A Novel Printable Tag of M-Shaped Strips for Chipless Radio-Frequency Identiﬁcation in IoT Applications

: There is a growing interest in chipless radio-frequency identiﬁcation (RFID) technology for a number of Internet of things (IoT) applications. This is due to its advantages of being of low-cost, low-power, and fully printable. In addition, it enjoys ease of implementation. In this paper, we present a novel, compact, chipless radio-frequency identiﬁcation (RFID) tag that can be read with either vertical or horizontal polarization within its frequency bandwidth. This increases the sturdiness and detection ability of the RFID system. In addition, the di ﬀ erence between the vertical and horizontal responses can be used for tag identiﬁcation. The proposed tag uses strip length variations to double the coding capacity and thereby reduce the overall size by almost 50%. It has a coding capacity of 20 bits in the operating bandwidth 3 GHz–7.5 GHz, and its spatial density is approximately 11 bits / cm 2 . The proposed tag has a 4.44 bits / GHz spectral capacity, 2.44 bits / cm 2 / GHz encoding capacity, a spatial density at the center frequency of 358.33 bits / λ 2 , and an encoding capacity at the center frequency of 79.63 bits / λ 2 / GHz. A prototype is fabricated and experimentally tested at a distance of 10 cm from the RFID reader system. Then, we compare the measured results with the simulations. The simulated results are in reasonable agreement with the simulated ones. bistatic system as it requires only one antenna and a less-complex setup. Figure shows the setup for the proposed tag intended for a monostatic RFID system.


