Sunteți pe pagina 1din 8

RFID TECHNOLOGY

Self-Powered Wireless
Temperature Sensors
Exploit RFID Technology
An embedded temperature sensor combined with passive RFID circuitry
enables a single-chip solution for commercial smart sensors.

E
merging RFID technology lets us A passive RFID solution
embed sensors into a very small chip, Our RFID chip has five parts: an analog front
creating a wireless sensing device. So, end, a digital controller, a charge pump, an
we set out to develop such a single- Eeprom, and the temperature sensor (see figure 1).
chip versatile temperature sensor. We
also wanted to be able to transfer our design to Analog front end
an implantable temperature sensor for an animal An external LC (inductor/capacitor) resonator
healthcare application with induces an AC voltage that generates DC volt-
Karn Opasjumruskit, Thaweesak minimal structural modifica- age through a fullwave rectifier and a system
Thanthipwan, Ohmmarin Sathusen, tion (in compliance with the clock through a clock extractor. An RF limiter
Pairote Sirinamarattana, Prachanart animal identification standard, guards the AC amplitude so that it doesnt
Gadmanee, Eakkaphob Pootarapan, ISO 11785), so we chose a exceed the breakdown operating voltage of the
Naiyavud Wongkomet, Apinunt robust operating frequency transistors. A regulator then uses the generated
Thanachayanont, and Manop between 100 kHz to 150 kHz. DC voltage to create a stable supply voltage for
Thamsirianunt In healthcare applications, the internal circuits. When the chip enters the RF
Silicon Craft Technology sensor generally requires a fields vicinity, the power-on-reset circuit gives
smaller temperature range (~35 a start-up signal to reset the digital controller to
to 45oC) and higher accuracy its initial state. To communicate with a reader
o
(~0.1 C) than in general applications. via the RF link, the chip uses a load modulator
Here, we discuss the implementation of our sen- in a back-scattering manner to send data back.
sor. The fully integrated complementary metal- It uses a demodulator to detect an encoded com-
oxide semiconductor (CMOS) batteryless device mand signal sent from a reader.
measures temperature and performs calibration to
compensate for the sensors inherent imperfections Digital controller
(see the Related Work in Temperature Sensors The controller comprises a state sequencer, bit-
sidebar on page 56). An RF link using passive rate generator, data encoder, and command
RFIDs backscattering technique wirelessly trans- decoder. The sequencer is a finite-state machine
mits the data to a reading device while extracting that handles all main chip operations including
power from the same airwave, letting the device reading and programming memory, passing
operate anywhere and last almost forever. The Eeprom data to the encoder, and timing the tem-
entire microchip, including the temperature sen- perature sensor. The bit-rate generator gives ref-
sor, consumes less than a few microamperes over erence timing for the encoder to use in RF data
a half a second, so the scanning device can capture transmission. The data encoder appropriately
data from longer read distances. We hope it acts encodes an NRZ (nonreturn to zero) data stream
as a model for future low-power smart sensors. from the sequencer before sending it to the load

54 PERVASIVE computing Published by the IEEE CS and IEEE ComSoc 1536-1268/06/$20.00 2006 IEEE
950 m
Analog front end Digital and memory
Electrical
Regulator Demodulator State sequencer Erasable
Data Temperature
Programmable
in sensor
Command decoder ROM
Rectifier
and limiter Modulator Data
out

1,500 m
Data encoder
External coil
and capacitor Power-on- Charge pump Digital
reset circuit Clock control
Clock
Bit-rate generator Eeprom

Clock Charge
Temperature sensor pump Analog front end
IPTAT
Proportional-to-absolute Sigma delta Bitstream Digital
temperature current and IREF analog-to-digital control
reference current generator converter

(a) (b)

Figure 1. The temperature sensor RFID chip: (a) a block diagram and (b) a photo showing each parts actual location. The die area
measures approximately 1.4 mm2.

