Simple and inexpensive microwave setup for industrial based applications: Quantification of flower honey adulteration as a case study

A simple and inexpensive microwave measurement setup based on measurements of magnitudes of transmission properties (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$|S_{21}|_{\text {dB}}$$\end{document}|S21|dB) is proposed for industrial-based microwave aquametry (moisture or water content) applications. An easy-to-apply calibration procedure based on normalization is implemented to eliminate systematic errors in the measurement system. As a case study, we applied this setup for the quantification of water-adulteration in flower honey. After validating this system by distilled water and pure flower honey measurements, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$|S_{21}|_{\text {dB}}$$\end{document}|S21|dB measurements of the pure flower honey with various adulteration percentages (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\delta$$\end{document}δ) up to 9% are conducted to examine the performance of the measurement setup for quantification of water adulteration. A multi-dimensional fitting procedure is implemented to predict \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\delta$$\end{document}δ using the proposed inexpensive microwave measurement setup. It is shown that it is possible to quantify an adulteration level with an accuracy better than \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mp } 1$$\end{document}∓1% by the proposed measurement setup and the applied multi-dimensional fitting procedure.


Methodology
Figure 1a illustrates a dielectric sample (honey samples in general possess dielectric property only) with a relative complex permittivity ε r and length L loaded into a rectangular metallic hollow waveguide operated in its dominant mode TE 10 .The wave-cascading matrix of the sample [M] in reference to empty (air) waveguide sections can be written as given in the study 49 where here Ŵ is the reflection coefficient at the air-sample interface, and T is the propagation factor within the sample.Z and Z 0 are the wave impedances of the sample-filled and air-filled waveguide sections.γ and γ 0 are the propaga- tion constants of the sample-filled and air-filled waveguide sections.µ 0 is the permeability of the free-space, ω is the angular frequency, f c and f are the cutoff and operating frequencies, and ε ′ r and ε ′′ r are the real and imaginary parts of ε r , which indicates the degree to which a material can be polarized and dissipation within a dielectric sample, respectively.
As shown in Fig. 1b, the first and third matrices in Eq. (1) denote impedance transitions while the middle one corresponds to the amplitude and/or phase change.The matrix [M] is a function of the measured S-parameters as implemented in the study 49 where S 11 , S 22 , S 21 , and S 12 are the forward and backward reflection and transmission S-parameters.
From Eqs. (1) and (5), it is possible to write S 21 as shown in the studies 14,36 Our goal is to measure |S 21 | and detect and quantify the water adulteration in honey samples by means of |S 21 |.

Measurement setup
A simple microwave measurement setup, as shown in Fig. 2, was constructed to measure |S 21 | of a waveguide section entirely loaded by flower honey (pure or adulterated).The setup was positioned vertically to conveniently perform |S 21 | measurements of the examined honey samples.The process of pouring or loading the sample into the measurement cell will be discussed in Subsections 3.1 and 3.2.This setup consists of simple microwave equipments including a microwave source, an adaptor, an attenuator, a waveguide measurement cell, a diode-detector, and an analog dB meter.The source from Flann Microwave Instruments (FMI) with model 449X operates at X-band (between 8.2 and 12.4 GHz) producing an output signal of approximately 10 mW.An adaptor from FMI with model 16093 operates as a guide to direct signals from the source to the attenuator and waveguide. (1) ( Compared with the setups used in earlier works 20,21,27,28,40,45,47 , the measurement setup in Fig. 2 can be considered to be appreciably less expensive since it does not require any expensive VNA instrument and even any directional couplers as shown in the study 14 .Furthermore, the proposed setup is as relatively inexpensive as the setups in the studies 14,[37][38][39] , which use a calibrated power sensor for transmission power detection in charge of detector and dB meter.
Because the setup is based on |S 21 | measurements, a suitable yet accurate calibration procedure should be implemented to eliminate some systematic errors in the measurement system.Since the measured signals are proportional to the square of the transmitted signals through the sample (i.e., the principle of square-law detection), and provided that the amplitude of the transmitted signal is above the noise level and is not too large, as detailed in the book 50 , we applied the calibration procedure in the study 36 where T s (ω) is the measured transmitted signal when the measurement cell is loaded with pure or adulterated flower honey, T f (ω) is the measured transmitted signal when the measurement cell is unloaded (no sample); and T m (ω) is the measured transmitted signal when a metal plate is located at the end of the measurement cell.It is noted that T m (ω) corresponds to noise signals when no energy is present and is included into Eq. ( 7 to improve the measurement accuracy.

