin effect or magnetic viscosity.
• The coil coupling coefficient RX1 to target determines the amplitude of the resistive response.
• The coil coupling coefficient TX to RX1 determines the amplitude of the reactive response.

Here is a setup similar to the one I used in the Triangular Wave thread. Ideal TX coil with an additional constant current period so that the decay curve after the driving pulse is visible as well. The TX on/off ratio is limited to 2:1 (50µs to 25µs) to keep the amplitude differences small. The coupling coefficients K24 and K14 will be changed in the different simulations. With K24 = K25 there is no reactive RX signal, and with K14 = K15 there is no resistive RX signal. Only the target with TC=10µs is being used.

First one with no target response (K25 and K24 = 0.01), but with a reactive signal as if you moved a ferrite core up and down over the RX1 coil (K14 varied). Soil with magnetic susceptibility and mechanical coil instability would show the same response:



Now for comparison a target with a resistive response, like a gold ring (K14 = K15, K24 changed to 0.012):



And finally the crown cork. I measured 9-10µs TC horizontal over the RX1 coil center, and 12-13µs vertical (real hardware, scope readout). As expected, the reactive response is much larger when it is vertical. In the simulations below I modified the coupling coefficients so that the simulation results are very close to the scope readouts.

Left side horizontal crown cork with K14 = 0.001001 and K24 unchanged at 0.012. Right side vertical crown cork with K14 = 0.001004 and K24 unchanged at 0.012. I did not change the target TC (both simulations 10µs) as the real values do not differ so much.



Both curves for the crown cork have basically the same shape as the curve for the gold ring with the same TC. They are just more or less shifted due to the reactive component during the TX on/off periods. To make the curves easier to compare, the amplitudes of the underlying decay curves are identical in each of the above plots. If you rotate a real iron target at the same distance from the coils, the vertical response may be larger, especially if it is a long target like a nail.

Thomas

[Ferric Toes]

The crown caps are not what I expected. I tested two, which appear to be from different manufacturers, and neither show any magnetic lag (X). They are, however, strongly attracted to a magnet. The total decay is 100uS when perpendicular the the TX field and 75uS when parallel to the field. In other words it is behaving like a non-ferrous target. No wonder they are a problem.

The UK steel cored coins are just as much of a problem here, as 2p and 1p denominations have been made this way since 1992. With them, there is a clear magnetic signature (long decay when // to field), so discrimination should be easy by observing the TC change when scanned, or large X response from a balanced coil.

I can only assume that the crown caps that I have are a steel alloy which has constituents that prevent a magnetic lag.

Eric.

[Tinkerer]

The crown caps are not what I expected. I tested two, which appear to be from different manufacturers, and neither show any magnetic lag (X). They are, however, strongly attracted to a magnet. The total decay is 100uS when perpendicular the the TX field and 75uS when parallel to the field. In other words it is behaving like a non-ferrous target. No wonder they are a problem.

The UK steel cored coins are just as much of a problem here, as 2p and 1p denominations have been made this way since 1992. With them, there is a clear magnetic signature (long decay when // to field), so discrimination should be easy by observing the TC change when scanned, or large X response from a balanced coil.

I can only assume that the crown caps that I have are a steel alloy which has constituents that prevent a magnetic lag.

Eric.

Thank you for the feedback and the TC data.

If we look at the response of steel as a combination of a ferrite and of a low conductivity conductor, we can simulate a similar response. Ferrites come in different grades, according to the frequency at which they are going to be used.

A ferrite core for a high frequency transformer can not have a lag, because then the impedance would be great.

On the other hand, a ferrite sleeve on a 50Hz power line is designed to present low impedance at that frequency.

We could term this to be frequency susceptibility.

Steel can be designed to be frequency susceptible. Adding silicon to the alloy makes for better 50Hz transformer cores.

Adding 14% of Manganese, makes the steel non magnetic and extremely hard, to be used for rock drills. Since such rock drill bits are not attracted to even a very powerful electromagnet, they present a problem for the rock crushers in the mining business.

Some crown caps are plated with zinc or tin or cadmium or whatnot metals. This can make things a bit more complicated still. It is a fact that copper plating on a steel rod changes the HF frequency conductance considerably.

All that may mean that my choice of a crown top as a universal test target turned out to be not universal.

But, crown tops are a big pain when detecting. So it still may be worth defining why they behave as they do.

In one of my PI designs, I tried to differentiate the crown tops by enhancing the X response. This seemed to work OK and might work in the field, if this feature could be switched on only to verify a doubtful target response.

By the way, I use the X response denomination as the response that shows as 180 degrees from the R response, while there is current running in the coil. This is very easy to see during the TX ON time.
When we switch OFF, we still have current flowing in the coil, until the decay curve reaches 0. However, at some time on the decay slope, the X response becomes less than the R response. I call the point where the response crosses over the coil decay curve, the pivot.

Sampling before the pivot enhances the X response. Sampling after the pivot enhances the R response.

Tinkerer

[Tinkerer]

Hi,

basically, the response for iron targets is similar to that of non-iron ones, except that depending on shape, orientation and alloy some reactive response has to be added in case of an IB setup.

The problems with crown corks have been discussed already – and I think this actually makes a useful simulation with LTspice impossible. As proposed, you may limit the simulation to 3 target positions (plus 90° rotation this would make 6), but to really simulate the influence of target position and rotation you would need a suitable software, like the 3D coil simulation software from Aziz, but with a life target response plugin With LTspice, you would have to use experimental values for target TCs and coupling coefficients for each target position and rotation angle.

