Chipless-RFID: A Review and Recent Developments

In this paper, a review of the state-of-the-art chipless radiofrequency identification (RFID) technology is carried out. This recent technology may provide low cost tags as long as these tags are not equipped with application specific integrated circuits (ASICs). Nevertheless, chipless-RFID presents a series of technological challenges that have been addressed by different research groups in the last decade. One of these challenges is to increase the data storage capacity of tags, in order to be competitive with optical barcodes, or even with chip-based RFID tags. Thus, the main aim of this paper is to properly clarify the advantages and disadvantages of chipless-RFID technology. Moreover, since the coding information is an important aspect in such technology, the different coding techniques, as well as the main figures of merit used to compare different chipless-RFID tags, will be analyzed.


Introduction
Radiofrequency identification (RFID) is a wireless technology for identification and tracking of objects, animals, persons, etc. which has moderately penetrated the market, as compared to optical identification systems (based on barcodes or QR-codes) [1][2][3][4][5][6][7]. Despite the fact that RFID offers significant advantages over optical systems, such as larger reading distances without the need of line-of-sight, high data storage capacity, and the possibility to read multiple tags simultaneously, the cost of traditional RFID tags (i.e., tags equipped with a silicon integrated circuit or chip), is excessively high for many applications involving low-cost items. This applies even to passive tags, fed by the interrogation signal, where batteries are not present. Thus, the typical cost of passive UHF-RFID tags is situated in the vicinity of several eurocents, much beyond the cost of optical barcodes or QR codes. Since the cost of chipped RFID tags is mainly dictated by the presence of the chip, the implementation of planar encoders, able to replace the chip in the tag, has been an object of intensive research in the last years, giving rise to the so-called chipless-RFID technology [8][9][10][11][12][13]. Thus, in chipless-RFID tags, the tag is equipped with a planar encoder (typically a printed pattern containing the ID code) and sometimes with an antenna for communication with the reader. By using low-cost conductive inks, the price of the tag can be substantially reduced. However, with chipless-RFID it is not possible to be as competitive as with chipped-RFID in terms of system performance (read range, data capacity, etc.). Indeed, the main research activity relative to chipless-RFID technology in the last years has been oriented towards increasing the data storage capacity and density of the tags. In this review paper, the main strategies for the implementation of chipless-RFID systems are discussed, and the latest or more relevant implementations are highlighted. Such approaches include (i) time domain systems, (ii) frequency domain systems, and (iii) hybrid systems.