Results and discussion
All the measurements of distilled water, pure flower honey, and water-adulterated flower honey in this section were implemented at ordinary laboratory conditions (temperature: 23∓1 • C and relative humidity: 50-60%).

Validation
Before carrying out water-adulteration analysis by the simple and cheaper microwave measurement setup in Fig. 2, we performed some preliminary measurements to validate the simple calibration process based on averaging in Eq. (7), and to examine the accuracy of the setup.For this goal, we measured |S 21 | (dB) (henceforth denoted as |S 21 | dB ) of distilled water ( L ∼ = 5.0 mm) poured in the opening of the cell with length L g = 10.16 mm.Here, a thin adhesive tape was applied to cover the entire opening at the bottom of the cell.In this way, there is no need to use a dielectric plug/window to hold the liquid sample in place in the cell as exercised in the study 51 .Figure 3 illustrates the measured |S 21 | dB of the distilled water (shown by a red square symbol) averaged from three independent measurements at some discrete frequencies at X-band.In order to compare the accuracy of measurements, calculated |S 21 | dB values of the distilled water were also calculated by using the Debye model with one-pole as implemented in the studies 14,27,38 where ε s and ε ∞ are, respectively, the relative permittivity at considerably small (theoretically zero) and consid- erably high (theoretically infinite) frequencies, and τ is the relaxation time (rearrangement time).For distilled (7)    2)-( 6) are in good agreement at the discrete frequencies, with differences not exceeding 2%.These results partly show that the measurement setup has sufficient accuracy to quantify the water-adulteration level within flower honey samples.
In order to examine whether the measured |S 21 | values at discrete frequencies are in good agreement with the curve of the Debye model over the entire frequency band, we applied an exponential curve fitting for the measured |S 21 | dB ( |S f 21 | dB ) using the following expression where a, b, and c are the curve fitting parameters, and f GHz is the frequency in GHz.Utilizing the measured |S 21 | dB at discrete frequencies in Fig. 3, the parameters a, b, c were determined as shown in Table 1.This table also presents the value of R 2 which means how well the measured data fit to the fitting expression in Eq. ( 9) ( R 2 = 1.0 means the best fitting).It is noted from Fig. 3 that the measured data fitted by these parameters follow the data of the Debye model very closely.

Sample preparation and environmental conditions
Pure flower honey blended from different origins of Anatolia regions in Türkiye, purchased from a local market, was used to examine the efficiency of the proposed method.Details about the physicochemical properties (including standard maximum or minimum limits) of the tested flower honey along with measurement techniques are presented in Table 2. Tested flower honey was stored within its original glass container throughout all tests to minimize the effect of storage conditions on microwave measurements.Honey samples were prepared at ordinary room conditions (a temperature of ∓23 • C and a relative humidity of approximately 55%).The water adulteration level was arranged on the mass-to-mass basis procedure by using precision scales (hodbehod SF-400C) with 0.01 g accuracy and 600 g maximum capacity.It is noted that distilled water and pure flower honey in proper amounts were mixed in a 50 mL glass beaker by a stirrer with a constant rotational speed of approximately 30 seconds to mitigate any formation of any air bubbles, especially for mixtures with higher water adulteration.Such a step is necessary because viscosity of distilled water is considerably lower than the viscosity of pure flower honey as discussed in the studies 20,21,27 .While adulteration levels of δ = 3 , 6, and 9 (each of which corresponds to the percentage of adulteration calculated by the mass-to-mass ratio of water to pure flower honey) were considered in the analysis of the fitting procedure, adulteration levels of δ = 4 , 5, 7, and 8 were utilized for testing the valid- ity of the fitted formula used for the adulteration quantification procedure, as discussed in Subsection 3.4.We restricted our analysis to the values of δ less than 10% because any value of δ over or around 10 not only produces a considerable change in viscosity of pure flower honey but also results in a substantial difference in its color and taste 53 .For each adulteration level, three sample sets were prepared to minimize any possible measurement errors (a total of 24 samples including the pure flower honey).The masses of each pure or water-adulterated flower honey sample used in each experiment were set at approximately 10g.We paid attention to the determination of this mass value by considering that it would be relatively much so that its value measured by the used scales was sufficiently accurate.Figure 4a shows water-adulterated (9%) flower honey within a beaker on the used scales.www.nature.com/scientificreports/A thin and widely used adhesive tape was applied to fill the entire opening at the bottom of the cell.Then the prepared mixtures of adulterated flower honey samples were carefully poured into the opening (hollow waveguide) of the measurement cell, as shown in Fig. 4b.Because of the relatively high viscosity of the flower honey samples, special care was exercised in cleaning the measurement cell with acetone after the mixture was poured back into the beaker for future use and after the adhesive tape was removed.Thereafter, the measurement cell was left at ordinary room conditions for about one minute in order for the acetone to evaporate.Such a two-step cleaning procedure was observed to be effective in reducing the effect of drift of the first experiment to the last one for repeated measurements as implemented in our previous study 27 .

