Ground Balance Theory

[Ferric Toes]

Didn't mean to kill the thread. Looking at Eric's plots the obvious way to cancel ground is. Sum two samples into an integrator. Preamp out at a delay of say 4 to 8 and inverted out with a adjustabe gain around -1 at a delay of maybe 64 to 128. It can't be that easy or everyone would be doing it. Is the problem with ground response not being predictable, with hardware or some other variables? Received a copy of ITMD a couple weeks ago. Maybe I'll get smarter and ask better questions.

I've been out of the loop for a bit, working on a high power boat towed detector. That's now finished and delivered so lets think about GB again. Yes Green, in principle it is that easy. Not perfect, but an effective GB for most soils nonetheless. You don't want to go that far out with the GB delay though as you will need considerable amplification to achieve balance, with resulting increase in noise.

I missed the comments by Zed and I wonder what detectors show this large variability in ground signal? Are they PI or CW? I have accumulated a lot of measurements on different soils and rocks and the decay is remarkably constant. I could set the GB control to cancel a housebrick in England and it would only require a minor adjustment in Australia. This is of course with PI.

Eric.

[mickstv]

Didn't mean to kill the thread. Looking at Eric's plots the obvious way to cancel ground is. Sum two samples into an integrator. Preamp out at a delay of say 4 to 8 and inverted out with a adjustabe gain around -1 at a delay of maybe 64 to 128. It can't be that easy or everyone would be doing it. Is the problem with ground response not being predictable, with hardware or some other variables? Received a copy of ITMD a couple weeks ago. Maybe I'll get smarter and ask better questions.

The other problem you have in relation to GB is that some targets depending on size and TC can fall into whats called an electronic hole, depth on these targets will be reduced or not detectable at all.

[Prospector_Al]

Or more like Exxon Valdez, not completely sink.


One small update of one method I proposed here and some results I get with it. Yes, it works, this is effective GB mechanism, but as usual, having it's own good and bad sides.


Technically, method is similar to what ML call DVT, generating one relatively long (50uS in this case), and one relatively short (adjustable, but at least 10 times shorter, even down to 0.5uS) pulse of opposite polarity, releasing same amount of energy. Achieved by constant current control and two different voltages to produce required pulses without need for external timing, voltages are same polarity but two sections of the coil connected in opposite directions to get bipolar field, fairly simple circuit built around SG3525 chip. Idea is, long pulse around target TC will excite it and ground, short one will excite ground but target very little excited, then substract.


Result: it works, having no “holes” and rejection zones in response, less substraction is needed, only some practical aspects limit it's usability. Problem is with short\long pulse ratio and two effects they produce. First one is, despite no dead-zones, sensitivity to very small, low TC targets fall rapidly as it's TC approach short pulse width. Another is practical duration of this short pulse. It is possible to generate very short ones, uS or below, but with very short pulses another issue appear. Now total pulse width is comparable to flyback release time, this waveform produce different effect in soil compared to long pulse, alone cannot be used to establish reference. Limiting practical pulse ratio to say 10\1 , like 50\5uS or so (using longer “long” pulse will not help much, target TC consideration is for short one). Then this short pulse became comparable to target TC, reducing “contrast” it ment to achieve and sensitivity, again lot of substraction needed. Workable but not effective, loss of sensitivity, overally not better than GS approach, instead of response hole, now small TC sensitivity suffer. Alone, not a good solution. But was interesting to try. When time allow, I will modify this setup to try few different tricks. What I have to try is, first, true log-amp front end running in parallel with normal one, to see how it can be exploited best, and some playing with nonlinear inductance (saturable core) in the circuit, to get current ramp different from linear. By all chance, this may end up like nice DIY project, sort of bipolar pulsing (maybe logarithmic) GS incarnation. With time available, maybe finished before end of days. Any suggestion before I disassemble this prototype? I have something different, apparently workable and relatiely new on my workbench, will post more, at least idea.

Hi Tepco,

The method you describe works. It has been used in industrial "tramp metal" detectors for about ten years.

There are no "holes" in the response and in this application, the loss of sensitivity to small targets is immaterial since only targets that are large enoough to damage the crusher need to be detected.

Good thinking, but a bit late for patent applications...

All the best,

Allan

[Tepco]

There is nothing about patent applications or anything similar, just hobby and some hardware hacking, best way to learn about something is to try to actually build it. I never sold even a single hand made detector, put aside anything else, not interested in this for any commercial reason. I supposed all this is already done before, just tried on my own. Never figured out that similar approach can be used to actually separate different sized objects.


