Measurement 47 (2014) 483–495

Contents lists available at ScienceDirect

Measurement

journal homepage: www.elsevier .com/ locate/measurement

Wireless sensor network for smart composting monitoringand control

0263-2241/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.measurement.2013.09.026

⇑ Corresponding author. Tel.: +351 218418485.E-mail addresses: jaime.oscar.casas@upc.edu (O. Casas), marga.

lopez@upc.edu (M. López), marcos.quilez@upc.edu (M. Quílez), xavier.martinez-farre@upc.edu (X. Martinez-Farre), gemma.hornero@upc.edu(G. Hornero), crovira@iris.cat (C. Rovira), mpinilla@iris.cat (M.R. Pinilla),pedro.m.ramos@ist.utl.pt (P.M. Ramos), bborges@lx.it.pt (B. Borges),hdsm@lx.it.pt (H. Marques), psgirao@ist.utl.pt (P.S. Girão).

Oscar Casas a, Marga López a, Marcos Quílez a, Xavier Martinez-Farre a, Gemma Hornero a,Carlos Rovira b, Mirta R. Pinilla b, Pedro M. Ramos c,⇑, Beatriz Borges c, Hugo Marques c,Pedro Silva Girão c

a Universitat Politècnica de Catalunya, Parc Mediterrani de la Tecnologia, Campus del Baix Llobregat, C/ Esteve Terradas, 7/8, 08860 Castelldefels, Spainb Iris – Innovacio i Recerca Industrial i Sostenible, Parc Mediterrani de la Tecnologia, Avda. Carl Friedrich Gauss n� 11, Castelldefels, Spainc Instituto Superior Técnico-UL, Instituto de Telecomunicações, Av. Rovisco Pais, 1, 1049-001 Lisbon, Portugal

a r t i c l e i n f o a b s t r a c t

Article history:Received 27 June 2013Received in revised form 2 September 2013Accepted 12 September 2013Available online 20 September 2013

Keywords:Wireless sensor networkTemperature measurementsMoisture measurementsContactlessContactless battery recharging

This article describes a wireless sensor network (WSN) designed and implemented to mon-itor and assist composting processes. After framing the problem, the objectives and char-acteristics of the WSN are discussed and a detailed description of the hardware andsoftware of network components are presented. The article ends with the presentationand discussion of results obtained with the network.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

The increase of the world population during the last50 years (from about 3 billion people in 1960 to over 7 bil-lion at present) and of the consumption of goods and prod-ucts led to the huge production of waste resulting fromhuman activity. The stress of this waste on the environ-ment is particularly important since it includes importantquantities of non-bio degradable and even poisonous ele-ments (e.g., gaseous products, plastics, metallic and glasspackages and containers, chemicals, etc.). Global warmingand pollution events affecting namely water supplies inall continents have been instrumental in the increase of

the political consciousness of dealing with all the problemsthat can endanger the world in which we live. On the otherhand, the increase in the demand of raw materials led notonly to the increase of their prices, but also to the eco-nomic viability of product recycling for reuse. It is in thiscontext that, in developed countries, local, regional, andnational authorities promote or even enforce solid waste(both industrial and urban) management policies with re-sources recovery as one of the objectives.

There are multiple types of organic waste produced bydifferent activities. This includes biowaste (e.g., plantmaterial, food leftovers, food waste), manure, sewagesludge from wastewater treatment, etc., that can be recov-ered through composting in order to decompose the organ-ic matter and transform the waste into a stable product,useful for agricultural or landscaping purposes. SinceDirective 1999/31/EC on the landfill of waste promotedthe reduction of biodegradable material in landfill, biolog-ical treatments were enhanced.

484 O. Casas et al. / Measurement 47 (2014) 483–495

In [1], compost is defined as an ‘‘organic soil condi-tioner that has been stabilised to a humus-like product,that is free of viable human and plant pathogens and plantseeds, that does not attract insects or vectors, that can behandled and stored without nuisance, and that is benefi-cial to the growth of plants’’. To obtain this product, bothquality of the raw materials and management of the pro-cess are of main interest; on one hand, the presence ofinorganic pollutants can limit the use of the productregarding regulations. On the other hand, even when thematerial is free of pollutants, if initial conditions (mix-tures, water) are not properly defined to ensure microbialactivity and there is no control of the process through dif-ferent parameters, the final quality will probably notachieve the standards of stability and hygienisation tobe safely used.

There are different systems that allow obtaining ade-quate compost. Generally, they are based on the monitor-ing of some basic parameters, such as temperature andmoisture, to ensure the quality of the final product andprocess efficiency. This paper presents a wireless systemto control temperature and moisture in composting pro-cesses. The system is composed of multiple nodes, placedwithin the composting pile. Each node measures the tem-perature and moisture and relays these values to a pile-hub device located outside the pile.

2. Composting

2.1. The composting process

Composting is an aerobic, biologic and thermophilicprocess that allows transforming a biodegradable materialinto a stable product, the compost, useful to improve thequality of soil as long as the process has been controlledand raw materials have enough quality. There are severaldefinitions of composting, more or less complex, but theaforementioned concepts include the main keys to be con-sidered in a composting process to produce quality com-post. This transformation involves the reduction inweight and volume through the aerobic conversion of theorganic compounds into CO2 and water, and the stabilisa-tion of the material.

