"About those tapes: the card plant in Endicott got enormous rolls of card stock from the paper mills. For regular card manufacturing they slit the rolls to three-inch width (card height). For the SSEC they furnished rolls eight inches wide (card length). The resulting rolls weighed 400 pounds, and had to be hoisted onto the SSEC with a thoughtfully-provided chain fall! For the Stallion, we pushed the rolls up a ramp.
"The punch stations, slightly modified from standard IBM reproducer components, punched two round sprocket holes at the edges, and 78(!) regular IBM rectangular holes in between. The sprockets drove the tape one line at a time, and drives under separate program control fed the fresh or pre-punched tape under ten 78-brush reading stations. The tapes hanging down could lengthen and shorten, and for program tapes and the table lookup unit we cemented the tape end-to-end into short loops (yes, someone had had to provide the jig). There were three of these monsters at the end of the machine room.
"Up to 36 of the fixed-length tape loops could be mounted on the separate table-lookup unit, which in later years was also sometimes used for program reading. For the lunar calculation, I used 24 loops to make lookup time as short as possible, and we got programming from the main tape readers."
"The SSEC tape drive [above] consisted of a four hundred pound reel of card stock which had to be lifted with a chain hoist. It is sliced, unlike the average roll of card stock, to the length of the punch card rather than to the width. In other words, it is eighty columns wide. It passes through a punch unit in the square box at the upper left and is punched with two round sprocket holes and up to seventy-eight conventional IBM rectangular holes. One line of this punching constitutes a line of instructions or data. It could then go to as many as desired of the succeeding ten stations or loops used as subroutines. Sixteen feet high racks of wire contact relays (equivalent of what we would call core or central memory) were behind the scenes. The electronic memory in the arithmetic unit was only a few words, since the tapes were used as input and longterm memory devices." (This paragraph, also by Herb Grosch, is from a Computer Museum lecture given October 22, 1982.)
The following picture and text are from Eckert's 1948 Scientific Monthly article : "Storage for 20,0000 20-digit numbers is provided in the form of punched tapes, each line on a tape corresponding to one number (Fig.9). There are 3 punching units and 66 reading units; the punching, selection, and moving of the tapes are all controlled electromagnetically. ... Thirty tape-reading units are associated with the 3 punches, so that a number recorded by the punch may be read back later as it passes the successive reading stations. Each punch is followed by 10 reading stations, or the tape from the first may pass to all 30 reading stations. Each reader can handle either continous tape or closed loops. Thirty-six other tape readers are associated with a searching mechanism, so that any desired data may be quickly located. This unit has a capacity of 100,000 digits, and a complete search may be made in an average time of 1 second. This unit is used for consulting large mathematical tables containing such data as sines, exponentials, cube roots, etc."
|Columbia University Computing History
|Frank da Cruz / firstname.lastname@example.org
|This page created: May 2003
|Last update: 1 April 2021