A handheld moisture content sensor using coupled-dipole antennas

J. Mearnchu^†, T. Limpiti^†, D. Torrungrueng^‡, P. Akkaraekthalin* and M. Krairiksh^†

^† Faculty of Engineering, King Mongkut's Institute of Technology Ladkrabang, Bangkok 10520, Thailand.
s6060210@kmitl.ac.th, s8060939@kmitl.ac.th, kkmonai@kmitl.ac.th
^‡ Department of Electrical and Electronics Engineering, Faculty of Engineering and Technology, Asian University, Chon Buri 20150, Thailand. dtg@asianust.ac.th
* Faculty of Engineering, King Mongkut's University of Technology North Bangkok, Bangkok 10800, Thailand.
prayoot@kmutnb.ac.th

Abstract - This paper presents an analysis and design of a handheld moisture content sensor. A dielectric property determination technique was used to measure the magnitudes of the reflection and the coupling coefficients of coupled-dipole antennas. These coefficients were plotted and their intersection determined to obtain the values of and ( and , respectively, are the real and imaginary parts of the relative complex permittivity of the dielectric of interest). Moisture content measurements for paddy at various moisture content levels are shown. Comparison of measurements made by this technique with those made by conventional transmission measurement technique yielded a compensation scheme for error reduction. This sensor is useful for controlling the quality of paddy dried in a continuous microwave drying system process.

Keywords - Moisture Content Of Paddy. Magnitude Measurement. Coupled-Dipole Antennas. Moisture Content Sensor

I. INTRODUCTION

Granular materials can be considered the fourth state of matter because of their abundance both in nature and in several industries (Jaeger and Nagel, 1996). Dielectric characterization of a given material reveals its unique electric signature (Venkatesh and Raghavan, 2004; Seker and Abatary, 2005). At high frequencies, measurement of the dielectric properties of materials provides a basis for developing method of indirect determination of their physical properties, particularly for materials containing water. For decades, moisture content determination of granular and other materials from measurements of their dielectric properties at microwave frequencies have been the focus of several studies (Meyer and Schilz, 1981; Menke and Knochel, 1996). The intrinsic nature of the dielectric properties and their high correlation with water content at these frequencies make them useful for developing indirect sensing methods for nondestructive determination of the water content as well as other physical properties of materials (Hasted, 1973; Nyfors and Vainikainen, 1989; Kraszewski, 1996).

The dielectric properties must be accurately measured, and explicit relationships between the dielectric properties and any given internal property must be identified. These can be done theoretically or empirically. However, although water governs the dielectric response of substances containing it, two major limitations in defining a unique relationship between the dielectric properties and the moisture content have influenced the use of these techniques. The first relates to the bulk density changes for granular material. It produces a similar effect to that of the moisture content on the relative complex permittivity (Hippel, 1954). The second relates to the need for an individual moisture calibration equation for each material. Both of these limitations complicate the design, cost, and maintenance of a microwave moisture sensor based on the principle of permittivity measurement.

In recent years, a permittivity-based method for moisture determination in granular materials which is independent of bulk density changes and of the kind of material was proposed and developed (Trabelsi et al., 1998, 1999, 2001, 2006). This method allows moisture determination of grains of different structures and compositions from a single calibration equation, hence alleviating the two mentioned limitations (Nelson, 1991, 2006).

As Kandala et al. (2007) demonstrated earlier, the method of measuring the capacitance and phase angle of a parallel-plate system provides accurate moisture content measurements because of its closed boundary. However, the need to fill the sample up to the whole volume limits the capability of on-line monitoring. One of the measurement techniques that determines both real () and imaginary () parts of the relative complex permittivity of the dielectric of interest is the free space technique that measures reflected and transmitted waves (Ma and Okamura, 1999). The advantage of this technique is that the properties of the object under test along the propagation path can be readily obtained. It is used in dynamic measurement such as cement-based variation during the days (Kharkovsky et al., 2002). The dielectric property may be determined by obtaining the values of and from an intersection of the plot of measured transmission and reflection coefficients. Although free space technique has the advantage of non-contact measurement of objects lying on a conveyer belt, its sample preparation is very complicated. A sample must be large in size in order to prevent multiple reflection and edge effects and to comply with plane wave condition; therefore, it needs sufficiently large space and is not suitable for measurement of grains in a small container. Contact-measurement is preferable in this case.

