Instructions: To add a question/comment to a specific line, equation, table or graph simply click on it.
Click on the annotations on the left side of the paper to read and reply to the questions and comments.
Sputnik was the first artificial satellite to orbit the earth. It w...
![Mars Probe]( *Picture of the Mar...
A sextant is an instrument used to determine the angle between an o...
Battin is talking about a round trip time of three years, during wh...
J. Halcombe Laning was a computer pioneer that was heavily involved...
You can take a look at the report [here](
President Dwight D. Eisenhower signed the National Aeronautics and ...
Core rope memory had a bit density of 72KB per cubic foot. In contr...
Gravitational slingshot or gravity assist maneuver is a movement wh...
The idea is that you are trying to follow a certain route and you n...
William Proxmire was a United States senator from Wisconsin notorio...
![Apollo Guidance Computer](
Vol. 25, No. 1, January
February 2002
Some Funny Things Happened on the Way to the Moon
Richard H. Battin
Massachusetts In stitute of Technology
Cambridge, Massachusetts 02139
In the Beginning
X installed in my
MAC II, I begin the von K´arm´an Lecture for the AIAA.
The date is December 21, 1988, exactly 20 years to the day from
the launch of Apollo 8. Eleven years earlier, the world had b een
enthralled by the Russian Sputnik and the Space Age had begun.
I have at my ngertips several orders of magnitude more comput-
ing powe r than the Apollo Guidance Computer which was carried
onboard the Apollo spacecraft. And this marvel of modern technol-
ogy sits on my desk at home!
This is the year for celebrating Apollo 11—the 20
of the rst lunar landing. But the earlier ight of Apollo 8 was also
a dramatic milestone. It was the rst manned space ight beyond
an earth orbit. The astronauts, Frank Borman, Jim Lovell, and Bill
Anders, were the rst human beings ever to see the entire earth as a
ball. Said Norman Cousins
“On the rst i ght to the mo on we really discovered the
Indeed, who can ever forget that picture of the earth rising above
the lunar landscape?
To many of us who were part of the Apollo program, it was
the most thrilling ight of all. We demonstrated the feasibility of
onboard, self-contained space navigation for the very rst time.
It all began on October4, 1957 when the Russianslaunchedthat rst
satellite. I had started life at MIT— rst as a student, then working
with Hal Laning at the MIT InstrumentationLab. But on that fateful
day I had been away from MIT for a yea r working in industry.When
I learned t hat Hal had a simulation of the solar system running on
the IBM 650 and was “ ying round trips to Mars, I could hardly
wait to rejoin him.
A report by Hal Laning, Elmer Frey, and Milt Trageser on the
feasibility of a photographicreconnaissance ight to Mars had just
been published at the laboratory. They were dead serious that such
a ig ht was possible within the next ve to seven years.
Richard H. Battin received an S.B. degree in electrical engineering in 1945 and a Ph.D. in applied mathematics in
1951—both from the Massa chusetts Institute of Technology. He received an Honorary Doctor of Science Degree in
1999 from Texas A& M University. Currently, he is a Senior Lecturer in the MIT Aeronautics and Astronautics De-
partment. He retired in 1987 from The Charles Stark Draper Laboratory,Inc. In 1972, he and David G. Hoag were
presented by the AIAA with the Louis W. Hill Space Transportation Award
now called the Goddard Astronautics
“for leadership in the hardware and software design of the Apo llo spacecraft primary control, guidance,
and navigation system which rst demonstrated the feasibility of onboard space navigation during the historic
ight of Apollo 8.” He received the AIAA Mechanics and Control of Flight Award for 1978, the Institute of Naviga-
tion Superior Achievement Award for 1980, the AIAA Pendray Aerospace Literature Award for 1987, and the von
K´arm´an Lectureship in Astronautics for 1989. He was presented by the American Astronautical Society with the
1996 Dirk Brouwer Award and the inaugural2 000 Tycho Brahe Award by the Institute of Navigation.For his latest
book An Introduction to the Mathematics and Methods of Astrodynamics, Revised Edition, pu blished in 1999 by the
AIAA in their Education Series, he will receive the AIAA Summer eld Book Award for 2002. He is an Honorary
Fellow of the AIAA and a Fellow of the American Astronautical Society. He is a member of the NationalAcademy of
Engineering and the International Academy of Astronautics. “In recognition of outstandingteaching” the students
of the MIT Department of Aeronautics and Astronautics honored him in 1981 with their rst Teaching Award.
