Authors:  Jimmy Chang, Paul Chow, Alexander Reiss, Anthony Tran, Tony Yang Ki

Advisor:  Jan Van der Spiegel

The project focused on learning the original architecture of the ENIAC
and implementing the main units (accumulators, master programmer, and
cycling unit) on an Field Programmable Gate Array while preserving as
much as possible the original operation.  The Xilinx Foundation software
was used to write and implement the VHDL design code, simulate the parts,
and download the design on to the hardware.  The final design, consisting
of two accumulators, the master programmer and the cycling unit, was
constructed by wire-wrapping the Xilinx XC4044XL BG352 chip and socket.
The PC is used as the input device and seven segment LEDs display the
accumulator outputs.

Programmable Electronic Cochlea for Voice Recognition

Authors: Albert Cheng and Brad Edelman

Advisors : Jan Van der Spiegel, Paul Mueller, Ahmed Abdelatty and Ilija Hadjic

Traditional attempts at speech recognition have been plagued by
critical problems such as understanding different voices and adapting to
larger vocabularies.  This new approach, modeled after the human cochlea, breaks
down sound into separate frequency components before matching these components to form
phoneme blocks in a neural computer.  This system improves speed and phoneme level
recognition accuracy in presence of noise by using biologically-oriented signal processing.

The electronic cochlea is composed of a cascade of 36 fixed
low-pass and 36 programmable band-pass filters, tuned to specific frequencies from
300-7800 Hz.  Through a software interface, the user can send band-pass filter settings to an
FPGA located on an ISA card to the main filter board.   Outputs from the filter system can then
be multiplexed by an A/D converter for further processing by a neural computer.

Imaging with Near Infrared Light

Author: Seelig Sinton

Advisors:  Arjun Yodh, Joseph Culver

Imaging with near infrared light is a new technology.  Reconstructing
the image is a very tedious process that requires switching between
multiple programs and making all the numbers match.  A GUI (Graphical
User Interface) created in Visual Basic enables the user to specify the
parameters required to perform the image reconstruction.  The parameters
as well as the measured data are saved as text files where they can be
accessed by Matlab.  Matlab has superior mathematics and graphics
capabilities which make it an ideal medium for performing image
reconstructions.  By typing a single command, the image will be
reconstructed and displayed for viewing thus simplifying the entire
image reconstruction process.

 Nanometer Resolution Position Detector

 Author: Alexandru Cernicanu

 Advisor: John Crocker
 
 A nanometer resolution position detector was designed and built
 on a Zeiss Axiovert 135 microscope.Making use of the laser trap of the
 microscope, the system is able to detect the brownian motion of a
 trapped micrometer sized particle.The output of the circuit (a normalizing
 differential amplifier) was interpreted using a spectrum analyzer, and
 information about the motion of the particle and the trap was derived.

This electronic detection instrument adds to the imaging detection
scheme, and is a important tool in researching the dynamic behaviour of
micrometer sized colloids.

Authors: Sidhartha Mathai, Jonathan Sherry

Advisor: Arjun Yodh

Modern medicine has developed numerous reliable methods to detect tumors in
the human body. However, many of these methods are inconvenient, expensive,
and sometimes even harmful. In addition, these devices are not readily
capable of performing real-time imaging. The Multi-Channel Near-Infrared
Patient Monitor is a device that will lessen the effects of these negative
factors. First, lasers with a wavelength of about 750 nm will be used as the
sources. Next, the circuitry will be in a rather small box that is capable
of being easily transported. Finally, the lasers and detectors will be
mounted on an ergonomically engineered pad that will lie against the skin.
The quick interaction between the sources and detectors will allow for an
immediate image of the environment to be calculated, eliminating the wait
for hardcopy results. The main difference between this product and others is
its speed. Therefore, the purpose of the Patient Monitor will be to provide
a very quick method of determining problems without having to wait for a
long time. This design is to be placed on the arm to detect blood saturation
levels only located in the forearm. However, once completed, the design will
have the ability to be modified to accommodate other parts of the body where
tumors are more prevalent, such as the head or chest. Furthermore, the
accuracy of the test grows exponentially with respect to the number of
sources and detectors. This project includes six lasers and twelve
detectors, but could be modified in the future to operate with many more
lasers and detectors.

