Ground Balance Theory

[Ferric Toes]

This below is very interesting target response (of relatively large ring) to different pulse widths, 3 and 50 uS in this case, generated with same peak current, so magnetic field straight and stored energy is identical. Top is 50uS, middle is 3uS, lower is idle setup response for reference, 10 uS\div.


Response varies not only in magnitude, as expected, but also in shape of decay curve (anticipated, but now i captured it). Response is interesting, increasing pulse width, it rises up to one point when it remains fairly constant for any further increase. Perhaps the easiest way to determine target TC. Note this is only response waveform, integrated measurement produce even more dramatic change. This is relatively short TC, with longer TC objects change is even greater. Unlike this, ferrite response remain almost unchanged, i suppose soil will be somewhere between, also salt water may be interesting to see. This is bit problematic to measure with normal MD setup, only varying pulse width will also change peak current, and stored energy, so initial conditions are different. In my version TX time is variable continuously over 50 to 1 range keeping other conditions almost unchanged.


Basic idea behind this is to utilize difference in response of soil and metal to two different pulse widths, one very short, below target TC and one longer, above TC to establish GB. If it fails in ground, idea may be seaworthy, enabling very fast sampling detector to operate underwater without being affected. All this is just at theory, but even if it fails completely, i will at least gather enough data to know exactly why. Sometimes this make more fun than actual detecting...

What you are seeing is exactly right. With the short pulse you are only exciting skin effect currents as there is not time enough for the pulse to fully energise the ring. As the TX pulse width is increased the field penetration goes deeper and you reach a point where the ring (or any other target) is giving its full eddy current response. Theoretically this would be when the TX width is 5x the target TC. Ferrite response should not change much whatever the pulse width, although some would say differently. Salt water will be interesting. I have found that you get no response from a plastic bucketful, yet go to a wet beach and even at 10uS there is a strong response. All to do with the volume that is energised.

Try a thin plain ring such as a wedding band. Compare 9ct and 18ct gold.

Eric.

[Davor]

Salt water will be interesting. I have found that you get no response from a plastic bucketful, yet go to a wet beach and even at 10uS there is a strong response. All to do with the volume that is energised.

That's another nail in the "search cone" coffin. In case of a beach a whole half space is exposed. That's why I'm into a differential coil and 4-quadrant discrimination with CW IB.

[Ferric Toes]

Back to the matter of viscous decay in soils and rocks; Aziz is questioning (on Doug's forum) why the decays he has do not conform to a 1/t law and whether there is non-linearity either in the material or a function of the electronics. It is easy with devices available today to make a system that is linear and with enough bandwidth to make accurate measurements suitable for practical field useable metal detector design. Looking at the huge amount of material, both theoretical and practical that has accumulated since at least the 1960's, by defense organisations, universities, and private companies, I don't see point of reinventing the wheel on that side of the fence. Now we know what we are dealing with, (and for PI that is short term SPM magnetic viscosity), and the laws which govern the decay are very well known and readily available, let's concentrate on what can be done electronically to minimise the effect.

I have been concentrating for a number of years in doing just that, and I believe that I have reproducible results that are nearer to the theoretical value than have been obtained previously. Even on a sample of Australian ironstone the preamp is not even a quarter of the way to saturating at 10uS delay and any remaining coil LR response or lack of flatness is cancelled at early time for each measurement. The front end of my viscosity tester is a NE5532A with a total gain of 500x, and I can (with a clock mod) get down to 5uS delay. I have substituted a LME49990 dual and I see that it recovers faster, but then the 5532 is fast enough anyway. Plotting a graph with either type showed no difference. One big reason for keeping the 5532 is that two of LME49990s draw a total of 20mA more from my AA battery power supply.

Below is a plot done today for a sample of Red Hill soil. This is nothing exotic, but just a normal soil that has a moderately high viscous response. One that would give you problems unless you had a reasonable GB facility. Instead of starting the plot at 10uS, I have now started it at 5uS. No deviation from a straight line shows on the log log plot and the slope calcuates at -1.047, so I can safely assume that there are no non-linear effects either in the sample or the electronics. Even if there were they do not affect the result. The calculated slope for the graph is -1.047.

My workshop is a bit noisy from sources of adjacent premises, so there is a small amount of scatter at the late time end of the decay. I am planning on a noise cancelling sensor which will improve this.

Eric.

[Aziz]

Back to the matter of viscous decay in soils and rocks; Aziz is questioning (on Doug's forum) why the decays he has do not conform to a 1/t law and whether there is non-linearity either in the material or a function of the electronics. It is easy with devices available today to make a system that is linear and with enough bandwidth to make accurate measurements suitable for practical field useable metal detector design. Looking at the huge amount of material, both theoretical and practical that has accumulated since at least the 1960's, by defense organisations, universities, and private companies, I don't see point of reinventing the wheel on that side of the fence. Now we know what we are dealing with, (and for PI that is short term SPM magnetic viscosity), and the laws which govern the decay are very well known and readily available, let's concentrate on what can be done electronically to minimise the effect.