Presented as the von K´arm´an Lecture at the AIAA 27th Aerospace Sciences Meeting, Reno, Nevada; Jan. 9
12, 1989, Paper 89-0861; received Jan. 9 , 1989;
revision received March 8, 1989. Copyright
2001 by Richard H. Battin. Published by the American Institute of Aeronautics and Astronautics, Inc., wi th
permission. Copies of this paper may be made for personal or internal use, on condition that the copier pay the $10.00 per-copy fee to the Copyright Clearance
Center, Inc., 222 Rosewood Drive, Danvers, MA 01923; include the code 0731-5090/02 $10.00 in correspondence with the CCC.
The Draper Lab and the Instrumentation Lab are the same. Only th e name was changed in 1970 as a tting tribute to its founder.
In those days, the contract between MIT and the Ballistic Missile
Division of the US Air Force contained a clause which allowed
our laboratory, within certain limits of course, to work on whatever
struck its fancy—a sort of government IR&D program.
The few of us lucky enough to be involve d were very excited
about the Mars probe. We studied the problem intensely for a year
or so and produced a three volume report together with a full-scale
wooden mod el of the spacecraft.
Today, that model is displayed in the lobby of The Charles Stark
Draper Laboratory.
Navigation data f or the Mars probe was to be gathered by an
onboard sextant and processed by a spacecraft digital computer.
Observation data would be used to determine vehicle position and
a correction to the onboard clock. Periodically, changes in velocity
would be made by a small propulsion system a s directed by the
computer. Spacecraft attitude was to be maintained by momentum
wheels also undercomputer control. Power would be obtained from
the sun using solar panels which would unfold like venetian blinds
after launch.
To ensure reasonablelaunch velocities, the round-trip ight time
was to be three years. Computer activity would be minimal during
the long coasting periods between velocity c orrections. Most o f the
time the computer would be asleep” to conserve power. Indeed, the
principal requirement for the computer was that it have a long shelf
Only on e picture of the Martian surface was to b e taken to elim-
inate the mechanical pitfalls of a lm transport system. The space-
craft would make one pass by the planet a few thousand miles above
the surface, orient itself for the photograph, open the shutter to
expose the lm plate, and coast back to earth.
The capsule ho using the lm had a shape similar to the Apollo
Command Module and, in like manner, would dive through the at-
mosphereto splash down in the Gulf of Mexico. A radio transmitter,
a yellow dye, a nd a ashing light would aid in its recovery. A repel-
lent would hopefully discourageany shark who might think of it as
All in all, it was a rather sophisticated project considering the
state-of-the-art. Milt Trageser wa s and is to be congratulated for
his engineering prowess and ingenuity. The program was well con-
ceived and carefully planned. But, as has happened to so many other
“best laid plans, it too went astray.
In the Doldrums
The Mars probe preliminary design was complete in the summer
of 1959. The Air Force had been its sponsor, but a new government
agencythe “National Aeronautics and Space Administration”
would control its destiny.
With view graphs, reports, and the wooden spacecraft model, we
arrived in Washington on the same day as Nikita Khrushchev. Our
presentation was well received. But the high-level NASA audience
we had anticipated, including Hugh Dryden, was busy entertaining
the Russians.
NASA did not immediately write us a blank check for the Mars
probe, but we were promised some future study money. Our small
team survived but much of the original enthusiasmdid not. Now we
were simply doing “interplanetary navigation system studies. We
had no reason to anticipate what lay ahead.
To support the Marsproject,we developedappropriatetrajectories
with ight times of roughly three years, and launch dates in 1962
1963 time frame. In th is case, the spacecraftmakes two orbits of the
sun while the earth does three. Later we found round-trip missions
to Venus having ight times of only a y ear and a quarter.
I also discovered on January 26, 1961 the rst multiple yby
orbit—earth to Venus to Mars to earthwhich is traversed, at least
theoretically,without additional propulsion. This game of celestial
billiards is played by proper control o f the orientation of the orb ital
plane an d altitude during the swingby of each planet.
Today, the Voyager spacecraft on its Grand Tour of the solar sys-
tem is a spectacular demonstration of such missions. Soon, hope-
fully, the Galileo spacecraftwill be making dozens of similar close
encounter ybys of the Jovian moons.
The tools we needed for this work were not easily achieved. One
did not study Celestial Mechanics unless he planned to be an as-
tronomer and astronomers were not designing orbits for missions
to Mars. On the contrary, in 1956 the British Astronomer Royal
declared “Space travel is utter bilge!”
Some astronomers, whom we did consult, had reservationsabout
the success of the project. “How do you expect to se nd a spacecraft
to Mars when you don’t know exactly where it is?”
In those days
the uncertainty was several thousand miles.
I suppose it was not
easy to abandon their familiar earth-based reference coordinates.
Navigating the Mars probe consistedof measuringangles between
planets and stars. We linearized those measurements about a refer-
ence point and used Gaussian weighted least squares to obtain the
celestial x.