GPS Emergency Alert System

Authors: Harris Kyriakopoulos, Pelopidas Nicolaides

Advisor: Phil Farnum

The purpose of the project is to design the GPS Emergency Alert System.
It is a personal security system, composed of a
Portable Activation Device (PAD) and a Home Base. Customers carry the
PAD, and in the event of an emergency they can
instantaneously notify the system operators of their location, via the
Home Base.  The operation of the system is based on
receiving the location coordinates of the customer with a GPS receiver
and transmitting this location to the Home Base through
a high frequency radio link. Appropriate software at the Home Base
allows the system operator to  view the coordinate location
on a map.

Wireless Headphones

Authors: Patrick Cahill, William Martin

Advisor: Philip Farnum
 
Perhaps the fastest growing technology in the modern electronics industry
is the area of wireless communications.  The advent of cellular
telephones, pagers,and other forms of transmission devices are growing in number and rising
in capability and quality.  The goal of this project is to bring wireless transmission
to an activity that almost everyone enjoys- listening to music.  People already enjoy
listening to theradio while they engage in other activities, such as yardwork or cleaning
the house.

Now they can do the same activities but instead of listening to only
radio, they can listen to a favorite compact disc.  The user simply connects the output
of their compact disc player to our transmitter system.  The music will be broadcast via
an FM station to a portable receiver, which ouputs to a pair of headphones.  The user can
enjoy the music he likes, without DJ talk or commercials.

Bicycle Location System

Authors: Brian Jessup and Baehyun Sung

Advisor: Phil Farnum

There are several different methods of locating an RF signal, some of
which include: null-seeking antennas, measurement of phase, and GPS.  The
goal for this project was to develop a direction finding switched
parasitic antenna array.  The array consists of a switching system to
collect data from each of 4 monopoles, a receiver, and a micro-controller
to perform the switching and to transfer data to a computer.  The array
detects signals from a portable transmitter and relays that information
to a computer.  The angle of signal origin is then determined by the
ratio of the two strongest signals.

Motion Control System of an Autonomous Robot Vehicle

Author: Hun Choo

Advisor: Vijay Kumar

An autonomous robot capable of ascending a flight of stairs was constructed,
and its components are described in detail in this report.  Construction of
the robot entailed the combination of a RC Servo driver and four RC Servo
Motors for climbing movements, two motor drivers and four geared-down DC
motors for driving and steering capabilities, a microcontroller, and code in
Assembly language to synchronize the systems.  A distance sensing system,
developed by another design group, will be interfaced with the robot for
vision capabilities to allow the robot to run and climb stairs autonomously.
The entire system will be powered by a 12-Volt battery, and it was
determined the robot will run for approximately an hour.  The speed of the
robot's driving movement was determined to be one foot per second, and its
climbing movement to be one step in thirty seconds.

Robotic Distance Sensing System

Authors: Sevile George, Michal Levi

Advisor: Vijay Kumar

A distance sensing system was developed for a robot that can travel over
unstructured terrain. The system consists of four ultrasonic
transmitter/receivers strategically placed on the front of the robot so
as to convey useful information for motor control decision making. A
microcontroller is used to control the ultrasonic devices by
consecutively firing and receiving signals from each of the four sensors,
in continuous one-second time intervals. The time differences between the
firing and reception are then used to determine the distances detected by
the sensors. The distance data is transmitted in real time from the
microcontroller to two sources. It is sent through a parallel interface
to another microcontroller used to control the motors on the robot. In
addition, the data is transmitted serially through an RS-232 link to a
graphic display on a personal computer.

The Parking Assistant

Authors: Deepak Sindwani, Brian Yoshino

Advisor: Phil Lopresti

The parking assistant is an integrated automobile positioning system
designed to assist drivers when parking.  Damage to cars and trucks
caused by low-speed collisions are frequently costly and annoying.
Furthermore, it is often inconvenient for truck drivers to require
additional manpower to assist them when parking in reverse into loading
docks.  The parking assistant addresses these problems and provides a
cost-effective solution that eases the task of parking.