On-Off Keying (OOK)
On-off keying (OOK) encoding is considered to be the simplest encoding method in time domain-based chipless-RFID systems. Basically, the logic states '0' and '1' are determined by the absence or presence, respectively, of a reflection of the interrogation signal in a specific time interval, as illustrated in Figure 2.
The OOK encoding method was reported for the first time by Zhang et al. [21] in 2006. The tag is composed of a meandered coplanar waveguide transmission line (CPW) with surface mount device (SMD) capacitors inserted in parallel at predefined positions (see Figure 3a). Such capacitors or discontinuities act as reflectors. A pulse of short duration, specifically 2 ns, which propagates through the transmission line CPW, is sent by the reader. The presence of a capacitive discontinuity at certain positions produces a mismatching that results in a reflected pulse which is transmitted back to the reader. The presence of such reflection is interpreted as the logic state '1', whereas its absence is associated to logic state '0'. An example of a measured 4-bit tag is shown in Figure 3b.  Zheng et al. [22] and Botao et al. [23], in 2008 and 2010, respectively, replaced the SMD capacitors by patches, in order to obtain a fully planar structure. In this case, such patches were printed during tag fabrication and disconnected from the microstrip transmission line, as can be seen in Figure 4a. Consequently, the identification code of the tag is programmed by means of the connection of certain patches to the delay line. In order to reduce the size of the tag by approximately a factor of four, using the same substrate as in [21], the pulse duration of the interrogation signal was reduced to 0.4 ns. An example of a measured 8-bit tag is shown in Figure 4b. In [24,25] a different strategy to implement chipless-RFID tags based on time domain with OOK coding was reported. The main novelty is the fact that the delay line is used in transmission, rather than in reflection. As shown in Figure 5a, the input signal is divided into two branches, a short and straight line (without delay), and the other one a longer and meandered line (with delay). The two branches are connected by a circulator (which acts as an isolator) to deflect the second pulse at the appropriate time. Thus, the temporal position of the pulse is modified with the position of the isolator, and consequently, the encoding of the tag. In order to clarify the encoding strategy, Figure 5b,c show the principle of code generation, based on the superposition of delayed signals, and a simulated code example, respectively. On the other hand, in 2012, a novel strategy based on delay lines able to support magneto-inductive waves (MIW), was proposed by Herraiz et al. [26][27][28]. As depicted in Figure 6a, MIWs are slow waves that propagate through a chain of magnetically coupled resonators [26,29]. Such slow waves exhibit a group velocity (v g ) around c/100 (where c is the speed of light in vacuum), hence maximizing the group delay (τ g ). Thus, for the same delay, shorter delay lines can be achieved (as compared to conventional lines). For the implementation of the delay lines, split ring resonators (SRRs) with rectangular topology were used to maximize the coupling. A microstrip transmission line coupled inductively to the first resonator of the chain of SRRs, acts as a transducer between the microstrip mode and the MIW mode of the delay line. Finally, the additional microstrip transmission lines are used as reflectors, and coupled to specific SRRs of the chain in order to generate a specific identification code. Thereby, the position of the reflectors is different for each tag (see Figure 6b), giving rise to different temporal responses, as appreciated in Figure 6c. The main limitation of this strategy is the fact that the high losses of the MIW delay lines prevent the implementation of tags with high information capacity.

Pulse Position Modulation
Pulse position modulation (PPM) encoding consists of the variation of the position of a pulse within a certain temporal window. Specifically, PPM encodes n bits with a single pulse, which occupies one of the 2 n slots of a temporal window (see Figure 7). The advantage of this encoding, with respect to OOK, is that it requires a smaller number of reflectors. Nevertheless, a longer time span is necessary in order to achieve the same information capacity. Most of the works reported in the literature that use PPM encoding use dispersive structures which exhibit a frequency-dependent group velocity. Thus, a broadband signal propagating through such structures experiences a different group delay for each spectral component [31,32]. This structure can be implemented by means of C-section all-pass networks [33], which are composed of two coupled lines (see Figure 8a), whose electrical length is λ g /4 at the frequency where the group delay is maximum. Therefore, the frequency position of such maximum depends on the length of the structure, as depicted in Figure 8b. Consequently, in the time domain, for the corresponding frequency, a pulse delay equal to the group delay was implemented. In 2010, a 3-bit tag based on cascaded C-section dispersive structures was implemented for the first time by Gupta et al. [35] (see Figure 9). Such a tag was implemented in order to provide maximums in the group delay at 3, 4, and 5 GHz. In this case, each pulse can provide two different positions within a temporal window, depending on its encoding ('0' or '1'). Other examples based on this concept can be found in [34][35][36][37].

Phase Modulation
In order to increase the information density of the tag, a different coding method based on the phase modulation of the input pulse to the tag was proposed by Mandel et al. [38,39]. Moreover, such tags include the use of composite right/left handed (CRLH) artificial delay lines, also used in [13,40]. The use of these lines is justified by their lower group velocity, as compared to conventional delay lines. This aspect allows for the implementation of shorter delay lines for a specific delay. In this case, complex loads of different values are used for the purpose of modifying the phase of the reflected pulses, as can be seen in Figure 10. As a result, with four different elements it is possible to achieve four different phase values (2 bits for each reflector). The number of codes that can be generated is, therefore, 4 k (where k is the number of reflectors of the delay line).