Composting can be divided into two phases, decompo-sition and maturation, basically according to the evolutionof the material temperature. During the first phase, intense

Time

Tem

pera

ture

(°C

)

60

40

Mes

ophi

lic

Thermophilic Cooling

Stabilisation

Decomposition Maturation

Fig. 1. Temperature evolution during the two phases of the compostingprocess.

degradation is produced due to high microbial activity andthe increase in temperature is rapid (see Fig. 1). After thisperiod, the reduction in biodegradable material availableto microorganisms reduces the activity and the tempera-ture slowly reduces.

The decomposition phase typically lasts over 30 days inwindrow composting, while in vessel composting it can beshorter (about 15–20 days) due to more intensive produc-tion conditions. During this stage, most of the compostingwindrows have a height of 1.5 up to 3 m, depending on theturning machinery, but can even be higher, between 3 and4 m high, when they are not turned, as in extended piles.Other static systems different from in-vessel can also sup-ply the air without turning through perforated slabs, so thesize of the heaps can vary significantly.

The moisture content should ideally be around 60% atthe beginning of the process, when the different organicwastes are mixed. If moisture content exceeds 60%, thereis a loss in porosity and thereby the procedure becomesanaerobic. Fermentation will set in and odours will beemitted from the material as well as methane as a resultof the anaerobic decomposition of the biodegradablewastes. If the moisture content decreases below 50%, therate of decomposition decreases because nutrients mustbe in solution to be used by microorganisms. Maintainingthe correct moisture level during the thermophilic phaseof composting can be difficult, especially in an open-airwindrow system, due to dry or wet climatic conditions.When the composting activity is greater and temperaturesin the core of the composting pile are very high, watermust be added at regular intervals to support and maintainmaximum activity. It is critical to maintain the moisture le-vel near 60% in the initial period of composting. If themoisture level falls below 50%, composting activity willslow down and eventually cease as the material continuesto dry.

The second stage of the composting process is the mat-uration phase that, in most of the cases, lasts between 30and 60 days. There are however cases where this stagelasts longer. Along this phase, temperature will diminishand stable structures appear. The dimensions of the com-posting piles during this phase are not very different fromthe ones of the decomposition stage. When maturationphase starts, a decrease in biodegradable matter from thedecomposition can be observed. Even though the microbialpopulation needs oxygen to maintain its activity to achievestabilisation, turning or forced aeration is needed less fre-quently than during decomposition.

Complete monitoring and control of the compostingprocess involve the measurement of several quantitiesnamely, temperature, moisture, oxygen, pH, carbon-to-nitrogen ratio [2], and even volatile organic compounds[3]. Oxygen must be guaranteed since composting isan aerobic process, but minimum moisture content mustbe also assured to allow microorganisms to developtheir degrading activity, where temperature is its directconsequence. So, oxygen, temperature and moisture arethe most important indicators of how the compostingprocess is evolving. They must be correctly monitoredif one aims to improve the quality of the resultingcompost.

O. Casas et al. / Measurement 47 (2014) 483–495 485

2.2. Composting monitoring

In general, the monitoring and control of the compost-ing process are based only on the measurement of the tem-perature and moisture at few locations inside the piles.

2.2.1. Classical manual methodsCommercial temperature measuring systems for com-

post typically consist of a stainless steel rod that is intro-duced in the material by an operator. Two sensors in therod allow the characterization of the temperature at thecore and at the periphery of the material. The result is dis-played on an indicator or is sent via a wireless link to theplant control system.

The moisture content of a composting pile is usuallymeasured by the oven drying method, which is a gravimet-ric method. It consists in the constant temperature dryingof a material sample and the computation of the ratio ofthe evaporated mass (water) with respect to the initialmass of the sample. This ‘‘offline’’ measurement is expen-sive due to the high labour needed to take the sampleand the subsequent lengthy oven-dry method procedure.Actually, when online measurements are necessary, micro-wave sensors [4–5] or Time Domain Reflectometry (TDR)based systems [6–8] are the normal solutions. However,these systems are very expensive and their energy con-sumptions are very high to enable their use in one auton-omous sensor. To solve these limitations, newdevelopments have been introduced. The possibility touse the correlation between the changes of conductivityor permittivity of the material and the moisture has beenstudied and, in soil moisture measurements, it is possibleto find commercial systems that use this method. Oneexample is the Theta Probe ML2x of Delta-T Devices(www.delta-t.co.uk). The Theta Probe measures volumetricsoil moisture content by the well-established method ofresponding to changes in the apparent dielectric constant.The probes generate a radiofrequency signal of 100 MHz,which is applied to a specially designed internal transmis-sion line that extends into the soil by means of an array offour rods. However, determining moisture levels in com-posting media is more challenging than doing so in soil.The causes are: (a) the range of moisture levels in compost-ing situations is much broader than in soils, (b) conditionswithin a composting pile change considerably from onelocation to the next, and (c) the moisture held by organicmaterials is difficult to detect because of the complexarrangement of pores.

2.2.2. Automated methodsManually collecting data – measurement of tempera-

ture and moisture in the piles – is time consuming andmay expose persons collecting the data to harmful patho-gens. The development of wireless communicationsopened the door to distributed measuring systems com-posed of several measuring sites that transmit the col-lected data to a centralized location [9]. Such systems aredesignated by wireless sensor networks (WSN) and theapplication of the concept to composting monitoring andcontrol allows composting piles temperatures and mois-tures to be continually recorded and logged, improving

efficiency, reducing the time needed to complete a com-posting cycle, and minimizing human exposure and poten-tial risks.

