Metal detection and discrimination

• Input amplifier
• Synchronous detector and DC amplifier
• AC channel . Motion filter
• Target detection
• Metal Discrimination
• earth balance

Unfortunately, after looking through the mountain of information, we can draw a sad conclusion – the topic has almost exhausted itself. Recently, there are almost no new patents in this area, and manufacturing companies produce clones of well-known devices. The peak of activity falls on the 80s of the last century. Another surge of interest in metal detectors, but less so, is associated with the transfer of old circuits to microprocessors.

All this concerns the search for new principles of operation and new circuitry for devices. This has little to do with practice, since it is at the present time that we are witnessing the peak of interest in metal detectors from amateur search engines. This is understandable – no one wants to work, they fool around in the lottery, but here you can dig right out of the ground for bread and butter 🙂

Most of the schemes can be found on the Internet by visiting sites and forums of “treasure hunters”. The most interesting is hidden in the patent database – for example, http://patft.uspto.gov/ . If you know technical English, then go ahead, there are a lot of interesting things. Or take on the repair of metal detectors – and you will learn a lot of new things, and you will get hold of the schemes 🙂

Input amplifier  

Let’s immediately discard circuits in “loose” – this is how devices assembled on individual transistors are called in the jargon of circuit engineers – when integrated circuits can be used. The requirements for metal detectors are not so high, so that it is impossible to pick up a cheap and reliable operational amplifier (op-amp).

As a rule, the input amplifier (aka receiver) is assembled on one or two low-noise op-amps. The switching circuit depends on the design of the sensor, its frequency and supply voltage.

The figure below shows a typical input amplifier circuit for a sensor with a resonant receiver circuit.

The circuit is a first-order band-pass amplifier (BPF1), whose maximum gain must match the operating frequency of the sensor. But not necessarily, since often phase shifts are regulated by capacitances to balance the LED.

With slight variations, such a scheme migrates from one device to another. Only the values ​​​​of resistors and capacitors change, which determine the gain ( K u003d R 2: R 1) and the frequency response in the low ( C 1) and high ( C 2) frequencies. The R3C3 chain additionally attenuates the gain in the low and infra-low frequencies and eliminates the constant component of the signal. Very often there is no such filter at all.

Often there is a two-stage amplifier. A typical circuit of such an amplifier is shown in the figure.

A typical frequency response (AFC) of a receiver looks something like the one shown in the figure below.

At the output of the receiver, we have a signal, which is described by the following expression:

Here ULrc is the signal on the receiver coil.

d is the phase shift, depending on the design of the sensor.

r – phase shift of receiver circuit elements.

As a rule, when designing a metal detector, a single-stage amplifier circuit is taken as a basis … to be honest, I wanted to say – one to one is copied 🙂 . In principle, there is nothing wrong with this if the amplitude-frequency characteristic of the amplifier corresponds to the parameters of the sensor.

But, as always, there are nuances and pitfalls …

Please note that the receiver coil is turned on so that the input current of the op-amp flows through it. As you know, when searching, the sensor constantly hits rocks, branches and other objects. There is a mechanical effect on the system of sensor coils, which leads to modulation of the fundamental frequency of the sensor by a “shock” signal of low or even infra-low frequency. Or such a signal penetrates through the input circuits of the amplifier. And if the frequency response of the receiver has insufficient attenuation in the low-frequency region (units and tens of Hertz), then the metal detector will respond to strikes with a “squeak”, like metal. Such an effect is possible even when the sensor mounted on the rod is wiggled.

The same effect occurs with an incorrectly designed Faraday screen. If the screen has a low resistance, then it begins to absorb too much transmitter energy due to the occurrence of Foucault currents. And when hit, the receiver modulates. Only a screen change will help here.

There are many ways to deal with the phenomenon of low-frequency modulation. For example, completely fill the sensor coils with epoxy resin to give it sufficient rigidity. But such a solution will significantly weight the sensor. As a rule, this is exactly what amateur designers do.

