Improving the Near-Metal Performance of UHF RFID Tags

Transcription

1 Improving the Near-Metal Performance of UHF RFID Tags Daniel D. Deavours Information an Telecommunications Technology Center University of Kansas, Lawrence, KS Abstract It is well-known that UHF RFID tag performance egraes when place near metal. While the mechanisms for how ipole performance egraes near metal is known, it is generally not known how the parameters of the T-match change in the presence of metal, an what, if anything, can be one to improve near-metal performance. In this paper, we evelop a set of expressions that escribe the affect of antenna parameters on the input reactance of the antenna near metal, an a set of esign principles that can be use to minimize the near-metal impeance mismatch. We conclue by emonstrating these principles with a simple antenna moel yiels a 12.7 meter free-space rea istance, an a rea istance of 7 meters when separate by a 3.2 mm HDPE foam ielectric spacer from a large groun plane. Inex Terms Antennas, RFID, Antenna fees, Microstrip antennas, Impeance matching. I. INTRODUCTION It has long been known that UHF RFID tag performance egraes significantly when place near metal [1], [2]. While numerous microstrip-base RFID tags have been evelope (e.g., [3] an references therein), they ten to be too costly for use in supply chain an other cost-sensitive applications where the near-metal problem remains problematic. It is wellknown that raiating resistance of the ipole ecreases when close to metal [4], [5], but most RFID tags use some kin of T-match [6]. The T-match as an extra egree of freeom or complexity that can, potentially, improve the near-metal performance of a ipole. To ate, there has been no systematic escription of the mechanisms by which the T-match ipole antenna performance egraes near metal (primarily through changes in impeance), nor what can be one (if anything) to mitigate this problem. In this paper, examine the parameters of a T-match RFID tag an how those parameters are affecte by the presence of a groun plane. Our results show that, for example, ifferences of over 1 B in near-metal performance can be observe through very minor changes in free-space performance of essentially the same tag, an that by careful planning, one can achieve more than five-fol increases in near-metal rea istance of stanar commercial RFID tags. Inexpensive supply chain tags are often use as low-cost asset tags by incluing a thin foam spacer, often about 4.8 mm thick. This is commercially interesting because of the ability to manufacture such tags in high spees an at low cost. While the near-metal performance of such tags have been extensively measure an ocumente, there has been little effort to esign tags for improve performance in such This work was supporte by the Information an Telecommunications Technology Center at the University of Kansas. an environment, primarily because there is little unerstaning of the mechanisms that govern the near-metal performance of RFID tags. In this paper, we perform a etaile investigation of how the ipole antenna using a T-match performs near metal. To o this, we rely on the Ua moel of the T-match [7], [4], an the embee-t match, a kin of T-match that is especially simple an amenable to analysis, which we present in more etail in Section II. We present in Section III a simple experiment in which we vary the length of the antenna an observe how the near-metal performance changes. We fin that there is a wie variation in performance, an that there are two lengths of antenna that give near-maximum performance. Next, in Section IV, we evelop a set of analytical moels that escribe the near-metal behavior of the various components of the T-match ipole. Using this, we are able to analytically escribe the impeance behavior of the tag, incluing making several preictions that are valiate an are useful for eveloping simple esign principles. In Section V, we use this new knowlege to esign a tag that presents a conjugate impeance match in air an a peak -3 Bi near-metal performance using a 3.2 mm HDPE foam ielectric spacer. We summarize an conclue our finings in Section VI While the results of this work oes not eliminate the nearmetal problem, it oes for the first time show the factors that are funamentally responsible for the problem, an techniques that can be use to ramatically improve near-metal performance of RFID tags. II. BACKGROUND In this section, we give an overview of the essential backgroun material that is use for the remainer of this paper. A. Impeance matching It is well-known that the maximum power is elivere to a complex loa when the loa is the complex conjugate of the source impeance. This is particularly useful with RFID tags when matching the antenna an IC impeances. Let Z in be the antenna impeance an Z ic be the input impeance to the antenna, where Z in = R in + jx in an Z ic = R ic + jx ic. Generally, the power transfer efficiency can be expresse as follows [8] τ = 4R inr ic Z in + Z ic 2. The IC impeance tens to be reactive because of the ioe (or equivalent) rectifier RF front en, which tens to provie a significant parallel capacitance. A typical impeance might