The wireless sensor network described in the next sec-tion was designed within the framework of the EC fundedproject Compo-ball [10]. Prototypes of the system, trade-marked as SensoBall, were assembled and tested for dem-onstration purposes by the project research team takinginto consideration national and European legal require-ments for composting monitoring and the contributionsof the project industrial partners.

3. Network design

The objective of the Compo-ball project was to developa novel on-line wireless system for the measurement oftemperature and moisture at various points in a compost-ing material. With this objective in mind, the design andimplementation of a sensor network whose nodes areindependent and have temperature and moisture mea-surement capabilities was developed. The data gatheredin the nodes is transmitted via a hub-gateway link to abase station where it is processed.

Due to the harsh chemical and mechanical environmentin which the nodes have to operate, the sensors must beencapsulated in an inert material capable of withstandingthe harsh composting environment. This means that thenodes cannot have external wires to allow the transmis-sion of the data they acquire. Therefore, the network mustbe wireless and the nodes must be equipped with a trans-ceiver and the adequate antenna. The autonomy of thenodes, and the fact that all the electronics is encapsulatedwithout external access, means that the nodes must bepowered by a long-lasting rechargeable battery, and thecircuitry that enables its contactless recharge must be in-cluded in each node.

Fig. 2 illustrates the wireless sensor network: the nodes,hubs, gateway and server (base station).

4. Network components

The wireless sensor network, described in this section 4,consists of:

� A series of encapsulated nodes, each of them with atemperature and moisture sensor, a long lasting battery,and a charging coil for the inductive recharge of the bat-teries of the system, together with an antenna for thewireless communication, and an integrated PCB.� A charging plate for the external recharge of the nodes.� A communication network that includes a communica-

tion protocol between the node and a pile-hub, which isa gateway for the collection of data from the WSN andfor the transmission of the data to a central server.� A web-based user interface.

4.1. Nodes

Fig. 3 shows the block diagram of each node. The wire-less system block includes the transceiver and antenna

Fig. 2. Wireless sensor network for compost monitoring and control.

Fig. 3. Block diagram of a network node.

486 O. Casas et al. / Measurement 47 (2014) 483–495

that allows a node to communicate with other nodes andwith the network base station.

4.1.1. Temperature sensorIn order to optimize the accuracy and linearity of the

temperature system, a Resistive Temperature Detector(RTD) sensor (Pt1000) was used. As sensor conditioningcircuit, a direct sensor-to-microcontroller interface waschosen. This conditioning circuit has low energy consump-tion and, in front of classic amplitude conditioning, it re-duces considerably the number of components, which isa very important characteristic for an autonomous sensornetwork. As can be seen in Fig. 4, this conditioning consistson measuring the time that it takes a capacitor to becharged and discharged through the resistive sensor usingthe digital inputs and outputs of a low-cost microcontrollerunit and one of its timers to measure the discharge time.This time directly depends on the resistive value of thesensor (Rx)

Tx ¼ RxC � ln V1 � V2

VTL � V2

� �: ð1Þ

This method has some sources that contribute to themeasurement uncertainty. The main ones are:

(1) The tolerance and variability of the high state volt-age supplied by the MCU (V1), the low state voltage(V2), the low level threshold at which the MCU

considers that the level is low (VTL) and the capacityof the capacitor (C).

(2) The internal resistance value of the MCU pins.(3) The limitation due to the base clock used in the

counting of the discharge time (Quantization noise).This may cause that a small change in the resistanceof the sensor might not be reflected as a change inthe measured time.

To compensate the uncertainty introduced by V1, V2, VTL

and C, and to reduce the influence of the internal resistanceof the digital MCU pins, an auto-calibration procedure isused, the three-signal method proposed in [11]. It consistson charging the capacitor and discharging it through threedifferent pins (Fig. 5). One of the pin lines contains a refer-ence resistor Rref of 1.5 kO, another is short-circuited andthe last one is connected to our resistive sensor Rx. Theduration of each discharge cycle is measured and, usingthis value, the resistance value of the sensor is determinedusing

Rx ¼Nx � Noff

Nref � NoffRref ð2Þ

where Nx is the time of discharge through the resistive sen-sor, Noff is the time of discharge through the short-circuit,Nref is the time of discharge through Rref.

To reduce the quantization noise, the value of C maybe decreased. When doing so, some problems due to theslew rate may arise and affect negatively the speed ofthe circuit, extending the propagation times. For thisapplication, the optimal value of C is 1 lF. To reduce theinfluence of the noise in the voltage comparison, whichreflects on the precision of the timing process, a Schmitttrigger digital pin (P1 in Fig. 5) is used. The value of Ro

is dimensioned to limit the maximum current in the dis-charge stage.

Using a calibrated system, the resistance of onePt1000 sensor is measured, the maximum absolute erroris 0.1 O (about 0.03 �C), and the effective number of bitsis 13 (which is translated to an accuracy of 0.10 O or0.025 �C). Since in our application the calibration of eachtemperature node is undesirable, very low tolerance(0.05%) and low temperature drift components(10 ppm) are used. With this design it is possible to ob-tain 0.3 �C accuracy and 0.1 �C resolution in all the mea-suring range.

