Taking a look at GROUND BALANCE on the bench

[Tinkerer]

The "time-domain" vs "frequency-domain" debate is, IMO, much ado about nothing. Metal detectors are inherently time domain processors and I tend to do my design thinking (and simulations) in the time domain. But FD is also useful in some cases, such as the fundamental frequency response of targets. But once you've selected the frequency you want to use, it's TD pretty much to the end.

In my former life designing opamps, I primarily used FD sims for cases when the opamp was in a happy linear state; but when it came time to analyze large signal, overvoltage, and slew performance, an FD sim was pretty useless (I designed some fantastic opamps in the FD, that were total flops in the TD!). FD simulation is pretty useless for VCOs, PLLs, SHAs, and mixers. The point is, use the right tool for the right job.

Also, in traditional PI a single sample doesn't tell you squat about the tau of the target. It doesn't matter whether you sample at 5us or 50us, you have no idea what the heck is in the ground, except that at 50us it probably isn't a sub-gram nugget. Theoretically, you can take 2 or more sample points and get a good idea about tau. One of my engineers built up a bench test using Labview and got a PI system with target ID working. And it worked pretty good, until you added ground, and then it went to hell in a handbasket. Like my opamp designs, a lot of ideas work great, until...

Yes, I agree, the ground is the problem. This is why I started this discussion, to see if our joint brains can find a solution.

First, we need to agree if the above very hot ground simulation is good enough to be used as a base for the discussion. I showed the ground and several different targets together so that we can decide on that.

I can pump the coil above this ground to see if a Ground Balance scheme is working.

If we can agree, then we can move to the next step: what could be tried with this setup, to make a Ground Balance.
I have a few ideas, I am sure others have probably better ideas. We can try them and see what works.

Tinkerer

[Qiaozhi]

Qiaozhi, your alias shows that you are a legal agent of CHINELAB Unlimited in this forum.
I know that your company needs cicuit diagrams for free, but now we need money for something serious.
Let your company buy a PDF file for $2:

Very funny, but your assumptions are incorrect.
You need to read some of the other parts of Geotech, such as here, for example ->
www.geotech1.com/itmd
plus clues scattered throughout the other forums.

In my former life designing opamps, I primarily used FD sims for cases when the opamp was in a happy linear state; but when it came time to analyze large signal, overvoltage, and slew performance, an FD sim was pretty useless (I designed some fantastic opamps in the FD, that were total flops in the TD!). FD simulation is pretty useless for VCOs, PLLs, SHAs, and mixers. The point is, use the right tool for the right job.

I tried to explain this in a different thread, but it was conveniently ignored. No doubt the above statements will also fall on stony ground.
So, repeating myself once again for the benefit of MikeBG:
SPICE frequency analysis has another name ... it is also called "small signal analysis". This is because the models are linearised around the current operating point prior to the start of the simulation, and is the reason why FD simulations always run fast. Note the words "small signal". Large signal simulations need to be done in the time domain. For RF design you can use a shooting Newton algorithm and/or harmonic balance, but both of these methods mainly exist so that you get to see the simulation results before you reach retirement. However, both of these techniques are only really useful in certain specific cases, and are certainly not applicable to PI design.

[Midas]

Here you go Tinkerer,
Ground is original green
Ground + ferrite is purple
Ground + foil is blue
Ground + silver dollar is dull red
All side effects of some fancy filters and gradient stuff. Not perfect but I've already spent too long on it..

[Davor]

Davor, this one is for you. Could you help me measuring the time of when this target signal merges with the Ground signal? Maybe just using a comparator to start with?

First thing's first. There is a high probability that either ground or the target tau shall meet far beyond the pulse duration, hence you can't count on a far end junction. But there is your pivot at the beginning.

Your diagrams are showing a pivot in all it's glory. It is obvious that the pivot will become some kind of ... a pivot. Next thing is to establish a tau relationship. Again, it is obvious that shorter taus respond sooner and the first up-going hump is more prominent for them, also the second down going hump is more prominent for the longer taus. But up to the pivot these curves go along. Trouble is a reference that is prone to offset errors. A whole curve, or infinite number of curves would have a nett value of zero, or the integral of the whole shebang would ideally be zero. I'd suggest you to make an integrating servo to keep the average of the signal at precisely zero.

