Chaotic Metal Detector

It was very good weather last week so I got the strimmer (frullino/weed whacker/whipper-snipper) out. I’d only taken a few steps before the cord spool flew off. One of the cord guides disappeared into the undergrowth. A small alloy part, about 2cm square. I searched the nearby area, sign of it at all. What I need is a metal detector.

So I did a bit of reading, a bit of thinking and a bit of breadboarding I have one to try, once it’s stopped snowing…

It only uses 8 standard components!

[youtube https://www.youtube.com/watch?v=RKsjotjh-Bk&w=560&h=315]

Circuit & description below.

Most current designs fall into one of three basic design types: BFO, PI and IB/VLF. All three rely on one or more coils acting as an inductor, the target metallic item affects the magnetic field which is somehow indicated.

There are loads of designs available online, a good few and some background theory at Geotech. Some of the best write-ups can be found in the magazine articles archived at americanradiohistory.com (an awesome site for any electronics tinkerers).

To give an idea of how things things generally work, here’s a bit of background. Note that the terminology can vary quite a bit, commercial designs emphasizing different aspects.

Beat Frequency Oscillator

Probably the oldest design, the same technique can be found in radios.

This relies on two fairly high-frequency oscillators, usually sine wave. One is a fixed frequency call it Ff. the other, call it Fv, has it’s frequency determined by an LC tuned circuit. The L in that is the search coil, the C being variable (either a traditional variable capacitor or a varicap diode).

The output from the oscillators is mixed by a subcircuit that isn’t simply additive, it typically approximates Ff x Fv. This leads to an output which will include not only the input frequencies but also components at Ff + Fv and abs(Ff – Fv). So it Ff is 100kHz and Fv is 101kHz, the output will include some 201kHz and some 1kHz. The latter is audible.

If the inductance of the search coil L is varied, so Fv will vary and so will the difference, leading to a pitch change in the output.

(An interesting trick I found in one design was to make the variable oscillator produce square waves with a fixed oscillator running at a harmonic of the the variable’s fundamental.)

The circuit can be relatively simple although sensitivity is limited. Physical setup should be straightforward, maybe a bit of tweaking to get the frequencies close. Note the search coil should have a Faraday shield to minimise capacitive effects. (Wrapping the coil in aluminium foil – with a gap in the loop – is the usual technique).

Inductance Balance

Sometimes known as VLF (Very Low Frequency, think audio range) this typically relies on two coils, one acting as a transmitter and the other as a receiver of magnetic signals. Some designs have the receiver coil perpendicular to the transmitter.

A VLF signal is sent to the first coil and picked up by the second. The received signal will be influenced by the environment and differ in phase to the original. The phases are compared and some indication generated from this. If I understand correctly, different metals will produce kinds of shift, allowing some kind of discrimination between them.

The circuits for this type don’t seem especially complex, but the physical construction seems quite critical. It is claimed they are more sensitive than the BFO type.

Pulse Induction

This design uses yet another trick for detecting changes in inducted fields. A brief, relatively high-powered pulse is sent to a coil. Immediately after the pulse has ended, a detector coil picks up any environmental field. If there’s a piece of metal nearby it will be stimulated by the impulse and effectively echo it. Most designs use signal switching to enable the same coil to be used for transmission and reception.

The circuits for these can get quite fancy, thanks mostly to the timing/switching required. Physical aspects shouldn’t be too critical and again these are claimed to be sensitive.

Chaotic

My main requirements are simplicity and sensitivity, main constraints time and the components I have at hand. As it happens I already had an eminently suitable coil I wound for another project. I’ve got a bunch of standard components including op amps and transistors. I also happen to have a brand new (joy!) breadboard, so I don’t have to strip down some other work-in-progress.

So initially I had a play around, trying to get a induction balance design working. I hacked something together using the existing search coil and a coil I salvaged from a transformer ages ago. Ok, I only tried a few experiments, but I couldn’t get anything with remotely enough sensitivity.

The effect of a small metal object a few cm away from a coil is tiny.

Meanwhile, I couldn’t help thinking of the phrase sensitive dependence on initial conditions. Why not try a chaotic circuit? An oscillator can be simplified down to some kind of amplifier with some kind of delayed feedback. The simplest chaotic circuits are just this with a bit of non-linearity thrown into the loop. The delay (/phase shift) part can be achieved using an LC circuit. Vary the L …

So I started with a version of Chua’s Circuit that used an inductor :

I could not get this to work. As noted at StackExchange, these things are sensitive to component values and may only have a very limited operating range. But I have played around with chaotic circuits in the past and did get simple ones running.

Short story shorter, I played around with one of the old circuits until I got it to detect metal. The circuit used in the video above is simply this:

(The coil measured about 1.5mH, 8 Ohm).

This produces nothing on some settings, a fairly high-pitch periodic wave on others and on some, something that sounds like filtered/distorted noise, the genuine chaotic signal.

I hadn’t really thought through the sensitive dependence… thing. If you pick an instantaneous value at some point in the chaotic signal and then after some tiny time delay, sure, changing the circuit parameters a tiny amount will yield a wildly different value at the second point. But that’s kinda impractical here.

But what does work, is turning the controls near to an audible bifurcaction point in the system’s behaviour.  The tiny change in inductance sends the circuit into a different path. You can here it in the video above, this is the same kind of thing represented visually:

I had no doubt someone had played around with something like this before. Sure enough, there’s : Chaotic Oscillators as Inductive Sensors: Theory and Practice. But this kind of thing doesn’t seem to have been explored very much. I guess a big reason is that due to the designed-in sensitivity, making replicable systems would be inherently difficult.

Remains to be seen if it can find the strimmer part.