Capacitance-based Wireless Sensor Mote for SnailPest Detection

D. Garcıa-Lesta, E. Ferro, V.M. Brea, P. Lopez, D. Cabello, J. Iglesias*, J. Castillejo*Centro de Investigacion en Tecnoloxıas da Informacion (CiTIUS)

Departamento de Zoologıa y Antropologıa Fısica*University of Santiago de Compostela

Santiago de Compostela, SpainEmail: daniel.garcia.lesta@rai.usc.es

Abstract—Pests due to terrestrial mollusks cause serious dam-age, both economic and ecological, in various types of agriculturalplantations. In this paper we develop a low cost capacitivesensor that wirelessly communicates with the base, to monitorthe activity of land snails. Once implemented physically, it hasbeen tested in a controlled miniplot with favorable results.

I. INTRODUCTION

The terrestrial snail is one of the possible pests to manytypes of agricultural plantations. To fight them, molluscicidesare periodically used, causing both huge economical andenvironmental damage [1]. Detection in an early stage easesthe correct use of pesticides, improving the quality of theproduct and decreasing cost and environmental impact.

Given that the snails look for refuge to rest and protectthemselves from their predators, providing them with sheltersand measuring their occupancy levels constitutes a goodstrategy to assess their presence in a crop. Wireless sensornetworks (WSN) are very suitable for this task.

Multimedia WSN with a camera and a companion chip ariseas a natural solution for agriculture monitoring [2]–[4]. Theirprogrammability permits to detect pests of many differentsizes and shapes under different environmental conditions. Thedrawback is that the design of computer vision algorithmsmight be lengthy, and the hardware is usually costly and powerhungry.

Sensors capturing 1D data are less expensive and draw lesscurrent. Ultrasound, photoelectric and capacitive solutions areamong the sensors that can be adapted to our shelter shapefor snail monitoring. Such a shelter is shown in Fig. 1. It isa PVC pipe allocated at tenths of centimeters above groundwith 25 cm height and 5 cm of diameter.

Based on our experience, commercial ultrasound sensorshave been discarded because their emitting beam producesmultiple reflections on the tube walls, leading to many falsepositives.

On the other hand, previous work of the authors in thefield of snail detection with WSN lies in the design of aphotoelectric sensor that triggers detection by the reflectionof IR light from the snails [5]. Although this solution worksvery well in darkness or on cloudy days, it fails during sunnydays due to the high IR background radiation.

Fig. 1. Shelter and WSN mote for snail detection.

This paper addresses this issue with a capacitive sensor thatworks during night and day conditions, designed to make partof a WSN. A similar capacitive sensor for counting bees thatcome in and out of the hive was presented in [6]. Althoughthe work in reference [6] does not provide data on powerconsumption, their conditioning circuitry based on an ACbridge leads to a relatively complex system with large footprintand high power consumption. In this paper we introduce acapacitive to digital converter (CDC) approach as conditioningcircuit. Its compactness allows fitting the conditioning andcommunication circuits at the top of the shelter. Its low powerbudget permits autonomy for more than 100 days with a 2000mAh LiPo battery without recharging.

This paper is organized as follows. Section II outlines thecapacitive sensor and its conditioning circuitry. Section IIIaddresses the design of the geometry of the capacitive sensor.Section IV describes the calibration of the system and showsexperimental results. Finally, we draw the conclusions fromthe paper.

II. SENSOR AND CONDITIONING SYSTEM

The sensor is an open plates differential capacitor at theentry of the shelter. Fig. 2 shows such a structure with aview of the cross section of the lower part of the tube.The capacitor is implemented with three copper electrodesinserted into three dents around the inner wall of the tube. An

978-1-4799-6117-7/15/$31.00 ©2015 IEEE

This full text paper was peer-reviewed at the direction of IEEE Instrumentation and Measurement Society prior to the acceptance and publication.

Fig. 2. Cross-section view of the lower part of the PVC pipe. dxp is theremainder thickness of the PVC; dxc, the thickness of the copper; d, thedistance between electrodes; and dy is the length of the electrodes.

additional outer wall of PVC with a copper film shorted toground protects the sensor from electromagnetic interferencesand avoids the detection of snails when they are creepingup the tube along the outside of its outer wall. The top andbottom spaces between the inner and outer walls of the tubeare sealed against water. The presence of a snail over copperelectrodes modifies the differential capacitance between themand it triggers detection due to the different dielectric constantof the snail when compared to that of the air. Although asingle capacitance would be enough to detect the presence ofthe snails, the measurement of the difference of capacitancesis necessary to discriminate whether the snail is going in orout of the shelter.