Another important factor to quantify is the effect of theprotective material on the sensor. This material produces adelay in the temperature measurement. The model ob-tained is a first order response with a response constant(s) that depends of the type and thickness of the material.For the rigid-polyurethane material used to encapsulatethe node, the measured delay constant is 12 min. This va-lue limits the sampling frequency of the system to around1 sample/hour. It is possible, at the expense of node power,to increase the sampling frequency using a simple softwaremethod to estimate the final value of the first order re-sponse sensor [12].

(a) (b)Fig. 4. Temperature sensor: RTD and Direct interface. (a) Charging state. (b) Discharging and time measurement stage.

MSP430Microcontroller

Pt1000

Rx

Rref

Ro

C

Digital pin. P1

Digital pin. P2

Digital pin. P3

Digital pin. P4

Fig. 5. Direct sensor-to-microcontroller interface circuit with the threesignal measurement.

O. Casas et al. / Measurement 47 (2014) 483–495 487

4.1.2. Moisture sensorThe possibility of using the correlation between the

conductivity or permittivity changes of a material andthe moisture has been studied and tested [13]. It is alsopossible to reduce energy consumption and cost by usinglow frequency radiofrequency signals (10–100 MHz). Inthe case of permittivity, it is possible to reduce the workingfrequency without reducing the sensitivity of the method[14]. These permittivity changes can be measured withthe study of the attenuation of the RF signal transmissionor with capacitive methods. The main problem in the useof the transmitted signal is that the measurement dependson the exact knowledge of the distance between balls andthis knowledge requires having accurate systems of loca-tion, which makes the consumption of the sensor systemsto increase enormously. For this reason, a capacitive mois-

ture measurement system is more attractive and was usedin this work.

Recent work [15] has also shown that the capacitivemethod is a good alternative method for moisture mea-surement with autonomous sensors in soils because of itslow energy consumption. Capacitive sensors consist essen-tially of a pair of electrodes (either an array of parallelspikes or circular metal rings), which form a capacitor withthe compost or soil acting as the dielectric in between. Thiscapacitor works with an oscillator to form a tuned circuit,and changes in compost water content are detected bychanges in the operating frequency. Generally, capacitivesensors enable more freedom of choice in the design ofthe electrode geometry and operating frequency than inthe case of TDR systems. The influence of the changes inpermittivity due to the density changes in the compostmaterial will be compensated by the calibration of thesystem.

The conditioning circuit used is based on a square signalfixed-frequency oscillator where the changes in permittiv-ity are measured as changes in the phase between the cre-ated signal and the capacitor voltage. These phase changesare directly proportional to the changes in material permit-tivity and have a good correlation with compost moisturechanges when a 50 MHz frequency is used. This fixed-fre-quency design needs two circuits: an oscillator and a phasecomparator. In order to reduce the number of componentsand the system energy consumption, a simple crystal oscil-lator was used. The block diagram of the circuit is shown inFig. 6.

The phase between the measured signal (B) and theoscillator signal (A) is measured with an exclusive-Or Gateand converted to a DC-value with a passive first-order low-pass filter. The filter output voltage is connected to the ADCinput of the microcontroller to measure the DC value. Allthe selected components are low-tolerance and have low

50 MHzOscillator circuit

Phase comparator

circuit

Low-passfilter

Measuringelectrodes

B

A

Fig. 6. Fixed-frequency moisture sensor.

488 O. Casas et al. / Measurement 47 (2014) 483–495

temperature drifts. The values of the components used inthe measuring and phase circuits allow adjusting the dy-namic range of the output signal.

As the measuring electrodes are not in contact with thecompost, the protective material (with width d and dielec-tric contact er) creates a capacitance between the elec-trodes and the material. This capacitance is located inseries with the equivalent capacitance caused by the sur-rounding compost media and thus reduces the sensitivityof the implemented system. Therefore, an extensive effortin their design and validation was required to optimizethe sensor sensitivity. Three alternative shapes were stud-ied: rectangular, interdigitated and ring sensors. Sometests were carried out in order to see which shape of theelectrodes would be more suitable for the specific applica-tion of measuring compost moisture.

Rectangular copper electrodes were firstly used butthese electrodes were discarded since sensitivity to humid-ity changes increases in the interdigitated with a factor of10 compared to them. The main problem of interdigitatedelectrodes is the effect of the fringing factor, which de-pends on the geometry of edges of the electrodes. This ef-fect can be neglected when the distance betweenelectrodes is small enough. The main drawback is thatthe sensitivity and the capacity order are reduced. Anotherproblem is to adjust the shape in the circular form of oursystem. To solve this last problem, ring electrodes can beused. Many sizes were studied with COMSOL simulatedsoftware and the most sensitive case was used. The finalelectrodes location in the node, as well as a front side viewof the electrodes, are shown in Fig. 7. The electrodes con-

(a)

Electrode 1Electrode 2

Fig. 7. Concentric electrodes inside the node to measure the compost moisturcapacitive sensor for moisture measurement.

sist of a circle with 30 mm of radius with a central positiveelectrode (23 mm radius) – Electrode 2 in Fig. 7, and exter-nal ground electrode (5 mm width) – Electrode 1 in Fig. 7.

The transfer function of the transducer (sensor and con-ditioning circuit) was obtained in a laboratory environ-ment measuring the filter output DC voltage for sevendifferent moisture values in the desired moisture range.From these values, a lookup table is used within each nodeto estimate the moisture content from the filter output(transducer inverse model). Intermediate values are esti-mated through linear interpolation.

