Content uploaded by Moo Sung Chae
Author content
All content in this area was uploaded by Moo Sung Chae
Content may be subject to copyright.
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 57, NO. 10, OCTOBER 2009 2597
Wideband Communication for Implantable
and Wearable Systems
Mehmet Rasit Yuce, Member, IEEE, Ho Chee Keong, and Moo Sung Chae
Abstract—This paper presents the feasibility of applying an
ultra-wideband (UWB) wireless scheme to both high data rate im-
plantable neural recording and low data rate wearable biomedical
applications. An extensive analysis on the UWB signal generation
for biomedical application is discussed. A CMOS UWB trans-
mitter has been designed, fabricated, and used in a high data
rate neural recording system. A method to readily assemble a
flexible UWB transmitter for wearable physiological monitoring
system is also presented. An eight-channel low data rate recording
system for monitoring multiple continuous electrocardiogram
and electroencephalogram signals has been designed, and its test
results are presented.
Index Terms—CMOS pulse generator, implantable electronics,
ultra-wideband (UWB), wireless body area network (WBAN).
I. INTRODUCTION
THERE HAS been a growing trend of employing wireless
technologies for measuring physiological signals for both
implantable and wearable systems. Two such areas that are of
significant interest for deploying wireless technologies are the
implantable multichannel neural recording system and wearable
wireless body area network (WBAN).
Advances in microelectrode arrays (MEAs) and multichannel
neural recording systems [1]–[6] improved the quality of the sig-
nals recorded, but at the expense of higher numbers of channels.
Those systems, due to a large number of recording channels, nat-
urally produce a huge amount of data that should be transmitted
out of body to be processed or analyzed further. For example,
a 128-channel neural recording system with 8-bit resolution at
the Nyquist sampling rate generates 20 Mb/s when recording
extracellular action potentials whose signal energy spectrum is
up to 10 kHz. As the number of channel increases, the data rate
will reach a speed of 100 Mb/s and higher. This large amount of
data is desired to be transmitted to a receiver that is located out-
side the body wirelessly. Using wires is not preferable in neural
recording applications because those wires restrict the move-
ment and behavior of animals or humans to be observed, but
Manuscript received March 13, 2009; revised June 23, 2009. First published
September 18, 2009; current version published October 14, 2009. The work of
M. R. Yuce and H. C. Keong was supported by the Australian Research Council
(ARC) under Discovery Projects DP0772929.
M. R. Yuce and H. C. Keong are with the School of Electrical Engineering and
Computer Science, University of Newcastle, Callaghan, N.S.W. 2308, Australia
(e-mail: mehmet.yuce@newcastle.edu.au; chee.ho@newcastle.edu.au).
M. S. Chae is with the Department of Electrical and Computer Engineering,
University of California at Santa Cruz, Santa Cruz, CA 95064-1077 USA
(e-mail:gomdori07@gmail.com).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TMTT.2009.2029958
also increase the possibility of infections. Furthermore, wire-
less capability is also preferred at the cosmetic point of view.
The allowable channel bandwidth of the medical implant
communication service (MICS): a wireless scheme approved
for use in an implantable environment is only 300 kHz; it is un-
able to support the high data rate required. Current state-of-art
wireless neural recording chip [4] is able to provide recording
for 100 channels, but due to bandwidth limitation, it can only
support one channel of raw data transmission. Therefore, it
is quite obvious that there is an immediate need for higher
bandwidth data transmission for neural recording telemetry
systems.
As the FCC assigned the spectrum from 3.1 to 10.6 GHz
for unlicensed use of UWB devices to support high data rates,1
it provided an opportunity for a wideband wireless telemetry.
Impulse radio UWB (IR-UWB) uses simple short pulses for
sending data, it makes the transmitter design very simple, small
area, and low power. This paper presents an IR-UWB trans-
mitter that is designed in National’s 0.35- m CMOS process
and fabricated for neural recording application [7], [8].