It is easier to further attenuate the gain of the receiver at low frequencies, which will give the same effect. To do this, you must correctly select the values ​​​​of the elements, be sure to install the R3C3 chain and, possibly, use an input isolation capacitor. In addition, the penetration of interference can be significantly reduced by using a more complex (differential) circuit of the synchronous detector. Here is a diagram of such a receiver.

Here, the value of the capacitor C3 is reduced, an isolation capacitor Cs and a resistor Rp are introduced, which ensures the operation of the op-amp with direct current. You can also shunt the circuit of the receiving coil of the sensor with a resistor Rs.

Below is the frequency response of the receiver, which illustrates what has been said.

The yellow diagram corresponds to the case when the R3C3 chain is not installed. Blue – too high C3 and red – at optimal C3. The green diagram corresponds to the case when the coupling capacitor Cs is used. As you can see, at the main operating frequency, the gain of the receiver does not change.

The figure shows the signal at the output of a synchronous detector when the sensor hits a stone.

The blue diagram illustrates the impact signal penetration into the metal detector amplification path for a non-optimized receiver circuit. The red diagram corresponds to the response to the same shock for the upgraded circuit. And although the signal from the impact is small in amplitude (as a rule, a few millivolts), it is enough for the identification circuit to work and an audible signal to sound. This is a real signal, I observed it on the screen of the communication terminal with a metal detector (computer oscilloscope). The signal had to be amplified 20 times.

It is believed that to eliminate this unpleasant effect, it is sufficient to use a more complex synchronous detector assembled according to a differential circuit. But practice has shown that it does not completely suppress the spurious signal. This requires the use of parts with selected parameters.

As you can see, without much modification to the receiver, you can avoid the messy and difficult operation of pouring epoxy into the sensor.

conclusions

If everything is in order with the sensor, then you should use the receiver circuit shown in the first figure. It is quite suitable for the microprocessor version of the metal detector.

Elements Rs-Cs-Rp should be used only in case of increased response of the sensor to shocks.

Alternating current (AC) preamplification is used in almost 100% of metal detectors. I know of only one device in which there is no receiver, and the signal is amplified only by direct current.

You have seen circuits that use old ICs. No doubt they are still in use. But only when the MD is powered by a voltage of 8 V and above. However, modern devices are powered by 5V, sometimes 3.3V. For these circuits, there are cheap and good ICs, such as TS971, TS 972, TS922, LMV721 and others. Their price is about 4-6 UAH, but not 50 UAH, like Analog Device , for example.

Synchronous detector and DC

The second, almost obligatory for use :-), a metal detector assembly. This is not surprising – it is the synchronous detector that allows you to highlight the useful signal against the background of very strong interference. This is such a typical node that I have come across the only solution where the synchronous detector circuit is not used in its usual form. It is widely known in a narrow circle of amateurs 🙂 Robert Hoolko’s microprocessor-based metal detector project. However, now many metal detectors use Digital Signal Processing (DSP) – there is only a receiver, all other nodes are implemented in software.

Most circuits use the simplest synchronous detector. As far as I know, it has only one significant drawback – it passes the constant component of the signal to the output, due to the non-symmetry of its circuit. However, this scheme is used both in simple and in the most complex MDs. I also use it and did not notice any problems associated with the penetration of the DC component into the DC channel .

The synchronous detector is a key (element U1 D ) that closes synchronously with the detected signal. Hence its name. The key is controlled by the drS signal. If we close the switch at exactly known time periods – for example, when the sinusoidal signal Urc from the receiver output is positive – and apply it to the R4C4 integrating circuit to smooth out the rectified voltage ripples, we will get a synchronous half-wave rectifier.

The rectified signal is fed to the amplifier, which produces additional smoothing of ripples and is a conventional first-order low-pass filter (LPF). It would be more correct to use a filter of a higher order, but everything works that way, so no one bothers with complex J circuits .

The same scheme is used for the second channel of the metal detector. If the control signals are shifted by 90 ° , then at the outputs of the synchronous detectors we will receive separated signals for the active and reactive components. In the practice of metal detection, the indices X and Y are added to such signals, respectively. Control signals can be called, for example, drX and drY. The output signals are sdX and sdY .

