Weather Radar on a PDP-8 in Antarctica
[If you’re only interested in the tech, you may skip ahead.]
Background
bill75 who wrote the excellent posting at Timesharing on the MIT Weather Radar PDP-8/IX was my fraternity brother, a senior while I was a freshman. He describes how in college he sought a computer job; anything better than retail! Same for me: I didn’t want to go back to stocking sleeping bags in an unairconditioned warehouse in New Jersey for the summer. Bill suggested I apply at Weather Radar, and I got a part-time job as an “undergraduate researcher”. I was uninterested in weather but very interested in computers, so writing software for the computers that the real researchers used was right up my alley. My minor contributions (positive and otherwise) to MIT Weather Radar are mentioned in comments on that thread.
At the end of my sophomore year, I realized that I missed the “real world” outside academia, so I left Weather Radar and looked off campus for a summer job. Then, a few weeks into my junior year, I did very poorly on a couple of tests, and realized that I was going to have to really buckle down and study hard for the rest of the term. So the obvious thing to do was to have a party, since there would be no partying after that! I invited a bunch of friends, including Bill (we had left the fraternity and were living in different apartments in the same neighborhood). At the party, Bill mentioned that the University of Nevada Reno weather radar group had gotten in touch with MIT Weather Radar, looking for a computer/radar tech to run a weather radar system in Antarctica for a year. I was super excited! My experience would be applicable, so I had a good chance to get the job. I loved to travel and hadn’t had many opportunities to do so. (I was intensely jealous of Bill’s trip on the Gilliss.) Bill warned me, though: this wasn’t just a three-week gig; you had to commit to a whole year. Twenty-year-old me thought that was just fine; how else would I ever get to Antarctica? I told my best friend Dan about it; he was happy for me but had to struggle to hide his disappointment at losing a friend for a whole year.
Long story short, I got the job. Flew out to Reno (second airplane trip ever), met the team and started to learn all about their system. My boss was Joe Warburton, a bona fide research scientist.
I spent January and early February 1977 in Reno. I liked my colleagues and loved the job, but never liked the city. Then all the equipment and yours truly was shipped off to Palmer Station, Antarctica. We drove it to Port Hueneme in LA where the equipment (and my personal gear, including my stereo and record collection) was loaded onto a ship. I then flew from LAX to Buenos Aires, Rio Gallegos, and Ushuaia, where we boarded R/V Hero for the 4-day voyage to Palmer Station. I was violently seasick the entire time. Hero was some NSF administrator’s dream of reconstructing the whaling vessel Hero that Nathaniel Palmer sailed as the first American to sight the Antarctic continent. It was a 125-foot wooden ship with no keel, so it could ride up on light sea ice and act as an icebreaker. In the Drake Passage, it rolled 40 degrees and pitched 10-15 degrees, in a corkscrew motion that I never got used to and hope never to experience again. I arrived on February 14, 1977.
Palmer Station in mid-winter. The Biolab building is the black one, the GWR the gray one. The radar is the black dome at the rear of the GWR building.
Abstract of the science
Air-sea-ice interaction investigations with X-band radar.
Dr. Joseph A.Warburton, Desert Research Institute, Reno, Nevada.
We are interested in how the Antarctic Peninsula, acting as a mountain barrier, affects the tracks of storms approaching Palmer from the southern ocean. We will continue to investigate storms within a 150-kilometer radius of Palmer Station to determine their motions, intensities, and structures, and their effects on sea ice and iceberg formations and movement. We are particularly interested in the structure of precipitating storms, apparently caused by considerable temperature differences between the surface of the ocean and the air masses passing over the ocean.
(He was particularly interested in this because the Sierras do a similar thing with storms coming off the Pacific, leading to the Nevada desert. He spent much of his career studying cloud seeding and other ways to increase precipitation in Nevada. Whether that was purely scientific or topics the Nevada legislature was particularly eager to fund was never clear to me.)
The Tech
There was a PDP-8/e with 8K (two banks) of nonvolatile core memory. This was one level up in integration, and considerably smaller, than the 8i at MIT Weather Radar. Still based on 7400-series TTL, but the CPU was on 3-4 boards instead of dozens of Flip Chips. No hardware mods.
