REQUEST: REPORT
ID: 59703
P-NAME: radix8
Fab-ID: T01SDD
P-P: /* I have on file the project password */
REPORT:

Functional Description:
The chip was a radix-8 multiplier that took two 9 bit inputs and produced
an 18 bit output. The inputs were given sequentially, 9 bits at a time,
and the output was multiplexed so that a signal selected the low 9 bits
or the high 9 bits.

Yield and Functionality:
Five out of five chips yielded correct results on all tests
performed.

Testing:
The same tests used in the pre-silicon simulations were used to test the
chips. These tests had already been determined to test all parts of the
algorithm, though not necessarily all internal wires. As an attempt to test
all internal wires without doing an exhaustive test, we made sure that we
tested all combinations of two numbers that were power's of two. The
reasoning was that this would ensure that we could force every possible
path internally to go both high and low. Ideally, with better test
equipment, it would have been possible to do more sophisticated
testing. Considering that our chip had very little state, we could have
feasibly done an exhaustive test with a more automated test setup.

Speed testing:
Using spice on the components on our longest path gave a rough estimate of
a maximum clock speed of 37 MHz. Our test equipment allowed us to test the
chip at speeds of up to 31 MHz. On all but the simplest cases, we got
incorrect results at 31 MHz. But all chips appeared to give correct results
at the next lower speed of 17 MHz. It would have been interesting to
determine at exactly what speed the chip would fail, but the test equipment
was not flexible enough.

URL: http://www.owlnet.rice.edu/~toddw/elec422/

SUBMITTED BY:
Eric Johnson    erj@rice.edu
Todd Wille      toddw@rice.edu

REQUEST: END