modulator. The command decoder inter- Temperature sensor emitter-junction voltages) thats pro-
prets demodulated data and tells the The temperature sensor comprises portional to the temperature. The sin-
sequencer to jump to the state being three major blocksa digital-control cir- gle-chip temperature sensor has evolved
instructed by the reader. cuit, a proportional-to-absolute temper- ever since.
ature (PTAT) current generator, and a Our temperature sensor circuitry is
Charge pump sigma-delta analog-to-digital converter based on Widlars concept. The PTAT
The charge pump multiplies the recti- (ADC). We designed the sensor to mea- current generator, directly representing
fied voltage and converts it to the pro- sure industrial-grade temperatures from the temperature, delivers a PTAT current
gramming voltage (Vpp) used to modify 40o to 120oC, with a maximum reso- and a reference current, keeping the lat-
data in the Eeprom. The charge pump lution of 8 bits. ters value constant over the temperature
also incorporates a voltage regulator to range of interest. The sigma-delta ADC
control the programming voltage at a The sigma-delta temperature then processes both currents into the dig-
suitable level during both erase and pro- sensor ital data stream.
gramming processes. Our temperature sensor plays two
major roles. First, it converts tempera- Digital control
Eeprom ture into a current using its PTAT cur- The digital controller handles the ana-
The Eeprom is nonvolatile; it stores rent generator. Then, a data converter log-to-digital conversion after the com-
data even if the power supply is inter- transforms this current into digital infor- mand decoder receives a temperature-
rupted. Normally, it contains a set of mation using the sigma-delta principle reading command. The process involves
configuration data for the digital con- so we call this the sigma-delta ADC. turning the PTAT and reference currents
troller and a unique data ID to serve a on and off, converting the bit-stream
primary RFID functionidentifying its The PTAT current generator data from the sigma-delta ADC, and
uniqueness. Our design includes an To some extent, temperature governs sending the converted temperature value
extended Eeprom section, where we can a semiconductors electrical properties. back to the main digital circuit.
store more useful data depending on the This principle led to the realization of
software and applications. A small part an electronic temperature sensor on a Sigma-delta ADC
of this extra memory stores process- silicon chip. In 1965, Robert Widlar pro- The sigma-delta ADC consists of an
dependent calibration parameters for posed a PTAT circuit, which produces a integrator and a clock comparator (see
each temperature sensor unit. vbe (the difference between two base- figure 2a, which also shows the PTAT

JANUARYMARCH 2006 PERVASIVE computing 55


RFID TECHNOLOGY

Related Work in Temperature Sensors

I n many applications, maintaining an appropriate temperature


level is essential. Rudimentary systems monitor temperatures
by employing copper wires as media, sending an analog or digital
accuracy (~0.1oC) than in general applications. For example, a
read-only RFID sensor, Bio-Thermo,2 was designed for an animal
healthcare application. It comes in an implantable glass-tube form
signal between the sensors and host using data communication and uses a 134.2 kHz carrier frequency and protocol that conforms
protocols. This method, however, is inconvenient when accessibil- to an animal identification standard (ISO 11785). However, it
ity is limitedsuch as in harsh environments or a living body. A doesnt provide a memory space to store user data.
viable solution has been to use wireless communication between In general RFID development, more has been focused on RFID
temperature sensors and an RF interrogator or RFID reader based application-specific integrated circuits using ultra high frequency
on passive RFID technology. This technology eliminates an exter- bands (900 MHz and higher).3,4 However, such circuits are usually
nal power supply from the RFID device, making it cheap and designed for general temperature measurement, not for a specific
maintenance free. application. Also, although UHF RFID offers lower production costs
A complementary metal-oxide semiconductor temperature sen- and longer read ranges, it suffers heavily from electromagnetic-field
sor is a strong candidate for passive RFID because it consumes little absorption near metal or water or in humid environments.5
power, letting us easily integrate it into a passive RFID system.
Using a CMOS process, we can also incorporate signal-interfacing
circuitries (for example, a current-to-voltage converter, an analog- REFERENCES
to-digital converter, and so on) within a single chip to minimize the
systems power consumption, cost, and size. This renders a very 1. New Low-Cost Temperature Sensor, RFID J., July 2002; www.rfidjournal.
com/article/view/28/1/1.
small, self-contained temperature-sensing system that can commu-
nicate with other reading and logging devices via the RF signal. The 2. S.J. Miller-Smith, New Chip Can Read Your Pets Temperature, Dar-
chip might require only an external LC (inductor/capacitor) win Veterinary Center, www.darwinvets.plus.com/topical/biothermo.
htm.
resonator.
Existing sensors such as TempSens1 and TELID (Telemetry ID, 3. F. Kocer, P.M. Walsh, and M.P. Flynn, Wireless, Remotely Powered
http://microsensys.de/unten-sensor.htm) can measure tempera- Telemetry in 0.25 m CMOS, Radio Frequency Integrated Circuits
Symp., IEEE Press, 2004, pp. 339342.
ture from 40 to 85oC at 1oC accuracy, operating at 13.56 MHz.
However, to function as temperature loggerssensing and mem- 4. N. Cho et al., A 8-W, 0.3 mm2 RF-Powered Transponder with Temper-
ature Sensor for Wireless Environmental Monitoring, Proc. Intl Symp.
orizing temperature at specified intervalsthey require external
Circuits and Systems (ISCAS 2005), IEEE Press, 2005, pp. 47634766.
power sources. In healthcare applications, the sensor generally
requires a smaller temperature range (~35 to 45oC) and higher 5. K. Finkenzeller, RFID Handbook, 2nd ed., John Wiley & Sons, 2003.