Method I described still can be rescued, only not in it's original form. I tried to achieve GB contrast using two very different pulse widths alone, but this is wrong approach. Using two different pulses, one longer than ground TC and one shorter, maybe 10:1 ratio (no need for 500nS TX and similar extreme solutions) works. Then processed in two independent channels, each GB-ed separately using two sample method. Now, each channel will have it's own “hole” in response, but for metallic object holes will fall in different positions. Adding outputs together, this is almost “hole free”, well, some variation in sensitivity but no notch in response.


Another version, with resonant waveform, is also promising, but on prototype ended up with 6 Mosfets, 3 H-drivers and extremely messy timing to command them high, low or open... Time needed to produce something really workable out of this unfortunately by far exceed my available resources.


Best regard

[Ferric Toes]

In the quest for higher accuracy in the measurement of magnetic viscosity in soil and rock samples, a new type of sensor coil was constructed. Instead of the usual monocoil type solenoid winding a bifilar centre tapped winding was used. The potential advantages of this type of coil winding having been discussed previously on this forum. Effectively half of the winding is used as a TX and the total winding as RX in a PI type circuit.

It is important that the soil or rock sample is measured in a uniform field, hence the solenoid type coil is roughly 2x the depth of the 10cc sample pot, with the sample centrally placed. Previously the single winding covered a length of 50mm with a diameter of 40mm. To achieve the same uniform field a bifilar winding was used which was so connected as to form a centre tapped inductor with one end grounded; the TX drive to the centre tap; and the RX forming the total winding from ground to the far end. The original single winding had a inductance of 200uH and this was also the case for the total winding of the Bifilar CT coil. Inductance from ground to the centre tap is therefore 50uH which is more favourable for a fast rise of current when the TX pulse is applied. The TX coil circuit resistance totals at 30 ohms, so the turn on TC is 50/30 = 1.66uS. The TX pulse is then flat from about 8uS to whenever the switchoff point occurs. In contrast, the single coil solenoid has a TC of 6.66uS and flat after 33uS. The small tradoff is that the ampere turns is halved for the BCT coil, and to compensate for this the TX pulse amplitude was doubled.

Simple resistive damping was applied to each half winding and consisted of a fixed resistor in series with a cermet preset. This arrangement would normally give trouble from the high voltage spike, but the half winding only generated 30V peak and did not come close to the avalanche voltage of the Mosfet used. Damping can be finely adjusted to give the waveform shown. The lower pulse is the delay interval which triggers the scope on its positive edge which coincides with TXoff. The delay is set at 10uS, but as can be seen, you could easily sample as early as 6uS, and with an improved preamp, even 5uS is possible.

A couple of viscosity decay curves have been plotted so far with this BCT coil and results are a bit closer to the theoretical t^-1.0xx but an interesting thing is beginning to consistently occur. For delays <20uS the decay appears to have another faster curve superimposed on it. This should become more apparent when I run a modified timer to observe signals at 5, 10, and 15uS where the discrepancies can be seen. Till now, I thought that it could be an instrumentation error, but as I see it with the faster coil, it may well be real. This has implications for those trying sub 10uS sampling, as the amplitude will rise even faster than it does with the fundamental decay.

Eric

[Reno Chris]

I missed the comments by Zed and I wonder what detectors show this large variability in ground signal? Are they PI or CW? I have accumulated a lot of measurements on different soils and rocks and the decay is remarkably constant.

This experience is with a minelab PI. The problem with high variability in ground mineralization is one that is probably almost unique to gold prospecting, especially residual type placers. I have experienced this hunting gold in Australia, California and Alaska. In some places it is actually possible to tell how close you are to where the gold is by how chattery the signal from your detector is. The more noise and frequently re-ground balancing is required, the closer you are to the gold. This is an effect of the hot solutions that actually deposit the gold because they are often also strongly iron bearing. It also occurs sometimes as iron minerals are heavy and concentrated along bedrock as is the gold. It is not universal in all goldfields, but as Zed pointed out and I agree, it is more common than one might expect. Single samples from any particular location will not demonstrate this, but would pretty much require a series of samples taken at maybe 6 inch to 1 foot intervals over a length of 30 to 50 feet.

[Ferric Toes]

This experience is with a minelab PI. The problem with high variability in ground mineralization is one that is probably almost unique to gold prospecting, especially residual type placers. I have experienced this hunting gold in Australia, California and Alaska. In some places it is actually possible to tell how close you are to where the gold is by how chattery the signal from your detector is. The more noise and frequently re-ground balancing is required, the closer you are to the gold. This is an effect of the hot solutions that actually deposit the gold because they are often also strongly iron bearing. It also occurs sometimes as iron minerals are heavy and concentrated along bedrock as is the gold. It is not universal in all goldfields, but as Zed pointed out and I agree, it is more common than one might expect. Single samples from any particular location will not demonstrate this, but would pretty much require a series of samples taken at maybe 6 inch to 1 foot intervals over a length of 30 to 50 feet.

