CONFIDENTIAL DRAFT

ODIN technologies lab – Reston, VA

Trusted, Turn-Key RFID Solutions

EPC Generation 2.0: The Next Step

Why is it Better?
and
What Should you do to Prepare for it?

12 January 2004

ODIN technologies laboratory

Reston, VA

-1-
 CONFIDENTIAL DRAFT
ODIN technologies lab – Reston, VA

Three Protocols + No Standard = Confused Users

The recent developments of the Electronic Product Code (EPC) system and the EPC
RFID protocol have created a tsunami of activity in a previously stagnant RFID industry.
Along with future promise comes present-day confusion: three distinct and mutually
incompatible versions of “EPC compliant” tags exist today, early 2004, plus an emerging
ISO standard. Most RFID end-users will want to uniquely identify and track their items
using the most cost effective and efficient means possible. Three protocols will militate
against such ideals and create unnecessary implementation problems in a global trading
environment.

The problems developing are apparent in the compliance and interoperability areas.
Using three different protocols would illogically create three different forms of
technology based on the same concept. Each protocol would need its own unique set of
readers, tags, and potentially networking methods. R&D costs would be unnecessarily
high for manufacturers and hardware would be expensive for integrators (as three
systems would be needed to serve each protocol). However, if one protocol is used for all
future RFID technology, these unnecessary costs will be eliminated, and most
importantly, global RFID adoption will occur at a faster pace. As of now, one global
standard does not exist. EPC tag protocol Class 1 Generation 2.0 is the solution to this
problem.

Primary Problem – Three incompatible standards

1. Class 0
2. Class 1
3. ISO Standard

The “acceptable protocols” today are defined as either Class 0 or Class 1 Identity tags
and are open standards that any vendor can use to identify products. Also emerging is
ISO 18000 standard. However, these protocol standards are not interoperable.

Currently, the Class 0 tag is a factory programmable tag disallowing an end-user to write
a new number to the tag. Manufacturers are being allocated specific blocks of EPC
numbers; and adding their own product codes and serial numbers to their assigned
manufacturer numbers to create a unique identifier. Factory written tags increase
administrative and logistics cost of affixing the correct tag to the correct item and
minimize a tag’s flexibility. Should the user attempt to use unused tags, they could be
stuck with a disorganized EPC address space, creating a host of network and
identification related problems. Therefore Class 0 is only acceptable for closed or tightly
managed systems.

-2-
 CONFIDENTIAL DRAFT
ODIN technologies lab – Reston, VA
Class1 tags allow end-users to write serial numbers to their tags, but current Class1
Generation 1.0 reader technology does not enable the communication with Class 0 tags.
The other problem with Class 1 Generation 1.0 tags is there is only 64 bits of memory on
the tag, limiting potential numbering schemes.

The emerging ISO standard has four primary components, but ISO-18000-6 is the one
dealing with UHF range. The ISO standard differs from the EPC in that the ISO standard
only addresses the air interface, how the tags and readers talk to each other and the EPC
standard addresses other components of the system. The good news is that it will be
compatible with the emerging Generation 2.0 protocol.

Generation 2.0 Why is it better?

The next generation of EPC protocol is better for three primary reasons:

1. It creates an interoperable, global standard,

2. There are additional features making it technically more advanced,
3. It uses more advanced anti-collision protocols for faster, more accurate
performance.

The Class1 Generation 2.0 protocol will be backward compatible for Generation 1.0
Class 1 and Class 0, and incorporate the specifications for both classes of tags. Class1
Generation 2.0 protocol will also operate with the emerging ISO18000-6 standard
protocol, creating one global standard and enabling an efficient solution to the current
lack of interoperability between Class 0 and Class 1 tags.

