Very Low-Cost 80-Bit Chipless-RFID Tags Inkjet Printed on Ordinary Paper

This paper presents a time-domain, chipless-RFID system with 80-bit tags inkjet-printed on ordinary DIN A4 paper. The tags, consisting of a linear chain of resonant elements (with as many resonators as the number of identification bits plus header bits), are read sequentially and by proximity (through near-field coupling). To this end, a transmission line, fed by a harmonic (interrogation) signal tuned to the resonance frequency of the tag resonators (or close to it), is used as a reader. Thus, during reader operation, the tag chain is mechanically shifted over the transmission line so that the coupling between the line and the functional resonant elements of the tag chain is favored. Logic states that ‘1’ and ‘0’ are determined by the functionality and non-functionality (resonator detuning), respectively, of the resonant elements of the chain. Through near-field coupling, the transmission coefficient of the line is modulated and, as a result, the output signal is modulated in amplitude (AM), which is the identification code contained in the envelope function. As long as the tags are inkjet-printed on ordinary DIN A4 paper, the cost is minimal. Moreover, such tags can be easily programmed and erased, so that identical tags can be fabricated on a large scale (and programmed at a later stage), further reducing the cost of manufacture. The reported prototype tags, with 80 bits of information plus four header bits, demonstrate the potential of this approach, which is of particular interest to secure paper applications.


Introduction
Chipless radiofrequency identification (chipless-RFID) has emerged in the last fifteen years as an alternative to RFID systems with tags equipped with silicon integrated circuits (IC) or chips [1][2][3][4][5][6]. State-of-the-art chipless-RFID systems cannot compete with chipped-RFID in terms of data storage capacity, size, and read distances [7,8]. However, one clear limitation of chipped tags is their cost, which is connected to the necessary presence of the silicon chip. Chipless-RFID tags solve this problem. Thus, chipless-RFID systems are of particular interest to applications devoted to low-cost tagged items.
Recently, numerous efforts have been dedicated to improving the data storage capability of chipless-RFID systems and to reduce the size of the ID encoders, typically consisting of a metallic pattern printed on a flexible dielectric substrate (which include plastic or paper, among others). There are two predominant strategies for the implementation of chipless-RFID systems: (i) time-domain 2. The Chipless-RFID System: Operating Principle, Reader, and Tag

Operating Principle
The operating principle of the time-domain chipless-RFID system, which is based on near-field and sequential bit reading was first reported in Herrojo et al. [50] and is succinctly reviewed in this paper for coherence and completeness. The tag consists of a linear chain of resonant elements (all identical) printed or etched at predefined and equidistant positions on a dielectric substrate (typically a flexible substrate, such as plastic, paper, etc.). The ID code is given by the functionality or non-functionality of the resonant elements, in which a resonator is not functional if it is detuned, namely, not resonating at its fundamental resonance frequency. Resonator detuning may be achieved, e.g., by cutting it. For tag reading, an element able to detect the functionality of the resonant elements is required. For such purposes, a transmission line, which is conveniently fed by a single tone signal tuned to the resonance frequency of the tag resonators or close to it, is adequate. Thus, by displacing the tag chain over the fed line, a coupling between the line and the resonant elements is expected (as long as the resonant elements are functional and the distance between the line and the resonant elements is small). Each time a functional resonator lies on top of the line, the coupling prevents the signal from transmitting through the line. The effect is a variation of the transmission coefficient at the feeding frequency due to near-field coupling; consequently, the output signal is amplitude modulated. Hence, the ID code is contained in the envelope function of the amplitude modulated (AM) signal, which can be inferred by means of an envelope detector, a simple circuit able to provide the envelope function where the ID code is contained. Indeed, the ID code is provided in time-domain, where the logic state '1' is given by the dips in the time response of the envelope function. The working principle of this unconventional time-domain chipless-RFID system is illustrated in Figure 1. When compared to frequency-domain chipless-RFID systems, the time-domain approach illustrated in Figure 1 exhibits a significant advantage: the interrogation signal is a harmonic (single tone) signal. Thus, the cost of the reader can be significantly reduced because a multi-frequency sweeping signal is not required. Indeed, the spectral bandwidth of the proposed tags is virtually null because all the resonant elements of the tag are identical (i.e., tuned to the same frequency). The difference between these tags and spectral signature barcodes is that the reading is based on frequency division multiplexing in spectral signature barcodes. In contrast, in the novel chipless-RFID systems proposed by us, time division multiplexing is used for tag reading.
Technologies 2018, 6, x FOR PEER REVIEW 3 of 10 domain, where the logic state '1' is given by the dips in the time response of the envelope function. The working principle of this unconventional time-domain chipless-RFID system is illustrated in Figure 1. When compared to frequency-domain chipless-RFID systems, the time-domain approach illustrated in Figure 1 exhibits a significant advantage: the interrogation signal is a harmonic (single tone) signal. Thus, the cost of the reader can be significantly reduced because a multi-frequency sweeping signal is not required. Indeed, the spectral bandwidth of the proposed tags is virtually null because all the resonant elements of the tag are identical (i.e., tuned to the same frequency). The difference between these tags and spectral signature barcodes is that the reading is based on frequency division multiplexing in spectral signature barcodes. In contrast, in the novel chipless-RFID systems proposed by us, time division multiplexing is used for tag reading.

