Deconvoluting Metal Detector

[G-set]

Hello, I heard of this detector called deconvoluting metal detector and ask how this detector works. I look here for info but can not find it. Is deconvoluted metal detector different to normal VLF or PI detector, and where can I learn about deconvoluted metal detector? Thankyou.

[Qiaozhi]

Hello, I heard of this detector called deconvoluting metal detector and ask how this detector works. I look here for info but can not find it. Is deconvoluted metal detector different to normal VLF or PI detector, and where can I learn about deconvoluted metal detector? Thankyou.

I've never heard of a detector that uses deconvolution directly.

Perhaps you're getting confused with an image processing technique that uses "deconvolution" to process images in the lab.
http://www.ndt.net/article/ecndt2006/doc/Tu.4.5.4.pdf

[Skippy]

I recall this subject has come up before on here. The suggestion is that when two targets are below the search-coil, the response is not simply the mathematical sum (addition) of the two responses, but a more complicated multiplication function, possibly with some non-linear element to it. Sort of like the difference between a mixer circuit, and a summing circuit. So to better seperate out multiple close targets, deconvolution would promise more than simple differencing/subtraction methods. It would also seem reasonable to suggest that ground signal gets convolved with target signal, too, so seperating a weak target from fluctuating ground level just might benefit from this technique.

Whether you can do it on a continuous basis, or if it only works best on a fixed set of data, I don't know.

[Silver Dollar]

A couple of things I found in a search showed some good results but said it was for lab use only so far.
Some of the mining guys are looking into this. Not sure why you cannot use it in the field? Processor too
slow? Need multiple data lines?

[Davor]

mikebg often mentioned deconvolution as a next big thing. Maybe it is his design.

[mikebg]

mikebg often mentioned deconvolution as a next big thing. Maybe it is his design.

No, the "thing" is not next. Deconvolution is quite old math operation used in geophysics for decades as method for target identification. Frequency domain was used to simplify calculation of “convolution” and “deconvolution” - integrating math operations in Time domain. However calculation is not problem at DSP.
DECONV is ь standard math operatioon in DSP. The design of deconvoluting metal detector is not mine. Only a search head with additional winding was designed by me because we need a kernel signal to calculate deconvolution.

Hello everyone!
I believe that many participants in this forum can become authors in writing a textbook for amateur designers entitled “DECONVOLUTING METAL DETECTOR”. The predominant part of information that will contain the future textbook is posted in thread

http://www.geotech1.com/forums/showt...Balance-Theory

The rest part of information for convolution and deconvolution is not a problem, because several posts contain links to it.
I have an idea for front cover of the textbook. It will contain two pictures. The first one is a glass of red wine with two different coins on the bottom. Convolution makes them red and dificult visible.
The second picture shows the first picture after deconvolution. Deconvolution makes the wine colorless. A silver dollar and a gold dollar are on the bottom, but the environment is colored in complement color.

While I do the front cover, you can make suggestions for table of content.
I expect your suggestions for the first page of the textbook which is a table of contents.

[G-set]

Hey MikeBG,


To design a modern metal detector, before everything we need to know more about how the received signals are created OUTSIDE THE SEARCH HEAD and why the ground makes aconvolution of the target signal. If we know that, we will be able to design the most suitable signal processing and deconvolution INSIDE THE METAL DETECTOR.


So I think that amateur designers like me need a textbook entitled OUTSIDE THE METAL DETECTORS. Please give a draft (your suggestion) for the contents of such a book and we will discuss it here in the forum.

[Tinkerer]

Hey MikeBG,


To design a modern metal detector, before everything we need to know more about how the received signals are created OUTSIDE THE SEARCH HEAD and why the ground makes aconvolution of the target signal. If we know that, we will be able to design the most suitable signal processing and deconvolution INSIDE THE METAL DETECTOR.


So I think that amateur designers like me need a textbook entitled OUTSIDE THE METAL DETECTORS. Please give a draft (your suggestion) for the contents of such a book and we will discuss it here in the forum.