The terms “estimator” and “state vector” were not in vogue so
we couldn’t yet say that we had designed a n “estimator” for a
four-dimensional “state vector”three for position and one for the
onboard clock correction.
Later, for the Apollo system, the state spacehad nine-dimensions.
In addition to position and velocity, we would also be estimating
the rendezvousradar antenna angle biases on the Lunar Module and
estimatinglunar landmarklocationsas observedfrom the Command
Gaussian least squares requires batch processing.All data is col-
lected beforethe computationbegins.In a small igh t computerwith
data gathered over long periods of time, the method is cumbersome
and impractical.
The situation was remedied by a recursive form of the estimator
which allowed measurements to be in corporated as they are made.
It was not important for the Mars probe, but it was essential for
Apollo navigation.
This recursiveestimationprocessis nowknown as a Kalman lter.
Today, it is widely used for all sorts of purposes. Every student of
control systems studies the subject in school.
But it was used for the rst time to navigate the Apollo 8 Com-
mand Module in cis-lunar space on its way to the moon!
With NASA funding, Hal Laning and Ramon Alonzo began, in
earnest, the design of the Mars probe computer with its unique
characteristics for space applications:
variable speed to save power
relatively few transistors
parallel word transfer
automatic counter incrementing
automatic interrupt
The program and constants were wired in a so-called “core
rope”a memory with unusually high bit densities which could
not be altered electronically.
This was the computer concept and architecture that would one
day take man to the moon.
In the Race
Our NASA contract ended and there was a nine month hiatus
before another six-month contract began in early 1961this time
for a preliminary design study of a guidance and navigation system
sponsored by the NASA Space Task Group.
Later that year on May 25, 1961, President John F. Kennedy in
his Special Message t o Congress on Urgent National Needs said:
“I believe that this nation should commit itself to achieve
the goal, before this d ecade is out, of landing a man on
the Moon and returning him safely to Earth.
Jim Webb, the NASA Administrator, knew Doc Draper and asked
him to develop the Apollo guidance and navigation system. Of
course Doc ag reed.
“When will it be ready?” asked Webb.
“When you need it, said Draper.
“How do I know it will work?” Webb p ersisted.
“Ill go along and operate it fo r you.
And he most certainly would have done so, had they only let him.
It all became of cial on August 10, 1961—exactly eleven weeks
after Kennedy’s speech.
The rst major Apollo contract awarded by the space
agency was to the MIT Instrumentation Laboratory!
It sounds incredulous but that’s the way it really happened.
Everyone who knew Doc Draper believes it and it is the story just
the way Draper a lways told it. Doc is no lon ger with us but his spirit
and inspiration live on.†
Today, ma ny people wonder “Was thereever reallya Space Race?”
It was certainly real in the beginning.There was a strong concern
that the Russians would inte rfere with an Apollo ight by jamming
the telemetry signals.
Our lab was noted for developingautonomous systems in missile
guidance—the self-contained backup system for the Atlas inter-
continental ballistic missile, the Thor IRBM, and the Polaris eet
ballistic missile guidance system. Of course, needless to say, the
Mars probe would have been self contained.
The challenge of providing an autonomous guidance and navi-
gation system for Apollo was right up our alley. So when Charlie
Frick, our NASA boss from the Apollo Spacecraft Program Of ce,
“Therewill be absolutelyno groundcommunicationswith
the Apollo spacecraft! Don’t even think about it!”
†On September 28, 198 8 the Charles Stark Draper Prize was created by
the National Academy of Engineering and funded by the Draper Lab Board
of Directors. It is a major new international award to recognize achievement
in engineering and technology.
The Prize is similar to the Nobel Prize. It will recognize extraordinary
engineering accomplishment in the service of human welfare and freedom,
and will emphasize those aspects of engineering that are essential to a better
Doc would certainly have been pleased.
it was dif cult to suppressour smiles. We felt j ust like Br’er Rabbit:
“Please don’t throw us in that briar patch, Br’er Frick!”
Security was also very real. Almost everything was classi ed—
even schedules.Fortunately,t hat particularphobiasoon peteredout.
At the other extreme,there were manywho advocatedcooperation
with the Russians.Dur inga paneldiscussionat an AIAA conference,
Wernher von Braun addressed the question: “Why don’t we work
together on the Apollo program?” His response went right to the
heart of the matter:
“If there were cooperation with the Russians on space-
ight, there wouldn’t be a program in either country.
To start o ur part of the race we had rst to assemble a team. But
ndingthe rightpeople was frustrating.We had themost challenging
guidance system imaginable to develop. Recruits should have been
pounding at the door. But that didn’t happen.