The Generation 2.0 tags will utilize four distinct memory banks:

1. Tag Identification
2. Object Identification (OID) – EPC Data
3. User Memory and
4. Reserved Memory

The OID memory stores the identifier of the object to which the tag is affixed and
consists of a 16-bit protocol-control parameter, a 16-bit Cyclic Redundancy Check
(CRC16), which ensures no errors in data have been communicated from a tag to a reader
(with an accuracy rate of 99.998%), and an object identifier that is an N-bit EPC code
(“N” is a valid EPC length). The tag identification memory and user memory are
incorporated within the Generation 2.0 Class 2 Higher Functionality tags that allow tag
and vendor-specific data storage. The tag identification memory stores the unique
identifier for the tag, and the user memory allows user-specific memory storage.
Reserved memory is used for system parameters such as password.

Generation 2.0 Identity tags incorporate simple privacy functionality in the form of the
Conceal function. Concealed tags will not communicate any of their memory contents

-3-
 CONFIDENTIAL DRAFT
ODIN technologies lab – Reston, VA
unless the reader authenticates itself by issuing the password stored on the tag. The tags
can also be unconcealed only if authenticated by the password. Adding to the strength of
the security is a 16-bit random number generator (RN16) used to generate numbers for
encryption of data communicated to the tag.

The flexibility of the EPC also ensures for future use and global uniqueness indefinitely.
A 96-bit number enables the theoretical identification of nearly a million, trillion, trillion
objects. To put this in perspective, if this many golf balls were placed next to each other,
they would extend far beyond the edge of the known universe. If these golf balls were
formed into a sphere, they would make an object six billion times larger than our sun.
Considering that tagged items are most likely to be larger than golf balls, the industry
believes that a 96-bit code will suffice for object identification.

Gen 2.0 Anti-Collision – No Hits, No Errors

Generation 2.0 EPC tags utilize an advanced and globally employable anti-collision
technology enabling readers to identify tags in significantly shorter time periods than
other existing protocols are capable of– this means faster, more accurate systems. When
multiple RFID tags are communicating to a single reader and are in close proximity to
each other, their signals can interfere with each other. This interference is called a
“collision.”

Anti-collision methods can be broadly categorized into three domains: space, frequency,
and time.1 The Generation 2.0 protocol development is based partially on a privately
developed protocol from BTG, a British technology company credited with discovering
interferon (a human protein) and magnetic resonance imaging. They have developed the
“SuperTag” family of RFID protocols, which operates on the Aloha anti-collision
principle, but incorporates further additions to make it amenable to RFID systems.

Passive RFID tags do not communicate with one another; therefore a tag cannot identify
a collision of its communication. The reader must control the tags to “notify” them if the
signal was successful or unsuccessful. In the SuperTags approach, tags continuously
retransmit their identifier at random intervals until the reader acknowledges their
transmission. After reception of tag data, tags can be muted or their repetition rate can be
slowed. This method of muting a tag allows the proper counting of many tags in the same
field. Another SuperTag variation involves the muting of all tags except the one being
read. This ensures that no collision occurs. After a certain period, the muted tags are
activated, one by one, until they are all counted. In other methods, the reader, by sending
a gap or power burst, prompts the tags to respond after a randomly generated delay.
Although high performance can be achieved via Aloha-based methods, they may not
function as well as binary tree searches in high tag density environments.

1
See Appendix A

-4-
 CONFIDENTIAL DRAFT
ODIN technologies lab – Reston, VA
Nearly all classes of anti-collision methods require the reader to detect tag
communication collisions. Without such capability, even the best algorithms are useless.
The new Generation 2.0 tags will use a slotted Aloha based algorithm to identify tags
According to MIT’s Auto-ID Labs, this Generation 2.0 algorithm will enable a reader to
interrogate greater than one thousand tags in less than a second with 100% accuracy, far
better than any globally available solution today.

The Generation 2.0 protocol will revolutionize the RFID industry. There will be a single
higher-performance protocol useable by all applications. If the technology is adopted, it
will enable the interrogation of thousands of tags in seconds and make the older tags,
which are still considered by the general public as advanced technology, obsolete.
Generation 2.0 is faster, more universally accepted and will allow a global user base
easier communication.