The quality factor and bandwidth of a synchronous detector vary over a very wide range, it is enough to change the parameters of the integrating circuit R 4 C 4. Using such a simple device, you can get a quality factor of several thousand and a bandwidth of fractions of a hertz. But let’s go down from the sky-high theoretical heights and talk about the sore.

And again, as in the case of the receiver, you can’t just solder the parts just like that, simply by “ripping” the circuit of some “Fisher”. It would seem that everything is simple – he waved the sensor, the device squeaked, – and dig out the pot with royal gold coins! Well, it doesn’t work that easy…

First, let’s understand why you need to know about the bandwidth of a synchronous detector at all. As we know, the separation of flies from cutlets :-), or rather, the interfering signal of the earth from the useful signal of the object, occurs according to the frequency principle. From practice, we know that the signal from different bumps and pits lies in the frequency range from direct current to 2÷3 Hz. The frequency range of the signal from the target depends on several factors (for example, the size of the sensor and the speed of the stroke) and lies in the higher frequency region, from about 4 to 16 Hz. Only this allows you to find coins in the ground without much difficulty.

Here is the parameter for calculating the synchronous detector integrator. Let’s plug this value into the bandwidth formula and see what happens.

Here is the formula for calculating bandwidth:

Let’s transform this formula…… replace R=R4, C=C4 and see what happens.

We take R4 u003d 51K or 51 10 ³ Ohm and substitute this value in the formula

Or 260nf. Let’s take C4=220nf.

In practice, this means that if you solder R4=100K and C4=470 n , then in order not to lose sensitivity, you need to reduce the sweep speed of the sensor four times. In addition, the signal will acquire an unpleasant surge of negative (in relation to the useful signal) polarity. This surge can cause a secondary operation of the MD circuit.

But smaller denominations (four times) cannot be used. Then the sensitivity of the MD will be in the zone of interference from the impacts of the sensor. It’s not all that simple…

Secondly, you need to make sure that the amplifier circuit of the synchronous detector is correctly calculated. For normal operation, the cutoff band of the low-pass filter must lie above the highest frequency signal from the target. Since our highest frequency is 12-16Hz, the low-pass filter should also be calculated for this frequency. To calculate filters, there are a great many paid and free, and recently even interactive programs. The diagram in green shows the correct frequency response for the selected component ratings.

If you want to increase the gain, then you cannot do this by increasing the value of R6. Otherwise, the cutoff frequency of the filter will decrease. Such a case is shown in the figure above, the diagram is in red. As you can see, the filter cutoff frequency has decreased from 16Hz to 4Hz. In this case, you must either proportionally reduce the value of C5, or reduce R5 without touching R 6 C 5.

Pay attention to the ratio of the capacitances of the input C3 and integrating C4 capacitors. As a rule, C3 has a capacitance an order of magnitude smaller than C4. This provides additional suppression at infra-low frequencies and a reduction in rectified voltage ripple.

Recently, the scheme of a full-wave synchronous detector has gained popularity.

An additional inverter on the op-amp U3 and SPDT switch U1 provide full-wave rectification of the input signal. Resistor accuracy requirements are reduced – in fact, only R5 and R6 need to be 1% accurate to ensure a gain of exactly -1.

Such a synchronous detector scheme is most often used in modern metal detectors based on microprocessors. Moreover, the R3C3 high-pass filter, as a rule, is absent – since such an LED suppresses the constant component of the signal quite well.

conclusions

It is advisable to use the diagram shown in the last figure. Although I personally use a simple SD scheme …

You can learn more about the principles of operation of a synchronous detector by reading G. Petin’s article “Key Synchronous Detector” in the third issue of the magazine “Schemotekhnika” for 2003.

AC channel . Motion filter

It would seem that the amplified signal from the synchronous detector is enough to apply to the comparator and, if the voltage exceeds a certain threshold, the metal is found. It is so, but it is difficult to call such a search comfortable. A slight change in the degree of mineralization of the soil or hole will cause the metal detector to give an erroneous signal. Therefore, the signal directly from synchronous detectors is used only in specific search modes – for example, pinpoint (precise determination of the location of an object).

