A Dual-Resonant Planar Microstrip Antenna Design for UHF RFID Using Paperboard as a Substrate

A Dual-Resonant Planar Microstrip Antenna Design for UHF RFID Using Paperboard as a Substrate Mutharasu Sivakumar, Daniel D.Deavours Information and Telecommunication Technology Center University of Kansas email: {muthus, deavours} Abstract Passive UHF RFID tags are commonly used in the supply chain, and generally use dipole antennas. It is well known that the performance of these tags degrade when placed near water or metal continents. Thus, it is desirable for tags to have consistent performance regardless of the contents of containers. Microstrip antennas offer a potential solution, but variations in the dielectric properties of corrugated paperboard with varying moisture content provide unique challenges. In this paper, we investigate whether microstrip antennas are appropriate for cardboard containers, and show a novel dualresonant microstrip antenna and matching circuit that provides good performance over a wide range of dielectric properties. The antenna design can be realized using low-cost, commodity inlay and smart label manufacturing techniques. T I. INTRODUCTION HE majority of passive UHF RFID tags available in the market are tuned for optimal performance on free space or when attached to corrugated paperboard/fiberboard (commonly called cardboard ). That is because the great majority of tags are destined to be integrated into smart labels for tagging paperboard for use in the supply chain (e.g., Wal-Mart mandates ). While stripline dipoles offer good performance and low manufacturing costs (as little as $.1 US in high volume as of 26), their primary limitation is that the performance is highly dependent on the materials to which they are attached. In the case of paperboard containers, the contents of the container can have a dramatic affect on tag performance. Recent advances in lower-power IC designs offer some improvements, but those are being offset by lower interrogator usage in order to limit the interrogation zones. The materials of greatest significance are metal and water. Metal in the form of canned goods are relatively easy to manage by finding the air gaps between cans. The most difficult products are those that contain metallic foil in their packaging, which is common, e.g., in dishwashing detergent. Water-based products are also difficult if they are packaged in rectangular-shaped containers so that there is no predictable air gap. In principle, these limitations can be overcome by the use of microstrip antennas [1]. There are a number of microstrip Manuscript received November 24, 26. antenna solutions for UHF RFID [2] [3] [4] and [5], but the main limitation of those approaches is the complexity in manufacturing those tags, and thus they are not amenable to use in the supply chain. Indeed, the cost of some microstrip tags can exceed the cost of a cardboard container. In contrast, [1] provides a solution that requires no connection from the antenna and the ground plane, and thus is potentially viable for use in the supply chain. The only modification is that a metal foil needs to be applied to the inside of the container, and the paperboard becomes the substrate of the microstrip antenna. We discuss the particular antenna design below. The primary technical challenge to this approach is that the dielectric properties of paperboard can change substantially with humidity and temperature. If the relative dielectric constant changes, for example, from 1.3 to 1.6 with increase in relative humidity, the resonant frequency of the antenna can shift from 915 MHz to under 8 MHz, and render the narrowband antenna practically unusable. Furthermore, the loss tangent increases substantially with the increased water absorption, resulting in poor impedance match. The purpose of this paper is to show that we can design microstrip antennas that take into account the changing dielectric properties of paperboard under different humidity conditions. We do this by designing a very simple dualresonant patch antenna and matching circuit. The elements are designed so that the first antenna and matching circuit resonates and provides a good match with the lower dielectric constant and loss tangent, and the second antenna and matching circuit resonates and provides a good match at the higher dielectric constant and loss tangent. The increased complexity of our proposed solution is the application of a metallic foil to a portion of the interior of the container. II. APPROACH In this section, we review known properties of paperboard and previous work on planar microstrip antennas. A. Properties of paperboard Corrugated cardboard or corrugated paperboard (hereafter paperboard) can be classified based on the number of paper

