Department of Communication Engineering by pengxiuhui


									                  Department of Communication Engineering

Title:End-to-End Techniques for Broadband Internet (1/3)
Investigator:Chung-Ju Chang
Sponsor:National Science Council
Keywords:Routing, Multi-Protocol Label Switch (MPLS), DiffServ, Edge Router,
               Core Router, Multi-Field Classifier, Scheduler, Traffic Control, Call
               Admission Control, Usage Parameter Control, Self-Similar

      With the blooming of Internet application services, the Internet traffic flow
increases dramatically and the traffic flow from inter-network transmission also
increases rapidly. Not only the basic data transmission but also some multimedia
services (such as: voice, real-time video services) are also carried on the Internet.
Henceforth, to provide the QoS-provisioning services, the larger bandwidth capacity,
effective QoS-provisioning service framework, and traffic control mechanisms are the
primary requirements in the design of the Internet.
      The three-year integrated project is intended to propose an effective solution to
provide end-to-end QoS guaranteed Internet services and the corresponding research
interestings are: the traffic characteristcs of the Internet services, a QoS-provisioning
service framework and the traffic control mechanisms. The subproject 1 is to study
the self-similarity of traffic characteristcs of the real Internet services and propose an
appropriate traffic model that will be the reference model for other subprojects. The
subproject 2 is to study the high speed Internet routing techniques, the
QoS-provisioning traffic control mechanisms and the network configuration. To speed
up the Internet routing, 2 possible techniques will be proposed to improve the
performance of the router: faster the routing path searching techniques and layer-2
high speed switching techniques for replacing the layer-3 routing techniques. A
precise and effective control admission control (CAC) and usage parameter control
(UPC) will be also proposed for providing differentiated services. The subproject 3 is
to study and design the multiple field classifier and scheduler for the routers that
provide differentiated services. Many possible configurations will be verified and the
most appropriate one will be proposed and implemented by hardwares and softwares.
NSC89-2213-E009-235 (89R313)

Title:Design and Implementation of Routing Techniques and QoS Traffic Control for
       Broadband Internet
Investigator:Chung-Ju Chang
Sponsor:National Science Council
Keywords:Routing, Multi-Protocol Label Switch (MPLS), DiffServ, Call Admission
             Control (CAC), Usage Parameter Control (UPC)

      With the blooming of Internet application services, the Internet traffic flow
increases dramatically and the traffic flow from inter-network transmission also
increases rapidly. Not only the basic data transmission but also some multimedia
services (such as: voice, real-time video services) are also carried on the Internet.
Henceforth, to provide the QoS-provisioning services, the larger bandwidth capacity,
effective QoS-provisioning service framework, and traffic control mechanisms are the
primary requirements in the design of the Internet.
     For the issue of larger bandwidth capacity, 2 solutions are proposed in this
project: faster routing and virtual routing. For the faster routing, a better routing path
search scheme is proposed to speed up the performance of the router. The virtual
routing technique is to apply the high speed layer-2 switching mechanism on the
router to replace the original layer-3 routing mechanism, such as: IETF’s MPLS
(multi-Protocol Label Switch).
      As to the QoS-provisioning services, the DiffServ (differentiated Service)
service model is considered. An IETF Working Group is dedicated to the research of
DiffServ service model. The goal of DiffServ is not to provide QoS-provisioning
services on the pre-flow basis; instead, on the class basis. The DiffServ defines
several service classes and each class has its own QoS requirements. Therefore, when
the DiffServ services are provided on the Internet, the routers must classify all the
input traffic and assign different scheduling priority level. Also, it is necessary to add
the call control mechanism that has been well-developed in the connection-oriented
Broadband Network Lab has devoted to the research of the traffic control mechanism
on the ATM network for a long time and gained some precious experiences.
     Based on those experiences, an appropriate call admission control (CAC)
scheme is proposed for the traffic control of the DiffServ. Moreover, the proposed
CAC scheme can also applied to other Internet QoS-provisioning services, such as:
IntServ, DiffServ, MPLS and etc. On the receiving of new connection request, the
CAC will check if there is available system resource for the new request. In such way,
network can provide QoS guarantee for both the existing connections and new
connection. Also, we will study the Internet usage parameter control (UPC)
mechanism that is not clearly defined in the IETF specification. By monitoring the
traffic flow, the UPC can make the validation of the connection, such that the CAC
will operate properly.
       Finally, we will apply the fuzzy logic control and neural network mechanism on

the CAC and UPC schemes. By choosing proper fuzzy logic controller and neural
network architecture, we will proposed the corresponding fuzzy and neural network
traffic control rules. To verify the proposed schemes, we will build system simulation
environment and simulate the schemes by software.
NSC89-2213-E009-236 (89R314)

Title:Medium Access Control and Call Admission Control for WCDMA Systems
Investigator:Chung-Ju Chang
Sponsor:National Science Council
Keywords:WCDMA, Medium Access Control, Call Admission Control, DSP, Power
               Control, Code Allocation, Resource Management

      In 3G systems, wideband technology for providing various services and higher
system capacity are most important issues. In order to provide different types of
services in the same infrastructure, the wideband technologies must simultaneously
integrate delay sensitive services, loss sensitive services, and multimedia services, and
guarantee the required QoS for each type of services.
       WCDMA technology is more suitable for smooth traffic applications. However,
it is not suitable for applications with more burstness because all transmitted packets
using the same carrier will be lost due to too large instantaneous multiple access
interference (MAI). This problem is even worse in the uplink. In order to satisfy the
requirements of 3G systems, in this project we will design more flexible and more
efficient uplink MAC control to support various types of services and guarantee their
QoSs. To do this, we are going to input data, output from rake receiver, or in the later
output from multiuser detector, into SIR estimator to determine the users which can
request resource in the next frame.
      Besides, in order to provide variable rate services, multirate transmission
capability is necessary. Many multi-code and single code with variable spreading
factor schemes were proposed. In WCDMA systems, an orthogonal variable spreading
factor code was proposed in ETSI UTRA. In this project, we will design an
appropriate channel code allocation scheme for users with corresponding transmission
rate to enhance the efficiency of the channel code utilization. Furthermore, we will
design a proper power control scheme to achieve good performance for the 3G
WCDMA systems.
     On the other hand, although good MAC control can achieve better system
performace, conditioning on the active users number, MAC control will still be out of
control as too many users in the systems. Therefore, we have to limit the reasonable

number of users in the system based on call admission control. In the project, a
sophisticated call admission control mechanism will be designed so that the required
QoSs of the new call and the existing calls will be satisfied and guaranteed.
NSC89-2219-E009-035 (89R470)

Title:The Script for "The Current Status and Future Trend of Wireless
       Communication and Network Technology"
Investigator:Chung-Ju Chang
Sponsor:National Science Council
Keywords:iWreless Communication, Network Technology,

      With the blooming of communication industry, it is very important and essential
to provide the public with current status and the future trend of network and wireless
technology, and it is also significant to enhance the background knowledge to the
telecommunication industry of the public. Therefore, we are planning to produce an
one-hour program to achieve these goals. In our conception, we will partition the
program into two parts in the first half-hour, we will deliver the evolution and basics
of the network technology; in the remaining half-hour, we will introduce the plentiful
world of the wireless technology.
NSC89-2515-S009-002 (89R554)

Title:Design and Fabrication of Thin-film Microwave Communication Passive
       Devices (1/3)
Investigator:Chi-Yang Chang
Sponsor:National Science Council
Keywords:Microwave Communication, Miniaturized Filter, SAW Filter, Microwave
                Hybrid, Coupler, Diamond Film, High Dielectric Constant Thin Film.

      This program involves three sub-programs. We plane to develop thin film type
microwave communication passive components mainly worked at 900 MHz to 2500
MHz frequency band within three years. The components include miniaturized
thin-film filter, diamond film SAW filter, miniaturized branch-line hybrid, and thin
film coupler with tight coupling coefficient, etc.
      Miniaturized dielectric thin film filters, preliminary diamond film SAW filters
will be developed in the first year. Miniaturized branch-line hybrids, second possible
configuration of thin film filters, and improved diamond film SAW filters will be

developed according to the improvement of processes in the second year. Thin film
couplers with tight coupling coefficient (typically 1 to 3 dB coupling), and further
improved diamond film SAW filters will be developed in the third year. We hope that
the proposed thin film type microwave communication passive components can be
directly used in the current personal communication devices after three years of
     We also plane to develop cost-effective and stable processes for the proposed
microwave communication passive components. A transmission line resonator can be
significantly reduced in size by properly configuring the transmission line structures.
The dielectric thin film is for reducing of the physical length of a transmission line
resonator. Diamond thin film is for increasing of the wave speed of a surface acoustic
wave, and it can increase the operation frequency of a SAW filter. The diamond thin
film and other types of low dielectric diamond like thin films are useful for the thin
film passive components as well for it broadening the design parameters.
NSC89-2213-E009-225 (89R304)

Title:Design and Measurement of Thin-Film Microwave Communication Passive
Investigator:Chi-Yang Chang
Sponsor:National Science Council
Keywords:Mcrowave Communication, Miniaturized Filter, SAW Filter, Branch-Line
               Hybrid, Coupler

     This program is responsible for the design, development, and measurement of
the proposed thin film type microwave communication passive circuit elements.
      We plane to develop miniaturized thin-film filter, diamond film SAW filter,
miniaturized branch-line hybrid, and thin film coupler with tight coupling coefficient,
etc. worked mainly in 900 to 2500 MHz frequency band within 3 years.
      Newly proposed high dielectric thin film miniaturized filters and preliminary
diamond film SAW filters will be design and developed in the first year. Except
continuing the first year’s study, miniaturized branch-line hybrids will also be
developed in the second year. The third year will develop thin film couplers with
tight coupling coefficient (typically 1 to 3 dB coupling) and continuing to improve the
performance of diamond film SAW filters.
     We hope that the methods of design and measurement of thin film type
microwave communication passive components can be developed after three years of