In the long ru n, though, the best p eople were alreadyat the Lab—
working in other divisions. Appropriate transfers were made. Nev-
ertheless, for a very long time to come, the software task was not
adequately staffed.
We were beginning to learn that “people problems were often
much more dif cult to solve than the technical ones.
Although we couldnt seem to get engineers,we did get advice—
both technical and theological:
“Apollo should be launched on a clear night when the
moon is full to provide the best possible target.
“If the Lord had intended man to go to the moon, He
would never have created Senator Proxmire.
In the Software Jungle
The MIT Instrumentation Lab was designing software for the
Apollo onboard guidance system even before the word “software
was invented.
I still remember the rst time I told my wife that I was in charge
of Apollo Software. She exh orted me: “ Please don’t tell any of
our friends!
I suppose real men do “Hardware just as real men dont eat
It was an attitude that prevailed a long time in many organiza-
tions. Salaries fo r computer programmers did not keep up with the
salaries of engineers.Engineers did engineering.The programming
or coding
was more menial work and should be left to others.
I wanted no such distinction. Our best engineers should design
program the software for the ight computers. The reasoning
was simple. A good engineer can learn to write programs. But a
computer software specialist would nd it far more dif cult to do
the engineering without considerably more training.
The basic architecture of the Apollo onboard computers was the
design of Hal Laning—certainly one of our most creative engineers,
and the early application pr ograms were written by our best system
engineers. We tried hard to keep our standards high.
Even so, with the best talent,computer programs seldom perform
as intendedthe rst time. If they work at all, they may do unexpected
or “funny things. We called them “FLT’s”—short for
Funny Little
This was before “Murphy discovered his famous law. He must
have been secretly watching our software development efforts.
The Apollo Guidance Computer, or AGC as it was called, evolved
from the design for the Mars probe. In 1961, when our Apollo
contract was signed, it had 4096 words of xed memory and 256
words of erasable.
A word was 16 bits
with one bit for sign and
one for parity check. Hence, double precision was required for most
calculations. The cycle time was modest, approximately 24¹sec,
which was more than suf cient for t he Mars application.
The xed memory was called a “core rope memory since the
early models clearly resembled lengths of rope. The high density
of storage was achieved by storing a large number of bits in
magneticcore. A stored bit is a “one” whenever a sensewire threads
a core and a “zero” when it fails to th read a core.
In the operationof the rope memory, a core is s witched which in-
duces a voltage drop in every senseline which threadsthat core. Six-
teen sense wires were connected to sense ampli ers to detect which
had voltage drops. With the addition of an appropriate switching
network, each core could then hold several words.
All this information was permanently wired at Raytheon by
“LOLs”—literally, “little old ladies who slowly and painstakingly
threaded the cores by hand. Late r an ingenious adaptation o f a tex-
tile loom was used for this purpose which was far faster and cer-
tainly more reliable. The loom was driven by a punched paper tape
created by the same program that produced the mission software.
Once a memory module was manufactured, not a single bit could
be changed, either intentionally or unintentionally.
There wa s a dichotomy of opinion regarding the xed memory
concept. Once the memory had been programmed, it took about six
weeks to manufacture. Add to that the time required for testing and
you discover the rule:
No ight-computer memory changes
within two months of a launch!!
It was far too late to change the design so the argumentwas purely
philosophical:“Was it bad or was it good?
Each side made its own case:
There must be a way to make last minute changes.
But, last minute changes may be ill-considered and/or hur-
riedly tested. They are dangerous.
In the long run, despite objections, it was good discipline and kept
everyone honest. If you know you can’t ma ke last minute changes,
then, by golly, youll be that much more careful in your design and
Inevitably,of course,someonewould discoverthat erasablemem-
ory could hold more than just data. You could actually write and
execute programs from the erasable memory.
By that time, though, strict NASA approval was required to
do so .
The physical sizeof the computerwasonecubicfoot.That could be
changed only at great cost since the spacecraft manufacturer,North
American Aviation,had allowed that much room and no more in the
Command Module. If we needed m ore capability, it would have to
come fro m advances in computer technology.
Eldon Hall, wh o had designed the Polaris missile computer, was
in charge of the Apollo Guidance Computer development.I can still
remember the day he asked if I could use twice a s much memory.
He ha d gure d out how to stuff twice as many sense wires through
a core.
I was ecstatic. Our prayers had been answered.
Charlie Frick, however, was not so pleased:
“You told us that 4000 words was enough. And now you
want to double it!!
He was, obviously,a man of experiencein dealingwith unscrupulous
contractors.Doubling th e size would surely mea n doublingthe cost.