The parking assistant consists of a car-mounted collision avoidance
system and a wall-mounted guidance system.  Both functions are achieved
through the use of an ultrasonic distance detection system.  The
collision avoidance system broadcasts the distance from the bumper of
the automobile to another object (i.e. another car or a wall) to the
driver on an LED display located in the cabin of the automobile. The
distance display is updated at a frequency of 1 Hz (1 cycle per second)
to ensure timely and accurate distance readings. A warning buzzer goes
off when the distance becomes less than 2 feet to warn the driver of an
imminent collision.  The guidance system incorporates a wall-mounted
device that employs left/right directional arrows to assist drivers with
aligning the automobile in parking spaces. The parking assistant has a
range from 1 foot to 35 feet and assists the driver to realize an „on
center‰ lateral accuracy to within 0.5 feet.

Electronics Control System for a Robotic Arm

Authors: Peter Carpi, Matthew Olster

Advisor : Phil Lopresti

A previously constructed robotic arm, for use by handicapped people
limited to a wheelchair, needed a desktop computer running C code to
control its movement.  Since the objective is to create a portable system
that can be mounted on a wheelchair, a desktop computer is not efficient
in terms of space, power, or cost.

A vast reduction in size, power and cost can be realized by replacing the
computer (processor and hard drive) with a microcontroller (with RAM or
EEPROM on-board).

Through use of assembly language the previous control system has been
reduced to a set of chips, currently housed on various test boards, which
can be reduced further onto one printed circuit board for increased
manufacturability, cost and size reduction.

Videoconferencing Using an Omnidirectional Camera and Audio Speaker Localization

Authors: Laura Hornbeck, Bjorn-Steven Lindgren

Advisor: Kostas Daniilidis

In two-way videoconferencing involving multiple persons, an improvement
in the quality of resolution by means of transmitting the image of just
one speaker needs to be made.  To do this and to avoid using a rotating
camera or multiple cameras, a panoramic 360* Field of View camera will be
used.  Image processing will undistort the images of each speaker,
separately, given the location of the speakers.  The goal of this project
is to determine speaker localization using an array of microphones.
Finding the speaker's location consists of two steps: estimating the time
delay of arrival to the array of microphones using cross-correlation and
computing the speaker's location by intersecting the hyperbola arising
from each microphone pair.

Blind Separation of Speech Signals

Authors: Wah Kheng Tan, Su Lynn Wong

Advisor: Saleem A. Kassam

Blind Source Separation is a method for extracting the original source
signals from a random mix of two signals without any extra information from
the source. Given the scenario of two randomly mixed speech signals, this
project demonstrates that it is possible to separate these two signals even
though the receiver has no prior knowledge of how they were mixed. By
assuming that the two sources are independent, an algorithm that maximizes
the statistical independence of the sources can separate out the individual
signals. This process of blind source separation was successfully
implemented on a Motorola DSP 56303 EVM board. To accomplish this, much
research was conducted into the existing algorithms, and those algorithms
most appropriate for speech separation were simulated on MATLAB to test
their performance and robustness. The selected algorithm, proposed by
Amari, Cichocki and Yang, was chosen for its ability to perform uniformly
well under a wide range of input Signal-to-Noise Ratios. This good
performance was repeated on the Motorola DSP board, consistently separating
randomly mixed input speech samples.

Blind Source Separation has significant applications for
telecommunications, where the perpetual objective is to maximize the use of
available channels for information transfer. One example is that signals
modulated with different techniques can be separated in a similar fashion,
without the need for extraneous modulation type information.