4.1.3. Wireless communicationSystem operation is based on the measurement of the

compost temperature and moisture in each node and thenon the nodes relaying their measurements among them-selves according to the current flexible network topologyuntil they communicate with the pile-hub (Fig. 8) that islocated near each pile. Communication between the nodesand the pile-hub is done at 433 MHz in the industrial, sci-entific and medical (ISM) radio band, which is a limited,low-power, unlicensed internationally band reserved forthese purposes. The transceiver chosen for this applicationis the CC430 from Texas Instruments.

The data of each pile are gathered in each pile-hubwhich consists on a simplified computer system board(ALIX 3d3) with an AMD LX800 processor operating at800 MHz, 256 MB of RAM, multiple USB, LAN and serialports (Fig. 9 left). This computer runs a Linux installationwhere a Java developed program communicates with anidentical node PCB (Fig. 9 right) through a serial connection

(b)

Compost media surrounding the

node

e. (a) Top view of final implementation. (b) Principle of operation of the

Fig. 8. Pile-hub ready to work in harsh environments.

Fig. 9. ALIX base station on the left and PCB node on the right, both aretogether inside the pile-hub box.

O. Casas et al. / Measurement 47 (2014) 483–495 489

and by Wi-Fi with the plant gateway computer. The gate-way computer can gather the data from multiple piles ineach plant in an internal database and publish the dataon the internet for everyone with authorized access tomonitor the composting processes.

The CC430 family of ultra-low-power microcontrollersystem-on-chip with integrated RF cores consists of sev-eral devices featuring different sets of peripherals targetedat various low-power wireless applications. The easy-to-use MSP430 microcontroller architecture, with up to sevenlow-power modes, is optimized to achieve extended bat-tery life in portable applications. The CC430 devices aresub 1 GHz RF + microcontroller system-on-chips combin-ing the excellent performance of the CC1101 <1 GHz RF

transceiver with the MSP430 CPUXV2 microcontrollercore. They have up to 32 kB of in-system programmableflash memory, up to 4 kB of RAM, two 16-bit timers, ahigh-performance 10-bit or 12-bit A/D converter witheight external inputs plus internal temperature and batterysensors. This device is used in each node and in the pile-hub for communication with the pile nodes.

4.1.4. Power supplyOne of the most important issues in the development of

a wireless sensor network system is the power budget ofeach individual node. In the specific case of the system de-scribed in this paper, the issue is even more relevant sincethe nodes must remain in operation throughout the com-plete composting process. This is due to the fact that it isnot practical for plant operators to find all the nodes inthe middle of the process just to recharge them. In thiscontext, the node power supply and its budget were oneof the initial issues addressed in the system development.The first requirement is for the nodes to be able to operateat least for 90–120 days. The second requirement is thatthe battery should be able to operate well at 60 �C andwithstand operating up to 80 �C. The third requirement isthat each node should be recharged contactless and within8 h. Node operations were divided into four categories: (i)measurement mode; (ii) communication mode: (iii); sleepmode and (iv) off mode. Sleep mode corresponds to thetime when the nodes are in the compost but are neithermeasuring nor communicating and are waiting to awakeup in the next schedule time, measure and communicatewith other nodes. Off mode corresponds to the operationafter battery charging and before the use of the node with-in the compost. Measurement mode corresponds to theactivation of both sensors (temperature and moisture),acquisition of their outputs, and estimation of the temper-ature and moisture values. The communication mode cor-responds to the mode where the node is communicationwith the other nodes and eventually the pile-hub.

Given initial estimates of the power consumption in themeasurement mode (20 mA at 2.5 V during 2 s), communi-cating mode (80 mA at 2.5 V during 20 s) and sleep mode(0.5 mA at 2.5 V during the rest of the time), and assuming24 measurements and communications each day (1 everyhour), common use rechargeable batteries that have capac-ity of 0.75 Ah would last around 30 days. In order to im-prove system autonomy, several battery manufacturerswere contacted and the final choice was the SAFTMP144350 which has a capacity of 2.6 Ah and can operatewell at 60 �C. Further tests demonstrated that the batterycould operate within specifications at 80 �C. With this bat-tery, predicted system autonomy is around 100 days. In thecircuit, after the battery, a voltage regulator ensures thatthe system control processor is supplied with 2.5 V. Theregulator (ADP121-25) also has an enable port which isused to put the node in off mode using a wired OR connec-tion and a digital output port of the microcontroller. Thismethod is used to ensure no current consumption in theoff mode and proper operation after node activation. Thebattery is charged using a dedicated integrated circuit(MCP73831). The constant current stage of the batterycharging process is limited to 0.5 A to ensure full battery

Fig. 10. Schematic circuit of the contactless charging system.

490 O. Casas et al. / Measurement 47 (2014) 483–495

charge within 8 hours. Battery voltage is monitored duringnormal operation by the microcontroller using one of itsanalogue inputs. The contactless battery recharging systemmust provide 5 V and 0.5 A at the charging IC input.

4.2. Base station

The base station is composed of the main pile-hub and aserver where all data are saved in an SQL database. Thebase station is a computer (pile-hub) that receives and pro-cesses data from all the nodes. The pile-hub is able to workin extreme weather conditions (IP67), saves all the infor-mation transmitted by the nodes, and sends data to theserver. The server collects all data from the different pile-hubs of the composting plant and processes them to dis-play on the internet the information pertinent for com-posting process control and management. The databasestores the relevant fields characterizing the system nodes(Node ID, Node Temp, Node Moisture, Node Battery Level,Node Signal Strength, etc.), which can be accessed directlyfrom the web-based interface.