Tag protocols influence data transmission

The focus of this paper is on the tag protocol and the next generation of technology,
however consideration must be given to how that data travels to various nodes. Effective
long-term RFID architecture should be forward thinking in its design. For global
adoption to be successful, users must implement a homogeneous network communication
system using globally unique identifiers like the EPC protocol.

A much talked about, but so far still emerging solution is the “EPC network” using
current Internet Protocol as its transmission layer. One credible alternative to EPC is to
use address space from IPv6, the latest Generation of Internet Protocol. This address
space is designed to use 128 bits. This lends the possibility of using IPv6 numbers as
identifiers instead of EPC numbers; either 64-bit or 96-bit section would be allocated
from the IPv6 address space specifically for use by EPC, or alternatively, companies
could write tag IDs to RFID chips from their own IPv6 assignments. Regardless of
which approach is taken, IPv6 would provide globally unique address space suitable for
RFID tagging.

Another simple strategy under consideration would be to create globally agreed upon
headers that would identify the rest of the IPv6 number as an already-existing numbering
scheme. For example a header of “01” may mean the next number will be a GTIN
followed by a serial number, a “02” header could mean the Military’s Universal
Identification protocol would be next. Through international registries of Internet
numbers the header blocks would be assigned to ensure interoperability. This option is
also under consideration within an EPC numbering scheme – adding just an identifying
header followed by the UID, and is similar to how Wal-Mart wants their data formatted
by applying specific values (16=GTIN, 17=SSCC, 18=GLN) as the header followed by
GTIN, UPC, etc.2

2
ODIN technologies has patents pending for algorithms to automatically convert GTIN, SSCC, GLN, UID,
etc. to EPC numbers and can provide software to fully automate the translation process.

-5-
 CONFIDENTIAL DRAFT
ODIN technologies lab – Reston, VA
The Department of Defense has mandated that computing systems be IPv6 compliant by
4, 2006, and the Japanese government has mandated all Japan-based businesses to
compliant by Q4, 2005. Much of Europe has also begun adoption of IPv63 making the
protocol a credible global alternative to the current IPv4.

Although the future of technology is always unknown, EPC or a combination of EPC and
IPv6, will be the most probable solution to RFID object labeling. The more bits a tag
uses, the more expensive the tag is to produce. Therefore, it is imperative to create an
economical global standard protocol for RFID technology; and the Class 1 Generation
2.0 EPC Identity protocol is the most realistic solution to the deadlines set by government
agencies and large retailers.

User Recommendations
Based on what we know today about the change in protocol ODIN technologies
recommends that as you refine your project plan for compliance, you pose the following
four questions:

1. Have I started a dialogue with the older more established chip

manufacturers?
2. Does my vendor(s) have a demonstrable path for upgrading to the
Generation 2.0 protocol?
3. Does my contract with my vendor(s) have a Service Level Agreement,
which protects me from additional cost associated with Generation 2.0?
4. Do I have a plan for translating my existing numbering scheme to EPC
numbers and what is the cost in time and dollars for that migration?

Some estimates predict EPC Generation 2.0 tag technology will be available for
production in less than a year. However since a new chip design is required, which is a
significant process, that timing is unlikely. ODIN technologies’ lab estimates that broadly
obtainable Generation 2.0 tags will not be available until sometime in early to mid ’05.

Established chip manufacturers will be the ones leading the way with Generation 2.0 tags
since there is significant cost associated with designing this Generation 2.0 chip ($1-5
million according to various estimates). This design cost is much easier for a multi-billion
dollar company to absorb than a new, venture-backed company. Therefore ODIN
recommends opening a dialogue with the larger chip manufacturers if your project
strategy calls for early deployment of Generation 2.0 tags. The newer companies, which
have shown better tag performance, are working hard to be at the forefront but may be
“out spent” by the larger better capitalized competitors.

With Generation 2.0 tags likely becoming available after several looming deployment
deadlines (Jan ’05 for Wal-Mart and potentially DoD), careful consideration should be
3
ODIN technologies Director of IT, Nick Hilliard, deployed the first country-wide IPv6 network in Ireland
in 2002.

