Loadstar 242/242.d81/t.mos3

Text File | 2004-01-01 | 14.2 KB | 474 lines

u
            MOS Technology

                 from
             On The Edge:
       The Spectacular Rise and
          Fall of Commodore
           by Brian Bagnall

               Part III

Making Chips

    Chuck Peddle, Will Mathis, and Rod
Orgill would collaborate to design the
initial architecture for the new
microprocessor. "It was just the
perfect product, the perfect time, the
perfect team," says Peddle.

    The architects' task was similar
to designing a small city, except the
streets in this city would be paved
with metal. Electrons would inhabit
their city, traveling the streets
until they reached a transistor.
Timing within this little city would
be critical, otherwise traffic would
halt, causing the chip to lock up.

    Peddle and his group intentionally
numbered their chips starting with
6500, so it would sound similar to the
Motorola 6800. "It was a cheaper
version of the 6800 and there was
intended to be a whole string of
them," he explains. "In hindsight,
with many years and lawsuits behind us
now, it was designed to sound like the
6800."

    The first chip in the series was
the 6501, which could drop into a 6800
slot. "It was definitely not a clone,"
says Peddle. "Architecturally it's a
6502. The only difference is it plugs
into Motorola socket."

    Peddle explains the 6501 strategy.
"We were competing in a market where
we were selling to people who might
have bought the 6800," he says.
"Having a plug-in compatible version
was just a marketing game."
Unfortunately, socket compatibility
would later provoke Motorola.

    The centerpiece of their project
was the 6502 microprocessor. "The 6502
was what we were driving for," he
says.

    To create the architecture of the
chip, the three engineers created a
simple diagram to represent the
structure of the chip. "We would start
with a basic block diagram," says
Peddle.

    Some of the most important design
work took place away from MOS
Technology. "We put some of the more
significant stuff in while drinking
booze at Orgill's house one night,"
says Peddle. "The way to do really
creative work is to work on it and
then sometimes you've got to let it
alone. If somebody gets a bright idea
at a party, you take time out and you
go argue about it. We actually came up
with a really nice way of dealing with
the buses that came out of a
discussion at Orgill's."

    Al Charpentier was one of the
calculator chip designers at MOS
Technology. He witnessed Peddle
driving his team to build the new
processor. "Chuck was an interesting
character," he recalls. "He could be a
bit pompous, but he had a vision and
he was pushing that vision. Chuck was
the visionary."

    Peddle created a concept called
pipelining, which handled data in a
conveyor belt fashion. Instead of
stopping while the microprocessor
performed the arithmetic, the chip was
ready to accept the next piece of data
right away, while internally it
continued processing data. This
feature would make the chip faster
than anything produced by Intel or
Motorola at the time. A one-megahertz
6502 was equivalent to a four-
megahertz Intel 8080.

    The semiconductor team not only
developed a microprocessor, they also
developed the supporting chips. The
first was the 6520 PIA chip, which was
a clone of the Motorola 6820 PIA. One
chip, called the 6530, contained 1
kilobyte of ROM, 256 bytes of RAM, a
timer, and two IO ports. This allowed
engineers to assemble a complete
computer using only two chips. The
team also developed 128-byte 6532 RAM
chips.

    One by one, the architects passed
their designs to the layout people.

    The layout team consisted of two
main engineers: Bill Mensch and Rod
Orgill. A third engineer, Harry
Bawcum, aided the layout artists. It
was their task to turn an abstract
block diagram into a large-scale
representation of the surface of the
microprocessor. Orgill was responsible
for the 6501 chip, Mensch the 6502.

    Chuck Peddle originally hired
Mensch at Motorola after Mensch
graduated from the University of
Arizona. "Mensch was literally right
out of school," says Peddle. One of
eight children, Mensch grew up in a
small farming community in
Pennsylvania. According to Mensch, "I
lived on a dairy farm, got up at 4:30,
milked the cows, and went off to
school."(4)

    At Motorola, Peddle was impressed
with Mensch's natural talent. "He was
just spectacular doing N-channel
design and layout. He was the worlds
best layout guy," raves Peddle.

    Mensch was dependable, which made
him a favorite with MOS engineers.
"Bill was a good guy," says
Charpentier. "He was very
knowledgeable and knew what he was
doing."