The soil/ground is a very large target. It has conductivity, mostly caused by salts, that changes locally. This conductivity gives it properties not unlike metal. It has a skin effect too. So it also has a time constant TC.
The general soil matrix often has inclusions like pockets of soil with different characteristics than the surrounding matrix.

In the general conductive soil matrix there are also patches and areas with a great variety of magnetically susceptible particles.

To really know how all this affects a metal detector we need to build a detector that is capable of measuring/quantifying all these different effects.

[Davor]

Magnetic susceptibility and conductivity of the soil are mutually orthogonal phenomena. Soil clusters and various targets lie next to each other. Convolution happens as a part of our sampling process, so their responses become intertwined. Multiple targets and soil matrix get blurred together.
Deconvolution is also happening during the signal processing (of any kind) in our detectors, and more importantly our brains. The real advantage of any full deconvolution approach would be in graphical representation of detected grounds. Otherwise our brains may do much better than that.
Deconvolution, in optimistic scenario, should separate the two (or more) by some practical means. Small trouble is that the noise is also there. Once the toothpaste is out of the tube, it's hard to get it back in without any dirt or losses.
I'm curious as the next guy what principle is behind this story, and what method is going to be applied.

[mikebg]

FRONT COVER and INTRODUCTION PAGE

Here's a proposal for front cover of the future textbook and a draft for introduction page:

INTRODUCTION
„Deconvolution is nearly impossible to understand in the time domain, but quite straightforward in the frequency domain.“ - Steven W. Smith in
http://www.dspguide.com/ch17/2.htm

This textbook is intended to amateur designers. We - the authors of the textbook, are also amateurs. We tried to simplify and to visualize the explanations so that even professionals to understand the learning material.

The color is a concept that belongs to the frequency domain, so let's use a color photograph to explain convolution of target signal with the ground signal. We wll use other color picture to explain what can be achieved by deconvolution. The front cover contains such pictures to illustrate convolution and deconvolution.

The left-hand photograph is a glass of red wine. Two different coins are on the bottom. The coins are almost invisible when the glass is lit from above. The spectral transparence of red wine makes convolution with spectral response of each submersed coin. If the illumination is from the top, the reflected from coins luminous flux is passed twice through the layer of red wine - once reaching the coins, and then returning to the observer. We see that the red wine makes the color of both coins to seem dark red and even makes them dificult visible.

The right-hand picture shows the left one after deconvolution using red color as kernel for filtration. The software makes the red wine to seem colorless. Note that the environment appears in irreal colors, but this is not important because we see true color of artefacts on the bottom. There are a gold coin and a silver coin.

Convolution occurs at metal detecting. The electromagnetic energy reflected from buried target has passed twice through a layer of ground. The transfer function of ground makes twice convolution with reflecting function of target. If the target is deep buried or when the soil is “bad”, the conventional metal detector can not identify the target or even can not detect it despite the “ultimate ground balance”.

Main advantages of a metal detector with deconvolution are greater depth of detection and correct identification of deep buried targets. Of course, the deconvolution will not get the opportunities which the detector shows at AIR TEST, but it will be closer to them.

Another advantage of deconvolution is that it can sharpen the blur images. This is applicable in metal detector with relative large search loop. Using suitable kernel, we can make it to detect small targets and to pinpoint them as if it operates with a small loop.

[mikebg]

.....So I think that amateur designers like me need a textbook entitled OUTSIDE THE METAL DETECTORS. Please give a draft (your suggestion) for the contents of such a book and we will discuss it here in the forum.

Hi G-set.
I think you are not an amateur. Your ugly avatar seems as a thug - citizen of the Isis (Islamic state). However, I guess you're Chinese and employee of CLU (Chinelab Unlimited:-). Perhaps your alias "G-set" means GROUND BALANCE in Chinglish language. For needs of your company, let’s make OUTSIDE THE SEARCH HEAD as first chapter of the future textbook on deconvolution in metal detectors.
If you really do not have enough knowledge on signal processing, I recommend you to visit
http://terpconnect.umd.edu/~toh/spectrum/TOC.html
Downlad a file and find the phrase “buried treasure” in it. It is Appendix C on page 114.