-6-
 CONFIDENTIAL DRAFT
ODIN technologies lab – Reston, VA
given today to choosing a reader systems. Time should be spent over the 6-10 months
designing and implementing the optimal reader architecture and tag placement to meet
the 100% case and pallet read requirements that Wal-Mart and the DoD are mandating.

As part of a production system architecture, companies should choose reader

manufacturers that can demonstrate today that their readers can be upgradeable via
firmware to accept the new Generation 2.0 protocol. Several large companies have
already made significant capital investments in hardware that is soon to be obsolete.

A system upgrade provision should also go into the purchase contract Service Level
Agreement (SLA) either with the hardware manufacturer, or the integrator performing the
installation, to ensure the timeliness of being Generation 2.0 compliant.

Once systems have been designed, tested, and deployed in the field with 100% accuracy,
a structured plan should be implemented for translating existing numbering schemes
(UPC, GTIN, SSCC, UID, etc) to EPC and then creating an infrastructure for
management of the numbers and monitoring of the system.

The last step should be to evaluate how EPC will impact your current systems and what
the costs of integration and scalability are. ODIN technologies will continue to leverage
its portfolio of patent- pending technologies to create solutions to make these tasks easy
for the end users.

-7-
 CONFIDENTIAL DRAFT
ODIN technologies lab – Reston, VA

Appendix A: Anti-Collision Domains

A more detailed explanation of how anti-collision works
Space Domains

In space-domain methods, tags are placed in specific locations to achieve isolation. Tags
in space-domain methods are identified by variation of reader range (variation of power
is transferred to passive tags) and/or by using directional antennas (as opposed to
omnidirectional antennae). Triangulation, using ultra-wide band communications and
positioning methods, poses another possibility of locating individual tags. RFID
identification using only space-domain methods drastically hinders the effectiveness of
the technology. Directional antennas limit the space in which readers can communicate
with tags; and without directional antennas, isolating a tag emitting a broad three-
dimensional electromagnetic field limits the amount of tags that can be placed in a certain
area. Space-domain methods rely on the amount of tags in a reader’s range—if there are
too many tags in the area, collision will result and reduce the reader’s ability to interpret
signals.

Frequency Domains

Frequency-domain anti-collision methods allow for robust wireless communications, but

can add excessive complexity and cost to the tag. Frequency Division Multiple Access
(FDMA) systems divide the total available bandwidth into fixed width channels. FDMA
is costly because it requires accurate frequency sources and band-pass filters. One
company has implemented a system combining FDMA with Time Domain Multiple
Access (TDMA). While this system would undoubtedly perform well, the cost-
effectiveness of this system is unclear. Code Division Multiple Access (CDMA) systems
have many advantages over FDMA systems. CDMA offers better adaptability to varying
traffic load, increased capacity to read tags, and processing gain. CDMA and other
Spread Spectrum (SS) methods are currently difficult and costly to implement because of
their increased complexity. Additionally, there are bandwidth limitations, and for this
reason SS methods (including Frequency Hopping (FF) and Direct Sequence (DS)) are
generally confined to operation in UHF and microwave bands.

Time Domains

Most RFID anti-collision methods are time-domain. In these methods, fractional

communications from tags are varied in time. Time-domain methods can be classified
into synchronous and asynchronous schemes. Synchronous schemes are those where a
reader transmits a query to a specific tag using its UID (unique ID). This is an effective
anti-collision method because tags do not have to “take turns” communicating to the
reader, and tags do not have to rely on a complete “uncollided” transmission to be
identified. A reader can poll through a list of tags, but the polling method, also known as

-8-
 CONFIDENTIAL DRAFT
ODIN technologies lab – Reston, VA
tree walking or binary tree, is relatively time-consuming and depends on the tag’s UID
number being known. Binary tree searches4 use binary code (groups of 0’s and 1’s) for
communication and basically involve the reader actively sending a signal of a 1 or 0 to a
tag. If the reader sends the correct number, the tag acknowledges it by transmitting the
signal back to the reader. If the reader sends the wrong signal, the tag mutes itself and
awaits another signal. Eventually, the computer deciphers the code of the tag.

