Universal PI discrimination module.

[johnyp]

Hello. I have some questions in order to try make an mc controlled universal discrimination module for all PI detectors. I read that a ferous metal near the coil makes the discharge curve to go higher and a non ferous the opposite and change the same time the full discharge time. Is this right ? Should i measure the time that the curve takes to touch the zero or should i go with the voltage in every period/3 with referense the one with no metal near the coil ? What is your opinion or maybe your better proposal ? Thank you.

[johnyp]

Anyone?

[dfbowers]

Anyone?

You may be referring to something like below.

I don't think it's quite that simple or it would have been done long ago.



http://www.lammertbies.nl/electronic..._detector.html

[johnyp]

Yes that is exactly what i had in mind. Why can't my idea work ? I will just divide the decay time in 4 parts, store the voltages in every division when the coil is away from any metals and then i will always compare the voltages against the stored. What do you think ?

[Tinkerer]

Yes that is exactly what i had in mind. Why can't my idea work ? I will just divide the decay time in 4 parts, store the voltages in every division when the coil is away from any metals and then i will always compare the voltages against the stored. What do you think ?

The problem is that the soil response varies from place to place and each target of different shape or metal or orientation will give a different response.

In short, too many variables.

Tinkerer

[homefire]

Here's some more things to think about!

Got this off here someplace.

[johnyp]

Yep... You are right! No can done. But please tell me this: How discrimination is done in an IB MD using the phase comparator method ?

[msysta]

I am thinking this PI issue also. I have 18 bit ad converter and I am thinking of making the analysis very accurate with some MPS level sampling. I see the soil makes it harder but it still is always the same order with the curves, no signal being in the middle in time domain. So why not make that no signal as a reference that is updated for every pulse, thus making it to follow the change in soil? No massive computing needed there.

About the IB machines, it is done the same way; interpreting phase and amplitude variations, phase variations being the key for discrimination. But it's different field, RX coil always produces similar kind of signal per target type, no matter the soil (if not exteremely contamined with black sand or salt).

Here's good project for discriminating IB principle walkthrough:
http://www.geotech1.com/pages/metdet...um/magnum1.pdf

[Tinkerer]

Here's some more things to think about!

Got this off here someplace.

This is a very good experiment, but it needs to be repeated with targets of exactly the same size and shape.
Then it needs to be repeated with different, but equal shapes.

For example, take 10mm diameter spheres each of the metals described above and run the test series.

Then take 20mm diameter disks of the same metals and run the test series.

Then take the same disks and run the series with the disks at 45 degrees and 90 degrees.

Then take a black sand test bes and repeat all the tests above.

Then make a salt test bed and repeat.

Then mix a salt and black sand test bed and repeat.

Then compute all the results and see if you can detect a common denominator.

There is a lot that can be done but it is a lot of work.

Only with a lot of work you will break through the "PI discrimination wall"

Good luck.

Tinkerer

[mikebg]

Yep... You are right! No can done. But please tell me this: How discrimination is done in an IB MD using the phase comparator method ?

Johnip, I know the answer in frequency domain. You should calculate the spectral characteristic of target as system of two functions:
magnitude vs. frequency and phase vs. frequency.
The DSP is described in patent US4,506,225.
Below are given frequency responses of different targets in complex plane. At single frequency metal detectors, the response is represented only by one point in complex plane.

[johnyp]

In general speaking in ferous objects frequency increeses and vice versa. I am just trying to find a way to discriminate between ferous and not with a pi md only cause they have more depth ability. Thank you.

[mikebg]

I think the color legend given in posting #3 is not correct. According signals in complex plane (posting #10), the color legend should seem so:

[Tepco]

Lovely graph, but this will never happen in real world. We are talking about monocoil PI, not to stretch discussion further. Under normal conditions, coils and normal targets, all we can have from coil output is superposition of coil signal (remember that you always have a lump of metal in search coil, well, coil itself) and that from target eddy current decay. And it will never going negative as shown in this graph, no matter target is Fe or non-Fe. So there is no "magic" sampling point that can "discriminate" between materials of different magnetic properties. Only decay rate will depend upon target shape, size and material (conductivity), being fast for small and conductive, and longer for large targets.

(I'm aware of some patents using "passive dumping" for coil etc, i tried different configurations, uS sampling, no magic point observed, only some 40-50uS will make difference point between what you wont to dig or not, see GS and some Bulgarian made detectors.)

For uC controlled PI, there are at least two approaches: One, already known for DIY is taking more than two samples, and calculating decay time using some form of lookup table or something. Good examples are http://www.pulsdetektor.de/, or Russian http://www.fandy.vov.ru/Chance.htm. Later uses 5 samples to make calculated guess of target nature (discrimination next to none, not to mention).

Another approach is to use even more samples. Something like every uS, (or even more, sampling 2 or more pulses), some 64 samples or more, so you need decent 16bit 1MSPS capable ADC and some processing power. Average this, to get rid of noise, and "normalize" to no metal response of coil. And you will find that decay is not strictly exponential, one part of decay is faster or slower, and there is "intercept" point whit coil decay in most cases. Whit some cleaver software and more data available (intercept time and total decay duration) you can decide target type more closely.

And not to mix up this whit Minelab, or VLF machines, or anything else that use balanced coils. In those, system of balanced coils actually provide discrimination, signal from coil disbalance always go to "right" side for Fe or non-Fe, machine just interpret this signal,one way or another. Monocoil PI so far at best can discriminate target decay time, and this is, more or less, indication of target size, not magnetic properties.

(And again, sorry if i'm misfired something here, this is my 2¥)

[mikebg]

SIGNAL ELIMINATION WITH INTEGRATING WINDOW
There is an old PI method to suppress signal from ferrous conductor (iron or mineralized soil) with difference integrator. Integrating window is shifted so that unwanted signal in output of integrator becomes zero.
The principle is illustrated in the figure with colored areas. Between moment t1 and moment tc the input signal for integrator is positive. Then the output of integrator represented as red area is positive. Between moment tc and t2 the input signal is negative and output represented as blue area is negative. When blue area is equal to red area, the output of difference integrator remains zero.

[Tepco]

Nice for low conductivity magnetic material (mineralized soil), when eddy current is negligible, but material is still capable of storing some energy in it's magnetic field (GS use this among others). But for more conductive ferromagnetic targets, whit significant eddy current present, bottlecaps etc will be of little use. Adjusting time between samples, or using two channels whit different delay and compare output will easily achieve this kind of signal elimination. But this is still way too far from true Fe-nonFe discrimination.