An Actuator for a Process Control Valve Using LTCC Tape

Authors: Olivia Koentjoro, Kelum Pinnaduwage

Advisor: Jorge Santiago

Increasingly materials such as LTCC (Low - Temperature Co - fired
Ceramic) tape are thought of as an alternative to silicon, especially in
the fabrication of three-dimensional Meso (intermediate) and MEMS (Micro
Electro Mechanical System) devices. Therefore, a significant amount of
research on LTCC is called for in order to realize it's potential in
such applications. An Electro Magnetic actuator that could be used in
the creation of devices such as process control valves was built. It
consists mainly of two parts, namely a diaphragm and an electromagnet.
Diaphragms were initially fabricated using both LTCC and Photoformable
tape. However, due to the fragile nature of the fired tape, the etched
diaphragms did not provide the functionality sought. Therefore, a
diaphragm etched in Molybdenum was used to build and test the actuator.
The electromagnet was created using screen-printing technology. Silver
paste was printed on layers of unfired LTCC tape. These layers were
inter-connected together using a lamination technique and then fired.
Using these two components a functional actuator of dimensions 17 mm *
16 mm * 4 mm, was built. This leads to the conclusion that LTCC tape can
be used in fabricating at least parts of Meso - EMS (Electromechanical
System) devices successfully.

Ferrofluid Pump

Author: Jason Aron

Advisor: Jorge Santiago

A design was completed and the first steps achieved in producing a pump
which uses no solid moving parts.  This is possible using a ferrofluid, a
solution of magnetic particles, to produce the propelling force.  In most
mechanized systems the lifetime of a product can be reduced by vibration
from nearby devices.  This project attempted to reduce the vibration and
noise produced by a fluid pump by eliminating its moving solid parts.  A
unique system was designed in which a contained ferrofluid transforms
magnetic energy into a fluid-propelling force.  By sequentially driving a
series of electromagnets the ferrofluid can be shaped into a pulse, driving
water through a tube. The design included shaping the dynamic magnetic
field, arranging solenoids to achieve this field, and constructing a stable
but flexible barrier between the water and the ferrofluid.

A problem was encountered late in the project, which could not be overcome
using available equipment.  A large enough magnetic gradient could be not
generated with which to form a ferrofluid pulse.  Although the system is not
fully operational, many questions proposed by the design have been answered
by observing the partial pulses achieved.  These conclusions have reaffirmed
the feasibility of the final design, and suggested some improvements that
should be implemented in future projects.

Clock Distribution System for CDF

 Authors: Sung Hwan Chung, Hyun Suk Lee, James Lin

 Advisor : Matthew Jones

 The Collider Detector at Fermilab (CDF) is a huge particle detector which
 helps physicists make discoveries about the nature of matter and energy.
 The High Energy Physics (HEP) group at Penn has been working on an
 upgrade of this CDF, named Time Of Flight (TOF). Our project focused on
 clock distribution aspect of TOF, which requires an accurate measurement
 of time between two pulses with 25 pico second precision. The project
 was divided into three parts.

 The first part focused on studying drift of alpha. Alpha is the extra
 delay added to the time between two pulses so that a Time to Digital
 Converter (TDC) can read it out with more precision than the TDC's
 resolution would allow. A 50 ps resolution TDC was used to measure the
 delay of a pulse around a 40 feet twisted pair cable. A programmable delay
 chip was used to add an extra delay to the returning pulse. The TDC was
 read out by a PC. After extensive measurements, drift of alpha was computed.

 The second part was monitoring temperature along this cable. Highly
 sensitive temperature sensor chips were placed along the cable and
 interfaced to a PC. Their temperature were read out together with
 measurements of the TDC. Finally, correlation between temperature and
 change of alpha was computed.

 The third part focused on producing a prototype of a TDC to be used in
 TOF. This TDC consists of a Time to Analog Converter (TAC), infinite
 impedance amplifiers, a 12 bit ADC, a clocking circuitry, and latches.
 The TDC yields a minimum resolution of 25 ps, smaller than what typical
 commercial TDCs would yield, and a range of 100 ns.

Tools, Techniques and Devices for Silicon-on-Insulator Process Integrated Circuits

Author: Yasmin Beg

Advisor: Rick Van Berg

The BiCMOS capabilities and radiation hardness of the DMILL Silicon-on-Insulator (SOI) process make it ideal for application in high energy physics research. The Amplifier-Shaper-Discriminator-and-Baseline-Restorer (ASDBLR) ASIC, fabricated in the DMILL process, exhibits oscillations at the output due to the floating back substrate, a property of the SOI process. Each device on the chip is coupled to the back substrate through the parasitic capacitance. The potential of a shielding layer fabricated underneath each resistor and capacitor can be controlled in such a way as to reduce the coupling effects of the parasitic capacitance. By examining the semiconductor physics of the DMILL process, this project developed computer-aided-design tools for more accurate simulations and created shielded resistive and capacitive devices that reduce oscillations on the output of the ASDBLR.