A method to determine the absolute value of impedances of two dipole antennas has also been developed in order to determine the relative complex permittivity of an object. Dipole antennas are laid over a homogeneous media (Antonyk et al., 2004) or a heterogeneous media (Shostak et al., 2002) to measure its dielectric property by ground penetration radar (GPR). This technique measures the values of the amplitude of the antenna impedances. A calibration curve is constructed from the ratio of the impedances of antenna above the sand and those above the metal plate. Although Shostak et al. (2002) provided data of complex permittivity, Antonyk et al. (2004) provided only the real part of the complex permittivity. However, for measuring the absolute value of the impedances, an expensive vector network analyzer is necessary.

Recently, the author has developed a continuous microwave heating system for paddy drying (Sangdao et al., 2006). Accurate prediction of water diffusivity can be accomplished by modeling volume changes in paddy (Aguerre et al., 2008). It was found that information about the moisture content of the paddy is needed in order to monitor its quality. Therefore, this paper is about a handheld sensor for determining the percentage of moisture content via and . The contribution of this work to the research community is that we have used a sensor with two dipole antennas which is mounted along the interface of air and paddy. By measuring only the |S₁₁| and |S₂₁| and plotting them on a surface curve, and can be readily obtained. A simple calculation scheme is introduced so that lengthy computation is not necessary.

II. METHODS

A. Principle of a coupled-dipole sensor

Two half-wave dipoles of length L lie in parallel with separation d located at height h above a dielectric half space (,) as depicted in Fig. 1. Antenna #1 is excited by a transmitter via a dual directional coupler so that measurements of the incident power (P_i) and the reflected power (P_r) can be made. The value of or |S₁₁| represents the magnitude of the reflection coefficient. Antenna #2 couples the wave transmitted from antenna #1 and scattered from the dielectric half space. A power detector for measuring the coupled power (P_c) is installed at the input of antenna #2. The magnitude of the coupling coefficient is denoted by or |S₂₁|. Electromagnetic fields in this problem can be calculated by using either the classical electromagnetic theory for the half-space problem (Kong, 1990) or by using an electromagnetic simulator. The magnitudes of the reflection and coupling coefficients are calculated as |S₁₁| and |S₂₁| at different values of and . Then, |S₁₁| and |S₂₁| are plotted in the plane. The projection from the cross point of |S₁₁| and |S₂₁| provides the values of and on the horizontal and vertical axes, respectively.

Fig. 1. Diagram of a coupled-dipole sensor.

According to the data in (Nelson et al., 2001), paddy with 10 % to 20 % moisture content had the values of its relative complex permittivity in the ranges of 2.0 ≤ ≤ 4.0 and 0.2 ≤ ≤ 0.6. A number of surface curves of versus were constructed in terms of |S₁₁| and |S₂₁| for determining and at different dipole lengths. Note that h and d were equal to 0.00 and 0.76 cm, respectively. Three values of L of a half wavelength were utilized: one in free space (= 1, = 0.0), another in paddy (= 3, = 0.4) and another one in the composite medium of an infinite half space of free space and paddy (= 2, = 0.2). It was found that the corresponding |S₁₁| were 0.608, 0.602 and 0.592, respectively, and the corresponding |S₂₁| were 0.351, 0.452 and 0.553, respectively. Note that this composite medium yielded lower |S₁₁| and higher |S₂₁| compared to those of free space and paddy. This is a desirable property of this coupled-dipole sensor. However, when antenna parameters were chosen based on this composite medium, optimum sensor performance was not always achieved.

The lowest |S₁₁| and the highest |S₂₁| (the sensor can transfer power from the transmitter to the dielectric object under test efficiently) took place when the dipole length L was equal to the half wavelength in the composite medium of free space and paddy, which was equal to 4.33 cm.