Time Division Multiplexing Based Tags
Time division multiplexing is a novel and unconventional strategy to implement chipless-RFID systems based on time domain, which is of particular interest in secure paper applications. Nevertheless, the working principle is completely different than those of the aforementioned chipless-RFID systems. Such a strategy was reported for the first time by Herrojo et al. in 2017 [41]. With this approach, the number of achievable bits is only limited by the area occupied by the tags and the ID code is contained in the amplitude modulated signal generated by the tags. Such tags typically consist of a chain of either resonant elements [42][43][44][45][46][47][48][49] or metallic strips [50][51][52], etched or printed at predefined and equidistant positions on a dielectric substrate. Basically, in a tag reading operation, the tag is mechanically displaced over the sensitive part of the reader. The reader is a planar microwave structure, typically a resonant element, coupled to a transmission line and fed by a harmonic signal. By positioning the resonator of the sensitive part of the reader at a short distance from the tag, the tag motion modulates the amplitude of the feeding signal at the output port of the transmission line (a sketch of the working principle of the proposed chipless-RFID system is depicted in Figure 11). This amplitude modulation is due to the electromagnetic coupling between the resonator of the reader and the resonators or metallic strips of the tag, which modulate the magnitude of the transmission coefficient at the frequency of the feeding signal. With respect to encoding, it can be done either by cutting those elements of the tag associated to the logic state '0' or by its presence/absence at predefined equidistant positions. The former strategy is important in order to reduce costs, since it is possible to fabricate thousands of identical tags with a single mask and programming such tags in a later stage [46]. In [48], such a strategy was demonstrated since the authors implemented an 80-bit tag by means of inkjet printing on a conventional paper (see Figure 12) with all bits set to '1' logic state, and then programmed with different codes by cutting those resonators associated to the logic state '0'. Such an approach provides, in terms of performance, high capacity and relatively small size at the expense of near-field tag reading. Nevertheless, the required proximity between the reader and the tag is not an issue in certain applications, such as secure paper (of interest to avoid counterfeiting or copying, e.g., of official documents, etc.).  Recently, an alternative approach based on all-dielectric electromagnetic encoders was proposed, avoiding the use of metallic elements (either resonant element or strips). In this case, the AM signal is generated by means of permittivity contrast. Specifically, in [53], the functionality of the system was demonstrated by implementing a chain of 10 apertures on the dielectric substrate RO4003C with permittivity and loss tangent of 3.55 and 0.0021, respectively. The sketch of the working principle is depicted in Figure 13, whereas the normalized measured envelope functions of three different codes are shown in Figure 14.

Retransmission-Based Tags
The working principle of retransmission-based tags is depicted in Figure 15. Such tags were designed, implemented, and patented by Preradovic et al. [55] for the first time. Typically, these chipless-RFID tags are equipped with a transmitting and receiving antenna, which are cross-polarized in order to communicate wirelessly with the reader. Such antennas are used for the reception of the interrogation signal and the transmission of the spectral signature of the tag. A tag to be highlighted is the one implemented by Preradovic et al. As shown in Figure 16a, the tag was implemented by a meander microstrip transmission line (to reduce the size) loaded with multiple resonators, specifically 35 spiral resonators, hence encoding 35 bits in a bandwidth and a surface of 4 GHz and 88 × 65 mm 2 (including antennas), respectively. With regard to the frequency response of the tag (see Figure 16b), it presents 35 transmission zeros in the magnitude of the spectral signature, introduced by each of the elements which are part of the tag.

