Let's made a PC-base metal detector with usb interface !!!

[Aziz]

Hi Simonbaker,

Excellent, more types of info combined is good approach!

I like idea of measuring detuning of TX -- a little different from BFO because only phase shift, not freq -- that is why I prefer MD designs where TX coil is not part of oscillator, rather drive it from simple relaxation oscillator, etc. I think Lobo like that.

I am interested in how well you can make a metal detector with just TX coil alone, very high Q, and measure phase, amplitude changes -- just to see.

Regards,

-SB

Did you try the following experiment?
Combining the off-resonance and BFO type MD. Well, I didn't yet. Maybe this could give an interesting result.
The coils will be driven in micro-power mode (very low energy on transmitter side). So the sound-card will have enough energy (driving current) for this. The TX coil has its own resonant tank (LC). There is an another RX coil with same diameter, which is operated not in balanced mode (very easy to build). TX and RX coils share the same position (direct coupled) but they are galvanic decoupled (seperate coil). RX coil has also resonant tank (LC) with much more windings to amplify the TX signal to adapt to dynamic range of input signal (ADC). The RX signal level should be at 90 % of dynamic range of ADC. Transmit frequency equals to the receive frequency (matching of two resonant tanks).
Any metal objects nearby the coils will detune the resonant tanks, changing the phase and changing the RX signal level (due to losses). The coils should have much higher Q (Q = L / R) to give more sensitivity. The reference signal is generated internally in the computer.

Just a crazy idea. Well, I will try this experiment soon. If this works good, then the coil are very easy to build. No more balancing necessary.

Aziz

[simonbaker]

Hi SB,

my balance findings relates only to Laptop Digital Lock-in Amplifier VLF MD. On the synchronous detectors used in standard VLF MD's may differ from this.

There is a very significant difference between my and the standard VLF's:
My reference is not changing whereas the standard VLF reference is changed on metal targets nearby the search coil (changing relative permeability of the coils will change the operating frequency slightly). So I am measuring also the detuning of the transmitter side (like the BFO).

My reference signal is generated internally and tiny fractional parts of phase changes from reference to RX signal can be measured (1/100 of degree). So the behaviour for discrimination is quite obvious using very accurate measurement systems like the lock-in amplifier.

Aziz

Excellent, more types of info combined is good approach!

I like idea of measuring detuning of TX -- a little different from BFO because only phase shift, not freq -- that is why I prefer MD designs where TX coil is not part of oscillator, rather drive it from simple relaxation oscillator, etc. I think Lobo like that.

I am interested in how well you can make a metal detector with just TX coil alone, very high Q, and measure phase, amplitude changes -- just to see.

Regards,

-SB

[Aziz]

Hi SB,

my balance findings relates only to Laptop Digital Lock-in Amplifier VLF MD. On the synchronous detectors used in standard VLF MD's may differ from this.

There is a very significant difference between my and the standard VLF's:
My reference is not changing whereas the standard VLF reference is changed on metal targets nearby the search coil (changing relative permeability of the coils will change the operating frequency slightly). So I am measuring also the detuning of the transmitter side (like the BFO).

My reference signal is generated internally and tiny fractional parts of phase changes from reference to RX signal can be measured (1/100 of degree). So the behaviour for discrimination is quite obvious using very accurate measurement systems like the lock-in amplifier.

Aziz

[simonbaker]

I don't see yet that precise nulling is critical for phase detection.

Start with synchronous detector reference: square pulses with phase = zero degrees, freq = w.

Let r == nulling residue signal at RX coil after nulling, == Rsin(wt + a), were R is amplitude, a is phase relative to reference, t is time.

Let g == target signal at RX coil, == Gsin(wt + b)

Rx signal is sum of residue signal plus target signal == r + g == Rsin(wt + a) + Gsin(wt + b)

Now let us put Rx signal through Synchronous Detector and integrate over one pulse, one half cycle of square wave, same as integrating x from 0 to pi.

Integral [ r + g] = Integral [r] + Integral [g]

Depending on phase a of signal r, the integral will be a constant, we don't care what, call it A == integral of residue signal.