Note that h and L were equal to 0.00 and 4.33 cm, respectively. In addition, d varied as shown in Fig. 2. Fig. 2a) depicts a surface curve when d was equal to λ/16 for 0.30 ≤ |S₁₁| ≤ 0.66 and 0.43 ≤ |S₂₁| ≤ 0.61. The values of |S₁₁| and |S₂₁| when d was equal to λ/8 and 3λ/16 are illustrated in Fig. 2b) and 2c), respectively. The |S₁₁| and |S₂₁| in Fig. 2b) were in the ranges of 0.32 ≤ |S₁₁| ≤ 0.72 and 0.31 ≤ |S₂₁| ≤ 0.55, and those in Fig. 2c) were in the ranges of 0.34 ≤ |S₁₁| ≤ 0.72 and 0.21 ≤ |S₂₁| ≤ 0.46. It should be noted that both |S₁₁| and |S₂₁| should not be too low so that less sensitive detectors could be employed. In this context, Fig. 2b) with d = λ/8 was a trade-off between |S₁₁| and |S₂₁|. Hence, d equals λ/8 was used, which was equal to 0.76 cm.

Fig. 2. Surface curves of two coupled dipoles at the frequency of 2.45 GHz. a) d = λ/16, b) d = λ/8, c) d = 3λ/16.

B. Application in paddy moisture content determination

A coupled-dipole sensor design based on the previous section was fabricated on an FR-4 PCB substrate with a dielectric constant of 4.4. Each dipole was fed by a coaxial cable having a characteristic impedance of 50Ω. They were connected to port 1 and port 2 of an HP8530A

HP8530A network analyzer which measured |S₁₁| and |S₂₁| when the sensor was laid on the paddy surface contained in a 40 cm × 65 cm × 30 cm plastic container. The skin depth of paddy at 2.45 GHz was equal to 6.25 cm, which was sufficiently deep for measurements to be performed in the above container. The five levels of paddy moisture content, ranging from 10 % to 20 %, were prepared. Measurements were made at 15 different positions of each sample of various moisture content levels at 27°C temperature. For better accuracy, each position was measured three times and the results averaged. The system was calibrated by placing the sensor in air (surrounded by RF absorbers) and measured |S₁₁| and |S₂₁| of free space. Then, it was compared with the simulation results by IE3D (Zeland Software, Inc.). The simulated |S₁₁| and |S₂₁| were 0.601 and 0.359, while the corresponding values from measurements were 0.611 and 0.369, respectively. Therefore, the calibration factors of |S₁₁| and |S₂₁| were 0.984 and 0.972, respectively. These factors were multiplied to the measured values of |S₁₁| and |S₂₁| before they were plotted on the surface curve in Fig. 2b). The paddies were stored for bulk density and moisture content measurements by the standard ASAE (1998) method. Since this method was devised for wheat measurement, modification for paddy measurement was required as follows: the paddy was measured for 10 g by an electronic balance with an accuracy of 0.0001 g and was filled into 30 heat resisted glass covered plates. These samples were heated in an oven at 130°C for 19 hours. Then, they were cooled in a desiccator and then their weights were measured and weight losses determined. The procedure was repeated after 6 hours heating and was stopped when weight change was less than 0.005 g.

Figure 3 clearly shows that the relative complex permittivity of paddy increased as the moisture content increased. The moisture content was determined by the mehod of Nelson et al. (2001). The relative complex permittivity normalized by the bulk density could provide a clear relationship between  and  as shown in Fig. 4. The slope a_f and the constant k were determined from the graph in Fig. 4, which were equal to 0.2428 and 6.5474, respectively. The bulk density was calculated, but it is not shown in this paper.

Fig. 3. Argand diagram of the effective relative complex permittivity of paddy.

Fig. 4. Argand diagram of the normalized relative complex permittivity of paddy.

The loss tangent normalized by the bulk density, which is a density-independent function of moisture content when the product ka_f is constant, is defined as a predicted moisture content, ξ, mathematically given by

(1)

The  increases linearly with the percentage of moisture content as shown in Fig. 5. At each moisture content value, the data points corresponding to different Bulk densities were nearly congruent and lay along a straight line. A linear regression was used to fit the data; i.e.,

(2)

Fig. 5. Variation of  with the moisture content of paddy.