[G-set]

Hi G-set.
I think you are not an amateur. Your ugly avatar seems as a thug - citizen of the Isis (Islamic state). However, I guess you're Chinese and employee of CLU (Chinelab Unlimited:-). Perhaps your alias "G-set" means GROUND BALANCE in Chinglish language. For needs of your company, let’s make OUTSIDE THE SEARCH HEAD as first chapter of the future textbook on deconvolution in metal detectors.
If you really do not have enough knowledge on signal processing, I recommend you to visit
http://terpconnect.umd.edu/~toh/spectrum/TOC.html
Downlad a file and find the phrase “buried treasure” in it. It is Appendix C on page 114.

Thanks Mike. Mike, I’m not a Chinese, but I think this is a Chinese forum because an admin has Chinese avatar and Chinese alias. But serious, despite my ugly avatar, this is a great forum! If you have table of contents for the future textbook on deconvolution, please post it. Then I will know what to read in the Web and what to ask in the forum. Thanks in advance.”

[mikebg]

TABLE OF CONTENTS (Draft for collaboration)
0. Introduction
0.1. Target inside the matrix. Convolution
0.2. Deconvolution. What to read.
0.3. What to experiment
0.4. Discrimination or identification?

1: Outside the search head
1.1. Transfer function of mutual inductance and self-inductance
1.2. Transfer function of ferromagnetism
1.3. Transfer function of conductivity
1.4. Harmonic analysis of eddy currents
1.4.1. Timeconstants and specific frequencies of relatively small target
1.4.2 Timeconstants and specific frequencies of conductive halfspace
1.5. Transfer functions of conductive ferromagnetism. Resonance frequency.
1.6. Convolution of target signal with transfer functions of matrix
1.6.1. Target in conductive nonmagnetic halfspace
1.6.2. Target in ferromagnetic nonconductive halfspace
1.6.3. Target in conductive ferromagnetic halfspace
1.7. What depends on diameter of TX loop

2: Inside the search head
2.1. Modulation index
2.2. Convolution with transfer function of search head
2.3. The optimal excitation
2.4. TX modulation by ground
2.5. Passive metal detector. Suppression of AM.
2.6. Convolution with transfer function of shielding
2.7. Loop configurations for suppression of ground signal
2.8. Loop configurations for deconvolution
2.9. Search head for passive metal detecting

3: Inside the metal detector
3.1. Synchronous demodulation of convoluted signal
3.2. Deconvoluting kernel, apperture function and point spectrum function
3.3. Blind deconvolution
3.4. Block diagram of metal detector with deconvolution
3.4.1. Dynamic compensator
3.4.2. Narrow band deconvolutor
3.4.3. Wide band deconvolutor
3.5. Software for deconvolution
3.6. Noise at deconvolution

4: Q & A
4.1. Deconvolution vs Ground balance
4.2. Spectrum of excitation.
4.3. Automatic induction balance
4.4. “Color” coefficients of convoluted target signal
4.5. S/N ratio and modulation index at deconvolution
4.6. Patents for deconvolution in metal detectors
4.7. Terminology: “deconvoluted” or “deconvolved”
4.8. Caves, voids, hot rocks and meteorites

5: Deconvolution in GPR

6: Design of DECON metal detector
6.1. Design of search head for deconvolution
6.2. Narrow band DECON detector
6.2.1. Block diagram
6.2.2. Circuit diagram
6.3. Wide band DECON detector
6.3.1. Block diagram
6.3.2. Circuit diagram
6.4. Display and controls

7: Battery
7.1. NiMH or Li-ion cells?
7.2. Supervisor for discharge of each Li-ion cell
7.3. Individual charging of each Li-ion cell

8: References

[WM6]

Hi mikebg

You deconvolute to much. Can we shrink to block scheme?

[6666]

Hi MIkebg are you going to publish a book ?
If so when will it be available and what format ?
Thanks