The memory size was, indeed, doubled. And it doubled twice
again after that before it went to the moon. The nal count was
36,864 si xteen-bit words for the xed memory and 2048 for the
erasable. The cycle time had also been cut in half to 12¹sec. Bu t
the physical size never did change.
We were able to cope with limited memory, limited instruction
repertoire, and short word length for mission programs by using a
powerful interpretive language.
Charlie Muntz created the “Interpreter to manipulate the 28 bit
data words. Memory was conserved
dramatically, to be sure
now the time for numerical computationswas measured in millisec-
Double-precisionadd—0.66 millisec
Double-precisionmultiply—1.1 millisec
Double-precisionsquare root—1.9 millisec
Double-precisionsine—5.6 millisec
Although the data words now had suf cient length, they were
expressed with a xed decimal point. We co uldn’t afford the luxury
of a oating-point arithmetic. Fixed point is much faster and many
of the mission programs had to function in real time.
The Master Control of this software maze was the Executive and
Waitlist Program skillfully designed and written by Hal Laning. It
was a sophisticatedpiece of software that:
permitted time-sharing of erasable memory
allowed orderly interruption of programs by those of higher
accommodated as many as seven programs in suspended an-
imation with suf cient information saved to enable each to
resume at a later time as though nothing had happened.
It was a supremetriumphof HalLaning’s ingenuityandperspicacity.
Remember that this was 1961 and control computers were barely
in their infancy.
I can still remember when Hal rst tried to explain to me just how
all t his was supposed to work. It was mind boggling. Hal, I said,
“This is much too complicated. There has to be a simpler way. He
responded with profound authority:
“There is no other way to do this job!”
And, of course, he was right.
But the same mind which could conceivethis logical masterpiece
preferred not to cope with endless meetings, in large conference
rooms, with too many people, and too much contention. This was
to be our lot for years to come and Hal avoided it like the plague.
Fortunately for us, he was always there for advice and counsel.
Mission programming began in earnest as soon as the software
tools were in place. We had no rm requirements from NASA but
we knew, generally, some of the things that had to be done.
We would hav e to navigate with onboard sensors; to make course
changes outside t he atmosphere; to reenter the atmosphere at the
proper angle; and we would have to guide the Command Module
safely to its splash-down site in the ocean.
These activities started before the decision favoring Lunar Orbit
Rendezvous and the invention of the Lunar Module. In fact, the
Grumman Aircraft Engineering Corporation wasnt even identi ed
as the LM contractor until the end of 1962.
In June 1964, Robert C.
Duncan, Chief of Guidance and
Control in Houston, directed that spacecraft autopilot functions be
performed digitally in our computer. It was rather late for such a big
change. By that time, we had been on the job for three years and a
lot of code had been written.
To accommodate these new and time-critical functions, he au-
thorized a design change in the computer—we could double the
speed and increase the repertoire of codes.
At that time the com-
puter had but eight basic instructions and “divide” was not one of
But the software people were not to worry. Th eir carefully-
crafted programs would s till work on the new computer. “Upward
Compatibility was to be the name of the game.
But the new computer could be programmed much more ef -
ciently with the new codesand our precious computer memory must
never be squandered.Everythingwas redonefor the Block II system.
Unfortunately,the Block I system didnt go away. It would still be
used for the rst orbital ights. Now we had two different s oftware
systems to maintain.
That wasn’t the worst either. Soon we would have both the LM
computeras well as the CM computerfor which to provide software.
The two were the same but they had radically different tasks to
The Apollo software job was escalating rapidly. The question
was: “Were we really up to it?”
In the Trenches
The rs t manned Mercury ight was May 5, 1961 and the last
on May 15, 1963. The Gemini program, which was to proof-test
the concept of orbital rendezvous, had its rst ight on March 23,
1965—the last took place November 11, 1966.
During this period, NASA was totally focused on the success of
those missions. The next ight always has everyone’s undivided
At o ne time during those Gemini ights, a software change was
required in their reentry guidance program. IBM, the software con-
tractor, told NASA the cost of the change would be on e million
George Mueller was appalled. As the new Associate Administra-
tor for Manned Space Flight he wondered “If it was that costly to
change Gemini software, how about Apo llo? Who was d oing th e
Apollo software anyhow?”
It was our rst indication that top NASA management ha d ever
thought about the MIT software effort. Until then, we were virtually
unimpeded by outside in uences. It was pure heaven! So many
things were decided by one or two engineers which, today, would
take many trade-off studies and large committees.
When the spotlight nally fell on us, the conceptualpart of the job
and many of the mission related programs were essentially nished.
Later, I reminded George Muellerof the incidentand asked How
much did it really cost to change the Gemini reentry program?” His
answer: “One million dollars.