Water-adulterated measurements
After validating our measurement apparatus using transmission |S 21 | dB measurements for the distilled water, we continued with |S 21 | dB measurements of pure flower honey with and without water-adulteration.Presented results here and in the following subsection were obtained by averaging |S 21 | dB from three sets for each δ level.Figure 5a illustrates the variation of |S 21 | dB at some discrete frequencies (from 8.5 GHz to 12 GHz with 0.5 GHz increments) for different adulteration levels of δ = 0 , 3, 6, and 9.As expected, it is seen from Fig. 5a that |S 21 | dB decreases with an increase in δ , since water has a loss factor considerably greater than that of pure flower honey tested in our study, as validated by earlier works 27,28 .| dB at some discrete frequencies (8.5:0.5:12.0GHz) for adulteration percentages of δ = 0 , 3, 6, and 9, (b) extracted ε ′ r and ε ′′ r of the pure flower honey by using the method discussed in the study 27 , and (c) measured |S 21 | dB (Meas) of the pure flower and fitted |S 21 | dB using a, b, and c values in Table 1 (Fitted).
for the Case-II is greater than that of the Case-I.This implies that the optimum fitting could be implemented by using larger number of samples.
It is instructive to examine whether a all , b all , c all , and k values for Case-II (Table 3) for δ = 0 produce results similar to those calculated by using a, b, and c values (Table 1).For this goal, we obtained |S 21 | dB values for the pure flower honey ( δ = 0 ) shown in Fig. 7.It is seen from Fig. 7 that |S 21 | dB values calculated by a all , b all , c all , and k values for the Case-II (Table 3) and by using a, b, and c values (Table 1) are in good agreement.
Finally, to evaluate the performance of the proposed relatively inexpensive microwave measurement setup and multi-dimensional fitting process through the expected functional behavior in Eq. ( 10), we performed |S 21 | dB measurements of the tested flower honey with δ = 5 and δ = 8 .Figure 8a,b demonstrate the measured |S 21 | dB of the tested flower honey with δ = 5 and δ = 8 because they were not taken into account in the curve fitting analysis.In addition, these figures show fitted |S 21 | dB values using a all , b all , and c all values of the Case-II in Table 3 (All-Fitted) for δ = 4 , 6, 7, and 9.It is noted from Fig. 8a,b that the measured |S 21 | dB with δ = 5 and δ = 8 are, respectively, lying within 4 < δ < 6 and 7 < δ < 9 adulteration ranges predicted by the a all , b all , and c all values.This indicates that our proposed measurement setup, along with the fitting model, can predict accurate δ values within ∓1 % δ range for adulteration levels up to δ = 10 .This means that by drawing a grid of spectral |S f 21 | dB dependencies, each of which corresponds to different levels from δ = 1 to δ = 10 with 1% increment using Eq. ( 10), it is possible to predict water adulteration level within ∓1 % δ range.Table 3. Curve fitting values a all , b all , c all , and k using the measured |S 21 | for different δ values.Case-I and Case-II refer, respectively, to the analysis performed for δ = 0 , 3, 6 and the analysis performed for δ = 0 , 3, 6, and 9.It should be emphasized here that in our analysis the fitting process implemented and the expression used in the fitting process may not be the best choice but were selected with the sole purpose of demonstrating the applicability of the proposed simple and inexpensive measurement setup.If needed, other fitting processes and fitting expressions can be used to improve the accuracy of quantification.Besides, it is noted that our proposed microwave measurement setup was validated, as a case study, by only one type of honey (flower honey).Various honey types different than flower honey such as highland and thyme honey are available, and yet physicochemical properties of even one certain type of honey can change due to its location and origin.Nonetheless, previous works in the literature 20,21,27,28,40,45,47,48 and sensitivity of microwave signals to a change in water content inside various samples (aquametry) could be partially and potentially considered as concrete bases for our microwave setup to be actively used for quantification of water-adulteration of various honey samples (detection of minimum δ may possibly change from honey type and origin).It is also important to discuss that different from or in addition to the validation of our measurement setup by distilled water measurements (before its application to the quantification of water adulteration of flower honey), the Karl Fischer titration method 54 could be potentially and equally used for evaluation of our measurement setup for quantification of water adulteration of flower honey.However, such a measurement technique requires expensive apparatus currently not available in our laboratory.On the other hand, the implemented multi-dimensional fitting process is limited to X-band measurements for the tested honey type only.If measurements were performed at another waveguide frequency band such as 3.95-5.85GHz (WR187) for the same honey type or if measurements were conducted at the same band for a different honey type, then the proposed multi-dimensional fitting process must be re-implemented, which is considered as a disadvantage of our proposed technique.Besides, it is instructive to make a comparison of our study with studies on microwave quantification/detection of honey adulteration in the literature.Table 4 illustrates such a comparison of our work with those studies in the literature 20,21,27,28,40,45,47,48 in terms of measurement type, overall cost, and analysis type.It is noted from Table 4 that while the method in the study 45 is a resonant method, methods in those studies 20,21,27,28,40,47,48 and our work are non-resonant methods.Besides, the studies 40,48 focus on detection of adulteration in honey only whereas the studies 20,21,27,28,45,47 and our work concentrate not only on detection but also on quantification of adulteration in honey.Finally, while the measurement setups in the studies 20,21,27,28,40,45,47,48 use expensive VNA instruments, the setup is in the present study is relatively inexpensive without requiring any VNA instrument.The disadvantage of our measurement setup, however, is that it can perform |S 21 | measurements at discrete frequencies only.