current [IPTAT] and the reference current S1 opens, so VINT increases again. This jitter. In other words, noise and power-
[IREF]). The converters operation tightly process repeats until the conversion supply rejection requirements of the inte-
correlates with the clock signal. Figure phase ends. The total sum of the deci- grator and the comparator in the sigma-
2b displays the waveforms of signals, sion result (that is, the number of times delta ADC are less stringent when using
depicting the converters operation. that 1 occurs in the decision result) is this overresolution technique. After the
In each clock period, the integrator directly proportional to the PTAT cur- ADC completes the conversion, the dig-
integrates IPTAT to raise its output volt- rent and proportional to the tempera- ital control gets only the most significant
age (VINT). At the end of each period, the ture. 8 bits from the overall 16 bits to repre-
clock comparator compares the integra- Our prototype sensors incorporated a sent the averaged temperature data; it
tor output (VINT) to the reference volt- 16-bit ADC despite a requisite resolu- neglects the remaining 8 bits.
age. The decision (bitstream) defines the tion of only 8 bits. Given a fixed clock
on and off state of the switch S1 in frequency, the sigma-delta ADC needs Acquiring the temperature
the next period. If the integrator output more time to convert bitstream data into Communication between the RFID
is higher than the reference voltage, the 16-bit resolution than into the 8-bit tag and reader is based on a simple pro-
decision is 1 and S1 is closed, allow- counterpart. At a glance, the extended tocol. After initialization, a tag auto-
ing IREF to flow. This current will reduce time seems like overkill, but it helps aver- matically sends its modulated serial ID
VINT until the comparator detects that age out several imperfections of the tem- repeatedly until a companion reader or
VINT is falling below the reference level. perature sensor circuit, such as power scanner interrupts the RF field with a
Then, the decision is 0 and the switch supply noise, thermal noise, and clock command. The tag then executes the

56 PERVASIVE computing www.computer.org/pervasive


Usage Model

command and sends the requested data


(see the Usage Model sidebar).
W arehouses, cold-storage facilities, and produce distributors can have miniatur-
ized RFID tags attached inside styrofoam packaging, on a tray that holds the
food, or in a freezing chamber. To read the tags temperature, the RFID reader beams
To probe for a targets temperature, a the RF field to power up an RFID chip, simultaneously sending a command to activate
companion reader or scanner sends a the microchip and wirelessly collecting the data from it. A single operation can read
command to the RFID tag using a 100 both the manufacturer lots ID and the goods temperatureanytime, anywhereto
percent amplitude shift keying (ASK) or ensure that each lot has been treated and maintained at the proper temperature.
on-off-keying modulation (see figure 3).
In the tag, the demodulator detects the
RF envelope and converts this envelope read temperature command, the dig- the temperature sensor is inactive or in
signal into digital data. The command ital circuit performs only a basic RFID the power down mode.
decoder then decodes this data to deter- operation, such as reading from or writ- If the sensor unit receives a read tem-
mine the command type. If its not a ing to the tags memory. In this mode, perature command, it converts the tem-

Figure 2. The sigma-delta converter: (a) a block diagram and (b) signals at each converter node.

