Inventory management systems now rely on passive radio-frequency- identification (RFID) technology for automated identification of products in real time. For many applications, the use of RFID provides an acceptable return on investment. These systems must capture information on all present inventory in real time, requiring an RFID system to read 100 percent of all tagged items. An RFID system's read capability is a function of many variables, including tag size, orientation, and placement as well as interrogator antenna (IA) design. Unfortunately, "holes," which are locations where tags are not read, are present for all single IA designs. By analyzing and identifying those holes, a method has been developed for 100-percent read performance through the use of diversity in ISO 15693/ISO 18000-3 (13.56-MHz) item-level systems.

High-frequency (HF) RFID systems, such as smart carts/cabinets, are functioning in the field and are available from numerous manufacturers and solution providers. These affordable systems use passive RFID tags that are mass produced (less than $0.25 in quantity) and available from many manufacturers. The technology has great potential for tracking highvalue clinical items, some of which have limited shelf lives. For example, one common cabinet found in a hospital's cardiac catheter laboratory stores over 250 stents with an estimated total value of $375,000. Depending upon the size of a hospital, as many as four such cabinets may be in use, with products typically replenished three times per year for an annual throughput of $1,125,000 per cabinet. Implantable cardiac defibrillators (ICD) are also high-valued items in a hospital. They are small (about 3 x 4 x 6 in. in packaging) but valued from $10,000 to $20,000. They are usually stored in secure areas such as locked cabinets. Using RFID in such applications can lower the costs of understocking or overstocking certain items, as well as greater control over the whereabouts of expensive items.

A basic RFID system1 consists of a host system and RF components (Fig. 1). The RF components consist of an interrogator (reader and antenna) and tags. The purpose of the interrogator is to communicate to the tags in the field and to also (for passive systems) power the tag through the transmitted RF signal. The interrogator is responsible for protocols, supplying power to the tags, reading tag information, writing information to the tags, and ensuring message delivery and validity to the host system.

The ISO 15693 standard specifies passive tags that may only become active if placed in an RF field. In order for the tag to become active the voltage induced from the RF field (VTag) must be adequate to achieve the minimum level requirements of the RFID chip embedded on the tag. The level of VTag is a function of the tag size/orientation and the magnitude of the magnetic field strength and for an ideal loop may be expressed by2:

VTag = 2Ï€ f0NQB(Scosa) (1)

where

N = number of windings in the tag coil,
Q = the tag quality factor,
B = the magnetic field strength,
S = the area of the tag coil, and
a = the tag orientation angle.

The magnetic field strength (B) is generated by a circular interrogator antenna (IA) and may be expressed by Eq. 2:

B = (0INa2)/2r3 (2)

where

I = the IA coil current,
N = number of windings in the IA coil,
a = the radius of the IA coil,
0 = the permeability of free space, and
r = the distance from the IA.

From these equations, it is possible to derive an indication of the relationship between tag size and orientation and that of the magnitude of the field induced along the axis of the IA. Tag-to-interrogator coupling when in close proximity though is established from the complex reactive near-field relationship among the tag and the interrogator and is poorly represented by theses equations and difficult to predict accurately especially when r << a and in locations off the axis of the interrogator. In practical item-level applications, the tag is commonly in close proximity to the interrogator antenna and for this reason the choice is not to rely entirely upon these predictions.

The mechanisms critical to the understanding of RF holes are a function of both the IA and the tag designs and their interaction. HF tags are available in many designs and sizes and generally fall into two categories, planar and three-dimensional (3D) designs. Planar tags are the more common thin paper carrier types while 3D types incorporate a ferrite and are significantly smaller. The tags used in this study were of the planar type. Since the performance is a function of both the tag and IA, three commonly used tags of difsingleferent sizes were investigated, with the smallest shown in Fig. 2. Two distinct IA designs of various sizes (Fig. 3) were also investigated. The reader response was recorded with both a single tag and with multiple tags in the field. This would be representative of actual applications in which there are numerous products in close proximity. From these measurements, it was the intention to map out a 3D space that would be representative of an actual system and would locate any RF holes if present.

Information about the location of RF holes could be used to position additional antennas for use in a "holefree" diversity system. Commonly used diversity systems (Fig. 4) incorporate single-pole, multithrow switches to route multiple antennas to the RFID reader. Such systems, which are designed to switch often among a large number of antennas, rely on PIN diode switches with significantly longer mean-time before failure (MTBF) than mechanical relays for RFID systems with a single movable antenna. Diversity RFID systems integrating a reader with multiplexing circuitry (some capable of handling as many as 256 interrogator antennas) are commercially available and relatively affordable.