The Reader
As mentioned in the previous subsection, the active part of the reader is a transmission line fed by a harmonic signal. Particularly, the structure under consideration is a microstrip line loaded with a split-ring resonator (SRR) in bandpass configuration, as depicted in Figure 2 [54]. This structure exhibits two singular frequencies: (i) a reflection zero frequency (where the signal between the input and the output port is transmitted), and (ii) a transmission zero frequency. Accordingly, a large excursion of the transmission coefficient is achieved when a resonant element identical to the SRR of the line is aligned and oppositely oriented to it. The net effect is an overall displacement of the transmission coefficient to the left, as depicted in Figure 2b. By properly tuning the resonator dimensions as well as the vertical distance between the line and the oppositely oriented resonator (air gap), it is possible to optimize the variation of the transmission coefficient at the frequency of maximum transmission of the unloaded SRR-loaded line, fc, as seen in Figure 2b. Thus, the tags will be implemented as chains of SRRs, printed on a different substrate, and tag reading will proceed by mechanically shifting the tag chain over the SRR of the line, with the SRRs of the tag oppositely oriented and aligned with regard to the SRR of the line. In practice, it is difficult to keep the air gap uniform, but some tolerance in the air gap distance is acceptable as long as the variation of the transmission coefficient at fc is significant, even when the air gap deviates from the optimum value ( Figure 3).