Integral [g] = integral [Gsin(wt + b)] = G * integral [ sin(wt)cos(b) + cos(wt)sin(b)] (0 to Pi) = G * (2 cos(b) + 0) = 2G*cos(b) .

So the voltage from the nulling residue is a constant A, and the voltage from the target signal is 2G*cos(b), where b is the phase angle between the reference signal and the target signal. So "relative" (motion detector) phase is very important; residue is constant, we just want to track change in detected signal due to target, don't care about nulling voltage.

The point is that this math says it doesn't matter where you null the signal for detecting phase, it will be some constant. However, because the Tx and Rx coils pick up correlated noise, careful nulling should reduce the common noise.

We actually have two reference signals in a quadrature system like TGSL, so we get two voltages: 2G*cos(b) and 2G*cos(b + phi). We know what phi is because we control it, typically around 90 deg. We will also have two residue voltages for each channel, A1 and A2, but they are constant. We can solve for G and b as usual (at least Aziz can because in software).

Bandpass filter gets rid of A1 and A2, but has to be chosen carefully. That is the secret why exact null point not so important (maybe, many different opinions I'm sure).

That's all I see. I don't know why gold is different unless it makes same phase shift as something else. I don't see why details of nulling so important except to reduce noise, based on math.

But when I look at real signals, and I null like Aziz says, right down at crossover point, signal seems to shift very fast with target. But I guess the change in voltage through the detector is the same.

This is just a start to make discussion. I don't know how true to reality or if important consideration was left out.

Cheers,
-SB

pretty boring, I know

[Aziz]

Hi Simonbaker and others,

I have now the optimum balance position for the DD coils. It is really at the minimum residual RX coil voltage for best iron/non-iron discrimination. I have tested different type and shape of metals including gold ring and a small gold coin. I was not observing the absolute phase angles but the phase direction change. The optimum balance position is for iron/non-iron the turning point of phase change. The RX coil voltage will increase at any type of metal at this turning point. Only the phase direction change will give the discrimination information. It doesn't matter, if the coil can not be nulled exactly to zero volt. Any residual voltage should work but it must be the best minimum possible one. Shielding, cables, coil wires, coils itself and any other metal type objects nearby the search coil may prevent the RX residual voltage to balance to zero volt.

If the coil is not balanced well, you may have same behavior for iron and gold (I was quite confused)! Even if you can discriminate iron and aluminium or copper, gold/iron may not work. So pay attention to correct balancing.

Ground effects may shift the phase of the RX signal. Working on relative changes instead of absolute phase angles, gives a better and easier implementation.

I know, nobody can make a coil at the minimum residual RX coil voltage. So a balancing network is a must have feature. I will include the previously introduced balancing network in this thread to the coil arrangement. If any mechanical or assembly distortions occur to the balanced coils, this can be corrected later. Also the turning point of phase change can be adjusted correctly. So this will give the best discrimination option.

The very difficult part of the whole project begins now.. The damn search coil.

Aziz

[simonbaker]

Hi Simonbaker,
...The most phase sensitivity occurs on minimum RX coil voltage...
Aziz

That is what I'm trying to figure out. In one sense, it seems obvious that better null makes more phase sensitivity. But on other hand, each part of signal (null residue and target signal) adds a voltage when detected by Synchronous Detector. These voltages should be independent in a linear system. Since null residue is constant, shouldn't the bandpass filter remove it and we get the target signal the same, regardless of null residue size?

Regards,

-SB

[gwzd]

The impedance of the RX signal. If there is more residual voltage, the high impedance signals will suffer and the residual voltage ....

The impedance of the RX signal??!! That was a good one!
Regards
R.A.

[Aziz]

Hi Simonbaker,

I just thought of something I should have thought of before - but need to check with calculation.

It's this -- null residue + coin signal = RX signal. So I worry because combined phase depends on relative amplitudes as well as phase of the two signals that combine.

