DEEPER PI DETECTION DEPTH

[Tinkerer]

Sounds great.

Probably hard to get hands on nice samples. I'd prefer thick sheets of copper, aluminum, zinc or brass, tin, and maybe lead. The more the merrier. Iron would be interesting comparison.

For my interests, I'd start with coin thickness if possible. Cut in squares say 1 x 1 cm, 2 x 2 cm, 4 x 4 cm, 8 x 8 cm, etc.

Then cut circular ones with same area for comparison.

Then we try different thickness test. Maybe stacking tight would be fair test, not sure. Otherwise, we need a blob of metal -- got to simulate nuggets! If you have a nugget, all the better.

Photos of analog scope would probably be fine if you can get them nice and clear.

Seems we need to see some real data at this point.

Regards,

-SB

About stacking tight:

This is how transformer iron cores are made. The transformer core is made up with many thin sheets of silicon steel, with a thin dielectric coating between. The coating does not actually isolate the layers, it just makes a bad contact. The reason is to contain the eddy currents within the individual sheets. Sometimes we see such a transformer core welded on one edge. This does not seem to cause a big difference with the eddy currents.

If we compare a tightly stacked pack of aluminium foil, we have a similar occurrence. Aluminium has a thin dielectric oxide coating. Although the aluminium oxide is a very good dielectric, because it is so thin, it gets easily scratched or abraded, for example with the stress of being tightly stacked. But, any contact with air generates a new oxide layer.

So what kind of target do we have with a tight stack of aluminum? We have some continuity, but we also have individually insulated sheets.

Do the eddy currents in this stack add up?
Do the eddy currents partially cancel each other?
Anyway, this is not a usual target, we better use test targets that are more similar to the targets we are searching for.

The targets that we are searching for, have a very great variety. If we can define a few categories and then classify the targets into these categories, we already have taken a great step forward in the endeavor to ID the targets.

So let's look for a answer to a first question: What percentage of the target response is generated by the target surface area.

Tinkerer

[simonbaker]

This could be interesting.

Some time ago I made tests with thickness of targets. I wanted to find out if I can design and manipulate a TX pulse that will give me information about the thickness of a target. http://www.geotech1.com/forums/showp...5&postcount=26

The copper targets were of the same diameter and presented the same way to the coil. Only the thickness changed. Well, almost. When you see the response of the copper coin, you notice that it does not respond the same way.
So there is copper and copper.
The thicker targets were all cut from the same rod. (actually a "pin" from an old 18th century, shipwreck. The purity of the copper is not known).

I still have these targets and could re-do a series of tests.

First I would like to know about the skin effect on the thickness of the targets. Does the slope of the switch off transient have an influence on the response of targets of different thickness?

Then I would like to know what part of the response is due to the surface area presented to the coil.

There are quite a few other questions that we could explore, using the same targets.

Once we have all the questions answered, we can make a similar target test series out of a different metal to see what differences we find.

As far as the digital scope goes, I need to repair it. If I manage to get it working again, we are in business.

Tinkerer

Sounds great.

Probably hard to get hands on nice samples. I'd prefer thick sheets of copper, aluminum, zinc or brass, tin, and maybe lead. The more the merrier. Iron would be interesting comparison.

For my interests, I'd start with coin thickness if possible. Cut in squares say 1 x 1 cm, 2 x 2 cm, 4 x 4 cm, 8 x 8 cm, etc.

Then cut circular ones with same area for comparison.

Then we try different thickness test. Maybe stacking tight would be fair test, not sure. Otherwise, we need a blob of metal -- got to simulate nuggets! If you have a nugget, all the better.

Photos of analog scope would probably be fine if you can get them nice and clear.

Seems we need to see some real data at this point.

Regards,

-SB

[Tinkerer]

It's fun to try to read your mind Aziz, but I thought maybe I'd start on something simpler like deriving Maxwell's equations from scratch...

Actually there is an important issue to resolve first... Qiaozhi and others say that the target TC (time constant) depends both on target metal (conductivity) and target shape/size.

If that is the case, then TC is useless for discriminating metals. So all the graphs and processing in the world that enable us to estimate a target TC won't give us what we want (unless you want a target TC ).

There is an exception: if we know the soil only contains coins, and they are all similar size, then the size/shape parameter is roughly constant and the TC will be adequate to discriminate metals.

I have never seen such a world -- mine is full of soda cans buried a foot deep and old roofing nails and bits of power cable and picnic foil.

I would however like to analyze further the dependence of TC on size, shape, and metal. I have a feeling there is some data on this forum. However, it is always worth repeating tests. Maybe Tinkerer could take a whack at it. I would propose say 3 metals that are available in 3 different sizes and maybe 2 different thicknesses.

