Tests of big depth metal detectors ( TR and PI systems )

[Ferric Toes]

Download this paper where pages 5 - 8 covers the subject of current diffusion. http://www.geonics.com/pdfs/technicalnotes/tn7.pdf

As Riss says, "metal detectors are surface geophysical systems,they use methods and principles originally used in geophysics".

Eric.

[bbsailor]

Download this paper where pages 5 - 8 covers the subject of current diffusion. http://www.geonics.com/pdfs/technicalnotes/tn7.pdf

As Riss says, "metal detectors are surface geophysical systems,they use methods and principles originally used in geophysics".

Eric.

Eric,

I recall from your forums that you simplified these technical paper key points by stating this about the pulse characteristics. The full stimulation of a target comes from the time constant (TC) of the coil current turn off slope which should be one fifth of the target TC (rule of thumb). The target TC defined by this rule of thumb means that the coil discharge TC is the coil inductance divided by the effective value of the damping resistor. If a desired delay is 10 uS, then a target needs a TC longer than 2 uS as it's energy will fall to near zero when the sampling at 10 uS occurs. Now, let's assume that we use the same 2 uS target but can drop the delay to about 7.5 uS, then we might be able to detect that 2 uS target but we now need to consider fully stimulating it. Using the five times shorter than the target rule of thumb, we now need a coil discharge TC of 2/5 or 0.4 uS. For a 300 uH coil to have a 0.4 uS discharge TC, the effective damping resistance value needs to be 300/0.4 or 750 ohms. In mono coils, however, the clamping diodes put the input resistor to the first amp stage effectively in parallel with the damping resistor value until the voltage across the diode drops to below about 0.6V. The consequence of this makes it difficult to fully stimulate very low TC targets by following the information in this technical paper to it's logical conclusion. It also stimulates some of your early innovative Pulse induction designs such as using very short TX pulses, making the RX window the same as the delay, using 3K to 10K PPS rates and integrating many RX signals to improve the signal to noise ratio.

Did I connect all these dots correctly?

Joseph J. Rogowski

[green]

Download this paper where pages 5 - 8 covers the subject of current diffusion. http://www.geonics.com/pdfs/technicalnotes/tn7.pdf

As Riss says, "metal detectors are surface geophysical systems,they use methods and principles originally used in geophysics".

Eric.

Thanks for the link to the paper. Read it a couple times. Decided I needed a paper(magnetics for dummies)to start with. Did a search and found some but so far haven't answered my questions. Charted targets a little different. Tx, 160usec 6250 amps/sec. Rx two 200mm round coils figure eight, IB. Blue trace, preamp out(gain about 450)target recording-no target recording. Red trace, amplifier out(gain=9, total gain about 4000)(target recording-no target recording)/9. Added a line to coil off trace where it appeared to be straight on the lin log chart and calculated time constant. Added another line parallel to the other line thru the coil on decay. The decay slope was similar with all targets but the lead ball. Ground, ferrite and some hot rocks I have chart an amplitude change with little or no decay slope when coil is on. A nail depending on orientation charts an increase in slope when coil on. Coil current shape and on time effect the coil off decay. Does the target look like an inductor with resistance when the coil is on? What controls the decay slope when the coil is on with a constant rate current? The paper explains the coil off decay.

I've made a lot of measurements. Knowing something about what I'm trying to measure and what to expect makes for better measurements. Some time things don't repeat as close as I think they should, probably because I don't know what to expect. Any suggestions would be appreciated.

How much if any would reducing the Tx from 1000pps to 100pps effect the Rx signal amplitude or decay shape?

[green]

Tried a spice simulation to understand what is happening with a constant rate current Tx pulse. The target(coupled LR circuit) volts and current look like they increase exponentially. The signal decreases on my charts. Don't know how to do a IB simulation. Maybe that would help explain or maybe someone could explain what is happening. Thanks

[Davor]

It is not that complicated. Although physically more correct is the paper Eric linked, you may observe this via skin effect as well. The skin effect is frequency dependent. After a transient excitation, the immediate response at time t1 is bandwidth limited from below at about 1/t1, and the high frequency corresponds to thin skin. Later on at time t2>t1 the bandwidth limit goes down to ~1/t2, and that's the reason thick chunks of good conductors appear on PI machines to have taus exceed values expected by their conductivity.
IB-s have no such problem because there is no frequency dependence - frequency is set. Therefore with a given material skin depth is constant, and only when a sample is considerably thinner than skin size, like with aluminium foil, you get off effects.

As for detecting deep anomalies, I'd say "cable locator" kind of a detector would be in a considerable advantage over normal metal detectors.

[green]

It is not that complicated. Although physically more correct is the paper Eric linked, you may observe this via skin effect as well. The skin effect is frequency dependent. After a transient excitation, the immediate response at time t1 is bandwidth limited from below at about 1/t1, and the high frequency corresponds to thin skin. Later on at time t2>t1 the bandwidth limit goes down to ~1/t2, and that's the reason thick chunks of good conductors appear on PI machines to have taus exceed values expected by their conductivity.
IB-s have no such problem because there is no frequency dependence - frequency is set. Therefore with a given material skin depth is constant, and only when a sample is considerably thinner than skin size, like with aluminium foil, you get off effects.

As for detecting deep anomalies, I'd say "cable locator" kind of a detector would be in a considerable advantage over normal metal detectors.