A very basic simulation that shows the principle could be like this:

• IB setup so the reactive response can be seen as well (double D with two identical RX coils RX1 and RX2 and the TX coil around both). The response of a conventional PI is of course also visible after the flyback pulse.
• Target TC can be varied to simulate in which angle the TX field strikes the target. For simplicity only one single TC, although real iron targets may have several TCs plus other effects that may influence the response, like skin effect or magnetic viscosity.
• The coil coupling coefficient RX1 to target determines the amplitude of the resistive response.
• The coil coupling coefficient TX to RX1 determines the amplitude of the reactive response.

Here is a setup similar to the one I used in the Triangular Wave thread. Ideal TX coil with an additional constant current period so that the decay curve after the driving pulse is visible as well. The TX on/off ratio is limited to 2:1 (50µs to 25µs) to keep the amplitude differences small. The coupling coefficients K24 and K14 will be changed in the different simulations. With K24 = K25 there is no reactive RX signal, and with K14 = K15 there is no resistive RX signal. Only the target with TC=10µs is being used.

First one with no target response (K25 and K24 = 0.01), but with a reactive signal as if you moved a ferrite core up and down over the RX1 coil (K14 varied). Soil with magnetic susceptibility and mechanical coil instability would show the same response:

[ATTACH]23067[/ATTACH]

Now for comparison a target with a resistive response, like a gold ring (K14 = K15, K24 changed to 0.012):

[ATTACH]23068[/ATTACH]

And finally the crown cork. I measured 9-10µs TC horizontal over the RX1 coil center, and 12-13µs vertical (real hardware, scope readout). As expected, the reactive response is much larger when it is vertical. In the simulations below I modified the coupling coefficients so that the simulation results are very close to the scope readouts.

Left side horizontal crown cork with K14 = 0.001001 and K24 unchanged at 0.012. Right side vertical crown cork with K14 = 0.001004 and K24 unchanged at 0.012. I did not change the target TC (both simulations 10µs) as the real values do not differ so much.

[ATTACH]23066[/ATTACH]

Both curves for the crown cork have basically the same shape as the curve for the gold ring with the same TC. They are just more or less shifted due to the reactive component during the TX on/off periods. To make the curves easier to compare, the amplitudes of the underlying decay curves are identical in each of the above plots. If you rotate a real iron target at the same distance from the coils, the vertical response may be larger, especially if it is a long target like a nail.

Thomas

Thomas, thank you for the simulations and explanations. I will need some time to analyse all that.

For now I have just one comment:

My reason for making 3 simulations, with the target
1) over the center of the coil, the magnetic field lines cut through the target vertical, at 0 degrees, south to north
2) over the rim of the coil, the field lines cut through the target at 45 to 90 degrees, depending how close to the rim.
3) Off to the side of the coil, the field lines cut through the target at 180 degrees, the X response of a ferrite should be 180 degrees different from 1).

Therefore, we do not need to rotate the target, changing the position relative to the coil, while maintaining the same orientation should be good enough.

Somewhere I posted a picture showing the field lines of a TANDEM COIL. This coil arrangement is designed to enhance the response of flat targets at different angles.

Tinkerer

[Ferric Toes]

I have come to the conclusion that simulating ferrous objects is going to be a nightmare. I have just made some test targets all the same diameter as a 2p coin but different thicknesses. They are made from steel shim stock and of thickness 2, 4, 6, 8 and 12thou. As you might guess the shim stock is rather old. All display like the crown cap with no apparent viscosity lag. Perpendicular to the TX field the decay time appears to be proportional to thickness i.e. the 4thou is twice the time of the 2thou. The other orientation starts shorter and also doubles. It appears that when the objects are parallel to the TX field, there is a certain permiability effect which has the effect of magnifying the smaller X section eddy currents. Bring up a strong ferrite magnet and the decay can be neutralised, indicating saturation. The 0.5mm wire on a 1N4148 diode gives a strong response that disappears when a strong field is applied. Not so the 2p coin but maybe my magnet is not powerful enough.

Eric.

[PiTec]

I have found two typos in my previous post:

1. Horizontal and vertical TCs swapped, i.e. it should read ‘I measured 9-10µs TC vertical over the RX1 coil center, and 12-13µs horizontal …’. So this is of course in accordance with your measurements, Eric.

2. K14 = 0.001001 / K14 = 0.001004 should read K14 = 0.01001 / K14 = 0.01004

My reason for making 3 simulations, with the target
1) over the center of the coil, the magnetic field lines cut through the target vertical, at 0 degrees, south to north
2) over the rim of the coil, the field lines cut through the target at 45 to 90 degrees, depending how close to the rim.
3) Off to the side of the coil, the field lines cut through the target at 180 degrees, the X response of a ferrite should be 180 degrees different from 1).

Therefore, we do not need to rotate the target, changing the position relative to the coil, while maintaining the same orientation should be good enough. ).

1) and 2) should be very similar (horizontal at the center same as vertical over the rim and vice versa), but 3) may have other reactive component amplitudes.

I have about 20 different crown corks here, and they are all quite similar regarding their two main TCs (9-10µs when parallel to the TX field, 12-13µs when perpendicular).They all show a similar and strong reactive response when they are parallel to the TX field, but also a less stronger response when perpendicular to the TX field. The amplitude ratio is approx. 3:1 for equal distances to the TX coil. For comparison, I measured a 5 Eurocent coin (also copper plated steel). Here the ratio is approx. 8:1, so they they are more difficult to identify as iron when buried horizontally.

Here is the LTspice file:
iron_target.rar

Thomas