Cases
Finally, the proposed microwave sensor based on waveguide measurements, which is applied for quantification of water-adulteration in flower honey as a case study, can find applications for low-lost microwave aquametry applications of granular or liquid samples such as moisture detection of grains, soil samples, and food products 55

Conclusion
A simple and relatively inexpensive microwave measurement setup is introduced for industrial-based applications especially for microwave aquametry applications.The setup uses only |S 21 | dB measurements, which could be realized by a typical source, an adapter, an attenuator, a waveguide measurement cell, a diode-detector, and an analog dB meter.As a case study, this setup was tested for the quantification of water-adulteration of pure flower honey.Systematic errors in the measurement system were removed by an easy-to-apply calibration procedure based on normalization.In the quantification process, because the one-dimensional fitting procedure is limited for our analysis, a multi-dimensional fitting procedure based on the 'fit' function of MATLAB© along with the metric function in Eq. ( 10) involving an exponential decay with some offset is applied for evaluating the performance of the proposed measurement system.It is observed that the proposed measurement system and the implemented fitting procedure allow accurate quantification of water-adulteration of the tested pure flower honey up to ∓1 % within an adulteration limit of 10%.

Figure 1 .
Figure 1.(a) Configuration of sample-filled waveguide section and (b) representation of the sample-filled waveguide in terms of impedance transition and amplitude and/or phase change 49 .

Figure 2 .
Figure 2. A picture of the microwave measurement setup.

21 |Figure 3 .
Figure 3. (Measured (shown by a red square symbol), calculated (shown by a blue solid line-Debye model), and fitted (shown by a black dashed line) |S 21 | dB of the distilled water at X-band by the proposed setup.

Figure 4 .Figure 5 .
Figure 4. (a) Water-adulterated (9%) flower honey within a beaker positioned on the used scales and (b) sample pouring into the measurement cell.

Figure 7 .
Figure 7. Measured |S 21 | dB (Meas) of the pure flower, fitted |S 21 | dB using a, b, and c values inTable 1 (Fitted), and fitted |S 21 | dB using a all , b all , and c all values of the Case-II in Table3(All-Fitted). .

Table 2 .
Physicochemical properties (including standard maximum or minimum limits) of the tested flower honey along with measurement techniques.

Table 4 .
Comparison of our study with studies on microwave quantification/detection of honey adulteration in the literature.