You can, of course, introduce feedback and monitor the level of soil mineralization, but this feedback will turn the amplifier into a high-frequency filter (HPF, HPF). I myself once made a similar MD. He had normal sensitivity, and in the air and in the soil it was the same. The soil signal was passed through the LPF and its signal was used to compensate for the SD. The result was a high-pass filter with a settling time of several seconds.

And it is also problematic to obtain an acceptable gain using cascades with galvanic coupling – the DC bias of each cascade will interfere, no matter what it is caused by, the temperature instability of the op-amp or a change in soil mineralization. Here, constant balancing of the cascades for direct current (tracking) is also required.

In general, whatever one may say, it’s easier to immediately apply the HPF to amplify the useful signal from the target. In addition, as already explained earlier, the separation of the soil signal interfering with the search from the target signal is carried out according to the frequency characteristic.

So, we came to the conclusion that the main amplification of the target signal and filtering of interfering signals in modern metal detectors is carried out by high-frequency filters. To isolate a useful signal – recall, its frequency range is located approximately from 3 to 16 Hz – you need to apply a band-pass filter (PF, BPF ) to this frequency. And to remove high-frequency interference, it’s not bad to add a low-pass filter with a cutoff frequency of at least 16 Hz. Collectively, this is called a “motion filter”.

In general, a set of different filters can be seen in the vast majority of metal detectors. And the wider this set, the higher the class of the metal detector. Moreover, it does not play a special role, these filters are assembled on discrete elements or implemented in software. They must be mandatory. In digital MD there are the same filters, only they are “not visible”, they are implemented in software.

The filter parameters depend on the geometric dimensions (diameter) of the sensor and the speed of its movement during the search. Of course, if you change the swing speed, then these values ​​u200bu200bwill have to be recalculated.

So, let’s consider practical schemes of motion filters. A diagram of the simplest filter used in Whte’s Classic series metal detectors is shown in the figure below.

It is a bandpass filter with a maximum gain at 8Hz. Its frequency response is shown in the figure below.

One of the disadvantages of such filters is a very small dynamic range. With small input signals, on the order of a few millivolts, everything is fine. But as soon as the input signal increases slightly, the signal is clipped. For an entry-level metal detector that does not have a VDI indication, this would not be scary – you still only need to “squeak” at the metal. But with large signals, the filter settling time also increases. The figure shows such a situation.

In the figure, the green color diagram corresponds to a normal amplitude input signal with a duration of about 200 ms. After 300÷330ms the filter is ready for operation again. But if the signal is large, then the signal is limited (steps in the red diagram) and the filter does not restore the normal mode for a long time, i.e. enters saturation mode. In the diagram above, the settling time is 600ms.

Why is it bad? Let’s count a little. Assume that the swing speed is 1m per second. This means that in 330ms the sensor will describe an arc 33cm long. At this time, the metal detector identified the object and gave a signal – this is 200ms and another 130ms, or 13cm, it “calmed down”. This is a dead zone, and if there was another item under the coil at that time, it will be skipped.

For a large signal, the situation is even worse. The duration of the pulse from the object remained the same, but the settling time increased to 400ms. In total, we get that the dead zone has increased to 40 cm, that is, it is twice as much as the useful one. Imagine how many small artifacts will be missed.

The situation can be somewhat improved by artificially limiting the gain for large signals. For example, by introducing diodes D1 and D 2 into the feedback circuit of the amplifier (see Fig. 3.11). Such a circuit for limiting the output voltage has several varieties. For example, the circuit in the figure above only works for negative signals. If the signal is positive, then the polarity of the diodes must be reversed. For bipolar signals, the counter-connection of diodes is used. Their number depends on the level of the input signal. Recall that the polarity of the signal is evaluated relative to the virtual ground of the op-amp.

For more “advanced” metal detectors, a different solution is used. The filter is made multi-link, consisting of several stages with a small gain. For the first time such a filter was used by White’s in metal detectors of the 3900 and 6000 series. This solution turned out to be so simple and successful that it is still used in one form or another, even in microprocessor devices. Here is the filter.

