Apollo Guidance Computer Activities

AGC - Conference 1: Joe Gavin's introduction

Apollo Guidance Computer History Project

First conference

July 27, 2001


Joe Gavin's introduction

JOE GAVIN: Well, my name is Joe Gavin, class of '41 at MIT in aero-astro; in the last class that even worried about biplanes. That gives you some idea of the times.

After the Navy I went to Grumman in '46, spent 40 years there. And Grumman in the 50's was a pioneer in putting together the really early electronic systems in naval aircraft; the attack airplanes, the reconnaissance early warning airplanes, involving radar, neutral platforms, and extensive communication.

And I think it was that background, and at the same time we made the transition from pilot muscle to servo-controls and so on, and high speed airplanes. And I had been a project engineer on a couple of airplanes at the time that the space age came upon us. The first thing we did of any consequence was the opening of Astronomical Observatory, which I think really was the thing that made us ready to bid on Apollo.

Now, we have a late start, because as you go back and look at history, you'll find that everybody else started a year before the decision for lunar orbit rendezvous came along. And we had abetted the acceptance of lunar orbit rendezvous. Tom Kelly and some of the guys had done a study to verify the weight of damages of having a separate device to go to the surface of the moon and return.

The NASA competition for the lunar module, as I say, came roughly a year late compared to everything else. And it was not a competition for design, it was answer 20 questions and we'll see if you know what you think you're going to do. We must have answered it more or less right, because we did eventually negotiate a contract which allegedly provided an award fee system based on trading off performance, schedule and cost. It took us about 90 days to figure out that there was no trade off. You absolutely had to have the performance. You did your best to meet the schedule, and the cost came very much to the disgust of many of the bean counters. And we did get our priorities right. We were behind, and over cost all the way until we got to the basic missions. We had no award fee in those years. We were successful on every mission and did very well on the award fee from that point on.

I envied the MIT people that they could use integrated circuits. We spent thousands of hours penalized because of having to use discrete devices and communications gear, and so on. NASA insisted that we not use integrated circuits in so many other places, and we had so much trouble as a result of that. The communication system is a prime example.  The system may not use integrated circuits. We paid thousands of hours of testing and fumbling around on account of that. I do know something about how the system was hooked up on a complete lunar module system test at Bethpage. There is a reference in one of these documents to that set-up, but this is something we did with aircraft during that time period, so it was not that unique.

The unique thing about it was that we got to use it, because I think the Draper people wanted dearly to run that test here. And we, as being responsible for the overall module, said, in effect, in a pig's eye. Our name is on that vehicle and we're going to run that test, and we did.

And then, of course, we were involved with procuring the back-up system which was a strap down affair, which was deliberately a different technology from the Draper approach. There was a period where we looked at the question of should you have two different technologies between the primary and the back up, or should you just have duplicates of the same thing? And I have forgotten the details, frankly, but we came up with the idea that it was better to have something completely different, and we did.

I should explain that as integrator, and that's what Grumman was. We build the airplane, and we integrated a lot of pieces and parts. We were responsible for procuring the rocket engines. And now half of the effort was subcontracted out of Grumman, but we were responsible for putting it all together and making it work. We did some very elaborate testing procedures, none of which were truly visualized at the beginning. Because, you see, we had a machine that could not be flight tested prior to the mission. A fundamental difference from our experience with aircraft.

So over a period of several years, they have developed an extraordinary ground test set up where, to be quite frank, we didn't know what we were doing when we started. We probably did a lot of tests that in hindsight were not necessary. We got smarter as time went on. But we did a lot of exhaustive ground testing because there was no other way.

Each vehicle, when it was launched, was brand new. Propulsion had never been fired in the vehicle. There had been a lot of electronic tests for the system, but -- And of course, we never got a vehicle back to see how it survived. So this was a little bit like doing things blindfolded.

It seems to me, one of the things that you really need to emphasize is the overall state of the art at the time. You people, looking across the room here, are young enough, so you perhaps don't remember those big mainframes, IBM mainframes, that we all ran, or the matter in which we used them, which is quite different from the way as I understand it, today's engineers function. Not being today's engineer, I'm not sure I know how today's engineers function, but I'm sure it's different from the way we did it.