IREF Reference level

VINT
S1 Bitstream
(to digital)
VINT

Bitstream
IPTAT
Integrator Clock
Clock Clock
comparator
Time
(a) (b)

Figure 3. The RFID reader generates the RF field that powers the RFID tag. After initialization, the reader sends a command to
invoke the temperature-sensing operation.

Power Eeprom read + Demodulate and Turn on temperature Data encode and modulate
on reset modulator command decode sensor (ASK Manchester, RF/64)

Reader's
RF field
off

off

1 1 1 0 0 1 0 0 1 0

Tag's
RF field

Temperature Frame Frame


Tag's ID sensor enable synchron- 8-bit temperature synchron-
data
operation code izer izer
Frame data
Measurement time 0.51s

JANUARYMARCH 2006 PERVASIVE computing 57


RFID TECHNOLOGY

0 1

Operation
command code Frame Frame
synchronizer 8-bit data RF/64, Manchester synchronizer
(4 slots)

(a) (b)

Figure 4. The captured waveforms at a tags antenna: (a) The read temperature command interrupts normal data modulation, and
the temperature sensor is later turned on. (b) At the end of the sensors operation, a tag transmits framed data repeatedly.

perature into 16-bit output data, repre- times in each frame, mimicking a 32-bit method is to eliminate the errors using
senting a value between 60 and 135C. data read from a block of the Eeprom calculations performed in the digital
However, the actual sensing range is nar- organized in 32 bits  32 blocks. The domain, or digital calibration.
rower owing to the sigma-delta ADCs modulated data frame repeats until the For our initial study, we used a one-
inherent inaccuracy. The temperature chip receives a new command. Through point digital calibration method because
sensor can precisely sense the tempera- postprocessing firmware or a supported its simple yet well understood. This
ture from 40 to 120C. software application, users can read the method can cancel an offset error at the
During the temperature measurement temperature measurement after the reference point used for calibration but
and data conversion phase, the RF field RFID reader decodes the received data. cant eliminate a gain error and nonlin-
isnt modulated to avoid the power sup- earity error.1 So, reading the temperature
ply fluctuation that could deteriorate Errors and calibration near the calibration point produces less
the sigma-delta ADCs accuracy. The All temperature sensors more or less error than reading it from further away
measurement time lasts for 216 clock inherit nonlinear properties, which can points.
periods or approximately 0.51 second cause errors in the output data in terms In our experiment, we left calibration-
at 125kHz. After the conversion com- of the offset and gain. We approximated related duties to the reader unit; the
pletes, the temperature sensor turns off, these errors using the data from the fab- RFID tag just kept calibration data in the
and the converted 16-bit data is stored rication foundry, such as device model- Eeprom. First, we put the RFID tag into
in a temporary register. Finally, the ing and parameter matching. The circuit a well-controlled temperature chamber
main digital circuit serially reads the simulation showed that the worst-case set to a fixed reference temperaturewe
eight most significant bits of stored data output error was 10oC; the main cause used 40C as the calibrating point. The
and encodes them using Manchester of error was transistor mismatches. reader obtains the temperature data
RF/64 (the bit rate is 1/64 of the RF fre- To ensure accuracy, designers should from the tag and compares it to a refer-
quency) encoding. first calibrate each sensor using one of ence value to determine if an error
The modulator sends the encoded several possible schemes. Analog cali- occurred. Then, the reader writes the off-
temperature data through the RF link bration techniques usually relate to set correction data back to the RFID
with ASK modulation (see figure 3). A adjusting the primitive values of the tags Eeprom block allotted to keep the
modulated data frame consists of a selected circuit components or parame- calibration value. In practice, the reader
frame synchronizer and the encoded tersfor example, bias current, resis- must acquire from the RFID tag both
data. The synchronizer pattern defines a tance, capacitance, or even transistor size. temperature data and the calibration
beginning for each frame so that the The designer must routinely tune these value and simultaneously perform sub-
reader can mark the data packages start. combinations until he or she obtains the traction to compensate the offset and
The 8-bit temperature data repeats four correct measurement. A more convenient display the final temperature.

