Metal DiscriminatorGirish O

Dept. of Electrical EngineeringIIT Bombay

Mumbai, Indiaogirish@ee.iitb.ac.in

Pranav AgarwalDept. of Electrical Engineering

IIT BombayMumbai, India

pranavg@ee.iitb.ac.in

Abstract—Metal discriminators are electronic devices whichcan detect the presence of metal and discriminate the typeof metal. The metal detectors are present in airports, officebuildings, schools, government agencies, prisons, etc helps toensure safety while consumer-oriented metal detectors providesopportunity to discover hidden treasures, none of them can dis-criminate the type of metals. Here we are focusing on consumer-oriented metal detectors which can detect and discriminate thetypes of metals. We use the fact that the inductive property of ametal is proportional to its conductivity to discriminate the typeof metals.

Index Terms—Metal detector, Metal discriminator, D-shapedcoils, Eight shaped coils, High Q active band pass filter, Magni-tude independent phase detection.

I. INTRODUCTION

Metal Detectors generally use one of the three technologiesmentioned below.• Beat Frequency Oscillator (BFO)• Pulse Induction (PI)• Very Low Frequency (VLF)

The most basic way to detect metal uses a technology calledbeat-frequency oscillator (BFO). [1] In a BFO system, thereare two coils of wire. One large coil is in the search head, anda smaller coil is located inside the control box. Each coil isconnected to an oscillator that generates thousands of pulsesof current per second. The frequency of these pulses is slightlyoffset between the two coils. As the pulses travel througheach coil, the coil generates radio waves. A tiny receiverwithin the control box picks up the radio waves and creates anaudible series of tones (beats) based on the difference betweenthe frequencies. If the coil in the search head passes over ametal object, the magnetic field caused by the current flowingthrough the coil creates a magnetic field around the object.The object’s magnetic field interferes with the frequency ofthe radio waves generated by the search-head coil. As thefrequency deviates from the frequency of the coil in the controlbox, the audible beats change in duration and tone.

The simplicity of BFO based systems allows them to bemanufactured and sold for a very low cost, but these detectorsdo not provide the level of control and accuracy provided byVLF or PI systems.

[1] PI uses a single coil as transmitter and receiver. Thistechnology sends powerful, short bursts(pulses) of currentthrough a coil, each pulse generating brief magnetic field.When the pulse ends, the magnetic field reverses its polarity

and collapses very suddenly, thus resulting in a sharp electricalspike lasting for few micro seconds. If the metal detector isover a metal object, the pulse creates an opposite magneticfield in object, which causes the magnetic field due to aparticular pulse to decay slowly, so by measuring the widthwe can determine the presence of metal. These metal detectorsare not good at discrimination since reflected pulse length ofvarious metals are not easily separated. They are useful inareas that have highly conductive material in the soil or generalenvironment such as salt water exploration.Here we are using VLF technology to develop our metal de-tector. Here there are two coils in which one acts as transmitterand other as receiver. The transmitter and receiver are arrangedin such a way that net EMF (Electromotive Force) generatedin the receiver due to transmitter is zero in normal condition. Ifthe coil passes over a metal object, the magnetic field inducesan eddy current in the metal object, which generates its ownfield inducing voltage in receiver coil, from this voltage wecan detect the presence of metal. Due to the inductive natureof the metal there is a phase difference between the inducededdy current and the transmitter signal which remains constantfor particular metal. Since inductive nature of the metal isproportional to its conductivity, thus the phase differenceobserved too is proportional to its conductivity. Conductivityfor a metal depends on its composition and remains constant atparticular environmental conditions. Thus, the phase differencecan be used to discriminate between metals. [4]

II. DESIGN IMPLEMENTATION

Here we are transmitting 940 Hz sine wave which is gen-erated using an oscillator circuit and then amplified by poweramplifier. The received signal is amplified using an instru-mentational amplifier followed by a Band Pass Filter (BPF) toremove out of band noise signal. The phase difference betweenthis filtered signal and the transmitted signal is estimated usinglock-in phase detector. The traditional lock-in amplifier wasmodified to make its output independent of magnitude of inputsignal. Along with the new modification in lockin amplifierwe are checking the phase at particular signal level, thisreduces error in phase measurement due to magnitude. Theseinformation is finally sent to the microcontroller through ADCwhich has an hard coded table for phase difference shown byvarious commonly available metals.

Rx Coil IN Amp Active BPF

Full Wave

Rectifier

Lock-in AmpAC coupled

ComparatorLPF Level Shifter

MCP3008

ADC

89C5131A

Microcontroller

Display

LPF Comparator

LED

Tx CoilPower

Amplifier

Wien Bridge

Oscillator

Reference

Clock

Generator

Fig. 1. Block Diagram for Metal Discriminator

Thus the metal discriminated is display directly on LCD.Figure 1 shows the overall block diagram of our the MetalDiscriminator.