    Rod Orgill, the youngest member of
the team, worked at Motorola on the
fabrication process of the 6800. Out
of everyone on the team, Orgill had
the most diverse set of abilities.
Peddle relates, "Rod was a combination
of chip designer and architect." For
the first time in his life, Orgill
would acquire layout abilities as an
understudy to Mensch.

    Peddle claims the 6501 was a
marketing game, but Rod Orgill
believed the 6501 would be more
successful than the 6502. According to
Mensch, "We made a bet and said who's
going to have the highest volume and
Rod says, 'There's no question:
following Motorola's marketing, the
6501 will surpass your (6502) design
and yours won't even have a chance.'"

    The small group of young engineers
worked in a small room on the second
floor containing several large art
tables. Here, Mensch and Orgill
brooded over thick sheets of vellum
paper. The layout consisted of
thousands of polygons, each a specific
size and shape. Thin lines called
traces connected the polygons,
creating a complex circuit.
Incredibly, the engineers created the
layout in pencil, one component at a
time. The task was formidable, with a
completed diagram containing
approximately 4,300 transistors.(5)

    Near the end of the design
process, disaster struck. The
engineers realized their architecture
would not fit within the allotted area
of the microchip. "When we sat down to
optimize the system, we discovered we
were 10 mills too wide," says Peddle.
"The design was almost done. Mathis
and I put a big piece of paper down on
a table and sat there and optimized
every line until we got rid of 10
mills."

    The engineers were on a tight
deadline to have the product ready for
the upcoming Wescon show in September.
They obsessively searched for ways to
recycle lines in the schematic, thus
reducing the area. Peddle grimly
recalls, "Mathis and I had to keep
redoing the architecture to make sure
they stayed within that area."

    To print the microchips, the
engineers used a process called Metal
Oxide Semiconductor, or simply MOS.
This process used six layers of
different materials, printed one on
top of the other, to build the tiny
components on the surface of a silicon
wafer. This meant the layout artists
had to create six different diagrams,
one on top of the other.

    The process required incredible
precision because the layers had to
line up exactly. The surface of the
chip was necessarily dense in order to
fit everything into a small area, so
the artists squeezed transistors and
pathways close to each other. If a
single layer deviated by more than a
few microns, it could touch another
pathway and create a short circuit.

    After the layout was completed,
the engineers faced the soul-draining
task of rechecking their design. The
most sophisticated tool in this
process was a small metal ruler, or
more accurately, a scale. Herd
recalls, "They would take their scales
out of their pocket - don't call them
a ruler V and they would measure for
months! They would measure each
transistor and make sure it was two
millimeters by point seven."

    Mensch, Orgill and Bawcum sat
bleary-eyed over their drawings,
sometimes for 12 hours a day,
painstakingly measuring every point on
the layout. They measured the size of
components, the distance between
components, the distance between
traces, and the distance between
traces and components. With a touch of
sympathy in his voice, Herd explains,
"You could be a really talented
designer but if you couldn't check
your design with the mind-numbing
repetitiveness, your stuff didn't work
and you would get a bad reputation."

    Mensch and Orgill kept small cots
in the room so they could work for
long uninterrupted periods followed by
a few hours rest. "With the
semiconductor guys, that tends to be
something you do when you are doing
that at a certain level of design,"
recalls Peddle. "You tend to just keep
going."

    Even today, Peddle is still in awe
of Mensch's ability as a layout
engineer. "Bill has this unique
ability to look at the requirements
for a circuit, and he can see how it
is going to layout in his head," he
says. "He's just totally unique.
Nobody matches Mensch."

    In June 1975, the chip design was
ready. It was up to the process
engineers to imprint the design onto
tiny silicon wafers. Months earlier,
Pavinen promised Peddle he would have
the N-channel process ready. Pavinen
was true to his word. "He gave me
everything I wanted," says Peddle.

    The procedure to shrink a large,
dense design onto something smaller
than a thumbtack is both mysterious
and under-appreciated. In many ways,
it is also the most important step
and, if intelligently planned, it can
reduce the cost of a microchip
dramatically. Engineers simply call
this step "the process."