The RFID reader is considered the most critical item in the test setup and was specified for operation with the ISO 15693/ISO 18000-3 Mode 1 protocols. This ISO standard is mature and has global acceptance and offers the availability of both a large number of reader models and tags sizes from many established manufacturers. Due to the high quantity of products that may often be present on a single scan, the reader chosen was also specified to have the capability to read a minimum of 100 RFID tags per scan. The reader chosen for the test system has a 1-W (RF) output and was available from a reliable manufacturer. Lower-power (200 to 250-mW) readers were also evaluated but were found to have poor read range performance for practical item-level applications. Higher-power readers are available to 10 W but did not offer a significant performance improvement. In addition, higher-power levels in combination with the proposed IA would exceed regulatory emission levels. Also, these higher-power readers approach a cost that is a magnitude higher than the lower-power models.

Due to actual-use models, where a large number of tags are to be placed in very close proximity, there was a concern that interrogator detuning effects would degrade the reader's ability to properly read tags. The measured returnloss response (S11) of an interrogator antenna alone (Fig. 5) is close to 50 , which is a match to the characteristic impedance specified by the reader. Figure 5 also shows the S11 responses of the interrogator in the presence of tags of different sizes. The larger tags, which couple very well with the interrogator, have a significant effect on the S11 response and place it outside the requirements specified by the reader. Some reader models would not perform at all with tags in close proximity and the manufacturers indicated that the high mismatch would overwhelm the receiver circuitry to the point where tags could not be detected. The readers used in this study performed well under these conditions, however. Single tag degradation (relative to multiple tags) to the interrogator S11 were slight except for the case in which the tag was very close to the interrogator PCB traces. It was hoped that RF holes found with single-tag tests would be similar to those found with multiple-tag tests, to expedite the verification process for future interrogator designs.

A simple passive RF probe was found to be quite valuable in pretesting (Fig. 6). The probe consists of a tag in which its RFID chip is replaced with a light-emitting diode (LED) in order to indicate the presence of an EM field; three such probes were assembled using different sized tags. Although this one-dollar test tool is crude, it is effective as a real-time probe that could locate RF holes. This probe located RF holes that were not intuitively obvious at the time. RF holes were revealed when the tag was in very close proximity to the interrogator and positioned symmetrically about the loop PCB trace. This was verified by the interrogator's S11response in which no change was observed with the tag placed in this location or S21 measurements recorded with a small loop probe.

This was an indication that RF holes may be observed using a vector network analyzer (VNA) and observing changes in the S11 and S21 responses with tag or loop movement. Further testing with different-sized tags and interrogator antennas all indicated holes in this same location about the printed-circuit-board (PCB) trace. The area of poor performance was found to be significantly large and located all about the loop in an area where a tag would most likely be placed.

RFID tag test-bed configurations (Fig. 7) of numerous tag designs and orientations were prepared. Test-bed configurations consisted of as many as 77 tags and were oriented in the x-y plane (parallel to the interrogator plane) and orthogonal to the interrogator plane. Each chip in every RFID tag incorporates a unique identifier that can be retrieved as part of the reading process; it was used to map its test-bed position. The reader response (ability to read a tag) was recorded at fixed increments above the IA (the z axis). In addition, recordings were also made with the cards moved in small increments in the x-y plane of the interrogator. The x-y plane movement is important since it allows for symmetric alignment of tag and interrogator trace and reveals the RF holes previously encountered during pretesting.

It was found that the single- and multiple-tag tests collate well for both parallel-plane and perpendicularplane measurements. The perpendicular- plane measurement agreement degraded when tags had a spacing of less than 0.4 in., where high tag-totag coupling is prevalent. Multiple-tag tests in the perpendicular case consistently read more tags; this is thought to be the outcome of an active tag, and not the interrogator directly, exciting a tag in a RF hole due to their high level of coupling.