A. Oscillator

−

+22pF

1.303kΩ 1.303kΩ22pF

Vout

10kΩ

+5V

10kΩ 10kΩ 10kΩ

−5V

Fig. 2. Wien Bridge Oscillator

The oscillator is Wein Bridge oscillator(Fig. 2) with ampli-tude stabilization. [12] By connecting two diodes as per thediagram, it ensures that until the diode is forward biased (thatis |vout| < 0.7) the gain of the circuit is large, which helpsin oscillation to build up. When either of the diode becomeforward biased the gain goes to 3 which is the minimumrequired to sustain oscillation.

B. Power Amplifier

We cannot connect an inductor directly to the output of os-cillator, since impedance of the oscillator will change causingoscillations to cease.

−

+

10kΩ

10kΩ

+12V

−12V

Vout

+12V

−12V

Vin

Fig. 3. Power Amplifier

Also we cannot connect the output of the oscillator throughbuffer because peak current flowing through the coil is closeto 1 A which will damage the opamp. If we want to bufferthe output of oscillator then we have to use power opampwhich is very expensive and it is not available in lab. So,we used the Fig. 3 circuit which can be implemented usingcommercial opamp. The beauty of the circuit lies in the twoMOSFETs which provides necessary current needed by thecoil by keeping the current flowing through the opamp tosafety limits. Here two diodes are connected to introducea small change in switching point to ensure that both theMOSFETs won’t switch on at the same time.

C. Coils

Designing the coils was one of the biggest challenge, coil isthe most critical part in the design. Deciding the dimensionsof the coil like shape, number of turns and also number ofcoils required for effective operation are critical. After somesearching [2] we found that D-shaped coils are best suited

for our application since D-shaped coils penetrate more thanthe circular coil of same dimension. Since we are using VLFmetal detector we have to use at least two coils, one actingas transmitter and other receiver, three coils configuration isalso possible where there is two transmitter and one receivercoil. Since we are using D-shaped coils it is very difficult tocalculate the field at a particular point mathematically, so fortesting purpose we made 2 pairs of coil of diameter 9 inch (70turns) and 5.5 inch(140 turns) to get an idea of the range andmagnitude of received signal. The reason for using coils having2 diameter is detection range increase with diameter and thesize of the smallest object that can be detected degrades withdiameter.This is due to the fact that smaller coils have moreconcentrated fields, which can detect smaller objects. [3]

Another way to wind the coil is in form of Eight as shownin Fig 4. Here the two halves of eight shaped coil acts astwo individual transmitter coil which carries equal current inopposite direction. So, the major advantage we observed byusing this structure is easy cancellation of the transmitter fieldin receiver coil. But we cannot use this coil for discriminationbecause the phase shift observed is 180 out of phase whenthe metal is brought near to each half of eight.

Winding surface for both the shape i.e. D and 8 were madeby stacking card-board pieces cut in shape Fig. 4.

Next bigger task is choosing the frequency of operation.If frequency of operation is smaller then penetration depthwill be larger but the sensitivity is small. So if we want todetect smaller objects then we have to increase the frequencyof operation but it will degrades the depth of penetration. Moreover we cannot increase the frequency to MHz range since weare using breadboard to test our circuit, which adds parasiticsat these frequency range. So we chosen the frequency ofoperation around 1K(940 Hz), to eliminate the above shortcomings.

Next step is to determine the inductance of coil. For thatwe connected a capacitance of 1µF in series and variedthe oscillation frequency until voltage across the inductor ismaximum. That frequency is the resonance frequency andfrom that inductance of the coil can be calculated. This methodof determining inductance is best when compared to othermethods, since we have to model the coil as a inductor withseries resistance.

Now we need to create resonance at these frequencies. Inthe transmitter side we connected a capacitor in series withcoil to cause the resonance at operating frequency. The reasonfor connecting the capacitor in series is, at series resonanceimpedance offered by the circuit is lowest (here DC resistanceof coil) so, the current is maximum. Since the field producedby the transmitter coil is directly proportional to current, atresonance transmitter field will be maximum.

Fig. 4. Coils

Similarly a capacitor is connected in parallel to the coil atthe receiver. These resonator circuits acts a BPF which worksas EMI rejection circuits and also eliminates noise outside thebandwidth of BPF. We also twisted the coil ends(upto the inputof instrumentation amplifier ) so that it forms a twisted paircable, which also reduces EMI.