walls. Cardboard commonly comes in three forms: single faced, in which flutes have paper on one side, single walled, in which flutes are sandwiched between paper walls, and double walled, in which two layers of flutes are contained by three paper walls. Single-walled paperboard is the type most commonly used in the supply chain. Paperboard can also be classified in the type of flutes and number of flutes per unit length. The C flute is the most common, and the flute density of 42 flutes every 12 inches, i.e., approximately one flute per 2/7th of an inch. For this paper, we use a sample of C flute, 42 flutes per 12 inches, and 15 mils (.15 inches, 3.81 mm) thick. Table I: Sorption / desorption of paperboard [2]. Temp. 4% RH 9% RH 1 C 8 g/1g (Sorption) 16 g/1g (Sorption) 1 g/1g (Desorption) 18 g/1g (Desorption) 4 C 6 g/1g (Sorption) 14 g/1g (Sorption) 1 g/1g (Desorption) 18 g/1g (Desorption) The main limitation of using paperboard as a substrate for a microstrip patch antenna is the fact that dielectric properties of the paperboard can change over time depending on the water content in the cardboard. Water content in cardboard is typically measured in grams of water per 1 grams of solid substance. Water content follows a sorption / desorption isotherm based on the relative humidity and temperature [6]. When the relative humidity increases, the sorption rate is used to calculate the moisture content, and when the relative humidity decreases, the desorption rate is used. Table 1 illustrates a typical sorption / desorption isotherm. As Table I indicates, water content can vary between 6 and 18 g/1g. Equilibrium moisture content of single walled paperboard at 2 degree Celsius and 54.4% relative humidity (approximately room temperature conditions) is 9.34 grams per 1 grams of solid substance [4], although we measured a lower rate for our paperboard sample (see Section 3). Considering temperatures between 1 and 4 degree Celsius and RH between 4 and 9 percent, the range of water contents we are looking for would be from 6 to 18 grams per 1 grams of solid paper substance. To be conservative, we consider water content rates between 5 and g/1g. In Section 3, we measure the dielectric properties for paperboard with varying moisture content. B. Microstrip Antennas Traditional microstrip antennas for RFID tags (e.g., [3]), require some direct electrical connection between the antenna or matching circuit and the ground plane. For lowcost packaging, introducing a via is simply not cost effective. A completely planar microstrip antenna design in which the antenna, matching circuit, and RFID IC are all on the same plane (e.g., [1]) is ideally suited for such an application. Using this method, the only additional step is to include a ground plane in the paperboard, which can be performed relatively easily at manufacturing time, either by integrating a metal foil into the paper wall, or by applying a metal foil tape to the interior of the box before it is formed. The microstrip inlay (antenna and IC on a PET substrate) can be converted to a smart label (label with inlay affixed) and applied using the usual label-application method. To exercise this approach, what is necessary is to design a planar microstrip tag for the paperboard substrate. The primary difficulty of this approach is that the dielectric properties of paperboard can change substantially with changes in relative humidity. What we will show is that a simple, rectangular antenna and matching circuit that is designed for relatively low water content becomes practically unreadable at higher moisture contents. Thus, we need to develop a microstrip antenna and matching circuit that is able to match to the range of different dielectric properties of the paperboard. III. DIELECTRIC PROPERTIES OF PAPERBOARD If we can find the percentage of water by volume for various values of water content then we will be able to obtain a theoretical bound on the dielectric constants we need to operate over, provided we know the dielectric properties of water over the range of temperatures and water behaves the same way electrically when bonded with the paper molecules and at free state. Dielectric properties of water have been evaluated extensively and available in literature [7]. Values of relative permittivity and loss tangent for water at different temperatures are tabulated in Table II. Table II: Dielectric properties of water at 915 MHz. T ( C) ε r tan δ 87..95 78.2.44 4 73.1.31 We found that the density of our dry cardboard sample has a density of.14 g/cc. It is well known that water has a density of 1 g/cc. A simple calculation shows that water content between 6 and 18 g/1g yields.84% to 2.52% water by volume respectively. Assuming a simplistic model of proportionality, and assuming that the dielectric constant of cardboard is 1 (air), we could estimate that the dielectric constant of cardboard varies between 1.66 and 2.97, respectively. Based on measurements we show below later, this simplistic model is not valid, and clearly the water is not behaving electrically the same way as it does in liquid form. Thus, we must obtain direct measurement of the dielectric constant of the cardboard in order to design to that material.

