I don’t watch much television , but when I do, it’s because my Tivo’s found something cool. Like: last night’s introduction to RFIDs by Dr. Chris Diorio, chairman and CEO of Impinj. His presentation is available online, so I will just summarize some of the things I learned.
Who wants them?
Business! Among the benefits:
- For the supplier, RFIDs provide a way to track a product through its complete lifecycle.
- For the retailer, it offers quick counting. UPC codes often require you turn over the object to find the code. (This is one of the annoying aspects of self-checkout.) RFIDs do not require a line of sight to communicate.
- For both, there’s better inventory control.
- RFIDs are rewritable, allowing the supplier to update information as the product passes through the chain.
- By having an RFID reader on the shelf, periodically interrogating the items around it, a retailer can determine when products are running out. (For high-churn things, this may not make sense.)
- It would be possible to triangulate mis-shelved items in a store.
- Anti-shoplifting. For example, if you go into Barnes and Noble, the tags are placed in random books. The reader is at the door. The RFID is deactivated at the register.
- A supplier can isolate where “shrinkage” occurs. (Shrinkage is also known as “the object fell off the truck.”)
- A broader catalog. Whereas UPC codes are limited to 100k manufacturers with up to 100k items each (which is more restrictive than it sounds because products can be phased out), RFIDs’ Electronic Product Code fields offer enough items for individual serial numbers, a marketer’s wet dream. The 12 bytes break out into:
- 8 bits for a header
- 28 bits for the company (268,435,455 companies)
- 24 bits for the object class (16,777,215 groups)
- 36 bits for the serial number (68,719,476,735 items)
The 96-bits could be extended pretty easily, apparently, accommodating the number of Big Macs sold if necessary.
- Harder to spoof. There are apparently scams where people print UPC codes on stick-on labels to fool scanners. (I suppose this is a variant of the black marker scheme. I don’t know if this is as common as people believing what they read on CraigsList.)
The potential market for RFIDs was described in the Trillions (Texas “T”).
Passive tags are of most commercial interest.
An active tag has a battery: it’s essentially a radio. They’re larger (120 x 100 x 50mm), operate at higher power, considered more reliable and better suited for water-based environments like people or large objects that would be interrogated from far away (think: shipping containers). Battery life is 5-10 years.
Because passive tags do not have batteries, they’re smaller, cheaper, and have a potentially unlimited lifespan. Their power comes from induction in the antenna occurring in certain frequency ranges. Interrogation ranges vary from a few meters up to 100+ meters on the newer chips. The tag is a tiny chip about 0.75 mm attached to a piece of paper. The antenna is silk-screened on with conductive ink.
Wikipedia mentions semi-passive tags that use a battery to power the chip, but inductance to send information back.
The grain-sized Impinj ZumaRFID chip field-writable, with an 25′ read/18′ write range. (The protocols limit the reader to identifying 500 tags per second.) They get about 40,000 of these chips in one 8″ semiconductor wafer. The sample wafer shown had a lot of empty space on it, presumably for testing chips.
The chip has 41,798 transistors (compared to 29,000 on the original Intel 8086). Power consumption is 8 microwatts, or 1/500,000th the original 8086. Subsequent chips have 50% more transistors in 30% less area, but with similar power consumption.
The antennas are the largest part of the package. The reader can process up to 500 RFIDs per second. However, success rates in the field are still in the 80th percentile, slowing the adoption by large agencies like Wal-Mart and the Department of Defense.
Tag, you’re it!
The communication protocol was described as being analogous to the reader having a flashlight and tag having a mirror. To send information to the tag, the reader modulates the “light” sent to the tag. To receive information, it holds the light constant while the tag uses the “mirror” to reflect or not reflect. (The tag returns its electronic product code by reflecting incident RF.)
Then there are the logistic problems… The reader must be able to identify all the tags present in the room. The tags are essentially blindfolded, nearly deaf, talk by whispering, and cannot hear each other. The reader is blindfolded, doesn’t know who’s in the room (because the tags, if present, are all powered off), can’t listen to more than one tag at a time, and needs to yell a million times louder than the tags can respond. Furthermore, since the reader broadcasts with enough power to be observed outside the retailer’s premises, it cannot say the tags’ names, lest the retailer’s competitor learn what she’s selling.
The example algorithm he presented was showed how the set of numbers was exchanged without the reader calling the chip by name. Once identified, a chip is asked to go to sleep so it doesn’t cause collisions. (He didn’t elaborate on the mechanism to handle multiple responses, though it sounded like it was a variant of the Ethernet CSMA/CD scheme: if there are collisions, both parties wait for a random, exponentially increasing period before trying again.)
There were additional tricks needed to deal with the physical issues with the silicon. For example, very cold conditions can make the RFID chips stay “asleep.” If something was scanned before being loaded in a truck in a typical Minnesota winter, it might take half an hour for it to warm up enough to turn on.
Many of the passive RFID examples assume they’re on a non-metallic and/or dry surface. RFIDs being implanted (!)
From a pure technology and future, benevolent application perspective, it’s pretty cool.
There are a lot of privacy[9,10] and security[8,11] concerns. I haven’t read through this literature; however, I have the urge to cough *bullshit* whenever I hear “uncrackable” mentioned. Such as . I don’t know how much to make of the problem, though, as people apparently put highly personal stuff on Facebook because they believe it’s secure.
And health concerns. RFID reader transmission power is in the 900MHz band, and of the power of 2 cell phones, but is not held against your head. Diorio asserted that this background radiation “acceptable health limits.”
There needs to be a critical mass for these to work. Some retailers have run into slow adoption. For example, Wal-Mart has only 600 suppliers out of 20,000 using RFID technology. Sam’s has recently imposed a fee for suppliers not using them.
Sources and further reading:
-  Chris Diorio, UW Television, “RFID: The Next Big Little Thing,”April 28, 2005.
-  Radio-frequency identification, Wikipedia
-  RFID Security, Frank Thornton, ISBN 1597490474.
-  “Hot Topics: RFID,” Stanford Graduate School of Business, Jackson Library.
-  “About RFID,” Impinj.com
-  “Wal-Mart Rethinks RFID,” Information Week, March 26, 2007.
-  Technovelgy.com
-  “Arphid Watch: Find Own Foot, Aim Hastily, Pull Trigger,” Bruce Sterling, wired.com
-  Electronic Privacy Information Center.
-  Spychips.com
-  Gildas Avione, list of papers on RFID security and privacy.