How it works

[Tec]

The main thing that is different about this detector is the way it samples.

It doesn't sample.
After the pulse finishes most PIs (pulse induction detectors) take a sample, or some samples at a certain time, for example; after 25 uSec.

This detector instead measures the time for the coil to discharge below a set voltage level.

That way, it is measuring a slight shift in time, rather than a slight amplitude change.

Which achieves the same thing as any other PI detector.
But there are subtle differences, advantages, and disadvantages.
(I don't know what they all are)

[Davor]

I figured that much. I also thought of this concept when I played with the log amp that enables linearisation of the otherwise quite complicated curve. Log amplification is quite useful because input signal ratios are easily achieved as simple differences.

There is a phenomenon that Tinkerer calls "the pivot". It is the point on a decay curve after which some distinction between targets can be observed. There is no point sampling any earlier because there is no useful information there, and also not much later because of the noise. If I remember correctly, I found pivot in my simulations to be somewhere above 100mv.

[WM6]

That way, it is measuring a slight shift in time, rather than a slight amplitude change.

Sampling slight shift in time, can be usable by Relic (militaria big Ferro parts) hunting, but not in prospecting where all changes happen in mentioned "pivot" area, with hard detectable changes in time.

[Davor]

Actually with pivot above 100mV you have almost two decades of useful signal span to do your window compares. With log amp pre you'd have even more than that, and it is something. The other problem is time granularity because you'll have a few microseconds full span of all samples and measuring these directly by a PIC is not promising much of a precision.

[Tinkerer]

I figured that much. I also thought of this concept when I played with the log amp that enables linearisation of the otherwise quite complicated curve. Log amplification is quite useful because input signal ratios are easily achieved as simple differences.

There is a phenomenon that Tinkerer calls "the pivot". It is the point on a decay curve after which some distinction between targets can be observed. There is no point sampling any earlier because there is no useful information there, and also not much later because of the noise. If I remember correctly, I found pivot in my simulations to be somewhere above 100mv.

Davor,

about "the pivot":

What I call "the pivot", is the point in time, where the target signal crosses the coil decay curve.
If we look at the coil decay curve with no target, as the 0 base line, then there is a point in time when the target response signal crosses this base line.
This means, that at this point in time, comparing the coil decay signal and the target response signal, we get 0, or no change.
At any time there is current running in the TX coil, the response of a ferrite target, (only X response) will be 180 degrees from the response of a purely R target.

Purely X response targets do not exist in nature. All kinds of steel have different magnetic properties (or permeability) and also resistance that causes eddy currents (R), and therefore the phase angle is different from 180 degrees.

The pivot, or point in time where the target signal crosses the base line, will therefore be varying with the phase angle.

So, in other words, "THE PIVOT" can be considered an indicator of phase angle.

To what degree can a PI detector define or indicate a phase angle? In the time domain? I think it could be interesting to investigate this assumption.

It might be a good idea to start a PIVOT THREAD, where we try to analyze the pivot and it's implications and the possible benefits of observing it might provide.

Tinkerer

[Davor]

Be assured that I'd be investigating this myself if only I had a decent scope, so my "experience" here comes from playing with the sims. These may be deceptive, but also may lead you to conclusions that you may test in a real life.

Anyway, I seen the pivot point in sims where I stepped the target tau, and there clearly is a point in time from where different targets' trajectories part. It was imposible to miss in case of a log amp.

I did not analyse what influences the pivot, is it damping, or coil inductance, pulse voltage or else, and I'm not sure wether sims would yield useful information on that. Given constant environment, e.g. rig components, the pivot is a fairly predictable in delay and value. It could be used for the initial sample as a baseline, but the real detection starts a microsecond or two later.

[Tinkerer]

Be assured that I'd be investigating this myself if only I had a decent scope, so my "experience" here comes from playing with the sims. These may be deceptive, but also may lead you to conclusions that you may test in a real life.

Anyway, I seen the pivot point in sims where I stepped the target tau, and there clearly is a point in time from where different targets' trajectories part. It was imposible to miss in case of a log amp.

I did not analyse what influences the pivot, is it damping, or coil inductance, pulse voltage or else, and I'm not sure wether sims would yield useful information on that. Given constant environment, e.g. rig components, the pivot is a fairly predictable in delay and value. It could be used for the initial sample as a baseline, but the real detection starts a microsecond or two later.

Actually, I have been looking mostly at the time before the pivot, but either side, before and after, give interesting information about the target.

I never looked at it with a log amplified signal. Could be very interesting.

With a DSP, my idea is to detect the time of the pivot of the ground by using an algorithm and then setting the pivot as baseline. This would constitute a ground balance.

Similarly, a look-up table could be made for typical target responses and the look-up table be used for an approximate target ID.

I first observed the pivot on the oscilloscope, using real coils and circuits. Later I tried to reproduce the results observed, with simulations. THE PIVOT IS OBSERVABLE EITHER WAY.

What produces the pivot?

How can we read the pivot?

How can we control the pivot?

I will try to answer these questions once I fully understand every aspect of it.

By the way, the pivot is not my invention. ERIC FOSTER mentioned and explained it many years ago on a public forum.

Tinkerer

[Davor]

Makes sense. Considering the voltage values at hand you most probably need no amplification to do that at all. Garden variety ADCs are already balanced devices so your balanced frontend could do just fine, even without amplification. You could even observe the "on" period for zero crossing and establish ground balance directly from there. With bipolar pulsing and your fast repetition approach. Yeah, perfect.