Introduction
The Internet of things (IoT) allows objects and people to be connected at any place and time through the Internet. However, this technology needs some important components to enable communication among these objects. Radio-frequency identification (RFID) is a wireless technique that can be used for sensing, identification, and tracking. Therefore, RFID tags and sensors have great potential for use in IoT applications such as smart cites and industrial and tracking systems [1][2][3][4]. The RFID system utilizes electromagnetic (EM) waves instead of the optical waves which are used with quick response (QR) codes and barcodes. Furthermore, many materials are penetrated by the EM waves, and thus the RFID tags could be read even when they are isolated by other low-loss materials [5,6]. Conventional RFID tags utilize integrated circuit (IC) chips that limit their applicability due to their prohibitive costs. Chipless RFID tags have been recently proposed to reduce the cost of the tags by removing the expensive IC chips, thus producing fully printable tags [7,8]. Chipless RFID tags generally consist of the following: (1) the chipless RFID tag, which contains a multi-resonance structure for object identification, and (2) the RFID reader that is used to interrogate the tag and read the retransmitted or backscattered response from a tag [9][10][11][12].
In the literature, different approaches have been proposed for encoding RFID tags with data such as time [13,14], frequency [12,[15][16][17], phase [18], image [19], and hybrid domains [20,21]. In frequency domain-based tags, the RFID reader uses radio signals to interrogate the tag, which then retransmits or backscatters the frequency signature back to the reader. These tags exist in either retransmission mode or backscattered mode varieties. Retransmission-based tags consist of multiple resonant structures and one or more ultra-wideband antennas for transmission and reception of the data [22][23][24]. On the other hand, in backscattering mode, no separate antennas are required [25][26][27].
To date, many different RFID structures have been demonstrated, including dual-polarized I-slots [10,11], U-shaped strips [9], L-shaped strips [28], square loops [29], triangular loops [30], circular loops [31], open loops [32], and inverted M-shapes [33]. High-impedance surface-based square loop resonators were proposed by Filippo Costa et al. [34]. The authors used a frequency selective surface (FSS) with an overall size of 4.5 cm 2 × 4.5 cm 2 and a coding capacity of 5 bits. Another work that used a quarter wavelength slot resonator was reported in [31] to encode 24 bits within the size of 2.4 cm 2 × 2.4 cm 2 . In [35], a 16-bit dual polarized chipless RFID tag was presented. Two rectangular patches were placed in horizontal polarization, and two were placed in vertical polarization to double the number of bits within the same frequency bandwidth.
There are two types of frequency-selective surface-based chipless RFID system configurations: monostatic and bistatic systems. Bistatic systems use two applicators: one for interrogating and the other for reading the RF signal. In contrast, monostatic systems use only one applicator for both reading and interrogating. This type of system is generally considerably less expensive than a bistatic system as it requires only one antenna and a less-complex setup. Figure 1 shows the setup for the proposed tag intended for a monostatic RFID system. Electronics 2020, 9, x FOR PEER REVIEW 2 of 16 multi-resonance structure for object identification, and (2) the RFID reader that is used to interrogate the tag and read the retransmitted or backscattered response from a tag [9][10][11][12].
In the literature, different approaches have been proposed for encoding RFID tags with data such as time [13,14], frequency [12,[15][16][17], phase [18], image [19], and hybrid domains [20,21]. In frequency domain-based tags, the RFID reader uses radio signals to interrogate the tag, which then retransmits or backscatters the frequency signature back to the reader. These tags exist in either retransmission mode or backscattered mode varieties. Retransmission-based tags consist of multiple resonant structures and one or more ultra-wideband antennas for transmission and reception of the data [22][23][24]. On the other hand, in backscattering mode, no separate antennas are required [25][26][27].
To date, many different RFID structures have been demonstrated, including dual-polarized Islots [10,11], U-shaped strips [9], L-shaped strips [28], square loops [29], triangular loops [30], circular loops [31], open loops [32], and inverted M-shapes [33]. High-impedance surface-based square loop resonators were proposed by Filippo Costa et al. [34]. The authors used a frequency selective surface (FSS) with an overall size of 4.5 cm 2 × 4.5 cm 2 and a coding capacity of 5 bits. Another work that used a quarter wavelength slot resonator was reported in [31] to encode 24 bits within the size of 2.4 cm 2 × 2.4 cm 2 . In [35], a 16-bit dual polarized chipless RFID tag was presented. Two rectangular patches were placed in horizontal polarization, and two were placed in vertical polarization to double the number of bits within the same frequency bandwidth.
There are two types of frequency-selective surface-based chipless RFID system configurations: monostatic and bistatic systems. Bistatic systems use two applicators: one for interrogating and the other for reading the RF signal. In contrast, monostatic systems use only one applicator for both reading and interrogating. This type of system is generally considerably less expensive than a bistatic system as it requires only one antenna and a less-complex setup. Figure 1 shows the setup for the proposed tag intended for a monostatic RFID system. The main obstacle in deploying RFID systems remains the high cost of the tags. The cost has to be extremely low to justify mass production. Therefore, the present study aims to develop low-cost, chipless RFID tags by (1) reducing the overall size and the bandwidth-per-bit of chipless RFID tags and (2) increasing the spectral capacity (bits/GHz), spatial density (bits/cm 2 ), and spatial density at the center frequency (bits/λ 2 ).
In this paper, we present a novel chipless RFID tag consisting of overturned M-shaped strips. This compact tag uses dual polarizations to make the reading procedure more robust and employs length variation encoding to double the coding capacity from 10 to 20 bits. We have simulated the proposed tag as an infinite structure, using a 3D full-wave electromagnetic (EM) simulator. The The main obstacle in deploying RFID systems remains the high cost of the tags. The cost has to be extremely low to justify mass production. Therefore, the present study aims to develop low-cost, chipless RFID tags by (1) reducing the overall size and the bandwidth-per-bit of chipless RFID tags and (2) increasing the spectral capacity (bits/GHz), spatial density (bits/cm 2 ), and spatial density at the center frequency (bits/λ 2 ).
In this paper, we present a novel chipless RFID tag consisting of overturned M-shaped strips. This compact tag uses dual polarizations to make the reading procedure more robust and employs length variation encoding to double the coding capacity from 10 to 20 bits. We have simulated the proposed tag as an infinite structure, using a 3D full-wave electromagnetic (EM) simulator. The proposed tag has Electronics 2020, 9, 2116 3 of 16 a 4.44 bits/GHz spectral capacity, 2.44 bits/cm 2 /GHz encoding capacity, a spatial density at the center frequency of 358.33 bits/λ 2 , and an encoding capacity at the center frequency of 79.63 bits/λ 2 /GHz. Note that processing multiple RFID tags may cause collision. In such cases, tag-to-tag anti-collision can be implemented by one of the earlier presented techniques in [36][37][38][39][40][41][42][43].
The remainder of this paper is structured as follows. In Section 2, the tag design methodology is presented. Section 3 involves the simulation results for some configurations. In Section 4, one configuration of the proposed tag is experimentally validated. Finally, some concluding remarks are presented in Section 5.