Apart from implantable, deployment of wireless technology
for wearable medical monitoring has improved patient quality
of life and efficiency of medical staff. Aging population,
shortage of medical staff, and high demand of hospital re-
sources are problems faced in many countries around the
world. A wireless medical sensor network facilitates remote
monitoring and allows the medical staff to detect early symp-
toms. Timely medical intervention significantly improves the
patient’s chances of recovery. Several proprietary solutions
based on Bluetooth, ZigBee, and wireless local area networks
(WLANs) are available, but they are not optimized for body
sensor networks and lacks interoperability. Therefore, there
is a need for standardization to ensure interoperability and
provides an optimized solution for WBANs. A task group
(IEEE802.15.6) was formed in November 2007 to develop a
standard for WBANs. The targeted data rate for WBANs is
10 Mb/s. Low data rate UWB is one of the potential candidates,
to overcome the bandwidth limitations of current narrowband
systems, and to improve the power consumption and size. An-
other advantageous is that UWB does not have any interfer-
ence on other wireless medical devices when used in a medical
environment due its low-power transmission. An eight-channel
wearable physiological signals monitoring system has been de-
signed and tested to show the feasibility of low data rate UWB
for a WBAN application.
1Federal Communications Commission (FCC), Washington, DC. [Online].
Available: http://www.fcc.gov
0018-9480/$26.00 © 2009 IEEE
Authorized licensed use limited to: University of Newcastle. Downloaded on October 6, 2009 at 19:12 from IEEE Xplore. Restrictions apply.
2598 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 57, NO. 10, OCTOBER 2009
Fig. 1. Narrow pulse generator.
II. UWB SIGNAL GENERATION FOR LOW-POWER
BIOMEDICAL CIRCUITS
There are two common transmitter categories used for UWB
technology. The transmitters in the first category include a pulse
generator and an up-converter that uses a mixer and a local oscil-
lator (LO) to convert the baseband signal into the UWB band [9].
The transmitters in the second category consist of a short pulse
generator and a pulse-shaping circuitry that makes the band-
width of the generated pulses directly fall in the UWB band.
In these transmitters, there is no need for a mixer and an LO,
significantly reducing the complexity and power consumption
of the transmitter [10]. Since the transmitter in our application
does not require a complex multiple-access communication and
the power consumption is the most critical design specification,
the second type of transmitter design technique is preferred for
biomedical system.
A. UWB Pulse Generation Techniques
There are various methods to generate UWB pulses. Among
all methods, using the delay-and-AND gate or delay-and-XOR
gate is the least complex way in CMOS integrated circuit (IC)
technology [11]. The delay unit can be realized using digital
gates such as inverters, analog differential delay cells [12], flip-
flops, and controllable capacitors [13]. A general scheme for
such pulse generations is given in Fig. 1. Herein we apply a
bandpass filter as a pulse shaper due to its simplicity, which is
desired for low-power biomedical applications.
B. Analysis of Band-Limited UWB Pulses
In this section, we will first analyze pulse generation schemes
in both the time and frequency domains. The method described
here can be applied to different applications in order to meet the
spectral mask of the UWB band.
The data signal and the delayed replica are passed
through an XOR gate or an AND gate to obtain a UWB narrow
pulse (e.g., . ). A narrowband square
wave can be represented by
(1)
where is the bit period and
(2)
where is the amplitude of the pulse and is the width of the
(3)
Fig. 2. Timing diagram for UWB pulse generation.
Fig. 3. Effects of null on UWB. (a) Signal with null. (b) Signal without null.
UWB pulses obtained from the delay element, as depicted in
Fig. 2. Note that when code schemes like Manchester nonreturn
zero (NRZ) is used for data, there will always be a transition
repeated every bit period . Assuming there is a square pulse
repeated in every bit period, the Fourier series of the signal in (1)
is given as (3) [14]. The signal includes a dc term and the fun-
damental frequency together with harmonic frequencies. The
least complex method of generating a UWB compliant pulse
is passing the narrow square pulse through a pulse-shaping cir-
cuitry, as shown in Fig. 2. The pulse-shaping circuitry can be a
high-pass or bandpass filter. Since the square pulse is band-lim-
ited after passing through a bandpass filter, it can be represented
by (4) as follows:
(4)
where , , , and and are
the lower and upper cutoff frequencies of the bandpass signal.
The band-limited signal described by (4) assumes ideal zero
rise and fall times, which is applicable for low-frequency system
analysis, but in a high-frequency 1 GHz scenario, finite rise/
fall time causes an additional null in the band of interest, as
shown in Fig. 3. A null occurring in the band of interest may
cause the signal to be unstable after going through a bandpass
filter. In the worst case scenario, as illustrated in Fig. 3(a), when
a deep null occurs at the center frequency [see Fig. 3(a)] of the
transmission band, it causes the noise floor to increase signifi-
cantly due to excessive ringing.
Authorized licensed use limited to: University of Newcastle. Downloaded on October 6, 2009 at 19:12 from IEEE Xplore. Restrictions apply.