Continue to page 2

Page Title

Testing with the tags orientated in the same plane as the IA revealed common RF hole locations for all tag sizes and interrogator designs. As observed earlier, the RF holes occur all about the IA PCB trace when there is symmetry with the tag. As detailed in the mapping response Fig. 8(a)>, the RF holes are clearly indicated when shifting the card to the left and right in which an entire column of tags achieves symmetry with the interrogator antenna trace. As the height is increased, there is loss of the perimeter tags and the overall 3D read zone is representative of that of a pyramid. The read-zone size was also found to be proportional to the interrogator size and tag size.

Tag read performance, when oriented orthogonal to the interrogator, was expected to be poor since the tag and interrogator fields are orthogonal and do not couple well. Tag mapping in this orientation Fig. 8(b)> reveals that tags read well when in close proximity and parallel to the interrogator PCB trace, but poorly everywhere else. No RF holes were found in the vicinity about the interrogator PCB trace, contrary to the results found for the symmetrical alignment in the parallel plane orientation. Read performance, as a function of height, fell off quickly (relative to the parallel plane response) especially for smaller tags; these results justified that only larger (ISO) sized tags should be used in this orientation. Even though the overall read performance in this orientation is poor, it may improve for multiple antenna designs in a diversity system.

As the tests show, no single planar interrogator antenna will achieve 100- percent read performance over its entire surface, and there is a significant area of RF holes all about the interrogator. The results were considered promising in that they appear to offer the potential to achieve 100-percent read performance for the appropriate combination of IAs. There is also an indication that 100-percent read performance can be satisfied over the entire surface of a properly designed diversity system, without limiting the product count. These results apply to only a sampling of the possible different tag/interrogator combinations available, some of which may achieve the desired performance.

A dual-loop IA design was also investigated. It was apparent that, for the small tags, RF holes were prevalent in the center of the larger single-loop IA designs. Modeling results of the dualloop IA indicated an improvement in reading tags in the center relative to that of a similarly sized single-loop design. Although a dual-loop IA was not commercially available, one was fabricated and tested. The results showed a significant improvement in reading the centrally located tags but also suffered the same PCB trace symmetry effects in this location.

DIVERSITY ANTENNA
These test results were then applied toward the design of a diversity system, with the goal of achieving 100-percent read performance for inventory-management applications. Another goal was to provide a planar design without significant modification of existing hardware (shelving, cabinets, etc.), and that would not diminish product space and be aesthetically pleasing. The design must also account for any factors that might degrade performance, such as packaging. A great deal of testing has been performed with stent and catheter products with cardboard outer packaging and the product enclosed in a foil-lined bag. Numerous tests were also performed in tagging and reading library books in smart shelf applications. ICD testing was also performed and found to be the least problematic for their relatively large packaging, permitting the use of a large ISO tag in a parallel orientation with the RFID interrogator.

For item-level applications with stents, it was considered appropriate to tag the product such that the tag was located as close to the interrogator as practically possible. For the stent product, the tag Fig. 9(a)> was placed along the bottom edge; even with a foil-lined package, the 100-percent read requirement was met. Tests performed with the larger ISO tag placed in an orthogonal orientation were also performed and found to achieve the 100-percent read rate only if the foil package was removed. Poor results were obtained with the ISO tag in a perpendicular orientation with the foil present, for it would sandwich the tags between metal and this will detune the tag and also reduce the RF field reaching the tag.

For library books, the ISO-sized tag was placed on the inside lower edge of the front cover Fig. 9(b)>. Even with the tag at right angles to the IA, 100-percent read performance was achieved with the larger tag as long as the book width was greater than 0.2 in. When the book width is too small, it permits the tags to be in close proximity, in effect detuning the tags so that reading them becomes difficult. It should be noted that randomly placed tags can hinder any effort to achieve a 100-percent read rate and this study focused on properly placed tags.

In the early development of the interrogator design, it was understood that the combination of all single loops may be problematic in that they would strongly couple with one another and their discrete performance obtained from testing would no longer be valid. Numerous combinations of loops were modeled and it was found that the combination of single- and dual-loop (commonly referred to as a "figure-8") structures performed to complement each others' coverage, which was also indicated in the earlier tests. In addition, the coupling of concentric loop/"figureeight" pairs was predicted to be low and was later verified by measurements to be better than -20 dB.

