SPI-EG4 !

[Tec]

The latest circuit has alot of changes, so I have made a new thread..

I will try different coils (Induction Balance) as MikeBG has recommended in the future, but currently I am working on the circuit; discrimination.

The last build (SPI-EG3) was a PI detector like others, with adjustable sample delay, sample width. I set it up with the coil to be most sensitive to a 20c coin (happened to be sample delay ~ 35uSec, sample ~ 10uSec).

I tried it in two locations:
Old tree in a park,
Sisters house.

Old tree: countless ( 8 ) pulltabs, 4 bottle caps, misc rusty junk, and one 1946 shilling, half a day of digging and I didn't even get around the whole tree.

Sisters house (100 years old),: couldn't walk one foot, without a target, hunks of iron, spanner, one cent piece, got tired of digging.

My conclusion: must have accurate discrimination.
If not discrimination between ferrous and non ferrous, then definitely discrimination between a pull-tab and a shilling, large target and small..
Not possible with PI ?, rubbish.

So..



Main improvements - SPI-EG4:

2. SIMULTANEOUS SCAN of 16 sample delays.
Instead of sampling in two positions on the discharge curve, now sampling at 16..
a 1 to 16 multiplexer switches the signal to 16 capacitors..each capacitor represents 5uSec of sample, at a different sample delay.
This can be used to either quickly plot (if PC ocnnected) or change the tone, target proximity will change the volume...target conductivity will change the tone.
Resolution will hopefully be enough to hear (see) different target types based on where the peak signal occurs (across the discharge curve).

So this could be like having 15 detectors (the 16th sample is reference) all running simultaneously, each with a slightly different sample delay.


Other improvements:
6V Zener diode on the mosfet between gate and source, stops the mosfet being switched on momentarily by the high voltage discharge spike.


Have attached the circuit diagram.
Will assemble the PCB in 1.5 weeks.


Other:
Has anyone on here used Hall sensors on a metal detector ?

Tec

[6666]

Hi
a fine project you have there, I guess you are here in Oz seeing you are using jaycar bits

[goldfinder]

hall sensor = not useable

A hall effect sensor is not sensitive enough to do any good. A fluxgate magnetometer will detect down to at least 1/50,000 of earth field but from my experiments it is worthless also. For small pieces of iron (nails, etc.) there is not enough interaction with the earth's magnetic field to be detectable. A fluxgate is only useful for large objects that will affect the earth's magnetic field. Hall effect sensor and fluxgate are passive.

Now you might be able to combine the fluxgate and the PI decay curve and get some synergistic action that would make detailed discrimination possible but your faced with DSP of the data in real time.

You best option is if you are collecting 16 points on the decay curve of the PI coil is to run a lot of tests to see how various objects affect the shape of that decay curve. There should be enough differences to make at least a ferrous vs non-ferrous discrimination. Further, you might have to do some kind of cross-correlation analysis and digital signal processing to get more discrimination.

There was someone on this topic who used some PIC micro-controller doing DSP and achieved a very good level of discrimination but I don't remember the thread. The person sort of evaporated off the board. What he had was probably so good he cashed in on his discovery.

Good Experimenting - there is pony somewhere in all that BS!
Goldfinder

[mikebg]

Tec, no need of 16 samples. Four samples are enough to calculate color coefficients of ground and of target:
http://www.geotech1.com/forums/attac...5&d=1185776553

[Tec]

Thanks for the info on Hall effect sensors.

And the PD3200 manual.. I found the circuit diagram in another post
http://www.geotech1.com/forums/showthread.php?t=12736
I will study it.

[mikebg]

Thanks for the info on Hall effect sensors.

And the PD3200 manual.. I found the circuit diagram in another post
http://www.geotech1.com/forums/showthread.php?t=12736
I will study it.

Tec, I placed the last circuit diagram of PD3200 in other thread:
http://www.geotech1.com/forums/showp...0&postcount=70

[Tec]

Been working on it

I haven't given up..
I have been working on it, like a machine .
What a complicated nightmare...but.
After some circuit changes (that I will post), it is starting to work.

I have it auto-zeroing so.. must be waved over a target to detect it

I was able to give it a quick test tonight (outside on ground), it is very quiet, but gives a nice clean signal on a tiny Aluminium rivet head (~10cm distance)

Looking like an improvement on the other system. It picks up smaller targets mostly due to the coil and sample timing.

The coil is 'Faster': Less turns but slightly larger in the same style as the last coil.
(will post pic)



Sampling 7 points (not 16) on the discharge curve.
Each sample is 10uSec wide (total 70 uSec)
The first sample happens 20 uSec after the coil is switched off.

I think taking multiple samples might be better than taking one wide sample...
for various reasons, that I can not explain without a diagram.
But it will give less noise..maybe 7 times better than one sample ?

because each sample is subject to its own noise and ADC cycle, so the noise tends to average out..i.e taking 7 samples on one curve gives an averaging effect just like taking 7 samples on different pulses would..