We found the dielectric properties of different materials by constructing a resonant rectangular microstrip antenna and measuring the S-parameters using a network analyzer. We start by estimating a value for dielectric constant and loss tangent of the paperboard. Using the assumed dielectric properties, we constructed a microstrip patch antenna, measured the impedance characteristics over a broad frequency range, then used data fitting and a method of moments simulation package to estimate the dielectric properties of the paperboard. The antenna design was fabricated using the paperboard as substrate as shown in Figure 1. Figure 1: Microstrip test apparatus for paperboard. The reverse side is laminated with a metal foil. Table III: Measured dielectric properties of paperboard. Water content (g/1g) ε r tan δ ~ 1.2.1 1.24.14 1 1.3.2 1.45.5 Five samples of paperboard of same dimensions were taken. The samples were dried completely, weighed, and stored in an air tight bag. After about 24 hours, they were weighed again to make sure there was no change in water content. Then we added water to obtain a 1g, and g of water per 1g of dry paperboard. The samples were kept for about 24 hours in an air tight bag to reach uniform moisture content throughout the paperboard. Finally, the dielectric constants of all the samples were measured using the microstrip method. The dielectric constant and loss tangent that we obtained are tabulated in Table III. Dielectric constant and loss tangent of paperboard with g/1g water content was 1.45 and.5 respectively. Dielectric constant and loss tangent of paperboard with no water content was 1.2 and.1 respectively. These specify the upper and lower bound of the dielectric properties of the paperboard respectively. For accuracy, we repeated the procedure a number of times until we were satisfied with the accuracy. Relative permittivity and loss tangent of paperboard at room temperatures conditions found using the above procedure were 1.24 and.14 respectively. Note that the measured moisture content at room temperature and approximately 55% relative humidity was g/1g, which differs from [8]. We can conclude that different paperboard behaves differently, and thus results from this paper should not be overly generalized. IV. SINGLE RESONANT ANTENNA DESIGN Once we know the dielectric properties of the substrate, we first investigated whether a single-resonant patch antenna such as [1] can be used to give the required performance. A. Design We began by attempting to design the antenna with roomtemperature conditions (i.e., εr = 1.24 and tan δ =.14). We chose the patch antenna to have a width of 4 mm, which can safely fit in a 2-inch label. As illustrated in Figure 2, we shaped the antenna so that the matching circuit could safely fit within the 4 mm width. We determined the length to be 128 mm. The matching circuit was constructed with 3 mm transmission lines with 3 mm spacing, and the matching circuit is placed in a 32 mm by 12 mm cavity. This yields a resonant frequency of about 932 MHz with a resonant impedance of 89 + j111 for g/1g paperboard. The resonant frequency was intentionally high and the impedance intentionally too large in order to obtain good performance over the range of dielectric constants. 5-5 -1-15 -2 88 89 9 91 92 Figure 2: Simulated antenna gain of single patch for various moisture levels. One novel aspect of the design in Figure 5 over [1] is that that [1] forces relatively high current densities through 1 mm transmission lines, while Figure 2 uses 3 mm transmission lines. Using wider lines yielded a measurable increase in performance. The matching circuit was designed to provide a good match to 65 j11 Ohms. For the performance criterion, we chose a gain of -5 dbi, which should be readable reliably to about 16 feet (about 5 meters), and found that this design gives about 44 MHz of bandwidth with that criterion. The simulated gain values are shown in Figure 3. When we varied the dielectric properties of the material, the simulation results showed the gain curves in Figure 2. While 1g/1g still gives acceptable performance, g/1g is completely unacceptable. Experimentation with 93 94 95 96 1

simulation indicates that about 15g/1g is enough to reduce the gain to unacceptable levels. Thus, the change in dielectric constant due to 9% relative humidity is sufficient to render this tag as below-adequate performance. 