Then use a PI to record the target response, fairly close so it is large and clear (Tinkerer's balanced coil might help). Of course a digital storage scope would be dandy!

I'd like to see the data.

-SB

This could be interesting.

Some time ago I made tests with thickness of targets. I wanted to find out if I can design and manipulate a TX pulse that will give me information about the thickness of a target. http://www.geotech1.com/forums/showp...5&postcount=26

The copper targets were of the same diameter and presented the same way to the coil. Only the thickness changed. Well, almost. When you see the response of the copper coin, you notice that it does not respond the same way.
So there is copper and copper.
The thicker targets were all cut from the same rod. (actually a "pin" from an old 18th century, shipwreck. The purity of the copper is not known).

I still have these targets and could re-do a series of tests.

First I would like to know about the skin effect on the thickness of the targets. Does the slope of the switch off transient have an influence on the response of targets of different thickness?

Then I would like to know what part of the response is due to the surface area presented to the coil.

There are quite a few other questions that we could explore, using the same targets.

Once we have all the questions answered, we can make a similar target test series out of a different metal to see what differences we find.

As far as the digital scope goes, I need to repair it. If I manage to get it working again, we are in business.

Tinkerer

[simonbaker]

Hi all,

have anyone had a look at the AC analysis response yet? Frightened? Shocked?
Or are you all after the "free lunch"?

But before you get some "free lunch", you all have to work a little bit.
So, what the f...ing h.ll the simple AC analysis response says?
Comeon, it isn't much difficult to interpret the response.

Aziz

It's fun to try to read your mind Aziz, but I thought maybe I'd start on something simpler like deriving Maxwell's equations from scratch...

Actually there is an important issue to resolve first... Qiaozhi and others say that the target TC (time constant) depends both on target metal (conductivity) and target shape/size.

If that is the case, then TC is useless for discriminating metals. So all the graphs and processing in the world that enable us to estimate a target TC won't give us what we want (unless you want a target TC ).

There is an exception: if we know the soil only contains coins, and they are all similar size, then the size/shape parameter is roughly constant and the TC will be adequate to discriminate metals.

I have never seen such a world -- mine is full of soda cans buried a foot deep and old roofing nails and bits of power cable and picnic foil.

I would however like to analyze further the dependence of TC on size, shape, and metal. I have a feeling there is some data on this forum. However, it is always worth repeating tests. Maybe Tinkerer could take a whack at it. I would propose say 3 metals that are available in 3 different sizes and maybe 2 different thicknesses.

Then use a PI to record the target response, fairly close so it is large and clear (Tinkerer's balanced coil might help). Of course a digital storage scope would be dandy!

I'd like to see the data.

-SB

[Aziz]

Hi all,

have anyone had a look at the AC analysis response yet? Frightened? Shocked?
Or are you all after the "free lunch"?

But before you get some "free lunch", you all have to work a little bit.
So, what the f...ing h.ll the simple AC analysis response says?
Comeon, it isn't much difficult to interpret the response.

Aziz

[Qiaozhi]

Using Spice to design and test radical new circuits is hit-or-miss... you can easily get sim results that make you drool, followed by a built circuit that makes you cry.

That is so true!

After the crying stops though, it often leads to a better understanding of the design.

[Aziz]

Aziz: why don't you just go ahead and say what's on your mind? We're all ears...

-SB

That would mean a lot of work for me to explain it.
The learning effect will be less too.

Ok, more food:

Look at the following items in the Bode plot (magnitude only):
- all different target responses tend to go to the same spectral response energy density level (asymptote)
- every different target response starts increasing response with different starting frequency
- the upper limit frequency (roll off barely visible)
- the response surface area for the largest target TC response (integrate spectral energy density over frequency)

We aren't going to modify the initial AC analysis spice simulation until we know some fundamental basics.

Cheers,
Aziz

[Carl-NC]

Spice is a double-edged sword. You really gotta know what you're doing to use it effectively.

I was in a hurry, so lemme 'splain better...

Spice is as good as the models you use, and the circuit you design, and the sim method you choose. A little wrong on any of these can give you a lot of wrong in the sim results. So to effectively use Spice, especially in complex circuits, you really need to know circuit design.

I remember in college we had to design an audio amp, run it in Spice, then build & test it. I did this elaborate fully complementary design that worked wonderfully in Spice, but not when built. My first lesson in statistical component variation. Then, for kicks, I designed a model train controller which looked good in Spice, but promptly burned up my F9 Diesel and then insulted me further with a literal flare shooting out of the power MOSFET. I still own the Ace breadboard with a melted spot on it. My first lesson in modeling inductive loads.

In chip design we simulated the hell out of circuits, because the cost of making circuit changes was about $25,000. I got pretty good with Spice and Spectre RF. Model accuracy was super critical and a lot of manpower was put into characterization and modeling.