Tag Design
The geometry of the proposed tag is shown in Figure 2. It consisted of 10 inverted M-shaped strips implemented on a Rogers RT5880 substrate, with a thickness of 0.256 mm and a dielectric constant ε r = 2.2. In Figure 2, w and l are the width and length of the proposed tag, respectively, s x is the horizontal separation between two vertical elements, which was fixed for all strip resonators, s y is the vertical separation between two slanted elements, which was different from one strip to another, and w r is the width of each strip. Each inverted M-shaped strip structure could be utilized in either an active or a passive mode of operation. The existence of a null in the frequency response denoted the active mode (logic 1), while the nonexistence of the null denoted the passive mode (logic 0). proposed tag has a 4.44 bits/GHz spectral capacity, 2.44 bits/cm 2 /GHz encoding capacity, a spatial density at the center frequency of 358.33 bits/λ 2 , and an encoding capacity at the center frequency of 79.63 bits/λ 2 /GHz. Note that processing multiple RFID tags may cause collision. In such cases, tag-totag anti-collision can be implemented by one of the earlier presented techniques in [36][37][38][39][40][41][42][43]. The remainder of this paper is structured as follows. In Section 2, the tag design methodology is presented. Section 3 involves the simulation results for some configurations. In Section 4, one configuration of the proposed tag is experimentally validated. Finally, some concluding remarks are presented in Section 5.

Tag Design
The geometry of the proposed tag is shown in Figure 2. It consisted of 10 inverted M-shaped strips implemented on a Rogers RT5880 substrate, with a thickness of 0.256 mm and a dielectric constant = 2.2. In Figure 2, w and l are the width and length of the proposed tag, respectively, sx is is the horizontal separation between two vertical elements, which was fixed for all strip resonators, sy is the vertical separation between two slanted elements, which was different from one strip to another, and wr is the width of each strip. Each inverted M-shaped strip structure could be utilized in either an active or a passive mode of operation. The existence of a null in the frequency response denoted the active mode (logic 1), while the nonexistence of the null denoted the passive mode (logic 0). Note that the proposed tag was different from the structure developed in [44]; it was based on the inverted M-shaped strips implemented on the dielectric substrate, while that of [44] was based on U-shaped slots etched in a metallic background.  Note that the proposed tag was different from the structure developed in [44]; it was based on the inverted M-shaped strips implemented on the dielectric substrate, while that of [44] was based on U-shaped slots etched in a metallic background.  We have optimized the proposed structure to obtain deep notches at suitable frequencies in the operating frequency band from 3 GHz to 7.5 GHz. Essentially, each single-strip resonator was designed to operate at one resonance frequency (one notch). Then, another strip resonator was added to obtain two notches in the operating frequency band. In this way, the number of resonance frequencies could be controlled by removing or adding resonators. Our structure utilized 10 optimized strips to obtain 10 resonance frequencies (notches). The fundamental and theoretical analysis of these types of tags were discussed by F. Costa et al. in [25,34] and by N. Karmakar et al. in [9][10][11][12]30,45]. The total length of each strip resonator was the sum of the lengths of the vertical and slanted elements: 9,10) (1) Table 1 lists the total length of each strip resonator. The first strip, which was the longest, resonated at the lowest resonant frequency and corresponded to the most significant bit. On the other hand, the last strip, which had smallest length, gave the highest resonance frequency and corresponded to the least significant bit.  We have optimized the proposed structure to obtain deep notches at suitable frequencies in the operating frequency band from 3 GHz to 7.5 GHz. Essentially, each single-strip resonator was designed to operate at one resonance frequency (one notch). Then, another strip resonator was added to obtain two notches in the operating frequency band. In this way, the number of resonance frequencies could be controlled by removing or adding resonators. Our structure utilized 10 optimized strips to obtain 10 resonance frequencies (notches). The fundamental and theoretical analysis of these types of tags were discussed by F. Costa et al. in [25,34] and by N. Karmakar et al. in [9][10][11][12]30,45]. The total length of each strip resonator was the sum of the lengths of the vertical and slanted elements: 9,10) (1) Table 1 lists the total length of each strip resonator. The first strip, which was the longest, resonated at the lowest resonant frequency and corresponded to the most significant bit. On the other hand, the last strip, which had smallest length, gave the highest resonance frequency and corresponded to the least significant bit. We analyzed the proposed tag as an infinite structure using Floquet boundary conditions. When an exciting port sent an RF signal to the tag, a surface current was induced around each resonator at its resonant frequency. The proposed structure therefore exhibited frequency-selective performance, with deep notches at the desired resonant frequencies. Figure 4 shows the simulated response for all 10 strips (each shown separately).
Electronics 2020, 9,  We analyzed the proposed tag as an infinite structure using Floquet boundary conditions. When an exciting port sent an RF signal to the tag, a surface current was induced around each resonator at its resonant frequency. The proposed structure therefore exhibited frequency-selective performance, with deep notches at the desired resonant frequencies. Figure 4 shows the simulated response for all 10 strips (each shown separately).