    When Pavinen and his two partners
founded MOS Technology, it was their
explicit goal to be the best process
company in the business. "MOS
Technology's business premise when
they started was that they knew how to
process better than other people,"
says Peddle. Engineers at the time
documented very little of what they
did, and most process engineers stored
the process in their heads.

    In order to print the transistors
and other components to a silicon
chip, the engineers had to create a
mask. The mask blocks out everything
except for the parts of the chip they
want, much like a stencil blocks spray
paint to produce letters. The mask
relied on the principles of
photography and light.

    To transform the circuit diagram
into a mask, the engineers used a
material borrowed from the graphics
industry called Rubylith. Rubylith is
a sheet of acetate film with a red
base covering the surface. Since the
semiconductor industry was in an early
stage of development, the tools to
transfer the diagram were outrageously
primitive.

    According to Bil Herd, "They were
doing chips by cutting Rubylith with
razor blades. They would kick their
shoes off, push some tables together,
and jump up on them." It was up to
engineers Mike James and Harry Bawcum
to perform the tedious task of cutting
out pieces from the Rubylith to form
the mask.

    According to Bob Yannes, who
arrived at MOS just after the Rubylith
years, "I can't imagine using that
stuff. You're looking at this huge red
plastic thing in front of you and
you're supposed to peel off the parts
that are supposed to stay and leave
the parts that are supposed to go
away. Unless you were very careful,
you got the two confused and you ended
up peeling off the stuff that is
supposed to go away. Then you start
taping it back down again."

    With engineers crawling all over
the huge sheets of acetate film, it
was vital sharp toenails were not
exposed; otherwise they would drag
over the surface and slice into the
acetate. Engineers were not known for
their attention to appearance and it
became vital to keep pairs of fresh
socks available. "Everyone would wear
fresh socks with no holes in the toes
for getting on the table," explains
Herd with some amusement.

    Orgill and Bawcum created six
Rubylith masks for the 6502 chip, one
for each layer. Once completed, the
engineers photographically reduced
each of the large sheets of Rubylith
to create a smaller negative.
Engineers chemically etched a tiny
metal mask using this negative. The
technicians would eventually use this
mask, almost like a rubber stamp, to
create thousands of microprocessors.

    Precise robotic machines used the
tiny metal mask to duplicate the
pattern over the entire surface of the
silicon wafer. In the early seventies,
the metal mask made contact with the
surface of the silicon so the
electrons could flow through the mask,
imprinting the design to the surface.
"People used to have what they call
contact masks, which were pretty
destructive on the mask," recalls
Peddle. "They actually put the mask on
the chip and it got worn out very
quickly." Every time a mask wore out,
the designers had to go through the
laborious process of making a new
mask.

    At MOS Technology, John Pavinen
pioneered a new way to fabricate
microprocessors. "They were one of the
first companies to use non-contact
mask liners," says Peddle. "At that
time everybody was using contact
masks."

    With non-contact masks, the metal
die did not touch the wafer. Once the
engineers worked out all the flaws in
the mask, it would last indefinitely.

    Pavinen and Holt handed off the
completed mask to the MOS technicians,
who began fabricating the first run of
chips. Bil Herd summarizes the
situation. "No chip worked the first
time," he states emphatically. "No
chip. It took seven or nine revs
(revisions), or if someone was real
good they would get it in five or
six."

    Normally, a large number of flaws
originate from the layout design.
After all, there are six layers (and
six masks) that have to align with
each other perfectly. Imagine
designing a town with every
conceivable layer of infrastructure
placed one on top of another. Plumbing
is the lowest layer, followed by the
subway system, underground walkways,
buildings, overhead walkways, and
finally telephone wires. These
different layers have to connect to
each other perfectly; otherwise, the
town will not function. The massive
complexity of such a system makes it
likely that human errors will creep
into the design.

    After fabricating a run of chips
and probing them, the layout engineers
usually have to make changes to their
original design and the process
repeats from the Rubylith down. "Each
run is a couple of hundred thousand
(dollars)," says Herd.

    Implausibly, the engineers
detected no errors in Mensch's layout.
"He built seven different chips
without ever having an error," says
Peddle with disbelief in his voice.
"Almost all done by hand. When I tell
people that, they don't believe me,
but it's true. This guy is a unique
person. He is the best layout guy in
world."

     [Continued in MOS - Part IV]