To simplify the method of binary tree searching, imagine there is one tag in the field of an
antenna. If the reader sends the correct number, either a 1 or 0, the tag acknowledges it
by transmitting a signal back to the reader, and awaits another signal from the reader.
The reader has two commands when sending out bit numbers—“DOWN” and
“TOGGLE.” The “DOWN” command prompts the reader to send the same bit number as
the last bit number, and the “TOGGLE” command prompts the reader to send the
opposite bit number. The reader interrogates the tag randomly, resetting to zero when it
sends the wrong bit. However, it only has to interrogate the tag 8 bits at a time, because
after each byte (set of 8 bits) has been identified, the memory is stored, and the tag sets
the next bytes first number to zero.

The binary tree searching method is more advanced when multiple tags are in an
antenna’s field. Both the reader and tags act differently. In this method, a tag does not
acknowledge the reader’s signal if the reader sends the wrong bit number. The reader
will recognize this lack of communication and the tag will mute, reducing the chance of
collision. When a tag is muted with the “FLAG” command, it sets a pointer to the
beginning of the last byte identified. This way, the reader can partially interpret tags, 8
bits at a time. If a reader tried to decipher Generation 2.0 tags, 96 bits in a row, it would
take an extraordinary amount of time. Eventually, through the random binary tree
process, one tag remains un-muted, and the reader deciphers the code of this tag based on
the method previously explained. Once this tag’s code is deciphered, it mutes itself and
the reader subsequently sends an “UP” command to tell all muted tags (except the
identified tag) to become active again. Another tag is discovered and the cycle repeats
until all tags are read. Some binary tree searches use a more complex method of
signaling groups of bits to detect tags. In deciphering code, this algorithm is surprisingly
fast when compared to many time domain methods. Matrics tags use this anti-collision
algorithm and ODIN technologies lab research has shown it to be an effective method for
interrogating multiple tags at a fast rate. Matrics tags use this anti-collision algorithm
and ODIN technologies lab research has shown it to be an effective method for
interrogating multiple tags at a fast rate largely because of the wide bandwidth Matrics
utilizes. This creates a problem, however in Europe where the constraints on the
frequency band is much greater.

Asynchronous schemes are those in which tags in the reader’s field respond at randomly
generated times. This helps to reduce the chance of collisions. The “Aloha5” scheme is
asynchronous and involves a node transmitting a data packet after receiving a data
packet. If a collision occurs, a node becomes saturated and transmits the packet again
4
See Appendix B for diagrams
5
See Appendix B for diagrams

-9-
 CONFIDENTIAL DRAFT
ODIN technologies lab – Reston, VA
after a random delay. The reader transmits continuously until a collision does not occur.
In “slotted Aloha6,” transmission is performed in slotted times by making small
restrictions in the transmission freedom of individual data packets. When packets collide
under slotted Aloha protocol, they overlap completely instead of partially, and this
significantly increases the efficiency of data transfer.

Aloha is more easily understood with an analogy. For the purposes of the analogy,
imagine there are fifty tags in an interrogation zone. Imagine fifty railway stations as
tags, a railroad car as a tag’s transmission, and two tunnels which represent the
antenna/reader. A train gets loaded with supplies at its respective station (which is
representative of the data being delivered) and once supplies are delivered through the
tunnel, they are no longer needed, and thus the next train does not need to leave the
station. One tunnel (the reader output tunnel) has trains coming out on one track, one
behind the other, at constant speeds (this represents the electromagnetic waves being
emitted by the antenna). At the fifty stations, the trains stop simultaneously, pick up their
supplies, which takes a random amount of time (analogous to the random delay of Aloha
protocol), and take off for the next tunnel. The tracks all converge at the one tunnel,
however, and the trains cannot touch each other while merging, or the “collision” will
destroy them; and thus no supplies will reach the other side (no data will be transferred to
the reader). Trains will continue to leave stations until the supplies are delivered, and the
station, receiving word from the “antenna/reader” tunnel, shuts down (representing a tag
shutting off, or “muting” itself). The trains are long, and even though they move
extremely fast, they sometimes overlap by random distances. The more trains there are
leaving at random times, the more chance for collision.