YUCE et al.: WIDEBAND COMMUNICATION FOR IMPLANTABLE AND WEARABLE SYSTEMS 2599
A realistic pulse with finite rise and fall times can be repre-
sented by (5) as follows:
(5)
(6)
where (7)
A finite rise time adds an additional null to the circuit. A null
due to rise/fall times occurs every , the null arises from the
pulsewidth and appears every interval. Therefore, the
format of the transmitted pulse should be arranged to ensure that
the null does not appear at the frequency band of interest. It is
desirable to keep the spectrum peak of the sinc envelope at the
center frequency. Satisfying these two criteria would enhance
the signal-to-noise ratio (SNR) of the system.
The rise time can be optimized using (6), where 0.5 corre-
sponds to a 4-dB loss. The peak of the spectrum (i.e., the sinc
envelope) can be put at the center frequency by using (7), where
represents the sidelobes, “1” is the main lobe and “3” is the
first sidelobe. Optimizing the nulls and peak of the spectrum en-
hances the performance of the transmitter, which is achieved by
adjusting the pulsewidth.
Apart from controlling the pulsewidth and rise time, the
power spectrum is optimized to meet the UWB emission mask
by controlling both and . Using (5), the amplitude of the
spectral lines become inversely proportional to their frequen-
cies (i.e., amplitudes ). Lower data rates contain
more spectral lines, but are lower in amplitude. For example,
for a system with a 2-GHz bandwidth spanning from 3.1 to
5.1 GHz and a data rate of 100 Mb/s, the maximum possible
number of spectral lines is 21 and the amplitude is multiplied
by a factor of , where ranges from 30 to 51, while a
system with a data rate of 10 Mb/s contains 200 spectral lines
and ranges from 300 to 500. Therefore, the FCC emission
limit of 41 dBm/MHz can be met by either increasing the
pulse repetitive frequency (PRF) or increasing .
Maintaining the data rate, while increasing the PRF (i.e.,
sending multiple pulses per data bit), results in a processing
gain, and therefore, eliminates the need for a power amplifier,
but high PRF system is limited by the FCC average power re-
quirement, and therefore, is not optimized for a communication
range. To optimize the communication range, the transmitted
power should be close to both the peak power limit and average
power limit. The relationship between the FCC’s peak limit
and average limit for a 1-MHz bandwidth is illustrated in
(8). The FCC peak limit for 1-MHz resolution bandwidth is
17 dBm and the average limit is 41-dBm effective isotropic
radiated power (EIRP). The pulsewidth for this system is
determined by the bandpass filter. For example, a pulsewidth
of 1 ns, the optimal PRF to achieve a maximum peak power
limit of 20 W for this system is 3.75 MHz. For a low PRF
Fig. 4. Simplified block diagram of typical multichannel neural recording sys-
tems.
of 3.75 MHz, the processing gain is low. Therefore, a power
amplifier is required to boost the transmission power (higher
) in order to be near to the emission limit. The use of a power
amplifier would increase the power consumption. Therefore,
there is a tradeoff between lower power consumption with the
use of the processing gain and that of the communication range.
(8)
III. UWB TRANSMITTER FOR NEURAL RECORDING
A general multichannel wireless neural recording system
consists of preamplifiers, filters, an analog multiplexer, a second
amplifier, an analog-to-digital converter (ADC), and a wireless
telemetry, as shown in Fig. 4. Preamplifiers are usually designed
to have input dc blocking capacitors due to a large dc offset
voltage present in electrode–electrolyte interface. The cutoff
frequency of the low-pass filters following the preamplifiers
are set to 10 kHz for antialiasing because the signal spectrum
of the extracellular action potential can be as high as 10 kHz.
Time multiplexing is performed by the analog multiplexer and
the second amplifier provides additional gain for the proper
operation of the following ADC. The overall gain of the am-
plifiers is set to over 60 dB for the recording of extracellular
action potentials whose typical amplitude is around 100 V.
Since all the recording channels share one second amplifier,
the power consumption of the second amplifier is not critical
even though it should have relatively high bandwidth and slew
rate compared to the preamplifiers. The ADC converts the
time multiplexed analog signal into digital data and then the
digitized bit streams are fed to the wireless telemetry.