Some early CRO shots show the translation of the long tail up or down with the applied target. To remove ground, I'd play with the relationship of zero crossings past the pivot. Establishing a zero crossing is easy by balancing the sample before and after the actual zero crossing, but to do that I'd need to remove offset, hence an integrating servo. Once in action, the zero crossings can self-balance to the ground, disbalance proportion says target tau, and amplitude it's size/proximity.

It is also possible to integrate windows between the zero crossings to establish amplitude more directly.

Please note that such ground balancing will not be able to establish tau directly as a relationship to ground balanced zero crossing on difficult terrains. Direct relationship could be possible only in air tests. Some further refinement will be necessary for target ID on difficult ground.

[Davor]

When I think about it, with 10,000 pulses per second, zero crossings are easily obtained by a mere sign() function e.g. a comparator. Pulses obtained that way can be balanced with some PLL-like mechanism (correlating result with prediction, correcting prediction in a loop).

[Ferric Toes]

Important things to know for the GB designer.

Hi all,

I've been away in France for a week in a place with no internet access, so it is good to see that there has been a lot of activity, and particularly a thread for ground balance has been started. Soil magnetism is one of my pet subjects and how to balance out its effects in a metal detector is a fascinating offshoot.
The reason that many ground types and rocks give a signal is a bit more complicated than one might first think, and it is important to get a bit of basic understanding if one is going to design an effective electronic GB system. A good starting point is to read the Hitchhiker's Guide to Magnetism which is available as a free .pdf download. The factors which come into play with soils and rocks are the minerals haematite, magnetite, maghemite, and their respective grain sizes. Alignment of domains, remanent magnetism, and conductivity have been mentioned in the above posts but are not the primary cause of the ground response. The primary culprits are superparamagnetic grains of <30um diameter. It is these that give rise to the 1/t^-1 decaying signal that we see in the offtime of a conventional PI. These grains are too small to accomodate a domain and there is no remanence or coercivity and they have an extremely thin hysteresis loop. In response to the applied field there will be a net statistical alignment of SPM grains with the foregoing growth and relaxation times. With larger nanometer size grains, single and multi domains can be supported which exhibit remanence and coercivity.

Australian rock, and housebrick, contains an assemblage of different grain sizes, so they can exhibit different behaviours side by side i.e. my Oz test rock has high susceptibility, high viscosity, and will strongly repel a compass needle. A housebrick, or piece of earthenware pottery, can do the same as a result of firing in a kiln and cooling within the earth's field. Larger multidomain grains get locked in, so to speak, and a net permanent field is created, but this does not affect the (important for GB) SPM grains.

The great thing is, that unlike metal objects, the decay law is practically the same for Oz rock, fired brick and pottery, volcanic rock, and much inland soil; in any quantity. It is only the amplitude that varies. Metal objects, ferrous and non-ferrous are all over the place in decay law and amplitude.

Ferric (not Ferrous). The Hitchhiker's guide will tell you why.

[Qiaozhi]

A good starting point is to read the Hitchhiker's Guide to Magnetism which is available as a free .pdf download.

To save everyone searching for it, here is the PDF file that Eric mentioned ->
http://www.irm.umn.edu/hg2m/hg2m.pdf

[bklein]

A few weeks ago I went on a planned outing to a running mining operation in Randsburg, CA. They enlightened me in a way that I hadn't thought about prospecting. They called themselves "black sand prospectors" not "gold prospectors". Find the black sand concentrations and you find the best material to run through a drywasher - and I'd imagine a metal detector. I was too anxious and greedy - I was trying my GMT for the first time and wanted nuggests but all I found were bullets. If however one was to first look for black sand concentrations then gold it likely would pay off better than just hoping for nugget finds. So then I thought about our detectors and how they really "hate" finding black sand concentrations - they "want" to find the gold in the concentrations. These guys use more expensive equipment to map out areas then target key areas with the detectors or just dig. So, question is, should/could we come up with a modification of our PI designs to map out black sand first - then switch to targeting mode? (the GMT, not a PI, has a mineralization level bar and should be fine for what they do)

