Five-Cent Tags -- Then What?

There are a whole new set of challenges if 5-cent RFID tags become a reality. Today, at a cost of 50 cents each or more, worldwide demand is about 400 million tags per year, say researchers. Auto-ID Center research shows that demand for a 5-cent tag will be much, much larger. And if tags reach the 1-cent range -- a real possibility in the long term -- demand may well equal that of bar codes. Now, as many as 5 billion bar codes are scanned daily worldwide. If Center sponsors decided to place the new tags on every item they manufacture, demand for these ICs could total several hundred million daily. That would mean a large portion of worldwide IC manufacturing would be dedicated to producing these tags. Also, consider that manufacturing tags in large volumes would require new processes, new equipment, and new suppliers. This in itself may delay availability of RFIDs in large volumes for several years. But don't put the cart before the horse, says Sanjay Sarma, Auto-ID Center director. In that vein, the Auto-ID Center prefers to focus on creating the 5-cent tag first, then leave scaling, commercialization, and final pricing to vendors.

Believe it or not, long checkout lines may one day be relegated to the dustbin of history. The reason: Super-cheap smart tags that will let packages ring themselves up as you leave the store. Though such a development might revolutionize trips to the mall, the biggest improvements will be behind the scenes. Smart tags may let manufacturers track, in real time, every product they make from cradle to grave.

These scenarios hinge on development of both inexpensive Radio Frequency Identification (RFID) tags and an infrastructure of standards to surround them. Much of the effort toward low-cost RFID is underway at the Massachusetts Institute of Technology's Auto-ID Center (www.autoidcenter.org). The industry-funded research center is working with a host of major companies, including Procter & Gamble and The Gillette Co., to create this next-generation bar code. First on the agenda is the design of an open network to identify and globally track individual products through the supply chain.

The Center's ultimate vision? A universal environment where computers understand the world without human help, identifying any object, anywhere, instantly. If perfected, such a system could help save businesses billions of dollars in lost, stolen, or wasted products, say researchers. Consumers and the stores they shop in would also benefit. Besides shorter lines, smart store shelves would tell manufacturers when to restock items. Potential buyers could also easily look up product features by, say, simply pulling out a scanner-equipped cell phone and linking to the manufacturer's Web site.

But before such smart tags become commonplace their price needs to radically drop. RFID tags currently cost more than 50 cents each because most still combine large ICs, wired or etched-copper antennas, and time-consuming manufacturing processes. The Auto-ID Center and its sponsor companies believe new technologies and manufacturing approaches in the works will bring the first 5-cent tags.

The backbone of a new global auto-identification system developed by the Auto-ID Center relies on five essential elements: Electronic Product Codes (EPC), ID systems (radio-frequency readers and tags), an Object Name Service (ONS), Physical Markup Language (PML), and Savant software technology.

The nuts and bolts
Though bar codes are cheap, reliable, and fairly easy to produce, they have limitations smart tags don't. Like bar codes, EPCs are made up of a series of numbers identifying the manufacturer, product, version, and serial numbers. EPCs, however, contain an extra set of digits that identifies each distinctive item.

The EPC (a 96-bit code of numbers) is embedded in a smart tag and typically applied to products during manufacturing. A smart tag communicates its EPC via radio waves to various readers in plants, warehouses, stores, and shelves. The readers then transmit the product-identity code to the Internet. EPCs work with a Physical Markup Language and an Object Naming Service. According to Auto-ID researchers, PML is a new standard language, similar to HTML, for describing physical objects to the Internet. The ONS tells computer systems where to find information on the Internet about a product with an EPC code. The ONS is partly based on the Internet's Domain Name Service, say researchers, but will likely be much larger, acting as a superfast post office that finds data for every single product with an EPC code. Software technology called Savant created by the Auto-ID Center manages and moves the information around the EPC network to avoid overloading existing networks.

Smaller IC is key
One of the first steps toward cheaper RFID tags is a smaller IC that does less work. The goal of Auto-ID researchers is a tag carrying only the EPC; this limits the memory required on the chip, cuts the cost of the tag, and also adds flexibility and security because any product data can be stored and accessed online. Through ONS, the EPC is mapped to an Internet Protocol address where information about the tag can be written and accessed. This redirection reduces space and power requirements on the tag, according to researchers.