Hi Chris,

There is of course the possibility that the TX waveform of the Minelab PI is not the best for generating a consistent ground signal. Theoretically the decay of the SPM non-conducting particles in mineralised ground obeys a certain law within very small limits. However, this can be adversely affected by a non-ideal pulse shape. My viscosity meter utilises as near to a true rectangular pulse as is possible of exactly the same amplitude whatever the pulse width. By comparison ML pulses are nearer sawtooth shape. Using the rectangular pulse waveform I see very small variations in decay in samples from different areas and countries, but nothing that would cause more than a minor adjustment in the ground balance setting, and this has been proved in the field in UK, Australia, and USA. There are different design philosophies for metal detectors and good reasons for doing something one way often has a tradeoff in other areas.

Eric.

[bbsailor]

Hi Chris,

There is of course the possibility that the TX waveform of the Minelab PI is not the best for generating a consistent ground signal. Theoretically the decay of the SPM non-conducting particles in mineralised ground obeys a certain law within very small limits. However, this can be adversely affected by a non-ideal pulse shape. My viscosity meter utilises as near to a true rectangular pulse as is possible of exactly the same amplitude whatever the pulse width. By comparison ML pulses are nearer sawtooth shape. Using the rectangular pulse waveform I see very small variations in decay in samples from different areas and countries, but nothing that would cause more than a minor adjustment in the ground balance setting, and this has been proved in the field in UK, Australia, and USA. There are different design philosophies for metal detectors and good reasons for doing something one way often has a tradeoff in other areas.

Eric.

Eric,

Picking up from your statement "There are different design philosophies for metal detectors and good reasons for doing something one way often has a tradeoff in other areas". Can you envision computer controlled metal detectors being developed and used to evaluate the ground conditions and variations within one sweep width to recommend an optimal (1) pulse width, (2) amplitude, (3) sweep speed and other TX or RX parameters for optimizing searching in that specific physical location for specified targets with preset TC ranges?

Thanks

bbsailor

Joseph Rogowski

[Paul99]

There is of course the possibility that the TX waveform of the Minelab PI is not the best for generating a consistent ground signal.

In saying that Eric, do you think if Minelab were to change their TX waveform pulse shape that, while they might get a consistent ground signal, it could result in them losing the significant depth advantage they currently have in the mineralised ground regularly encountered in the Australian goldfields?

Theoretically the decay of the SPM non-conducting particles in mineralised ground obeys a certain law within very small limits. However, this can be adversely affected by a non-ideal pulse shape. My viscosity meter utilises as near to a true rectangular pulse as is possible of exactly the same amplitude whatever the pulse width. By comparison ML pulses are nearer sawtooth shape. Using the rectangular pulse waveform I see very small variations in decay in samples from different areas and countries, but nothing that would cause more than a minor adjustment in the ground balance setting, and this has been proved in the field in UK, Australia, and USA.

Are you able to compete on depth with the current generation of Minelab detectors in Oz mineralised ground? If so, when do you expect to have a detector in the market place to compete on depth with Minelab , and do you have any idea of the final price?

There are different design philosophies for metal detectors and good reasons for doing something one way often has a tradeoff in other areas.

Eric.

Is it possible that Minelab are trading off some consistency in the ground signal in some ground to achieve the depth advantage they obviously have?

[Aziz]

Hey many thanks!, the interesting thread is continuing!
Looking forward to the outcome.
Aziz

PS: Ground balancing / mineralized ground always reduces the detection depth (compared to air test).

[Ferric Toes]

Eric,

Picking up from your statement "There are different design philosophies for metal detectors and good reasons for doing something one way often has a tradeoff in other areas". Can you envision computer controlled metal detectors being developed and used to evaluate the ground conditions and variations within one sweep width to recommend an optimal (1) pulse width, (2) amplitude, (3) sweep speed and other TX or RX parameters for optimizing searching in that specific physical location for specified targets with preset TC ranges?

Thanks

bbsailor

Joseph Rogowski

Hi Joseph,

No reason why it should not happen. Mine detection technology is already moving in that direction. The tradeoff in other areas that I referred to was really a downside effect. Improve one aspect e.g. depth, and other performance aspects suffer.

Eric.

[Ferric Toes]

In saying that Eric, do you think if Minelab were to change their TX waveform pulse shape that, while they might get a consistent ground signal, it could result in them losing the significant depth advantage they currently have in the mineralised ground regularly encountered in the Australian goldfields?