NSC89-2213-E009-226 (89R305)

Title:Design of Wideband Multicode Spread-Signature CDMA Transport
       Architecture for Wireless Error-Resilent Secure Motion Image
Investigator:Po-Rong Chang
Sponsor:National Science Council
Keywords:2D Chaotic Motion Encryption, MPEG-4, Wideband Multicode
              Spread-Signature CDMA System

      Recently, standardization of third-generation mobile communication systems is
rapidly processing in all major regions of the world. These systems, called
International Mobile Telecommunications -2000 (IMT-2000) in the International
Telecommunication Union (ITU) are designed to support a wide variety of
multimedia services with diverse statistical characteristics and quality of service
(QOS) requirements at cell and call levels. To realize IMT-2000 systems, a new
wideband wireless access technology incorporating as many recent technology
developments as possible is necessary. The most promising candidate, wideband
CDMA(WCDMA), is being developed throughout the world. Unfortunately,
wireless network are known for their susceptibility to tampering and eavesdropping.
In a wire network, the fact that a user must be physically connected to the network,
and that information is transmitted within protected physical links(wires), offers some
measure of security. However, in a wireless network, there are no such quarantees as
anyone with a simple radio transmitter can pretend to be a valid network user. In
addition, the transmission medium is the open air which implies that anyone with the
appropriate radio scanner can eavesdrop as well. This project is aimed at developing
a new cryptographic method for encrypting the 4th generation MPEG-4-like motion
images which are the important component of wireless multimedia communications,
and then provides the secure transmission of image information via wireless WCDMA
networks. Meanwhile, since the encryption process is highly nonlinear, both the
multipath fading and multi-user interference on WCDMA radio channels tend to
cause significant transmission error, and the encrypted MPEG-4 motion images are
very vulnerable to these errors. To tackle this difficulty, our new encoding
mechanism has been developed based on a combination of both the MPEG-4 video
object plane (VOP) segmentation and the motion-compensated wavelet
multiresolution wavelet encoding. For the MPEG-4 motion-image encoding, each
picture frame of an input video is segmented into a number of arbitrarily shape image

regions (video object plane) and each of the regions may possibly cover particular
image or video content of interest ,i.e., describing physical objects or content within
scenes. Each VOP is further encoded using the motion-compensated wavelet
multiresolution wavelet encoding. This would lead to one most-important low
resolution prediction error subimage and a number of least-important detail prediction
error subimages. The MPEG-4 video is particularly suitable for the multicode
CDMA system in order to achieve the high data transmission rate via the bandlimted
wireless channel since each VOP generated from the MPEG-4 video may be assigned
to a Walsh-Hadamard(WH) orthogonal codes whereas the MPEG-4 itself is assigned
to a specified PN code. In other woeds, it is creates more than one virtual channel
for transmitting the VOPs of the MPEG-4 video parallely. Furthermore, a two-layer
highest security protection mechanism is applied to the most-important low resolution
prediction error subimage. This two-layer security protetion mechanism includes (1)
2D Chaotic phase scrambler and (2) Chaotic image pixel scanning order encryption.
On the other hand, single layer chaotic image pixel scanning order encryption is
applied to the least-important detail prediction error subimages for providing class
two security protection. To increase the error resillience for those encrypted
prediction error subimages via the wireless CDMA channel, an adaptive channel
optimized scalar quantizer(COSQ) and optimal codeword encoder is developed for
the encrypted subimages according to the value of signal-to-noise/interference(SNIR).
The prediction value of SNIR on the CDMA channel can be obtained by piplined
recurrent neural network with real-time capability. Since the low resolution
prediction error subimage is actually arbitrarily marginal probability distributed, the
subimage should be converted to be a memoryless Gaussian-distributed source signal
by using 2D chaotic phase scrambler. Thus, the SNIR-based adaptive Gaussian
COSQ is applied to the chaotic phase-scrambling subimage to combat the channel
noise and interference . On the other hand, the detail prediction error subimages are
identified as a generalized Gaussian-distributed source signals with a approximately
fixed shape parameter=0.7.          Then, a SNIR-based adaptive generalized
Gaussian(shape parameter=0.7) without phase scrambling is applied to the remaining
detail prediction error subimages. Unfortunately, there is a major drawback in the
SNIR-based adaptive COSQ. It is only suitable for the binary symmetric
channel(BSC).      To overcome this difficulty, this project applies Wornell’s
Spread-signature CDMA-based Rayleigh-to-Gaussian channel transformation
technique to our multicode CDMA system. Thus, from the perspective of the coded
symbol stream, the Rayleigh fading channel looks in effect like an ideal additive
white Gaussian noise channel.
NSC89-2213-E009-159 (89R283)


Title:Investigation of CIC Filter for Decimation and Interpolation for Application in
       Digital Modulation and Demodulation
Investigator:Wen-Thong Chang
Sponsor:National Science Council
Keywords:Cascade-Integrator-Comb Filter, Decimation, Interpolation, Modulation,

       To facilitate the high ratio decimation and interpolation required in the
DSP-based implementation of the digital modulation/demodulation, a two-stage
filtering structure is investigated in this project. The two-stage filter structure consists
of an integrator and an one-tap comb filter. The structure is obtained by decomposing
a multi-tap comb filter. The two-stage structure is called a cascaded-integrator-comb
filter (CIC). The advantage of the decomposition is the reductions of the amount of
multiplication and the dynamic range of the register required to store the data. Since
the integrator and the one-tap comb filter are all with unity gain, the multiplication
can be replaced with addition.
      The ratio of the decimation and interpolation is determined by the ratio between
the carrier frequency and the base band signal bandwidth. In the digital
implementation of modulation, a necessary task is the match of the sampling rates
between the carrier and the base band signal in order to perform the multiplication
required for modulation. Thus a high ratio interpolation is to be done. For
demodulation, successive decimation is performed to implement the bandpass
sampling to extract the base band signal from the carrier.
      The CIC is essentially a low pass filter used to eliminate the aliasing and
imaging errors due to the decimation and the interpolation. Many CIC filter sections
are usually cascaded together to fulfill the required pass band gain and the stop band
attenuation and are usually called as high order CIC filters. The structure of the high
order CIC is also very simple. All the integrators can be grouped together as well as
all the one-tap comb filter. There are three parameters to be determined in a CIC filter.
They are the differential delay of the one-tap comb (This corresponds to the original
tap length of the M-tap comb filter) , the order of the filter and the
decimation/interpolation ratio. The relationship among these three parameters will be
jointly studied and their effects on the digital implementation of the
modulation/demodulation will be investigated.
NSC89-2213-E009-189 (89R294)

Title:Quantization Effect Reduction with Adaptive Wiener Filter in Low Bit Rate
Investigator:Wen-Thong Chang
Sponsor:National Science Council
Keywords:Video Compression, Quantization Effect, Blocky Effect, Adaptive Wiener
             Filter, Local Statistics.

      In low bit rate video transmission environment, many methods have been
proposed to control the output bit rate by adjusting the quantization stepsizes. These
methods are mainly based on the statistics of the transform coefficients or the
distribution of the block pixels. In more complex algorithms, quantized coefficients
from several blocks are taken into consideration simultaneously. All of the above
methods eventually result in the adjustment of the quantization step sizes to control
the quantized coefficient level. After these coefficients are inverse transformed,
blocky effects, discontinuity in block boundaries, or noise inside the blocks often
appear in the reconstructed signal.
     In this proposal, we consider a general post-processing procedure to solve this
problem. In a DCT-based image coder, the quantization stage inevitably introduces
noise to the transform coefficients and hence causes signal degradation after
reconstruction. The quantization effects in the DCT-based image coder may appear as
blocky effects , ringing effects, or other artifacts, etc. In order to reduce such effects,
many spatial domain and frequency domain post-processing approaches have been
proposed. Due to the computational complexity and iterative characteristics, Several
methods need much computational cost and may not be applicable for real-time
processing. We propose an approach based on the use of adaptive Wiener filtering on
the transform coefficients to reduce the quantization error and hence the quantization
effects. Although this approach operates on the transform domain, it is feasible for
real-time applications due to the use of local transform coefficients. Simulation results
show that this method not only reduces the quantization effects, but also increase the
reconstruction PSNR from 0.1 to 1.3 dB.
NSC89-2213-E009-232 (89R310)

Title:Long-Distance Voiced Website for Hearing-Impaired Users
Investigator:Wen-Whei Chang
Sponsor:National Science Council
Keywords:Sensorineural Hearing Loss, Hearing Aids, Sinusoidal Model, Computer

               Telephony Integration

      Approximately 7.5 percent of the population has some degree of hearing loss,
but only 15 percent of them are actually using hearing aids. Of the two major types of
hearing loss, conductive and sensorineural, only the former can be adequately
compensated using normally available hearing aids. This motivates our research into
trying to devise an aid that properly deals with the restricted bandwidth and reduced
dynamic range present in sensorineural hearing impaired. The strategy applied here is
based on a sinusoidal model, in which speech waveforms are represented as the sum
of sinusoids with specific frequencies, amplitudes, and phases. The motivation for this
representation is two-fold. First, frequency-scale modification can easily be applied
on sinusoidal parameters to place speech in a desired bandwidth as an aid to the
hearing impaired. Secondly, compression of sinusoidal amplitudes can serve to
compensate for recruitment of loudness incurred by the reduced dynamic range of
hearing. For practical use, we will implement the speech enhancement for hearing
aids through the computer telephony integration so that the hearing impaired can use
telephones to access all kinds of voiced services within the internet. This aid will be
evaluated by hearing-impaired listeners. The test will consist of four different tasks:
detection, discrimination, recognition, and comprehension, representing four levels of
the listening process. The results of this evaluation will shed light on the various kinds
of benefit our future users with different degree of losses can derive from this aid.
NSC89-2614-E009-003 (89R270)