The Reader
As mentioned in the previous subsection, the active part of the reader is a transmission line fed by a harmonic signal. Particularly, the structure under consideration is a microstrip line loaded with a split-ring resonator (SRR) in bandpass configuration, as depicted in Figure 2 [54]. This structure exhibits two singular frequencies: (i) a reflection zero frequency (where the signal between the input and the output port is transmitted), and (ii) a transmission zero frequency. Accordingly, a large excursion of the transmission coefficient is achieved when a resonant element identical to the SRR of the line is aligned and oppositely oriented to it. The net effect is an overall displacement of the transmission coefficient to the left, as depicted in Figure 2b. By properly tuning the resonator dimensions as well as the vertical distance between the line and the oppositely oriented resonator (air gap), it is possible to optimize the variation of the transmission coefficient at the frequency of maximum transmission of the unloaded SRR-loaded line, f c , as seen in Figure 2b. Thus, the tags will be implemented as chains of SRRs, printed on a different substrate, and tag reading will proceed by mechanically shifting the tag chain over the SRR of the line, with the SRRs of the tag oppositely oriented and aligned with regard to the SRR of the line. In practice, it is difficult to keep the air gap uniform, but some tolerance in the air gap distance is acceptable as long as the variation of the transmission coefficient at f c is significant, even when the air gap deviates from the optimum value ( Figure 3).  The reader consists not only of the SRR-loaded line but also the envelope detector, plus additional electronics that are necessary to generate the harmonic (interrogation) signal. In numerous studies by Herrojo et al. [50][51][52][53][54][55], the envelope detector was implemented by means of a Schottky diode (Avago HSMS-2860), an active probe which acts as a low-pass filter (with R = 1 MΩ and C = 1 pF) and an isolator to prevent reflections from the diode. In this work, an integrated circuit (Analog Devices ADL5511), able to provide the envelope function (hence reducing cost), was alternatively used. The envelope signal was then driven to a data acquisition card (National Instruments myRIO) connected to a computer, where the sampled data could be viewed. The harmonic feeding signal was generated by means of the Agilent E44338C function generator. An important part of the reader is the mechanical system, which is needed to displace the tag over the transmission line of the reader in close proximity to it. To this end, a printer was adapted and equipped with an adhoc guiding channel to absorb the tag (printed encoder). Good alignment and uniform air gap distance between the tag chain and the resonator of the reader was achieved with this system. Nevertheless, our system supported some tolerances in the vertical direction (air gap) as well as lateral misalignments, as discussed in Herrojo et al. [54]. With regard to the speed of reading, the system is limited by the sampling rate of the reader. However, for any practical mechanical guiding system used to displace the tag over the reader, the achievable speed did not represent an issue in this regard. The photograph of the complete reader is depicted in Figure 4, where the different reader parts are indicated.  The reader consists not only of the SRR-loaded line but also the envelope detector, plus additional electronics that are necessary to generate the harmonic (interrogation) signal. In numerous studies by Herrojo et al. [50][51][52][53][54][55], the envelope detector was implemented by means of a Schottky diode (Avago HSMS-2860), an active probe which acts as a low-pass filter (with R = 1 MΩ and C = 1 pF) and an isolator to prevent reflections from the diode. In this work, an integrated circuit (Analog Devices ADL5511), able to provide the envelope function (hence reducing cost), was alternatively used. The envelope signal was then driven to a data acquisition card (National Instruments myRIO) connected to a computer, where the sampled data could be viewed. The harmonic feeding signal was generated by means of the Agilent E44338C function generator. An important part of the reader is the mechanical system, which is needed to displace the tag over the transmission line of the reader in close proximity to it. To this end, a printer was adapted and equipped with an adhoc guiding channel to absorb the tag (printed encoder). Good alignment and uniform air gap distance between the tag chain and the resonator of the reader was achieved with this system. Nevertheless, our system supported some tolerances in the vertical direction (air gap) as well as lateral misalignments, as discussed in Herrojo et al. [54]. With regard to the speed of reading, the system is limited by the sampling rate of the reader. However, for any practical mechanical guiding system used to displace the tag over the reader, the achievable speed did not represent an issue in this regard. The photograph of the complete reader is depicted in Figure 4, where the different reader parts are indicated. The reader consists not only of the SRR-loaded line but also the envelope detector, plus additional electronics that are necessary to generate the harmonic (interrogation) signal. In numerous studies by Herrojo et al. [50][51][52][53][54][55], the envelope detector was implemented by means of a Schottky diode (Avago HSMS-2860), an active probe which acts as a low-pass filter (with R = 1 MΩ and C = 1 pF) and an isolator to prevent reflections from the diode. In this work, an integrated circuit (Analog Devices ADL5511), able to provide the envelope function (hence reducing cost), was alternatively used. The envelope signal was then driven to a data acquisition card (National Instruments myRIO) connected to a computer, where the sampled data could be viewed. The harmonic feeding signal was generated by means of the Agilent E44338C function generator. An important part of the reader is the mechanical system, which is needed to displace the tag over the transmission line of the reader in close proximity to it. To this end, a printer was adapted and equipped with an adhoc guiding channel to absorb the tag (printed encoder). Good alignment and uniform air gap distance between the tag chain and the resonator of the reader was achieved with this system. Nevertheless, our system supported some tolerances in the vertical direction (air gap) as well as lateral misalignments, as discussed in Herrojo et al. [54]. With regard to the speed of reading, the system is limited by the sampling rate of the reader. However, for any practical mechanical guiding system used to displace the tag over the reader, the achievable speed did not represent an issue in this regard. The photograph of the complete reader is depicted in Figure 4, where the different reader parts are indicated.

The Tags
As previously mentioned, the tags consist of a linear chain of SRRs, which is identical to the one of the reader line. To demonstrate the viability of the system with very low-cost tags, we opted to use ordinary DIN A4 paper as tag substrate, with measured dielectric constant and loss tangent of εr = 3.11 and tanδ = 0.036, respectively. Such measurements were carried out by means of the resonant cavity 85072A. Clearly, the response of the SRR-loaded line of the reader, which, in turn, was loaded with the tag, depended on the tag substrate, but this dependence was relatively soft as long as tag substrates with similar thickness and dielectric constant were considered. Note also that the dielectric constant value of the substrate of the reader line was not related to the substrate of the tag. Such value (the dielectric constant of the reader substrate) was chosen relatively high to reduce the size of the resonant element of the reader and, consequently, the resonant elements of the tag (provided they are identical). Moreover, we implemented the tags by inkjet-printing using Dupont TM PE410 conductive ink, with a measured conductivity of 7.28 × 10 6 S/m. One layer (with measured thickness of 2.6 μm) suffices, thereby optimizing tag cost. The number of resonant elements was set to 80 for the ID code (corresponding to 80-bit tags), plus four SRRs for the header bits. Such header bits were needed because the tag could be displaced over the reader starting from the two extremes (corresponding to tag SRRs either face up or down). The photograph of the fabricated tag with all ID bits set to the logic state '1' (i.e., with all resonators functional) is depicted in Figure 5. The total length and width of the tag (including the header bits) was 267.2 mm and 3.35 mm, respectively. With this size, the tag could be accommodated along the longitudinal side of DIN A4 paper. For security and authentication of this type of document, the limit was 88 bits (plus the header bits) because this is the number of bits achievable along the longer side of DIN A4, taking into account that the tag period was 3.36 mm.