The resolution of the ADC is determined based on the SNR
of the signal to be recorded, which is mostly affected by the
noise from the recording electrodes [15]. For the extracellular
recording with MEAs whose electrode–electrolyte interface
impedance is 1 M , the SNR is around 40 dB, resulting in a
conservative 9-bit resolution of the ADC [8]. The sampling rate
is set to 40 ksample/s/channel, which is four times higher than
the cutoff frequency of the low-pass filters. Therefore, the raw
data rate produced by the analog front-end with 128 recording
channels is 46 Mbit/s. However, this raw data should be ex-
panded with redundant bits for channel separation and other
purposes required by the wireless communications. Therefore,
for 128 channel neural recording systems, the UWB telemetry
Authorized licensed use limited to: University of Newcastle. Downloaded on October 6, 2009 at 19:12 from IEEE Xplore. Restrictions apply.
2600 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 57, NO. 10, OCTOBER 2009
Fig. 5. Block diagram of IR-UWB transmitter for wireless telemetry.
should be able to support a data rate up to 100 Mb/s for the
lossless simultaneous recording [7], [8].
Power consumption is also critical in implantable devices be-
cause the excessive heat produced by the device can easily give
damages to the tissue surrounding the implanted devices. Even
a temperature increase by 1 C will cause serious heat damages
for cortical signal recordings. The wireless telemetry unit is the
most power-consuming circuit in wireless neural recording sys-
tems and for this reason, IR-UWB is the best choice because the
IR-UWB transmitter consumes very small power and is easy to
integrate on-chip by the nature [8].
A. CMOS UWB Transmitter Design
For pulse generation, the method described in Section II is
followed. Fig. 5 is the block diagram of the IR-UWB trans-
mitter. The first stage of the transmitter is an encoder. The en-
coder is used to convert the baseband data into different formats
(e.g., NRZ Manchester). It can also be used to enable the re-
ceiver to recover clocks directly from the encoded data, as well
as to distinguish the data from different channels. After config-
uring the transmitter to a modulation scheme, the encoded data
is then passed to a narrow pulse generator. The pulse generator
circuit used is shown in Fig. 6. In this circuit, a pulse width is
adjusted from 200 to 900 ps by the external control voltage .
Generated pulses are passed through the pulse-shaping filter to
fit them into the FCC emission mask and to eliminate the trans-
mission of unnecessary frequencies. Unlike other UWB appli-
cations, power amplifiers are not necessary due to the low trans-
mitted power and the short-distance range in neural recording
applications. Instead, a wideband matching filter (acting as a
bandpass filter) is used to regulate the transmitted power. The
matching circuit together with the parasitic components from
the output pad and the package pin contributes extra deriva-
tion (e.g., further filtering in frequency domain) to the UWB
Gaussian pulse that makes the signal more suitable for an UWB
transmission [10], [12].
Our transmitter can be configured to different pulse modu-
lation schemes: on–off keying (OOK), pulse-position modula-
tion (PPM), and binary phase-shift keying (BPSK). A signal
OOK_in is generated by passing the NRZ and Manchester NRZ
baseband signals through an AND gate. As shown in Fig. 7, when
the signal OOK_in is given to the pulse generator circuit de-
picted in Fig. 6, will be an OOK modulated signal [see
Fig. 7(d)]. During bit “1,” a pulse is transmitted, and meanwhile
there is no pulse during bit “0.” The PPM UWB signal is gener-
ated by using the Manchester NRZ signal and the OOK UWB
Fig. 6. Circuit used for pulse generation and modulation selection.
Fig. 7. Time diagram for PPM and OOK modulation.
Fig. 8. Fabricated UWB transmitter chip.
signal. First the Manchester NRZ is passed through the pulse
generation circuit. The resulted narrow pulses are added (using
OR gate) to the OOK UWB [see Fig. 7(d)] to obtain PPM UWB
in Fig. 7(e). As can be seen, the pulse position for bit “1” is
different than that of bit “0.” The bits are positioned such that
the bit detection will be easier at the receiver site. To generate a
BPSK signal, the pulse is simply inverted by 180 when the bit
is “0.”
B. Circuit Measured Result
The transmitter circuit in Fig. 5 is designed, laid out, and fab-
ricated using National’s 0.35- m CMOS process and the chip
photograph is shown in Fig. 8. The circuit consumes 24 A for
a 5.12-Mb/s OOK UWB signal generation and 243 A for a
100-Mb/s OOK UWB signal generation at the power supply
Authorized licensed use limited to: University of Newcastle. Downloaded on October 6, 2009 at 19:12 from IEEE Xplore. Restrictions apply.
