SIMULATIONS

[Tinkerer]

Often the simulations do not produce quite the same output as the real component circuits. But even then they can give us an approximate idea what to expect from the real circuit.

Of course, if the simulations are different, it is because we have not taken account of all the involved factors. Therefore, we try again and after a few tries, we get closer to the real results. It is a learning process.
Learning is good, but takes time.

Could we accelerate our learning process by Crowd Sourcing PI simulations? There are a few people on this forum, with lots of knowledge. Others, like myself, are just trying to learn something.

Would the knowing ones teach the students? I hope.

How can we simulate an FE target? Let's take the response of a real target as goal. The infamous bottle top or crown cork. It is a tricky target, because it shows very different characteristics, when presented to the coil at different angles.
In the vertical position, the X response is dominant, but there is still a little R component too.
In the flat position, the R response is dominant, but the x response is still there to some degree.

If we want to find ways to perfectly discriminate this pesky target, we must fully understand the reasons for the different responses.
If we fully understand the reasons, we have a better chance to design a perfect discrimination.

Let's simulate!

Tinkerer

[Pete the Builder]

Slightly off-topic, but... I'd like to ask what simulation software you guys use?

I've been looking around for software to run on my Macbook Pro, but the few (4 or 5) open source (i.e. free) simulators all have pretty steep learning curves.

There's more available for Windows (I have a number of very high-end Windows machines that are perfect for simulation), but that just complicates things!

I have Protel DXP (version 7.0), which has great Spice simulation, but it's a real handful to get configured, especially since I use many non-standard library devices, many of which have no Spice model attached. I feel like a Cessna pilot sitting in the cockpit of an FA-18 fighter, looking for the key!

If anyone could perhaps point me in the direction of a good simulator package that a new user could use to get started, and perhaps that others here use, it would be much appreciated! The more common the better, at the moment - this would allow a lot of interchange between members. If that's on anyone's priority list!

I don't mean to hijack this thread, so PM would be acceptable if it doesn't detract from the real purpose of this thread.

Sorry for the interruption. Normal service will resume in 3...2...1...
PtB

[Ferric Toes]

Often the simulations do not produce quite the same output as the real component circuits. But even then they can give us an approximate idea what to expect from the real circuit.

Of course, if the simulations are different, it is because we have not taken account of all the involved factors. Therefore, we try again and after a few tries, we get closer to the real results. It is a learning process.
Learning is good, but takes time.

Could we accelerate our learning process by Crowd Sourcing PI simulations? There are a few people on this forum, with lots of knowledge. Others, like myself, are just trying to learn something.

Would the knowing ones teach the students? I hope.

How can we simulate an FE target? Let's take the response of a real target as goal. The infamous bottle top or crown cork. It is a tricky target, because it shows very different characteristics, when presented to the coil at different angles.
In the vertical position, the X response is dominant, but there is still a little R component too.
In the flat position, the R response is dominant, but the x response is still there to some degree.

If we want to find ways to perfectly discriminate this pesky target, we must fully understand the reasons for the different responses.
If we fully understand the reasons, we have a better chance to design a perfect discrimination.

Let's simulate!

Tinkerer

Let's deal with the real world first. When you scan over a horizontal steel bottle cap, you get a large change in TC. As the coil edge approaches the cap the field is parallel to the plane of the cap (X direction). The response is primarily a viscosity one, amplified by the permiability, and is a long decay. As the cap passes under the coil centre the response shifts to a conductivity one (R direction) with only a small X component, as you say. Steel is not a very conductive metal so the decay is fast. This can be seen quite nicely by taking an early sample and a later sample and dividing the early by the late. This gives a result that is virtually independent of amplitude but changes for different TCs. A horizontal non-ferrous coin will give a constant reading dependent on conductivity, while a ferrous bottle cap will give a variable reading as scanned. Vertical bottle cap will reverse the responses but you will still see a change.

Eric.

[Davor]

There is more to Fe targets than the PI Rx can show, mainly because it neglects the time zero behaviour. I made some Fe targets for simulation that are less than perfect for two reasons. I have no PI machine to compare my findings with simulation, and second, due to Spice limitations I was able to simulate a target with split personality - one side links to Tx, while the other links to Rx.
Besides, my "target" so far never crosses the 90° barrier, while real world Fe targets do at higher frequencies, where eddy currents overcome ferrous behaviour. Perhaps it would be wise to split the Fe target in simulations in several targets to achieve the obvious - dissimilar behaviour at low against the high frequencies due to completely different modes of operation.