Are you able to compete on depth with the current generation of Minelab detectors in Oz mineralised ground? If so, when do you expect to have a detector in the market place to compete on depth with Minelab , and do you have any idea of the final price?

Is it possible that Minelab are trading off some consistency in the ground signal in some ground to achieve the depth advantage they obviously have?

These look like loaded questions and my answers will be brief and clear.

What ML do or not do is up to them. They no doubt have good reason to use the waveform they have. Em noise, powerline and ground noise do however limit the performance of ML detectors in many areas such that the maximum detection capability is not realisable.

I am not trying to compete with Minelab and have no plans to produce detectors for general sale. I just do research, development, consultancy and a few one-off specials mainly for the industrial market.

Eric.

[Paul99]

...my answers will be brief and clear.

Em noise, powerline and ground noise do however limit the performance of ML detectors in many areas such that the maximum detection capability is not realisable.

...

Eric.

Eric, my understanding is that increased emi, powerline and ground noise amplitude affect the depth capability of all high performance detectors. Could I ask you to make one or two points more clear for me please....

Am I right in understanding your statement that "Em noise, powerline and ground noise do however limit the performance of ML detectors in many areas such that the maximum detection capability is not realisable" to infer that it is possible to design a usable detector that will realise full detection depth in the presence of emi, powerline and ground noise?

And are you aware of any detector that is unaffected by emi and powerline noise that has depth performance comparable to the Minelab GPX series of detectors when working under powerlines?

The reason I ask is because there are many thousands of km of powerlines throughout Australian gold bearing areas, and your statement that powerline noise will "limit the performance of ML detectors in many areas such that the maximum detection capability is not realisable" infers that ML machines are not the ideal machine for working these areas. However, my experience is that other makes of detectors in the hands of experienced users have demonstrated less depth capability under powerlines than ML detectors achieve.

Your advice on a more appropriate detector for these areas would be appreciated.

[Tepco]

For delays <20uS the decay appears to have another faster curve superimposed on it. This should become more apparent when I run a modified timer to observe signals at 5, 10, and 15uS where the discrepancies can be seen. Till now, I thought that it could be an instrumentation error, but as I see it with the faster coil, it may well be real. This has implications for those trying sub 10uS sampling, as the amplitude will rise even faster than it does with the fundamental decay.

Eric

[ATTACH]25626[/ATTACH]

This is interesting, I noticed something similar. By all chance, there are actually 3 exponents. First, usually fastest one is dictated by coil and changes only with coil, another one, lasting to about 10uS is target related. Varying with shape and material, in not quite logical and predictable way. Also, strongly dependent on pulse width, gradually fading out at very short pulse excitation. After that, 15us or so later, everything is quite regular single exponent. How this can affect real detector is less important, very fast setup is needed for this effect to be visible. I'm using old and obsolete LM733 (NE592) video amplifier, in fully differential connection, but at least no speed limit and recovery issues with it, measurement after only few hundred nS is possible.


Another interesting effect I noticed is related to flyback release time and shape. Mostly related to “wet sand effect”, I can be wrong with this, but here is my conclusion, may be relevant to soil samples too, so please correct me if i'm wrong. I never figured out what is exactly going on with this wet sand stuff, salt water alone, or sand alone produce no response, but in right combination (and right sand type, ordinary sand for construction work will not respond, something grain size related probably), problem appears at short delays. Now, effect that looks like very short TC object happens. Only, problem is by all chance not initiated by TX pulse, independent on width, but by flyback cycle dV\dT in relatively low conductivity medium. Slower rate of change produce less effect, exactly opposite of what fast coil designer is trying to achieve. Not only longer sampling delay, but intentionally slower coil seems to help here.


I played with this effect, and figured out that intentionally underdamped coil, with flyback release in form of one complete ringdown sine cycle significantly reduce response. With flat spiral coil, each cycle is only 1.5-2uS, so delay is reduced only by 3-4uS, still fast enough, but more stability with same delay compared to “optimally” dumped coil. Even, at some empirical adjustment, consisting of 5-6 cycles of exponential ringdown, large, some 500g ferrite is completely ignored by detector, everything else being unaffected?!? Unfortunately, not soil sample or red brick, so nothing about “passive” ground balance. Still this may affect response from low conductivity soil samples, also may be interesting for underwater detectors. Any idea, what i'm missed, my measurement error or something else, these effects are real. Devil is in the waveforms , positively.

[green]

I've been playing with a PI circuit with a DD coil looking at coil on and off signals. I tried a piece of landscaping lava rock and some clay from the yard. I get a signal during on time, not much if any during off time. Exponential, 300usec TC. Should I expect the on slope to be different than the off slope(power)? Are my samples not typical? Excel plots, target - reference.