SPI-EG Detector

[Tec]

Hi,
I have designed a PI detector.
SPI-EG-1

It has a USB interface (but doesn't require PC to operate).

The design is different to most PIs I have seen
1. There is no pre-amp, or knob to threshold adjust or set gain, ..no conventional OP-amp amplification at all.
The signal is sampled directly at high resolution through a simple and stable comparator / timing system.
The signal threshold and amplification is all performed in software.
(Press a button and software will set the threshold).

Any part of the discharge curve can be sampled.
The sample is very narrow (fine)
The whole curve can be scanned and displayed on a PC (if you wanted to)

I have built prototype-1
Coil: 0.5mm wire 25 turns around a 200mm pram wheel, no fine tuning of damping resistor.

Results:
Sensitivity: Australian 20c piece - 160mm in air.
Works well in soil:
Found: decimal coins $2, 50c, 10c, 2c, copper bracelet, plenty of junk also at range of depths.
(Test Area is less than 20 years old consisting of a small park, beach including black sand, salt water)

Was noisy and unusable in wet salty sand..changing coil angles gave a signal.


I will be assembling the first production PCB next week, and testing 'production version 1' ASAP.

Tec

[TH Jim]

Hi there, your work sounds very interesting! Can you post a schematic of the circuit and the source code for your software?

Thanks Jim.

[Tec]

Hi Jim,


I have attached the circuit diagram,...

The main gist of it is: two resistor arrays (DAC) to set a voltage on the comparators.
One is a timing pulse, the other is the pulse from the coil.
The pulses out from the comparators clock a flip-flop, the flip flop output is 1 or 0 depending on which pulse is fastest.

One voltage sets the point on the coil curve to detect.
The other voltage sets the timing pulse.

The point at which the coil curve crosses the voltage threshold, is found by adjusting the timing pulse.
i.e, The time for the coil to discharge to a certain point is timed very accurately.


The processor is a Cypress PSOC, and the code is written in C.
The PC interface is just serial commands, so any language can will be able to talk to it.
There isn't any PC software yet.. (I will start with just a normal spreadsheet, that pulls the data from the detector using a Macro)



Tec

[TH Jim]

Hey Tec,

Thanks for sharing your work with us. Yes it is unlike any PI i have seen, thus very interesting. Will take me a while to digest it! Keep us up to date with any future developments please.

Cheers, Jim.

[Tec]

Making Iron dissapear

I have built and played around with version 1 ( SPI-EG-1 ), there were plenty of mistakes and things that needed to be fixed and it ended up looking like a messy prototype

I've learnt alot and made some changes, the most interesting thing I have learn't is how to achieve reliable discrimination between iron and non iron... I think.
See measurements attached (they basically show a coil discharging from right to left)

'The affect of pulse width on ferrous and non ferrous targets'
What is important is the affect of pulse width on the nail and wing-nut (both ferrous) compared to the copper coin (aus 2c) and nickel copper coin (aus 20c).

There are a few explanations around on the on google about how a PI detector works, and I think some of them are confusing or maybe wrong, I will give my explanation so far, I'm sure there are quite a few experts on here that can correct me, or already know this:

1. Metal targets do not create 'echos' like sonar, they interact with the magnetic field.

2. The way a Ferrous (Iron) object interacts with the coil pulse is quite different to non iron objects.
Iron is a poor electrical conductor, but has good magnetic conductivity (100 times more than non-iron metals),..ie Iron lets a magnetic field easily flow through it.

Non Iron metals are better electrical conductors, but have high magnetic resistance (same as plain air).

2. Non-iron objects do not get 'Charged up' by the magnetic field.
A current will flow in them only when the field is changing: at the start and end of the pulse,..i'e the edges of the pulse.

3. It is the increase and decrease of the field that causes (induces) a current (eddy current) in the non-iron object, this eddy-current is like a 'fly-wheel' that is 'kicked' by the changing field. The eddy current continues to flow for a while, as it decays it induces a current back into the coil prolonging the discharge curve slightly.. like a fly-wheel continuing to turn for a while.
The amount of current (the size of the fly wheel) depends on the conductivity and size of the metal object.


4. Iron objects increase the inductance of the coil. They allow magnetic field to flow more easily ( increasing the magnetic field strength). This affects the length of the coil discharge for a different reason than non-iron objects.
Iron objects ARE affected by the pulse width.
Because Iron has a 'magnetic inertia' it takes a while for it reach its maximum magnetic field conduction (it 'charges up'), so shorter pulses mean less field strength and less affect on the inductance of the coil, and on the discharge curve.



Summary
Reliable discrimination between iron and non-iron should be possible by adjusting the pulse width.

[Vague]

Great project. Your results are very interesting. Is the schematic you posted correct?
Thanks for posting this, it has given me a lot to think about. I think I will start building

[Tec]

Updated circuit

Latest circuit. A few changes, shows component values.

