Target models for simulation

[Davor]

We are already using target models for rig simulations, so why not keep them in one place, and improve them according to new discoveries.
I am attaching a selection of models that also include two specific models of my own, Alu-foil and iron.
The common models, such as copper , silver etc., I translate tau directly to inductance by using 1ohm resistor in parallel. E.g. bulk aluminium has tau ~90μs, and it is modelled as 90μH in parallel with 1ohm. Easy.

Here are some of my estimates, feel free to amend the list:
Silver: 165μs
Copper: 135-155μs
Gold (bulk): 105μs
Aluminium: 90μs
Nickel, brass: 15μs
Bronze: 8μs
Aluminium foil 16μm @ 10kHz: 1.7μs
same foil @ 100kHz: 5.4μs
...
sea water: ~0.3μs (estimate for 25 cm thick layer)

I modelled aluminium foil as a little bit of a network that incorporates a capacitor, in order to accommodate a change in tau from 1.7μs to 5.4μs for frequencies from 10kHz to 100kHz. It has a somewhat steeper phase change.

Iron ... I have no measured values for iron, so it is a bit of my imagination at play. It had to be in a correct quadrant, and I didn't want to play with saturation etc. and I certainly didn't want to play with gyrators, negative reactances... so this is it. It is in a correct quadrant. I would be amazed if it works correctly in wideband systems, but it is a start.

Have fun, but please contribute your models as well

[Tinkerer]

Designing a target model, we need to look a the various characteristics of the target. For example a resistive target:
The size
The shape
The resistivity
The homogeneity

The response of a ring is different than the response of a sphere. How can we model this difference?

We can have a target of large size, but short TC, like a alu beer can. It gives a high amplitude response, but has a short TC.

Or we can take a blob of pure silver that has a small volume, but a long TC.

Why is the response of a gold nugget different from a 999 pure gold bar? Because the gold bar is pure and homogeneous, while the gold nugget is not.

For good simulations, we must design the sim to be able to take all these factors into account. I am not there yet, but, maybe with some help, we may all get there soon.

Sum of exponentials.zip

The attached files are zip for the .asc file, screen shot of cct and screen shot of simulation traces.

This is an IB coil with a transient in continuous wave form.

The yellow trace is the sum of exponentials of the ground X+R+coil etc. as you can see on top of the picture.

The other traces are of different targets.

Tinkerer

[Tinkerer]

PS. the target named 500us is actually a 50us target that is much larger than the target named FE target, which is also 50us. Interesting that the higher amplitude of the large target meets the same pivot point as the smaller target, due to the same TC.

Tinkerer

[Davor]

I seem to have a candidate for a variable voltage controlled coupling (K) for LTspice. I'll post it as soon as I test it a bit more. I also have some ideas on how to implement some lossy ferrous materials as well. Cross fingers...

[Davor]

Here is my take on a model of a transformer with variable coupling that is achieved by a voltage controlled node. It is simple and idealised, so any parasitic perks must be added as lumped elements.

It is a nested hierarchical model. Therefore files that start with h_Vari-K-xformer and varicoil are in fact the model. You can play with those as well - there are no secrets. I can make it as a standard model and symbol combo, and it is surprisingly simple to make from a hierarchical model, but I have chosen to keep it this way for the time being. Point is that you can play much easier and more intuitive this way than tweaking it in a text editor, wondering what's wrong.

The aim of this model is to finally be able to simulate the response of the gain blocks as a function of simulated target sweep. So now I can serve the K node with some windowing function that will mimic the target underneath the coil, with amplitude and phase changes, and see what else can be done with my gain block filters etc.

Just use your imagination

Next up will be a model with a core to mimic ferrous targets. Hope it works out the way I expect. Don't hold your breath though...

[Davor]

I was playing with viscous soil model, and it seem I got one. I get a straight line in log/log scale and near perfect 1/t response from 100ns to 10ms. It is too soon to say if it is also valid for VLF, but by following the logic of the viscous soil response it seem to be OK.
I also observed some of the familiar curved responses at short delays related to a short "charging" period. Most probably the differences in exponents are merely a result of a charging pulse interfering with a flyback pulse.
It is also interesting to see this effect in frequency domain. Guess the future of discrimination is in frequency domain processing. In a way the GB mechanisms in PI detectors are doing that very thing already.
I'll run some more sims to make sure everything is OK, and I'll post the model here.