Then we hit my problem No. 1 - I do not have real world PI samples to compare my simulations with.

[joop]

Slightly off-topic, but... I'd like to ask what simulation software you guys use?

Judging from all the electronics simulations I see on this forum, LTspice is your friend. See http://www.linear.com/designtools/software/#LTspice
It has good documentation and an active user community at Yahoo Groups. There you can also find plenty of tutorials.

[Sargon]

I'd like to ask what simulation software you guys use?

There are many Spice-based programs out there. So far, I have worked with PSpice (now Cadence OrCAD), LTSpice (free), TinaTI (free) and currently NI Multisim is my simulator of choice. There's also a low-priced student edition available. What I like about Multisim is the GUI, the large database of elements and the instruments.


Greetings and happy hunting!

Sargon

[Tinkerer]

Let's deal with the real world first. When you scan over a horizontal steel bottle cap, you get a large change in TC. As the coil edge approaches the cap the field is parallel to the plane of the cap (X direction). The response is primarily a viscosity one, amplified by the permiability, and is a long decay. As the cap passes under the coil centre the response shifts to a conductivity one (R direction) with only a small X component, as you say. Steel is not a very conductive metal so the decay is fast. This can be seen quite nicely by taking an early sample and a later sample and dividing the early by the late. This gives a result that is virtually independent of amplitude but changes for different TCs. A horizontal non-ferrous coin will give a constant reading dependent on conductivity, while a ferrous bottle cap will give a variable reading as scanned. Vertical bottle cap will reverse the responses but you will still see a change.

Eric.

Eric, thank you for the very precise description of the bottle cap response. This is exactly what I have been observing, but have not been capable of describing.

Could you give us some idea about the TC of the responses.

Now I want to make a LTSpice simulation showing that behavior.

I am thinking of designing 3 different bottle top targets,
-to simulate the target at the center of the coil,
-the target over the rim of the coil
-and the target at some distance beside the coil.

Personally, I prefer a concentric co-planar IP coil configuration. With this configuration I can sample during TX, which shows the R and X at 180 degrees.
However, most of the people are more familiar with the traditional PI configuration, so we should design both kinds of configurations ans show them side by side.

Tinkerer

[Tinkerer]

Slightly off-topic, but... I'd like to ask what simulation software you guys use?

I've been looking around for software to run on my Macbook Pro, but the few (4 or 5) open source (i.e. free) simulators all have pretty steep learning curves.

There's more available for Windows (I have a number of very high-end Windows machines that are perfect for simulation), but that just complicates things!

I have Protel DXP (version 7.0), which has great Spice simulation, but it's a real handful to get configured, especially since I use many non-standard library devices, many of which have no Spice model attached. I feel like a Cessna pilot sitting in the cockpit of an FA-18 fighter, looking for the key!

If anyone could perhaps point me in the direction of a good simulator package that a new user could use to get started, and perhaps that others here use, it would be much appreciated! The more common the better, at the moment - this would allow a lot of interchange between members. If that's on anyone's priority list!

I don't mean to hijack this thread, so PM would be acceptable if it doesn't detract from the real purpose of this thread.

Sorry for the interruption. Normal service will resume in 3...2...1...
PtB

Thank you for the feedback.

LTSpice is easy, free and does the job. It has a large component inventory, with most of what we need. I do miss an IGBT though, but maybe somebody knows how to import a model?

Tinkerer

[Tinkerer]

There is more to Fe targets than the PI Rx can show, mainly because it neglects the time zero behaviour. I made some Fe targets for simulation that are less than perfect for two reasons. I have no PI machine to compare my findings with simulation, and second, due to Spice limitations I was able to simulate a target with split personality - one side links to Tx, while the other links to Rx.
Besides, my "target" so far never crosses the 90° barrier, while real world Fe targets do at higher frequencies, where eddy currents overcome ferrous behaviour. Perhaps it would be wise to split the Fe target in simulations in several targets to achieve the obvious - dissimilar behaviour at low against the high frequencies due to completely different modes of operation.

