Plans for my own PI

[sled]

Hi,

I'm a impassionated "treasure hunter" since 2 years. I haven't many electronic skills (except the stuff from the physics lesson). But I'll try to design my own microcontrolled-PI with a good perfomance. I already have skills in programming (especially C/C++, Java, Perl, PHP).

I studied a few schematics of PI's (hammerhead, goldpic and the one from oldradiobuilder.com). I also read a technical book about metal detectors.

So here's my idea:

I want to use 12V as main input (perhaps from a lead accumulator). I can boost it up with a step-up converter, so that I can drive my modules (square wave generator, sample processing and user inputs).

A big capacitor gets loaded during the delay. Then the transistor (what kind of transistor should I use? Nmos? Pmos? or just a normal Bipolar Transistor?).

The pulse is damped by the damping resistor. There's also a transistor after the coil, which controls the way to the signal processing module. After the pulse, there should be a high voltage peak, caused by the induction of the fast disconnect from the power source (am I right?).

I want to analyse the signal after this peak. Therefore I have 2 timers. The first timer is responsible for the main pulse (actually the frequency of the detector). The second timer is for the sampling window. Timer 1 and 2 are working delayed.

The voltage curve, which is induced by the metal object in the ground gets amplified by the opamp and then the signal will be converted by the ADC. After the ADC is the tricky point, shall I use a DSP for doing a FFT? Because I want to discriminate iron... In any case the ATmega will analyse the signal and show some information about the signal on a 2-line LCD. How fast has the ADC and DSP to be?

I read that you can discriminate an object by its voltage curve.

The first "idea" I had is attached to this posting. What do you think about it? The third picture is a scan of the book which I read, it show the voltage curve of different metals.

I would be really thankful if you guys could give me some advice


Thanks and best regards from Switzerland

Simon

[Geo]

Hi sled.
The third picture is very useful if you plan to make discrimination. It is about the same if the objects are small, but when the object is very big then the fall of curve is difference and you may make mistake between ferrous and no ferrous. I am not sure (i am not a PI guru) but as i know the manufactures that makes PI machines with discrimination use these curves as references point. With a good software you are OK.
Good Luck

[TheWizard]

How FFT help you to make discrimination... Can you explain?

[sled]

@geo: I think if the object is big and near, the amplitude will be very high, so that I can't discriminate it. Also VLF detectors have this problem, my Tesoro Cibola detects big iron objects although the disc is on.

@theWizard: I thought I could use FFT for a frequency spectrum analysis. But I don't know how useful it is. Without a DSP the project would be cheaper I think.... So if the DSP (FFT) doesn't improve the performance, I'll skip it.

[sled]

ok.... I was crawling through the internet for a while and I've found an interesting idea about the damping resistor.

Usually a fixed shunt resistor is placed across the coil. It's used to dampen or suppress coil oscillation caused by the LCR circuit formed by the coil.

The larger the shunt (damp) resistor, the greater the current dissipation and the faster the current decay. Fast current decay allows for small targets to be more easily detected since the coil has a higher bandwidth. But if the damping resistor value is set to high, current is forced into the coil where the capacitance and the inductancce combination causes voltage/current oscillation: the oscillation will mask small metal target signals. A small damping resistor slows down the coil decay and lower the sensitivity of the coil to small metal targets.

The solution is: a variable damping resistor!

The variable resistor can be made from a MOSFET transistor. The gate voltage of the MOSFET should vary. This, in turn, affects the resistance between the source and drain of the MOSFET.

The ultimate solution would be a variable resistor which is a function of time during the metal detection process.

Raising the resistance during transmitter coil operation will increase the kick-back voltage while reducing the resistance will decrease the kick-back voltage. When the coil is operating in receiver mode, raising the resistance will increase the coil bandwidth while lowering resistance will decrease the coil bandwidth. Increased sensitivity will optimize the metal detector for metal classification via time decay analysis.

If the first time decay estimate measures a target with a long time decay then the AVR (processor) can select a low damping resistor value from a pre-programmed look-up table. So that the time decay of the receiver coil more closely matches the target's time decay.

I could use a DAC which controls the gate voltage of the MOSFET that in turn controls the resistance of the source-drain path of the MOSFET. This implementation acts as a fast variable resistor. The fast resistance response of the MOSFET allows to dynamically adjust the damping resistor during the coil current change.

What do you think about this idea? Has anyone already made some experience he/she wants to share?

Thanks

sled

[sled]

I read some articles about operational amplifiers today. They seem very handy for the job of amplifing the TX and also RX signal.

I decided to use a TL071 low power opamp. I don't know if it's useful for amplifing the TX signal, perhaps the amp is too weak.

The gain of the RX amplifier should be between 100 and 1000.