D. Instrumentation Amplifier

[7] The receiver coil is again AC coupled to the input ofinstrumentation amplifier to eliminate the low frequency noise.The idea behind using instrumentation amplifier to eliminatenoise in the received signal by amplifying the differentialvoltage between the terminals of coil. Since noise can beconsidered as common mode signal, in-amp will improve theSNR of signal. The gain of the in-amp is set to around 50,if the gain of the in-amp is larger it is difficult to cancel theeffect of transmitter field on receiver coil.

E. Band Pass Filter

The output of the instrumentation amplifier is fed to activebandpass filter (Fig. 5)to further filter the noise outside signalband. The centre frequency of the BPF is chosen to be 940Hz with a Q of 9 and gain of 2. [11] This is the last stagefiltering before feeding the signal to lock-in amplifier.

−

+

0.1µF7.6kΩVin

30.47kΩ

0.1µF

Vout

100Ω

+5V

−5V

Fig. 5. Band Pass Filter

F. Lock-In Amplifier

−

+

1kΩ

1kΩ

1kΩ

1kΩ

φ

φ

φ

φ

Vps

Vpsd

Fig. 6. Lock-in Amplifier

1) Switching Circuit: The output of the BPF is fed to theswitching circuit which is an important part of the lock-inamplifier.The switches are implemented using CD4066 whichis an analog switch.

−

+φ

φ

10kΩ

+5V

Vin

+12V

−12V

Vref = −5V

Fig. 7. Reference Clock Generator

2) Reference Clock Generator: [8] The reference clock re-quired for the switching circuit is generated by Fig.7 circuit. Itconsists of a comparator followed by inverter to get actual andcomplemented reference output. The input of the comparatoris the reference sine wave from oscillator.

3) Subractor: The switches are connected to the subtractorcircuit as per the fig. 6. The operation of the circuit can beexplained in two phases.

Phase1:

−

+

1kΩ

1kΩ

1kΩ

1kΩ

Vps

Vpsd

Fig. 8. Equivalent Circuit Phase1

In Fig. 8 phase the subractor acts as noninverting amplifier.The gain of the circuit is

Vout = Vin

[R

R+R

] [1 +

R

R

]=

(Vin2

)× 2

VoutVin

= 1

Phase2:

−

+

1kΩ

1kΩ

1kΩ

1kΩ

Vps

Vpsd

Fig. 9. Equivalent Circuit Phase2

In Fig.9 phase the circuit acts as inverting amplifier whosegain is -1. So basically this circuit performs multiplication ofreference signal with the received waveform.

−

+1µF

Vin

1kΩ

Vout

5kΩ

+5V

−5V

+5V

−5V

Fig. 10. Comparator

4) Comparator: [9] The output of subtractor is fed to theAC coupled comparator to make the output of phase detectorindependent of received signal amplitude. The reference inputof the comparator is adjusted to some milli volts such that itcompensates the offset error in opamp. So depending on thephase difference between input and received signal the dutycycle of the comparator output changes. Here the comparatorsare implemented using LM318N (Fig. 10) which is a very highspeed op-amp.

G. Lowpass Filter

−

+

15kΩ

15kΩ

Vin

Vout

1µF

Fig. 11. Low Pass Filter

Since the output of comparator is periodic waveform, wecan represent it as sum of sinusoids known as Fourier series.The first term represents the DC value and others will besinusoids at frequency f1, 3f1,etc where f1 is the frequencyof square waveform. So to extract the DC value we can usea lowpass with a reasonable low cutoff frequency. Here wechosen the cutoff frequency to be 10Hz. If the received signalis represented as Acos(ωt+φ), where φ is the phase differencebetween transmitted and received waveform, then the outputof lowpass filter is

V0 =1

2π

2π∫0

Acos(ωt+ φ)× ψ(ωt)d(ωt)

where

ψ(t) =

1, if 0 < ωt < π

−1, if π < ωt < 2π

=1

2π

π∫0

Acos(ωt+ φ)d(ωt) +

2π∫π

(−cos(ωt+ φ))d(ωt)

= −2A

πSin(φ)

So, if we make the output of LPF independent of the amplitudeof received signal we can discriminate the metal. Here we usedop-amp OP27G for implementing the Lowpass filter (Fig.11)to minimize the offset error to 10µ V.

H. Control Signal Generator

−+

10kΩFrom BPF

10kΩ

+5V

−5V−+

10kΩ

+5V

−5V

20kΩ

20kΩ

−+ 1µF

5kΩ

+5V

−5V

+5V

−5V

100ΩTo µC

Fig. 12. Control Signal Generator Circuit

To remove dependence on magnitude of received signal forphase difference estimation, in addition to the AC coupledcomparator after lockin amplifier, we are checking the phasedifference only when the magnitude of input signal reachessignificant value i.e. when the output of the BPF is 1 Vpp. Toimplement this we precisely full rectify the signal and afterlow pass filtering we compare it with a pre-defined voltagesuch that, we get high output when there input from BPF is< 1Vpp and low when it is > 1Vpp which is further fed tomicrocontroller as control signal.