Title:Chinese Dialect Identification Based on Acoustic-Phonotactic and Prosodic
Investigator:Wen-Whei Chang
Sponsor:National Science Council
Keywords:Dialect Identification, Hidden Markov Models, Speaker Adaptation,
               Discriminative Training

     Automatic language identification is one of the active studies within an overall
target towards multilingual information access. The interconnections between Chinese
dialects are in fact as complicated as those among the family of Romance languages.
This motivated our research into trying to devise a system for identifying three major
dialects spoken in Taiwan, Mandarin, Holo, and Hakka. The first step toward
realization is to use a 3-state Markov chain to model the sequential statistics of the
acoustic segments such as the syllable onset, nucleus, and ending. This model is said

to have a composite structure; it is a large Markov chain in which each state is built
from a bank of elementary left-to-right hidden Markov models (HMMs). Each
elementary HMM corresponds to one of the five broad phonetic classes (BPCs)
including the stop, fricative, affricate, nasal, and vowel or diphthong. In order for
prosodic features to be integrated with the composite model, we also propose to use
tonal BPCs rather than BPCs as the basis for dialect identification. Since pitch is
defined for voiced speech, we need only superimpose distinct tonal notations onto the
BPC equivalents of the syllable nuclei in order to obtain a set of BPC variants that
share the same BPCs but have distinct lexical tones. The dialect identification system
will operate in two phases: training and testing. In the training phase, a composite
model is trained for every dialect to be recognized. During recognition, a test
utterance is classified by extracting mel-cepstral and pitch features and by calculating
the likelihood that these features were produced in each of the dialects. It is believed
that the most likely dialect is optimal with respect to some combination of both
acoustic and prosodic information. Finally, we will apply discriminative training
algorithms for speaker adaptation so that the system can properly deal with new
talkers whose utterances are not used during the training phase.
NSC89-2213-E009-188 (89R293)

Title:Resolving and Modeling the Self-Similar Nature of the Packet Traffic on the
       Real Networks
Investigator:Po-Ning Chen
Sponsor:National Science Council
Keywords:Network Traffic Model, Self-Similarity, Long-Range Dependence

      Recent empirical studies of network traffic measurements for packet-train
sources have convincingly demonstrated that the actual network traffic is self-similar
in nature. In other words, they are bursty over an extremely wide range of time scales.
This is sharply contrary to the traditional traffic models, which only possess
short-range dependency. As a consequence, those analysis and simulations based on
the traditional traffic models, such as Poisson and Bursty Poisson, become doubtful in
their physical implications to true networks. Two questions are therefore raised due to
this observation: (1) what the interpretation is for the observed self-similar nature of
the measured traffic of today's networks; (2) how self-similarity impacts on the future
network protocol design, as well as router/switch architecture. In this project, we seek
to answer the first question by means of providing a protocol/structure-based
statistical model for individual source. In our view, the self-similar nature of today's

networks is somewhat due to the packet network structure and protocols employed
nowadays; and hence, such a statistical model seems plausible, if our interpretation
matches the actual phenomena. We will substantiate our model by comparing a
time-scale superposition of many such statistically modeled sources with the
measured data from others. Parameter fitting for networks with varied degree of
long-range dependence will be examined next. To resolve the second question, we
will try to collect some basic rules in future protocol designs, which result in a better
network performance. Considerations of backward compatibility, complexity and
feasibility in implementation will be included.
NSC89-2213-E009-237 (89R315)

Title Integrated Circuits Design and Implementation of Medium Access Control and
       Baseband Tranceiver for High-Speed Wireless Networks
Investigator:Po-Ning Chen
Sponsor:National Science Council
Keywords:High-Speed Wireless Data Networks, CSMA/CD, OFDM, CCK, FHSS,

       The recent growth of wireless communication market and the wide-spread of
portable 3C applications make high-speed wireless transmission a promising future
technology. The current transmission techniques over wireless LAN are primarily
based on Direct Sequence Spread Spectrum (DSSS) and Frequency Hopping Spread
Spectrum (FHSS); and the mediums are mostly chosen the ISM band. In this project,
we will first aim at the performance analysis for the ISM band modulation
techniques transmitted over channels with diverse characteristics. The channels
considered will include AWGN channel and Rayleigh Multipath Fading channel. The
ISM band modulation techniques will cover the FHSS specified in Bluetooth system,
and CCK and OFDM employed in WLAN. We will then investigate the performance
degradation due to quantization, as well as synchronization. On the other hand, from
the system development that we have experienced in the past, the design of the
baseband tranceiver shall be tightly related to other system modules, such as RF/IF
and MAC. Therefore, we will perform the system-level integration of MAC and
Baseband transceiver in this sub-project. As learned from the MAC design of the
existing IEEE 802.11 products, a CPU-based approach is widely-adopted, which
implements the CSMA/CA module in terms of the embedded coding. Two known

examples are the 80188-based AMD79C930 and the DSP-based HFA3841. Although
a firmware-based implementation is easy in its maintenance and more flexible to
standard revision, it unavoidably limits its transmission speed by the CPU
performance, and hence, render a dilemma between speed (CPU performance) and
cost, especially when high-speed becomes a demand. Most importantly, a
firmware-based MAC design will harden the joint simulations, as SOC is the target,
together with RF/IF and Baseband modules. We therefore propose an alternative
cost-effect partition between the specifically designed components for MAC and its
associated software modules executed at host (in a form of a driver). Our specifically
designed MAC modules will include the CSMA/CA Unit, Control Frame Handling
Unit, High-Speed CRC Unit, and Double-Buffering Memory Unit. The targeted
transmission speed is 100Mbps. In the end of the sub-project, we will integrate the
MAC and Baseband in one chip; and, if possible, we will further incorporate RF/IF
module designed by other sub-projects.
NSC89-2218-E009-052 (89R421)

Title:Development of Advanced Mandarin Speech Recognition Systems (1/3)
Investigator:Sin-Horng Chen
Sponsor:National Science Council
Keywords:Mandarin Speech Recognition, HMM, Robust Training, Prosodic
               Modeling, Pre-rocessing, RNN

     This is a three-year project. The goal is to develop advanced Mandarin speech
recognition systems using both the MAT-2400 telephone-speech database and a
300-speaker microphone-speech database. The study will base on a newly developed
right-context dependent initial/final HMM recognizer and emphasize on the following
four issues : robust training, prosodic modeling, pre-processing, and system
integration. Firstly, a new robust training method will be developed by simultaneously
considering the compensations/eliminations of channel bias, speaker cluster bias,
noise interference, and the energy contour effect. It aims at obtaining a set of compact
HMM models. Secondly, a new prosodic modeling method using initial/final
durations, syllable pitch contour, syllable energy level, and inter-syllable pause
duration will be discussed in detail. Thirdly, we will develop an advanced RNN-based
pre-processing to extract sophisticated acoustic cues for helping the following
acoustic decoding. Lastly, advanced speech recognition systems for both
telephone-speech and microphone-speech will be developed via integrating all above

NSC89-2213-E009-187 (89R292)

Title:Designs of Near-Millimeter-Wave Array Antenna and Retrodirective Array
       Reflector for Vehicle Collision Avoidance System (2/2)
Investigator:Shyh-Jong Chung
Sponsor:National Science Council
Keywords:Vehicle Collision Avoidance System, 24 GHz Microstrip Antenna Array,
               24GHz Amplifier, 24 GHz Active Retrodirective Array Reflector

      A 24 GHz microstrip antenna array and a 24 GHz active retrodirective array
reflector will be designed and demonstrated in three years. The antenna array is to be
used as the front antenna of a vehicle collision avoidance sensor, and the reflector is
proposed to be equipped on the vehicle body to enhance the radar echo over a wide
range of wave incidence angles. The microstrip antennas, which possess a planar
configuration, are adopted as the unit elements for both the designs. Also, the
microstrip lines are utilized for the feedings and connections of the antennas. In the
first year, a single 24 GHz microstrip antenna and various microstrip-line
discontinuities for the feeding network of the antenna array have been designed. In the
second year, an 88 microstrip antenna array and a 24 GHz small-signal amplifier
will be developed. The array is expected to possess a main-lobe beamwidth of about
10 to 15, a gain of 20 dBi, and side-lobe levels lower than -25 dB. The small-signal
gain of the amplifier is about 9 dB. In the last year, an active retrodirective array
reflector is to be completed. The reflector is composed of several antenna pairs, each
including two microstrip antennas and two small-signal amplifiers, connected by
microstrip lines. By suitably arranging the positions of the antennas and the lengths of
the connecting lines, the antennas in the array reflector can coherently reradiate the
received illuminating wave toward the wave-incidence direction, over a large range of
about 100 of the incidence directions.
NSC89-2213-E009-242 (89R340)

Title:Study on Acoustic Echo Cancellation Based on Maximum Length Correlation
Investigator:Shih-Fu Hsieh
Sponsor:National Science Council
Keywords:Adaptive Filter, Acoustic Echo Cancellation, Double Talk Detection

