For the PDF version of this article, click here.

Radio-based identification has appeared in a variety of forms in the past decade — from keyless entry badge readers to automatic toll collection to smart cards. Yet, RF identification has been around much longer, tracing its roots to military IFF (identification friend or foe) systems that appeared during the Second World War at about the same time as radar. Later, the technology was used for railroad car and military asset tracking, and even livestock management through tag implants.

RFID systems are often classified as passive (deriving power in the tag solely from rectifying the incident RF power) or active (battery-assisted). Many of these systems are based on a low-frequency standard at 13.56 MHz, with read distances on the order of one centimeter. Longer-range systems often use high transmitted power, higher frequencies, and/or active or semi-passive transponders, for range enhancement.

A key difference among systems involves the physics of the air interface; low-frequency systems, with their long wavelength energy, tend to communicate in the electromagnetic near-field, whereas microwave systems deal with radiated energy in the far field. In the former case, sensing occurs from detecting the load on a transformer primary in the presence of a “secondary winding” (the tag), whereas far-field sensing involves changing the tag's radar cross section, effectively modulating the return echo of a continuous wave (CW) signal originating at the reader. In fact, microwave RFID systems strongly resemble Doppler radars.

RFID is poised to move into a new class of high-volume uses that will cause this technology to become ubiquitous. Made possible by the Internet and its underlying information infrastructure, it is actually not the tag or reader that is the significant development, but rather the information itself — about the location and status of goods worldwide, to manufacturers, distributors, and retailers simultaneously — that makes RFID an enabling technology. This has become the motivation for retailers such as Wal-Mart, as well as the U.S. Department of Defense, to mandate the use of RFID by their top suppliers, with deployments already well under way.

Typical scenarios for this application of RFID are illustrated in Figure 1. Goods can be automatically marked and inventoried in many real-time situations, including manufacturers' conveyor lines, loading and unloading of trucks at dock doors, and handling palletized loads within warehouses or distribution centers. Labor savings from avoidance of manual bar code scanning or keypad entry makes RFID attractive for this use alone.

However, bar code replacement is just the beginning. The real economic benefits come from higher-level uses. These include theft and loss prevention, streamlined inventories, reduced turnaround time, and avoidance of unnecessary handling. More sophisticated use involves production adjustment in response to inventory levels as manufacturers gain access to real-time information on products further downstream, and even to take responsibility for replenishing goods on demand at the distribution center or retail store level. Thus, while the impetus for RFID is coming from retailers to benefit distribution efficiencies, there is also the expectation that manufacturers will become more efficient and improve their own productivity, further lowering product cost to the retailer.

The move into the UHF frequencies to exploit range benefits through the use of unlicensed industrial, scientific, and medical (ISM) bands has led to the creation of several first-generation protocols that have sought legitimacy as de-facto standards (Table 1). While they have been useful for proving system concepts, they lack the features, communications reliability, and throughput to adequately serve a growing number of applications — particularly when taking worldwide operability into account.

To address the need for worldwide interoperability, a central role was played by EPCglobal, a not-for-profit, standards-driven organization that represents the needs and capabilities of end users and solutions providers alike. Descended from the Uniform Code Council, the body that standardized bar code labels through the Universal Product Code (UPC), EPCglobal facilitated creation of a single open standard that would 1) create an environment of interoperability and international regulatory compliance, and 2) raise the bar on RFID system performance in a significant way. These two values formed the foundation of what became the Gen 2 standard in December 2004.

User's perspective: RFID systems for the supply chain emphasize tagging of pallets, cases, and in certain situations, individual items. End users view tags as merely a vehicle for data collection, and readers as the collection point. Systems must be foolproof and robust, building on existing business processes that are today centered on optical bar codes governed by the UPC format, even as RFID overcomes several limitations of bar codes:

• Bar codes require optical line of sight and are blocked by many materials that are transparent to RF.

• Bar codes are fixed at the time of printing.

• Bar codes can be rendered useless by defacement or smudging.

• Bar codes can be spoofed or easily defeated by any malicious individual having a laser printer at his disposal.

RFID can thus lead to more complete automation of material handling, with greater overall efficiency.