Simulation Results
We designed and simulated seven different passive tags with seven different codes to validate the concept. Figure 3a shows the first code, where all 10 strips were in active mode, which corresponded to the structure code 1111111111. Figure 3b shows the second code, where all strips were in passive mode, representing the code 0000000000. The other five bit codes we chose were 1010101010, 0101010101, 1010000000, 1010001111, and 0000011111. The simulation reflection coefficients (S11) for these different codes are shown in Figures 5-8.

Simulation Results
We designed and simulated seven different passive tags with seven different codes to validate the concept. Figure 3a shows the first code, where all 10 strips were in active mode, which corresponded to the structure code 1111111111. Figure 3b shows the second code, where all strips were in passive mode, representing the code 0000000000. The other five bit codes we chose were 1010101010, 0101010101, 1010000000, 1010001111, and 0000011111. The simulation reflection coefficients (S 11 ) for these different codes are shown in Figures 5-8.     The developed structure could be excited by either a horizontally or vertically polarized plane wave. For each polarization, the response waveform had 10 notches in the given operating bandwidth. Table 2 presents the frequency differences between the two responses obtained for excitation with vertical and horizontal polarizations. The maximum variance in dBs between the  The developed structure could be excited by either a horizontally or vertically polarized plane wave. For each polarization, the response waveform had 10 notches in the given operating bandwidth. Table 2 presents the frequency differences between the two responses obtained for excitation with vertical and horizontal polarizations. The maximum variance in dBs between the resonant frequencies for the two polarizations was around 140 MHz, while the minimum variance was almost 14 MHz. A shift in resonant frequencies between the two polarizations was necessary to  The developed structure could be excited by either a horizontally or vertically polarized plane wave. For each polarization, the response waveform had 10 notches in the given operating bandwidth. Table 2 presents the frequency differences between the two responses obtained for excitation with vertical and horizontal polarizations. The maximum variance in dBs between the resonant frequencies for the two polarizations was around 140 MHz, while the minimum variance was almost 14 MHz. A shift in resonant frequencies between the two polarizations was necessary to achieve sufficiently sharp peaks for the two responses [25]. These sharp peaks could also be used for encoding data, because all the sharp peaks were present in the all-ones code. For this state, there were thus 10 sharp peaks, as shown in Figure 9. Figures 10 and 11 show the frequency responses and the differences in dBs between the vertically and horizontally polarized S 11 for the codes 1010001111 and 1010000000, respectively.