2 Z in (1+ ) 2 :1 P2 1mm W2 W P1 S L W1 Z Z c Fig. 1. The embee T-match with the input elta-gap port (P 1 ) an a secon faux elta-gap port use for analysis (P 2 ). Fig. 2. Circuit moel for Z in using the Ua analysis. be Z ic = 1 j15. Most UHF RFID antennas are some type of ipole antenna. Because of economic pressures, the antennas ten to be both physically an electrically short, an so also have a capacitive input impeance. To to transfer the capacitive antenna impeance to one that is inuctive, an to ajust the resistance to provie a conjugate impeance to the IC, a T-match is often use. For this paper, we assume that a conjugate IC impeance is the esire antenna impeance, an that τ is the esire metric to optimize. We are aware that other impeance matching schemes may be more esirable in some circumstances, for example, impeance matching to maximize the backscatter signal strength [9], [1], an in those cases an analysis similar to the one given here coul be performe. B. T-match an embee T-match To transform the capacitive antenna impeance to inuctive, an to ajust the input resistance if necessary, practitioners commonly use some variant of the T-match [7], [4]. In this paper, we will use a special form of the T-match that we call the embee T-match, shown in Fig. 1. The embee T- match is analogous to the stanar T-match except for the T- section is inscribe into the printe rectangular ipole, instea of solering on external wire sections, as was historically common. The embee-t has been use commercially in such tags as the Alien M tag (now the G tag [11]) an the Avery Dennison AD-21 [12]. For our purposes, it has a number of avantages over the more traitional an common T-match antennas, which are summarize below. The analysis of the T-match is well-known an first propose by Ua [7]. This involves ecomposing the antenna into two moes: a ifferential moe acting as a short-circuite transmission line, an a common moe that operates as a common ipole. The two moes are then combine to etermine the input impeance. The net result is the common moe impeance Z c scale by a factor (1 + α) 2, seen in parallel with the ifferential moe impeance Z. The splitting factor α can be viewe as the ratio of the current flowing through Port 2 to Port 1 when the antenna is riven in the common moe. Equivalently, it can be viewe as the ratio of voltage source in Port 1 to Port 2 when the antenna is riven in the ifferential (transmission line) moe. For etails, see e.g. [4]. Let Z c be the impeance seen in the common moe. Let Z /2 be the input impeance seen in one half of the antenna when riven in the ifferential moe. Note that the ifferential moe resembles a shorte transmission line of length S/2. Thus, we can estimate Z 2 = jz tan ks 2 with Z being the characteristic impeance of the coplanar strips an k the guie wave number. Finally, the two moes can be combine to calculate the input impeance. Z in = (1 + α)2 Z c Z (1 + α) 2 Z c + Z (1) The equivalent circuit moel is shown in Fig. 2. Note that Z c, α, an Z can be compute irectly from the two-port parameters when Port 2 is a elta gap (c.f. Fig. 1). Z c = Z 11Z 22 Z 2 12 Z 11 + Z 12 2Z 12 α = Z 11 Z 21 Z 22 Z 12 Z = Z 11 + Z 22 2Z 12 These equations can be useful when using simulation moels in orer to irectly evaluate Z c, α, an Z from a two-port moel. We chose to use the embee-t antenna for this analysis for several reasons. The justifications for each of these are outsie of the scope of this paper, but are summarize below. The small number of parameters (4) that yiel a large, representative esign space. The splitting factor α is nearly always real, while with popular commercial RFID antennas, α can be complex an sometimes with a large phase angle. While the circuit moel in Fig. 2 is still correct, analysis with a complex α is significantly more challenging. The embee-t makes efficient use of space, an an tens to have significantly less conuctive losses near metal than narrow strip ipole. The common moe impeance Z c is almost completely etermine by L an W, an is virtually unaffecte by W 1 an S. The splitting factor α is almost completely etermine by the ratio of W 1 to W 2, an is virtually unaffecte by L an S. The ifferential moe impeance Z is a function of the characteristic impeance of the transmission line, which is solely a function of W 1 an W 2, an the length of the transmission line S. Thus, the geometry of the antenna has a strong an wellefine relationship with the equivalent circuit moel, an thus ieally suite for analysis. C. Microstrip an near-metal performance Of course, there are numerous asset tags that have been evelope for UHF RFID tags. References in Rao et. al. [3]