      An adaptive acoustic echo canceller (AEC) is typically implemented using a
finite impulse response (FIR) to account for any changes in room impulse response
(RIR). There are three major concerns in AEC: The first one is that the impulse
response of the RIR is generally very long, sometimes several thousands for the
teleconferencing system. This renders a complex implementation. The second one is
that RIR is very sensitive to object movements in the room enclosure so that the
adaptive algorithm must be fast in order to prevent the adaptive filter from being
misadjusted. The last one is to provide echo cancellation during double-talk.
Double-talk occurs when a near-end speech source other than the loudspeaker is
active within the room. In this case, the adaptive algorithm attempts to cancel both the
echo and the near-end speech. This invariably leads to divergence of the adaptive
filter from the optimum solution.
       Current techniques attempt to effectively turn off the adaptation during
double-talk. A critical question is that merely measuring the prediction error will not
discriminate between double talk and echo path changes.
      Another technique that robust to estimate RIR is the maximal length correlation
(MLC) algorithm . This is done by adding a low level, wide-band maximal length
sequence (MLS) to the far-end audio before it is sent into the room. However, the
MLC algorithm are interferences due to the near-end and far-end signal.
       In this project, we propose an iterative MLC (IMLC) algorithm without the need
for double-talk detection. We try to cancel the far-end speech first before the RIR is
NSC89-2213-E009-191 (89R296)

Title:Design, Simulation, and Verification of a Hybrid Image Rejection Receiver
Investigator:Chia-Chi Huang
Sponsor:National Science Council
Keywords:Image Rejection Receiver, Low IF, Digital Image Rejection

     An image rejection receiver plays a very important role in a low intermediate
frequency (IF) receiver architecture. In this project, we will develop a hybrid image
rejection receiver architecture, which is suitable for VLSI implementation. This
hybrid image rejection receiver architecture consists of two parts. One part is a RF
front-end architecture and the other part is a baseband digital image rejection receiver
architecture. We plan to verify and analyze the performance of our designed receiver
architecture based on both computer simulation and experimental measurement. At
the end of this project, we will provide both the algorithms and the block diagrams of

this hybrid image rejection receiver architecture for future IC design of a low-IF radio
NSC89-2213-E009-182 (89R287)

Title:Research of X/Ku band High-Efficiency Wideband Active-Integrated Antenna
       Arrays (1/2)
Investigator:Christina Jou
Sponsor:National Science Council
Keywords:Quasi-Optical, Asymmetric, Dual-Beam Circular-Polarized, Power
               Combining, Leaky-Wave Antenna, Coplanar Waveguide, Back Lobe

      Current research interest in millimeter wave system is motivated by the
advantage such as smaller antennas, wider bandwidths, and better resolution for
imaging and radar system. However, this natural progress from the microwave to
millimeter wave band has been hindered in many cases by the lake compact, reliable,
high power solid-state source at these wavelengths. The advances in device
technology make wafer-scale integrated power combining sources very attractive. In
addition, free-space power combining is more efficient at high frequencies than power
combining in guided wave structure, and transmitting and receiving system based on
monolithic implementation have the potential to be smaller, lighter.
      Quasi-optical techniques spurred the development of active integrated power
combiner in 1980's 1990's. As the operating frequency increases, the available power
from solid state device decrease. Therefore, power combining of solid-state devices
using quasi-optical techniques in the millimeter wave region became an important
issue. There are two different approaches for quasi-optical power combining. One is
the active antenna approach and the other is the grid approach.
      The advantage of active antenna is: (l) increasing the effective length of short
antenna (2) increasing the band-width (3) decreasing the mutual coupling between
array elements (4) improving the noise factor. From the forgoing advantages, we can
see active devices and passive patch antenna fabricated on the same substrate can
improve antenna performance. The technology of monolithic microwave active
integrated circuit is mature, and makes the approach promising. Because the patch
antenna has the drawbacks of the narrow bandwidth and the smaller scanning angle,
we decide to use another types of antennas, including the traveling antenna and the
leaky-wave antenna, to improve the above drawbacks of patch antenna. In our
1998-1999 NSC project, we have some success studies of active leaky-wave antenna

arrays.They have the advantages of power combining, the large scanning angle, high
efficiency, and noise bandwidth. The results have been published in the international
journals, such as IEEE,IEE ,IEICE ,and other famous journals.
     The results are described in the following paragraphs:
   (1)Coplanar Waveguide to Coplanar Strips Fed Active Leaky-Wave Antenna[1]
   (2)An Asymmetric Feeding Active Frequency-Scanning Leaky-Wave
   (3)Asymmetric Feeding Active Frequency-Scanning Leaky-Wave Antenna
   (4)A New Two-Beam Scanning Active Leaky-Wave Antenna[8-12]
   (5)Active Feedback Microstrip Leaky-Wave Antenna Synthesizer Design With
      Back Lobe Radiation[13]
   (6)Active Aperture-Coupled Leaky-Wave Antenna[14-15]
In the next two year NSC project, we hope and try to enhance the wideband
characteristics, scanning angles, and circular polarized property arrays. In the first
(1) Two-dimensional beam-scanning active leaky-wave antenna:
    We want to demonstrate an asymmetric microstrip line fed linear active phased
antenna array with two-dimensional beam-scanning capability without using phase
shifters. The coupling transmission line enhances the mutual synchronization of the
active antenna array. The controlling of the free-running frequency is achieved by
tunning the varactor’s DC bias, which also gives an additional advantage of fine
tunning the scanning angle of the array.
(2) Dual-beam asymmetric-scanning active leaky-wave antenna:
   In the past project from 1998-1999,m we have fabricated a two-terminal feeding
leaky-wave antenna integrated with a Voltage-Controlled Oscillator (VCO). The
proposed antenna circuit can create a dual-beam symmetric-scanning radiation pattern.
When the operating frequency changes, the main beam can scan symmetrically. In the
new project, we wish to demonstrate a novel active leaky-wave whose beams can scan
asymmetrically. Active and passive mixers, a power divider, VCO, local oscillator
(LO), and a two-terminal feeding leaky-wave antenna are integrated on a substrate.
With the frequency variation of LO, only one beam can be changed into another angle
and the other beam is still fixed.
    In the second year, we hope to achieve these goals: power combining efficiency,
polarization, and different switchable beam forming (sum and difference and normal
(1) Beam-forming switchable active antenna array:

      In 1996, Professor Tzuang, published the paper ”Short antenna of sum and
difference patterns”, in IEE Electronics Letters, to perform the sum and difference
radiation pattern. But it must use two different structures to switch the main-beam
direction. In our project, we will try to use a two-terminal feeding leaky-wave
antenna to create a beaming-forming switchable radiation pattern electronically, not
mechanically. Applying the phase difference of two inputs, the sum, difference, and
normal pattern can be created. After the first-step calculation and simulation, a
theoretical result can be obtained (See Fig 4.)
(2) Circular-polarized active leaky-wave antenna:
      In the available papers, the circular-polarized research has not been studied yet.
In this project, in order to find out the quasi-optical power-combining efficiency and
circular-polarized phenomenon, we want to use the simple structure leaky-wave
antenna to demonstrate a circular-polarized active antenna array. This topology can
be used in the applications of the satellite communication, the LMDS system, and
military communication system.
      The project leader, Professor Jou, has worked in the MIC and MMIC research
for ten years and many solid-state source circuits were developed. The post doctor, Dr
Chien-Jen Wang, is a skillful engineer in the topic of active antenna arrays and
power-combining networks. It is believed that the active antenna has the advantages
of broad bandwidth, high efficiency and large scanning angle will be demonstrated
successfully in this project.
NSC89-2213-E009-210 (89R301)

Title:Development of X/Ku Band 4 Bits MEMS Phase Shifter (1/2)
Investigator:Christina Jou
Sponsor:National Science Council
Keywords:MEMS, Phase Shifter, Microswitch

      This project is three years of second year. In the past year, we had a lot of
research results, it will to recount in table C018.This project is to develop new
prototype Ku-band 4 bits phase shifter with low insertion loss, using the
microelectromechanical systems (MEMS) technology which have promoted exciting
advancements in the field of microwave switching recently.
      Phase shifter can be used to control the phase of RF signal through switching of
different transmission paths. RF MEMS switches will be used to replace traditional
p-i-n diode or FET devices in phase shifter circuit in this study. It can be actuated to
ON or OFF state through electrostatic force, and the RF signal path will be connected

or disconnected. Figure 1 shows the basic operation principles of RF MEMS switch.

       RF Input    Top Membrane            RF Output

                                  Si Substrate

        Post                  Control Pad

                  (a) OFF State

      RF Input                            RF Output
                   Top Membrane

                            Control Pad

                  (b) ON

                   Figure 1. Basic operation principle of RF MEMS switch

  As a example of phase shifter, figure 2 is a 3 bits circuit, S1 to S6 represent the RF
MEMS switches.

           IN                                                               OUT
                                  S1                   S3        S5


                  Figure 2.Schematic of 3 bits phase shifter

      RF signal enter the circuit from input port, when S1 , S3, S5 are actuated to ON
state, and S2,S4,S6 are actuated to OFF state, the RF signal will have the minimum
phase relative to other definitions of the 3 bits switches. Totally, there are 8 different
phase states which the input RF signal can be tuned.
      In this project to develop MEMS switch measurement system in microwave is
our emphasis. We will to develop two prototypes MEMS switch, and will meet
difficulties to measure frequency responses in microstrip line MEMS switch, because
on wafer probe measurement is CPW structure. When we will to measure microstrip
line MEMS switch through CPW toμstrip adapter, but CPW toμstrip adapter

substrate is too expensive(about NT eighty ten thousand).So we will to develop this
high frequency adapter by ourselves, and accede to TRL measurement
technology to make measurement is exact.
     This study will cooperate with Professor N. C. Luhmann from University of
California, Davis in USA. He is the chairman of electrical engineering department at
UCDavis. He was also the advisor of this author. Professor Luhmann has been
working in RF area over 20 years, the researches in his laboratory cover almost all the
important issues about traditional RF switches and its system applications, such as
switch delay line, quasi-optical phase shifter, phase antenna arrays, their efforts has
acquired much respect from the world. We will work together to develop various high
performance RF MEMS phase shifters. The process of the device will cooperate with
Professor and chairman Star Ruey-Shing Huang from NTHU.
      The MEMS phase shifter will be fabricated at the facilities of Semiconductor
Research Center and National Nano-Device Laboratory in NCTU. The process of RF
MEMS technologies will to be manufacture by our research group.
     This project will to connect investigated result of past year and to develop other
types MEMS Switch (Ex CPW MEMS Switch , show as figure 3)at next two year . In
the past year, we had concentrated on design, layout, simulate, and we have already
finished microstrip line MEMS switch. In the next year, we will concentrate on
building the test facilities for the measurement of DC and frequency domain
characteristics; base on the results from the measurement and simulation, we will
design and simulate the MEMS phase shifter. In the second year, we will concentrate
on characterization of MEMS phase shifter, and performance evaluation in the
integrated front-end subsystem with antenna and LNA devices.