I have been concentrating for a number of years in doing just that, and I believe that I have reproducible results that are nearer to the theoretical value than have been obtained previously. Even on a sample of Australian ironstone the preamp is not even a quarter of the way to saturating at 10uS delay and any remaining coil LR response or lack of flatness is cancelled at early time for each measurement. The front end of my viscosity tester is a NE5532A with a total gain of 500x, and I can (with a clock mod) get down to 5uS delay. I have substituted a LME49990 dual and I see that it recovers faster, but then the 5532 is fast enough anyway. Plotting a graph with either type showed no difference. One big reason for keeping the 5532 is that two of LME49990s draw a total of 20mA more from my AA battery power supply.

Below is a plot done today for a sample of Red Hill soil. This is nothing exotic, but just a normal soil that has a moderately high viscous response. One that would give you problems unless you had a reasonable GB facility. Instead of starting the plot at 10uS, I have now started it at 5uS. No deviation from a straight line shows on the log log plot and the slope calcuates at -1.047, so I can safely assume that there are no non-linear effects either in the sample or the electronics. Even if there were they do not affect the result. The calculated slope for the graph is -1.047.

My workshop is a bit noisy from sources of adjacent premises, so there is a small amount of scatter at the late time end of the decay. I am planning on a noise cancelling sensor which will improve this.

Eric.
[ATTACH]24315[/ATTACH]

Hi Eric,

thank you very much for clarifying this issue. I'm very happy about the fact now and this makes the GB much more simple.
I wonder, whether the decay exponent will vary (slope != -1.0).
It would be very fine, if the slope could stay at -1.0 and the ground response function would be simply G(t) = a / t.

Cheers,
Aziz

[deemon]

Below is a plot done today for a sample of Red Hill soil. This is nothing exotic, but just a normal soil that has a moderately high viscous response. One that would give you problems unless you had a reasonable GB facility. Instead of starting the plot at 10uS, I have now started it at 5uS. No deviation from a straight line shows on the log log plot and the slope calcuates at -1.047, so I can safely assume that there are no non-linear effects either in the sample or the electronics. Even if there were they do not affect the result. The calculated slope for the graph is -1.047.
Eric.

By the way , Eric , I have a question - does the viscous material behavior depend on unipolar or bipolar pulse exitation ?

[green]

Questions and observations
Is the plot correct for a first sample at 5 usec? If the ground and a target signal are 120 mv do they add to 240 mv? Can I slide any of the traces vertically and the plot be correct? The ground signal is greater than all target signals with a TC of 30 usec or less at 100 usec.

[Ferric Toes]

Hi Eric,

thank you very much for clarifying this issue. I'm very happy about the fact now and this makes the GB much more simple.
I wonder, whether the decay exponent will vary (slope != -1.0).
It would be very fine, if the slope could stay at -1.0 and the ground response function would be simply G(t) = a / t.

Cheers,
Aziz

Hi Aziz,

The Bosnar patent (US 6326 791 B1 of 2001 describes one way of subtracting out the ground response. They suggest a slope of -1.3, but they do question it. The slope is bound to vary a bit either side of 1.0. How much I will try and find out, but the measurements I have done so far suggest 0.99 - 1.1. The theoretical slope of 1.0 is derived from the classical Neel theory of well dispersed non-interacting single domain grains (Dabas and Skinner) but most soils and rocks are never going to be the theoretical ideal. Probably the Yucca Mountain Tuff comes closest.

One unknown at present is whether the slope varies with increased magnetic flux. Le Borgne quoted an upper limit of 2000uT for linearity, however Dabas and Skinner measured over the range 200 - 2300uT and found no change. The tests I do are in a calculated field of 350uT with the sample inside a solenoid coil 40mm diameter and 50mm long.

What would be nice is a user variable rejection window centred on -1.0 but with the capability to be widened out to encompass outlier values if the need arises.

Eric.

[Ferric Toes]

Questions and observations
Is the plot correct for a first sample at 5 usec? If the ground and a target signal are 120 mv do they add to 240 mv? Can I slide any of the traces vertically and the plot be correct? The ground signal is greater than all target signals with a TC of 30 usec or less at 100 usec.

Is the plot correct for a first sample at 5 usec? Yes, you can start wherever you like.

If the ground and a target signal are 120 mv do they add to 240 mv? Yes, simple addition.

Can I slide any of the traces vertically and the plot be correct? Yes you can do that too.

The ground signal is greater than all target signals with a TC of 30 usec or less at 100 usec. This one is a No. If you slide the ground signal down, the ground signal origin will come down too and the metal decays will come further and further clear. In reality a target with a short TC will have a higher starting amplitude than one with a long TC.

Eric.