Unique identifiers ensure no ambiguity. A 96-bit electronic product code (EPC) field corresponds to 10²⁹ objects, which is approximately 2×10¹⁹ for every human alive. This provides absolute uniqueness not now possible with the UPC system, again offering opportunities for traceability of goods not otherwise possible.

Regulatory environment: Widespread deployment of RFID relies on availability of either dedicated or unlicensed ISM bands. Current interest is in the UHF frequencies, which offer a good balance between antenna size and path loss. However, the requirements for these bands vary widely around the world, frustrating attempts to deploy systems in an era of global trade.

North America: In the United States, the FCC provides unlicensed spectrum in the 902 MHz to 928 MHz band, as governed by Part 15, Section 247 regulations. These rules permit radiated power up to 1 W total, 4 W effective isotropic radiated power (EIRP). For electromagnetic interference mitigation, spread-spectrum techniques are used, either direct sequence or frequency hopping, with channel separation of 25 kHz and out of channel emissions 20 dB down in the latter case. Since emissions limits govern the primary transmitter, when applied to RFID, these regulations do not directly address backscatter emissions of passive tags.

Europe: With a more crowded electromagnetic environment, regulations in Europe are far more constraining, effectively limiting the range of RFID systems. ETSI EN 302 208-1 governs such systems, providing a narrow band of frequencies allocated to RFID in the range of 865 MHz to 868 MHz, with channel spacing of just 200 kHz. As such, European RFID deployments tolerate much less interference and require much tighter spectral control than Gen 1 systems were able to deliver. ETSI EN 302 208-1, therefore, includes the definition of spectral masks for both reader and tag emissions, and multiple power classes, recognizing the potential for interference with critical uses at the band edges. These classes include:

• Power Class 11: 100 mW ERP (865 MHz to 868 MHz).

• Power Class 12: 500 mW ERP (865.6 MHz to 868 MHz).

• Power Class 13: 2.0 W ERP (865.6 MHz to 867.6 MHz).

Asia: Japanese UHF RFID spectrum regulations are not fully defined at this time. The Ministry of Public Management, Home Affairs, Posts and Telecommunications (MPMHAPT) is investigating UHF band RFID spectrum allocation, with a temporary allowance in the band of 950 MHz to 956 MHz. Preliminary information suggests use of frequency-hopped spread-spectrum techniques with channelization of 1 MHz, a duty cycle limit of 16.67%, reader out-of-band emissions of -36 dBm, and no regulation of tag emissions. Limits on radiated power have not yet been established, and it is probable that RFID uses of spectrum will be reconsidered at some future date.

Similarly, RFID regulations in China and Korea are in transition. Currently, neither China nor Korea permits the use of the spectrum for RFID purposes or for other short-range devices, but both countries are showing interest in the technology. China is now considering use of spectrum at 433 MHz for active RFID and close to 900MHz for passive RFID usage, with tentative regulations expected before the end of 2004. Likewise, Korea is likely to commit spectrum for RFID at 433 MHz and 910 MHz to 914MHz bands in the coming year.

Given often-conflicting global constraints, with the implied requirement to recognize tags wherever goods might flow, the challenge is to build in support for multiple data rates, modulation formats, and interference environments through a flexible and programmable air interface.

Link and power budgets: In passive backscatter systems, range is often set by the forward link (reader-to-tag), through the radiated power available at the tag. Theoretical range is determined from the Friis equation:

where P_tag is power required at the tag antenna output, G_tag is the tag antenna gain, and λ is the wavelength of the RF carrier.

With only microwatts to work with at the tag antenna and rectifier efficiencies on the order of 20%, tag circuits must operate from only a few microamps of current at under a volt:

where K_loss is the modulation loss parameter and η_rectifier is the power conversion efficiency.

Power in the return link (tag-to-reader) accessible to the reader front end is on the order of -25 dBm to -65 dBm, which is easily detectable by inexpensive means.

Protocol: Signaling across the air interface represents a careful balance between communications efficiency and RF power loss at the tag rectifier through modulation. Amplitude modulation or equivalent approaches are regularly used to simplify detection of the forward link signal. This avoids the power burden of circuits required for coherent detection with frequency or phase modulation, but comes at the cost of power loss during the symbol transmission.