Anyway:
The total signal level is the sum of the samples like this:

Signal =

((Sample1 / sample1zero) -1 )+
((Sample2 / sample2zero) - 1) +
((Sample2 / sample2zero) - 1) +

....
etc



The conductivity can sort of be calculated also, but haven't got it working well yet.
small fast resistive targets have bigger response early (sample1)
And bigger, slower more conductive targets have bigger response later.


I have found for anyone that doesn't already know the most important things in a PI system beside the coil are:


Damping resistor: Make it perfectly matched (use an oscilloscope, I just got a USB digitech scope from Jaycar on special, it's great)

Mosfet gate resistor: If the mosfet switches on TOO quickly can get oscillations and spikes and ringing, if it switches on too slowly, then the mosfet will get warm.
100 ohms was about right on mine


I'm going to do a better test tomorrow.

[Tec]

Field Test (circuit attached)

Tested in same place I have tested previously (small sandy beach),
found a new set of smaller (faster) targets.
20c, 5c, tiny sinker.

Pictures attached.
Circuit attached.

[Tec]

mmm, no comments on my brilliant work .

Well here's a thought, I have been thinking about 'coil shielding'.

Honestly, I don;t know what this is supposed to achieve. I can see no reason to put conductive or semi conductive tape etc around your coil..

But, I have not tried it.

When I rest the coil on wet-sand or push it against the ground, it does get a signal..
Is shielding meant to get rid of this ?

The only think I can see that shielding would achieve is shorting out electrical (static) difference between the ground and circuit ground (0v)...
in which case, a single wire close to the coil would achieve the same as carefully wrapping stuff around the coil.

i'e, a wire wrapped around the loop, to dischare static (like a lighting rod), but not a closed loop of wire.

Another thing wrapped shielding would do I guess, is increase the capacitance of the coil slightly, this would make it less sensitive to higher frequency signals..(ground noise ?).

So what is it ?

[Vague]

The shielding is to act like a faraday shield to reduce pickup from stray electric fields. i built a circuit similar to one of your earlier circuits using a comparator. I thought it was working great because every time I put a piece of metal in the coil I got a huge change. Then I tried with just my hand I got just as big a reading. I had built a brilliant hand detector!! I shielded the coil and the response disappeared. Like you I thought that putting metal in the coil of a metal detector would be worse than useless but I see now that provided that it is not too good a conductor it could reduce noise substantially.
I have been following your thread with great interest and find your circuits really inspiring.
Please keep posting about your progress
Thanks
Vague

[moodz]

Sheilding

Very good work. I agree with comments above though ... shielding for this type of configuration is essential.
Moodz.

[Tec]

Discrimination

Hi..
I have attached the circuit showing the revolutionary guts of the detector as it is right now. Most of you with electronics skills will understand it.

I have been working on Dual audio : Two tones with independent volume (circuit also attached).

In Australia,.. typical targets in order of conductivity (lowest to highest) are:

Foil
5c
pulltab
Bottle cap
Aus 20c
Aus 50c
copper nickel (AUS $1, Aus $2)
Sinker
Copper coins


So...
I Set up the audio like this:

Tone 1:
Frequency1: if pulltab or smaller
Frequency2 Low if 20C or larger

Tone2:
Frequency3 if copper coin or smaller
Frequency4 if really large target.


Frequency 1 and 2 and 3 are in harmony, frequency 4 is slightly off..

So this means that there are distinguishable sounds for different targets, I will post a movie / demo soon.

But, I tested it and got the following results:

Could recognize pulltabs and bottle caps, sinkers, and 50c pieces
Couldn't handle being wet or touching wet sand.
Worked ok on iron stone, (detect a small target in iron stone).

The circuit is there so you can see how it works..
The op amp method has potential.. don't tell anyone.


Tec

[6666]

Great project you there Tec

Mosfet gate resistor: If the mosfet switches on TOO quickly can get oscillations and spikes and ringing, if it switches on too slowly, then the mosfet will get warm.
100 ohms was about right on mine

interesting about the ringing, what voltage are you driving it with ?
at what point in the cct did you see the ringing ?

[Tec]

Swithcing the mosfet

Hi 66666,

I didn't measure when the ringing stopped which is a bit lazy, but.

I am switching with 10v from a mosfet driver through 100 ohms = 100mA.

When I was using 27 ohm gate resistor it was ringing. Now it is not.

The mosfet is the IRF720S everyone seems to use.
It has a gate charge of 63nJ,..so apparently if I switch it with 100mA, then 63nJ / 100mA = 0.63 uSec.

So.. switching it on and off in around half a micro second is fine I think,..the pulses are 100uSec so the switching time is small in the context of this application.
The switching frequency is no more than 1kHz and the mosfet is not getting warm.

So I figure it is fast enough, and not ringing, it looks nice and clean on the scope.

I would like to try to make it as fast as possible..find the perfect resistor, and also understand why there was the ringing/spikes.. but now

the pulse is 120uSec, the discharge takes about 100uSec and the on /off edges look straight up and down, nice and clean....

[6666]

Thanks for the info Tec, much appreciated ,
I will add this to my list of things to check with my experimental projects
cheers