12 Real Impedance (Ohm) 1 8 6 4 2 88 89 9 91 92 Figure 3: Simulated input resistance of single patch for 93 94 95 96 1 Figure 5: Image of single patch design for paperboard. Assuming minimum transmit power obtained provides the minimum turn on power for the chip (a known quantity), we can find gain of the tag antenna using Friss transmission equation. What we obtain is a result that includes the product of the tag antenna gain (Gt) and the power transmission coefficient (τ). We call this product the realized gain. The results for the realized gain are shown in Figure 6 for g/1g. Note that for g/1g, we were not able to achieve a single read at any frequency at 1 meter. Imaginary Impedance (Ohm) 16 8 1 -.5-1 -1.5-2 -2.5-3 -3.5 88 89 9 Also, it was found that for higher values of loss tangent the port impedance reduced drastically which lead to poor impedance matching. Figures 3 and 4 show the simulated real and imaginary input impedance for the tag, which we found to have good agreement with measured impedance values. Note that the g/1g and 1g/1g show a good impedance match over a portion of the band, but g/1g has unacceptably low real impedance over the FCC band. B. Measurements 91 92 Figure 4: Simulated input reactance of single patch for Next, we constructed a rectangular tag, attached the RFID IC, and measured its performance. An image of the constructed tag is shown in Figure 5. Note that a metal foil was attached to the reverse side. We used an instrumented RFID reader in which we varied the frequency and power settings from 93 to 927 MHz in order to find the minimum power setting required to stimulate the tag. The tag was placed a fixed distance (1 meter) from the reader antenna. Note that the turn-on power measures both gain and impedance match. 93 94 95 96-4 93 96 99 912 915 Figure 6: Measured antenna gain of single patch antenna at g/1g. At g/1g, the tag was unreadable. C. Interpretations The tag shows excellent performance at g/1g, and we would expect similar performance at 1g/1g. This tag should be readable well over 2 feet away. However, with higher moisture contents, the tag becomes completely unreadable. The tag experiences very poor performance at the higher moisture levels both because the antenna is not resonant at the higher dielectric constants and because it experiences poor impedance matching that happens when the antenna is non-resonant and because there is a significant reduction in the resonant resistance due to the higher tan δ. From these results, we conclude that a suitable microstrip patch antenna is not suitable to address the variance in dielectric constant for paperboard. Traditional methods of making the antenna more broadband, such as making the tag wider or thicker, are not practical for this problem. 918 921 924 927 If we are to achieve good performance over a considerable

variation in the dielectric constant of the paperboard, we must develop new techniques to address this unique challenge. V. DUAL-RESONANT ANTENNA DESIGN In Section 4, we concluded that a single-resonant patch antenna with a single, balanced feed structure is not suitable to work suitably over the range of dielectric properties of paperboard. We noted two factors that caused problems: poor antenna efficiency due to non-resonant behavior, and poor impedance matching. Any solution will need to adequate address both of those challenges. Here, we propose the use of two resonant patch antennas. Unlike previous work of [1], we use a single, edge feed off of the non-radiating edge of each antenna. Each antenna is designed to resonate at a significantly different frequency. We note that the non-resonant impedance of a patch antenna is small in magnitude, and we use that small-magnitude impedance as an approximation for a reference to ground. A simple transmission line length is used to increase the reactive component of the impedance in order to provide good impedance matching. By using this approach, we are able to not only match to two different frequencies, but to two different impedances as well. A. The Design For comparison purposes, we choose width of the antenna form factor to be the same 4 mm as in section 4. The two patches are offset in terms of resonant frequency, one has a resonant frequency of about 9 MHz and the other patch has a resonant frequency of about 973 MHz. The lengths of these patches were 148 mm and 136 mm respectively. The patches were 15mm wide and were placed 1mm apart. A rectangular patch antenna fed offset from center of the non-radiating edge has a small reactive impedance. This reactive impedance can be increased as required for matching by changing length and width of the transmission line. Again, we make use of the fact that at frequencies in which one patch is resonating; the other patch has very low impedance and acts as a virtual ground, eliminating the need for any via. Adjusting lengths and offsets of transmission lines feeding the two patches, we can control the impedance for both the resonances separately. This proves to be a great advantage because for higher values of water content, both the relative permittivity and loss tangent goes up resulting in the lower resonant frequencies and a decrease in Q. Hence we can adjust the impedance of the resonances so that one resonance maintains a good match and a good gain for lower values of relative permittivity and the other resonance maintains a good match and reasonable gain for higher values of relative permittivity and loss tangent. A dual patch antenna was designed taking all these factors into consideration. The design with dimensions is shown in Figure 7. Figure 7: Dual-resonant antenna design with measurements. Next, we present simulation results for the antenna. Figure 8 shows the simulated antenna gain for varying moisture rates. We note that the -5 dbi gain bandwidth at g/1g is approximately 81 MHz, almost twice as large as the single rectangular patch. 5-5 -1-15 -2 88 895 91 9 94 Figure 8: Simulated antenna gain of dual patch for various moisture levels. Real Impedance (Ohm) 35 3 2 15 1 5 88 895 91 9 94 Figure 9: Simulated input resistance of dual patch for Next, we show the simulated real and imaginary input impedance for varying moisture contents in Figures 9 and 1. Note how the tag is able to maintain excellent impedance matching (to 65 j11ω) over the entire band for all the tested moisture contents. The second, large peak at 988 MHz for dry paperboard is the same peak that gives an almost ideal peak of 83Ω at 919 MHz with g/1g. The 955 955 97 97 985 985 1 1 1 1