[Vague]

thanks Tec . Thats great

[Tec]

Varying pulse width to ID targets

I need to delete my post about, pulse width and iron etc.
It seems that iron targets look more like normal metal targets from a distance, quite hard to discriminate

I am interested to know if there is already a known technique of identifying targets by varying pulse width.

By varying the pulse width ( coil = 27 turns 20cm diameter) pulse varing 40uSec to 440usec.

Different objects have distinct step changes in signal at certain pulse widths.

The pattern for an object varies with depth.
But a 2c coin doesn't look like a 20c coin at any depth...
If that makes sense.
Maybe it is possible to scan an object at various depths and use the signature to ID it.

[Tec]

Improvements - reduce noise, use mosfet driver.

Don't use an opto coupler to switch a mosfet.

I have replaced the opto-coupler with a mosfet driver.

This has improved stability.
The opto coupler was causing drift and noise and was also affected by temperature changes (Because the opto coupler is slow to switch and has low current output).

Now that the mosfet is being switched by a mosfet driver, the signal is more stable (less low frequency noise and drift).,..because the pulses are now more stable.

I threw the opto coupler in the bin.
The detector signal is much more stable..added 5 centimetres detection depth to a 20c coin in air.

[Qiaozhi]

I need to delete my post about, pulse width and iron etc.
It seems that iron targets look more like normal metal targets from a distance, quite hard to discriminate

I am interested to know if there is already a known technique of identifying targets by varying pulse width.

By varying the pulse width ( coil = 27 turns 20cm diameter) pulse varing 40uSec to 440usec.

Different objects have distinct step changes in signal at certain pulse widths.

The pattern for an object varies with depth.
But a 2c coin doesn't look like a 20c coin at any depth...
If that makes sense.
Maybe it is possible to scan an object at various depths and use the signature to ID it.

The response of the target will also vary depending on the orientation. Try plotting the same graph for the 2cent coin with it placed at an angle to the coil. The huge number of variations in target composition, orientation and distance from the coil, will probably screw up your master plan, but don't let that put you off trying. If you could log all the data and analyze it later, then a pattern might emerge.

[bklein]

1. Any chance you can add a gold nugget to your test mix?
2. I'm curious as to if/how you designed your coil. What design rules or technique did you use to determine optimum inductance/series resistance/turns? Or did you just go with what seems popular...
3. Will you be publishing the PSOC code?
4. I'm wondering how much shorter pulse width TX pulse at higher frequency would compare. What is the frequency of your TX signal now?
5. Your theory is that an iron signal interacts with the coil and affects its inductance. This should be measureable during the TX phase then. Any intentions of experimenting with this?
6. Some part numbers for the schematic?

Barry

[Tec]

Yes, The interesting curve I got had a lot to do with errors in the Digital analog converter and interference. There would be many different orientations and depths to map out. Probably I am chasing ghosts. I have left the pulse-width-varying for now.

And looking at measuring resistance (conductivity)... This should be independent of depth, orientation and size of target.

It seems well known that the resistance of the target affects the response in the following way: Less resistance means longer signal, more resistance means faster shorter signal.

Measuring resistance (conductivity) seems feasible (results attached).
Method was the following:

Pulse width = 80uSec.
Measure discharge curve in two sections: between 40uSec -> 50uSec (Block A) and 50uSec -> 90uSec (Block B)

The difference between Block B and Block A should be affected by the resistance of the target (ratio Block B / Block A).
A calibration measurement was done with no target (BlockB / Block A)
this was 1.00

So according to the results,

2c copper coin is more conductive than a 20c coin.
Both coins are more conductive than the nail and wingnut.

at any depth in air form 0cm to 12cm I can tell the difference between a 20c coin, and a 2c coin, and a nail or wingnut.
Can't tell the difference between a nail and wing nut.

[Tec]

Hi bklein,

1.
I would like to add a typical nugget, I don't have one, but I have a friend who has, it would be a good test target, I may try and borrow it, or maybe steal it.

2.
For the coil I went with what seems popular for other designs.
I am currently doing testing and development with an easy to remember: 25cm, 25 turns, 0.5mm
The resistance accross it (damping resistor) has not been finely optimized it is 3 x 500 ohms in parallel = 166 ohms.

3.
Yes, once it is stable I will put the source code up.

4.
The pulse frequency is currently 200Hz, and the pulse width is 80uSec.


5. Maybe, I hear there was once a PI detector that did discrimination like that.
..probably could measure the pulse on the way up, worth looking in to...

[Vague]

Hi Tec
Can I ask you if the values in the second schematic are correct? C1 and R34 look too large. This would give a time constant of about 1/100th of a second and it would take about 1/20th S to charge capacitor fully or am I misunderstanding how the circuit works.
Keep up the good work
Thanks