Scanners built to the Auto-ID spec must be able to read multiple tags and sort them into collision-free channels. This is called tag anticollision. Auto-ID Center Research Director Sanjay Sarma says tags carrying an EPC need only be able to quickly read several tags and narrow the search down to certain versions, manufacturers, or product numbers. The anticollision system must be realized on the least amount of silicon, and the Center and its sponsors believe they've developed such a scheme optimized for size and speed.

According to Sarma, the target size of the new IC is 0.25 mm2, possible because the new chips need less capability than ICs currently used. But conventional techniques would have a hard time fabricating these mini ICs. Trying to "shoehorn RFID ICs into existing manufacturing techniques for standard ICs is suboptimal," says Sarma. New manufacturing technologies in the works at the Center and its sponsor companies may bring about a prototype 5-cent tag within a year.

One such idea is to eliminate full die testing in favor of wireless testing after the RFID is built. That would do away with costly mechanical motion systems currently used in testing. "Wireless testing is conveniently available in RFID," says Sarma. "Since the RFID chip we are designing is very simple, yield will tend to be high after the process is tuned. Furthermore, the anticollision scheme we've developed is very rapid when the distribution of EPC codes is serial. Read rates expected are in the range of several hundred a second, and the cost of testing equipment will be low." Researchers recommend testing at the final packaging and conversion stage.

New wafer-processing methods producing thinner ICs with narrower streets (gaps between adjacent dies on a wafer) can also lower RFID costs. A key in achieving this is making IC wafers thinner. Wafers can be thinned much more economically thanks to advances in grinding, chemical-mechanical polishing, and wafer handling. Also, improvements in etching have cut costs further. According to Sarma, the wafer is first etched from the front side using dry or wet etching to create deep but narrow (5 micron) grooves. Next comes a grinding operation on the wafer back. A final chemical-mechanical polishing step results in separation of dies from the wafer. This new process permits cost-effective die separation on a large scale and results in higher-quality dies, while eliminating the traditional and less precise diamond-sawing operation.

Better assembly
Cheaper RFID manufacturing relies heavily on a technique called roll-to-roll processing, where a continuous flexible substrate is used as opposed to current batch-oriented fabrication with rigid circuit boards. This offers improved efficiency and lower costs that come with high-volume production. One such technique showing promise is called Fluidic Self Assembly (FSA) from Alien Technology Corp., Morgan Hill, Calif. (www.alientechnology.com). The FSA process, originally developed at the University of California-Berkeley, suspends trapezoidal-shaped (due to the etching process) ICs called Nanoblocks in liquid and then flows them over the substrate surface. The Nanoblocks range in size from 10 to several hundred microns on a side. Correspondingly shaped holes or receptors on the surface catch the self-aligning Nanoblocks as they pass by. Bulk processes such as this, says Sarma, have the potential for high yield, low costs, and the ability to scale to massive volumes. Most importantly, it works roll-to-roll. Alien recently partnered with another Auto-ID Center member, Rafsec Ltd., Finland (www.rafsec.com), to develop the first RFID IC that operates according to the Center's specifications. Nanoblocks packaged and interconnected using FSA will arrive on small straps to Rafsec for conversion into finished tags with antennae.

Another bulk-processing approach for getting ICs to a substrate comes from M.I.T. and uses vibration. Chips march down channels to the tune of a 300-Hz vibration and are aligned and oriented along the way. On-wafer processing is yet another cost-cutting idea. Scientists at the Frauhofer Institute and at FlexChip AG want to eliminate some of the handling required for single chips after the die is separated. One approach, explains Sarma, uses a flip-chip-type connection to attach contact-elements directly to the dies that are still on the wafer. "These leads can effectively act as interposers, and are expected to allow large-scale integration of ICs into RFID packages," he says.

Testing, testing
While bugs in the technologies still must be worked out, the Auto-ID Center and its sponsors have begun testing the concept in phases. Phase 1, initiated last fall, tracked inventory pallets embedded with two commercially available RFID tags from distribution centers to their final destination, a Sam's Club in Tulsa, Okla. Phase 1 testing evaluated technologies such as EPC, ONS, PML, and the Savant software, and also added stress to the system to see if these technologies could handle multiple entry points within the supply chain. Though the Auto-ID Center components worked as designed, readers couldn't pick up all of the tags. (Multiple tags per item boosted identification hits.)

This past summer Phase 2 testing began. Unlike previous tests that just tracked pallets, Phase 2 tests put ID tags on individual cases of items to see how the system handles the additional data.