The program in the base station connects directly via aWiFi link created in the composting plant to a server wherethe user interface is hosted. A Java program reads from theRS232 serial port the values that the Compo-ball PCB getsfrom the nodes and stores all the information directly inthe server database. It also stores network info such aswhen the user has pressed the reset button or when anew node has joined the network.

In addition to real-time monitoring, another objectiveof this project is to preserve all information measured bythe nodes for further processing and eventual certification.A graphical user interface (GUI) allows real time monitor-ing of temperature, moisture, approximate location, andbattery level of all active nodes. This information allowsthe operator to act at any time on the composting plantin order to optimize the composting process.

4.3. Battery recharging system

As previously explained, due to the adverse conditionswhere it must be immersed, the electronic circuitry of eachnode is encapsulated without external access. The nodesare therefore powered by a long-lasting rechargeable bat-

tery that is recharged by a contactless system specificallycreated for this application.

A contactless energy transfer system enables the transferof electromagnetic energy between two circuits that do nothave any mechanical contact among them by means ofinductive coupling (loosely coupled transformer). In thepresent application, the contactless charging system hastwo independent circuits: the transmitter (chargingplate – that includes a high frequency DC-AC drive con-verter, a compensating circuit, and a primary inductor wind-ing), and the receiver (each node – that has a secondarywinding, an output rectifier, a capacitive filter, and a voltageregulator before the IC charging device). Fig. 10 presents thecircuit of the contactless charging system. The load repre-sents the circuit from the voltage regulator onwards.

Due to the distance between the transmitter and thereceiver (air-gap), contactless energy transfer systems usu-ally have worse performance than non-contactless ones.This is mainly due to the weak magnetic coupling betweentransmitter and receiver inductors that requires the pro-cessing of a greater amount of energy (reactive) whencompared to traditional wired solutions.

Physically, the magnetic coupling coefficient k evi-dences how effective the magnetic coupling is and isrelated to the values of the mutual inductance and theself-inductances, varying from 0 to 1 (0 < k < 1) [16–17].Theoretically, the maximum value for the magnetic cou-pling coefficient is 1, when considering an ideal trans-former, when there are no leakage inductances of theprimary and secondary windings. Usually, practical trans-formers exhibit lower, but very close to unity, magneticcoupling coefficients, because of the dispersion of themagnetic flux lines.

In contactless energy transfer systems, due to the highvalue of the air-gap, this coefficient can be very low, reach-ing values that are prohibitive. As a consequence, in whatconcerns the coupled inductors equivalent circuit, theleakage inductances values increase, leading to the in-crease of the value of L1x shown in Fig. 10. The magneticcoupling is very sensitive to the winding dimensions andgeometry, especially when the magnetic path is air. Inthe cases where ferromagnetic cores are used, their mag-netic properties and dimensions must also be considered.In particular, the relation between the dimensions of themagnetic cores and the dimensions of the air gap can

Fig. 11. Loosely coupled transformer: (a) ferrite cores with large relative dimensions air-gap; (b) electromagnetic simulation result showing the magneticflux lines closing on the primary and only a few reaching the secondary.

Fig. 12. Coil configurations for optimizing the magnetic coupling coefficient when dealing with large air-gaps: (a) planar spiral, (b) globe spiral and (c)solenoid.

Table 1Coils physical dimensions.

Magnetic parameters [D = 20 mm]

Spiral planar Spiral globe Solenoid

Rec. Trans. Rec. Trans. Rec. Trans.

Inner diam. [mm] 10 10 0 0 85 125Outer dia. [mm] 70 70 62.5 102.5 85 125High [mm] 0 0 31.25 51.25 20 20Nr. of turns 20 20 20 20 20 20Coil length [m] 2.5 2.5 2.5 4.1 5.3 7.9Coil area [cm2] 37.7 37.7 61.4 165.0 54.4 78.5Coil volume [cm3] 0 0 63.9 281.9 113.5 245.4

O. Casas et al. / Measurement 47 (2014) 483–495 491

determine which type of solution must be used: air or fer-romagnetic core. Normally, the nature of the applicationand the amount of energy to be transferred may definewhich type of solution is needed.

For the application described in this paper, a 100 mm diam-eter spherical case to encapsulate the electronic circuitry of thenode, together with the receiver circuit and the rechargeablebattery [16], was firstly considered. Another initial specifica-tion was an air-gap of 20 mm to take into account the neces-sity of robustness for the mechanical structure. Varioustransformers characterizations, simulations, and experimen-tal tests were conducted considering ferrite cores. However,the extremely high air gap caused the majority of magneticflux lines to never reach the secondary, as is shown in

Fig. 11. Therefore, for this application the use of coreless induc-tors to avoid saturation problems was analysed.

In accordance, three geometric configurations for thecoils were studied and prototypes were built [17]: planarspiral, globe spiral and solenoid – Fig. 12. The dimensionsare presented in Table 1.

The first approach for the design considered the case oftwo concentric planar spiral windings, Fig. 12a). The mutualinductance M was calculated to verify the accuracy of theexperimental characterization. However, due to the rela-tively high separation distance, the resulting transformerpresents a very weak coupling coefficient. The obtainedmagnetic parameters are presented in Table 2 for the threecoil configurations [17].

Table 2Measured magnetic parameters.

Magnetic parameters [D = 20 mm]

Spiral planar Spiral globe Solenoid

Rec. Trans. Rec. Trans. Rec. Trans.

