UHF Gen2 RFID Where do we go from here?

April 27, 2009 Chris DiorioImpinj, Inc.

[email protected]

UHF Gen2 RFID Where do we go from here?

2

A Brief History of UHF Gen2

First Gen2 chip

Gen2 ratified (Dec 2004)

First DRM reader

First fielded NF UHF

First prod’n reader chip

Ancient history (ISO 18000-6 A/B & Auto-ID C0/C1)

Key RFID mandates

Privacy & security

< 2003

2003

2004

2005

2006

2007

2008

2009

I have omitted innumerable milestones

and key events…

3

Step Back to 2003…

• These entities (and many others) joined EPCglobal– They created user requirements– They wanted a worldwide standard– They wanted RFID products that worked

4

2003: RFID Issues

• Existing UHF RFID simply didn’t perform– No dense-reader capability– No worldwide operation– Poor spectral efficiency– Ghost reads– Low speed– Lots more

• UHF RFID technology was ancient

Example: A “difficult-to-detect” response from an RFID tag

5

2003 Tradeoff: T->R Signaling• Proposed Gen2 FM0 Encoding

– Linear modulation with memory– Biorthogonal basis functions

• Class-1 F2F Encoding– Linear modulation w/o memory– Orthogonal basis functions

A

-A0 T

s1(t)

A

-A0 T

s2(t) = -s1(t)

A

-A0 T

s3(t)

A

-A0 T

s4(t) = -s3(t)

Data 0 Basis Data 1 Basis

A

-A0 T

s1(t)

A

-A0 T

s2(t)

Data 0 Basis Data 1 Basis

T

FM0

F2F

1 1 1 0 0 1 0

6

T->R Bandwidth Efficiency

• FM0 requires approximately 0.5× the bandwidth of F2F– FM0 : 80% power bandwidth ≅

1.25/T – F2F : 80% power bandwidth ≅

2.45/T

• FM0 has a lower symbol BER than F2F at a given Eb/No– Exploit memory within FM0 waveform (MLSE)

0 1 2 3 4 5 6 7 8-80

-70

-60

-50

-40

-30

-20

-10

0

10

Normalized Frequency (fT)

Nor

mal

ized

PSD

(dB

/Hz)

Power Spectral Density

F2F (Simulation)FM0 (Simulation)FM0 (Theory)

0 1 2 3 4 5 6 7 80

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Normalized Frequency (fT)

Frac

tion

of T

otal

Inte

grat

ed P

ower

Integrated Power

F2FFM0

7

Introduce Modern Radio TechniquesA recoverable signal for a properly designed Gen2 reader

8

2004: UHF Gen2 Ratified

User Requirement Gen2 Capability

Global regulatory compliance Europe, North America, others

Memory access control 32-bit access password, memory locking

Fast read speed > 1000 tags/sec peak

Dense-reader operation Dense-reader operating mode

Kill security 32-bit kill password

Memory write capability > 400 tags/minute write rate

Bit masked filtering Flexible Select command

Optional user memory Vendor option

Low cost Multi-vendor availability

Industry certification plan EPCglobal™ certification

9

Simulated Gen2 Throughput• Absolute peak performance

– Assumes 100% detection of tag collisions and empty slots– Assumes tags do not lose power during an inventory round

0 1 2 3 4 5 6 7 80

200

400

600

800

1000

1200

1400

1600

Number of Tags/Number of Slots

Ave

rage

Thr

ough

put (

Tags

/sec

ond)

Peak Throughput

25/25 kbps40/40 kbps40/80 kbps40/160 kbps62.5/62.5 kbps80/160 kbps160/160 kbps160/640 kbps

10

FilterFilter

Dense-Reader Mode

• Goal: 10’s or 100’s of readers operating simultaneously– Using only a few MHz of bandwidth

• Solution: Dense-reader mode using Miller backscatter – Separates tags and readers in frequency– Prevents reader—tag interference– Eliminates need for LBT

Readers collide with readers but not tagsReaders filter interfering readers from their tag responses

11

Dense-Reader OperationEliminate signals

outside filter

-500 0 500

Interfering Reader into Rx

Reader noise is low inside filter

12

2005 Gen2 Testing

Test SetupUnits under test: 2 readers, each controlling 2 antennasInterferers: Up to 13 Interferers per side, all transmitting

Interfering reader(s)(simulating adjacent door)

Interfering reader(s) (simulating adjacent door)