Thanks Davor, I think I'm starting to understand the coil off decay. The coil on I hadn't. I keep forgetting coupling is rate of change. Don't see skin effect on the coil on trace. With spice the target current had an exponential increase in current during coil on with a constant rate Tx pulse. Faster at start so the IB Rx trace should be higher at the start and decay exponentially? Maybe all wrong? Still would like to know how to do an IB simulation with spice.

[green]

Eric,

I recall from your forums that you simplified these technical paper key points by stating this about the pulse characteristics. The full stimulation of a target comes from the time constant (TC) of the coil current turn off slope which should be one fifth of the target TC (rule of thumb). The target TC defined by this rule of thumb means that the coil discharge TC is the coil inductance divided by the effective value of the damping resistor. If a desired delay is 10 uS, then a target needs a TC longer than 2 uS as it's energy will fall to near zero when the sampling at 10 uS occurs. Now, let's assume that we use the same 2 uS target but can drop the delay to about 7.5 uS, then we might be able to detect that 2 uS target but we now need to consider fully stimulating it. Using the five times shorter than the target rule of thumb, we now need a coil discharge TC of 2/5 or 0.4 uS. For a 300 uH coil to have a 0.4 uS discharge TC, the effective damping resistance value needs to be 300/0.4 or 750 ohms. In mono coils, however, the clamping diodes put the input resistor to the first amp stage effectively in parallel with the damping resistor value until the voltage across the diode drops to below about 0.6V. The consequence of this makes it difficult to fully stimulate very low TC targets by following the information in this technical paper to it's logical conclusion. It also stimulates some of your early innovative Pulse induction designs such as using very short TX pulses, making the RX window the same as the delay, using 3K to 10K PPS rates and integrating many RX signals to improve the signal to noise ratio.

Did I connect all these dots correctly?

Joseph J. Rogowski

In mono coils, however, the clamping diodes put the input resistor to the first amp stage effectively in parallel with the damping resistor value until the voltage across the diode drops to below about 0.6V. The consequence of this makes it difficult to fully stimulate very low TC targets by following the information in this technical paper to it's logical conclusion.

Connecting the coil to the -input and using the input resistor for the damping resistor can solve that problem.

[Davor]

Unless you use an induction balanced coil, say DD.

[green]

Unless you use an induction balanced coil, say DD.

My IB coil. Rx two(200mm round coils, figure eight) Tx(oval over Rx coils). A picture I posted awhile back.

[Davor]

Very nice. And what is the setup you use to get the response diagrams?

[Davor]

...Still would like to know how to do an IB simulation with spice.

Easy. You simply connect a sine voltage source to a coil L1 of, say 1mH, and series resistance of 1 ohm to avoid Spice complaining. Parallel capacitor is not required, as you are not trying to recuperate any current here.
A target is a coil L2 of, say 100uH for copper in parallel with a 1 ohm resistor, and one side of this contraption goes to ground.
At Rx side you mimic a Rx front end with a coil L3 in parallel with a capacitor and a resistor. Pick a typical resistance of 10k and a coil inductance in some reasonable range, say 16mH. Pick a capacitor that causes resonance to overshoot Tx frequency (or in case you simulate whites undershoot).
The most important part, add coupling statements like K1 L1 L2 0.01 and K2 L2 L3 0.01
That's it.

[green]

Very nice. And what is the setup you use to get the response diagrams?

I use a RIGOL 1052E to record the data(low cost scope but good enough). Use IB coil reply #39 potted with foam with graphite shield. Two stage amplifier, first stage(gain about 450) second stage(gain=9 for total gain about 4000). Scope channel 1 connected to first stage amplifier out, channel 2 connected to second stage amplifier out, scope trigger connected to coil mosfet driver input to trigger at Tx off(zero usec at turn off). Record a no target scan, record a target scan(target height adjusted for near full scale on channel 1). I use a similar coil(Rx about 38mm diameter) for smaller targets to get enough signal. Copy the no target recording and the target recording into Excel. Multiply the time recording by 1E-6 for usec. Subtract channel 1 no target recording from channel 1 target recording. Subtract (channel 2 no target recording from channel 2 target recording)/9. Chart the data on a XY graph. Change the XY scales to desired range, linear or log.

[green]

I use a RIGOL 1052E to record the data(low cost scope but good enough). Use IB coil reply #39 potted with foam with graphite shield. Two stage amplifier, first stage(gain about 450) second stage(gain=9 for total gain about 4000). Scope channel 1 connected to first stage amplifier out, channel 2 connected to second stage amplifier out, scope trigger connected to coil mosfet driver input to trigger at Tx off(zero usec at turn off). Record a no target scan, record a target scan(target height adjusted for near full scale on channel 1). I use a similar coil(Rx about 38mm diameter) for smaller targets to get enough signal. Copy the no target recording and the target recording into Excel. Multiply the time recording by 1E-6 for usec. Subtract channel 1 no target recording from channel 1 target recording. Subtract (channel 2 no target recording from channel 2 target recording)/9. Chart the data on a XY graph. Change the XY scales to desired range, linear or log.

Multiply time scale by 1E6 not 1E-6.

[green]

An attempt at a spice simulation.

[Davor]

Looks good.
As you see, you can't simulate the thickness-dependent behaviour that is relevant for PI. I made a viscous soil model, but so far I did not try simulating the thickness.