Android Phone Based Metal Detector - FCMD

[Aziz]

Hi Paul,

The driver is based on the work I did here : https://www.geotech1.com/forums/foru...ansmit-circuit

OMG, this is a nice transmitter.

But I tend to use a very very simple TX-circuit having
- one mosfet only (n-ch, high voltage)
- simple switch timing
- power efficient (with energy recovering)
- overall simple and kiss design
- still wide band spectrum response on the RX side
- bipolar coil current (polarity change of coil current)
- the ability to push enough power into the TX coil regardless of TX coils reactance (max +1/-1 A is enough for beginning)

Oh well, this is my wishing list for the X-mas. I'm confident, that Holly Santa Claus has the gift for me in his bag.
I think, we already have such a TX-circuit somewhere here.

Cherrs and Marry X-mas to everybody.
Aziz

[Aziz]

Hi all,

I have looked at my old frequency response measurements in Excel table. Pity pity, It's a shame I didn't take more measurements with more infos. The measurements I have made are not convenient to show you. I can't remember, what I have done at that time. The great GB method could be lost forever. The only fact I know is, that it is based on linear combination factors with lot's of equations and variables and you even can't solve it. This is the reason for the lossy GB method.

I have to make the measurements once again sometime. With more infos and different target types and ground conditions.
Guess what! I need a new improved test board or have to search for the right one out of the dozens I have made so far. Where are they? I don't know.
Yep, Alzheimer.

Anyway, a good chance to improve everything. A new GB method too. At least I can peer review Pauls work.
Cheers,
Aziz

[Aziz]

Welcome back again Aziz ! Hey I connected a USB sound dongle to the phone USB port and android recognises it and you can write code ( using AI ) to do DSP in less than 10 minutes. You can even use the codec inside the phone ( no dongle ). Then I realised that instead of doing all the DSP in the phone .. I could do the same in a STM32 with onboard 16 bit ADC and just use the phone to display whats going on in the chip but also the chip will run fine if you disconnect the phone. That means that we can build a detector using the chip and few components that will fit inside a matchbox or a dongle ( plus the coils of course ). At the moment the chip does a fast syncronous sweep over any frequency range from 100 hz to 100khz at 2 to 50 discrete points then does a statistical best fit to calculate the phase slope ( degrees / khz ) and this gives us the relative phase shift caused by ground and targets at all frequencies. The ground slope ( obtained in a ground calibration sample ) is subtracted leaving a target slope that will indicate a VDI and tone to give us the target ID.
The AI is the answer to the maidens prayer .... it compressed months of coding into days .. if it could build hardware I would have this done by now.

That's fine. I prefer hand coded stuff.

I would like to peer review of the GB stuff shown here (but on my platform). We should get the same results I think.

BTW, the multi-frequency GB does even work with amplitudes only (A = sqrt(I²+Q²) ) . You have to look at the absolute value of amplitude changes (change = fabs(current meauserement - reference)) . So leaving the phase angles completely. This is practical, if you don't have any TX reference signal to correct the phase angles in the demods. But it s a lossy GB method. Just like PI GB, where portions of target signals can be subtracted out. I have to show the GB response change spectrum and the linear combination factors sometime.

So this is my motivation for a better GB scheme to minimize the target signal losses in heavy mineralized grounds.
Aziz

[moodz]

Hi Paul,

that's really an interesting project. I have enjoyed reading a bit (but not thoroughly).
I am looking forward to read more.
Cheers,
Aziz

Welcome back again Aziz ! Hey I connected a USB sound dongle to the phone USB port and android recognises it and you can write code ( using AI ) to do DSP in less than 10 minutes. You can even use the codec inside the phone ( no dongle ). Then I realised that instead of doing all the DSP in the phone .. I could do the same in a STM32 with onboard 16 bit ADC and just use the phone to display whats going on in the chip but also the chip will run fine if you disconnect the phone. That means that we can build a detector using the chip and few components that will fit inside a matchbox or a dongle ( plus the coils of course ). At the moment the chip does a fast syncronous sweep over any frequency range from 100 hz to 100khz at 2 to 50 discrete points then does a statistical best fit to calculate the phase slope ( degrees / khz ) and this gives us the relative phase shift caused by ground and targets at all frequencies. The ground slope ( obtained in a ground calibration sample ) is subtracted leaving a target slope that will indicate a VDI and tone to give us the target ID.
The AI is the answer to the maidens prayer .... it compressed months of coding into days .. if it could build hardware I would have this done by now.