3 TABLE I VALUES FOR EMBEDDED-T MATCH ANTENNA VARYING L. 4 6 W 1 S W 1 S τ M (B) contain an excellent survey of papers an proucts that aress this particular problem. Note that these antennas are typically microstrip-base antennas, an thus incorporate a groun plane an a ielectric substrate. This is istinct from the problem we are aressing here; we are consiering a simple T-match ipole antenna that provies a conjugate impeance match in air, an seeking to minimize impeance mismatch losses ue to the introuction of a groun plane. Ukkonen et. al. [13] presents a large, patch-like antenna for use with objects containing metallic foil, but the antenna is too large to be practical. Mohamme et. al. [14] presents the kin of antenna we escribe here, but gives no explanation as to the mechanisms involve or any esign principles for systematically arriving at a suitable solution. III. PEDAGOGICAL EXPERIMENT A. Description of Experiment In this section, we perform a simple experiment that illustrates some of the important factors for goo near-metal performance. We use the simple embee-t antenna esign for all these experiments. Recall that the embee-t has four inepenent variables: L, W, W 1, an S. Let us (arbitrarily) fix W = 2 mm, which may be slightly large by moern commercial stanars, but comparable to the Alien G tag [11], for example. Since we fix W, then Z c is now solely a function of L. The experiment is this: vary L from 85 mm to 17 mm in steps of 5mm. Then, for each L, we select W 1 an S that provies a conjugate impeance match for the antenna in free space for Z ic = 12 j133 Ohms [15]. For moeling simplicity, we assume that the antenna is being fe by a elta gap of with.2 mm, an the conuctor has zero thickness. (We present a more realistic example in Section V.) The values for W 1 an S are summarize in Tab. I. These values were foun using a MoM solver with a suitably fine mesh size, an were chosen so that the return loss at 915 MHz was less than 2 B (i.e., better than 99% efficient). B. Near-metal performance varying L Let us take the same set of antennas escribe in Tab. I an place them 3 mm from an infinite groun plane an observe how the input impeance changes with L. For this, we use a FEM simulation package, which we foun gives more accurate Fig. 3. Power transfer efficiency of antennas as a function of L. solutions than the MoM solver for this scenario. We assume a copper conuctor in both the antenna an groun plane to moel conuctive losses, but further assume zero thickness (in both MOM an FEM moels) to reuce the numerical complexity. We plot the power transfer efficiency τ, in B, as a function of L, in Fig. 3. Fig. 3 shows some remarkable features. Clearly, there are choices of L that yiel better near-metal performance than others. We see that there are two lengths that give maximum performance, an that there is as much as 15 B of ifference between the best an worst choices of L. For the remainer of this section, we give some more ata that gives more insight into the in-air an near-metal behavior of the embee-t antenna. In Section IV we evelop a rigorous theoretical basis that escribes an preicts this behavior, which leas to simple esign methos. C. Comparison of free-space an near-metal antenna parameters To gain a better unerstaning of how the antennas are behaving, recall the Ua moel presente in Section II. There, we showe that we can obtain the input impeance as a function of the common moe impeance Z c, the splitting factor α, an the ifferential moe impeance Z. Using a two-port moel, we are able to calculate those values for varying L both in air an near metal. The common moe impeance for the antenna is given in Fig. 4 using a Smith chart normalize to 5 Ohms. For the antenna both in air an near metal, the impeance moves clockwise with increasing L, an a marker is place at every 5 mm change in length. Because α is almost completely real an Z is almost completely reactive, we present the real value of α an the reactance of Z in Fig. 5 an 6 respectively. From the figures, we can make the following observations. The common moe resistance on metal is much lass than the resistance in air, as expecte. The common moe reactance iffers substantially at lower frequencies, but is nearly ientical when the antenna is near resonant-length. The frequencies of similar reactance may change with W. The splitting factor α is relatively stable in air an near metal, an iverges milly at the largest values of α.