Fig. 3 shows the WSN mote with the sensor and itsconditioning and communication circuits. There are manyoptions to measure the change in differential capacitances [7].Capacitance to digital converters (CDC) are very compact andlow power solutions. The circuit chosen in this work was theAD7153 from Analog Devices [8]. This is a low power circuitthat features 12 bits of resolution with a typical current of 100µA for a supply voltage range [2.7, 3.6] V. The CDC chosenreaches 0.25 fF of resolution with 0.05% of linearity with afast enough selectable conversion rate for snail detection of 5,20, 50 and 60 ms. Also, it has two operation modes with 4selectable ranges, namely, a single-ended mode with a rangeup to 0.5 pF, 1 pF, 2 pF and 4 pF, and a differential mode with4 possible values of ±0.25 pF, ±0.5 pF, ±1 pF and ±2 pF. Thechip accepts up to 5 pF common-mode capacitance. This canbe balanced by a programmable on-chip digital to capacitanceconverter (CAPDAC). As will be shown in Section III theinput range of capacitances of the CDC is compatible withthe capacitances of the differential structure inserted into thePVC tube shown in Fig. 2.

The WSN mote is communicated with a base station throughthe combination of an XBee module with a low cost ArduinoFio running at 8 MHz and powered at 3.3 V [9], [10]. Thecommunication between the CDC and Arduino is implementedthrough the I2C protocol. Also, the calibration phase requires amultiplexer, in this case an SN74LV40153A. As Fig. 4 showsthe CDC and the multiplexer with all their required passive

Fig. 3. Schematic of the sensor with its conditioning and communicationcircuits.

Fig. 4. Conditioning system integrated on a PCB mounted over ArduinoFio.

Fig. 5. Copper electrodes and coaxial cables over the wall of the inner PVCpipe.

components are integrated onto a PCB which is directlymounted over the pins of the Arduino Fio, resulting in verylow footprint for the conditioning and communication circuits.Such circuits are inside a waterproof box at the top of theshelter (see Fig. 1), and connected to the copper electrodesthrough three coaxial cables that go between the inner andouter walls of the shelter (see Fig. 5). Finally, the system ispowered with a 2000 mAh LiPo battery, which is rechargedwith a 15 cm x 6 cm solar cell.

0 2 4 6 8 10 12 14 16

0,80

0,85

0,90

0,95

1,00

Max

imum

diff

eren

ce o

f cap

acita

nce

(A.U

.)

Distance (mm)

Distance between electrodes (d) Length of electrodes (dy)

Fig. 6. Simulation results of the maximum difference of capacitances betweenthe electrodes as a function of the distance between them (d) and their length(dy).

III. DESIGN OF THE SENSOR GEOMETRY

The working principle of the capacitance sensor is basedon measuring the difference of capacitances between certainelectrodes on the sensor. To do so, three electrodes arenecessary. As apparent, their sizes as well as the locationand relative distance among them will have an impact onthe sensor response. In order to obtain the optimal response,the geometry of the sensor has been optimized by means ofsimulations using the software FEMM [11]. The simulationswere performed varying the parameters dxp, dxc, d and dy(Fig. 2), and leaving the rest of parameters, like the size ofthe pipe or the width of the wall as boundary conditions.

The software FEMM solves electrostatic problems in twosymmetries: cylindrical and translational. As the snail goesinto the pipe, the cylindrical symmetry is broken, so wesimulated a planar situation just to optimize the response. First,we simulated the system in translational symmetry, taking thesnail as a rectangle of helix aspersa’s typical size (3 cmx 1.5 cm), with a relative dielectric constant of 75, loopingthe position of the snail all over the sensor. To calculatethe capacitances of the capacitors, we applied 5 V and -5 Vrespectively on the bottom and top electrodes, and 0 V on thecentral electrode. Extracting Q from the simulation, and usingthe equation C = Q/V , we calculated the capacitances atevery loop, and recorded the maximum value of the differenceof capacitances.

As expected, the results show that the bigger the distance be-tween the electrodes and their length, the bigger the maximumdifference of capacitances. This is because as the distance isbigger, more dielectric material is between the electrodes, andas the size of the electrodes increases, more density of freecharge is available to polarize.

In Fig. 6, we can see that the maximum value of capacitanceafter the snail has passed increases very strongly at low valuesof dy and d and, at a certain point, this response saturates.Then, these parameters will be selected at the saturation range,taking into account as well other considerations. For example,the smaller these parameters are, the smaller margin the snailwill have to stop in the middle of the sensor, giving with that

0,0 0,5 1,0 1,5 2,0 2,5 3,00,3

0,4

0,5

0,6

0,7

0,8

0,9

1,0

Max

imum

diff

eren

ce o

f cap

acita

nce

(A.U

.)