In detector design, I tend to use Spice in 3 ways:

1, as a tool to look at difficult-to-measure signals, like transient gate currents.
2, as a quick check on mundane things like filter responses.
3, as a 'what-if' tool, making changes to circuits I know work.

I've done the transformer target modeling thang but don't care for it, so when I need target results I build the durn circuit and use real targets. Using Spice to design and test radical new circuits is hit-or-miss... you can easily get sim results that make you drool, followed by a built circuit that makes you cry.

[bbsailor]

Spice is a double-edged sword. You really gotta know what you're doing to use it effectively.



Well, that's no reason to quit posting! I enjoy your participation, please reconsider.



According to Eric, the 70-pound silver bars from the Atocha have a TC of 8ms.



Target TC is heavily dependent on thickness due to skin effect.

- Carl

Carl, and others

Eric Foster posted this little gem a while back on his forum. I have taken the liberty of putting in bold Eric's relevant comment about the relationship of the target TC and the coil discharge TC or ramp down slope. If a target has a 5us TC than the coil discharge TC should be 1us to obtain 90 percent of the potential target signal. That means that the Coil discharge TC should be at least 1us which is easy to obtain with a 300 uH coil and at least a 300 ohm damping resistor. But wait there is more.

The coil discharges in three stages.

Stage 1. The flyback spike is clamped by the MOSFET voltage rating
Stage 2. The Input resistor and two clamping diodes on the first amplifier input stage are in parallel with the damping resistor while the diodes are conducting down to about 0.7V and this is equivalent to 231 ohms (300 ohms in parallel with 1000 ohms or the input resistor)
Stage 3. Only thereafter (below 0.7V) is the damping resistor value of 300 ohms contributing to the final coil discharge slope by itself.

Then we must wait for the op amp to come out of saturation before the earliest possible sampling can occur.

If accurate modeling is to occur, as Carl suggests, then you got to know what you are putting into the model and that includes knowing the actual coil discharge characterics as it relates to the actual target TC. Remember that during the delay time the eddy currents in the target are beginning to ramp down and you need to sample early enough to catch them before they reach the noise floor. Large targets are no problem as their TCs last well into the millisecond range but small gold targets are what people want and each is a different size, metal consistancy and shape. As they get smaller with lower TCs, the coil design, MOSFET capacitance, coil coax cable length and capacitance and total circut capacitance governs the actual discharge slope of the coil. That slope defines the smallest target that you are going to detect. Modeling should be able to tell you what that minimum target TC size is with known circuit components and simple measurements to confirm the model on a working PI circuit and coil. Small gold link chains are good low TC targets.

See how Eric puts this into perspective below.

Thanks

bbsailor

PS Sean, Keep on posting. More keen minds are better!

2010 Lesson 1 - Flyback
Posted by: Eric Foster
Date: January 01, 2010 05:46AM

A common misconception is that the voltage flyback pulse has a direct bearing on the signal from a metal target. PI theory shows that it doesn't. The voltage pulse occurs simply because the coil becomes almost open circuit at the end of the TX drive and the collapsing magnetic field tries to maintain the current at the level it was just before cut off. This voltage rises till it exceeds the avalanche voltage of the Mosfet and then stays at this level until the magnetic energy has dissipated such that the voltage falls below the avalanche rating, and then the remaining energy is absorbed by the damping resistor. If you insert a small resistor in the ground end of the coil, the coil current waveform can be observed, which exactly matches the magnetic field waveform. There is no spike, but just a smooth ramp down of the current and field. Now, the ramp down is important in that it must be significantly faster than the target time constant, otherwise signal will be lost. Theory states that provided the ramp down is faster than 1/10th of the object TC, then it is as if the field were removed instantaneously. So, objects with a short TC need a fast ramp down and objects with a long TC can get away with a slower ramp down. Also, for maximum signal you have to factor in the duration of the TX pulse. This too has to be significantly different to the object TC to achieve maximum signal, in this case longer. Again the factor of 10 comes in, but having it 5x longer will still get you around 90% of the signal. Looking at the collapsing magnetic field from the targets viewpoint, because there is also a small inductance and parallel resistance involved, an induced voltage spike like a micro flyback occurs. It is this induced emf that drives the eddy currents until they are dissipated in the internal resistance in the target.

Eric.

[Carl-NC]

Anyway I'm starting to realize the danger of simulations... you can spend hours tweaking them never really learning anything with confidence.

Spice is a double-edged sword. You really gotta know what you're doing to use it effectively.

In light of the offence I have cause, this will be the LAST post I shall me making on this forum!

Well, that's no reason to quit posting! I enjoy your participation, please reconsider.

In AZIZ's simulation he use targets with time constants of 1us, 10us, 100us and 1ms. These all seem very reasonable values to use, except perhaps the 1ms target.