4 ΛΥ ΗΩ Ο Fig. 4. Antenna common moe impeance (Z c) in air an near metal as L varies from 85 to 17 mm. Fig. 5. α Reactance (Ohms) Antenna splitting factor α in air an near metal X A X M.77 X A α A α M The ifferential moe reactance is scale by a nearly constant factor of approximately.77 when place near metal. This is expecte since the characteristic impeance of the transmission line ecreases with the ae capacitance of the groun plane. Since the impeance is also a function of the antenna s environment, we inclue a superscript A or M to enote in-air or near-metal. E.g., Zin A an ZM in is the input impeance in air an near metal, respectively. We will use the above observations when eveloping a theory. Specifically, we can evelop the following approximation. Z M = βz A, (2) where β =.77. To valiate, we plotte βz A in Fig. 6, which shows goo agreement with the approximation. The splitting factor α is stable, though iverges for large α. Since we are not aware of any simple explanation for the change in splitting factor, an for simplicity of analysis, we make the assumption that α is inepenent of its environment, i.e., is the same in air an near metal. Thus, we approximate α M = α A. (3) While this is obviously not accurate for large α, it is suitable for our current nees for eveloping a theory. IV. THEORY Since Z c is always scale by (1 + α) 2, an because with the embee-t antenna α is real-value, we can temporarily efine a new term, the antenna impeance Z a = (1 + α) 2 Z c. We can then split Z in into the real an imaginary values. R in = (R ar X a X )(R a + R ) (R a + R ) 2 + (X a + X ) 2 + (R ax + R X a )(X a + X ) (R a + R ) 2 + (X a + X ) 2 X in = (R ax + R X a )(R a + R ) (R a + R ) 2 + (X a + X ) 2 (R ar X a X )(X a + X ) (R a + R ) 2 + (X a + X ) 2 Of particular interest is X in. We know that Rin M is ifficult to control an will generally be quite small ue to the small raiating resistance near metal. However, it appears as if one can control X in to a significant egree. In general, for a fixe small R in, τ M will be maximize when Xin M = X ic. Notice that R A is negligible (i.e., Z can be approximate as a lossless transmission line), so we can simplify (5): (4) (5) X A in = X (R A a ) 2 + X A a (X A a + X A ) (R A a ) 2 + (X A a + X A )2. (6) For the tag near metal, both Ra M an R M are negligible (so long as X a X Ra M,R M ), so can simplify (5): Fig Antenna ifferential moe reactance (X ) in air an near metal. X M c X M in = XM a X M X M a + X M. (7) Next, we evelop analytical approximations to Rc A, Xc A, an as a function of L in units of millimeters. We use a quartic

5 R c A simulate R c A analytic X c A simulate X c A analytic X c M simulate X c M analytic Impeance (Ohms) Fig. 7. Comparison of numerical an analytical approximations for R A c, X A c, an X M c. (9), we can further express α M as a function of Z ic an L. Specifically, (1 + α M ) 2 = R L 4 Z ic 2 R ic (R 2 L8 + (m A L + b A ) 2 ) (15) Earlier in (1) we gave Xc M in terms of constants an L. Finally, recall (2) where we mae the approximation X M = βx A. For a conjugate match in air, we use (13), which we note is a function ofz c an Z ic. Again, we can reuce Z c into functions of constants an L by (8) an (9). Thus, X M = βx A R L 4 Z ic 2 = β R L 4 X ic + R ic (m A L + b A ) (16) From (14) an substituting using (15), (1), an (16), we can now express X M in entirely as a polynomial function of R ic, X ic, L, an constants, an is given in here for completeness. to approximate R A c, an linear regression to approximate X c as a line. Generally, we foun the following. R A c (L) = R L 4 (8) X A c (L) = m A L + b A (9) X M c (L) = m M L + b M (1) where it was foun for this particular example where W = 2 mm an for L between 85 an 17 mm that R = m A = b A = m M =.7358 b M = (11) To valiate, we plot the estimates against the values foun through the numerical methos in Fig. 7. We can see that these