Apollo Guidance Computer Activities

AGC Conference 3 - Reliability

Apollo Guidance Computer History Project

Third conference

November 30, 2001

Detail of AGC module


ELDON HALL: I just mentioned to Dave Mindell that the key issue is the reliability of it.

JACK POUNDSTONE: Well, I believe what Raytheon and Draper did a lot of pioneering in reliability in that program.

ELDON HALL: Yes, we did.

JACK POUNDSTONE: I know what I did from my end. In a lot of ways we competed with each other, which was healthy. But we started out with a policy that if you're going to analyze every stinking failure down to the nitty-gritty, you got to understand it all. We started building up a failure analysis lab. I bought the first sim. out there in Sudbury and all sorts of fancy materials equipment and developed a very good materials laboratory that never existed before. And Draper started doing the same thing.

ED DUGGAN: They forced us to do it in self-defense.

DAVE BATES: We were hired by North American to do an analysis of some of the parts. We did an awful lot of outside analysis for people because that facility was established.

JACK POUNDSTONE: But I assume other people in other programs were probably doing similar things. But I think we were probably a year or two ahead of everybody then, in terms of the depth.

DAVE HANLEY: Not the same that I've ever seen, because what we used to do is we’d take all the failures out of a lot and then we’d analyze them, then we used to put a weighting factor on them and with the weighting factor, you could just reject the whole lot. I don’t think I've ever seen that.

JACK POUNDSTONE: But I think there was a general understanding, whether it was in the spec or not I'll be damned if I know, but any time there was a failure, you just got to find out why.

DAVE HANLEY: Oh, yes, the failure analysis had to be done.

ELDON HALL: And eventually it got into the specs. It started much earlier.

ED DUGGAN: There was actually a lot of burn in experience, a lot of burn in components.

JACK POUNDSTONE: All that burn in that we did and so forth, years later I went on up to where he (pointing to ELDON HALL ) came from, Oak Ridge. It was Dullville in those days. They were building 18,000 computers a day up there for automobiles. And they had the damnedest burn in rack you ever saw. It was just Apollo multiplied by about a million. But I think a lot of the principles came out of what we were doing.

CLINE FRASIER: Initially they did, I think, 18 days burn in on those computers.

DAVE HANLEY: Remember when we had leaky packages? I think that's what Ed is referring to. And what Eldon and I did is we got some devices that we knew were really bad and we shoved them through the line to see how many would be picked up. They didn’t get picked up. And then after that, we tried all kinds of different systems. And finally, there was the weighing method. We had all these marble platforms with these balances on them.

JACK POUNDSTONE: I think Aaron come up with that one. That was the technique. We’d pressurize them then bleed them to find leaky packages. Yes, I remember that.

DAVE HANLEY: When I first went into the room and I saw all those scales, I just went, "My God, what have we done?"

JACK POUNDSTONE: Well, as I said, there-- I think a lot of the reliability techniques and practices that were developed in that program spun off to other programs

ELDON HALL: We spun off most of the work to the military spec 1833 and the failure analysis that's done on Apollo goes directly into that military stuff.

DAVE HANLEY: They copied them verbatim off of Apollo.

CLINE FRASIER: The thing I wish I had still is the color slide of the one flat pack that had had purple plague that got downed in a computer. I had to then stand up before the flight readiness review board at the Cape, and explain this was a failure we had not been able to correct, it was the only one we’ve ever seen like this, and there probably aren’t any more in there!

DAVE HANLEY: That was because of the leakage, I'm sure. Remember when they shut down the whole line and Brady, Hugh Brady, was yelling and screaming at us and getting us fired and everything else? I can’t remember.

HUGH BLAIR-SMITH: On the architecture side, I wonder if it’s still kind of extraordinary that we had what I call block zero, (AGC 3) the AGC3 block one and block two, and nobody worried about backward compatibility. That was extremely beneficial because everybody was learning everything about the project. And it was absolutely correct to throw away all that old shit and start over.

JACK POUNDSTONE: Well, you only got away with that because you didn’t have a spec.

SANDY BROWN: Folks, thank you very much from Professor Mindell and I for coming in. This has been extremely valuable and we hope that you continue your involvement with this project. There are a couple of people here who’ve been involved before. Thank you very much. This has been a very valuable session.

Home: Conference 3