But I forgot - that is only problem if detect phase using zero crossings. Using synchronous detector, integration turns phase into voltage. So assuming linear system, integral of sum = sum of integrals. So increment due to coin signal is same regardless of null residue signal. Since it is a motion detector, constant null residue voltage is removed by bandpass filter.

So maybe not worry so much about size of null residue as long as it not too big to go out of diff amp input range.

Is that on the right track or am I smoking something...?

Regards,

-SB

the phase dependency may not be linear behavior on the residual RX coil voltage. The most phase sensitivity occurs on minimum RX coil voltage. But there is another topic to be taken into account:
The impedance of the RX signal. If there is more residual voltage, the high impedance signals will suffer and the residual voltage will dominate and will therefore decrease the overall detection performance.
Generally, the better the coil adjustment, the better the sensitivity it will be.

But I will see soon more results on coming days.

Aziz

[simonbaker]

Hi Simonbaker,



I have not decided yet, where to set the residual voltage and phase of the RX coil. As far as I observed, there are many possibilties with some drawbacks. As a fact of lock-in amps, I need some residual voltage on the RX coil to keep the phase noise low. My residual voltage can be between 2% to 80% of the dynamic range of the ADC input.
I will investigate next week, which option is better.

Regards,
Aziz

I just thought of something I should have thought of before - but need to check with calculation.

It's this -- null residue + coin signal = RX signal. So I worry because combined phase depends on relative amplitudes as well as phase of the two signals that combine.

But I forgot - that is only problem if detect phase using zero crossings. Using synchronous detector, integration turns phase into voltage. So assuming linear system, integral of sum = sum of integrals. So increment due to coin signal is same regardless of null residue signal. Since it is a motion detector, constant null residue voltage is removed by bandpass filter.

So maybe not worry so much about size of null residue as long as it not too big to go out of diff amp input range.

Is that on the right track or am I smoking something...?

Regards,

-SB

[gwzd]

I have published this work some weeks ago in the internet to avoid people stealing the idea via patent pending. The idea belongs to all mankind.

Just check out in google:
http://www.google.de/search?hl=de&q=...tnG=Suche&meta=

This is the reason for, why I am not giving the source code. I don't want people registering patents for stolen ideas. There are some novel software solutions built-in in the software, which are not published yet. And I am going to keep them totally secret.

Aziz

way to go Aziz! But, does it work? Outside your room, that is.
Regards,
R.A.

[Aziz]

I have published this work some weeks ago in the internet to avoid people stealing the idea via patent pending. The idea belongs to all mankind.

Just check out in google:
http://www.google.de/search?hl=de&q=...tnG=Suche&meta=

This is the reason for, why I am not giving the source code. I don't want people registering patents for stolen ideas. There are some novel software solutions built-in in the software, which are not published yet. And I am going to keep them totally secret.

Aziz

Hi miki73,



I am sorry to say, that this project is not an open-source project. Software development is my daily bread and I can not give any source-codes. I also don't want some people stealing my good ideas and huge effort built-in in the software. But I am giving many hints and results to make this project possible to others.
Just start reading Win32 coding. It is well documented. You only need the wave input/output programming interface on the Win32 SDK.

Aziz

So let's make similar application as open source for the rest of us...?

Perhaps one of these could be used for a start:

Arecord:
ftp://ftp.alsa-project.org/pub/utils/

http://alsa.opensrc.org/index.php/Arecord

Sox:

http://sox.sourceforge.net/

[Aziz]

Hi Simonbaker,

Aziz -- this question about nulling interests me.

We null coil to get rid of TX signal. Let's pretend "perfect" case, perfect null, no TX signal, only signal from coin. Then phase of RX signal is constant no matter how deep coin is -- signal gets weaker, but phase same. Our "lock-in" amplifier will get same phase no matter how deep.

Now let's think of real case. RX signal = residual TX signal + coin signal.

So we are adding Asin(wt + phi1) + Bsin(wt + phi2), in other words, two signals with different amplitudes and phases. Result has a phase between the two I believe, depends on amplitudes.