But the context is really important, because a lot of the things that were done in Apollo -- and this is certainly true of this computer -- were pressing the state of the art to a degree that is hard to imagine unless you know the situation at the time. Some of these folks can talk more about what Draper had at hand to do this work, but I've gone back and looked at it myself and I've wondered how we could have been so lucky that it all worked.

CLINE FRASIER: To pick up on Joe's point, the state of the art at the time for commercial guidance systems and military aircraft in terms of reliability was about 15 hours mean time between failure. And you ask the question how could we all have been so audacious as to think you could make simply two orders of magnitude improvement in one leap to do something like Apollo?

JOE GAVIN: That's a very important point, because in looking at the whole vehicle, and using the information from then existing systems, you could -- a computer that would take 44 attempted landings to get one success with it. For somebody working for the president's scientific advisor, obviously that was an untenable situation. And we struggled with this at Grumman saying we know what our naval aircraft are doing, and any time the failure was not very encouraging.

So we made a very basic assumption, which we formulated in retrospect, because we stumbled into it. And that was there is no such thing as a random failure. If you start out on that premise, you go at your design quite differently. I just remember some "gee whiz" figures on that. In ten years of testing, ten years of testing -- This includes tests at the development base board level, the box level, the subsystem level, the system level, and the vehicle level, if you take all of those, we had something like 14,000 anomalies. And of those, there were only 22 that escaped analysis. Those, we just changed something anyhow. But if you make the assumption there is no such thing as a random failure, there has to be a reason for a failure, you do things differently.

Grumman had a unique culture, corporate culture. That's a whole story by itself. NASA never understood us. I'm sure we never understood North American, later Rockwell, but we did make it work. We stumbled -- 

Tom Kelly's book is, I think, a very good description of how we stumbled from crisis to crisis. It wasn't like an airplane. You could not flight test it. You couldn't examine it after the mission.  And we built a new culture in dealing with, particularly the electronics in Grumman, and it paid off in later times in our aircraft business. We made a major improvement in the mean time to failure of the tactical systems that we represented in the aircraft.

And some people say the space cadets took over the company. I got to be president mostly because the then-president had a heart attack. Nothing like having a little luck in some of these places.

And, of course, I think I have probably said enough. I know a lot about some of the people. I have spent a little time on the National Academy of Engineering -- Well, the Draper Prize Committee. I have done some compiling of some of the surrounding circumstances that attended the launching of Apollo. There are a lot of players involved, and there were many disagreements on how it should be done, and who should do it.

I really don't know anything about the details of the Apollo computer. I know how the system was supposed to work, and I know how it worked. And of course, we did integrate it into the spacecraft. That integration apparently worked very well. So I'm not sure how much I could really contribute, other than background information.

I recognize a number of the faces around the table. The fact that I guess we're still friends is probably indicative that it was one of the minor miracles of the whole system. But it's true. Grumman was a difficult contractor. NASA was much happier dealing with Rockwell, for a variety of reasons. We were arrogant. We were right most of the time. When we weren't right, my gosh, it was very obvious and very humbling. We had many cases of being, having to eat crow.

Amazingly enough, I think we all stayed more or less friends over the years. I'm sorry that a couple of the key people aren't here to talk about these things: Joe Shea, George Lowe for example.

We had some fascinating occurrences where it confirmed in my mind more than ever before that a lot of things can be analyzed. There are a few decisions that are based on the art of the situation. They come up with almost equal answers when some decision has to be made.

One of the key cases in point was the rocket engine for the descent stage. Because it had to be throttleable and that was a first, NASA had to start with two manufacturers. I can tell you that it was a difficult selection, and either one would have been justified on the basis of the technology. Grumman had to make that decision. NASA wanted to be in on it, but they didn't want to make it, because we were responsible for that contract. The one we picked worked, and I'm sure the other one would have too. This is why I say there are some things that don't fall out directly from an analysis.

Anyhow, it was a great job. I had a treat team. The interesting thing about Grumman at the time was that we had a core of people who had been with the company anywhere from 10 to 20 years. These were the core of the activity, and I can't say enough for the confidence that was there.

Dave Hoag's introduction

site last updated 12-08-2002 by Alexander Brown