Funny, but all of this is actually possible by mere window comparators, just as this SPI thing does.

[Tec]

The Pivot

Davor,

about "the pivot":

What I call "the pivot", is the point in time, where the target signal crosses the coil decay curve.
If we look at the coil decay curve with no target, as the 0 base line, then there is a point in time when the target response signal crosses this base line.
This means, that at this point in time, comparing the coil decay signal and the target response signal, we get 0, or no change.
At any time there is current running in the TX coil, the response of a ferrite target, (only X response) will be 180 degrees from the response of a purely R target.

Purely X response targets do not exist in nature. All kinds of steel have different magnetic properties (or permeability) and also resistance that causes eddy currents (R), and therefore the phase angle is different from 180 degrees.

The pivot, or point in time where the target signal crosses the base line, will therefore be varying with the phase angle.

So, in other words, "THE PIVOT" can be considered an indicator of phase angle.

To what degree can a PI detector define or indicate a phase angle? In the time domain? I think it could be interesting to investigate this assumption.

It might be a good idea to start a PIVOT THREAD, where we try to analyze the pivot and it's implications and the possible benefits of observing it might provide.

Tinkerer

I scanned this a while back, (the 400uSec pulse is irrelevant), but the response of different targets is the main thing I guess.
The X axis is 0V to 65mV threshold voltage.
It is not balanced, so is is measuring the curve lifting up as well as shifting sideways.
A balanced will look more interesting, I would like to plot one and see how it looks.

(X an R response of balanced systems is not my area at all, or the phase shift of sine waves vs amplitude change.

[Davor]

So your curves disintegrate below ~5mV. OK, it gives you a whole decade to work with. Now just apply some averaging of the samples, even addition should do, and plot the curves for 5 to 50 mV range. This looks a bit hairy.

BTW, did you optimise damping resistance?

[Tec]

So your curves disintegrate below ~5mV. OK, it gives you a whole decade to work with. Now just apply some averaging of the samples, even addition should do, and plot the curves for 5 to 50 mV range. This looks a bit hairy.

BTW, did you optimize damping resistance?

They are old plots from a different system (1 year ago)

Here are some recent plots (today) from the SPiMX.
All targets are 15cm from the coil in the center.

On a fast coil, 5 mV crossing may occur at 10uSec, on a slow coil 5mV crossing may occur at 50uSec

So, on this particular system and coil between 1mv and 8mV corresponds to 17uSec to 41uSec

I have set it up to use the signal at 30uSec (2mV) to find coins at the beach and park.


If you want to find a piece of gold the size of a match head in a pit of powdered iron, just make a faster coil and a use smaller pulse width.

[Davor]

Precisely! It all is about finding small stuff in a difficult soil.

I like these. IMHO it would be very interesting to have such measurements with a log amp in a front end. As learned recently these provide some integration as well, which is also beneficial as it accumulates target response from the pivot on.

Boy! I like paradigms being put upside down

[Tec]

Conductivity.

I would like to add the most important point I have found in measuring this way (using voltage threshold on comparators).

When measuring the change at one point compared to the change at a later point (when moving a target closer)...
The ratio is constant (fairly).


ie The conductivity is constant over distance.
Better that I attach the measurements and a graph, (it is the only way I can understand even basic concepts)

The graph means that a bottle cap always looks different to a one dollar coin regardless of distance, and the lump of iron looked like a 20c piece in this case, so I dug it up.
I can sweep over targets and know what fairly accurately their conductivity is regardless of depth (except for iron which is confounded by the magnetic field)

[Sean_Goddard]

Interesting you chose the voltage decay / time route.

Here's something I did about 15 years ago, though wasn't sure what I was seeing. I know better now.

Set your oscilloscope up to trigger on the NEGATIVE going edge or the Tx pulse (the Rx curve portion if you like) now look at the position of the leading edge of the pulse (OOPS, you need a DSO with pre-trigger for this) the leading edge move a LOT, one way for Ferrous and the other for NON-Ferrous (REPEATABLE). I agree noticed that the signal also seemed to, as you say, "pivot" around a point with ferrous "rotating" anti clockwise and lengthening the "tail off" or coil discharge, non-ferrous object seemed to have the opposite effect or clockwise "rotation".

It's hard to explain but I'll have a go; With clockwise (NF) rotation, the dv/dt of the INITIAL decay was slower, or the top of the decay curve "fatter" whereas the point below the rotation point was much lower than normal. See what I mean? If you have a DSO look at the output of a BALANCED PI coil (Tx and separate Rx coils) you will see what I mean.

I came up with the idea (in a previous post on the forum) of NEGATIVE pulsing the coil (using bipolar pulses) four or five times to “polarise” the target and integrating the result, THEN a single pulse on the POSITIVE and subtract that from the previous result, if the target is NF you should see a small residual if the target is ferrous, it won’t be able to decay sufficiently fast enough (so in a way you are measuring the tau of the target) – see BH hysteresis curve thus the target will give a large(ish) negative result.

Anyone want to play with this? Your existing PIC32 based PI could do this VERY easily. Maybe you could post results if you get anything of interest.

[scrungy_doolittle]

I know this is a very old thread, but I work on an instrument that uses a flyback transformer to produce high voltage pulses to be applied to the human body for pain relief.
Without the damping resistors, you have a damped sinusodial wave form. If you play with this, you will notice the the "swings" around the 0 point become wider, and fewer decreased impedence. That is the more conductive the target, the wider the swings....We use a voltage divider, feeding a couple of gates to convert these to a square wave, and I suspect that something similar will give you a bit of discrimination...