As you can see, it is much more complicated than the previous one. But it also has many more advantages. Due to the use of three stages, the filter gain was increased several times (more precisely, by several orders of magnitude). This, in turn, made it possible to shift the filter response to a higher frequency region – to reduce the “dead zone” to a few centimeters. In addition, this filter has dozens of times better suppression of the parasitic influence of the soil. This is perhaps the biggest advantage. Let’s take a closer look at the circuitry of this filter.

The first stage is a second-order HPF with +1 gain. This was done intentionally in order to immediately and permanently get rid of the soil signal. Recall that it can exceed the amplitude of the signal from the target by tens of times. The main amplification of the signal is carried out by the second stage, which is a band-pass filter with a center frequency of about 16 Hz. The third stage is a low pass filter. The C6R7 circuit eliminates the DC offset of the op-amp and is a first-order high-pass filter with a cutoff frequency of a fraction of a hertz. Without this chain, the metal detector would be inoperable, since the sensitivity of the metal comparator (the next stage) is a few millivolts, and the offset can be an order of magnitude larger.

Here is the resulting frequency response of the filter.

And here is the filter response to normal and large signals.

To analyze the filter response and calculate the settling time, a bell-shaped signal of different amplitude and duration of 250 ms was applied to its input. As you can see, the filter settling time does not exceed 150ms in the worst case. This means that the loss from the dead zone is no more than 15cm, with a useful signal of 250ms (25cm). Very good.

In the figure, the green diagram corresponds to the input signal of normal amplitude. The red diagram corresponds to a strong signal, the amplitude of which is 100 times greater than normal.

From all of the above, we can draw a conclusion that is not too related to the circuitry of the considered cascades – it is desirable that the target signal from the sensor be as short as possible, at the same swing speed and the same geometric dimensions. Together with the use of a filter with a short settling time, this will allow distinguishing closely spaced objects, which is quite common in search practice.

The sensor signal can be “shortened” by various methods. For example, for the “Ring” sensor, this is achieved by introducing several absorbers based on short-circuited turns. They are located so as to give the transmitter field a wedge shape. This method is described in more detail in US Patent 4,255,711 “Coil arrangement for search head of a metal detector”. At the same time, these absorbers reduce false signals at the edges of the sensor.

Target detection

Typically, the presence of metal is determined using a simple comparator. If the signal is greater than the threshold level (it is usually adjustable), then the metal is found. Seems simple. But no, and there are pitfalls. It’s not even that the signal from the output of the filters is “dirty” and bears little resemblance to the figure below. You also need to know what and how to analyze.

Consider a typical motion filter output signal. This is the shape of the signal for the White’s 6000 metal detector. Of course, the signal amplitudes are approximate.

The first, negative burst (green diagram) is hardly suitable for analysis – its amplitude is several times less than the second and third bursts. It is easily masked by interference and signal from the ground. The third one is also not a gift – it is very far behind the input signal (red diagram). It remains the second, positive. But then it is necessary either to reduce the gain for negative signals, or not to analyze them at all. In simple filters, a diode is used in the feedback of the amplifier for this – as in the MD Classic amplifier circuit . In microprocessors, positive pulses are simply analyzed.

For some MDs, the “signal delay” parameter is specified. This is what I think is the Z value in milliseconds – like in the picture above. Those. delay of the second filter pulse relative to the original signal from the synchronous detector. Typically, the delay value is between 50 and 150 ms.

By the way, in decent MDs, the “clean” signal of the acY channel is rarely used to detect a target (the acX channel is even more so). This is a synthetic signal, it is usually formed by a special channel G , in which the dcX and dcY signals are added with different weight coefficients. This allows you to get rid of the signal of the soil. This signal is then fed to filters similar in circuitry to the X or Y channel filters .

In fact, there are different methods for detecting a target, except for a purely threshold one. I also came across a method of analysis by signal increment. Indeed, if for a fixed period of time to perform the simplest operation

dG = acGnew –  acGold ,

it is easy to see that dcG will be significant for metals and small for soil. In addition, it makes sense to analyze only positive values ​​of dG . Now it remains only to set such a comparator threshold in order to exclude false positives. Typically, the time interval for analyzing the signal G is from 1 to 8 ms.

This method of target detection is used in my MD (along with the usual threshold). It makes sense to use it primarily for dynamic search, since it is nothing more than the simplest first-order high-pass filter.