[moodz]

So DDS to generate the waveform. Presumably you could generate any (repeating) waveform, not just sine or square waves, using suitable lookup tables?

Can you share the circuit for the constant current output vs frequency?

yep .. you can use any waveform you like from repeating lookup tables or you can calculate on the fly. The lookup table with DMA loading the DAC uses almost no CPU cycles.

The driver is based on the work I did here : https://www.geotech1.com/forums/foru...ansmit-circuit

Very easy to generate constant power spectrums upto frequencies way beyond 100 Khz that I will be using.
This method does introduce harmonics but the widest filter in the DSP is code is 100hz wide downto a few hertz so no problem ... plus all the harmonics are frequency / phase locked to the fundamental anyway.

[Aziz]

Hi Paul,

that's really an interesting project. I have enjoyed reading a bit (but not thoroughly).
I am looking forward to read more.
Cheers,
Aziz

[Repwoc]

So DDS to generate the waveform. Presumably you could generate any (repeating) waveform, not just sine or square waves, using suitable lookup tables?

Can you share the circuit for the constant current output vs frequency?

[moodz]

That sounds like a bit of a challenge. Are you using the STM32H7 on-chip DAC to generate the signal with an amplifier feeding the coil?

I use the onboard DAC driven by an NCO with .47 millihertz resolution.
The software can generate sine or square wave.
A 300uh coil has about 9 ohms of reactance at 5 khz and 188 ohms at 100 khz ...
Ideally we want at least a flat spectrum across 5 Khz to 100 khz a simple amplifier will give a spectral rolloff due to the increasing reactance of the coil with frequency.
However we can achieve a flat response by using an LR charging and energy recovery scheme in the driver so that the power spectrum will be flat across the desired frequency range.
So if our top frequency is 100 khz ( 10 microseconds ) then the minimum required rise time of the coil is 5 microseconds ( 5 t ).

Below are the plots of coil current at 100 khz and 5 Khz. If we sweep the frequency from 5 Khz to 100 khz the spectral energy is flat +/- 1 db.

What about the capacitance ? Yes it does affect the result but not in the way most people think ( resonances etc ) ... the capacitance effect is a function of the source resistance of the feeding power supply and the capacitance value.
A simple RC calculation where the time constant must also be be less than the time constant for the LR.



100 khz below




5 khz below.

[Repwoc]

That sounds like a bit of a challenge. Are you using the STM32H7 on-chip DAC to generate the signal with an amplifier feeding the coil?

[moodz]

How did it go with the modified coil?

Usually the hardware is quick and the software takes longer .. now I have found the situation is reversed. I am modifying the driver hardware to compensate for the increasing reactance of the coil with frequency so that the transmit power is flat across the spectrum from around 5 khz to 100 khz. The resonance point of the coil cant fall in this range so working on minimising the driver capacitance etc .... so getting there.

[Repwoc]

.. now we will need a broadband coil so we chop the capacitor out of our test coil Leaving us with two coils one at 3.2 mH and the other 2.6 mH ..
Its likely due to the high inductance that we will only be able to reach 50 Khz ...

How did it go with the modified coil?

[moodz]

If you click on "Facebook" you still can see the video clip. The other link seems to be broken.

Ivica could tell you if that guy has his deusII set up properly or not .

I dig everything ... VDI is a take it or leave it preference.

[GeoMax]

If you click on "Facebook" you still can see the video clip. The other link seems to be broken.

[moodz]

Target ID works?

Then you should watch this "learning video about targets and target ID's". This video actually confirms what I monitored about "traget ID's" in many metal detector videos.
Facebook link (www.facebook.com/reel/1280100617492805)

... the link seems to be broken. Personally I am happy with Iron / not iron.

[GeoMax]

Target ID works?

Then you should watch this "learning video about targets and target ID's". This video actually confirms what I monitored about "traget ID's" in many metal detector videos.
Facebook link (www.facebook.com/reel/1280100617492805)