YUCE et al.: WIDEBAND COMMUNICATION FOR IMPLANTABLE AND WEARABLE SYSTEMS 2601
Fig. 9. Spectrum of a transmitted UWB signal.
Fig. 10. Outputs time waveforms measured with oscilloscopes.
level of 3.3 V. During the PPM modulation, the circuit con-
sumes about twice as that of OOK because both “0” and “1”
require pulse generation. Although PPM consumes more power
than OOK in this IR-UWB implementation, it has an advan-
tage of easier synchronization at the receiver side because every
symbol always has a pulse.
Unlike other UWB applications, in our neural recording
system, the implanted device does not require a receiver circuit.
The transmission is unidirectional, as the information needs
to be recorded and monitored only. This greatly simplifies the
complexity of the implanted device and increases its lifetime.
Pulse shaping is performed by sending the generated pulses
through a 4-GHz bandpass filter (BPF) that has a bandwidth of
1 GHz. The measured UWB compliant spectrum at the output
of the transmitter is shown in Fig. 9 and the time waveform
at the output of the transmitter with an oscilloscope is pre-
sented in Fig. 10, respectively. The inter-symbol interference
(ISI), which is heavily dependent upon the characteristics of
the package, had a critical effect on the maximum data rate
achievable at the transmitter side in our experimental setup. On
the receiving end, the high-speed receiver is built from off-shelf
components using high-performance RF ICs (i.e., low-noise
amplifiers (LNAs), mixers) and high-speed field-programmable
gate arrays (FPGAs). Details of the receiver prototype system
developed to test the concept defined in this paper will be
covered in Section IV. Fig. 11 shows the measured result of the
Fig. 11. Generated PPM pulses with demodulated result.
Fig. 12. Power level arrangement for transmitted UWB signal in body.
generated PPM pulses together with the receiver demodulation
clock and the demodulated data. UWB chip antennas (Fractus
fr05–107) with the physical dimension of 10 10 0.8 mm
have been used for both Tx and Rx. The result is obtained based
on the transmission distance of 1 m in a direct line-of-sight
environment. Another test is carried out by enclosing the
transmitter antenna in salt-reduced corned beef silverside to
simulate the implantable environment. Fig. 12 illustrated the
result of the test, the deeper the penetration depth (i.e., the
deeper the transmitter is inserted into the human body), the
higher the transmission power required. Therefore, implantable
devices should be able to transmit more than 41 dBm/MHz
while meeting the FCC requirement.
C. Receiver Implementation
The received signal is passed through a bandpass filter
centered at 4 GHz with 1-GHz bandwidth to remove the nar-
rowband interfering signals. Three LNAs, each with a gain
of 14 dB, is used to amplify the signal, which gives a total
gain of 42 dBm. The amplified signal goes through a diode
detector, which translates the high-frequency components to
their low-frequency components. The recovered low-frequency
components are passed through a low-pass filter to form an
envelope of the data signal. The signal is further amplified by
an op-amp circuit by 60 dB before sending to the ADC.
The recovered data is converted to digital signal using
12-bit ADC embedded in the Stratix II development board.
Authorized licensed use limited to: University of Newcastle. Downloaded on October 6, 2009 at 19:12 from IEEE Xplore. Restrictions apply.
2602 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 57, NO. 10, OCTOBER 2009
Fig. 13. Multichannel UWB monitoring system block diagram.
The demodulation is carried out with an FPGA chip running
at 100 MHz (capable of demodulating data of 50 Mb/s). The
transmitter sends data in block of eight pulses with an addi-
tional pulse as a start bit. There is a blank interval between each
transmission, as shown in Fig. 10. Demodulation is performed
by synchronizing the demodulation clock with the start bit.
Once synchronized, the FPGA will generate eight clock cycles
that is aligned with pulse position for bit “1.” Pulses that occurs
within the demodulation clock period is bit “1,” those outside
are bit “0.” After the eighth cycle, the FPGA will goes into
sleep mode and wake up just before the next start bit.