Backscattered-Based Tags
In backscattered chipless-RFID based tags, the resonant elements provide the spectral signature through singularities in the response of the radar cross section (RCS). The main advantage of this strategy is the fact that the use of antennas is not necessary, hence the size of the tag is reduced. The working principle is illustrated in Figure 17. In 2005, Jalaly et al. [66] proposed a very simple solution consisting of multiple dipoles with a variable capacity implemented on a substrate. The main purpose of this strategy is to create multiple resonances in the bandwidth of the tag. Similar strategy, implemented by the same authors, consists of using dipoles with variable lengths [65], making the structure completely planar. The tags are implemented in the frequency range from 2.4 to 5.8 GHz and the information capacity is 5 bits (the size of the tag is not mentioned by the authors).
A tag to be highlighted is the one shown in Figure 18a and implemented by Vena et al. [69]. Such a tag is composed of 20 resonant C-shaped elements with resonance frequencies between 2 and 4 GHz, and a size of 25 × 70 mm 2 . As can be seen in Figure 18b, a peak appears in both the magnitude of the RCS and in the group delay at the resonance frequency of each resonator. With regard to the encoding, the logic state '0' is achieved by shifting the resonance frequency of the resonators (C-shaped resonator) to higher frequencies. Basically, such a frequency shift is achieved by short-circuiting the resonators. The fact that it is not necessary to remove them from the tag is an important aspect, since it opens the possibility of manufacturing all the tags with all the resonators short-circuited, programming them, and removing the short-circuit with a laser, for example. The main limitation of encoding information in the frequency domain is the bandwidth required to accommodate a large number of bits. Moreover, increasing the information capacity of the tags implies increasing the number of resonant elements, and, consequently, their size. In recent years, in order to minimize such limitations, different research groups have focused their efforts on designing tags able to encode information in more than one domain simultaneously, for example, frequency-phase, frequency-amplitude, etc. Such systems are known as hybrid systems, and the information capacity of the corresponding tags can be substantially increased, as compared to one of the frequency domain tags of this section.

Hybrid Tags
The main objective of hybrid systems is to assign more than two logic states to a single resonant element. In most cases, the approach to achieve such multi-state behavior is by encoding the information in the frequency [80][81][82][83][84]. Thus, the resonance frequency of each resonator can vary between different values within a predefined frequency window. Obviously, with this strategy, the number of states is limited by the bandwidth assigned to each resonant element. To clarify this concept, Figure 19 shows a chipless-RFID tag based on three circular rings. Such a tag operates in the frequency band from 3.1 to 10.6 GHz (i.e., in a bandwidth of 7.5 GHz which has been divided into 30 MHz slots). Therefore, there are 250 slots to be shared between three rings (80 for each one), and, consequently, 80 3 = 512,000 different codes (19 bits) can be generated in an area of 9 cm 2 . Another approach to achieve a multi-state functionality is to encode the information in the amplitude level of the transmission zeros introduced by the resonant elements of the tags [85,86] or in the level of the magnitude of the resonance peaks of the RCS [87][88][89]. An example of such encoding is shown in Figure 20. The information was encoded in the magnitude level of the resonance peak introduced by the C-shaped resonator. In this case, each resonator provides 2 bits (four states) of information. The bandwidth and the area of the tag are around 3 GHz and 7.2 cm 2 , respectively. In 2011, Vena et al. [90] presented a tag which encodes the information in two independent parameters, the phase deviation and the position of the frequency within a given frequency window.
Such a tag is constituted by five C-shaped resonators printed on a FR4 substrate (Figure 21a) with resonance frequencies within the frequency band from 2.5 GHz to 7.5 GHz. Each resonator exhibits a peak and a notch controlled, basically, by the resonator length and g/L ratio, respectively. In the example of Figure 21b, the logic state '00' is assigned to the response with a narrower phase deviation and a peak resonance at 2.5 GHz while the logical state '01' corresponds to the response with a wider phase deviation and maintains the same peak resonance. The logic states '10' and '11' correspond to a narrow and wide phase deviation, respectively, but in this case to the peak resonance at 3 GHz. It is clear that, with this example, 2 bits of information can be assigned to each resonator. However, varying the dimensions of the resonant elements, it is possible to increase the information capacity of each particle. In fact, in the work presented by the authors, each resonator provides 24 states. Therefore, it is possible to generate 24 5 = 7,962,654 different codes (equivalent to 22.9 bits) in a size of 8 cm 2 with only five resonators. Another possibility, proposed by El-Awamry et al. in [91], consists of a hybrid coding which combines the frequency position of a notch and its bandwidth. This coding technique consists of three different bandwidths (BW 1 , BW 2 , and BW 3 ) and three different frequency positions (f r , f r + ∆f r , f r − ∆f ), where f r is the resonant frequency of the resonant particle and ∆f is the frequency offset used to encode. In order to achieve three different bandwidths, the tag uses different coding elements: a dipole, a square ring, and a patch (see Figure 22a). Consequently, the control of the position in frequency of the notch is done by varying the dimensions of the resonant elements. With this strategy, 48 bits are achieved (4 bits per resonator) in a size of 20 cm 2 and using a bandwidth of 3 GHz. Polarization diversity is also useful to increase the information encoded in a given surface and bandwidth [92][93][94][95]. This method is used in [93] in order to multiply by a factor of two the information capacity of the tags in the same spectral bandwidth. Such tags consist of two sets of rectangular slot resonators etched on rectangular metallic patches with different polarizations (vertical and horizontal), and in turn, each set of resonators is divided in two patches. It is important to mention that the slots etched in each patch are conveniently placed in order to reduce the coupling between resonant elements. On the other hand, in the tag reading operation, the tag must be illuminated with a plane wave orthogonally polarized (Figure 23a). Therefore, the reader consists of two dual-polarized antennas, one to transmit the interrogation signal and the other one to receive the spectral signature of the tag.
With this strategy, as a proof-of-concept demonstrator, the authors encoded 16 bits within an area and bandwidth of 3.06 cm 2 and 5 GHz, respectively.

