Current Projects:
Nano-scale Molecular Electronics: Architecture, Test and CAD Issues
As CMOS technology for implementing digital integrated circuits continues to scale down it will face theoretical and also economical difficulties. Emerging technologies to overcome these obstacles are growing up in different fields. One of strongest threads of research in this area is based on using chemically created nano-scale devices like molecular switches, nanowires, nanotubes and nano-FETs in digital circuits. Limitations in assembly technology of these devices results in only regular structures like arrays. So designers will have to use these regular structure to design their architectures.
These emerging technologies also face another limitation. High defect densities are expected in them. So, suggested designs must be highly defect tolerant and testable. As design of architectures based on these technology develops, need for appropriate CAD tools in this field becomes obvious. Since the design paradigm in these technologies will be completely different in terms of architecture and device, proper CAD tools at all levels of abstraction will be required. This project focuses on Architecture, Test and CAD issues on emerging nano technologies. The aim here is to search and try to find solutions to the wide range of problems in each of these fields. Secure Design
Security has become a greater concern in the design and test of chips
recently. This has become more apparent with the advent of
cryptochips. Cryptochips perform encryption and decryption algorithms
at the circuit level. Many researchers have been able to show that
these chips are highly vulnerable to various power analysis, timing,
and fault-injection attacks if not specifically designed with
countermeasures in mind. If not considered carefully, strong
encryption algorithms that would take years to crack by brute force
can otherwise be crippled in a manner of weeks, days, or even hours
through these side-channel attacks.
Currently, the main objective in testing has been to control and observe a chip as much as possible in order to achieve high fault coverage and diagnosis on the CUT. As useful as these properties are for testing, they are completely contradictory to the objectives of security on a chip. In order to protect a chip from malicious users, a chip must reveal as little as possible while still considered useful to the end-user, but for reliability, a test engineer needs as much access to the chip as possible. Recently, scan test has been proven a security risk to the intellectual property (IP) on the chip. Some researchers were able to simulate an attack on the scan chain of a DES cryptochip to reveal the secret key with using only three plaintexts. Although the scan chains have only been exploited to find the secret key of a cryptochip, it is just as easy to uncover proprietary intellectual property through scan chains. Vital registers are part of the scan chains that are allowing high controllability and observability. This amount of access has been shown to reveal a lot more than the fabrication defects in a chip. In order to prevent IP theft, security measures must be implemented during the design phase. At-Speed Transition Fault Testing using Low-Cost Testers (Sponsored by SRC)
With today’s design size in millions of gates and working frequency
in gigahertz range, timing-related defects constitute a large
proportion of the defect population, and at-speed test is
crucial. Transition and path delay fault testing together provide a
good coverage for delay-induced defects. The
transition fault model (TFM) is widely practiced in the industry since
it leads to manageable number of faults. There are two transition
fault pattern generation methods, namely launch-off-last-shift (LOS)
and launch-off-capture (LOC). In LOS, the scan enable signal SEN must
change at-speed. In LOC, the at-speed constraint on the SEN signal is
relaxed and dead cycles are added after the last shift to provide
enough time for SEN to settle low. LOS provides higher fault coverage
with lower number of patterns compared to LOC. The LOC technique is
based on a sequential ATPG algorithm while the LOS method uses a
combinational ATPG algorithm. This will increase the test pattern
generation time in case of LOC and a high fault coverage cannot be
guaranteed. The main limitation for practice of LOS is that it
requires the SEN signal to be timing critical and is not applicable on
low cost testers. As the flop count increases, the SEN fanout becomes
excessive and routing area increases dramatically. Considering
thousands of flip-flops in today’s large designs, implementation of
an at-speed SEN in case of LOS can be as costly as a clock
synthesis. In contrast, LOC is easier to implement using low cost
testers due to its independency on the at-speed SEN signal; however it
provides lower test coverage. Other major challenges for both LOC and
LOS methods are as follows: the routing area must be decreased and the
test coverage must be increased to ensure test quality.
Power Supply Noise Effects in Deep Submicron
The power supply noise (PSN) reduces the actual voltage level reaching a device,
which increases the signal delay and results in signal integrity loss and performance d
egradation. It may also cause logic errors, degradation in switching speed
and hence timing errors. In this project, we propose pattern generation
algorithms that target power supply noise more accurately with less time by
reducing the pattern search space. We also investigate PSN effects on path
delay, crosstalk and IR-drop. As technology is shrinking, the PSN effects are
becoming more prominent.
|