Field rewritability: Uniqueness of the EPC data means that every chip must be personalized over the air interface or during the manufacturing process. In some cases, it is also desirable to write user information into the tag while in service. Uses include time stamping, expiration dates for goods with limited shelf life such as pharmaceuticals, or real-time data such as baggage handling. User data can greatly relieve network traffic to central database servers when tag information is locally relevant. User data also opens the possibility for more efficient tag identification and sorting.

Monza tag silicon represents the first implementation of the Gen 2 standard. The first-pass functional chip is manufactured in an industry-standard 0.25µ logic CMOS process. The chip's integrated non-volatile memory supports multiple field writes, providing a way to personalize the information-bearing content of RFID tags while simultaneously achieving considerably smaller die area. These benefits derive from a novel circuit approach that perfectly fits the unique requirements of RFID.

We have learned that various circuits can be more effectively implemented in standard CMOS processes, in a fraction of the space and power needed by conventional circuit topologies. This comes through rethinking the physics of floating-gate MOSFETs.

Floating gates are typically associated with Flash or EEPROM non-volatile memory (NVM) technology, which adjusts the electronic charge on an n-channel FET floating gate to store one of two digital values. Our approach, termed “self-adaptive silicon,” (SAS) differs radically from traditional floating gate technology in two fundamental ways:

• We fabricate floating gate devices in standard digital CMOS, with no additional process steps or mask layers.

• The floating gate MOSFET remains a fully functional transistor during programming or tuning, allowing us to store precise analog values on the floating gate.

SAS allows electrically tunable transistors with permanent analog characteristics, which have application to adjustable voltage and current sources, timing delay elements, RF matching, and a whole host of other analog and digital circuits. This comes with outstanding device reliability. Non-volatile memories exceed industry norms of 100,000 rewrite cycles of endurance and 10-year retention times by wide margins. Furthermore, SAS is process-independent and has been proven in hundreds of circuits and qualified in numerous foundry processes over the past 10 years.

The Monza design consists of a complete communications transponder incorporating modem, local oscillator, power management, memory, and digital controller functions, per the block diagram of Figure 2. As a point of interest, the chip is about twice as complex as an 8086 microprocessor, yet fits on a piece of silicon about the size of a large grain of sand. A photomicrograph appears in Figure 3.

A typical tag configuration consists of the Monza chip flip-mounted on a substrate that, together with one of a variety of printed antenna designs, constitutes an inlay. Each of the antenna designs, with their associated geometries, conductive materials, and impedances exhibits performance characteristics that are optimized for their particular application. Power levels for startup and read operation are around -11 dBm, or 154 mVrms from a 300 Ω source. The capacitive component of the rectifier input impedance can be resonated out by proper antenna tuning, with the maximum power transfer condition shown in Figure 4.

Implementing the air interface requires demodulation of the forward link from the reader and modulation of the reverse link by changing the antenna impedance and, therefore, the tag's radar cross section, allowing bidirectional transfer of digital data, even though the reader supplies the only active source of RF power.

Per the Gen 2 requirements, Monza tag silicon is able to demodulate any of a reader's three possible modulation formats: DSB-ASK, SSB-ASK or PR-ASK. The tag communicates back to a reader via backscatter of the incident RF waveform by switching the reflection coefficient of its antenna between absorptive (receiving power, or not backscattering) and reflective (backscattering) states. Backscattered data are encoded as either FM0 or Miller subcarrier modulation.

The demodulator serves to detect the reader's modulated waveform and output a digital representation of the envelope. The digital output pulses from the demodulator are passed to the tag controller logic, where symbol decisions are based on the relative spacing of pulses constituting 0s and 1s. Since the forward link data are encoded into the duration of the AM, the demodulator must accurately recreate these pulse widths for the digital symbol detection, yet the modest currents permissible from the limited power budget require especially careful circuit design.