Copper thickness (mm)

Fig. 7. Simulation results of the maximum difference of capacitances betweenthe electrodes as a function of the copper thickness (dxc).

TABLE IFINAL PARAMETERS FOR THE SENSOR GEOMETRY.

dxp (mm)) dxc (mm) d (mm) dy (mm)2 1.5 8 8

an anomalous signal. In addition, reasonable parameters haveto be selected to ease the fabrication process.

The behavior of the response as a function of the copperthickness dxc (Fig. 7) is similar, but with a not so clearsaturation. Thus, the copper thickness will be selected outof the strongly raising zone, resulting sufficiently thin tomanipulate the electrodes easily.

For the remainder parameter, dxp, the simulations show thatthis magnitude has little impact on the overall system response,so that a value thick enough not to break the PVC wall in theimplementation was chosen.

Table I shows the parameters obtained after the optimizationprocess. The resulting equipotential lines inside the sensor forthese values both in absence and presence of simulated snailsare showing in Fig. 8. As seen, the presence of a snail producesa significant perturbation of the electrostatic field which ismanifested as a change in the difference of capacitance, shownin Fig. 9. As we will show in Section IV, this simulatedresponse coincides with the experiments.

Once the geometry was optimized, we simulated the systemwithout the snail using cylindrical symmetry. In so doing, wewill know the total capacitance of the electrodes, and we cancheck whether the capacitance values are under the maximumaccepted by the CDC. This value was estimated to be 5.6pF, below the maximum value allowed by the CDC using theCAPDAC functionality.

IV. EXPERIMENTAL RESULTS

The CDC is programmed to work in differential mode withthe widest full-scale range, which in the case of the AD7153is ±2 pF. This is is done through the use of internal registersof the CDC. Table II lists the internal registers of the AD7153.

Fig. 8. Equipotential lines in absence (a) and presence (b) of the simulatedsnail.

-20 0 20 40 60 80 100 120 140

-1,0

-0,5

0,0

0,5

1,0

Nor

mal

ized

diff

eren

ce o

f cap

acita

nce

(A.U

.)

Position of the snail's center (mm)

Fig. 9. Output of the simulation with the snail modeled as a rectangle withεr=75.

The CDC saves the measurement data in the internal registerData. This value is calculated through (1) and (2), where Codeis an internal digital word of the CDC and, as their namesuggest, Offset Reg and Gain Reg are the values stored inthe Offset and Gain registers of the CDC. The measurementprocedure ends with Arduino reading the contents of DataReg and producing the differential capacitance of the sensorthrough (3), with Input Range being the widest possible full-scale in differential mode, namely 4 pF.

Data Reg = (Code-(Offset Reg-0x8000))Gain+0x8000 (1)

Gain =216 + Gain Reg

216(2)

C(pF ) =Data Reg − 0x8000

0xFFF0· Input Range (3)

Before any reading the system is calibrated only once. Thecalibration process accounts for the offset, the gain, and thecommon-mode capacitance. In so doing, the multiplexer, andthe internal registers of the CDC are used.

The offset is caused by the CDC chip itself, by all theparasitic capacitances from the metal lines on the PCB, fromthe coaxial cables along with the wielding that connects themto the copper electrodes, and from the pins of the differentICs on the PCB, as well as by any mismatch caused by themanufacturing of the electrodes that make up the differentialcapacitive structure.

The gain is calibrated with the Gain registers to reachthe full-scale range of differential capacitances (±2 pF). Inour case, we have employed 2.7 pF and 4.7 pF standardcapacitances (see Fig. 3).

The common-mode capacitance was determined experimen-tally to be around 5 pF, overcoming the maximum commonmode capacitance allowed by the CDC AD7153, which is 5 pFtoo. Thus, compensation is needed. This is done by subtractinga safety margin of 2 pF with the CAPDAC registers on theCDC.

Before the deployment of the final system in the controlledminiplot, lab tests were performed to check the sensor. Insteadof a snail, a material with high dielectric constant was used(organic materials, such as fruits or meat, have a similarvalue of εr given their high water content). Fig. 10 (top)shows the measured values for the different possible scenarios:absence of simulated snails, entrance and/or exit of one ormore simulated snails. As seen, when a snail enters the shelter,the ∆C signal reaches a maximum, and then a minimumvalue before returning to its reference value. This situationis reversed when a snail leaves the shelter. Given the system’soutput, a simple signal processing algorithm counts the numberof snails present in the shelter. First, a threshold value isestablished and compared to the mean value of the last 30minutes. When a minimum value of ∆C is found after amaximum, then the snail counter is increased one unit anddecreased when the opposite is true. This is shown in Fig. 10(bottom).