58 PERVASIVE computing www.computer.org/pervasive


Figure 5. Error plots for four prototype
12 2.5
microchips (a) before calibration and (b)

Error before calibration (C)

Error after calibration (C)


10 2.0
after calibration, with one-point digital 1.5
8
calibration at 40C. 1.0
6
0.5
4
0
2
Because were not limited to the sin- 0.5
0 1.0
gle-point technique, we can extend this 2 1.5
calibration method to compensate for a 4 2.0
higher order of errors. For example, if 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
we use two-point calibration, which can (a) Temperature (C) (b) Temperature (C)
calibrate both the offset and the gain
errors, the two calibration parameters Microchip 1 Microchip 2 Microchip 3 Microchip 4
can help determine the measured tem-
perature. The reader or scanner can
write both calibration attributes into can refer to the field strength required at shows the measurement results. Before
each microchips memory. When per- this distance to activate the tags as the calibration, chip-to-chip errors were as
forming a measurement, the reader then least-minimum field strength, or approx- wide as 100C. After a single calibration
computes the measured result on the imately 2.4 A/m2. An RFID system read at 40C, the error became less than
basis of these two-point calibrated val- range depends on many factors, such as 2.5C in a measurement range between
ues, for potentially better accuracy. antenna area (for both the reader and 0 to 100C. This indicates that we need
Generally, when sensors are in service tag), the antennas quality factor, or the to perform two-point calibration for
for a long time, reading accuracy deteri- number of antenna turns. When we better accuracy.
orates owing to errors introduced by the experimented with a larger square reader Figure 6 shows an RFID reader read-
devices inherent long-term drift. Users antenna (80  100 mm), the read range ing the temperature data from an RFID
can recalibrate our temperature sensors improved considerablyto a distance tag in an ice cube. This shows that our
anytime by rewriting the new calibration greater than 25 cm. microchip can operate and communicate
values into the Eeprom via the RF link. To evaluate chip performance, we put in ice- or water-filled packages or food
our RFID tag in a clear plastic package containers, in cold storage, or inside an
Experiments and dipped it in a beaker of water animals body, where RF field absorp-
Using chip characterization data, we (which helps stabilize the temperature). tion is highly likely.
determined that RFID tags consume, on We compared our tags temperature
average, 2A for read operations and data to a temperature read from a com-

W
10A to measure temperatures. We mercially available digital mulimeter e havent yet tested our
designed a companion handheld reader with an optional thermocouple (0.1C sensors full temperature
with a 60 mm-diameter antenna coil (60 resolution). We placed the sealed ther- range but hope to do so in
turns of winding) to perform a tempera- mocouple next to our RFID tag in tem- the near future. We intend
ture-reading experiment. Figure 4 shows perature-controlled water. Figure 5 to perform a detailed chip characteriza-
waveforms at the tags antenna when it
receives a read temperature command.
The average AC current through the
readers antenna is approximately 100
mA. The reader can communicate with
a ring-shaped tag (with a 365-turn, 27
mm-diameter antenna coil inside) at a
maximum distance of 10 cm when both The
antenna planes are perfectly parallel. We microchip

The RFID Tag

Figure 6. Our microchip can operate in


The reader
ice-filled packages: (a) our chip in an ice
cube and (b) the reader measuring the (a) (b)
temperature at the freezing point.

JANUARYMARCH 2006 PERVASIVE computing 59


RFID TECHNOLOGY

Figure 7. The next-generation microchip


Power on reset will perform the calibration algorithm
itself. The calibration reader will send
only the reference value to a tag, and the
tag will do the rest.

ID modulation

temperature with the reference value, cal-


culates errors, and programs them into
the Eeprom. During the read-tempera-
Receive a command ture command, the digital control circuit
loads the calibration data from the Eep-
rom and performs the arithmetic com-
putation to compensate for sensor inac-
Calibration Temperature reading curacies before modulating the corrected
Calibration/
temperature reading temperature back to the reader. So, the
reader doesnt need to perform data post-
processing, which incurs delays associ-
ated with the normal read/write data
Receive start Receive start
temperature command temperature command exchange through the RF field.
Furthermore, embedding the RFID
with other sensory circuitriessuch as a
humidity sensor, pressure gauge, or even
Receive exact Chips measure
temperature temperature
chemical detectorwill enable the dawn
of new age in measurement. Together
with an anticollision protocol (not sup-
ported in our current prototype), these
Chips measure Load error from ubiquitous sensors will use the power
temperature Eeprom
provided by a single RFID reader to con-
vert many physical quantities from mul-
tiple tags into figures that many control
Compare with exact Calculate exact and monitoring applications can under-
temperature (from reader) temperature stand. This emerging technology, dubbed
sensor RFIDs, will supersede traditional
sensor-wired equipment and let us mea-
Send exact temperature
sure environment parameters more
Write error to Eeprom freelywirelessly and without physical
to reader
or visual contact.