The resulting interrogator design is shown in Fig. 10. It has been designed to accommodate the practical storage of stents, ICDs, and books or any similar-sized products with the same makeup. The design originally consisted of three loop/figure-8 pairs. This configuration has been designed for the proposed application (stents/ICDs/books) to read any tag in the same plane as the interrogator or any tag that is both perpendicular to the interrogator and parallel to the side walls. Two figure-8 pairs were later added to evaluate additional tag orientations (perpendicular and parallel to back wall) that would provide the ability to read any orientation of a tag from a single planar interrogator arrangement. The layout and spacing of the interrogator antennas was based upon achieving optimal performance with a minimum number of antennas. As shown in Fig. 10, the layout allows for a significant number of traces to run the length of the PCB. These multiple long traces in close proximity favor the reading of orthogonally oriented tags over the entire surface.

Each interrogator antenna of the design was tuned such that the impedance match would favor the condition with tags present. The impedance requirement of the reader specified the IA to match the 50- characteristic impedance of the system. To tune the response with tags present, the performance without tags present would no longer be optimal. A compromise was established between the two cases such that the match would be equivalent to a VSWR of less than 2.0:1 with any number of tags present. Previous studies investigating the performance of readers related to VSWR did not show any significant degradation unless the match was significantly higher than a 5.0:1 VSWR. It is also worth noting from Eqs. 1 and 2 that, in order to double the read range, the current in the IA generated by the reader must be cubed. Since power is proportional to the square of the current, the reader power must be multiplied by a factor 64 to achieve twice the read range in the RFID system (where P is proportional to r6); a reasonable VSWR response in the system will not result in substantial loss in the induced voltage.

Test results were recorded with a commercial reader/ multiplexer and test setup (Fig. 11). The mapping response of all eight antennas is shown in Fig. 12. It is apparent from the mapping response that 100-percent read performance can be achieved for many tag orientations. In addition, there is significant redundancy in which a tag is read by more than one antenna and the number of antennas scanned could be reduced to as low as three while still achieving 100- percent read performance. The reader/multiplexer also had the functionality to multiplex as many as 256 interrogator antennas. Using common CAT5 cables as described by the manufacturer of the reader/multiplexer, a smart cart can be easily configured and assembled that could accommodate as many as 16 RFID shelves, each of which could switch among 16 antennas. Through the use of baluns, the RF energy was transmitted over one of the four (100- ) twisted pairs of CAT5 cable with the remaining six wires used for digital input/output (I/O). At 13.56 MHz, the losses in the CAT5 cables are relatively low and 100-percent read performance was achieved through 100 feet of cable.

There were a few issues encountered in this investigation that may be considered limitations for some applications, such as acquisition speed. With the multiplexing technology incorporated, switching takes place serially; the time for the reader to acquire tags is directly proportional to the number of tags in the field. The time required to read the 77-tag test bed with all eight interrogators was measured as 40 s. After reviewing the data sets, five entirely redundant interrogator scans were identified and removed, decreasing the overall scan time to 20 s. In the field, for the stent application, we have encountered no more than 50 products/ shelf and commonly 25, which was recorded to take no more than 15 s. The speed of the reader is a function of the transponder protocol (specifies the tag/sec rate), anti-collision algorithm and the readers ability to throughput the data to the host. The good news is that the new HF Gen2 RFID standard protocol is significantly faster and readers/ tags should be available in the near future. The time concern was not an important issue in this application in which inventory updates were only required three times a day at shift changes. There are some applications that track both product through RFID and personnel through biometrics (fingerprint) or ID cards that occur in busy arenas in which there may be short time intervals between cart/cabinet access and the time to scan will be an important parameter.

Another potential problem concerns narrow products, such as the stent; there is a tendency for such products to fall flat when a shelf is not fully stocked. Under this condition, the tag will orientate orthogonally with the interrogator and may be at a height at which it may not be read, especially when a small tag is used with the product. To remedy this condition, a removable plastic divider/bookend was incorporated that would prevent the occurrence of the product falling.

With the availability of affordable commercial-off-theshelf (COTS) hardware, cost-effective diversity systems can be configured to achieve the specifications required of many inventory-management applications. These systems can be modular in nature and configured for integration into existing systems without significant loss of product space or aesthetics. Item-level RFID that incorporates multiplexed technology offers an acceptable solution for many applications and is especially attractive for tracking highcost clinical items such as stents and ICDs.

REFERENCES
1. Infineon, Chip Card & Security ICs, SRF 55V02P Short Product Information Reference SRF55V02P_ShortProductInfo_2007-06.doc, www.infineon.com.

2. Youbok Lee, "RFID Coil Design, AN678," Microchip, 2002, www.microchip.com.