analytical approximations are quite accurate over a broa range of L. Recall again (1). Assuming Z c is given, i.e., fixe an outsie of our control, the process of esigning the matching circuit is to prouce the T-match with α an Z such that Z in = Zic. If we set (1) equal to Z ic an let Z c be constant, an if we assume α is real an Z is reactive, we can solve for the require α an Z. We use the subscript r to enote the require value to achieve Z in = Zic. R c Z ic α r = 2 R ic Z c 2 1 (12) R c Z ic 2 Z,r = j (13) R c X ic + R ic X c Finally, we can rewrite Xin M for a T-match ipole that has been matche to a conjugate impeance of the IC in air. X M in = (1 + αm ) 2 X M c X M (1 + α M ) 2 X M a + X M. (14) From (3) we assume that α M = α A = αr A, an from (12) we see that αr A is a function of Zc A an Z ic. Using (8) an Xin M A = R ic (B C) DE A =L 8 R β Z ic 2 (m M L + b M ) L 4 R β Z ic 2 B = R ic (m A L + b A ) + L 4 R X ic C = L4 R Z ic 2 (m M L + b M ) R ic ((m A L + b A ) 2 + L 8 R 2) D =(m A L + b A ) 2 + L 8 R 2 E =R ic (m A L + b A ) + L 4 R X ic A. Affect of L on Xin M Next, we valiate our formulation for preicting Xin M as a function of L. Recall in Section III-B we performe the experiment of eveloping free-space impeance-matche ipole antennas for L = 85 to 17 mm. We can compare the results Xin M of those antennas foun through FEM an the results of our analytical approximations, which we present in Fig. 8. The analytical moel tracks the simulate results quite well, which gives us confience in the moel. The exact values eviate slightly, but the analytical moel oes capture the values, trens, an poles. The analytical moel also successfully preicts two solutions in which Xin m = X ic, which are close to the best near-metal performance. One avantage that the analytical moel gives us is the ability to ifferentiate the expression. Specifically, we can examine how small changes in L affect Xin M, assuming that for small perturbations in L that W 1 an S are also change so that Zin A = Z ic. Numerical ifferentiation woul be nearly impossible to perform because of the ifficulty in fining exact solution at each L an the finite numerical accuracy inherent to numerical tools. We can therefore plot L XM in an plot this value as a function of L, which is given in Fig. 9. While the trens are obvious from inspecting Fig. 8, the values may not be. At L = 15 mm, every mm increase in L will ecrease Xin M by 3.7 Ohms, an at L = 14 mm, every increase in L will increase Xin M by 6.1 Ohms. To valiate, consier L = 13 mm. We mae small variations for L, then for each L selecte W 1 an S to provie conjugate impeance match in air. Then for each of those

6 6 5 X in M simulate X in M analytic TABLE II Z A in, ZM in, AND τm FOR SMALL VARIATIONS IN L GIVEN W = 2 MM. Reactance (Ohms) L W 1 S R A in X A in R M in X M in τ M Fig. 8. metal. Simulate vs. analytical input reactance of antennas place near /R ic X in m /L X in m Fig. 1. X R M ic in as a function of L. 8 Fig L XM in as a function of L. antennas, we place the antenna near a large groun plane an observe Xin M. The results are summarize in Tab. II. Analytically, we compute that L XM in at L = 13 mm is Ohms / mm. Observing Tab. II, we see that using very coarse numerical ifferentiation we woul estimate the value at This shows very goo agreement between theory an experimental observation. Errors are likely from an inability to provie an exact impeance match, an from error from the numerical solvers, as well as approximations in the analytical moel. It is also reassuring to see that the largest τ M occurs when L = 14 mm an Xin M is 135 Ohms, which is very close to X ic = 133 Ohms. However, Rin M oes change with L, so the optimum may not always be exactly where Xin M = X ic. B. Affect of Rin A on XM in Assume Z ic = j Suppose instea of matching R in to Ohms we instea provie an input resistance of 11.5 or 12.2 Ohms. Notice that R in can