Then we hit my problem No. 1 - I do not have real world PI samples to compare my simulations with.

This reminds me that I have a few old PI boards, fully populated and tested 20 years ago. Maybe some people would be interested in such a board for experimental purposes?

These are Non Motion boards.

Tinkerer

[ODM]

I love LTSpice. "freeware", fast and stable compared to some high priced commercial alternatives, and freely importable models. About adding parts: You can either add the part files to the root of your simulation file and use the "include" spice directive, or just cut and paste the actual model description as a spice directive. If the subcircuit terminals are wrong, you can open a symbol, rearrange the terminals, and then save the now-compatible symbol with a different name.

Another swift thing is using f4 (tag) for connecting multiple nets without wiring them together, and drawing diagrams by just tossing the parts first (l, c, r, g are quick keys for the components) and then drawing wire through the parts - it will clip into the terminals instead of short circuiting the part like some higher priced spice - even the UI is streamlined!

[Tinkerer]

I love LTSpice. "freeware", fast and stable compared to some high priced commercial alternatives, and freely importable models. About adding parts: You can either add the part files to the root of your simulation file and use the "include" spice directive, or just cut and paste the actual model description as a spice directive. If the subcircuit terminals are wrong, you can open a symbol, rearrange the terminals, and then save the now-compatible symbol with a different name.

Another swift thing is using f4 (tag) for connecting multiple nets without wiring them together, and drawing diagrams by just tossing the parts first (l, c, r, g are quick keys for the components) and then drawing wire through the parts - it will clip into the terminals instead of short circuiting the part like some higher priced spice - even the UI is streamlined!

Thanks for the tips. You seem to know LTSpice well. Could you help me with a problem?

When I simulate an IB-PI circuit, with target models, I end up with many different inductors with several different coupling (K) factors. LTS then tells me that the inductors are impossible to build. Is there a way to fix that?

Tinkerer

[ODM]

Interdependent K factors are hard to simulate, I haven't tackled that myself. It might be possible by using some "middleman" series-coupled L to couple the same current into different coils - haven't really looked into it, either, the sims I've done have mostly been about effects of (manually added) parasitics.

Even then, there needs to be some resistance in the inductors, even if it's just micro-ohms. Pure inductors in a multiple coupling circuit can't be simulated outright.

It's a few years too long since my physics and math courses. Shouldn't have let the rust build like that... Ah well. Back to text books!

[Tinkerer]

Interdependent K factors are hard to simulate, I haven't tackled that myself. It might be possible by using some "middleman" series-coupled L to couple the same current into different coils - haven't really looked into it, either, the sims I've done have mostly been about effects of (manually added) parasitics.

Even then, there needs to be some resistance in the inductors, even if it's just micro-ohms. Pure inductors in a multiple coupling circuit can't be simulated outright.

It's a few years too long since my physics and math courses. Shouldn't have let the rust build like that... Ah well. Back to text books!

Thanks for the feedback. I will post a simulation circuit and then maybe we can figure out how to do it or find some help.

[ODM]

Also, the behavioral current/voltage sources might be of some use. They can be made to reflect the current/voltage of various parts, or a function of such. I regularly use those for calculating the actual power across various components - and a RC lowpass filter gives an average of it.

[Davor]

True, and mostly because you can't use negative inductance in spice, but in both cases such simulations are very unstable, and the low point is that you must use two target "coils" for coupling to Tx and Rx respectively. Been there...
I started a similar discussion here: http://www.geotech1.com/forums/showthread.php?19872 and I placed my simulated targets as well. Now I know exactly where my Fe model fails, but so far I did not find a way to mend it.

To cut the long story very short - ferrous materials tend to cross the 90° (salts response) line at high frequencies, and there they appear as poor conductors. Therefore stainless steel is invisible for early sampling PI machines. Well, my model fails there because it never crosses the 90° line. However, I learned that you must include some extreme tau to simulate magnetisation.

However this may sound as if I have a clue, I don't. I managed to place a simulated target response to a proper quadrant, but that's about it. To make it really useful I'd need some real world materials' responses for comparison and tuning. The way it is now I can use for VLFs, but not so much with PI.