TABLE IIREGISTERS OF THE AD7153 AND THEIR SUBADDRESSES

Register Name SubaddressStatus 0x00

Data MSB 0x01Data LSB 0x02

Offset MSB 0x05Offset LSB 0x06Gain MSB 0x09Gain LSB 0x0A

Setup 0x0BConfiguration 0x0F

CAPDAC POS 0x11CAPDAC NEG 0x12Configuration 2 0x1A

Fig. 10. Results of the lab tests.

Fig. 11. Final system installed in the controlled miniplot.

Once the validity of the system has been established, webuilt a prototype ready to be installed in the field. Lab testsshowed that the system is very sensitive to external pertur-bations. Thus, to avoid noise sources, principally caused bythe rain, the sensor and the conditioning and communicationcircuits were totally shielded with a copper layer of 30 µmthick connected to ground. As said before, both the box thatcontains the circuits and the top and bottom parts of the pipewere sealed against water.

The system was installed in a controlled miniplot, where thesnails could not scape. As we can see in Fig. 11, the activityof the snails was recorded by an infrared active camera and,with this recording, a comparison between the signal collectedand the video can be made.

The results show that the system is totally robust to thepass of snails over the surface of the device, detecting signalvariations only when a snail goes into the shelter. In Fig. 12we can see two signals corresponding to different inputs ofsnails. Depending on the mass of the snail, the amplitude ofthe waveform will be different, which permits us not to onlydetect the presence of snails but also to estimate their sizes.

A possible source of unwanted signals is the random move-

0 10 20 30-1,0

-0,8

-0,6

-0,4

-0,2

0,0

0,2

0,4

0,6

0,8

1,0

0 10 20 30 40-1,0

-0,8

-0,6

-0,4

-0,2

0,0

0,2

0,4

0,6

0,8

1,0

Diff

eren

ce o

f cap

acita

nces

(pF)

Time (min) Time (min)

Fig. 12. Signal corresponding to two snails of different sizes entering theshelter.

0 1 2 3 4 5 6 7 8-0,4

-0,2

0,0

0,2

0,4

WATERING

0 20 40 60 80 100 120-0,02

0,00

0,02

0,04

0,06

(a)

(b)

Diff

eren

ce o

f cap

acita

nce

(pF)

Time (min)

Fig. 13. Noise due to snail movements inside the shelter (a) and watering(b).

ment of snails inside the shelter, passing over one electrodewithout going out. In this case there are two possibilities. Thefirst one is that the snail turns around and goes again intothe shelter, leaving a peak in the data, like in Fig. 13 (a). Theother possibility is that the snail stops over the electrode. Inthat case, the signal will be constant and different from zero.If this offset is small enough, so that when another snail leavesor enters the shelter the signal does not saturate, it will notconstitute a problem, because the signal processing softwarecompares the raisings or fallings with the mean value of thesignal at this moment. However, if the signal saturates, thedevice will only be able to inform that there is snail activityin the shelter, but not the precise number.

If two snails pass over the electrodes at the same time,the observed response would be the superposition of the twoindividual signals, and the algorithm could make a wronginterpretation. This does not constitute a problem due to itsvery low probability and the fact that in the worst possiblecase, one sensor inside the entire network would miss onesnail, and to monitor the snail activity in a plot the base stationcollects data from many sensors.

18 00 06 12 18 00 06 12 18

−2

−1

0

1

2

Time (hour)

∆ C

(p

F)

18 00 06 12 18 00 06 12 18

0

2

4

6

Time (hour)

Nu

mb

er

of

sn

ails

Fig. 14. Number of snails inside the shelter as a function of time during a3 days long experiment.

In order to assess the robustness of the system under differ-ent environmental conditions, the plot was watered everyday.In Fig. 13 (b) we can see that the noise signal due to thewatering is much lower than the signal corresponding to asnail entering the shelter.

If we apply the signal processing software to a long track ofdata (around three days), we obtain the results shown in Fig.14, where we can see the number of snails inside the refugeas a function of time, showing the robustness of the system.

Finally, the electrical consumption of the system was mea-sured taking into account these data:

• System in sleep mode: 200 µA.• System in operation: 4.1 mA.• Data transmission process: 215 mA.Also, the system measures data every 3 s (2.7 s sleeping

and 0.3 s running), and sends them every 80 measurements.Sending the data takes 0.2 s, so that taking 80 measurementsand sending them takes 240.2 s. The charge consumed withinthis time is calculated through (4), amounting up to 51.3 µAh.

