Android Phone Based Metal Detector - FCMD

[Aziz]

Hi Paul,

Might have nailed it ... found a new method based on very elegant maths that cancels ground (or anything really) whilst simultaneously boosting desired targets.

This sounds good. We would be interested in the elegant math.
What are your K1 and K2 factors?
Aziz

[moodz]

here the ferrite is is held about 10 cm from the coil and the target is waved twice at 80 cm ( 10cm x 10cm foil ). ... very hard to pick the foil without the magic !

[moodz]

Just to prove I am not doing magic tricks like gating or something ... I bring the ferrite near the coil and keep it there ... then wave the target 3 times ... then remove the ferrite from the coil ... you can see this in the top trace. In the bottom trace the method removes the ferrite response leaving the target response ... Magic without tricks

[moodz]

Might have nailed it ... found a new method based on very elegant maths that cancels ground (or anything really) whilst simultaneously boosting desired targets.
Below is the output from a python app that shows removal of a ground signal ( nasty ferrites ) whilst simultaneously boosting a wanted signal ( non ferrous ) in the time domain !
The pyhon app is taking a 5 second snap ( 250 samples ) from the STM32 ADC signal chain and doing a parallel execution of the method showing the complete elimination of the ferrite / ground signal but leaving the wanted signal boosted.

[moodz]

OK its a bit bigger than a box of matches .. the surface mount unit would be smaller.

I added a 20x preamp prior to the ADC and the 10cm x 10cm foil is detecting at just over the meter in air.
The coil current is 1 amp peak to peak true biploar square wave with a fall/rise time of 3 microseconds.
The power consumption for the whole detector is just over 1 watt. ( ie less than 100 ma from 12 volts ).

The code does simultaneous PI and VLF demodulation.

It can be this small because the detector is a true software defined metal detector ( SDMD ) consisting of CPU, bipolar transmitter, preamp and power supply
... there are no switches, demods, integrators or damping resistors.



transmit voltage waveform

[ivconic]

https://www.geotech1.com/forums/foru...-digital-audio

[ivconic]

[Aziz]

Hi ivica,

AI usually hallucinates too much.
I'm afraid that those mathematical dubios will show badly in the real field.
...
Look at those "tax leeches" in Cern; they have been achieving incredible feats for decades...none of which have any useful application in life!
I have the perfect AI task; to create a mathematical model for LRL!

well, I don't think so. AI gives us indeed some good tips, which can be checked and peer-reviewed in real world.
I like the idea, where coefficients and parameters become all complex numbers. Or all the GB math doing it in the complex plane.
It's not complex doing complex math. It is even easier do the math in the complex plane.

[ivconic]

AI usually hallucinates too much.
I'm afraid that those mathematical dubios will show badly in the real field.
...
Look at those "tax leeches" in Cern; they have been achieving incredible feats for decades...none of which have any useful application in life!
I have the perfect AI task; to create a mathematical model for LRL!

[Aziz]

Hi all,

if my IB-wideband test measurement platform is finished, I'll provide full complex frequency spectrum responses for different ground and target types.
The complex FFT delivers a complex frequency spectrum either. Each for TX reference and RX signal. So a coherent RX signal spectrum can be extracted.
All tracking parameters will be complex numbers.

We can ask the AI mate, how he would solve the GB problem in detail.

[Aziz]

Hi Paul,

The AI is on a roll ... here is an update with waveform analysis.

WTF!

I'm impressed. Can I borrow your AI mate?

Well, ok, we are going to calculate with complex numbers for the GB algorithm next time. I'll prepare my software doing so.
Thanks for your effort. Interesting reading...
Aziz

[moodz]

The AI is on a roll ... here is an update with waveform analysis.

[moodz]

my AI mate has provided this analysis ... he read your answer above Aziz and he knows how my detector works.

[Aziz]

Thanks for your comments Carl,

I see two flaws. First, dS = X(dS) + R(dS) is not true. It is



Second, "Grounds affect the X and targets R. Iron target both" is also not true. Ground is primarily -X but usually has some +R mixed in* (loss angle), and the loss angle can vary as you move around which is why ground tracking is needed. Tiny gold nuggets and foil are primarily +R but big silver coins are primarily +X, and other non-ferrous targets can have any mixture of +X and +R. Smaller ferrous targets (nails) have a mixture of -X and +R, and large ferrous targets can easily move into the +X region. All of this is, of course, frequency-dependent.

To make GB work you absolutely have to account for loss angle. If you only balance out the X component you will have a very noisy GB.

*In saltwater it is the opposite: mostly +R.

I didn't look at the phase angles yet. As I am using the fabs() for changes, both X and R effects lose their polarity. The term "dS = X(dS) + R(dS)" should had been read as that the signal change consists of X response and R response change.
But the GB works very well. Even it is simplified.
We can do a better GB next time. I am preparing the test platform for making lots of measurements.

Aziz

[Carl-NC]

I see two flaws. First, dS = X(dS) + R(dS) is not true. It is



Second, "Grounds affect the X and targets R. Iron target both" is also not true. Ground is primarily -X but usually has some +R mixed in* (loss angle), and the loss angle can vary as you move around which is why ground tracking is needed. Tiny gold nuggets and foil are primarily +R but big silver coins are primarily +X, and other non-ferrous targets can have any mixture of +X and +R. Smaller ferrous targets (nails) have a mixture of -X and +R, and large ferrous targets can easily move into the +X region. All of this is, of course, frequency-dependent.

To make GB work you absolutely have to account for loss angle. If you only balance out the X component you will have a very noisy GB.

*In saltwater it is the opposite: mostly +R.