How Engineers Designed The First Computer To Fly In Space

[catdlr]

May 24, 2023
The Gemini spacecraft carried the first fully functional computer in space, it ran with a clock speed of about 7kHz, and uses 26-bit data and 39-bit memory words, with 4096 words of memory. It was one of the oddest computer architectures ever developed, for a tiny niche, only ever used on the Gemini. Some of the design may have made it into the Saturn Launch Vehicle Digital computer, but

[laszlo]

Not so fast on that "First Computer to Fly in Space" stuff. Gemini 1 didn't fly until April of 1964. The Minuteman I missile began flight test 3 years before and became operational in 1962. Sometime in that period its D-17B computer passed the Kármán line into space.

The D-17B was a solid state computer based on DTL and RTL logic. It used 11- and 24-bit data and 29-bit memory words, with 5,454 words of single precision or 2,727 words of single precision memory and could perform an addition in 78.125 µs giving 12,800 additions per second. It had 48 input and 43 output lines. Bulk memory was a ferro-magnetic drum with a 78.125 µs cycle time. MTBF (mean time between failure) was 5.5 years.

I don't know if some other computer beat the D-17B into space, but it certainly beat the Gemini computer.

Edited to fix fat-fingering the Gemini 1 date

[Blackstar]

I want to know how people define "computer" in these examples. The Discoverer missions, first launched in 1959, had a control system that could be given commands in orbit.

[laszlo]

I'm going with a general purpose programmable processor with a stored program. The processor implements instructions consisting of computational functions and decision making capabilities which are not specific to any particular application. The program in the stored program device provides the instruction sequences to implement a particular application and can be changed without any mechanical changes at all or mechanical changes limited to replacing ROM (read-only memory).

The only information that I found about the Discoverer control system indicated that it was a programmable timer, not a true computer. If I'm interpreting it correctly, the hardware performed fixed operations in the blind, triggered by a timer which was programmed with the actual times that an operation was needed based on a determination of the actual trajectory conditions by ground stations. I'm sure Blackstar will have the correct detailed answer if this is wrong.

[edzieba]

GAMBIT-1 had extensive computer control so that headily beats Gemini by a year. CORONA had a rapidly changing computer system over its lifetime (KH-1 to KH-4B), so depending on how you define 'computer' (e.g. does the KH-1 executing the stored program to reorient and eject the film bucket when commanded from the ground count?) one of those would be a candidate for the first computer in orbit. Probes like Ranger, Mariner, Luna, etc would also be in the running, as well as the Agena stage itself (at least for later Agena models).

[Blackstar]

The only information that I found about the Discoverer control system indicated that it was a programmable timer, not a true computer. If I'm interpreting it correctly, the hardware performed fixed operations in the blind, triggered by a timer which was programmed with the actual times that an operation was needed based on a determination of the actual trajectory conditions by ground stations. I'm sure Blackstar will have the correct detailed answer if this is wrong.

That is a correct definition. Except that they could command it from the ground to skip pre-programmed steps. So if the next step was "begin the reentry sequence" they could tell it to skip that step.*

It did gain functions over time, so it is possible that the Agena B had more commandable capability and functionality than the A, and the D had even more. And of course the Agena D stayed in service from the early 1960s into the later 1980s, so I assume that it was substantially upgraded along the way, even if we don't know how.




*I'm probably describing this wrong. I don't think it was only programmable to skip steps, I think that additional time could be inserted to delay steps. So instead of telling it to skip reentry entirely, it could be told that instead of starting the reentry sequence in 30 minutes, it should start it in 90 minutes. But they really were gaining a lot of capability all through the early 1960s, so what was true in 1959 may have been upgraded a lot within a few years.

[zubenelgenubi]

Gemini 1 didn't fly until April of 1968.

April 1964.

[Proponent]

Gemini 1 didn't fly until April of 1968. The Minuteman I missile began flight test 7 years before and became operational in 1962.

Gemini 1 flew in April 1964, hence 2 years after Minuteman.

[john smith 19]

I want to know how people define "computer" in these examples. The Discoverer missions, first launched in 1959, had a control system that could be given commands in orbit.

Fair point.

A lot of the early stuff used what were called "Sequencers"

This is essentially a cam timer IE a set of disks on a common axis driven by a slow speed motor. As the stack of disks rotate they raise and lower electrical contacts, closing or opening various circuits. Cross connecting some of these together can give some logical decision making.

For a very long time this was the SoA in washing machine control. 

When digital chips started coming in they started being replaced, at first by simulated sequencers, a set of parallel channels being successively read out at a fixed rate and driving various valves and actuators. 

