IEEE Spectrum May, 2017 - 34

And there's one more piece of the CD
puzzle: how to translate the digital bits
into physical marks on the disc itself.
This piece, known as the channel code,
deserves much of the credit for making
CDs skip resistant without sacrificing
playing time. For this encoding scheme
and related optical-recording technology,
its creator, Kees Schouhamer Immink,
will receive the 2017 IEEE Medal of Honor.
Immink was a young electrical engineer working in the
laboratories of Philips, in
Eindhoven, Netherlands,
when the research effort that was to
become compact-disc technology got
under way in the mid-1970s. Initially,
Immink had nothing to do with the project. Rather, he worked on control theory,
assigned to a group of mostly optical engineers developing the analog laser videodisc. Philips introduced the LaserDisc in
1978, and it was a flop. This play-only system couldn't compete with video cas-

one to do measurements of the two systems, the quality, how they coped with
scratches, how they coped with imperfections of the disc. My job with the LaserDisc
was finished, so I said, 'Sure, I could do it.' "
Toward the end of the year, Sony
sent engineer Hiroshi Ogawa over to
Philips with a big box of electronics,
Immink recalls. Ogawa hooked up his
gear, Immink hooked up the Philips
prototype, and the two, using experimental discs made with each company's coding technologies, started their
tests. They messed up the plastic discs
with scratches and dust and, eventually,
crammed more bits onto the discs to see
how long they could extend playing time
before the discs became unreadable.
And then the two engineers packed up
their gear, went to Tokyo, and repeated
the experiments-because, Immink
notes facetiously, "Of course, the rules
of physics are completely different in
Tokyo than in Eindhoven." The whole
process took several months.

The impact of Immink's EFM
code didn't end with the CD.
It went into the MiniDisc...
the DVD...the Super Audio CD
settes, which could be used to record
and had a two-year head start in development. Come 1979, Immink was a
research engineer without a project.
Meanwhile, another research group
at Philips had developed a prototype of
a laser-based digital audio disc system,
while Sony engineers had demonstrated
a similar technology. The two systems differed enough in the details to render them
completely incompatible. But instead of
racing separately to market and triggering a format war, in 1979 the two companies decided to merge their engineering
talent and jointly settle on one design.
That's when Immink got involved. The
team at Philips, he says, "needed some34

MAY 2017

North AMericAN

Although Immink was acting as a test
engineer, he was a researcher at heart, so
he had his eye out for parts of the design
that could be improved. And he found
one: the coding scheme that specifies
how bits of data are translated into physical marks on the disc.
To understand what he came up with,
first visualize the shiny surface of an
optical disc. Just under that polycarbonate surface is a reflective layer of metal,
marked with a pattern of pits and "lands,"
the places where it's not pitted. The laser
hitting that metal reflects from the pits
and lands differently, creating a variation
in the light's intensity as it reflects back
and hits an optical sensor. The system

SPectrUM.ieee.orG

identifies a change in intensity as a 1; no
change over a set distance represents a 0.
There's a problem, though, when too
many pits or lands are bunched close
together, because the servo system-the
electronic and mechanical controls for
the optical head-relies on the trail of pits
and lands to stay on track. (A CD doesn't
have grooves like a vinyl record to serve
as a guide.) Immink explains this by referring to a classic fairy tale that goes by several names, among them "Little Thumb,"
in which children turned out into the
forest drop stones along the way to find
their way home (a scenario also played
out in "Hansel and Gretel").
"If you do not drop a stone often enough,
you lose your way," Immink says of the
path created by the pits and lands on the
disc. The problem is exacerbated by dust
or scratches that hide spots on the disc,
like a fallen leaf hiding a pebble.
Both Philips and Sony had come up
with different rules for translating digital audio data to sequences of pits and
lands, rules that took these considerations into account. But Immink thought
both approaches were investing too
much disc real estate in their efforts to
keep the servo system on track. Instead,
he thought, by designing better control
mechanisms for the servo itself, he could
place a few less "stones" along the path
it needed to follow. He also could reduce
the number of bits that separated each
8-bit block of data bits-so-called merging bits-as long as he selected them carefully in relation to the data bits around
them. These two adjustments would
allow 30 percent more data to fit into
the same physical space without causing the optical head to skip.
The encoding system Immink devised
came to be called Eight-to-Fourteen Modulation (EFM). Using a lookup table, it
translates each 8-bit chunk of data into a
series of 14 bits, with every binary 1 separated by at least two but no more than
ten 0s. This ensures that no sequence
of pits and lands is too long (problematic because the transitions between a
pit and a land generate a signal used for
timing recovery) or too short (problematic because short sequences are difficult to detect and prone to error). Three

http://SPectrUM.ieee.orG

Table of Contents for the Digital Edition of IEEE Spectrum May, 2017