Digital functions performed by the chip consume the greatest power, necessitating the use of circuit design styles that minimize switching energy. Operation in the sub-threshold region is used to minimize the overlap current, important not only to prevent energy waste, but also to limit power supply spikes applied to the relatively high impedance power source. A specially designed cell library was created for the chip, inasmuch as conventional standard cell library elements impose too great a power and area penalty.

Digital functions are partitioned into circuits for the tag controller and a separate NVM controller that handles the rather sophisticated algorithm used to write to the memory. The tag controller performs numerous communication, calibration, and maintenance functions, including:

• Decode symbols received from the demodulator, and encode reverse link symbols provided to the modulator.

• Interpret and process commands and data symbols to advance through defined states.

• Implement the various checks instituted by the protocol to ensure air-interface link robustness.

• Interface to NVM to perform READ and WRITE operations.

• Implement security functions such as NVM lock and tag kill.

• Manage built-in test functions needed to exhaustively test and calibrate the chip following manufacture. (Test support, while not fully self-contained, pays back the silicon area many times over in terms of overall manufacturing cost.)

The Monza non-volatile memory is formatted in a 240-bit block (15 rows × 16 bits per row) and organized into two segments: 1) EPC memory (up to 96 bits), and 2) reserved memory (which contains the kill and access passwords). The NVM offers low power (900 fJ/bit read, <400 nA standby current), relatively fast write times (<10 ms per row), 100,000 cycle write endurance (assumed uniformly distributed over the tag lifetime), and 10-year retention.

As with other applications of SAS technology, Impinj's NVM circuits are able to store information by modifying the charge stored on floating-gate pFET transistors. The memory cell state is then read by measuring the current of these devices when power is applied. The use of pFET in lieu of nFET devices offers better retention, higher endurance, and no additional complexity within standard logic CMOS processes. The fact that these circuits can be implemented in standard logic processes represents a substantial advantage in multiple ways:

• The process requires the least number of total manufacturing steps (about 30% fewer relative to EEPROM) and is higher-yielding,

• There is no compromise on using the process node (0.25 µm today) that offers the best value (wafer cost vs. die per wafer), and

• The process is generic, available from multiple manufacturing sources.

An industry consortium of major RFID system, chip, tag and reader suppliers has come together to advance a flexible standard extensible to higher tag classes. A remarkable achievement, it was crafted with the cooperation of more than 40 participating companies and driven by a shared vision of cross-vendor compatibility, worldwide interoperability, and significant improvements in performance, cost, and reliability over predecessor UHF protocols (Figure 5 provides some key metrics).

The Gen 2 standard offers a roadmap that extends well into subsequent generations, as future UHF standards will be built on this foundation. In fact, the specification is so widely endorsed that it is now under fast-track consideration as an ISO standard, slated for early 2006 ratification as ISO 18000-6C. Within the next few years, RFID will work into the supply chain mainstream, offering substantial economic benefit through directing the flow of goods worldwide in a robust, just-in-time manner.

Glidden, R. et al., “Design of Ultra-Low-Cost UHF RFID Tags for Supply Chain Applications,” IEEE Communications Magazine, August, 2004, Vol. 42, No. 8.

Rob Glidden is director of engineering at Impinj Inc., responsible for RFID technology at the Seattle and Newport Beach sites. He has been involved in development of analog, digital, and mixed-signal integrated circuit products for more than 20 years in bipolar, CMOS and BiCMOS processes. Prior to joining Impinj, he was a director of engineering at AMCC, responsible for SiGe projects. Before that, he held design engineering and management positions for RF and other communications-oriented integrated circuits at TDK Semiconductor, Silicon Systems and TRW. Early in his career, he was a naval officer assigned to the Division of Naval Reactors, responsible for reactor plant electronics for the Trident submarine. Glidden earned his BSEE degree from Cornell University in 1977 and MSEE from the University of Southern California in 1978. Glidden can be reached at rob.glidden@impinj.com.

John Schroeter is responsible for technical and product marketing communications at Impinj, Inc. Prior to joining Impinj, he held a number of product marketing management posts at UTMC, Seattle Silicon, and Fairchild Semiconductor. He is the author of the Prentice Hall book, Surviving the ASIC Experience.

RSS    Save to Del.icio.us  Digg This