vary between almost 1 an 14 Ohms (if X in = X ic ) an still maintain a 2 B return loss. Thus, we can change R in by a small amount by ajusting α an change Z appropriately to maintain X in = X ic without any appreciable affect on freespace performance. Next, we will examine how those changes in Rin A will affect XM in. We can again rely on the analytical moel to calculate the affects. The way we constructe the expressions above, all that is necessary is to ifferentiate with respect to R ic, specifically, R ic Xin M. Again, one can ifferentiate the expression an plot the value of the erivative as a function of L. We o this an plot the results graphically in Fig. 1. The results inicate that at L = 15 mm, every 1 Ohm increase in Rin A will increase Xin M by 6.9 Ohms, an at L = 14 mm, every 1 Ohm increase in Rin A will increase XM in by 7.5 Ohms. This again gives us new insight into how we can control Xin M to a moest amount. To valiate, let us consier again the embee-t antenna escribe earlier. As before, let W = 2 mm, an now let R ic Xin M L = 1 mm. At that length, = 12. Ohms / Ohm. We will vary W 1 over a suitable range (effectively varying R in in air), an for each W 1 we will select S so that Xin A = X ic. Then, we will place the same set of antennas 3 mm from a large groun plane an using a FEM tool, observe how the input impeance changes. Tab. III summarizes the results. We see that Rin A changes by 3.84 Ohms an that XA in varies by 72.2 Ohms, or about 18.8 Ohms per Ohm. While the two o not match as well as one woul like, the ifference is likely ue to the sensitivity to small changes in L, as shown in Fig. 1. The analytical moel successfully preicts the sign, magnitue, poles, an the shape of the curve, so it is quite useful for those purposes. It shoul be emphasize that all of these antenna geometries given in Tab. III have the same length, with, an a return loss in air of more than 2 B, but the ifference in performance near metal varies by 1 B. Clearly then, out of a set of nearly-

7 TABLE III Z A in, ZM in, AND τm FOR VARYING W 1 GIVEN L = 1 MM, W = 2 MM. TABLE IV Zin A AND ZM in FOR L = 1 MM, W = 2 MM, AND ǫr = W 1 S R A in X A in R M in X M in τ M W 1 S R A in X A in R M in X M in τ M equivalent T-match ipoles in air, some will perform much better near metal than others. C. Design methos Notice that the erivative in Fig. 1 is everywhere positive (for the range of L we consier). Also, by inspecting Fig. 8, we can see that L XM in is negative near L = 15 mm an positive with L = 14 mm. This gives us important insights into the esign process, which is summarize here. For L 15 mm, increasing the input resistance in air will increase the input reactance near metal, an increasing L will ecrease the reactance near metal. For L 14 mm, increasing the input resistance in air will increase the input reactance near metal, an increasing L will increase the reactance near metal. Thus, Rin A an XM in are always positively correlate, but L an Xin M are negatively correlate for short L an positively correlate for longer L. We can use this information to guie a heuristic esign process for selecting L an, if necessary, Rin A, to obtain the esire XM in. V. DESIGN EXAMPLE In this section, we apply the theory to a slightly ifferent setting that may be of some practical interest. In the inustry, foam-attache tags, or FAT tags, are common low-cost substitutes for asset tags, which are tags that are place on metal items. These FAT tags are normal RFID tags that have been esigne for free-space operation. While inexpensive, the challenge with using FAT tags are their relatively low performance compare to microstrip-base tags, an they ten to be less rugge than asset tags. Here, we will emonstrate the application of the theory to a practical FAT tag that yiels improve on-metal performance. The FAT tag inclues a thin foam separator commonly mae of either high-ensity polyethylene (HDPE) foam, which is a close-cell foam, or a polyurethane foam, which is an opencell foam. Open-cell foams have the avantage of being easily compressible, but they ten to absorb water an therefore change permittivity with the environment. HDPE foam has the avantage of being close-cell, inexpensive, an very hyrophobic (oes not absorb water). For our purposes here, ielectric stability is critical, an thus we chose to work with a HDPE foam. The foam that we have available was measure to have a ielectric constant ǫ r = 1.95 an a negligible issipation factor. The foam is 3.2 mm thick, an inclues a pressure-sensitive ahesive. To estimate the affect of the ielectric constant, we can roughly estimate an effective ielectric constant ǫ eff = ǫr+1 2, or The change in resonant length is approximately or.977. If the ieal length of the antenna in air was 15 mm, then equivalent length on foam is 12.5 mm. For convenience, we chose to start the esign process at 1 mm. We begin by creating a simulation moel the embee- T antenna with a 25 micron copper foil on large HDPE foam sheet. For this example we use an IC with packaging ǫ.5 eff parasitics that we estimate has Z ic = 12 j133 Ohms [15]. Tab. IV shows the imensions of the antenna that were teste in air an on a 3 mm 2 copper groun plane. The simulations show that the best performance is achieve with the imensions L = 1 mm, W = 2 mm, W 1 = 1.5 mm, an S = 3.5 mm. We explore lengths of 98 mm an 12 mm an observe only a few tenths of a B improvement at best. Because there was little ifference, we chose to experiment with W 1 = 9.5 an S = 31.7 because it gave the best conjugate impeance match in air. This resulte in a 6.9 B loss in power transfer efficiency near metal. Note that this is almost 3 B worse than the results shown in the previous section, which is likely ue to the ielectric constant of the substrate, an because we are using a real fee an not a elta gap. We constructe the tag, which is shown in Fig. 11. From experimentation, we foun that we neee to reuce S to 31.1 mm in orer to more closely match simulation results. The tag was then teste with a commoity RFID reaer by placing the tag two meters from the reaer s transmit antenna in an ieal orientation. The reaer transmit antenna an tag were polarization-matche. We varie the frequency an power setting of the reaer an plotte the results of the tag in air an when attache to a 3 mm 2 copper groun plane. To compute the gain an efficiency numbers, we measure the tag in air an assume the peak performance correspone to 2 Bi. (This was measurably better than the Alien AL- 954 Squiggle tag [16], which uses the same IC, so we believe that this is a reasonable assumption.) FEM simulation results inicate a 7.5 Bi in irectivity. The IC we teste ha an estimate turn-on power of -14 Bm [15], so the maximum free-space rea istance for a 2 Bi antenna is

8 Fig. 11. Picture of the teste RFID tag with 3.2 mm HDPE foam spacer mounte on the groun plane. B Gain Efficiency this paper, we present empirical ata that emonstrates that the electrical length of the antenna is a free variable that can be use to moify the near-metal antenna impeance, specifically the reactance, in orer to improve the near-metal impeance match. Next, we evelope analytical moels that escribe why this is so. These yiele expressions an values that were experimentally valiate, but more importantly were able to accurately preict trens an relationships between antenna variables an near-metal tag performance. The algebraic expression allows us to mathematically manipulate the moel, for example, to take a partial erivative, to gain insight an to evelop simple esign methoologies. Finally, we applie the new information to eveloping a realistic RFID FAT tag. We constructe a simple 1 mm by 2 mm RFID tag that provies better than 99% efficiency in air for a 12.7 meter free-space rea istance. The same antenna provies a -3 Bi peak performance