Then they moved up to various varieties of actual computer.

For anybody who's interested in this sort of thing I guess the key book is   Computers in Spaceflight: The NASA Experience  I think there's a specific book on the Apollo Guidance Computer by Eldon Hall as well. The NASA bookd does discuss sequencers as well.

 Keep in mind though these were not "monolithic" IC's as we know them but miniature ceramic circuit boards with individual transistors "flip chip" bonded and screen printed resistors into (at most) a pair of NAND (or NOR?) gates. IOW more like miniature hybrid circuits, for those aware of that technology.

[Jim]

https://history.nasa.gov/computers/contents.html

[Blackstar]

https://history.nasa.gov/computers/contents.html

Just from a quick look at this, NASA considers that the first "digital computer" to fly in space. That leaves open the possibility that an analog computer flew before it. And it suggests that maybe the thread title needs to be amended.

[joek]

Just from a quick look at this, NASA considers that the first "digital computer" to fly in space. That leaves open the possibility that an analog computer flew before it. And it suggests that maybe the thread title needs to be amended.

Certainly. And not even "digital computer" is definitive. At the risk of getting into the weeds, there are many "digital computers" of old which many would not recognize as such today (as in, not Turing complete). But they did their jobs. As to analog computers, think we can go back quite a bit further... darn near anything that went to space (or atmospheric for that matter) could claim to have such, unless it was blind-deaf-dumb open loop.

[Remes]

https://history.nasa.gov/computers/contents.html

Just from a quick look at this, NASA considers that the first "digital computer" to fly in space. That leaves open the possibility that an analog computer flew before it. And it suggests that maybe the thread title needs to be amended.

I also didn't read it in detail. But it talks about the computer for the first *manned* program which uses computers in all mission phases. So first computer in space and first computer in space for all mission phases in a manned space flight are two things. Geminis computer was only used in parts of the mission. And it was not critical.

An analog computer typically is not counted as a computer. Minimum for a computer is (among others) I believe reprogrammability. Whether this is a tape, a punch card or a board with connections or a core rope module.

That said, Mercury had a control algorithm implemented with transistors. (Astronauts could choose to activate the autopilot or steer by themselves). So discrete mathematical operations (I believe they would use transistors connected as an operational amplifier). I think it was about 1500 Transistors, something in that ballpark.

The poseidon missile guidance used digital logic, but not programmable. It calculated difference equations. ( y(t) = b1*y(t-1) + b2 *y(t-2)+ ... + a0*x(t) + a1*x(t-1) + ...)
y(t) beeing the output. y(t-n) last outputs. x the inputs and last inputs. Lot of things can be build with that. Integrators, differentiators, adders, low/high/bandpass filters, ...
People from Poseidon guidance would later end up nearly completely in the agnc program (all were at MIT).

The Saturn V had a mixture of a digital computer with an analog addendum. The analog part would close the control loops for e.g. tilt/yaw/roll of the rocket. Going even a step deeper, the TVC on Saturn V would have control loops closed in mechanics. Even on the shuttle this would be used, but now with 4 voting hydraulic loops which close the control loop by means of mechanic.The analog loop closure on Saturn V, I assume, is mainly to avoid dead time in the control loop, which lowers controlability.

Something about the autonetics minuteman-1 computer (the time offset doesn't work for me, go to t=232 for Minuteman I):

I said the same on Twitter, that the autonetics was the first space based computer. But okay. It's Scott Manley against ~5000 engineers, scientists, ..., who can't wait to find an error in his videos. For that he is really doing a great job. The Carl Sagans of our time.

[illectro]

I'm delighted to see everyone coming up with their own firsts - I'll be honest, I skipped a lot of qualifiers in the title. I really just wanted to focus on the Gemini computer because I had a really nice simulation in the form of Reentry
(https://reentrygame.com) - but also I'm a computer nerd and loved reading all the documentation.

The minuteman computer certainly qualifies for another story.

[john smith 19]

https://history.nasa.gov/computers/contents.html

Yes that was the book I linked to.

[catdlr]

I'm delighted to see everyone coming up with their own firsts - I'll be honest, I skipped a lot of qualifiers in the title. I really just wanted to focus on the Gemini computer because I had a really nice simulation in the form of Reentry
(https://reentrygame.com) - but also I'm a computer nerd and loved reading all the documentation.

The minuteman computer certainly qualifies for another story.

Thanks for chiming in.  I routinely post your special articles (videos) in this forum hoping to spur new thoughts and viewpoints.  Appreciate your work and content.
Tony