[Davor]

Here goes viscous soil model...
The whole idea is to model such a soil response by its effects on PI detectors, where viscous soil signal appears to follow hyperbolic law. There are no straightforward solutions hor hyperbola in electronics, and the closest thing to it is a piecewise linear approximation.A hyperbola is a function described as x^(const), where const is a negative number. Various sources give the viscous soil the value of const~-1 and many show some deviation from -1. It may be a genuine feature of the soil, or it may be simply a shortcoming of the equipment - I can't say. My goal was to make a reasonably simple model that can be used with spice and see if equipment design can influence the value of const, and it can.

Attached is a LTspice model with a simplified and idealised PI Tx to facilitate discovering the effects without non-ideal component influences. It is designed as a ridiculously long time constant current charged PI coil with a fast flyback to show effects down to nanosecond scale, and avoid seeing effects of the charging pulse. This model is nearly perfect over 5 decades.

If coil "charging time" is shortened to more realistic durations, it shows that a slight knee in response appears at just about the charging time.

[Davor]

Here is the same thing in IB configuration and frequency domain. It is interesting to note that phase is within 2° from 180° within over 3 decades, and the amplitude is also fairly constant.

[Carl-NC]

Wow, very nicely done.

[Davor]

Thanks, I had my moment of epiphany so I didn't want to waste it.

[Qiaozhi]

Here is the same thing in IB configuration and frequency domain. It is interesting to note that phase is within 2° from 180° within over 3 decades, and the amplitude is also fairly constant.

We had a power cut here nearly all day yesterday, and the entire village is now running on a backup generator. Apparently there's an underground fault somewhere on the local farm land in a mile long stretch of cable. Consequently I've only just got back on the internet, and managed to run both simulations successfully.

As you've probably noticed, Davor has manually set the X and Y axis ranges, which you can easily do by clicking the right-hand mouse button in the plot window, deselect Autoranging, and select Manual Limits. And ... one small tip for LTSpice users (if you don't know it already) is that you can save your plot settings to a file. Then, after re-running a simulation, you can select Plot Settings > Reload Plot Settings to reformat the layout.

Very interesting results.

[Davor]

I think I did it again.
Here is a plot obtained from a single universal ideal target set to behave as non-ferrous (blue), salt (orange) and ferrous (gray), obtained by export from LTspice, and converted to a polar plot in excel that is lacking in LTspice. I need to refine it a bit before posting, but the funniest part is that the same coil is inductively coupled to a Tx and Rx, and only a single parameter is changed by ".step param list" command.

[Davor]

So here is a preliminary scheme which I tested in frequency domain. It is based on a negative inductance obtained by a NIC circuit comprised of an inductor L4 and a voltage controlled voltage source E1. The gain of VCVS is selected to be 2 in order to obtain exact negative of an inductor L4 within NIC.
When negative inductance is smaller than L2, the total inductance is positive, and target is a good old non-ferrous one we know very well.
For negative inductance equal to a positive, the response is purely resistive, and phase is 90°.
Ferrous response is obtained with negative inductance overpowering the positive.

[Davor]

A little bit of cross-posting... I already posted this on a different topic, but it should stay here.

The model of a ferrous target in the above post is simple and beautiful, and it produces a proper impedance plane response, and it is wrong
However the impedance plane LOOKS fine, its frequency response in impedance plane is opposite from what it should. So for a single frequency it could be used, but beyond that the model is useless. its response was completely wrong. Instead of a response circling from left to right with frequency, it did the opposite. But at least the quadrant was correct.

Now I have a correction, and it seem OK. It is based on a sum of a positive and negative inductance, which effectively gives an equivalent of μr >> 1
Also it is rotating as it should, and it ends at high frequencies as mostly resistive, yet still on ferrous side. Fun part is that I can replicate almost any material with published real and imaginary plots vs frequency.

It should be properly tested against the measurements in reality, but here goes...