Soon we were made keenly aware of a basic fact of life: There is a
vast differencebetween gettingthe Apollo softwarejob and keeping
We frequently heard the opinion that MIT should not do the
Apollo software. Production software is not the kind of job fo r a
school. They dont have the resources,the right people, or the moti-
I guess nonpro ts c ant be rewarded or punishedat least
not in the traditional ways.
The NASA creed was evident everywhere—on desktops, on
posters, and bumper stickers:
Better is the Enemy of Good
But, “MIT doesn’t know when to quit designing, they said. “They
have a bunch of prima donnas who want to make everything perfect
regardless of how long it takes.
We seemed to be in a kind of trench warfareconstantlydefend-
ing our job against all assailants.
An amusing incident took place at a meeting between our MIT
software people and an industry group who was, obviously,probing
to nd our weak spots. I presided at the conference table and was
absentmindedly toying with a pair of scissors when an annoying y
ew past.
Suddenly, with one slash of the scissors, I cut that y in half in
midair. The visitors swallowed hard and the tone of the meeting
immediately changed. Once again, the day was saved.
George Mueller took a personal interest in the Apollo Guidance
Computer and it was no won der that he did. He had recently expe-
rienced the dire consequences of a p rogramming error.
A missing “hyphen” in a ight c omputer program caused the loss
of the Mariner I at t he start of its mission to Venus on July 21, 1962.
Although I never saw it, I understand that he had the symbol framed
and prominently displayed on his of ce wall lest anyone forget the
importance of ight software.
Early in his tenure at NASA, he formed the Apollo Software Task
Force with representatives from many of the Apollo contractors.
Fortunately, I was a member. When George saw computer memory
disappearing at a rapid clip, he solicited informal proposals from
Task Force members to program the Apollo computers using just
half the memory.
Inevitably, the “half-memory computer” would be privately la-
beled the “half-assed computer. At least, we hadn’t lost our senses
of humor.
But now we were really nervous. The door was being opened
to compete the Apollo software job and Laboratory policy did not
permit us to compete wit h industry.
We were con dent that the Apollo software requirements co uld
not be met with only half the memory. But we were also certain that
there were those who would say that it could be done. Bellcom and
IBM were among those interested in b idding.
The time soon came for proposal presentations from the various
bidders. When it was Bellcom’s turn, Gordon Heffron, the Bellcom
representative,said that MIT was doing a n outstandingjob and they
would not submit a proposal.
It all ended right there and then. Our job was secure—or so it
The biggest complaint NASA had about our software effort was
quite straightforward:
“MIT doesn’t have enough people!”
Some taskswere obviouslyoneman jobs. We used thoseexamples
in feeble attempts to fend off the criticisms. We loved the analogy:
“How many people do you have to squeeze into a booth
to make a phone ca ll?”
The Michelangelo analogy was another favorite:
“When he was painting the Sistine Chapel ceiling, would
Michelangelo have accepted help from a gang of house
painters ju st to nish on schedule?
Our real problem was paradoxical:
1. Goo d people are hard to nd.
2. If you nd them, they must be trained.
3. The people who must train them are too busy.
4. Why are they so bu sy?
5. Bec ause we don’t have enough people.
For the rs t Command Mo dule ight, Apollo 3, Alex Kosmala
was the of cial Rope Mother. With a small dedicated group, he
spent 15 months preparing the ight program, called CORONA.
The original estimate was 6 months. But until then, no one had
any idea just how dif cult and time-consuming the job would be.
Although we were 9 months late, NASA was even later in imple-
menting the Real Time Control Center. CORONA was released in
January of 1966 but didn’t y until August.
We were saved this time, but the specter loomed that one day the
Saturn V would be perched on its launch pa d carrying the Apollo
spacecraft and waiting for those MIT guys to deliver the ight pro-
We could well become the “long pole in the tent”—a colorful
label no one wanted pinned to
The people problem was so lved when the System Development
Corporation became our subcontractor. They supplied organized
teams with competentleadership.We didn’t have to worryabouthow
to employ 30 or 40 individuals, but c ould just deal with a few team
leaders. Other companies,Raytheon, AC Electronics,ARCON, and
CDC also supplied talent but not in s uch large numbers.
Flight Operationsin Ho uston under Bill Tindalltook charge of the
MIT software contract. His r st action was to deal with computer
memory problems. It was Friday the 13
of May, 1966—“Black
Friday” for many of our favorite programs.
Since the Russians probably wouldnt try to mess us up, the ight
would be controlledfrom the ground. We could just as well strip out
all thoselovelyalgorithmswhich madethe spacecraftself-suf cient.
But wait! Most of those programswere neededfor an other reason.