The slope a, the intercept b and the regression coefficient R² given in Fig. 5 were 0.0110, 0.4273 and 0.9553, respectively. The calibration equation of the percentage of moisture content was

(3)

The plot of the moisture content calculated from (3) versus the moisture content determined by the standard oven-drying method is shown in Fig. 6. The dash line corresponds to the ideal relation. The accuracy of the proposed sensor was validated by comparing its measurements with those in Nelson et al. (2001) as shown in Fig. 7. The results of  from the proposed sensor at the moisture content of 11 % were almost identical to those in Nelson et al. (2001) except at higher moisture content level. The values of  from Nelson et al. (2001) were higher than those of the proposed sensor at all moisture content levels. These differences was due mainly to the fact that the transmission measurement in Nelson et al. (2001) was obtained from waves propagating through the sample volume, while the proposed sensor dealt only with the region to which the sensor was attached. The non-uniformity of the moisture content throughout the medium was the main reason for the differences between these two sets of data. The deviations of  from the reference values in Nelson et al. (2001) ranged from 1.2 to 9.6 %, while those of  ranged from 13.7 to 23.5 %.

Fig. 6. Calculated versus measured moisture contents.

Fig. 7. Comparison of results in (Nelson et al., 2001) and of the coupled-dipole sensor.

III. ERROR COMPENSATION

The method for moisture content determination in paddy discussed in the previous section is based on measurements of reflected and coupled powers from the paddy. Bulk density and moisture content affect the relative complex permittivity in a similar manner. The accuracy of moisture content values can be predicted based on the accuracy of the computed values of  and  which are strongly dependent on the accuracy of the measured |S₁₁| and |S₂₁|. The discrepancy of the results comes mainly from the differences in the conditions between the simulation and the experiment. The simulated results are associated with the infinite half space, while the measured ones are related to a material of finite size. The slight variation of dipole length also affects the accuracy of the measured |S₁₁| and |S₂₁|. The disagreement between the results from this paper and those from Nelson et al. (2001) comes mainly from the different measurement techniques that were used; i.e., coupling measurement and transmission measurement, respectively. The differences of samples in experiments were also a source of error.

It can be seen in Fig. 7 that the versus moisture content curve has an error of less than 5.9% compared to the data from Nelson et al., (2001). However, the error is equal to 9.6 % when the moisture content is equal to 22 %. In addition, the error of  in Fig. 7 is higher than that of . The main reason for these errors was that simulations were performed by using an infinite half-space object while measurements were made on a finite size object.

To compensate for the error of the measured results, the data in Nelson et al. (2001) was used as reference data. The errors between the measured  and  ( and ) and those from Nelson et al. (2001) are plotted in Fig. 8. The curve fitting representation of  is

(4)

It has R² equals 0.8830.

Fig. 8. Differences of measured  and  from Nelson et al. (2001).

The representation of  is

(5)

where R² is 0.9807. Once  and  were measured, the errors of  and  were added to them to obtain the corrected values.

The measured |S₁₁| and |S₂₁| at different moisture content, in Fig.7, implied different values of  and . Figure 9 shows the relationship of  and  to |S₁₁| and |S₂₁|. It is evident that their relationship was almost linear.  and  may be represented by

	(6)
	(7)

where k₁ = c₁ + c₃ and k₂ = c₂ + c₄ are constants. By using the following relations,

	(8)
	(9)

the values of constants were found to be

Fig. 9. Relationships of  and  to |S₁₁| and |S₂₁|.

From the measured |S₁₁| and |S₂₁|,  and  were found from (6) and (7). The corrected  and  were obtained by adding to them  and in (4) and (5), respectively. Then, moisture content was calculated from (1) and (3). It is obvious that the moisture content may be obtained by a simple calculation scheme. This scheme enables a simple design for the handheld sensor.

In order to be able to use the sensor in practical situations, the measurement system must be simple and cost effective while providing sufficiently accurate readings. This section reports a low cost, yet very useful, measurement system that was experimentally tested and validated. The block diagram of the proposed system composing of a transmitter and a receiver is shown in Fig. 10. The transmitter module consists of a MAX2753 VCO and a LMX2347 PLL for generating the frequency of 2.45 GHz. The -8 dBm output power is amplified by a 13.5 dB gain MAX2641 low noise amplifier. The incident and reflected powers are sensed, through a dual directional coupler with 21 dB coupling factor, by MAX4003 RF power detectors #1 and #2, respectively. The transmitted signal is radiated by a transmitting dipole and coupled to the paddy to a receiving dipole. In the receiver, the received signal is amplified by a 13.5 dB gain of a MAX2641 low noise amplifier. The coupled power is measured by a MAX4003 RF power detector #3. The RF powers are converted to DC voltage by these detectors. The output voltages from the power detectors are converted to digital signals by a MAX1272 analog to digital converter. They are processed by the PIC16F876 8 bits microcontroller. The output values of the reflected and coupled voltages are displayed on an LCD.