near metal, which correlates to a 7 meter on-metal rea istance. This is about 14 B better near-metal performance than a popular commercial RFID tag using the same IC F (MHz) Fig. 12. Results of the performance test of the RFID tag on a 3 mm 2 groun plane meters (41.6 feet), which was verifie experimentally. Thus, from measurements, we can estimate the effective gain (the prouct of broasie irectivity, raiation efficiency, an power transfer efficiency). The results are given in Fig. 12. We observe the peak performance of the tag on the groun plane to be Bi. When place on the copper groun plane, the antenna gives a maximum rea istance of 7.3 meters (24.1 feet). The peak measure efficiency (incluing raiating an power transfer efficiency) is B. The simulation result inicates a -6.9 B loss in power transfer efficiency an -4.2 B loss in raiating efficiency for a total loss of B, or about 1 B worse than measure. Thus, simulation an measure results show goo agreement. The same experiment was performe with the ALR-954 [16] RFID tag, which is a 94 by 8 mm tag commonly use in the supply chain. We place the ALR-954 on a 3.2 mm HDPE foam on a 3 mm 2 groun plane an measure the tag s performance. The peak effective gain for the ALR-954 on the metal environment was -17 Bi, or 14 B less than the antenna shown in Fig. 11. We can conclue that with careful esign, one can increase the near-metal performance by more than an orer of magnitue, resulting in a five-fol increase in rea istance. VI. CONCLUSIONS Is generally well-known that RFID tag performance egraes significantly near metal, an it is somewhat known that some antenna geometries are better-performing near metal than others, but it is not generally known why this is the case. In REFERENCES [1] D. M. Dobkin an S. Weigan, Environmental effects on RFID tag antennas, in IEEE MTT-S International Microwave Symposium, Long Beach, CA, Jun. 25, pp [2] S. R. Aroor an D. D. Deavours, Evaluation of the state of passive UHF RFID: An experimental approach, IEEE Systems Journal, vol. 1, no. 2, pp , Dec. 27. [3] K. V. S. Rao, S. F. Lam, an P. V. Nikitin, Wieban metal mount UHF RFID tag, in Proc. IEEE Antennas an Propagation Symposium, San Diego, CA, Jun. 28. [4] C. A. Balanis, Antenna Theory Analysis an Design. New Jersey: Wiley, 25. [5] D. B. Miron, Small Antenna Design. Burlington, MA: Newnes, 26. [6] G. Marrocco, The art of UHF RFID antenna esign: impeancematching an size-reuction techniques, IEEE Antennas an Propagation Magazine, vol. 5, no. 1, pp , Feb. 28. [7] S. Ua, Yagi-Ua antenna. Tohoku University: Research Institute of Electrical Communication, [8] K. V. S. Rao, P. V. Nikitin, an S. F. Lam, Antenna esign for UHF RFID tags: A review an a practical application, IEEE Transactions an Propagation, vol. 53, no. 12, pp , Dec. 25. [9] J. C. Bolomey an F. Gariol, Optimization of passive RFID tag antennas, in Antennas an Propagation Socity International Symposium, San Diego, CA, USA, Jul. 28. [1] Y. Xi, S. Kwon, H. Kim, H. Cho, M. Kim, S. Jung, C.-S. Park, J. Kim, an Y. Yang, Optimum ASK moulation scheme for passive RFID tags uner antenna mismatch conitions, IEEE Trans. Microwave Theory an Techniques, vol. 57, no. 1, pp , Oct. 29. [11] Alien Technology, Alien-9654 G inlay, com/ocs/proucts/ds ALN-9654G.pf, Aug. 29. [12] Avery Dennison, Avery Dennison AD-21 RFID inlays, Prouct Brochure, Oct. 25. [13] L. Ukkonen, D. Engels, L. Syanheimo, an M. Kivikoski, Fole microstrip patch antenna for RFID tagging of objects containing metallic foil, in Proc. IEEE Antennas an Propagation Society International Symposium, Washington, DC, 25, pp [14] N. A. Mohamme, M. Sivakumar, an D. D. Deavours, An RFID tag capable of free-space an on-metal operation, in Proc. IEEE Raio an Wireless Symposium, San Diego, CA, Jan. 29. [15] Alien Technology, Higgs-2 prouct overview, Jul. 28. [16], Alien-954 prouct overview, ocs/proucts/ds ALN 954 Squiggle.pf, Aug. 28.