We could not, for instance, dispense with Navigation or “Return
to Earth or “Powered Flight Guidance. After all, we still could
lose contact with the ground for non-sinister reasons.
After that, when new and absolutely essential requirements sur-
faced, they could only be added when something else of lesser
importance was removed.
Things had gotten that serious.
Then came that terrible day of the re and the loss of the crew:
Gus Grissom, Ed White,and Roger Cha ffee. It was January 2 7, 1967
during ground testing of the Apollo spacecraft in Florida.
The launch schedule was now a shambles. The spacecraft was
completely redesignedunder the superb leadershipof Aaron Cohen.
Three unmanned guided ights, which had not been planned before
the re, were own and supported by our software teams before the
rst manned ight, Apollo 7, on October 11, 1968. Then we really
went into high gear!
I still can’t believe how fast events happened after that 20 month
Apollo 8 on December 21, 1968—To the moon
Apollo 9 on March 3, 1969—Flight-test the LM
Apollo 10 on May 18, 196 9The dress rehearsal
Apollo 11 on July 16, 1969—The lunar landing
NASA was shooting them off like Roman candles!
The rest is history. We never held up a ight; we never had a system
failure; and I am proud to say:
“The MIT Instrumentation Laboratory did all of the on-
board software from the rst Apollo guided ight on
August 25, 1966 through the three SkyLab missions in
1973 to the Apollo-Soyuz rendezvous mission with the
Russians on July 15, 1975.
In Route
Certainly the longest and most thrilling 5 minutes of my life
was the 5 minute burn of the S
IVB engine to boost the speed of
the Apollo 8 spacecraft to the 24,200 mph necessary to escape the
“You are on y our way, said Chris Kraft, from the MissionControl
room, “you are
on your way.
A few weeks before the launch, the Navigator Command Module
Pilot, Jim Lovell, spent a few hours practicing on the earth-horizon
sextant simulator at MIT. He consistently identi ed the “horizon”
about 20 miles above the real horizon.
Great! Jim Lovell could be calibrated and his bias number loaded
in the ight computer.
He was recalibrated in real time on the way to the moon. His rst
eleven sextant angle measurements, made early in the ight, were
compared with what they should have been accordingto the RTCC.
After the horizon calibration,came the rst midcourse correction
of almost 25 ft/sec. It was a fairly large one due to earliermaneuvers
to get the spacecraft safely away from the third stage of the launch
Following th e course change, the onboard computer’s version of
the state vector was made to agree with the value obtained from
ground tracking.
From that time on, Jim Lovell made dozensof star-elevationmea-
surements using both the earth and lunar horizons. These were pro-
cessed by the Apollo Guidance Computer using the recursive es -
timation algorithm. The nal set of 15 sightings was made about
35,000 miles from the moon.
The onboardandgroundtrackingestimateswere almost identical!
Chris Kraft even suggested th at we use the onboard state vector fo r
lunar orbit insertion. But the ight plan dictated that a state vector
update from the ground was required before any maneuver. And
there was no overriding argument to deviate from the plan.
Nevertheless, the evidence was conclusive.
The astronauts could have done it on their own without
any ground assistance whatsoever
Early in the morning of December24,Apollo 8 disappearedbehind
the Moon. For 34 minutes there was no way of knowing what had
happened. During that time a 247-second burn took place under the
controlof the MIT gu idancesystem and Apollo 8 was in lu nar orbit.
The astronauts announced their orbital parameters provided by the
AGC when voice contact was resumed. The Mission Control folk s
were obviously disconcerted:
“How do they know? We haven’t had time to track them
They were not yet experiencedwith self-containedinertial systems.
At 8:40 pm Christmas Eve, 1968, the Apollo 8 astronauts were
on television broadcasting from lunar orbit. “For all the people
on Earth, said Bill Anders, “ the crew of Apollo 8 has a message
we would like to send you. He paused a moment and then began
“In the beginning God created the Heaven and the Earth
: : :
and God saw that it was good.
The commander Fr ank Borman added:
“And from the crew of Apollo 8, we close with good night,
good luck, a Merry Christmas, and God bless all of you
all of you on the good Earth.
Days later, in the Washington Post, there appeared an editorial:
“At some point in the history of the world someone may
have read the rst ten verses of the
Book of Genesis
under conditions that gave th em greater meaning than
they had on Christmas Eve. But it seems unlikely
: : :
Christmas wil l always be remembered as the lunar one.
In Retrospect
The euphoriafollowing the successof Apollo 8 was quickly over-
shadowed by the departure of mo st of our top software talent. One
of the many “spinoffs of the Instrumentation Lab was happening
again. Like so many others before them, they were starting their
own company. It was a hard blow in deed.