I. ADC and Micro-controller

[5] The final part of the design is displaying the result. Itis done by interfacing ADC [6] to micro-controller 89C5131A

using SPI protocol. The results are displayed on 16× 2 LCDscreen attached to 89C5131A board.

III. RESULTS

We grounded one end of the instrumentational amplifierand connected other end to AWG channel to get a smallmagnitude sine wave such that the voltage at the input ofLock-in phase detector is 1Vpp. Now we changed the phaseof the input signal from −45 to 70 and the reading of LCDwas plotted against phase Fig. 13. Outside this range we can’tdetermine the phase because of the AC coupled comparatorwhich gives the maximum and minimum duty cycle waveformin this interval rather than −90 to 90. Above process wasrepeated thrice. Initially the power supply was kept ON for1 hr and then all the readings were take for increasing anddecreasing input phase, then power supply was kept ON for 2more hrs and then finally power supply was OFF for 2 hr. Fig.13 showes that there is no hystersis. Fig. 14 is plot of Driftbetween the second and first and thrid and first obervation. Weseen that there is less than 70 mV of drift. Next we plotted thepercentage error in our LCD reading w.r.t digiral multimeterreading. We observed maximum error of 6% (Fig. 15).

Now after the above characterization of the circuit weconnected D-shaped coil and cancelled the field at receiver byadjusting the position of the receiver coil w.r.t to the transmittercoil. Then various metal was brought on the overlapping area,which gave different phase for different metals. The observedphase and along with its conductivity [10] is listed in table I.

Fig. 13. Calibration Curve

Fig. 14. Offset Drift vs Phase

Fig. 15. Percentage Error vs Phase

TABLE IPHASE DIFFERENCE AND CONDUCTIVITY VS METAL

Metal LCD (inmV)

AbsolutePhaseDifference (inDeg)

Conductivity(in MegaS/m)

Copper 870 66 58Aluminium 1163 52 37Bronze 2520 26 7Carbon Steel 1640 24 6Stainless Steel 1771 18 1IronChromiumAlloy

1900 8 0.74

Fig. 16. Absolute Phase Difference and Conductivity vs Metal type

Fig. 17. Absolute Phase Difference vs Conductivity

IV. DISCUSSIONS

From Fig.16 and Fig.17 we verified our basic assumptionthat phase difference is proportional to the conductivity of themetal. Thus we calibrate our device to discriminate betweenmetal if we have the conductivity data which is easy avail-able and are accurate. Once the relation between phase andconductivity is established we can estimate the conductivityirrespective of the shape and size of the material. It alsoprovide a contact-less way of measuring conductivity. Herewe are try to implement every module using analog circuitsas far as possible due to which precision is less when com-pared to digital techniques. Moreover range of detection anddiscrimination can be increased by replacing analog processing

techniques with DSP techniques. If we implement the filtersin digital, then range of detection can be increased sinceeffective noise cancellation algorithms are available in DSP.Also we can make the PSD output independent of receivedsignal amplitude by implementing one more PSD circuit withreference signal are orthogonal to the transmitted waveformand dividing the two outputs using a micro-controller.

V. FUTURE WORK

Above idea for discrimination based on phase can be furtherbe applied to determine the exact composition of any materialand even to estimate its conductivity.

ACKNOWLEDGMENT

We would like to acknowledge Prof. Siddharth Tallur, VarunWarrier and Maheshwar G. Mangat for there constant supportand ideas to successfully implement our design.

REFERENCES

[1] https://electronics.howstuffworks.com/gadgets/other-gadgets/metal-detector5.htm.

[2] http://www.canadiantreasureseekers.com/index.php?l=product detailp=917.[3] http://www.garrett.com/hobbysite/hbby searchcoil tech sheet en.aspx.[4] https://www.youtube.com/watch?v=fnwgf5RrhTg.[5] http://pdf.datasheetcatalog.com/datasheets2/37/374479 1.pdf[6] http://ww1.microchip.com/downloads/en/DeviceDoc/21295C.pdf.[7] http://www.ti.com/lit/ds/symlink/ina118.pdf[8] http://www.ti.com/lit/ds/symlink/lm111-n.pdf[9] http://www.ti.com/lit/ds/symlink/lm318.pdf

[10] http://www.tibtech.com/conductivity.php[11] http://www.electronicshub.org/active-band-pass-filter/[12] http://www.electronics.dit.ie/staff/ypanarin/Lecture

Fig. 18. COMPLETE CIRCUIT DIAGRAM