IV. EIGHT-CHANNEL WEARABLE MEDICAL
MONITORING SYSTEM
In this section, an eight-channel UWB recording system for
monitoring of multiple continuous electrocardiogram (ECG)
and electroencephalogram (EEG) signals, as shown in Fig. 13,
is presented. One of the major differences between an im-
plantable and wearable body sensor monitoring system is the
location of the receiver. The receiver for an implantable device
is typically placed at a fixed location near to the where the
transmitter is implanted, but for a wearable monitoring system,
the distance between the transmitter and receiver is constantly
changing due to movement of the body. As the distance gets
further or when the direct path is blocked, more gain is re-
quired for the signal to be transmitted reliably. Therefore, the
transmitter need to adjust the gain constantly to ensure reliable
transmission is achieved. Gain can be increased using either
by sending more pulses per bit or increasing the pulse en-
ergy. A pulse generator that ultilizes the band-limited concept
described in Section II-B, to adjust the gain by changing the
PRF and amplitude is assembled using readily available com-
mercial-off-the-shelf (COTS) components. A microcontroller
PIC18F2320 is included in the transmitter to perform the task
of multiplexing, analog to digital conversion, and adding the
control bits. The use of COTS components helps to reduce the
prototyping time and allows for greater flexibility in changing
the transmitter configuration to adapt to various applications.
A. UWB Pulse Generator for Wearable System
Fig. 14 shows the block diagram of the proposed COTS
pulse generator. The pulse generator is able to generate pulse
width ranging from 300 ps to 4 ns by adjusting the VDDB1
Fig. 14. Block diagram of proposed COTS UWB pulse generator.
Fig. 15. UWB pulse generated by COT transmitter.
and VDDB2. The propagation delay of a buffer is given in (9)
[16]. The load capacitance and gain factor ( and )
for both buffers are approximately equal. Therefore, adjusting
the VDD will change the propagation delay, and a narrow pulse
is formed by passing the two signals through an XOR gate. The
adjustable voltage range is as shown in (10), which is dependent
on the specification of the selected components
(9)
where is
(10)
Apart from adjustable pulsewidth, the PRF and pulse ampli-
tude can also be varied to meet the different operating require-
ments. The PRF is varied by changing the oscillator output,
while adjustment of amplitude is performed by changing the
VDDC. Fig. 15 shows the UWB pulse generated using the pro-
posed pulse generator method. As seen, the generated pulse can
be less than 1-ns wide.
B. Analog Front-End and Microcontroller
Typically ECG and EEG signals has amplitude of less than
500 V with frequency less than 100 Hz. The front end of the
transmitter uses an instrumental amplifier (INA321) and an ac-
tive low-pass filter (LTC6081). The input signals are amplified
by 60 dB, INA321 produces a gain of 14 dB, while LTC6081
provides a gain of 46 dB with a cutoff frequency at 100 Hz. The
Authorized licensed use limited to: University of Newcastle. Downloaded on October 6, 2009 at 19:12 from IEEE Xplore. Restrictions apply.
YUCE et al.: WIDEBAND COMMUNICATION FOR IMPLANTABLE AND WEARABLE SYSTEMS 2603
Fig. 16. Transmit data frame.
circuit is running from a 3.3-V source with virtual ground set at
1.25 V.
The microcontroller is operating with an external 20-MHz
crystal. Each analog front-end output is connected to one of the
dedicated pins for the ADC module. The 10-bit ADC is pro-
grammed to sample at 4.8 kHz with each channel sampling at
600 Hz. The 10-bit data from each sample are separated into
2 B, as shown in Fig. 16. Low byte contains six least signifi-
cant data bits (LSBs) and two header bits. High byte contains
the four most significant data bits (MSBs), three channel bits,
and one header bit. The MSB for the low byte is always “1” and
“0” for the high byte. The three channel bits ranges from “000”
to “111,” each representing one of the eight channels. High and
low bytes are sent to the pulse generator through the universal
asynchronous receiver/transmitter (UART) transmit pin, where
a start bit, parity bit, and stop bit is added. The uutput data rate
can range from 1.2 kb/s to 1.25 Mb/s. The maximum data rate
is limited to 1.25 Mb/s because this is the highest UART baud
rate that the microcontroller can support with a 20-MHz crystal
in our current prototype.
C. Data Recovery for Low Data Rate UWB System
Front ends for both the high data rate neural recording system
and low data rate UWB-based wearable sensor system receivers
are similar, but different data demodulation approaches are used
for data recovery. A low data rate UWB transmitter sends mul-
tiple pulses per bit to increase the processing gain. The receiver
is designed to sample at a rate much higher than the data rate.
The information in the bit is determined, only after performing
several samples; this increases the reliability of the system. The
received signal after down conversion and low-pass filtering is
converted to a digital signal using a 12-bit ADC embedded in
the Stratix II development board. The demodulation is carried
out with an FPGA chip running at 100 MHz. The receiver is ca-
pable of demodulating data of 50 Mb/s.