According to Eric, the 70-pound silver bars from the Atocha have a TC of 8ms.

Why would metals with a conductivity range of 1:7 have orders of magnitude different time constants?

Target TC is heavily dependent on thickness due to skin effect.

- Carl

[simonbaker]

To find the time constant, simply divide L by R.

I can see what you were trying to do in your example, but you cannot replace R by the resistivity of the material. For one thing, resistivity is measured in ohm-meters, and for copper (for example) the resistivity is 1.68 x 10E-8, not 0.17. Also, the time constant of a target depends very much on its size, shape and orientation, and not just its resistivity.

All you can really do is use a set of targets with various time constants, as Aziz did in his simulation. Trying to define a target for a particular material is not possible, otherwise you could use this method to provide a PI with target discrimination, which you cannot.

Ok. I wasn't trying to calculate the actual resistance, just make relative targets (1.68 x 10E-8 is ohms per something).

But if there is data that shows that range of "time constants" is representative, that's what we have to deal with.

However, how do we relate the time constant of a target to its shape and metal? We need some kind of relationship for Aziz's graphs to be useful.

-SB

[Qiaozhi]

Yes, I just took a wild stab, I modified one of Aziz's target resistors slightly and made the others proportional.

Why would metals with a conductivity range of 1:7 have orders of magnitude different time constants?

I need education on target modeling. How do you come up with those TCs?

-SB

To find the time constant, simply divide L by R.

I can see what you were trying to do in your example, but you cannot replace R by the resistivity of the material. For one thing, resistivity is measured in ohm-meters, and for copper (for example) the resistivity is 1.68 x 10E-8, not 0.17. Also, the time constant of a target depends very much on its size, shape and orientation, and not just its resistivity.

All you can really do is use a set of targets with various time constants, as Aziz did in his simulation. Trying to define a target for a particular material is not possible, otherwise you could use this method to provide a PI with target discrimination, which you cannot.

[simonbaker]

Hi SB,

sorry, but you are onto wrong path (yet again). Don't go too fast. Just step by step. We want to reveal the basic fundamentals only. And we aren't doing any ground balance nor discrimination here.

What kind of information can we get from the response in the given configuration:
- in regards to the spectral energy of the response (the maximum possible response information)?
- in regards to the bandwidth of the response?
- in regards to target response of different time constants targets?
- in regards to selecting the TX pulse frequency?
(just to list some basic issues here)
...
(we can line up more questions here of course)

Don't think too complex. Just look at the Bode plot. You even do not need to look at the phase information.

Cheers,
Aziz

PS:
Thanks for clarifying this Qiaozhi.
Sean, don't go please!
Glad to see Dave here.

Aziz: why don't you just go ahead and say what's on your mind? We're all ears...

-SB

[simonbaker]

Simon - In AZIZ's simulation he use targets with time constants of 1us, 10us, 100us and 1ms. These all seem very reasonable values to use, except perhaps the 1ms target.
However, in your version you have the following:
silver = 625ns
copper = 588ns
gold = 435ns
aluminium = 370ns
zinc = 167ns

These all seem to be far too short.
Remember that TC = L/R.

Yes, I just took a wild stab, I modified one of Aziz's target resistors slightly and made the others proportional.

Why would metals with a conductivity range of 1:7 have orders of magnitude different time constants?

I need education on target modeling. How do you come up with those TCs?

-SB

[Aziz]

Ok, I'll bite...

I guess my first impression is that the spectral shape is very sensitive to target time constant but not too sensitive to RX/TX coil inductance, so you can use that information to differentiate targets.

However, you have pretty radical differences in target resistance. Is that realistic?

For fun, I tried substituting actual metal conductivity resistances for the target resistor values. I have no idea if that is valid. So let me know if that is meaningless or not.

Below is a comparison of the RX response for those targets.

Does it mean anything, or just BS?

-SB

P.S. Gee, the phases look nice and differentiated above 100 kHz -- maybe some sort of phase detector would be useful... what does that remind me of...

Hi SB,

sorry, but you are onto wrong path (yet again). Don't go too fast. Just step by step. We want to reveal the basic fundamentals only. And we aren't doing any ground balance nor discrimination here.

What kind of information can we get from the response in the given configuration:
- in regards to the spectral energy of the response (the maximum possible response information)?
- in regards to the bandwidth of the response?
- in regards to target response of different time constants targets?
- in regards to selecting the TX pulse frequency?
(just to list some basic issues here)
...
(we can line up more questions here of course)

Don't think too complex. Just look at the Bode plot. You even do not need to look at the phase information.

Cheers,
Aziz

PS:
Thanks for clarifying this Qiaozhi.
Sean, don't go please!
Glad to see Dave here.