Comparative Analysis between Coding Techniques
As described in the previous sections, there are different parameters (tag size, bandwidth, number of bits per resonator, etc.), which must be considered in order to determine the efficiency of the coding method used. As mentioned in the introduction, the main challenges in chipless-RFID technology are: (i) to encode the maximum information in the minimum possible area, and in turn, (ii) minimizing the bandwidth. Thus, it is interesting to define figures of merit (or criteria) in order to evaluate and/or compare the different coding methods as well as tag designs. Such figures of merit measure the information density per surface (DPS) in bits/cm 2 and density per frequency (DPF) in bits/GHz. Table 1 compares different chipless-RFID tags which can be found in the literature, including those shown and described in this paper. In general, in time domain-based tags where the interrogation signal is a pulsed signal, the DPS is very low due to the fact that it is necessary to implement transmission lines with large lengths in order to produce measurable delays. Conversely, those tags based on time division multiplexing (the interrogation signal is a harmonic signal), present, to the best of our knowledge, the largest DPS. Although this strategy provides a high data density per surface and bandwidth, it is at the expense of sacrificing read range. Nevertheless, this is not an issue in applications such as secure paper (where reading by proximity may be even convenient in order to provide major levels of confidence against eavesdropping or spying). On the other hand, typically, either frequency domain-based tags or hybrid systems need a large number of resonators in order to accommodate a significant number of bits. In fact, both encoding methods provide, in general, tags with similar DPS (either in terms of cm 2 or λ g 2 ) as can be appreciated either in Figure 24, where the historic evolution of DPS for different encoding methods is shown, or in the Table 1. However, hybrid tags provide larger DPF as compared to frequency domain tags. With regard to the read range (RR), it is difficult to compare both encoding systems (frequency domain and hybrid tags) due to the fact that, in most cases, relevant information, such as transmitted power or antenna gain, among others, are not given in the literature. Indeed, from the results of Table 1, it is difficult to identify a superior encoding system in terms of read range.  * The spectral bandwidth of this approach is virtually null due to the fact that the interrogation signal is harmonic. ** Bandwidth of the pulsed interrogation signal. *** In this approach the reader and the tag must be in contact.

Conflicts of Interest:
The authors declare no conflicts of interest.