But we didsurvive.By that time we had verystrongteam members
and they were ready for the promotions to follow.
Apollo 11 was a magni cent triumph! But the actual landing was
more exciting than we had counted on. Just before touch down the
Apollo Guidance Computer a lmost caused a near panic by display-
ing alarms. The NASA ight controllers remained cool and did not
order an abort.
During the short stay of Neil Armstrong and Buzz Aldrin on
the lunar surface, everyone who could possibly contribute to an
understandingof the problem was hard at wo rk. There was concern
about the upcoming liftoff and rendezvous in just a few hours.
The culpritwas an erroneousmode setting—the rendezvousradar
was transmitting pulses to the computer at maximum rate. During
landing, the computer has plenty of work to do. The additional task
of counting all those extraneous pulses was just too much.
There are still a lot of people, who should know be tter, who
describe that event as a computer malfunction. Nothing could be
farther from the truth. It was, in fact, a triumph of software design.
When the computer was operating near capacity and sensed an
overload condition, it was programmed to stop everything, clear
the table, and restart only the top priority jobs. The display of the
alarm was to tell the astronautsand ground controllersjust what was
The software designers had the foresight to anticipate such a
possibility and the skill to cope with it should it ever happen. No
one ever thought it would happen—except Murphy, of course.
Neil Armstrong remembered the Apollo computer when he was
later assigned to NASA headquarters to head the research and
development programs. One such program involved an F8 airplane
to be used for a y-by-wire experiment. NASA needed a reliable
computer to y that airplane. A computer failure would mean the
loss of the plane.
Neil suggested th e Apollo Guidance Computer:
“It’s the most reliable computer I know. It got me safely
to the moon and back.
And that is exactly the co mputer that was used. Phil Felleman, an
outstanding leader and systems engineer at the Draper Lab, was
given the job to do. And he did it well.
The Apollo Guidance Computer never failed in any ight. There
was concern when lightning struck Apollo 12 during launch. It
wiped out the erasable memory load and initiated a computer
restart. But the memory was reloaded in earth orbit and all went
The quality-control people couldn’t calculate the “mean-time-
between-failuressince it would have required dividing by zero.
Generally, en gineers stare in disbelief when told that the Apollo
computer had absolutely no built-in redundancy.There were many
possibilities of single-point failures which wou ld have disabled the
computer. But none of them happened.
Even the computer itself was not redundant. In the Space Shuttle
today there are ve guidance computers but there was only one
each in the Command Module and the Lunar Module.
There was
a back-up computer of limited capability in the LM for emergency
use, designed and built by TRW, but it was never needed.
It wasnt a matter of luck either. Reliability considerations were
uppermost in the design right from the start.
For example, only one type of integrated logic circuit, a three-
inputNOR gate, was used. This meant thatEldon Hall andRaytheon,
the computer manufacturer,could concentrateon quality control of
a single circuit rather than several. A wider variety of circuits would
have reducedthe numberof componentsper computerbut r eliability
would have bee n adversely affected.
In the wake of the Apollo 11 ight, Dave Hoag, th e MIT Program
Manager for Apollo, and I were invited to Russia as guests of the
USSR Academy of Sciences. We took with us the NASA lm of the
lunar landing.
It was an immediatesuccess. No matter where we went, everyone
wanted to see the lm. If there was n o screen, they would hang up
a bed sheet. When the lights went out, cleaning ladies, janitors, and
other extraneous folk would squeeze into the room. It was shown s o
many times on so many strange projectors that the sprocket holes
were damaged beyond repair.
During the question and answer period following the show, there
was always the question:
“When will the Americans go to the moon again?”
Fortunately, we knew the planned date of Apollo 12 a nd told them.
They were incredulous. No one ever announced launches ahead of
As it happened,we were dead right. Apollo 12 did y on schedule
November 14, 1969. That must have done a lot for our credibility
back in the USSR.
The Soviet penchant for secrecyhas always baf ed me. Just before
we ew to Russia, the So yuz 6 with two cosmonauts aboard was
launched on October 11, 1969. When we changedplanesin London,
we learned of the launch of Soyuz 7 on October 12. “Wonderful,
we th ought. “We will be there in person and learn all about this rst
As we drove toward Moscow from th e airport with our host,
I asked about t he purpose of the double launch. The reply was
an irrelevant comment about some housing project we had just
The next day, October 13, Soyuz 8 lifted off to j oin the other two.
Now there were three sp acecraftand six cosmonautsin orbita real
spectacular! We knew of this o nly from pictures in
. None
of us read Russian so we were at the mercy of our Russian host for
details. But, to our dismay, we again learned nothing. Any question
was countered by a quick change of th e subject.