Strip Length Variation Method
The resonant frequency of each strip resonator was inversely proportional to its physical length. Consequently, increasing the strip resonator length decreased its resonant frequency, and vice versa. Hence, each M-strip was able to represent one of the four probable codes: 00, 01, 10, and 11. In the first code, 00, the strip resonator i worked in passive mode (with open circuits). No resonance frequency could be detected in the operating bandwidth. In the second code, 01, the length of strip i ( ) decreased by a small factor ∆ , and the resonance frequency became instead of . On the other hand, the third code, 10, was achieved when the length of strip i ( ) increased by a small factor ∆ . Therefore, the resonant frequency became instead of . The last code, 11, came with the original strip, which had a length and corresponded to the resonant frequency . These four codes for the single strip resonator are summarized in Table 3.
In this way, the number of bits per single strip resonator increased to two bits. Therefore, the coding capacity (number of bits per tag) and coding density (number of bits per centimeter square) could be doubled within the same frequency range. That means the coding capacity increased to 20 bits for the same size when the method of variation length was used. This saved 50% of the overall size. For the all-ones configuration (ID: 1111111111), the simulated values of S11 for the last four cases listed in Table 3 are presented in Figure 12. Table 4 lists the corresponding frequencies for the M-strip tag with the original length and increased and decreased strip lengths. The optimized value of ΔL used in our structure was 0.33 mm. The tolerance value not affecting the tag performance was about 110 micrometers.

Strip Length Variation Method
The resonant frequency of each strip resonator was inversely proportional to its physical length. Consequently, increasing the strip resonator length decreased its resonant frequency, and vice versa. Hence, each M-strip was able to represent one of the four probable codes: 00, 01, 10, and 11. In the first code, 00, the strip resonator i worked in passive mode (with open circuits). No resonance frequency could be detected in the operating bandwidth. In the second code, 01, the length of strip i (l i ) decreased by a small factor ∆L − , and the resonance frequency became f + i instead of f i . On the other hand, the third code, 10, was achieved when the length of strip i (l i ) increased by a small factor ∆L + . Therefore, the resonant frequency became f − i instead of f i . The last code, 11, came with the original strip, which had a length l i and corresponded to the resonant frequency f i . These four codes for the single strip resonator are summarized in Table 3. Table 3. Possible states for a single overturned M-shaped resonator.

Structure
Strip Length State Resonant Frequency

Binary Code
Strip i with open circuits l i 1st state 0 00 The length of strip i (l i ) decreased by factor ∆L − The length of strip i (l i ) increased by factor ∆L In this way, the number of bits per single strip resonator increased to two bits. Therefore, the coding capacity (number of bits per tag) and coding density (number of bits per centimeter square) could be doubled within the same frequency range. That means the coding capacity increased to 20 bits for the same size when the method of variation length was used. This saved 50% of the overall size.
For the all-ones configuration (ID: 1111111111), the simulated values of S 11 for the last four cases listed in Table 3 are presented in Figure 12. Table 4 lists the corresponding frequencies for the M-strip tag with the original length and increased and decreased strip lengths. The optimized value of ∆L used in our structure was 0.33 mm. The tolerance value not affecting the tag performance was about 110 micrometers.

Experimental Results
One of the designed tags was fabricated and experimentally tested to validate the idea. This prototype had the all-ones bit code (1111111111). Although the simulated reflection coefficients were the results for an infinite array of the proposed overturned M-strip tags, practically, we reduced the fabricated tag to a finite 3 × 3 array of unit cells to decrease the tag size and the fabrication cost. Figure  12 shows the fabricated code with a 3 × 3 array of unit cells.
We measured the response of the fabricated tag shown in Figure 12 using the measurement setup described in Figure 13. We measured its frequency response using a two-port power network analyzer PNA-X(N5242A) and a wideband horn antenna operating from 1 GHz to 18 GHz. The tag was placed at a distance of 10 cm from the RFID reader system. The measured and simulated results for the all-ones code are compared in Figure 14. The measured results were in suitable agreement with the simulation, although there were slight shifts at some frequencies due to manufacture and testing mistakes.