In this case phase shift depends on how deep coin is and how well we null signal. If not nulled well, our discriminator will kill our weak signals because the residual signal will dominate.

Some people say don't completely null coil, but I don't know, it seems like the closer we come, the better. It seems really important to get good null when using discrimination if we want deep coins. For all-metal mode, not so important since lock-in can detect tiny shift.

Now, maybe we can use amplitude information of RX signal to change our discrimination -- if amplitude very small, then we should allow smaller phase shifts in our discrimination logic. However, amplitude is usually very noisy I think. But are you planning to do something about residual TX signal from imperfect nulling? What is your nulling strategy?

Regards,

-SB

I have not decided yet, where to set the residual voltage and phase of the RX coil. As far as I observed, there are many possibilties with some drawbacks. As a fact of lock-in amps, I need some residual voltage on the RX coil to keep the phase noise low. My residual voltage can be between 2% to 80% of the dynamic range of the ADC input.
I will investigate next week, which option is better.

Regards,
Aziz

[Aziz]

Again the external USB Creative X-Fi Surround 5.1 sound-card:

Excessive tests have shown the fact: The mentioned sound-card is too much noisy to break new detection records. The extracted signal is drifting too much in the low frequency region (0.05 .. 10 Hz) and therefore I have to set the detection threshold much above the standard deviation of the noise to avoid false signalling. My internal HD sound-card is much better. I will buy an another external sound-card next month. But the Creative X-Fi Surround 5.1 can be used with reduced sensitivity. 1 Euro coin can be detected save in the region between 40 .. 46 cm (air).
I will use this sound-card as a backup hardware until I find a better one. I know, there are still excellent sound-cards in the market. But I have not the resource to buy all and test them. So I have to wait a bit. In the mean time, I will develop the search coil and the software with the existing sound-card.

Aziz

[Aziz]

Hi bbsailor,

Aziz,

You have done well.

If you can find a simple way to build a phantom powered preamp stage (gain 10X) in the RX coil circuit (in the housing) you can improve the signal to noise ratio and eliminate the effect of cable capacitance from the RX circuit performance of PI detectors.

I see a new class of PI coils being developed that use an active RX circuit with small surface mounted components in a module built right in the coil housing in a location that is least affected by the TX radiation pattern.

It is nice to see you picking up on the lock-in amplifier posts I have made in the past. Next, look into how to use the boxcar integration instrument functions so that you can simulate the back end processing of a PI circuit.

You may be able to simply build the coil interface to connect the coil TX and RX circuits to a PDA (Personal Digital Assistant) for portable use. You will need an external MOSFET coil driver and first amplifier stage with a gain of a hundred or two with clamping diodes in the RX input. You could even use the PDA screen to observe the RX return signal to help identify the nature of the target.

Keep on pushing the envelope!!!

bbsailor

thank you very much. The success of this project is due to most of yours and others effort of calling my interest to lock-in amplifiers and boxcar integrators. Both techniques can be used well in a VLF metal detector.
But only few people have recognized the power of these techniques. Despite of driving the coil with very low power (milli Watt range), I can achieve impressive results compared to other VLF detectors. I am only scratching just the top of the hill what is possible and achieveable.

I have not tried to make active search coils. Phantom power supply is well used in microphone applications and maybe we can take over some good ideas. But I would spend a power supply line to the coil rather than making a phantom power supply. Active search coils will be tested, when theres is no more SNR to get. At the moment, I have lots of reserve of SNR not tested yet. I know, I can increase the coil driving power to get better sensitivity. Also amplifiying the RX signal to adapt to the dynamic range of the ADC in the sound-card. All active methods will be tested after I have realized a totally passive one: just the search coil, laptop and sound-card.

There is still too much work waiting for me. Particularly the field testing and not covered problems inherent to ground effects and hot rocks.

On the PI, the boxcar averager could be used rather than the lock-in amplifier. I will focus on PI technology next year whether the boxcar averager could be implemented easily. I am just collecting some experience on lock-in amps and boxcar averagers.

Regards,
Aziz