Mass storage was a 9-track 1600-bpi tape drive. Not supported by any PDP-8 operating system, so there was no operating system. More on that below. No other storage devices: no paper tape, no DECtape, etc.
The terminal was a Tektronix 4014 with a storage-tube CRT display. It was definitely nicer than Weather Radar’s TTY38 and separate storage-tube point display, since the 4014 could do characters and vectors along with points. There was a dream of displaying radar data on this screen, but that never happened.
Another “terminal” was a 1200-baud modem. (Vadic or Bell 201? Don’t remember.) More on that below.
Then there were the two custom radar interface boxes: digital and analog. The digital box took the radar echo return signal, ran it through an analog-to-digital converter and DMAed it into the PDP-8 memory. Similarly, the position of the radar antenna ran through a synchro-to-digital converter to provide a digital azimuth readout. The analog box controlled radar rotation speed, pulse width, and probably other stuff I don’t remember. It also drove an analog storage-tube display, for which I had a Polaroid scope camera. This was useful for real-time observation and occasional photographic recording, but was never intended to be a primary data recording mechanism. We also brought a Tektronix oscilloscope for hardware debugging purposes.
The radar room at Palmer. The tape drive is in the rack barely visible at far left. Then the PDP-8/e, digital and analog interface electronics, debugging scope, display scope and Polaroid scope camera.
The actual radar was a Bendix X-band unit designed for commercial aircraft; it was about 12”x12”x24” and very heavy. We had a spare, which I hand-carried back to Reno. That caused some consternation at US Customs, since equipment valued at over $10K is subject to paperwork nobody had thought to mention. Fortunately, they let me and it through.
This is me, installing the radar.
We had been given a room on the second floor of the GWR (Garage, Warehouse, and Recreation) building to house our equipment, some of which had been shipped earlier.
Off in a corner of the room was a rack-mounted Motorola VHF transmitter/receiver, used to communicate to NASA’s ATS-3 satellite via a large helix antenna outside the building.
Besides half-duplex voice communication and sending/receiving faxes, we hoped to use this to send some radar data back to Reno using the modem, rather than wait until the end of the Antarctic season (December) to ship the magnetic tapes back to civilization. I also had a secret dream of being able to connect to the Arpanet and email and chat with friends at MIT and Stanford and SRI. Alas, none of us at the time knew much about modem technology or about VHF radio propagation to geostationary satellites and back. Not surprisingly, it turns out that modems are very good at dealing with the kinds of signal interference encountered in the POTS telephone network. The kinds of interference experienced by satellite transmission are not similar. So even on the best days, our test transmissions were so garbled as to be unusable, and we likewise were not well-versed in then-novel technologies like error-correction codes and packetization. So we gave up on that. But I had fun typing a character, then switching the modem data input from the terminal to the modem data output, and watching my character oscillate between Antarctica and the satellite 30,000 miles away. It would gradually pick bits until it became a ~, then a rubout, then nothing.
We had a Xerox Telecopier fax machine, likewise hooked up to the satellite system. I still have some of the faxes we received over it. That saved my butt early on. The Tek 4014 terminal stopped working about a week before the end of the summer season, when the rest of the Reno team would head back and leave me to my own devices. We were able to trace the failure to the custom clock chip, which had stopped ticking, so to speak. We talked to Reno via satellite, and they talked to Tektronix in Beaverton, who were able to fax a timing diagram showing the two-phase clock signals this chip was supposed to output. My colleagues were able to construct a TTL implementation with some of our spare chips, which fixed the terminal.
Programming
So what about programming this stuff? All I had was a PDP-8 manual, interface descriptions for the magtape and radar peripherals, and the On-line Debugging Tool or ODT. ODT was a tiny program that was “online” in the sense that you could use a terminal to examine and change memory locations and jump to a location, all in octal. No operating system, no assembler, no editor, no radar control program. So the first thing I did was to write “Emacs”. Using the high 4K as my editor buffer, I implemented self-inserting characters and dispatched on control characters. (Aside: though I was familiar with the “gap” approach to editor buffers, I didn’t bother since my buffer size was limited to 4K.) A few control-characters became my OS: ^W to write the upper 4K to tape, ^R to read it, ^T to write EndOfTape, etc. I implemented all this by writing code on pieces of paper, hand-assembling it, and typing it in using ODT (then carefully saving it to tape). Sounds a lot like Bill’s description of Ken’s work on the TI-980!