Adaptive Brain Oxymeter

Author: Clara Dimas

Advisors: Britton Chance, Hong Ma

A frequency domain in-phase and quadrature (I&Q) oxymeter uses NIR light
to detect the hemoglobin concentration and saturation of tissue such as
the brain. The goal of this project is to minimize signal amplitude
interference in signal phase measurement through software algorithms. This
involved a characterization of the reproducibility, linearity, noise, and
tolerances of the system components. The corrected values of amplitude and
phase is a function of various factors including initial phase and
amplitude, voltage driving the PMT, and the above system characteristics.
The algorithm was tested with optical density filters attenuating
the signal through air, and with a blood sample.

The Wearable Sensor of Hematoma

Author: Jong Wan Kim

Advisor: Britton Chance

A portable hematoma sensor using near-infrared detection technique was
designed and fabricated for use by Emergency Medical Technicians (EMTs).
The detector has a volume of 12.0 cm X 6.7 cm X 3.3 cm and weighs 1.5
pounds. It displays three different evaluations of patients conditions:
* Hematoma detected - emergency treatment
* No Hematoma
* Possible small hematoma - not a emergency condition

It is very important to have early detection of intracranial hematoma
when patients have a severe head injury. Computerized tomography(CT) is
the standard imaging technique to detect these hematomas. However, in
certain areas where CT scanning is not available for immediate use,
portable size of the near-infrared spectroscopy can be used for early
detection of hematoma.

The portable hematoma detector is designed to perform three main
functions to detect hematoma. These functions are:

1. Near-infrared sourcing and detection.
2. Amplification and analog to digital conversion of input signal.
3. Dataprocessing

Online Video Conferencing Customer Support

Authors: Nina Yeuk-Wai Hung, Ting Fai Ng, Kittiphongse Yongkittikul

Advisor: Magda El Zarki
 
The goal of this project is to design a Cyber Brokerage, an online video conferencing application so that financial consultants can converse with their clients while gaining the ability to share portfolio or stock data and diagrams with them. The support features are Client Verification System, Client Queuing System and Automated Portfolio Summary with the integration of the Microsoft’s NetMeeting 2.1. Moreover, two separate web pages are specially designed to meet the different needs of clients and brokers. The clients’ web page consists of login, registration, company and contact information, and call-your-broker page, which activates NetMeeting automatically. The brokers’ web page is comprised of status form, call-back list (which activiates NetMeeting automatically) and check status form. The web pages are implemented using CGI programming, PERL, HTML, JAVASCRIPT and NetMeeting Objects.  A FoxPro database is developed for testing of analysis tools and the prototype of this project. Bandwidth issues with regard to application sharing were investigated and the result is that the application functions on a real time basis under Ethernet connections, but has significant delay problems when using slower networks, such as modem connection. The suggested future tasks for the project is to connect the developed application to real time stock quotes.

Data Acquisition System for the Mobile Anisotropy Telescope

Authors: Jonathan Bruzzi, Harvey Chapman

Advisor : Mark Devlin

The Mobile Anisotropy Telescope group's study of the evolution of the
universe involves the precise measurement of many small signals embedded
within a sky full of noise.  This requires the use of numerous
sophisticated and sensitive detectors.  To take full advantage of these
detectors requires a data acquisition system that is accurate, precise,
and possesses the capability of handling a large number of signals.  To
date, no commercially available product exists meeting the specifications
required by the MAT group.  This project attempts to satisfy the demands
of the MAT group in a manner that is flexible, scalable, and capable of
evolving to suit future needs.  It is autonomously capable of handling up
to 255 double-ended channels with 16-bit resolution.  Individual channel
sampling rates may be set from 2 Hz to 6 kHz.  Separate conversion boards
handle analog-to-digital conversion greatly reducing degradation of input
signals.  The conversion boards link via a 50-pin cable to a primary
control board housed inside a personal computer.  System software avoids
taxing the computer's resources through the use of 512 Kilobyte
First-In-First-Out data buffers and provides an easy-to-use interface for
specifying which channels are to be sampled and how often.