Fig. 10. Diagram of the proposed sensor.

From Fig. 2b), the lowest value of |S₂₁| was 0.31(-5.1 dB), and the minimum received power was obtained by multiplying this |S₂₁| with the transmitted power. The output power from the amplified PLL signal of -8 dBm through the 21 dB coupling factor of the dual directional coupler was -24.1 dBm, which was within the operating range (-45 dBm) of the MAX4003 power detector. The coupled-dipole sensor was connected to the measurement system to measure coupled and reflected voltages, which were related to coupled and reflected powers, respectively.

IV. SYSTEM DESIGN AND EVALUATION

The sensor was attached to the paddy surface contained in a plastic container (see Fig. 11). The system was calibrated by placing the sensor up in the air (surrounded by RF absorbers) to measure |S₁₁| and |S₂₁| of the free space. Then, the measured values were compared to the values from the simulation by using IE3D. The simulated |S₁₁| and |S₂₁| were 0.601 and 0.359, while the corresponding values from measurements were 0.991 and 0.088, respectively. Therefore, the calibration factors of |S₁₁| and |S₂₁| were 0.606 and 4.079, respectively. These factors were multiplied to the measured values of |S₁₁| and |S₂₁| before they were plotted on the surface curve in Fig. 12. The percentage of moisture content was calculated from the obtained relative complex permittivity according to the moisture content determination procedure in Nelson et al. (2001) and the compensation scheme in the previous section. The moisture contents of paddy were measured at 15 different positions for each sample of various moisture content levels by a network analyzer and by the proposed sensor, as shown in Table 1. It is relevant that, by using the standard oven-drying method as a reference method, the proposed sensor provides accurate results with slightly higher errors than those from the network analyzer ones.

Fig. 11. Photograph of the sensor.

Fig. 12. |S₁₁| and |S₂₁| from the proposed sensor and the network analyzer on the surface curve.

Table. 1, Comparison of measured moisture content between network analyzer and the proposed sensor (referenced to standard oven-drying method).

V. CONCLUSIONS

Due to the need of a sensor for determining moisture content in a continuous paddy drying process, a coupled-dipole sensor is proposed. It has a feature of compact design of which reflection and coupling coefficients are measured and plotted on a surface curve of |S₁₁| and |S₂₁| in the  plane. The sensor is designed and tested to operate at the frequency of 2.45 GHz. Evaluation of moisture content determination was carried out and compared to the data in the literature. Accurate  was obtained when the moisture content was less than 19 %. However,  had a higher error than that of  due to the different measurement techniques (transmission and coupling measurements). Furthermore, the design curve was plotted from the calculation based on an infinite half space, while the measurements were performed using a finite size object. This work proposed a compensation scheme to reduce error. A prototype sensor was designed and tested. Accurate moisture content readings can be made with this simple system that fulfills the need of quality control of a continuous paddy drying system.

ACKNOWLEDGMENT
The authors appreciate Dr. P. Sirisomboon, Department of Agricultural Engineering, King Mongkut's Institute Technology Ladkrabang, Thailand, for her assistance in moisture content measurement with the standard oven technique.
This work was supported by the Thailand Research Fund through the Royal Golden Jubilee Ph.D. Program (Grant No. PHD/0027/2547) and the Senior Research Scholar Program (Grant No. RTA - 5180002).

