The 2004 Mars Rover mission is one of the great
engineering triumphs of our time. So when the Silicon
Valley DVClub brought in a
NASA speaker for its quarterly lunch meeting, I knew it would be inspiring. And
it was -- even though it illustrated how different NASA's engineering effort was
from contemporary chip design.
DVClub, co-sponsored by Cadence, offers quarterly meetings
for IC verification professionals in Silicon Valley
and seven other locations. Michael Sims, research scientist at NASA Ames Research Center,
spoke on "The Life and Times of the Mars Exploration Rovers" at the Aug.
12 DVClub meeting in Silicon Valley. Sims
is the uplink lead for most of the rovers' cameras and is project lead for long
term planning for the mission.
You probably know that the Spirit and Opportunity
rovers, originally designed for 90-day missions, are still trudging along the
sands of Mars after six and a half years. And yet, these rovers were designed
manually, using techniques that Sims said were from the "pre-1980's era" in
comparison to automated VLSI design. They were mostly built from proven,
off-the shelf components, including an already-obsolete PowerPC processor. Verification
consisted of building a physical prototype and watching it make its way over
rocks and sand.
Not on the Leading
I already knew that NASA takes an extremely conservative
approach to engineering, preferring tried-and-true technologies over anything
close to the leading edge. In 2004 I wrote
an article about a full-custom CCD imaging chip designed for the rovers. The
chip designers faced some unique challenges, but the chip itself wasn't particularly
new technology. The rovers use 1 megapixel black and white cameras augmented by
The primary criterion for getting a project accepted at
NASA, Sims said, is that "you have to argue it's a safe, viable, doable
mission. That means we've done this before." So the rover team argued that
their project would be just like the 1997 Mars Pathfinder rover, only without a
lander, and would use the same PowerPC processor.
It turned out that the processor is much too slow to detect
obstacles in the path of the rovers, so "we close our eyes and walk," Sims
said. When the rovers get stuck, they spin wheels and slowly dig their way out.
Spirit, however, is currently stuck in the sand for the duration of the Martian
The rovers are programmed daily in what is "effectively"
assembly language, Sims said. How is the code verified? By getting a roomful of
people to look through it. "Reliability comes from throwing more people at it,"
he said. "Again, this is a good analogy to VLSI design in the 1970s."
Cause for Inspiration
And yet, and yet...those Mars rovers have done some stunning
science and have lasted far longer than anyone expected. This project could
have succeeded only with fantastic teamwork and dedication on the part of the
engineers and programmers over a period of many
years. There were many creative, ingenuous solutions to unique problems, such as 200-degree temperature variations between night and day.
And I think that's the takeaway for the attendees at DVClub.
You can't design a competitive consumer product the way the Mars rovers were
designed. But if you apply the same spirit of teamwork and creativity, and combine
that with up-to-date design and verification methodologies and tools, you can go a
long, long ways.
Nice post Richard! When Michael Sims talked last week about teams hand walking through code written to drive the rovers' next excursion, I had flashbacks of teams spreading chip plots on the floor of a cafeteria or other big room, looking for missing polygons and design-rule errors on giant sheets of plotter paper taped together. We've come a long, long way in EDA since then!