The radar control program was pretty simple: start it up with parameters like rotation speed and bucket size that could be adjusted with ODT, then save each “ray” (a header with date/time, azimuth, bucket size, maybe something else, then a vector of data samples) in the upper 4k, then to tape using a dirt-simple double-buffer which would store rays in 2k while the other 2k was being written.
Unlike Bill, I had no listing device so none of this code survives.
Palmer Station “technology”
The technology on the rest of the station was even more primitive. The radio room (in the Biolab building a hundred-odd feet away) had baudot teletype machines and HF transmitters and receivers which were used to communicate over US Navy radio frequencies. We could voice-chat most of the time with the folks at Siple Station and South Pole, but rarely with McMurdo and hardly ever with the US. So official teletype messages often passed from us to Pole to McMurdo to Christchurch NZ and on to the US, and back the same way.
The station manager would write up a “situation report” or SITREP once a week, with basic data like how many souls on station, what had been fixed or broken that week, visitors in the summer, etc. Fred the radio operator would type it in on an offline teletype, punching paper tape, then run the paper tape through an online teletype which would get printed out (and punched on papertape again) at Pole or wherever. There were other teletype messages from time to time for the rare things that couldn’t wait for the weekly SITREP. One such: President Jimmy Carter sent a message via this system to all Antarctic personnel at Christmas, thanking us for our service to the nation and wishing us happy holidays. We composed a suitably polite thank-you note and sent it back. It got to Pole and McMurdo and maybe as far as Christchurch before a Navy operator informed us that one does NOT send messages to the Commander-in-Chief unless he ordered you to do so.
Fred in the radio room with only a small fraction of the equipment visible.
The tedium of typing messages on teletype was most painful at resupply time. We had to make a shopping list of everything the next year’s crew would need for an entire year on station, along with enough spares to cover most any maintenance needs: food, fuel, etc. Those were long lists, requiring lots of typing. So Fred asked me if there was some way to make it easier. I introduced him to my “Emacs”, and he was smitten. “You can rub out without backing up the paper tape? That’s cool!” I wrote up a slip of paper documenting the “Emacs” commands and taped it to the terminal for Fred’s benefit. We ran a current-loop from the PDP-8 in the radar room to the radio room in the other building. I wrote a tiny ASCII-to-baudot translation program. The PDP-8 TTY interface let you set an arbitrary baud rate, so 75 baud was no problem. But it didn’t have provision for 1 start bit, 5 data bits and 1.5 stop bits. I had my baudot table output 1s for the unused data bits, effectively giving 2.5 or 3.5 stop bits. It worked, but the mechanical teletype machine made it clear via awful noises that starting up, typing one character, stopping, then restarting a fraction of a second later was going to (a) drive people from the room and (b) break the machine for sure. So I added a timing loop: after outputting a character, I’d loop for 1/10 of a second, then RST and send the next character. Perfect!
Return to civilization
The station “closed” in early March, when encroaching sea ice interfered with ship traffic. The last of the summer people left the station, and seven of us “winter-overs” ran things until early December when the sea ice retreated and Hero returned. I gave a quick tutorial to my successor, then shipped out on Hero (another 4 days of seasickness), then home to NJ. After the holidays, I returned to Reno, but told my boss I’d stay and help for no more than six months, since as I’ve said, I didn’t like the place. I helped write programs to read the tapes I’d written, and continued mentoring my successor via ATS-3 radio. I had a year’s salary (earned where you couldn’t spend any of it) burning a hole in my pocket, and lived walking distance to a flight school, so I got my private pilot’s license there (stories for another day and another forum).
Our replacements arriving, December 1977
In subsequent postings I’ll talk about some of the interesting intersections with other technology, and how this launched the rest of my career.
