|
|
Post by Jairo on Jul 20, 2010 16:04:49 GMT -4
What would be a fair criterium to gauge the capacity of the Apollo and STS computers, and to compare with modern ones? Clock, memory, operations per second?
|
|
|
|
Post by Czero 101 on Jul 21, 2010 1:22:22 GMT -4
|
|
|
|
Post by ka9q on Jul 21, 2010 8:18:54 GMT -4
What would be a fair criterium to gauge the capacity of the Apollo and STS computers, and to compare with modern ones? Clock, memory, operations per second? All are relevant, but you have to be very careful to compare apples and apples. Clock speeds are especially deceptive. There are fundamental tradeoffs in computer design between gate count and the clock count needed to perform some function. More gates allows fewer clocks. The Apollo guidance computer (AGC) was severely limited in gate count, so even simple operations took many clocks. E.g., there was no floating point hardware; it had to be emulated in software. Not only do today's computers have much faster clocks than the AGC, they do much more on every clock through parallelism. This is again made possible by the very high gate counts now available on modern ICs. Parallelism now takes several forms: multiple CPU cores; vector processing; super scalar execution; and pipelining. I don't think the AGC had any of those features. Intel got itself in serious trouble in the 1990s and 2000s by competing on clock speeds. Faster clocks mean higher power, a major system issue, so today there's a very deliberate trend to limit clock speed to improve energy efficiency and cooling. So again, just comparing clock speeds would greatly understate the difference between the AGC and a modern computer. Memory size is also obviously important to a computer's capability. There are more tradeoffs between gate count and performance, notably through caching, so again you have to be careful when comparing the AGC to the modern computer. Modern RAM is not only much bigger and faster than the AGC's read/write memory, but the modern CPU memory cache makes that RAM seem even faster. The bottom line is what matters; can you program a given computer to do what has to be done in the time available? Remember that the AGC, though very sophisticated for its day, didn't have to do any serious number crunching. It had two main functions: serving as a fly-by-wire control system ("digital autopilot") for the CSM or LM, and maintaining the spacecraft state vector through dead reckoning (the state vector is the spacecraft position and velocity vectors at some moment in time). All the real number crunching was done on the ground in the Real Time Computer Complex (RTCC), a room of big mainframe computers in Houston. Every maneuver was planned there and read up to the crew, who wrote them down. In addition to their planned course corrections the crew always had a number of abort scenarios on hand so they could return to earth if communications were lost. The transcripts are full of these "maneuver PADs". Today, all of this could easily be done in a small onboard computer. Although Apollo relied primarily on ground tracking to correct errors in its state vector, the crew could autonomously determine its state vector through scope sightings of relative star positions to the earth and moon. Apollo 8 (notably Jim Lovell) showed they could do as good a job as the ground, which is pretty remarkable. With the increased computing power now available, an Apollo mission could now navigate itself through an entire mission without help from the ground. |
|
|
|
Post by Jairo on Jul 21, 2010 11:42:29 GMT -4
Apollo 8 (notably Jim Lovell) showed they could do as good a job as the ground, which is pretty remarkable. Very impressive. Was it good enough for a landing or reentry? Wouldn't it become harder as the body's relative size gets bigger? |
|
|
|
Post by ka9q on Jul 21, 2010 12:50:48 GMT -4
Very impressive. Was it good enough for a landing or reentry? Wouldn't it become harder as the body's relative size gets bigger? Lovell's observations of stars against the limbs ("edges") of the earth and moon were good enough to generate state vectors that exactly matched those produced by Houston from radio tracking data. This meant that yes, they could have landed on Earth with them. Apollo 8 had no lunar module. I don't know if they could have landed on the moon with one using only their own observations. Lovell was CMP, which meant that on a regular Apollo flight he was only responsible for CSM navigation. The ground could produce radio-based orbital solutions much more quickly than the crew. During the first few landings this was vital after descent orbit insertion to ensure that the orbit would not intersect the moon (i.e., they wouldn't crash before powered descent initiation). They had to determine the LM's orbit ASAP after AOS (the burn was done on the far side) so the crew could still do an emergency wave-off burn if the pericynthion was dangerously low. A similar risk presented itself after lunar orbit insertion (the initial arrival at the moon) but was greatly alleviated by breaking the burn into two components. The first put the CSM+LM into an elliptical orbit, and the second circularized the orbit. This made the first burn far less critical and eliminated the risk of a slight overburn that would have them crashing on the near side a half orbit after LOI. |
|