The demodulation process is performed in two stages. In the
first stage, the demodulator waits for the start signals to arrive,
which is a low bit. Once the low bit is received, the demodu-
lator samples the start bit until half of the period before sending
a signal to start the demodulation clock. In the second stage, the
demodulation clock goes high once the start signal is received.
The first bit of data, which is the start bit, will be sent to the
serial port. The demodulation clock runs at the same rate as the
input data and samples the data at the middle of the bit. After
the demodulation clock has generated 11 clock cycles, it goes
low and wait for the start signal from stage one. The demod-
ulated signal is transmitted through the serial port to the PC.
Fig. 17. Recovered data.
Fig. 18. Multichannel ECG signal.
The purpose of this demodulation structure is to ensure the data
send to the PC is of the correct length to avoid any errors due
to synchronization. This is necessary as multiple pulses per bit
are sent, upon recovery there is a slight offset. The demodulated
data is shown in Fig. 17. A program written using Visual Basic
is used to decode the data; it performs filtering, as well as helps
to displays the received signals on the screen. A parity bit check
is performed on the received data to ensure all data received cor-
rectly. Once the received data is decoded, it is formatted back
into a 10-bit word and separated based on the information em-
bedded in the channel bits (see Fig. 16). Digital filtering is per-
formed on the received signal to remove the 50/60-Hz noise,
which comes from the power supply. The ECG signal in Fig. 18
is successfully monitored in our laboratory environment with
other wireless devices operating. The graphical user interface
(GUI) program can display any eight channels by changing the
button “channel selection” shown in the window.
V. C ONCLUSION
A band-limited UWB pulse generation technique has been
presented and analyzed. UWB transmitters based on the
band-limited technique have been developed using both CMOS
technology and from COTS components. The developed UWB
transmitters are applied to low-power biomedical applications.
The transmitters are suitable for both high data rate and low
data rate applications. In an implantable environment, higher
transmission power is required. For a low data rate system,
there is a tradeoff between the processing gain and communi-
cation range. A multiple-channel EEG/ECG monitoring system
using low data rate UWB technology has successfully been
implemented and tested. The receiver in the prototype success-
fully received and recovered the UWB modulated ECG/EEG
signals. The real time signals are displayed on a PC for body
area network applications.
Authorized licensed use limited to: University of Newcastle. Downloaded on October 6, 2009 at 19:12 from IEEE Xplore. Restrictions apply.
2604 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 57, NO. 10, OCTOBER 2009
REFERENCES
[1] G. A. Demichele and P. R. Troyk, “Integrated multichannel wireless
biotelemetry system,” in Proc. 25th IEEE EMBS Int. Conf., Cancun,
Mexico, Sep. 17–21, 2003, pp. 3372–3375.
[2] I. Obeid, M. A. L. Nicolelis, and P. D. Wolf, “A multichannel telemetry
system for single unit neural recordings,” J. Neurosci. Meth., vol. 133,
pp. 123–135, Feb. 2004.
[3] P. Mohseni and K. Najafi, “A battery-powered eight-channel wireless
FM IC for biopotential recording applications,” in Proc. IEEE Int.
Solid-State Circuits Conf., San Francisco, CA, Feb. 6–10, 2005, pp.
560–561.
[4] R. R. Harrison et al., “A low-power integrated circuit for a wireless
100-electrode neural recording system,” IEEE J. Solid-State Circuits,
vol. 42, no. 1, pp. 123–123, Jan. 2007.
[5] S. O’Driscoll, T. Meng, K. Shenoy, and C. Kemere, “Neurons to sil-
icon: Implantable prosthesis processor,” in IEEE Int. Solid-State Cir-
cuits Conf., San Francisco, CA, Feb. 2006, pp. 2248–2257.
[6] W. Liu, M. Sivaprakasam, G. Wang, and M. S. Chae, “A neural
recording system for monitoring shark behavior,” in Proc. IEEE Int.
Circuits Syst. Symp., May 21–24, 2006, pp. 4123–4236.
[7] M. R. Yuce, W. Liu, M. S. Chae, and J. S. Kim, “A wideband telemetry
unit for multi-channel neural recording systems,” in IEEE Int. Ultra-
Wideband Conf., Sep. 24–26, 2007, pp. 612–617.
[8] M. Chae, W. Liu, Z. Yang, T. Chen, J. Kim, M. Sivaprakasam, and
M. Yuce, “A 128-Channel 6 mW wireless neural recording IC with
on-the-fly spike sorting and UWB transmitter,” in Proc. IEEE Solid-
State Circuits Conf., Feb. 3–7, 2008, pp. 146–603.