The Tags
As previously mentioned, the tags consist of a linear chain of SRRs, which is identical to the one of the reader line. To demonstrate the viability of the system with very low-cost tags, we opted to use ordinary DIN A4 paper as tag substrate, with measured dielectric constant and loss tangent of ε r = 3.11 and tanδ = 0.036, respectively. Such measurements were carried out by means of the resonant cavity 85072A. Clearly, the response of the SRR-loaded line of the reader, which, in turn, was loaded with the tag, depended on the tag substrate, but this dependence was relatively soft as long as tag substrates with similar thickness and dielectric constant were considered. Note also that the dielectric constant value of the substrate of the reader line was not related to the substrate of the tag. Such value (the dielectric constant of the reader substrate) was chosen relatively high to reduce the size of the resonant element of the reader and, consequently, the resonant elements of the tag (provided they are identical). Moreover, we implemented the tags by inkjet-printing using Dupont TM PE410 conductive ink, with a measured conductivity of 7.28 × 10 6 S/m. One layer (with measured thickness of 2.6 µm) suffices, thereby optimizing tag cost. The number of resonant elements was set to 80 for the ID code (corresponding to 80-bit tags), plus four SRRs for the header bits. Such header bits were needed because the tag could be displaced over the reader starting from the two extremes (corresponding to tag SRRs either face up or down). The photograph of the fabricated tag with all ID bits set to the logic state '1' (i.e., with all resonators functional) is depicted in Figure 5. The total length and width of the tag (including the header bits) was 267.2 mm and 3.35 mm, respectively. With this size, the tag could be accommodated along the longitudinal side of DIN A4 paper. For security and authentication of this type of document, the limit was 88 bits (plus the header bits) because this is the number of bits achievable along the longer side of DIN A4, taking into account that the tag period was 3.36 mm.

The Tags
As previously mentioned, the tags consist of a linear chain of SRRs, which is identical to the one of the reader line. To demonstrate the viability of the system with very low-cost tags, we opted to use ordinary DIN A4 paper as tag substrate, with measured dielectric constant and loss tangent of εr = 3.11 and tanδ = 0.036, respectively. Such measurements were carried out by means of the resonant cavity 85072A. Clearly, the response of the SRR-loaded line of the reader, which, in turn, was loaded with the tag, depended on the tag substrate, but this dependence was relatively soft as long as tag substrates with similar thickness and dielectric constant were considered. Note also that the dielectric constant value of the substrate of the reader line was not related to the substrate of the tag. Such value (the dielectric constant of the reader substrate) was chosen relatively high to reduce the size of the resonant element of the reader and, consequently, the resonant elements of the tag (provided they are identical). Moreover, we implemented the tags by inkjet-printing using Dupont TM PE410 conductive ink, with a measured conductivity of 7.28 × 10 6 S/m. One layer (with measured thickness of 2.6 μm) suffices, thereby optimizing tag cost. The number of resonant elements was set to 80 for the ID code (corresponding to 80-bit tags), plus four SRRs for the header bits. Such header bits were needed because the tag could be displaced over the reader starting from the two extremes (corresponding to tag SRRs either face up or down). The photograph of the fabricated tag with all ID bits set to the logic state '1' (i.e., with all resonators functional) is depicted in Figure 5. The total length and width of the tag (including the header bits) was 267.2 mm and 3.35 mm, respectively. With this size, the tag could be accommodated along the longitudinal side of DIN A4 paper. For security and authentication of this type of document, the limit was 88 bits (plus the header bits) because this is the number of bits achievable along the longer side of DIN A4, taking into account that the tag period was 3.36 mm. Figure 5. Photograph of the 80-bit tag fabricated on ordinary paper, with all bits set to '1' and with header bits. Figure 5. Photograph of the 80-bit tag fabricated on ordinary paper, with all bits set to '1' and with header bits.
Notably, tags can be programmed and erased after fabrication. This is an important aspect which can lead to a further reduction in fabrication costs, provided all-identical tags can be massively manufactured (requiring only a single mask) and then programmed at a later stage (custom level). Tag programming can be achieved by cutting those resonators with logic state '0', hence detuning them. This can be done mechanically or by means of other systems such as laser ablation. Moreover, tag erasing can be achieved by simply short-circuiting those previously detuned SRRs (for instance, by means of inkjet-printing). Tag erasing (after programming) and reprogramming is unusual, but this possibility exists and suggests that a single tag can be recycled several times.