Q =1 h

3600 s· 80 · (2.7 s · 0.2 mA + 0.3 s · 4.1 mA) +

+1 h

3600 s· 0.2 s · 215 mA = 51.3 µAh (4)

This operation is performed 360 times a day, consuming18.45 mAh. This low consumption was achieved including aconnector between a digital output and the XBee socket toturn it off when the device is not sending data, increasing thebattery lifetime. As the battery has 2000 mAh, the autonomyof the system, without any external source of energy, is 108days. To increase the autonomy, Arduino Fio was connected toa solar panel with a maximum intensity of 100 mA. Then, withjust one hour a day working at 20% of its maximum capacity,the solar panel is enough to feed the system. Even more, thisestimation was made assuming that the system sends the data

to the base, and the processing is made on some other device.By integrating the signal processing on Arduino, it would onlyhave to send data if a snail has entered or exited, decreasingeven more the energy consumption.

V. CONCLUSIONS

In this work we have developed a wireless capacitivesensor mote for snail detection. After choosing the condi-tioning system, simulations were performed to optimize thegeometry of the sensor, obtaining the optimal parameters forthe implementation, taking into account also the viability ofthe physical implementation and the random behavior of thesnails. Tests were performed, obtaining good results. Thismeans that signals caused by any input or output of snailsis clearly differentiated from noise.

The requisites for the device were mainly three: energeticautonomy, wireless communication and low-cost. The first onewas achieved with the combination of low-power design and asolar cell. The wireless communication is accomplished withthe XBee socket, which has a range of 1.6 km. In addition,it could be programmed to have peer-to-peer communication,increasing even more its range in a WSN. Regarding the cost,the total price of this prototype is under 70e, which makes ita good candidate as a low-cost WSN mote.

ACKNOWLEDGMENT

This work has been partially funded by AE CITIUS(CN2012/151, European Region Development Fund, ERDF(FEDER)), GPC2013/040 ERDF (FEDER) and the TalentumStartups Program of Telefonica.

REFERENCES

[1] J. Castillejo, I. Seijas, and F. Villoch, “Slug and snail pests in spanishcrops and their economical importance.” in Slug & snail pests in agri-culture. Proceedings of a Symposium, University of Kent, Canterbury,UK, 24-26 September 1996. British Crop Protection Council, 1996,pp. 327–332.

[2] P. Tirelli, N. Borghese, F. Pedersini, G. Galassi, and R. Oberti, “Auto-matic monitoring of pest insects traps by zigbee-based wireless network-ing of image sensors,” in Instrumentation and Measurement TechnologyConference (I2MTC), 2011 IEEE. IEEE, 2011, pp. 1–5.

[3] V. Jelicic, T. Razov, D. Oletic, M. Kuri, and V. Bilas, “Maslinet: Awireless sensor network based environmental monitoring system,” inMIPRO, 2011 Proceedings of the 34th International Convention, May2011, pp. 150–155.

[4] R. Kays, B. Kranstauber, P. Jansen, C. Carbone, M. Rowcliffe, T. Foun-tain, and S. Tilak, “Camera traps as sensor networks for monitoringanimal communities,” in Local Computer Networks, 2009. LCN 2009.IEEE 34th Conference on. IEEE, 2009, pp. 811–818.

[5] E. Ferro, V. Brea, D. Cabello, P. Lopez, J. Iglesias, and J. Castillejo,“Wireless sensor mote for snail pest detection,” in SENSORS, 2014IEEE, Nov 2014, pp. 114–117.

[6] J.M. Campbell, D.C. Dahn and D.A.J. Ryan, “Capacitance-based sensorfor monitoring bees passing through a tunnel,” Measurement Science andTechnology, vol. 16, no. 12, p. 2503, 2005, http://stacks.iop.org/0957-0233/16/i=12/a=015.

[7] R. Pallas-Areny and J. Webster, Sensors and signal conditioning. J.Wiley, 2001.

[8] 12-Bit Capacitance-to-Digital Converter, Analog Devices, 2008, rev. 0.[9] Arduino Fio Board,

http://arduino.cc/en/pmwiki.php?n=Main/ArduinoBoardFio.[10] XBee, http://www.digi.com/products/wireless-wired-embedded-

solutions/zigbee-rf-modules/zigbee-mesh-module/xbee-zb-module.[11] D. Meeker, Finite Elements Methods Magnetics 4.2,

http://www.femm.info.