Active Color Spotlight

Author: Harris Romanoff

Advisors: Nader Engheta, Jim Ostrwoski

Theatrical productions, musical concerts, and other entertainment events
frequently use numerous individual spot lights to enhance their
performances.  Each light is manually adjusted before the show to set its
three fundamental features: color, direction and beam size.  This method,
however, requires that the lighting designer use a different light for any
given set of features.  For example, eight individual lights would be
required to produce eight colors.  In addition, once the performance is
underway, making manual adjustments to the lights is often hazardous, if
not impossible, because of their position over the stage.  To overcome
these difficulties, several companies have introduced computer controlled
intelligent lights that allow for remote adjustment of the above features.
However, these lights (each approx. $800 - $50,000) and the required
programmable controller (approx. $3000 - $50,000) are too expensive for many
small venues and individuals to afford.

The primary objective of this senior design project was to design and
build a comparable computer controlled light, but one which differed in
two fundamental ways.  First, instead of requiring the user to purchase a
specialized computer controller, the user would now use a standard PC,
requiring only the purchase of software.  Second, a light would be built
at a fraction of the cost by using standard parts, while still maintaining
the three fundamental features.  Both of these objectives were
successfully achieved with the production of the Active Color Spotlight,
now available for $350

Audio Digital Signal Processing

Author: Vinod Mapranath

Advisor: Santosh Venkatesh

Noisy audio signals with chalk scratches and unwanted dead space were
eliminated from a Digital Audio Tape.  Chalk scratches and dead space are
distinguished from a lecturer's voice or music by implementing an
algorithm achieved through programming of the Motorola DSP56303EVM Digital
Signal Processor in Assembly Language.  This algorithm for eliminating
chalk scratches and unwanted dead space is dependent on calculating
average values for half second data samples.  This average value is
compared to a predetermined threshold and classifies each half second as a
desired or undesired signal.  It has been determined that by sampling at
9.6 kHz with 16 bit resolution, a threshold in the DSP corresponding to
the hex number FFC000 will distinguish chalk scratches and empty space
from selected music.  This hex number coincides with over a 75 dB
difference in threshold for chalk scratches and selected music.

The Asynchronous Classroom

Authors: Jesse Leung and Victor Lam

Advisor: Dwight Jaggard

This project studied the development of a multi-medium, interactive,
asynchronous classroom designed for the student user.  The main development
tools for this project includes RealAudio, RealVideo, Javascript, PERL,
NetMeeting and HTML.  This classroom website incorporates an online
syllabus, online book, lectures on demand (in both RealAudio and RealVideo
formats), lectures slides, office hour meetings, and student communications.
This project and the information obtained in its creation will be used by
those people working on the Alfred P. Sloan grant in an effort to improve
the distance learning experience offered at the School of Engineering.

Motor Control in Ocular Studies

Author: John Horrocks

Advisor: David Solomon

In humans, the vestibular and ocular systems works together to help balance
the body as well as give a sense motion.  It is possible to gain
information about how these systems function by studying eye movements when
following an object.  In order for the data collected to be useful, it is
necessary that the position of the object relative to the subject be known.
 It is also useful to be able to program the object in order to study
different types of movement.  The problem is how to precisely position a
laser beam on a one meter by one meter screen.  As opposed to trying to
move the laser, this project uses two small servomotors to control mirrors,
which reflect the laser beam.  LabView, a software program, will be used to
send a signal to a motor control chip, which will define the paths of the
motors.  Using the defined allowable tolerances of the position of the
laser on the screen, the necessary tolerances for the mirrors was found and
used to design a carriage system for the laser and two servo motors.  Using
encoder feedback from the motors, LabView VI's were created to enable
precise control of the laser on the screen.  Ultimately, the hope is to
have the program advanced far enough so that the researcher can program the
path for the laser point without a lot of training on LabView.

Updated April 14, 1999
Return to EE441/442 Home Page
Return to the EE Home Page

Thsi site administered by Siddharth Deliwala, deliwala@ee.upenn.edu