My questions are:

- should I use an inverting amp for the RX signal? or a non-inverting? or something else?

- how can I protect the "sample module" from high-voltage peaks?

- do I need a RC/AC coupling ?? I mean do I have to convert the signal in a clean DC signal before amplifing it? The amplifier doesn't like high frequencies I think.

Thanks

sled


PS: perhaps everything is bull**** I made - don't know

[sled]

okk... I fixed some mistakes...

- added capacitor with 1000uF
- I mixed up the Vcc+ and Vcc- of the Opamps
- Offset nulling on the TX Opamp is not needed - I removed it
- Added the offset nulling on the RX amplifier, because it has a very high gain, so the offset voltage disturbs the signal.

[Carl-NC]

D2 & D3 need to go from R5 to GND.

IC2 probably doesn't have enough BW... you need 10MHz of more @ unity gain, I don't recall but doubt the 071 has more than 1MHz.

Not sure what you are attempting with IC1... you can't drive PI coils this way. The '071 doesn't have enough slew rate, and it also doesn't present an open-circuit "off" state.

- Carl

[Carl-NC]

The solution is: a variable damping resistor!

Recently patented, though I recall discussing it on Eric's forum years ago. Therefore, most likely, the patent is worthless.

- Carl

[bugwhiskers]

[quote=sled;54306]@geo: I think if the object is big and near, the amplitude will be very high, so that I can't discriminate it. Also VLF detectors have this problem, my Tesoro Cibola detects big iron objects although the disc is on.

Hi sled,

You can see from the first sample when the pre-amp is in saturation and you could use the micro to dynamically adjust the timing of the first sample so it is always just outside the saturation point.

regards
bugwhiskers

[sled]

D2 & D3 need to go from R5 to GND.

IC2 probably doesn't have enough BW... you need 10MHz of more @ unity gain, I don't recall but doubt the 071 has more than 1MHz.

Not sure what you are attempting with IC1... you can't drive PI coils this way. The '071 doesn't have enough slew rate, and it also doesn't present an open-circuit "off" state.

- Carl

Thanks for your answer Carl!

You're right, the bandwidth is really too small (~4MHz). So I decided to replace it with OP37 from Analog Devices (http://web.mit.edu/6.301/www/OP37b.pdf). It has a bandwidth up to 63 MHz at unity gain.

FEATURES
Low Noise, 80 nV p-p (0.1 Hz to 10 Hz)
3 nV/÷Hz @ 1 kHz
Low Drift, 0.2 uV/°C
High Speed, 17 V/uS Slew Rate
63 MHz Gain Bandwidth
Low Input Offset Voltage, 10 uV
Excellent CMRR, 126 dB (Common-Voltage @ 11 V)
High Open-Loop Gain, 1.8 Million
Replaces 725, OP-07, SE5534 In Gains > 5

An other alternative is the NE5534... And is one Opamp enough? Or do I need a multi-stage amplifier?

The first amplifier is stupid, I realised that. Because it would raise the voltage, but a higher voltage means no increase in the magnetic field strength. Only the current can boost the magnetic field, or a better coil of course.

Thank you very much for your help

best regards

Simon

[Unreg]

Recently patented, though I recall discussing it on Eric's forum years ago. Therefore, most likely, the patent is worthless.

- Carl

Even so, the idea with variable gain is as old as the mosfet or fet device itself.
USPTO are in decline indeed.

BJBJ

[Esteban]

Here a microcontroller project with the source code:

http://www.groegernet.de/schematics/metall1.html

[sled]

Hi,

thanks for your answers I've learnt a lot the last two weeks about electronics.... I'm going to use a uC to control the pulse and sample.

I use the ICL7667 for driving the MOSFET. The ICL7667 (Datasheet: http://www.maxim-ic.com/quick_view2.cfm/qv_pk/1485 ) is a power MOSFET driver which can be controlled by a 5V CMOS/TTL signal (output of the uC). It has a raise an fall time of ~20ns. It's also a cheap IC (2-3$).

I'll add the opamps etc. today, I hope I haven't made any fatal mistakes so far. I'd be very happy if someone could check the coil-part of the circuit. Did I connect the diodes / resistors correctly? Are they "grounded" correctly?


I'll add a plastic foil keyboard for adjustment.

Thank you very much

best regards

sled

[bugwhiskers]

HI sled,

If you plan to run the output of the amp stages to the micro's AtoD the circuit you have drawn won't work because the decay curve sits on top of your 5 volt rail and the AtoD will always read maximum.

If you change the MOSFET to a "P" channel device then the decay curve will go below ground and an inverting pre-amp will bring it in range of the AtoD.
You will also need a -5 volt supply for the ops amps and this can be generated by the spare gate in the MOSFET driver chip.



regards
bugwhiskers