Results and Discussion
To illustrate the possibilities of the presented approach, we have obtained the ID code of the fabricated tag with all ID bits set to '1'. Figure 6 depicts the inferred code, with tag SRRs face up and down. Clearly, the code was correctly read regardless of the relative orientation between the tag and the reader. We then programmed the tag by cutting certain resonant elements.
In particular, we considered the sequence (for the ID code) '111111111011111110111111111111100111111111100011111110101111111110111111111111'. In this case, tag reading provided the responses which are shown in Figure 6b, with tag SRRs face up and down, to demonstrate that the ID code is successfully obtained. Finally, we erased the tag and reprogrammed it with the ID code, '1111111111111111111101010111111111111001 00111111111111111111010111111111111000'. Results which were derived from the tag reading and are depicted in Figure 6c, revealed that, again, the ID code was correctly inferred, regardless of tag orientation. Indeed, we erased and reprogrammed several times, with a successful tag reading each time, indicating that the proposed system is robust. These results clearly suggest the following: (i) the number of achievable bits with this system is only limited by tag size; (ii) tag cost is very small as long as ordinary DIN A4 paper is used for tag substrate and a single layer of conductive ink suffices to correctly read the tags; (iii) tags can be programmed and erased as many times as needed. With these characteristics, the proposed chipless-RFID system is of interest for applications devoted to secure paper, in which, by printing the ID codes directly onto the document of interest, it is equipped with a secure code which is difficult to reproduce. Indeed, the codes can be photocopied, but tag reading requires a sophisticated system of the type shown in Figure 4. Moreover, it is possible to fabricate the tags through lamination, resulting in buried encoders difficult to reproduce. Banknotes, certificates, corporate documents, ballots, exams, medical recipes, etc. are different types of documents that may benefit of the proposed encoding system, providing them additional security against counterfeiting.
Future research will be to boost the number of resonant elements per unit length of the tag. Typically, in chipless-RFID systems, one benefit is the density of information per area unit. However, in the proposed chipless-RFID system, the tags exhibit an extremely long shape factor and, consequently, the interest is to accommodate the largest possible number of bits along the tag length (typically limited by the size of the document). Obviously, this can be achieved by increasing the resonance frequency of the SRRs (hence reducing its size), but this typically increases the cost of the associated electronics (harmonic signal generation, etc.). Therefore, the next challenge is to find novel resonator topologies, which are very short in the longitudinal direction of the tag chain, providing high sensitive tags that can be read through near-field coupling (sequentially) by means of a dedicated (adhoc) reader. Because linear half-wavelength resonators are very narrow, we are currently considering these resonant elements for tag implementation, where the idea is to transversely dispose the resonators along the tag. This strategy may increase the number of bits per unit length dramatically. However, it is necessary to find a suitable reader, which is able to reliably provide the ID codes in tags with extremely closed resonators. Work is in progress in this regard. Because the envelope function has been inferred from a data acquisition system, the abscise axis is the number of samples. Note that there is no difference in the reading results between face-up and facedown tags. The codes are read in opposite directions for face-up or face-down tags, but this is indicated by the header bits and, therefore, there is no confusion regarding this aspect. (a) code with all bits set to '1' logic state, (b) programmed tag and (c) reprogrammed tag.

Conclusions
To conclude, 80-bit chipless-RFID tags inkjet printed on ordinary DIN A4 paper have been presented here for the first time. These tags are of special interest for secure paper applications, where the encoders can be directly printed on the document of interest. Reading by proximity (through near-field), as required in the reported chipless-RFID system, is not an issue in such applications, where the small distance between the tag and the reader may provide a certain level of confidence against spying, snooping, or eavesdropping. Additional advantages of the system are the high number of achievable bits (only limited by tags size), the low cost of the tags, and the possibility to program and erase the tags several times.

Conclusions
To conclude, 80-bit chipless-RFID tags inkjet printed on ordinary DIN A4 paper have been presented here for the first time. These tags are of special interest for secure paper applications, where the encoders can be directly printed on the document of interest. Reading by proximity (through near-field), as required in the reported chipless-RFID system, is not an issue in such applications, where the small distance between the tag and the reader may provide a certain level of confidence against spying, snooping, or eavesdropping. Additional advantages of the system are the high number of achievable bits (only limited by tags size), the low cost of the tags, and the possibility to program and erase the tags several times.