[Ferric Toes]

By the way , Eric , I have a question - does the viscous material behavior depend on unipolar or bipolar pulse exitation ?

I have never observed a difference between the two. At high TX field strengths there may be a difference on some Australian ironstones which can support remanent magnetisation. Remanent magnetisation affects the susceptibility, which in turn affects the viscosity. Bipolar would prevent this.

Eric.

[Ferric Toes]

INSTRUMENTATION FOR MAGNETIC MEASUREMENT OF SOIL AND ROCKS.

Here is a picture of my soil magnetics test gear.



On the left at the back is the Bartington MS2 Susceptibility Meter with the MS2B sensor at the front left. The switch on the sensor sets the frequency to either 0.47kHz or 4.7kHz. The 10gm sample of Red Hill soil is in the measuring chamber and the reading is 871. This is 871 x 10^-5 SI units. Next to the MS2B sensor is an electronic scale for weighing the sample. The pots are 3.3gm so I fill them with the sample to measure to get 13.3gm.

The next measuring sensor is for the Viscosity Meter and is designed to accept the same 10gm sample. The control unit is bottom right and is displaying 140 units for a similar Red Hill sample. The viscosity meter is calibrated at 30uS delay to correspond to the difference in susceptibility between the LF and HF readings on the Bartington unit. Hence if you were to subtract 140 from 871 you would get 731, which would be the HF reading on the Bartington if you switched to that mode. That way it is a double check on the frequency dependent susceptibility, which equals viscosity.

Running through the delay settings on the viscosity meter you can then plot a graph, using suitable software, and get a reading for the slope.

Just thought a picture would clarify the method I use to come up with the figures and graphs which I post on this forum.

Eric.

[WM6]

Just thought a picture would clarify the method I use to come up with the figures and graphs which I post on this forum.

Eric.

Very illustative and clear display, thank you Eric.

[Ferric Toes]

In addition to the above equipment for measuring susceptibiity and viscosity (post 205), I have a small oven with timer and temperature control. This is used both for drying soil samples so that any water content does not distort the weight, and raising the samples in temperature to see how that affects the magnetic properties.

Eric.

[Ferric Toes]

These plots were done in 1968 by C. Colani and shows the advanced thinking at that time. All were done with a logarithmic front end amplifier so that exponential decays from non-ferrous targets appear as a straight line decay. Top left is a brass test plate 10 x 10cm. This is the noise free trace. The lower trace is a derived negative voltage level that is proportional to the slope, and that gets noisier as the signal decays. There is some non linearity at the bottom end of the decay, hence the curve. Top right is the trace from a nail which gives a log lin response similar to viscosity. The bottom left photo is for three exponential TCs - 50, 100, and 150uS and also shows the three derived levels. Further details can be seen on the .pdf (pages 8 - 10) attached to Post 65 on The PI History and Theory thread.

Eric.

[Aziz]

These plots were done in 1968 by C. Colani and shows the advanced thinking at that time. All were done with a logarithmic front end amplifier so that exponential decays from non-ferrous targets appear as a straight line decay. Top left is a brass test plate 10 x 10cm. This is the noise free trace. The lower trace is a derived negative voltage level that is proportional to the slope, and that gets noisier as the signal decays. There is some non linearity at the bottom end of the decay, hence the curve. Top right is the trace from a nail which gives a log lin response similar to viscosity. The bottom left photo is for three exponential TCs - 50, 100, and 150uS and also shows the three derived levels. Further details can be seen on the .pdf (pages 8 - 10) attached to Post 65 on The PI History and Theory thread.

Eric.

[ATTACH]24332[/ATTACH]

Hi Eric,

the attachment isn't working.

BTW, the old facts are really very interesting.
I'm referring to the Colani document now:
y1 = B/t (magnetic relaxation induction)
y2 = A*e^(-t/a) (target response exponential decay induction)
y(t) = y1(t)+y2(t)
Z = d(y(t)*t)/dt
Z = A*e^(-t/a) *(1 - t/a) and hence B is eliminated (ground balanced)

http://www.wolframalpha.com/input/?i...%29%29%29%2Fdt
( d(t*(B/t + A*e^(-t/a)))/dt )

Impressive.


Aziz

[Ferric Toes]

Hi Eric,

the attachment isn't working.

BTW, the old facts are really very interesting.
I'm referring to the Colani document now:
y1 = B/t (magnetic relaxation induction)
y2 = A*e^(-t/a) (target response exponential decay induction)
y(t) = y1(t)+y2(t)
Z = d(y(t)*t)/dt
Z = A*e^(-t/a) *(1 - t/a) and hence B is eliminated (ground balanced)

http://www.wolframalpha.com/input/?i...%29%29%29%2Fdt
( d(t*(B/t + A*e^(-t/a)))/dt )

Impressive.


Aziz

Link works for me, although I uploaded the picture to be visible with post 208. Here it is again.

Eric.