                         Figure 3. CPW MEMS               Switch       (a)Side view
                                                                        (b)Top view

NSC89-2218-E009-047 (89R418)

Title:Dual Network Twin Node Architecture for WDM Networks
Investigator:Ming-Seng Kao
Sponsor:National Science Council
Keywords:WDM Network, Network Architecture

      The continuing improvement on wavelength division multiplexing, tunable
optical transceivers and optical devices accelerates the spreading of optical fiber
infrastructure. An immediate question arises out of interconnecting many existing but
variant networks:how to link them by a simple and effective strategy? The question
may occur when connecting local or metropolitan area networks as well as when
building international links between countries.
      We had proposed a novel architecture for effective connection of existing
networks. Let Na , Nb , Nc , Nd and Ne be five local networks to be connected
into a large network. For reliability consideration, we choose two nodes from every
local network and name each pair of nodes as twin nodes, e.g. ( A, A') , ( B, B') ,
(C, C') , ( D, D') , and ( E , E ') are the selected twin nodes out of each local network.
Instead of constructing a single network, we cluster nodes A , B , C , D and E
into a network ( Net-I ) and construct another network consisting of nodes A' , B' ,
 C' , D' and E' ( Net-II ). Net-I and Net-II are independent networks, in fact, they
are always designed based on their respective considerations according to those nodes
for which they serve. The communications between Net-I and Net-II are
accomplished by the existing links between twin nodes. Net-I and Net-II together
with those existing links between twin nodes form the resulting network
interconnecting these five local networks. We named this special approach as the

Dual Networks-Twin Nodes (DNTN) architecture due to the existence of two subnets
and the many twin nodes.
       From our previous results, the DNTN architecture indeed brings several
advantages compared with conventional WDM networks in terms of reliability and
efficiency. In this project we will proceed to studying many interesting topics
accompanied with the DNTN architecture, which include :
           Possible combinations of the dual networks
           Performance analysis under practical constraints
           Wavelength routing algorithm
          Optimum wavelength assignment
          Other related issues
     The results of this project will further enhance the understanding of the DNTN
architecture, being critical for the applications of DNTN architecture in the future
high-capacity WDM networks.
NSC89-2213-E009-181 (89R286)

Title:2.4GHz CMOS RF Front-end Circuits Design(III)Quadrature Phase VCO
Investigator:Yao-Huang Kao
Sponsor:National Science Council

     CMOS technology applied to rf integrated circuits has received much attention
for the inherent properties of low power consumption and high integration. In this
third-year project the voltage-controlled oscillators (VCO) from cross-coupled
negative resistance configuration are developed. The tuning element is from CMOS
capacitor. To achieve small size, high integration, and low phase noise, the on-chip
inductor is especially investigated. Due to the nature of lossy substrate of Si in
standard CMOS process. The quality factor is poor. According to our previous study
the loaded Q is only about 3~4 at 2.4GHz, which is fatal to the required high quality
narrow band rf circuit. In this project a new technique by using micro-machining
method is applied to remove the lossy substrate beneath the spiral inductor such that
the quality factor is improved, and so does the self resonant frequency. The CMOS
process is from CIC and the micromachining is from Prof. C. M. Chang in Dep. of
Electronic engineering, NCTU.
     With the well-behavior inductor, the vco is then fabricated. Our goal is at 2.4GHz
with phase noise around 100dBc/Hz offset 100kHz, which is applicable to ISM band
and other modern wireless communication systems, such as home-rf and blue-tooth.
Besides that the quadratic phase shifter is built for the purpose of low IF receiving

without using image rejection filter. Base on the previous experience in LNA and
mixer, we are convinced to have the capability to achieve the goal and the integration
of the front end at 2.4GHz.
NSC89-2213-E009-245 (89R318)

Title:Linearization of High Power Amplifier For Mobile Communication
Principal Investigator:Yao-Huang Kao
Sponsor:Remotek Corp.
Keywords:Cross Modulation, Intermodulation, AM to PM Effect:

       This project research the RF power amplifier linearity. The RF power amplifier
is usually the most important and expensive component in a wireless communication
system. We study and implement the predistortion and feedforward linearizing
techniques for the RF PA to minimize the IMD under -30 dbc.This will decrease the
AM-AM and AM-PM effects in the output signals of the wireless communication
system.So can acheive the two tone test specification of modern wireless
communication system.
C89240( 89.07.01-90.06.30 )

Title:A Wideband Design of Parallel Coupled Microstrip Bandpass Filter
Investigator: Jen-Tsai Kuo
Sponsor:National Science Council
Keywords : Microwave Integrated Circuit (MIC), Microstrip Line, Wideband
               Microwave Filter

       We propose a design of a wideband filter based on the parallel coupled
microstrip filter structure. The philosophy of the design to achieve wide bandwidth is
to increase the coupling coefficient of the first and the last sections. The method of
implementation is to replace the coupled sections of both end stages with symmetric
three-line structures. Once the electromagnetic coupling is increased, the bandwidth
of the bandpass filter can be increased. In other words, the design releases the gap size
of the end sections. It can make the filter design work more flexible and raise the
NSC89-2213-E009-193 (89R297)

Title:Space-Time Processor for Multi-Carrier/OFDM CDMA Systems
Investigator:Ta-Sung Lee
Sponsor:National Science Council
Keywords:CDMA, Multi-Carrier Transmission, OFDM, Smart Antenna

     There have been a great deal of advances in wireless communication systems
development in recent years. In order to offer a higher data rate and larger system
capacity, the wideband CDMA (W-CDMA) technology has been proposed as a
standard for the third generation mobile communication systems. Unfortunately, due
to multipath fading effects, the W-CDMA system cannot offer a very high data rate
over a single carrier, and a RAKE receiver is typically needed to exploit multipath
diversity. The multi-carrier (MC), or OFDM transmission, can be employed to
alleviate the frequency-selective fading problem. This leads to the combination of
MC or OFDM with CDMA as a solution to next generation high speed wireless
       On the other hand, space-time processing is currently receiving much attention
in the area of wireless communications. The major merit of space-time processing
lies in the enhancement of effective signal-to-interference ratio via spatial division
multiple access, thereby achieving an improved communication quality and capacity.
In this project, various types of MC-CDMA systems will be investigated and
compared with the DS-CDMA system. An antenna array will also be incorporated to
offer the advantages of space-time or smart antenna processing. Finally, we will try
to realize the feasible algorithms using DSP firmware, in an attempt to move towards
the software radio realm.
NSC89-2213-E009-160 (89R284)

Title:Algorithm and Software Radio Realization of Space-Time RAKE Receiver
Investigator:Ta-Sung Lee
Sponsor:National Science Council
Keywords:Wireless Communications, CDMA, Smart Antenna, RAKE Receiver

     The smart antenna represents one of the emerging and key technologies in the
third-generation mobile communication systems (IMT-2000). The major function of
a smart antenna lies in the enhancement of effective signal-to-interference ratio via
spatial division multiple access, thereby achieving an improved communication
quality and capacity. Due to the rapid advancement of DSP chips, realization of
smart antenna technology via software radio is becoming feasible. Along these lines,

the goal of this proposal will be on the study of the algorithms and software radio
realization of smart antennas for wideband CDMA (W-CDMA) systems. Specifically,
the approach of so-called space-time 2-D RAKE receiver will be taken to integrate the
degrees of freedom offered by the smart antenna (spatial) and conventional RAKE
receiver (temporal) to develop a space-time joint processor. Various types of 2-D
algorithms have been investigated, and the feasible ones will be realized using DSP
firmware techniques.
     The DSP work in this proposal will be mainly based on our previous two years’
research results on W-CDMA smart antenna, 2-D RAKE receiver and interference
cancellation. Two sets of DSP modules (TMS320C6X) will be employed to simulate
the software radio structure. In particular, one of the module will be used as a signal
generator to produce simulated W-CDMA signals representing what are observed at
the base station antenna array. The other module will be used as a 2-D RAKE
receiver to perform signal reception and demodulation. The operation of the entire
DSP system can be controlled by a host PC or workstation, with the algorithms first
simulated and evaluated then down loaded to the board. This should enable the
seamless connection from simulation, debugging to final realization. Finally,
integrated experimental trial will be performed with other co-proposals to ascertain
the efficacy of the 2-D RAKE receiver.
NSC89-2219-E009-023 (89R509)

Title:QoS Control for Broadband Internet
Investigator:Tsern-Huei Lee
Sponsor:National Science Council
Keywords:Broadband Internet, Quality of Service, QoS Routing, Gigabit Router,
               Signaling, Policing, Scheduling, Network Planning, Management

      Continuous rapid growth of the Internet in recent years makes it the most
probable future integrated services network. However, current Internet architecture is
inadequate in providing real-time applications. It cannot guarantee delay bound
requirements of real-time applications. Moreover, non-real-time applications may be
terminated if real-time traffic causes congestion. Internet 2 is thus proposed to meet
future needs.
      Broadband Internet will undoubtedly face applications with wide quality of
service (QoS) requirements. In face of increasing traffic, to guarantee quality of
service (QoS) efficiently, it needs:
   (1) High-capacity (Gigabit) routers: to increase network capacity.