Reader

Antenna

Legend

Metallic StandFor reader and antenna mounting

3 m3 m

Pallet of Tagged ItemsTags buried within pallet

3 m

FCC environment

13

2005 Test Results – It Works!!• Reader on either side of dock door

• Tags on each of 40 boxes of Caress® soap

• All readers transmitting simultaneously

• There is a co-channel or adjacent channel reader 80% of the time

0%

20%

40%

60%

80%

100%

Inve

ntor

y Re

liabi

lity

Number of Interfering Readers

100%

100%

99%

97%

0 8 16 26

14

2006: EU Regulatory Issues SolvedETSI EN 302 208 RFID channel plan with no LBT in 2W channels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 CHANNELS

2 Watt

THE TEST36 dock doors

62 items/pallet36 pallet pulls (simultaneous)

36 readers transmitting

simultaneously98.2% accuracy

15

2007: Near-Field UHF for Item Tagging

Far-Field UHF RFIDElectromagnetic waves

Long range: 30cm to 10m

Attenuated by dielectrics

Near-Field UHF RFIDMagnetic or electric fields

Short range: 0 to 60cm

Unaffected by dielectrics

IMPINJ PROPRIETARY & CONFIDENTIAL

• Far field: 30cm to 10m range

• Near field: 0 to 60cm range– Reads on liquids and metals– The only difference is the antennas

16

Antenna Characteristics

–40

–35

–30

–25

–20

–15

–10

–5

1 10 50

Typical PatchBrickyard

Mag

netic

Fie

ld In

tens

ity H

, dB

(A/m

)

Z-axis distance from boresight (0,0,z), cm

Antenna Type Typical Read Range

Near field ~10 cm

Transitional near field ~ 60 cm

Far field 10+ m

17

Item Tagging @ Metro Galeria Kaufhof

18

2008: Reader Silicon => Many Products

8mm ×

8mm

19

Enter 2009…

• We have a handle on many hard problems– Die sensitivity– Broadband inlay tuning– Inlay orientation insensitivity– Tag interference rejection– Reader sensitivity, selectivity, speed

20

Best Die Sensitivity

2003 2005 2007 2009 2011

– 8

– 12

– 15

– 17???

Sens

itivi

ty (d

Bm

)

Sensitivity measured at a usable Rp < 1.5kΩDates are at the end of the indicated calendar yearVendor variation (best to worst performing) ~ 3dB

21

Why Sensitivity Matters• A cart containing 30 difficult-to-read items

– 36-pack soft-drink cans– 32-pack bottled water– Two large roasts– Aluminum foil– Multiple foil-lined bags– Lots more…

15 20 25 30 3550%

90%

99%

99.9%

Reader Transmit Power (dBm)

Pro

babi

lity

of A

ll Ta

gs R

ead

Cumulative Distribution Function as a Function of Reader Transmit Power

Perc

enta

ge

15 20 25 30 350.5

0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

1

Reader Transmit Power (dBm)

Prob

abili

ty o

f All

Tags

Rea

d

Cumulative Distribution Function as a Function of Reader Transmit Power

Perc

enta

ge

22

Broadband Inlay Tuning• Notice that the free-space tuning is high

– Tuning “pulls in” to 860 – 960 MHz for typical objects

• Other tags achieve similar range on metal/liquids– May require a 1 – 2 mm spacer

• The Propeller is a single-dipole tag– Dual-dipole designs add orientation insensitivity

Data courtesy Metro Group

23

Orientation Insensitivity

13dB

16dB

19dB

22dB

25dB

-150

-120

-90

-60

-30

0

30

60

90

120

150

180

Power Margin Pattern Response for inlay X41

865MHz915MHz954MHz

-15dB

-10dB

-5dB

0dB

-150

-120

-90

-60

-30

0

30

60

90

120

150

180

Normalized Pattern Response for inlay PH16

860MHz910MHz960MHz

Small inlay (4×4 cm)

2007 2009

24

Interference Rejection

1050-5-10

20

15

10

5

0

RF Power [dBm]

SIR

R[d

Bm]

Monza1aMonza2Monza3

Product

Scatterplot of SIRR[dBm] vs RF Power [dBm]Mode = DRM M=4 LF = 256KHz, Interference = CCI (Co-Channel Interference), Offset[MHz] = 0

1050-5-10

20

15

10

5

0

RF Power [dBm]

SIR

R[d

Bm]

Monza1aMonza2Monza3

Product

Scatterplot of SIRR[dBm] vs RF Power [dBm]Mode = DRM M=4 LF = 256KHz, Interference = ACI Adjacent Channel Interference, Offset[MHz] = 0.5

1050-5-10

10

8

6

4

2

0

-2

RF Power [dBm]

SIR

R[d

Bm]