reactive impedance gives excellent results except for the higher frequencies at g/1g, which will begin to degrade performance significantly. Imaginary Impedance (Ohm) 24 16 8-8 88 895 91 From the above graph we observe that up to a water content of g/1g the antenna maintains a gain over -8 dbi over all dielectric conditions. B. Measurement 9 94 Figure 1: Simulated input reactance of dual patch for As before, the design was etched into inlays on copper clad polyester. Figure 11 illustrates one of the tags used for testing. The inlays were then fabricated into tags using paperboard at room temperature and humidity conditions (g/1g) and with g/1g. The tag was tested using the same process described in Section 4.2. The performance results are shown in Figure 12. 955 97 985 1 1 C. Interpretations The measured results show generally good matching with simulated results, and validate this approach. At g/1g, the tag showed better-than-expected effective gain. It is possible that the moisture content was not fixed for the duration of the experiment, for example, and the tag took on excess moisture. It is also very possible that the paperboard samples were not uniform. At g/1g, the measured effective gain matches closely with simulation results. We expected to see a peak gain of about -8 dbi at lower frequencies with a decrease in performance at the higher frequencies due to poor impedance matching. The measured results show exactly this. As a basis for comparison, a tag designed for optimal performance at g/1g with a similar form factor as the tag in Section 4 will give a gain of about -6 dbi. This poor performance is primarily due the large loss component of the dielectric constant. VI. CONCLUSION In this paper, we considered the applicability of a microstrip patch antenna and its suitability for providing a consistently high level of performance regardless of the contents of the container, and with minimal impact on current processes. The proposed approach requires (1) the addition of a foil ground plane to the inside of the paperboard container, and replacing commodity RFID antennas (typically dipoles) with a planar microstrip antenna. We first characterized the dielectric properties of paperboard under different moisture conditions. We found that the dielectric constant may vary between 1.2 and 1.45, and the loss tangent between.1 and.5. Figure 11: Image of dual-patch design for paperboard. Our first attempt of using a single microstrip patch antenna based on [1] performed well at the lower moisture contents, but completely fails at higher moisture contents. -2-4 -6-8 -1-12 We developed a novel, dual-resonant structure that provides both frequency and impedance diversity so that we are able to achieve both good gain and good impedance over the range of dielectric constants. We present extensive simulation data as well as experimental results that show good consistency between the two. -14-16 93 96 99 912 915 918 921 924 927 Figure 12: Measured antenna gain of dual-patch antenna. REFERENCES [1] M. Eunni, M. Sivakumar, D. Deavours; A Novel Planar Microstrip UHF RFID. The 4th International Conference on Computing, Communications and Control Technologies, CCCT 26. [2] Ukkonen, L., Engles, D., Sydanheimo, L., and Kivikoski, M.; Planar Wire-Type Inverted-F RFID Tag Antenna Mountable on Metallic

Objects. Antennas and Propagation Society International Symposium 24 IEEE, vol.1, pp 11-14. [3] Kwon, H., and Lee, B.; 'Compact Slotted Inverted-F RFID Tag Mountable on Metallic Objects'. IEE Electrinic Letters, November Vol.1 pp 138-131. [4] Son, H.W., Yeo, J., Choi, G.Y., and Pyo, C.S.; A Low Cost, Wideband Antenna for Passive RFID Tags Mountable on Metallic Surfaces. Antennas and Propagation Society International Symposium 26, IEEE, pp 119-122. [5] Byunggil Yu; Sung-Joo Kim; Byungwoon Jung; Harackiewicz, F.J.; Myun-Joo Park; Byungje Lee : Balanced RFID Tag Antenna Mountable on Metallic Plates. Antennas and Propagation Society International Symposium 26, IEEE, pp 3237-324. [6] J. Marcondes: Corrugated Fiberboard in Modified Atmospheres: Moisture Sorption/ Desorption and Shock Cushioning. Packaging Technology and Science, Volume 9, Issue 2, Pg 55-12, December 1998. [7] K. Kupfer (Ed.): Electromagnetic Aquametry: Electromagnetic Wave Interaction with Water and Moist Substances. Springer Berlin Heidelberg New York,. [8] J. Marcondes: Cushioning Properties of Corrugated Fibreboard and Effects of Moisture Content. Transactions of ASAE 1992; 35(6): 1949-1953.