    (2) QoS routing: to improve network utilization.
    (3) Signaling: to characterize source traffic and reserve resources.
    (4) Admission control: to limit access of network resources.
    (5) Policing: to provide reliable QoS guarantee.
    (6) Scheduling: to provide different QoS.
    (7) Network Planning and Management: to monitor network performance and
       plan network evolution.
   In this project, we shall investigate all the above issues in six sub-projects. Our
main purpose is to develop techniques which can be used in Broadband Internet.
Results obtained in this project will be patented and/or published in international
conferences and journals.
NSC89-2219-E009-027 (89R463)

Title:Gigabit Router and Traffic Policing for Broadband Internet
Investigator:Tsern-Huei Lee
Sponsor:National Science Council
Keywords:IP Router, Crossbar Switch, Hardware Routing, Traffic Policing

      High-capacity router is an essential component for future broadband Internet to
handle rapidly growing traffic. The bottleneck for conventional routers is software
processing of routing information. Hence, to increase capacity, routing information
must be processed with hardware.
      In this sub-project, we shall investigate two key technologies of high-capacity
routers: switch architecture and longest prefix matching with hardware (or hardware
routing). Conventional bus-based architecture is not scalable because of clock rate
and pin count limitations. A capacity more than 12 Gbps is very difficult to be
realized with current technology.
      Space-division such as the crossbar architecture is thus necessary to build a real
large-capacity router. However, crossbar switches suffer from head-of-line (HOL)
blocking which limits the maximum throughput to about 50%. Longest prefix
matching is another bottleneck in IP routing. Unlike exact matching, hashing cannot
be easily utilized in IP routing. Putting routing table in silicon efficiently seems to
be the most widely acceptable solution in current stage.
      In this sub-project, we shall develop queue management and fast scheduling
algorithms to (partially) remove HOL blocking. In addition, we shall design
efficient hardware routing schemes to speed up routing decision. Traffic policing for
frame-based network and traffic classification are also investigated here. In traffic

policing, we shall focus on the leaky bucket algorithm. Our purpose is to design
efficient realization algorithm and implement it with FPGA technology. For traffic
classification, we shall design classifiers based on destination IP address, source IP
address, destination port number, source port number, and protocol ID。
      Results obtained in this sub-project will be patented and/or published in
internationally recognized conferences and journals. To enhance the hands-on
experience, we shall also develop chips and possibly prototype a simple switch router
for the rest of two years.
NSC89-2219-E009-028 (89R464)

Title:A TCP-Friendly ABR Flow Control
Investigator:Wei-Kuo Liao
Sponsor:National Science Council
Keywords:TCP-Friendly Flow Control, ATM ABR Flow Control, Maxmin Fairness,
               Gaussian Process

      A TCP-like source will decrease the window size substantially in response to the
detection of TCP segment losses in order to reduce the loss ratio. It has been reported
that this substantial decrement results in the network bandwidth underutilization. In
addition, it also leaves an advantage to the TCP-unfriendly sources, e.g., the one using
UDP, for obtaining more bandwidth. The network bandwidth underutilization is then
regarded as a unfairness problem by the network flow control society since at this
time in the Internet, only the flow using TCP flow control can react to the advises by
the network elements and thus the TCP flow is identified to be a responsive one. In
this project, we study how to protect the TCP-like source by the ATM ABR flow
control. The ATM ABR flow control is capable of doing the per-flow flow control and
thus can be recognized as the most powerful flow control scheme nowadays. In
addition, it is very easy for ABR flow control to provide fast maxmin fairness
convergence, which is very hard or impossible for other flow control techniques, such
as active queuing method. Nevertheless, up to date, there is no ATM ABR flow
control efficiently protecting the TCP-like source. Therefore, we set the main
objective of this project as to develop an ATM ABR flow control to maintain the
bandwidth utilization of TCP over ABR transmission, and in the meantime, the
maxmin fairness convergence is still approved.
NSC89-2218-E009-108 (89R542)

Title:Millimeter Wave Applications of Leaky-Wave Antenna
Investigator:Yu-De Lin
Sponsor:National Science Council
Keywords:Leaky-Wave Antenna, Millimeter-Wave Antenna, Planar Retrodirective
               Array, Slotline Antennas

      The goal of this project is to explore two possible millimeter-wave applications
of the leaky-wave antennas. One of them is the leaky-wave retrodirectivearray. The
other is the slotline leaky-wave antenna array. Using planar leaky-wave antennas as
the basic building blocks, the retrodirective array has the advantages of being planar
and simple design of routing network. This will make it a perfect candicate as a key
component in the future intelligent vehicle highway system (IVHS).
      This simple design method also leads to an easier design of active retrodirective
arrays and layer of retrodirective arrays. Using the slotline as the building blocks for
the antenna arrays has the advantages of resulting in the array being uniplanar and
wideband. After design of these arrays being completed, experimental realization in
the millimeter-wave band will be carried out to verify the design.
NSC89-2213-E009-195 (89R299)

Title:Theoretical and Experimental Study on Two-Dimensionally Periodic Structures
Investigator:Song-Tsuen Peng
Sponsor:National Science Council
Keywords:2D Periodic Structures, Band Gap, Millimeter-Wave Integrated Circuit,

      The guiding of waves by periodic structures has long been a subject of
continuing interest in the literature; in the past, however, most of the works had been
limited to the case of one-dimensionally periodic structures. Recently, the class of
two-dimensionally periodic structures that are also known as the photonic band gap
(PBG) structures has attracted considerable attention.                   For example,
two-dimensionally periodic structures have been demonstrated to be useful for the
suppression of surface waves in the strip-type transmission lines. So far, most of the
research are experimental in nature and the designs are based on the cut-and-trial
     The purpose of this work is to establish a solid foundation for the understanding
of wave phenomena in two-dimensionally periodic structures and to explore their
possible applications. We shall include both theoretical and experimental studies, so

that the results obtained can be confidently verified and employed for practical design.
Therefore, this research program is organized into three sub-programs:
1. Guidance characteristics of general two-dimensionally periodic structures.
This sub-program will carry out a theoretical study on the propagation of waves in 2D
periodic structures of infinite extent, as a rigorous boundary-value problem. Our
goal is to obtain reliable numerical data in order to establish benchmark results to be
published in the literature and to form a basis for the other two companion
sub-programs to develop potential applications in what follows.
2. Theoretical and experimental studies on strip-type transmission line with finite 2D
periodic structures. In this subprogram, we shall setup an experiment to measure the
phenomena of band gaps with surface wave under different incidence conditions. we
shall study the propagation characteristics of strip-type transmission lines under the
influence of finite 2D periodic structures, with particular attention to be focused on
the band-gap structures of surface waves and strip-type modes.
3. Theoretical and experimental studies on radiation from finite 2D periodic structures.
In this subprogram, we shall study the radiation characteristic of 2D periodic
structures and micro-strip line under the influence of finite 2D periodic structure.
This structure should be also considered as a case of radiation from shielded metal
plate with finite 2D periodic holes, which are designed for heat sink.
NSC89-2213-E009-223 (89R302)

Title:On the Study of RS-Code-Embedded Forward Error Correcting Systems
Investigator:Yu-De Su
Sponsor:National Science Council
Keywords:Error-Correcting Codes, RS Codes, GMD Decoding, Errors-and-Erasure

      Reed-Solomon (RS) codes belong to the prestigious class of maximum distance
separable (MDS) codes that possess powerful error-correcting capability. Besides the
use in deep outer space communications, RS codes have found numerous applications,
from compact disc player, HDTV, to ADSL and 3G wireless communication systems.
In addition to their burst-error-correcting capability, RS codes, as a subclass of BCH
codes, can enhance their error-control capability by employing the so-called
errors-and-erasures (EE) decoding method. Forney had presented a generalized
minimum distance (GMD) decoding method that can, with a proper erasure-insertion
scheme, achieve the performance of a maximum likelihood (ML) decoder with much

less decoding complexity. However, a complete GMD decoder still require a
complexity several order higher than that of a errors-only (EO) or a single-pass EE
decoder. Recent progress in soft-decision decoding of RS and BCH codes has greatly
shortened this complexity gap thus makes GMD decoding of RS feasible and ever
more attractive. In hostile communication environments such as those in wideband
wireless communication and military communication systems, soft-decision decoding
gain is usually much larger than that in AWGN and whence is a welcome design
alternative in relieving the often-very-tightened link margin.
      The aim of this proposed effort is to study the architectures and performances of
forward error-control systems that employ RS codes. Just like designing any
communication system, the difficulty lies in the receiving end, i.e., the design of an
optimal decoding system. The most critical issues are the derivation of the reliability
function of a received symbol, the erasure-insertion scheme and the corresponding
threshold value. These problems have to do with the modulation and demodulation
schemes and the channel statistics. We shall consider i) linear AWGN, multipath
fading and nonlinear satellite channels, ii) orthogonal and bi-orthogonal signals, iii)
coherent and non-coherent detection algorithms and evaluate the corresponding
performance. As in some cases, RS codes alone can not meet the system error-control
requirement, we shall also those cases when other error-correcting codes like
convolutional codes, TCM or turbo codes are used in conjunction with RS codes.
NSC89-2213-E009-183 (89R288)