This is representative of pure Aloha protocol. In slotted Aloha protocol, trains are only
allowed to leave at certain times (slots). Imagine the stations have stoplights, and that all
turn green and red at intervals just long enough to prevent partial train collision. If a train
is ready to go at green, it goes, but if it is not, it must wait for the next light. This ensures
that if two trains (signals) do go at the same time, they completely collide. If they
overlapped (as they do with pure Aloha), the lagging train would still have the ability to
collide with a train behind it. Now imagine the light flickers on and off multiple times
per second, and the trains instantly accelerate to light speed. It may be easier to see how
slotted Aloha is superior to pure Aloha. As stated before, once the train has delivered its
supplies, it sends a signal so no train leaves its respective station; thus ensuring a smaller
chance of collision.

6
See Appendix B for diagrams

- 10 -
 CONFIDENTIAL DRAFT
ODIN technologies lab – Reston, VA

Appendix B - Diagrams

Diagram 1.1 gives a visual representation of binary tree searches.

Diagram 1.1
ACK stands for “acknowledge”
(FLAG) means the tag mutes itself
The Up command prompts the reader to begin a tree search again
The Down command prompts to move down the command list (tree) and send a bit
The Toggle command prompts the reader to send a different bit

This diagram represents a hypothetical binary tree search when only two tags are
in the field. The first tag mutes itself when it fails to receive the correct bit
transmission from the reader. Once Tag2 is muted, it will not send any response
to the reader, and will not react to any signal except the “Up” command, which
will unmute it, and prepare it for data transmission. Note that the tags are only 8
bits long - any “Flag” command will bring the tag back to zero.

STEP 1

Tag 1 Tag2 Reader Command Tag1 Tag2

0 0 Down ACK ACK
0 11 Down ACK (FLAG)
0 0 Down ACK
1 1 Toggle ACK
1 1 Down ACK
0 0 Toggle ACK
1 1 Down ACK
0 0 Up ACK

- 11 -
 CONFIDENTIAL DRAFT
ODIN technologies lab – Reston, VA
Diagram 1.2
ACK stands for “acknowledge”
The Down command prompts to move down the command list (tree) and send a bit
The Toggle command prompts the reader to send a different bit
The Done command prompts the reader to stop sending signals
The “x” in the Tag1 bit locations represent that the tag is muted

STEP2

Tag1 Tag2 Reader Command Tag2

x 0 Down ACK
x 1 Toggle ACK
x 0 Toggle ACK
x 1 Down ACK
x 1 Down ACK
x 0 Toggle ACK
x 1 Down ACK
x 0 Toggle Done

- 12 -
 CONFIDENTIAL DRAFT
ODIN technologies lab – Reston, VA
Diagram 1.3- Representations of Pure Aloha Protocol
Black rectangle represents successful transmission
White rectangle represents collision

Rectangles are Transmissions from tags traveling toward a reader

Tag 1 Transmission

READER
Tag 2 Transmission

Tag 3 Transmission

Tag 4 Transmission

Time
No overlap No overlap No overlap

Diagram 1.4- Representations of Slotted Aloha Protocol

Black rectangle represents successful transmission
White rectangle represents collision

Rectangles are Transmissions from tags traveling toward a reader

Slot 1 2 3 4 5 6 7 8

Tag 1 Transmission

READER
Tag 2 Transmission

Tag 3 Transmission

Tag 4 Transmission

Time

These diagrams show the difference between Pure Aloha protocol and Slotted
Aloha protocol. To reiterate, in pure Aloha protocol, the transmissions overlap by
random distances during collision, leaving the chance for infinitesimally small distances
to completely disrupt tag interrogation. In slotted Aloha, transmissions overlap
completely or not at all, and this increases the efficiency of data transfer.

- 13 -
 CONFIDENTIAL DRAFT
ODIN technologies lab – Reston, VA

- 14 -