Experimental Results
One of the designed tags was fabricated and experimentally tested to validate the idea. This prototype had the all-ones bit code (1111111111). Although the simulated reflection coefficients were the results for an infinite array of the proposed overturned M-strip tags, practically, we reduced the fabricated tag to a finite 3 × 3 array of unit cells to decrease the tag size and the fabrication cost. Figure 12 shows the fabricated code with a 3 × 3 array of unit cells.
We measured the response of the fabricated tag shown in Figure 12 using the measurement setup described in Figure 13. We measured its frequency response using a two-port power network analyzer PNA-X(N5242A) and a wideband horn antenna operating from 1 GHz to 18 GHz. The tag was placed at a distance of 10 cm from the RFID reader system. The measured and simulated results for the all-ones code are compared in Figure 14. The measured results were in suitable agreement with the simulation, although there were slight shifts at some frequencies due to manufacture and testing mistakes.  It is observed from Figure 15 that there were ten frequency nulls in the backscattered signal from the fabricated tag. These ten resonance frequencies could be used to encode ten bits within the frequency band of 3 GHz-7.5 GHz.   It is observed from Figure 15 that there were ten frequency nulls in the backscattered signal from the fabricated tag. These ten resonance frequencies could be used to encode ten bits within the frequency band of 3 GHz-7.5 GHz.  It is observed from Figure 15 that there were ten frequency nulls in the backscattered signal from the fabricated tag. These ten resonance frequencies could be used to encode ten bits within the frequency band of 3 GHz-7.5 GHz.  It is observed from Figure 15 that there were ten frequency nulls in the backscattered signal from the fabricated tag. These ten resonance frequencies could be used to encode ten bits within the frequency band of 3 GHz-7.5 GHz.   The backscattered chipless RFID tags can be implemented in two forms: the strip-based structure proposed in this paper and the slot-based form, as given in [33]. Within the same frequency band and capacity, the proposed structure needs less copper material in its implantation; therefore, it is cheaper when compared with the slot-based structure. However, the proposed structure needs more optimization effort in order to reduce the mutual effect of adjacent lines, compared with the slot-based structure. Moreover, the distance between the tag and reader for accurate detection (60 cm) of the slot-base structure is greater than the distance between the tag and the reader of the proposed structure (10 cm). Although the proposed structure is lower in cost, these drawbacks may limit the applications of the proposed structure. Table 5 compares the performance of the proposed tag with those of some existing chipless RFID backscatter-based tags. The performance of a chipless RFID tag is defined not only by its encoding capacity (bits/cm 2 /GHz), but also by other metrics. These new metrics, defined in [23,46,47], include the spectral-capacity (bits/GHz), spatial density (bits/cm 2 ), spatial density at the center frequency (bits/λ 2 ), and the encoding capacity at the center frequency (bits/λ 2 /GHz).
In this table, the proposed tag exhibits high performance at the center frequency in terms of spatial density and encoding capacity. The proposed tag can be used to encode 11 bits/cm 2 , and it has a spectral capacity of 4.44 bits/GHz, a spatial density at the center frequency of 358.33 bits/λ 2 , an encoding capacity of 2.44 bits/cm 2 /GHz, and an encoding capacity at the center frequency of 79.63 bits/λ 2 /GHz. Other performance parameters of this device are listed in the last line of Table 5.

Conclusions
We have presented a novel and compact tag consisting of overturned M-shaped strips in this paper. We showed that the coding capacity of such a device could be doubled within the same frequency range of 3 GHz-7.5 GHz for a given tag size. We fabricated an all-ones tag and tested it at a distance of 10 cm from the RFID reader, and we compared the measured results with a simulation. The proposed tag can be used to encode 11 bits/cm 2 , and it has a spectral capacity of 4.44 bits/GHz, a spatial density at the center frequency of 358.33 bits/λ 2 , an encoding capacity of 2.44 bits/cm 2 /GHz, and an encoding capacity at the center frequency of 79.63 bits/λ 2 /GHz.