Metal Discrimination  

There are different methods for discriminating metals. Usually this is the suppression of the sound signal if the target hits a certain plug, set by the resistor knob or programmatically.

In microprocessor MDs, everything is simple – in the discriminator menu, the VDI values ​​u200bu200bof “from” and “to” are set. If the target’s VDI hits this plug, the speaker’s signal is blocked (in this case, I control the volume, usually lowering it to a barely audible level).

More difficult with simple analog MDs. Consider the operation of the Tesoro Lobo discriminator .

Analyzing the circuit, it is easy to see that this is an ordinary metal detector of the middle class. It was built according to one of the classic schemes that were widely used in the 80s of the last century. However, his scheme is not the worst.

So, the signal from the generator with external pumping, assembled on the IC U 11 and the transistor VT1 , is fed to the transmitting coil with parallel resonance. The transmitter receives a sinusoidal signal with an amplitude of about 6V (at a higher voltage , an overload of the U2B IC is possible ). It is fed to the control signal generator. Y channel shapers are assembled on the U5A and U6A ICs . The signals are out of phase, apparently so that one channel does not affect the other – the effect is possible due to the use of transistors VT2 , VT3 as switches of synchronous detectors. Both signals are phase-shifted as the variable resistor G rotates . b .

The signal from the receiving coil goes to the receiving amplifier, assembled on the IC U1.A, U1.B and then to the buffer stage on U2A . Between receiver and buffer there is a sensitivity regulator SENS . This is a good circuit, it has enough headroom and allows you to use a non-resonant receiver coil. From the output of the receiver, the signal is fed to synchronous detectors on transistor switches VT2 (dcY), VT3 (dcX) and VT4 (dcD) .

Ground balance                                                                                                      

Let’s see what happens to the dcX and dcY signals as the phase shift of the control signals increases relative to the transmitter signal. As a target, use the “hot stone”. At first, the signal in the dcY channel is very large. By increasing the phase shift (on the oscilloscope, the signal on the GB resistor “rides” to the right), you can see that the signal dcY decreases, and dcX increases. With some phase shift, and it depends on the phase shifts of the sensor and receiver, the signals dcX and dcYwill be related as 8:1. Such signals are shown in the first figure. This is a characteristic point often used in serious MDs. If we add these signals to the simplest resistive adder with the same ratio of input resistors, we get a zero signal at the output ( as in the third figure)

dcX : dcY = 8 : 1                                                                        dcX : dcY = 16 : 1

If we continue to shift the phase of the control signals, then the ratio dcX:dcY will increase and at a value of approximately 16:1, it is already possible to start searching on light soil. Next, there will come a moment of fragile balance of the soil, when the signal ratio dcY will be equal to zero when the sensor moves over the soil. This case is shown in the third figure. I called the balance “fragile” because everything affects the amplitude of the dcY signal: soil mineralization, and sensor “leaving” as a result of heating – cooling, and changes in the transmitter frequency … and the phase of the moon is also of great importance 🙂 Therefore, the GB knob must be periodically turned to restore the balance of the soil.

                                               dcY =0. Soil balance                                            dcX : dcY > 16 : – 1

Well, the fourth figure shows the maximum phase shift, at which it is still possible to search without missing targets. Of course, the ratio of 16:-1 is very conditional. I won’t mind 🙂 if someone says that even with 6:-1 a normal search is possible.

But back to the Lobo scheme . On the transistor VT2, an LED of the search channel Y is assembled . With the right tuning for any metals, we will get a negative impulse and no reaction to the soil. At the beginning of the search , the resistor G. b . “unscrews” counterclockwise so that when the sensor approaches or moves away from the ground, a signal sounds. Then the resistor is smoothly rotated clockwise, while continuing to move the sensor until the operation stops. If you “twist” the resistor more, then metal passes will begin. The ground balance is carried out in AllMetal mode .