Title:Autonomous Integrated Receive Systems (AIRS) for W-CDMA Base Stations
Investigator:Yu-De Su
Sponsor:National Science Council
Keywords:W-CDMA, Multipath Fading, DSP, RAKE Receiver

      Given multiple existing commercial wireless W-CDMA proposals, a reliable and
affordable receiver that suits more than one standard is both challenging and
appealing for communication system engineers. One way to achieve this design goal
is to replace the analog circuits by programmable devices. Thanks to recent advances
in wide band analog/digital converter and powerful high-speed digital signal
processors (DSP), the realization of such a versatile receiver has become feasible.
      The main purpose of this proposed effort is to present an FPGA/DSP-based
baseabnd architecture for a reverse link W-CDMA RAKE receiver. Related system

performance analysis/simulations shall also be given. We shall provide detailed
designs for the associated constituent subsystems that are responsible for multiple
access interference (MAI) cancellation, channel estimation, diversity combining, gain
control, soft decoding and all required synchronization mechanisms. Conventionally,
these subsystems are designed and optimized locally, taking only interfaces with
related functional blocks into account. We believe that these receiver functions should
be considered as an integrated unit, considering especially statistical signal
correlations at various stages of the detection process, so that the optimal overall
performance can be achieved.
      Based on our study and numerical optimization results, we hope to build a
prototype receiver for W-CDMA base stations in three years. The first half of the
second year shall concentrate on the system engineering and system analysis aspects;
present an overall system software and hardware architecture along with detailed
algorithms at the end of this period. In the remaining period, we shall direct our effort
toward the refining and prototyping of our design. The proposed effort shall be
divided into three tasks, namely, (1) system architecture, autonomous mode design,
channel estimation, decoding and overall performance evaluation, (2) code generation,
signal correlation, interpolation, RAKE combining and multi-user detection, (3)
synchronization (carrier recovery, code/symbol tracking) subsystems.
NSC89-2219-E009-024 (89R460)

Title:W-CDMA:Base Station Receiver System Architecture and Performance(II)
Investigator:Yu-De Su
Sponsor:National Science Council
Keywords:CDMA, Software Radio, AGC, Interpolation Filter, Spread Spectrum

      The phenomenon growth of the worldwide demand for mobile communications
in the past decade has created one of the world’s most dynamic industry sectors. The
International Telecommunication Union (ITU) has developed a concept known as
IMT-2000 for third generation. IMT-2000 represents a paradigm shift from the old
voice-centric design to true mobile multimedia systems. Of the 10 terrestrial radio
transmission technology (RTT) submissions to ITU, six are based on the W-CDMA
concept. Given multiple existing W-CDMA proposals, a reliable and affordable base
station receiver that suits more than one standard is a reasonable but challenging
engineering solution. Recent advances in wide band analog/digital converter and
powerful high-speed digital signal processors (DSP), however, have made the

realization of such a versatile receiver feasible.
      The proposed effort is part of a larger group effort to design a DSP-based
reverse link W-CDMA RAKE receiver. We shall be responsible for presenting an
FPGA/DSP-based baseabnd architecture and the implementation of some of its
functional blocks, namely, the interpolation filter, code correlation, acquisition, signal
and noise power measurement, and post-synchronization AGC subsystems. As we
plan to complete a prototype system by the end of this 3-year project, we shall start
testing DSP-based subsystem and designing demonstration system hardware
configuration in this year.
NSC89-2219-E009-025 (89R461)

Title:Development of Spatio-Temporal Radio Channel Models for Macrocellular
Investigator:Jenn-Hwan Tarng
Sponsor:National Science Council
Keywords:Wreless Cmmunications, Spatio-temporal Channel Model, Delay Spread,
              Angle Dispersion, Smart Antenna Systems

      The spatial properties of wireless communication channels are extremely
important in determining the performance of smart antenna systems. Classical
channel models provide information on signal power level and Doppler shifts of
received signal. However, modern spatial channel models build upon the classical
understanding of multipath fading and Doppler spread by incorporating additional
concepts such as time delay spread, Angle-of-Arrival (AoA), and adaptive array
antenna geometries. In this project, a physical-statistical spatio-temporal model
for macrocellular environment is proposed. The model combines a site-specific
model and a statistical model describing local scatterers around the mobile unit. The
site-specific model using a deterministic approach describes the effects of physical
large and stationary objects such as buildings and mountains. The latter model
employs a statistical approach to describe multipath components with excess delay
due to the local scatterers. The spatio-temporal properties of the radio channel
predicted by the proposed model will be compared with that predicted by other
models such as Raleigh’s model and Gaussian WSS (Wide Sense Stationary)
uncorrelated scattering model.
NSC89-2213-E009-194 (89R298)

Title:Multilayered Integrated Circuit (MIC) Multi-mode Field Solver
Investigator:Ching-Kuan Tzung
Sponsor:National Science Council
Keywords:MIC, GSM, Millimeter-Wave Circuits, Periodical Structures

      The electromagnetic simulation (EM) and modeling of microwave integrated
circuits (MIC) has improved significantly in the 15 years. While first-pass design for
microstrip MIC ahs been a common practice now, the multi-layered or cpw (coplanar
waveguide) MIC’s are often not the case. For frequency reaching millimeter-wave or
beyond, the departure of simulated results from the measurement is even worse. This
proposal aims to develop a comprehensive software package, allowing accurate
simulation and modeling of millimeter-wave circuits RF front-end circuits and hoping
that next-generation millimeter-wave circuits can be free of constrains imposed by the
softwares available now, either commercial or home-brew.
      This proposal present a main program followed by three sub-projects carried out
by three research groups, respectively. The objective is to develop a millimeter-wave
MIC computer-aided-design (MM-MIC-CAD) software package that, in the main
program, controls and executes all the necessary functions, (Libra, ADS) and
optimization (Libra, Matlab) routine that a designer often requests. Furthermore,
MM-MIC-CAD generates all the I/O needed, e.g., the surface-gridding database for
the three solvers to be developed by the three subprojects, thus greatly enhancing the
productivity of individuals, regardless of the projects that they engage on.
      Subproject I takes the multi-layered periodical structures and produces the
program called periodic-structures-solver, which will be called the main program of
this proposal by manipulating windows, icons and menus, Similary subproject II
develops the fast-algorithm-solver, when needed by the user facing a large problem at
scale at least tens or hundreds of wavelength dimension. To debug the multi-layered
MM MIC, the program GSM-Integral-Solver as developed in subproject III will be
called by the main program to assist designer with clear physical picture based on the
GSM (generalized-scattering-matrix) description of the MM MIC at various locations
(interfaces) demanded by the user.
      Successful implementation of this proposal will provide significant leverage
competing the world-class front-runners on millimeter-wave integrated circuit
technology development.
NSC89-2213-E009-222 (89R339)

Title:Techniques of Millimeter-Wave PCB Hybrid Module (3/3)

Investigator:Ching-Kuan Tzung
Sponsor:National Science Council
Keywords:Packaged Transistor, Hybrid Icrowave Assembly, Hybrid
               Millimeter-Wave Assembly, Millimeter-Wave PCB Hybrid Circuit,
               Millimeter-Wave PCB Hybrid Module

      The advent of high-performance millimeter-wave packaged transistor has been a
reality very recently. This low-cost, high-volume approach for packaging transistors
not only pushes the operating frequency of the discrete transistor to millimeter-wave
regime, but renders very cost-effective method for the printed-circuit board (PCB)
realization of modules and systems at millimeter-wave or upper microwave
      Two candidate circuits are planned for fulfilling the purpose of the feasibility
studies of incorporating the packaged transistors for millimeter-wave module designs.
The first is an oscillator built strictly by soldering iron and typical photolithography
technique. The oscillator is improved upon the internal research carried out at Chiao
Tung University, showing an impressive 20 GHz oscillator using the tools available
from any electronic bench. In the first year of the program the high-performance DRO
(dielectric resonator oscillator) will be developed at K- and Ka-bands. In the second
year the attention will focus on the PCB realization of the traveling-wave amplifier
(TWA) with appropriate noise and gain performances for use in a TR (transmitting
and receiving) module currently being deployed at Chiao Tung University for
millimeter-wave vision scanning application. The TWA was most often realized by
purely monolithic form in the past. This program will demonstrate the applicability of
PCB realization using discrete packaged transistors which normally occupy fairly
large area compared with the wavelengths, thus opening up the research into this
domain. At the third year, both oscillator and TWA will be tailored to fit into the TR
module system, making a portable millimeter-wave vision system a reality for
practical applications.
NSC89-2213-E009-243 (89R341)

Title: 高頻元件量測技術開發暨人員訓練計畫
Principal Investigator: Ching-Kuan Tzung
Sponsor: 三齊微電股份有限公司

      The development of wireless communication products is getting matured and

flourishing. In another word, the high frequency components will be greatly used in
related products. Although to produce these components is a must-do project, but how
to accurately show the function and characteristic of the components to those users is
also a major issue and can not be missed. The only way to achieve the goal of
showing the complete value of these components is based on the accurate
measurement method and technique. Therefore to develop the measurement
technology and well trained the staffs will be essential at this moment.
C89271 (89.09.01-90.08.31)

Title: Power Control for Wide-Band CDMA Wireless Transceivers(1)
Investigator:Li-Chun Wang
Sponsor:National Science Council
Keywords:WCDMA, Close-loop Power Control(CLPC), Antenna Array, Adaptive
              Spreading/ Modulaition, QoS