Self lnd. [lH] 13.5 13.5 11.5 17.1 54.0 88.9Mutual Ind. [lH] 3.37 5.57 24.59nx – turns ratio 0.25 0.49 0.46Llx [lH] 12.66 14.40 77.70Lmx [lH] 0.84 2.70 11.20k 0.25 0.40 0.36k/length [10 m�1] 0.99 0.99 0.97 1.59 0.45 0.66k/area [10 m�2] 6.62 6.62 2.41 6.48 4.52 6.65k/volume [103 m3] 1 1 1.41 6.22 1.45 3.13cx [nF] 270 22 150f0 [kHz] 86.05 285.44 46.72Vi [V] 3.8 10.2 8.5

(a)

(b)

Fig. 13. Output terminals characteristics for different frequencies with aligned and not aligned coils: (a) output voltage versus charging current and (b)output power versus charging current.

492 O. Casas et al. / Measurement 47 (2014) 483–495

Antenna Battery

Secondary coil

Primary coil

Charging plate box

Charging plate PCB

Charging plate cover

Charging plate alignment

module

Fig. 14. Layout representative of the charging plate and node placementduring charging procedure.

O. Casas et al. / Measurement 47 (2014) 483–495 493

With the objective of increasing the magnetic coupling,two spiral spherical coils were also tested, Fig. 12b), takingadvantage of the spherical shape of the node. This opti-mizes the windings useful area, and therefore increasesthe coupling factor, without compromising the devicedimensions.

The spiral globe configuration, although the one withthe best coupling factor, turned out to be unpractical forfabrication. Moreover, due to space limitations and EMI(ElectroMagnetic Interference) considerations, it wasdecided to use the solenoid configuration presented inFig. 12c), at the expense of a slight loss in the magneticcoupling coefficient (from 0.40 to 0.36).

The electrical circuit to drive the solenoid is representedin Fig. 10 and acts as an AC voltage source operating at a

Fig. 15. A data gathering tree and it

fixed frequency close to the resonant frequency definedby the inductance Llx and the resulting capacitance Cx.The capacitance Cx works as the compensator element forthe high leakage inductance Llx of the loosely coupledsolenoid.

Since the behaviour of the system is sensitive to thecoils alignment, parameters drift, and also to load varia-tion, some experimental results were obtained by varyingthe load for two different frequencies (174.2 kHz and172.2 kHz), and two alignment situations (aligned coilsand misaligned coils). The measured results are presentedin Fig. 13. They show that even if the alignment is compro-mised, the adopted solution satisfies all design restrictionswhen working at the proper frequency. To avoid unneces-sary misalignments, and therefore relaxing the necessaryaccuracy of the operating frequency, the geometry of thecharging plate was designed to assure proper alignmentbetween the charging plate and the nodes.

The built charging plate is shown in Fig. 14. The charg-ing plate box, cover, and alignment module form a cuboidwith 200 mm � 200 mm base and 132 mm height. On topof the plate, a not yet encapsulated node is shown to depictthe secondary coil location. After encapsulation, the twocoils are aligned.

5. Network management

The wireless system protocol between nodes is basedon the DMAC protocol [18]. The principal aim of DMAC isto achieve very low latency but good energy efficiency.DMAC can be summarized as an improved Slotted Alohaalgorithm where slots are assigned to the sets of nodesbased on a data gathering tree, as shown in Fig. 15. Hence,during the receive period of a node, all of its child nodeshave transmit periods and contend for the medium. Lowlatency is achieved by assigning subsequent slots to thenodes that are in succession in the data transmission path.

The robustness of the DMAC protocol is improved bythe inclusion of two mechanisms: (i) one to handle theproblem when each single source node has low traffic rate

s DMAC implementation [18].

Fig. 16. Example of temperature measurements in a composting pile.

Fig. 17. Example of moisture measurements in a composting pile.

494 O. Casas et al. / Measurement 47 (2014) 483–495

but the aggregate rate at an intermediate node is largerthan the basic duty cycle; (ii) another to solve the problemof the interference between nodes on the differentbranches of the tree. This last mechanism consists of a sep-arate control packet, named MTS (More to Send), thatmakes all the nodes on the multi-hop path to remain activein case of node failure due to interference.

6. Network testing: system demonstration

Prototypes of the several components described in theprevious sections needed for network operation and main-tenance were assembled and individually tested. The over-all network was afterwards tested in several compostingsites in The Netherlands, Austria and Spain. Figs. 16 and

17 show the webpage associated to the Senso-ball systeminstalled in Manresa (Spain) when temperature and mois-ture values were measured.

The temperature and moisture sensors proved to be ro-bust and with very good performances within their operat-ing ranges: temperature accuracy of 0.1 �C in the [0,80] �Crange, and moisture 5% accuracy in the 20–80% RH range.

The sensing nodes battery recharging system per-formed well according to the pre-defined requirements.During the tests, the autonomy of the nodes, reflected inthe time between batteries recharging, was smaller thandesigned. Possible explanations for the reduced autonomyof the nodes are the high temperatures of the compost,which may affect the performance of the batteries, andthe communication protocol, which may not maximise

O. Casas et al. / Measurement 47 (2014) 483–495 495

the amount of time that the microcontroller is in the sleepmode and thus consuming less power.