[9] J. R. Ryckaert et al., “Ultra-wide band transmitter for low-power wire-
less body area networks: Design and evaluation,” IEEE Trans. Circuits
Syst. I, Reg. Papers, vol. 52, no. 12, pp. 2515–2515, Dec. 2005.
[10] D. D. Wentzloff and A. P. Chandrakasan, “Gaussian pulse genera-
tors for subbanded ultra-wideband transmitters,” IEEE Trans. Microw.
Theory Tech., vol. 54, no. 4, pp. 1647–1655, Apr. 2006.
[11] N. Deparis, C. Loyez, N. Rolland, and P.-A. Rolland, “Pulse generator
for UWB communication and radar applications with PPM and time
hopping possibilities,” IEEE Int. Circuits Syst. Symp., pp. 661–665,
2006.
[12] P. Saha, N. Sasaki, and T. Kikkawa, “A single-chip Gaussian mono-
cycle pulse transmitter using 0.18 m CMOS technology for intra/
interchip UWB communication,” in IEEE VLSI Circuits Symp., Jun.
15–17, 2006, pp. 204–205.
[13] B. Jung, Y. H. Tseng, J. Harvey, and R. Harjani, “Pulse generator de-
sign for UWB IR communication systems,” in IEEE Int. Circuits Syst.
Symp., May 23–26, 2005, pp. 4381–4384.
[14] W. R. Bennett and J. R. Davey, Data Transmission. New York: Mc-
Graw-Hill, 1965.
[15] M. Chae, W. Liu, and M. Sivaprakasam, “Design optimization of inte-
grated neural recording systems,” IEEE J. Solid-State Circuits, vol. 43,
no. 9, pp. 1931–1939, Sep. 2008.
[16] N. R. Mahapatra, A. Tareen, and S. V. Garimella, “Comparison and
analysis of delay elements,” in IEEE Midwest Circuits Syst. Symp.,
Aug. 2002, pp. 473–476.
Mehmet Rasit Yuce (S’01–M’05) received the B.S.
degree in electronics engineering from Ankara Uni-
versity, Ankara, Turkey, in 1997, the M.S. degree in
electrical and computer engineering from the Univer-
sity of Florida, Gainesville, in 2001, and the Ph.D.
degree in electrical and computer engineering from
North Carolina State University (NCSU), Raleigh, in
2004.
He is currently a Senior Lecturer with the School of
Electrical Engineering and Computer Science, Uni-
versity of Newcastle, Callaghan, N.S.W., Australia.
From August 2001 to October 2004, he was a Research Assistant with the De-
partment of Electrical and Computer Engineering, NCSU. In 2005, he was a
Post-Doctoral Researcher with the Electrical Engineering Department, Univer-
sity of California at Santa Cruz. He has authored or coauthored about 50 tech-
nical papers. His research interests include wireless implantable telemetry, wire-
less body area networks (WBANs), and analog/digital mixed signal very large
scale integration (VLSI) for wireless, biomedical, and RF applications.
Dr. Yuce was the recipient of a National Aeronautics and Space Adminis-
tration (NASA) Group Achievement Award in 2007 for the development of a
silicon-on-insulator (SOI) transceiver.
Ho Chee Keong received the B.S. degree in elec-
trical engineering from the University of Newcastle,
Callaghan, N.S.W., Australia, in 2006, and is cur-
rently working toward the Ph.D. degree at the Uni-
versity of Newcastle.
From 2006 to 2007, he was an Electronic Design
Engineer with ET Designers, during which time
he specialized in LCD and touch panel design. His
research interests include UWB communication
schemes, biomedical applications, and system
designs for wireless communications.
Moo Sung Chae was born on June 8, 1975 in Pusan,
Korea. He received the B.S. and M.S. degrees in
electrical engineering from Seoul National Univer-
sity, Seoul, Korea, in 1998 and 2000, respectively,
and is currently working toward the Ph.D. degree at
the University of California at Santa Cruz (UCSC).
In 2000, he joined Samsung Electronics, Seoul,
Korea, where he was involved with circuit design
of high-speed dynamic random access memories
(DRAMs). His research interests are integrated
neural recording systems, neural prosthetic devices,
and retinal prostheses.
Authorized licensed use limited to: University of Newcastle. Downloaded on October 6, 2009 at 19:12 from IEEE Xplore. Restrictions apply.