The search channel is assembled on OS U3, U4 . An amplifier with zero balancing is assembled on the op-amp U3 . By choosing different balance time, it is possible to realize the modes ” Retune” – fast balance, “All Metal/Disk” – average balance time and “Pinpoint” – slow balance. On the op-amp U4.B, the second amplification stage is assembled, on the op-amp U4.A – a mixer of the signals of the search channel and the discriminator. The bias applied to the input of the op amp from the resistor “Tuner” , you can set the volume of the background sound. Then the signal goes to the ULF.

The Y     channel of the discriminator is assembled on VT3 (synchronous detector) and OU U7.A, U8.A, U 9 .A (bandpass filters, center frequency 6 Hz). At the output of U9.A , for any purpose, there will be a signal of positive polarity.

Channel X of the discriminator is assembled in a similar way – transistor VT4 , op-amp U7.B, U8.B, U 9 .B . But this is not an ordinary channel X, the phase of the control signal of which is shifted relative to the Y signal by 90 ° . Here, the phase shift varies from 0 ° to 154 ° relative to the transmitter signal. The shift is implemented using a phase shifter on the op-amp U2.B , an adder on the DISC resistor and a comparator U6.B. Let’s call this channel  dcD ( for “discriminator”).

At the output of the VT3 synchronous detector , when the resistor DISC is rotated , metals and mineral-containing soils will be sequentially “cut off”. If we arrange the metals in series (in two quadrants of the control signal drX ), then we get approximately the following chain: ferrite (light soil) – hot stones – iron – aluminum – foil – gold – copper. With a certain phase shift of the control signal, the polarity of the signal at the output of the synchronous detector will change to the opposite. Thus, if you set the phase shift so that the reaction to the foil disappears, then the device will not react to aluminum, hot stones and soil. Note that immediately after the foil comes the gold. So the discriminator must be used with care…

So, at the output of the channel dcD , with the “unscrewed” potentiometer DISK , we will receive a signal of negative polarity for all objects. Accordingly, the signal at the output of U9.B will be positive.

Both signals from the acX and acY channels are fed, respectively, to the non-inverting inputs of the comparators U10.A and U10 . b . A certain threshold voltage is applied to the inverting inputs so that there are no chaotic operations. The outputs of the comparators are combined so that a logical OR function is obtained. This means that only two signals of positive polarity will give a positive signal at the output. If one of the signals is negative, then the output signal will be negative.

Let me remind you that with the selected position of the DISK control , some metals will give a negative signal in the dcD channel (in the Y channel it is always positive). This negative signal will keep the output signal of the comparators at a negative level and, acting on the mixer U4.A , blocks the audio signal. This is how the discriminator works in analog metal detectors.

In more complex MDs, the ground balance is performed somewhat differently. As an example, consider a typical example of a ground balance circuit that was used in Whites’s 6000 series MD (4900 to 6000 to be exact).

The op-amp U3D implements an adder, from the output of which the signal dcG (or simply G ) is fed to the target identification filter. As you can see, the operation of the circuit is based on the assumption that the soil signal lies in the range dcX : dcY  from 5:1 (for the top position of the RGB slider) to >16:1 (for the bottom position). Ground balance, as in the previous scheme, is performed by raising and lowering the sensor above the ground and rotating the RGB slider . This rotation attenuates the dcX signal , increasing the ratio dcX : dcY . This is shown in the top picture. Blue signal – dcY, it is always positive. The blue, red and green signals refer to the dcX channel , for different positions of the RGB slider . It can be seen that when the ratio dcX : dcY is approximately 8:1, the signals are in balance.

Now, if you move the sensor closer or further from the ground, the level of signals in both channels will change, but their ratio will remain unchanged.

Why is such a land balance scheme applied? After all, the previously considered one works well, and the disadvantages of these schemes are approximately the same. The matter is that in MD White’s 6000 VDI is calculated . And, if you do not make the dcG channel separate from dcX and dcY , then turning the RGB knob will change VDI . Moreover, the scale will become non-linear. In general, the error reaches 9-10 VDI units .

In addition, if you examine both schemes for sensitivity to different targets, you will notice gaps and humps in the sensitivity of the first scheme (I remind you, as in Lobo ). The table below shows comparative characteristics of sensitivity.

sensitivity to the foil is alarming. But for gold it is 20% higher 🙂