As mentioned, the nodes of the network are inside thecompost, a medium whose composition and densitystrongly limits the communication between nodes to somemeters. The relative position of the nodes changes ran-domly when the compost is revolved. For practical andeconomic reasons, the number of nodes in a, for instance,100 m long, 3 m wide, and 2 m high pile should be limitedto around 10. Under these three constraints, it is only nat-ural that the major problem faced when testing the Senso-ball system has been the communication between the pile-hub and the nodes. The network management solutiontested demonstrated the viability of the system and itssuitability for some compost sites, but alternatives mustbe considered to improve the overall performance of thesystem.

7. Conclusion

This paper reports the development and implementa-tion of a wireless sensing network designed with the pur-pose of providing data to assist composting process controland management. The target application requires nodesextremely robust both mechanically and electrically,which presented the following design and implementationchallenges from the electrical/electronic point of view: (a)measurement of temperature and moisture of a mediumwith sensors encapsulated in a synthetic material; (b)wireless communications from/to within the capsulematerializing each node; (c) long term power supply ofthe nodes with contactless recharging capability. Theelected solutions, described in the previous sections, per-formed adequately, satisfying the design specifications.

The prototype of the system, which constitutes an inno-vative and currently the most advanced solution for com-posting control and management, tested very positivelyand although with margin for improvement, reached acommercialization phase. Communications between thepile-hub and the nodes must be refined to assemble a moregeneral purpose compost monitoring system with largerautonomy.

Acknowledgments

The research leading to the above-mentioned resultshas received funding from the European Union SeventhFramework Programme (FP7/2007-2013) under grantagreement no. 243625.

References

[1] Roger T. Haug, The Practical Handbook of Compost Engineering,Lewis Publishers, Boca Raton, FL, 1993.

[2] Saleh Ali Tweib, Rakmi Abd Rahman, Mohd Sahaid Kalil, ‘‘ALiterature Review on the Composting’’, in: 2011 InternationalConference on Environment and Industrial Innovation, IPCBEE, vol.12, 2011, pp. 124–127.

[3] A. Lay-Ekuakille, A. Trotta, Predicting VOC concentrationmeasurements: cognitive approach for sensor networks, IEEE Sens.J. 11 (11) (2011) 3023–3030.

[4] Anssi Toropainen, Pertti Vainikainen, Ebbe Nyfors, ‘‘Microwavehumidity sensor for difficult environmental conditions’’, in: 17thEuropean Microwave Conference, 1987, pp. 887–891.

[5] C. Kawalec, M. Pasternak, ‘‘Microwave saw humidity sensor’’, in:International Conference on Microwaves, Radar & WirelessCommunications, May 2006, pp. 662–664.

[6] Scott B. Jones, Jon M. Wraith, Dani Or, Time domain reflectometrymeasurement principles and applications, Hydrol. Process. 16 (2002)141–153.

[7] Holger Worsching, Rolf Becker, Stefan Schlaeger, AndreasBieberstein, Peter Kudella, ‘‘Spatial-TDR moisture measurement ina large scale levee model made of loamy soil material’’, in: Proc. TDR2006, Purdue University, West Lafayette, USA, September 2006,Paper ID 33, 15 p., <www.engineering.purdue.edu/TDR/Papers>.

[8] TDR Soil Moisture Meters, Eijkelkamp Agrisearch Equipment,<http://www.surechem.com.my/download/eijkelkamp/P1/P1-64e.pdf>.

[9] Ole Greena, Esmaeil S. Nadimib, Victoria Blanes-Vidal, Rasmus N.Jørgensenb, Ida M.L. Drejer Stormc, Claus G. Sørensena, Monitoringand modeling temperature variations inside silage stacks using novelwireless sensor networks, Comput. Electron. Agric. 69 (2009) 149–157.

[10] <www.compoball.eu>.[11] F. Reverter, R. Pallàs-Areny, Direct sensor-to-microcontroller

interface circuits. Design and characterisation, Marcombo,Barcelona, 2005.

[12] O. Casas, F. Rillo, Method for reducing response time in sensormeasurement, Rev. Sci. Instrum. 80 (8) (2009). 085102–085102-5.

[13] A. Broveli, G. Cassiani, Combined estimation of effective electricalconductivity and permittivity for soil monitoring, Water Resour. Res.47 (8) (August 2011) W08510.

[14] G. Campbell, R. Anderson, Evaluation of simple transmission lineoscillators for soil moisture measurement, Comput. Electron. Agric.20 (1998) 31–44.

[15] R.N. Dean, A.K. Rane, M.E. Baginski, J. Richard, Z. Hartzog, D.J. Elton, Acapacitive fringing field sensor design for moisture measurementbased on printed circuit board technology, IEEE Trans. Instrum.Meas. 61 (4) (2012) 1105–1112.

[16] Hugo Marques, Beatriz Borges, Pedro Ramos, André Martins,Contactless battery charger for composite humidity andtemperature wireless sensors, in: IEEE EUROCON – InternationalConference on Computer as a Tool (EUROCON), Lisbon, Portugal,April, 2011.

[17] Hugo Marques, Beatriz Borges, ‘‘Contactless battery charger withhigh relative separation distance and improved efficiency’’, in: IEEE33rd International Telecommunications Energy Conference(INTELEC), Netherlands, Amsterdam, October, 2011.

[18] G. Lu, B. Krishnamachari, C.S. Raghavendra, ‘‘An adaptive energy-efficient and low-latency MAC for data gathering in wireless sensornetworks’’, in: 18th International Parallel and Distributed ProcessingSymposium, April 2004.