Monza1aMonza2Monza3

Product

Scatterplot of SIRR[dBm] vs RF Power [dBm]Mode = DRM M=4 LF = 256KHz, Interference = Alternate Channel Interference, Offset[MHz] = 1

1050-5-10

10

8

6

4

2

0

-2

RF Power [dBm]

SIR

R[d

Bm]

Monza1aMonza2Monza3

Product

Scatterplot of SIRR[dBm] vs RF Power [dBm]Mode = DRM M=4 LF = 256KHz, Interference = FCI (Far Channel Interference, Offset[MHz] = 5

0 Hz offset

0.5 MHz offset

1 MHz offset

5 MHz offset

25

+30dBm Reader Tx power

R=>T link path loss

T=>R link path loss

–15dBm

–70dBmNext-gen (2009 tags) sensitivity = –74dBm

Tag Distance from Reader

Pow

er (

dBm

)

–74dBm

Tag loss ~10dB

Required reader sensitivity = –70dBm

Reader Sensitivity RequirementsIf reader meets sensitivity requirements, then link is limited by tag sensitivity

26

700 items moving at 1.5m/s

2009: Best Reader PerformanceParameter Performance Conditions

Receive Sensitivity –80 dBm 30dBm Tx, 8dB return loss, M=4

Interference rejection

74 dB CW interferer, adjacent channel66 dB Mod interferer, adjacent channel80 dB CW interferer, 2nd adjacent channel

Max. throughput 1000 tags/sec FCC, quiet environmentTyp. throughput 380 tags/sec FCC, 5 nearby readers

Speedway reader today can achieve >99% inventory accuracy on these pallets

27

Looking Forward

• Two biggest unsolved problems– Read-zone confinement– Consumer privacy

• Two solvable but open problems– Tag security and authentication– Inlay antenna design methodology

• Two problems being solved now– Battery, sensor, and I/O enabled tags– Combining RFID and EAS functionality on a tag

28

Read-Zone Confinement

• Stray reads have caused some retailers to abandon RFID

• How do we contain the RF field?– Antenna design?– RF phase?– Tag RSSI?– Statistical techniques?– Other?

• Even tag ranging is not good enough– Customers want to

confine the RF field in near-arbitrary shapes and sizes

Stray Zone

Read Zone

29

Consumer Privacy• An erroneous privacy example: “Police…

[will be] able to walk around with RFID readers and collect the serial numbers from people’s clothing…”

• IEEE Spectrum, July 2004

A cartoon from 2004 Educating consumers

in 2008

30

Security and Authentication

• Not hard to implement a security algorithm on a Gen2 chip– Early versions of the Gen2 spec (from

2004) proposed security!

• Plus, Gen2 tags have good RNGs– Rqmn’t: An RN16 drawn from a Tag’s

RNG 10 ms after powerup shall not be predictable with a probability greater than 0.025%

• Test data from 1 million RN16s– Predictability: Follows binomial distb’n– Correlation coefficient <0.006

• >100× better than needed to meet 0.025% prediction rqmn’t

-0.5 0 0.5 1 1.5 2 2.5

0.02

0.04

0.06

0.08

0.1

frac

tion

of b

ins

with

freq

uenc

y f

frequency per bin (f) normalized to 1/(2**16-1)

Theory

0 2 4 6 8 10 12 14 16-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1correlation coefficient

delay between samples

corr

elat

ion

coef

ficie

nt

Measured cc

Required cc

31

Inlay Design

• Many antenna designers try to apply their linear RF training to inlay design– But an RFID chip is not a linear system

• The proper methodology isn’t difficult1. Find optimum antenna impedance2. Design inductive loop3. Design radiating element4. Couple radiating element to loop5. Verify experimentally6. Iterate

• A well-designed inlay maintains a good impedance match with changes in– Frequency– Host materials– Nearby tags

Antenna

Chip

Chip || Antenna

Optimum Zs

1dB mismatch loss

2dB mismatch loss

32

Predictions

Problem Observations Prediction Requires research?

Read-zone confinement

• Unlikely to find a “silver bullet”

• A layered solution that combines physical and statistical techniques

Yes

Privacy• The problem is not

nearly as bad as the hype suggests

• Password-free tag anony- mization & range reduction will be “good enough”

No

Security & authentication

• Need to evaluate the trade space of security vs cost

• Add challenge-response security for those who need it

Yes

Inlay-design methodology • A black art today…

• Will remain an art until the academic community teaches inlay design

Yes (education)

April 27, 2009 Chris DiorioImpinj, Inc.

[email protected]

Questions?