|
|
|
The architecture is ingenious. There were so many unprecedented innovations in that cabinet that one hardly knows where to start listing them for importance.--John Gustafson Gary Sleege and John Erickson have collected nearly 500 vacuum tubes and ordered yards of cotton-rubber insulated wire, both relics of a time past. Now, if they can get their hands on an IBM punch, Christmas will come early for them. Sleege, Erickson and six other engineers are part of a team that is building a full-scale replica of the Atanasoff-Berry Computer (ABC). The ABC, the worlds first electronic digital computer, was built on campus by professor John Atanasoff and graduate student Clifford Berry from 1939 to 1942. The reconstruction, a tribute to the computer pioneers, requires the usual engineering skills, but also a considerable talent for scavenging. The IBM punch, which used perforated cards for data entry in the pre-laptop days of computers, is an example. There were thousands of these punches all across America, said Del Bluhm, manager of research and development engineering at Ames Lab and director of the replica project. Theyve all disappeared. The ABC replica project is not your Heath-kit variety nostalgia trip. The parts dont come easy -- and they dont come with detailed diagrams. The original drawings and design notes were discarded in the 1940s, as was the ABC itself. We thought it would be easier to build [the ABC] in 1995 than it was in 1939, Bluhm said. But we are re-engineering and recreating what they did in the 1930s, so all of the hours they spent in design and building, we will probably re-spend. Access to parts and subtle differences in how things are done get in the way all of the time, added John Gustafson, an Ames Lab computational scientist and replica team member. We cant use technology to shortcut a few things. It wouldnt be authentic. Which is why the team agonizes over the use of plastic-insulated wires of today or the cotton-rubber insulated wires of yesteryear. The team has assembled a varied collection of 1940s electronic paraphernalia from their visits with antique radio buffs (to acquire vacuum tubes), neon sign manufacturers (transformers), gear companies (gears and bearings) and specialty plastics manufacturers (a special material for the drums). Sometimes, theyre lucky enough to have copies of Atanasoffs original purchase orders to guide them. More often, they have only a grainy photograph and some fading memories. Its sort of like Sherlock Holmes detective work -- a piece of information here, a piece of information there, add a picture and come to conclusions as to what that component did, said Sleege. We guess a lot, Erickson added. HELP FROM FRIENDS An invaluable resource for the team has been contact with those either directly or indirectly involved in the original project. In the fall of 1994, Erickson and fellow team member Dave Birlingmair paid a visit to the Atanasoff family. Armed with a portable copy machine, they spent two days visiting with the family and copying original documents. Another big help has been Robert Mather, a retired physicist in Oakland, Calif. A graduate student in the late 1930s and early 1940s, Mather wired the original ABC under the direction of Berry. He still has a pretty good memory of how things were laid out, Gustafson said. This will be helpful for Sleege, who is in charge of replica electronics, because documentation of ABC electronics is spotty at best. Mather helped the team figure out where ABCs output was read, a mystery the team faced early in the project. Scientists read the computed answer by peering into an axle to find odometer numbers rotating into place. [Mather] showed us where the output was created, Bluhm said. We went back and looked at enlarged pictures of the ABC and said By golly, hes right! PAPER AND FIRE A formidable problem the team currently is grappling with is the electronic read and write system of the ABC. The system acted as sort of scratch paper for the computer as it worked its way through an equation. Atanasoff had to devise a way to record 1,500 bits of data per second and there was no method in the world for doing that at the time, Birlingmair said. IBM punches could handle 10, 20 or 30 numbers per second. So he devised a system of electrostatic punches in which ABC burned holes in paper with a spark. Later, the computer would find out where these holes were and read the data. Its incredible that it worked, Gustafson said. And they got it so that it failed only once every 10,000 to 100,000 times. The ABC had 30 such electrostatic punches working in unison. So far, Erickson has been successful zapping holes in paper, but only on one punch. Getting 30 to work in unison will be trickier. Zapping the paper without igniting it is tricky as well. The wrong paper, Bluhm said, will catch on fire. Atanasoff and Berry eventually found a type of paper that worked, but never documented it. After struggling in vain to find that paper, the team finally turned to Jean Berry (Clifford Berrys wife and Atanasoffs secretary). She knew the name of the paper and where we could buy it right off the top of her head, Gustafson said. THE SMELL OF COMPUTING When the ABC is completed (scheduled for August 1996), the team will operate the replica at least once, hoping to match the performance of the original. While designed to solve 29 simultaneous equations with 29 unknowns, an operation that would take nearly 50 hours of ABC computing time alone, its believed that the original only solved five equations with five unknowns. If we can solve five equations and five unknowns and get the right answer, Ill be satisfied that we succeeded, Gustafson said. The rest is cosmetic. We need to know what color the lights were, Gustafson added. We only have black and white photographs to go on. What did it sound like? I have a feeling that it probably produced an odor because of those sparks punching holes in the paper. It must have smelled like a thunderstorm just happened. LOOKING FOR A FEW GOOD PUNCHES As they work through building a replica of the ABC, the team of engineers would like to hear from anyone who has knowledge of the original machine or access to vintage parts like those used in its construction. Especially helpful would be leads on an IBM punch. The team is headed by Del Bluhm, manager of R&D Engineering at Ames Lab and includes Gary Sleege, Dave Birlingmair, Harold Skank, John Erickson and Terry Herrman of Bluhms group as well as retired ISU faculty member Al Reed, Ames Lab computational scientist John Gustafson and engineering student Jennifer Augenstein. |
加入最爱