I have the pleasure at work of learning about Synthetic Aperature Radar (all you techies out there are sure to be green with envy. Especially since I don’t actually have to learn the math. Just enough to brief the boss).
It is a cool technology and all that. I am enjoying myself.
What impresses me the most is something related-the computing power needed for SAR. It is considerable-very considerable. Why, even a gaming laptop might struggle with processing a SAR image. While back in the day: in 1978 NASA launched the first publicly acknowledged SAR satellite (SEASAT). At the time of launch the computer being developed to process the data hadn’t been finished. When this bleeding edge NASA computer-the supercomputer of the day-was finally ready it took 20 hours to process 18 seconds of data. And that was a great advance.
What amazes me is the original technology developed for processing these images. SAR was first demonstrated around 1960. But no digital method of processing a SAR image was conceivable. For the next 15 years the power of America’s universities, defense technologies, NASA, etc was focused on developing processing techniques. Bleeding edge optical analog computers, multi-channel scanners, advanced technology of all kinds were invented. All of which in retrospect looks like steampunk. But it was done because it was worth it and needed to do the job. Eh. today all that can be done in most people’s homes on their desk. Big deal. And that amazes me. The change that high-power computers have brought to us. And how little we realize the changes in our lives. Here we have technology that would and did amaze anyone who knew about it, that engaged the brightest minds, and now an undergrad probably implements the technique as a class project. Technologies are being routinely developed today that would have consumed huge resources 15 years ago (cell phones for example), that go from idea to market in 18 months. And we just accept it as normal. The change in technological ability is amazing. The human in the equation is still the smart part, still the driving force. But computers have enabled amazing feats. It shocks me is how we have come to accept what computers and computing speed have allowed us to do.
And remember: All of the time and effort going into the spellchecker isn’t going into any other aspect of the program, such as optimizing the commands used to control it.
I did my master’s in computer science at Penn, and during my first semester I had a class that met in the Linux cluster behind their ENIAC display. It was a little strange to look up from whatever I was working on and see a refrigerator-sized component that had less processing power than my phone (never mind the Linux box I was using).
This has been going on for quite a while. When I was in high school I learned to program on an LGP-21 - 4K of rotating memory, built in 1962. When I was in grad school, 1974, I was working on microprogramming, and I wrote an emulator for the LGP-21 in Lockheed SUE microcode, which I ran using a simulator running on a PDP 11/20. It still ran faster than the original machine. Now the 11/20 could be implemented in a tiny chunk of the microprocessors we’re working on now.
I did the thing mentioned in there, but for a tic-tac-toe program. I found that it would fit nicely if I switched a bunch of adds to subtracts between the computer move and the player move. Initialization involved making sure the adds were adds when you began.
BTW, the computer in the story is an LGP-30, which was the older version of the LGP-21 made with vacuum tubes, not transistors. It ran faster. I met a Harvard Business School professor who actually used one at the Aberdeen Proving Ground in the late 1950s.
You can find an LGP-21 programming manual (I have this, extracted from my high school) here.
While back in the day: in 1978 NASA launched the first publicly acknowledged SAR satellite (SEASAT). At the time of launch the computer being developed to process the data hadn’t been finished. When this bleeding edge NASA computer-the supercomputer of the day-was finally ready it took 20 hours to process 18 seconds of data. And that was a great advance.