      Wideband direct-sequence CDMA (W-CDMA) is emerging as the predominant
radio access technology for the third-generation cellular mobile standard. One of key
features for the W-CDMA is the efficient power control scheme to reduce the emitted
interference, thereby increasing system capacity [1,2]. The performance issues for
power control have been extensively studied in the second-generation digital cellular
mobile radio systems, but it is still an active research area for the third-generation
W-CDMA system. The major reason is because in addition to delivering voice
communications as current CDMA systems, the W-CDMA system is designed to
support high data-rate transmission (384 kbps with wide-area coverage and 2 Mbps
with local area coverage). The high data-rate transmission makes power control in
the W-CDMA system more challenging than that in the second-generation digital
cellular mobile radio system.
      This major goal of this project is to investigate how power control algorithms
can integrate with other emerging new techniques, such as adaptive antenna array and
adaptive spreading/modulation, to provide high-speed multi-rate wireless packet data
services with QoS provisioning. In this two year project, we propose to focus our
research in the following two directions:

(1) Measurement-based joint close-loop power control and adaptive antenna array on
    CDMA wireless multi-media communications:
     Power control and adaptive antenna arrays are two efficient approaches to
enhancing performance of wireless communications. In the literature, there are two

contradictory opinions on the effect of antenna diversity in an environment of
imperfect power control. In [30], by using an open-loop power control error model,
the authors analyze the performance of CDMA system with power control error.
Surprisingly, [30] concludes that antenna diversity will exacerbate the degradation
due to the power control error because the power control error statistics are
completely correlated on all branches of the Rake combiner at the receiver. On the
other hand, [33,35] by simulation and field test have shown that combining power
control and adaptive antenna array can significantly improve system capacity and
coverage performance in the TDMA system.
     In our project, we will develop a new antenna array close loop power control
(AA_CLPC) algorithm as shown in Fig. 3 and investigate its performance under a
channel with lognormal shadowing and frequency selective fading. Our results will
not only resolve the important but mysterious question about the performance gain of
combining antenna array and power control in the literature, but will also provide
many insights to what extent the performance improvements can be achieved by
exploring the synergy of antenna array and power control in practice.
(2) Throughput maximization and QoS provisioning for W-CDMA by joint power
    control, adaptive antenna array and adaptive spreading/modulation:
      In data communications, the major objective is to maximize system throughput
subject to certain quality of services (QoS) constraints, e.g. packet error rate. The
problem itself is no longer to introduce fairness as the function of power control in
voice CDMA system, but to improve capacity by a “best-effort” resolution. Since the
higher SINRs permit allocation of higher data rates to users, power control seemingly
plays no role in improving system throughput for variable rate systems, especially
when other rate adaptation techniques like adaptive modulation [14,15] are applied
together with power control.
      Recently, my previous works in [16,17] explored the potential of power control
for the variable rate systems in another direction -- QoS provisioning. In [16,17], it
has been proved that the problem of maximizing data throughput by adaptive
modulation and power control while meeting packet error requirement in the TDMA
system is NP-complete. However, a heuristic but efficient algorithm for integrated
link adaptation and power control is still can be found.
      In this project, we propose to develop a call admission policy that jointly
optimizes the power assignment, rate selection and QoS provisioning for the multirate
CDMA system. We will extend the joint adaptation algorithm in [17] to the CDMA
system and then integrate it with the measurement-based joint adaptive antenna array
and power control algorithm developed in the first year. Taking advantage of adaptive
antenna array for fast and accurate measurement of interference, our call admission

policy algorithm will fulfill the following ambitious objectives : throughout
maximization, robustness to power control error, and QoS provisioning.
NSC89-2213-E009-246 (89R524)

Title:A Method to Design the Reward Function in Keyword Spotting System
Investigator:Yih-Ru Wang
Sponsor:National Science Council
Keywords:Selective Training, Confusion Measure, Reward Function, Keyword

      In this project, a new Mandarin keyword spotting system using reward functions
to discriminate keywordsfrom background fillers is proposed. It first extends the
confusion measures between base-syllable pairsused in our previous study to be
confusion measures between keywords and the filler models. It then uses these
confusion measures to design reward functions to be added to keywords for standing
them out from background fillers. Keyword spotting can be regarded as a signal
detection problem under the false alarm constraint. The structure of the new keyword
spotting system is based on the optimal solution of the signal detection problem. The
project will first apply the selective training method to find a better set of HMM
models which is robust to erroneous syllable labelings existing in the training
database. The selective training will remove the influence of those outlier syllables
with low confidence. Then, the confusion measures of a11 syllable-pairs will be found
from a large verification data set. A keyword spotting system using 411 syllable
models as fillers will be constructed. The confusion measure between keywords and
the fillers will then be determined from these confusion measures of syllable-pairs.
Finally, the system structure that finds from the Max-Min simultaneous
detection/classification problem will be used in the new keyword spotting system. The
reward function of a keyword will be designed from the power of test of the keyword
and the best filler hypotheses. It is therefore not needed to retrain the system when the
keyword set is changed.
NSC89-2213-E009-190 (89R295)

Title:The Effects of Amplitude and Phase Ripples in Optical Fiber Gratings on
       Analog and Digital Subcarrier Multiplexed Lightwave Systems
Investigator:Winston Way
Sponsor:National Science Council

Keywords:Fiber Grating, Subcarrier Multiplexed Lightwave System, Dispersion

     The amplitude and phase ripples of a chirped fiber Bragg grating (CFBG) can
cause severe signal degradation in baseband 10 Gb/s transmission systems.
However, the ripple effects in subcarrier multiplexed (SCM) lightwave systems have
not been investigated thoroughly. In this project, we will investigate analytically and
experimentally the effects of CFBG amplitude/phase ripples in (1) direct-modulation
hybrid fiber coax (HFC) systems, and (2) external-modulated high-speed
telecommunication SCM systems. We will also provide a design guideline which
relates the nonlinear distortions, channel numbers, data rates, and transmission
distance with the phase ripple period and the peak-to-peak amplitude ripple variation.
NSC89-2215-E009-081 (89R495)

Title:Characterization of Embedded Passives on Multilayer PCB
Investigator: Lin-Kun Wu
Sponsor:National Science Council
Keywords:FDTD, MoM, Spectral Estimation, Embedded Passives, Multilayer PCBs,
               Conductor and Dielectric Losses, Parasitic Coupling, Finite-sized Solid

     A combined Finite-Difference Time-Domain method (FDTD)/Method of
Moment(MoM)/Spectral Estimation technique is proposed to determine broadband
characteristics of embedded passives for use with multilayer PCBs. The FDTD is to
be used to obtain a rough estimate of component's frequency response with conductor
and dielectric losses ignored. From which, particular frequency points will then be
selected for the MoM analysis. Effects of frequency-dependent material losses will be
accounted for in the MoM analysis. Finally, a spectral estimation technique will be
used to extend the selected frequency data provided by the MoM analysis into
continuous, broadband, characteristics of the component for use with circuit design.
Aside from the effects of frequency-dependent material losses, effects of parasitic
coupling between the component and the finite-sized solid plane(s) (e.g., ground
and/or power plane) inherent in a multilayer PCB will also be investigated.
NSC89-2213-E009-196 (89R300)

Title:Adaptive Channel Equalization Using Efficient Kalman Algorithms

Investigator:Wen-Rong Wu
Sponsor:National Science Council
Keywords:Channel Equalization, Kalman Filtering, Kalman Equalizer.

      It is known that the Kalman equalizer (KE) has superior efficiency than the
finite impulse response (FIR) equalizer. This is due to the infinite impulse response
(IIR) characteristics of the Kalman filter. Theoretically, the adaptive IIR filter can has
the similar performance with the KE, however, it suffers many problems such as slow
convergence, local minimum, and stability problems. In contrast, the KE converges
fast and stably to the global minimum. In general, the nonlinear decision feedback
equalizer (DFE) can outperform the liner equalizer. However, there are some
modulation techniques in which the DFE cannot be effectively applied (for example,
TCM, CPM, and OFDM). Because of the high computational complexity problem, .
the KE has not widespreadly used. This project proposes an efficient algorithm
overcoming this problem. Using a novel method, we derive a closed-form expression
for the steady-state Kalman gain. This steady-state Kalman gain is then used in the
filtering process. Using this approach, no matrix operations are required and a
tremendous reduction in computation is obtained. It can be shown that the
computational complexity for the conventional KE is on the order of O(N 2 ) where N
is the filter length, while that for the proposed KE is on the order of O(N). For the
time-varying channel, the steady-state Kalman gain is also time-varying. We then
propose an adaptive method to cope with the problem.
NSC89-2213-E009-161 (89R285)

Title Study and Implementation of the Correlator Subsystem and Multiuser Detection
     for W-CDMA Systems
Investigator:Wen-Rong Wu
Sponsor:National Science Council
Keywords:Correlator, Multiuser Detection, Multiaccess Interference Cancellation

     Due to its higher capacity and flexibility for variable rate uses, CDMA is
considered to be the most appropriate access method for the next generation wireless
communication system, known as IMT-2000. A essential device in the CDMA
receiver is the correlator. The correlator is to achieve symbol synchronization in the
code acquisition mode. And, to despread the input signal as well as provide timing
information in code tracking mode. It is known that the computational complexity of

the correlater is very high. Thus, it is not suitable to implement the correlater using a
DSP processor. In the first part of this project, we consider the correlator
implementation using FPGA. We use a highly efficient algorithm which not only can
greatly reduce the computations, but also allow a highly pipelining structure. In the
second part of this project, we consider the low-complexity multiuser detection
algorithms. It is well known that the multiuser detection can significantly increase the
system capacity. However, most existing algorithms are too complex to implement.
We propose a new parallel interference cancellation (PIC) algorithm. This algorithm
retains the low complexity nature of the existing PIC algorithm and improves their
performance. We will use FPGA as well as a DSP processor for PIC implementation.
Note that the operations realized using the FPGA are similar to those in the correlator.
Thus, they can be combined into the correlator resulting a correlator subsystem.
NSC89-2219-E009-026 (89R462)


To top