End
REFERENCES
1. G.v.d. Horn and J.L. Huijsing, An Inte-
grated Smart Sensor: Design and Calibra-
tion with a temperature chamber, and bration by transferring all calibration tion, Kluwer Academic, 1998
we anticipate accuracy better than 1C steps, which the reader currently per- 2. K. Finkenzeller, RFID Handbook, 2nd ed.,
using two-point calibration. However, forms externally, to the on-chip digital John Wiley & Sons, 2003.
to obtain an accuracy of 0.1C, well control circuit. Figure 7 demonstrates our
need to pay more attention to the ana- proposed autocalibration process. First,
log portion of the overall temperature the microchip receives a start command
For more information on this or any other comput-
sensor unit. and a reference temperature from the ing topic, please visit our Digital Library at www.
We plan to improve microchip cali- reader. Then, it compares the measured computer.org/publications/dlib.

60 PERVASIVE computing www.computer.org/pervasive


the AUTHORS
Karn Opasjumruskit is a Pairote Sirinamarattana Naiyavud Wongkomet is a
member of the research is a member of the research project director at Silicon
staff at Silicon Craft Technol- staff at Silicon Craft Tech- Craft Technology. His re-
ogy. His research interests nology. His research inter- search interests include elec-
include high-frequency and ests include video filters, trostatic micropositioners,
ultra high-frequency RFID sensor-processing units, pipelined and sigma-delta
tags and biosensors. He re- and long-range RFID read- analog-to-digital conversion,
ceived his MEng in electrical ers. He received his MEng in and antialiasing filters. He
engineering from Chulalongkorn University. electrical engineering from Chulalongkorn Uni- received his PhD in integrated-circuit design
Contact him at 196/103 Mu 1 Soi Kosumruam- versity. Contact him at pairote@sic.co.th. from the University of California, Berkeley. Con-
jai, Chaengwattana Rd., Donmueang, Bangkok tact him at naiyavud@sic.co.th.
10210, Thailand; karn@sic.co.th. Prachanart Gadmanee is
a member of the research Apinunt Thanachayanont
Thaweesak Thanthipwan staff at Silicon Craft Tech- is a project director at Silicon
is a member of the research nology. His research inter- Craft Technology. His research
staff at Silicon Craft Technol- ests include computer- interests include complemen-
ogy. His research interests aided-design-generated tary metal-oxide semiconduc-
include low-frequency RFID layout techniques, routing tor RF bandpass filters, active-
tags, microelectromechani- optimization, and physical- inductor circuits, translinear
cal sensors, and sigma-delta device characterization. He received his MEng phase-locked loops and direct
analog-to-digital conversion. in electrical engineering from Kasetsart Univer- digital synthesizers. He received his PhD in elec-
He received his MEng in electrical engineering sity. Contact him at prachanart@sic.co.th. trical and electronics engineering from the
from Chulalongkorn University; Contact him at Imperial College of Science, Technology and
thaweesak@sic.co.th. Eakkaphob Pootarapan is Medicine. Contact him at apinunt@sic.co.th.
a member of the research
Ohmmarin Sathusen is a staff at Silicon Craft Tech- Manop Thamsirianunt is a
member of the research staff nology. His research inter- director of product develop-
at Silicon Craft Technology. ests include anticollision ment at Silicon Craft Technol-
His research focuses on sili- algorithms, digital-circuit ogy. His research interests
con-based sensors. He re- optimization, and encryp- include wireline transceivers,
ceived his MEng in electrical tion techniques. He received phase-locked loops and volt-
engineering from Chula- his BEng in electrical engineering from King age-controlled oscillator and
longkorn University. Contact Mongkuts Institute of Technology Ladkrabang. gigabit I/O. He received his
him at ohmmarin@sic.co.th. Contact him at eakkaphob@sic.co.th. MEng in electrical engineering from Carleton Uni-
versity. Contact him at manop@sic.co.th.

JANUARYMARCH 2006 PERVASIVE computing 61