[green]

Interesting thread. I work in testing. I was taught to define what is required and what to expect when making a measurement . We need to get the no target signal low compared to the target signal. The target can be moved close to the coil to give a large response. I would hope the no target signal would be less than 1/5 the target signal. A couple options for measuring, mono coil or balanced Rx coil. Balanced coil: shouldn't have to wait for decay and the no target signal should be close to zero volts. Not sure what the target wave form should look like with different TC targets. Maybe someone could suggest a few target materials with different TC to look at. Maybe we could record some different ground responses.

[Carl-NC]

A few weeks ago I went on a planned outing to a running mining operation in Randsburg, CA. They enlightened me in a way that I hadn't thought about prospecting. They called themselves "black sand prospectors" not "gold prospectors". Find the black sand concentrations and you find the best material to run through a drywasher - and I'd imagine a metal detector. I was too anxious and greedy - I was trying my GMT for the first time and wanted nuggests but all I found were bullets. If however one was to first look for black sand concentrations then gold it likely would pay off better than just hoping for nugget finds. So then I thought about our detectors and how they really "hate" finding black sand concentrations - they "want" to find the gold in the concentrations. These guys use more expensive equipment to map out areas then target key areas with the detectors or just dig. So, question is, should/could we come up with a modification of our PI designs to map out black sand first - then switch to targeting mode? (the GMT, not a PI, has a mineralization level bar and should be fine for what they do)

You want a magnetometer for this.

[Tinkerer]

Eric, thank you for your help.

So, basically my ground model was wrong. Bricks are OK. Ferrite is OK. Cast iron is NOT OK.

Start over again.

Next try: use several pieces of ferrite together with the bricks, to produce a ground signal that is of greater amplitude than the largest target.

To make the PIVOTS easier to see, chose a target with a TC that decays to the baseline before the end of the cycle.

Tinkerer

[Prospector_Al]

You want a magnetometer for this.

Hi Barry,

Carl is right, you need a magnetometer. You can build one yourself. Popular Mechanics had plans for one, some time ago.

Here are some facts about these devices:

http://perso.infonie.be/j.g.delannoy...Technology.pdf

"Traps" in bedrock are full of black sand and nuggets. Owing to their specific gravity compared to the country rock, gold and blacks san are found together.

If you are looking for bigger nuggets, the correlation between magnetic mineral content and prevalence of nuggets may not be so good.

All the best,

Prospector_Al

[bklein]

I have a Schoenstadt survey pin finder that is awesome for this and meteorites...

[Ferric Toes]

Eric, thank you for your help.

So, basically my ground model was wrong. Bricks are OK. Ferrite is OK. Cast iron is NOT OK.

Start over again.

Next try: use several pieces of ferrite together with the bricks, to produce a ground signal that is of greater amplitude than the largest target.

To make the PIVOTS easier to see, chose a target with a TC that decays to the baseline before the end of the cycle.

Tinkerer

Bricks are OK but vary a lot in the amplitude department. You want about three layers with a small gap, say 1/10in, between layers to insert coins etc. The area should be at least twice the search coil area to be reasonably realistic. Ferrite material is not necessarily useful. Soft ferrites are not (transformer cores, many pot cores) as they are coarse grained and have very little viscous signal. Some fine grained hard ferrites, such as used for radio antennas have a strong viscous response. Soft and hard refer to magnetic properties rather than physical hardness. I used some very heavy "bricks" from an old electric storage heater at one time. These had a reasonable viscosity signal. Cast iron is definitely out.

Eric.

[Ferric Toes]

The attached report includes a comparison between housebrick and Wedderburn ironstone from Oz., plus other materials. All have similar slope on the log log graphs.

Eric.

Comparison Samples002.pdf