REFERENCES
1. Aguerre, R.J., M. Tolaba and C. Suarez, "Modeling volume changes in food drying and hydration," L.A.A.R.; 38, 345-349 (2008).         [ Links ]
2. Antonyk, N.A., I.S. Kruglov, V.V. Zagoskin, L.P. Ligthart, A.S. Shostak, V.N. Iljushenko and B.I. Avdochenko, "Measurement of the complex dielectric permittivity of homogeneous media with dipole antenna sensors," Proc. European Radar Conference, Amsterdam, 169-172 (2004).         [ Links ]
3. ASAE, "Moisture measurement-unground grain and seeds," 45^th Edition: American Society of Agricultural Engineering, 551 (1998).         [ Links ]
4. Hasted, J.B., Aqueous Dielectrics, London, U.K.: Chapman & Hall (1973).         [ Links ]
5. Hippel, A.R.V., Dielectric Materials and Applications, New York: Wiley (1954).         [ Links ]
6. Jaeger, H.M. and S.R. Nagel, "Granular solids, liquids and gases," Rev Mod Phys; 68, 1259-1273 (1996).         [ Links ]
7. Kandala, C.V.K., C.L. Butts and S.O. Nelson, "Capacitance sensor for nondestructive measurement of moisture content in nuts and grain," IEEE Trans Instrum Meas, 56, 1809-1813 (2007).         [ Links ]
8. Kharkovsky, S.N., M.F. Akay, U.C. Hasar and C.D. Atis, "Measurement and monitoring of microwave reflection and transmission properties of cement-based specimens," IEEE Trans Instrum Meas; 51, 1210-1218 (2002).         [ Links ]
9. Kong, J.A., Electromagnetic Wave Theory, Wiley Interscience (1990).         [ Links ]
10. Kraszewski, A.W., Microwave Aquametry, New York: IEEE Press (1996).         [ Links ]
11. Ma, Z. and S. Okamura, "Permittivity determination using amplitudes of transmission and reflection coefficients at microwave frequency," IEEE Microwave Theory Tech, 47, 546-550 (1999).         [ Links ]
12. Menke, F. and R. Knochel, "New density-independent moisture measurement methods using frequency-swept microwave transmission," Proc IEEE MTT-S Int Microw Symp Dig; 3, 1415-1418 (1996).         [ Links ]
13. Meyer, W. and W.M. Schilz, "Feasibility study of density-independent moisture measurement with microwaves," IEEE Trans Microw Theory Tech; MTT-29, 732-739 (1981).         [ Links ]
14. Nelson, S.O., "Dielectric properties of agricultural products-measurements and applications," IEEE Trans Elect Insul, 26, 845-869 (1991).         [ Links ]
15. Nelson, S.O., S. Trabelsi and A.W. Kraszewski, "RF sensing of grain and seed moisture content," IEEE Sensors, 1, 119-126 (2001).         [ Links ]
16. Nelson, S.O., "Agricultural applications of dielectric measurements," IEEE Trans Elect Insul, 13, 688-702 (2006).         [ Links ]
17. Nyfors, E. and P. Vainikainen, Industrial Microwave Sensor, Boston, MA: Artech House (1989).         [ Links ]
18. Sangdao, C., A. Roeksabutr and M. Krairiksh, "An applicator for uniform heating using perpendicular slots on a concentric cylindrical cavity excited by perpendicular waveguides," International Journal of Electronics, 93, 313-334 (2006).         [ Links ]
19. Seker, S. and H. Abatary, "New frequency-dependent parametric modeling of dielectric materials," Int J Electron Commun (AEU), Elsevier, 320-327 (2005).         [ Links ]
20. Shostak, A., V. Zagoskin, L. Ligthart and V. Iljushenko, "Analysis of a linear dipole antenna system for measurement the complex dielectric permittivity," Proc. the 14^th International Conference on Microwave, Radar and Wireless Communications, 294-298 (2002).         [ Links ]
21. Trabelsi, S., A.W. Kraszewski and S.O. Nelson, "New density independent calibration function for microwave sensing of moisture content in particulate materials," IEEE Trans Instrum Meas; 47, 613-622 (1998).         [ Links ]
22. Trabelsi, S., A.W. Kraszewski and S.O. Nelson, "Determining physical properties of grain by microwave permittivity measurement," Trans ASAE; 42, 531-536 (1999).         [ Links ]
23. Trabelsi, S., A.W.Kraszewski and S.O.Nelson, "New calibration technique for microwave moisture sensors," IEEE Trans Instrum Meas; 50, 877-881 (2001).         [ Links ]
24. Trabelsi, S., S.O. Nelson and O. Ramahi, "A low-cost microwave moisture sensor," Proc. the 36^th European Microwave Association, 447-450 (2006).         [ Links ]
25. Venkatesh, M.S. and G.S.V. Raghavan, "An overview of microwave processing and dielectric properties of agri-food materials," Biosystems Engineering, Postharvest Technology, Elsevier, 1-18 (2004).         [ Links ]

Received: October 12, 2008
Accepted: July 1, 2009
Recommended by Subject Editor: Jorge Solsona