A darkened mask-making room in a semiconductor company in Texas, a large light table glowing from beneath, on it a sheet of deep red translucent film 145 centimetres long, a plain white rectangular border strip inside its edge, within it a chip layout in bright white on dark red, the lower two thirds a dense regular field of fine vertical stripes, above and to the right irregular rectangular blocks, a hand in a pale knitted sleeve holding a slender pen-shaped cutter with a fine steel tip, the other hand with tweezers lifting a thin red strip from the clear backing, curled red scraps in a shallow glass dish.
What it was used for
Rubylith, Ulano's red masking film on a transparent backing, was cut by hand and peeled by hand. The maker Ulano's advertisements describe the sequence: outline the areas to be masked with a sharp blade without cutting through the backing sheet, lift a corner with the blade tip and peel off the film, leaving a mask that corresponds exactly to the desired pattern; the red layer is light safe, and light passes only through the cut-out portions. Its use in the mask shop of General Microelectronics in 1964 is documented: there layouts were drawn at 500 times size on quadrille-lined Mylar, the film was cut over it along the rails of a coordinatograph and peeled off. For Intel, a Computer History Museum photograph of about 1970 shows operators cutting, with the caption that the film was then optically reduced.
What was new
Knife-cut peelable stencil film is older than the chip industry: patent 1,781,834 of D'Autremont, filed 6 May 1929 and granted 18 November 1930, describes such a sheet for multi-coloured designs, and Ulano's own later application cites it as prior art. What Ulano's specification, filed 21 May 1948 and granted 27 November 1951, describes as the predecessor's defect is dimensional stability: the paper carrier expanded and contracted by as much as 3/8 of an inch in the sizes ordinarily used, which interfered with proper registry as soon as more than one screen was needed - and registry across layers is the precondition for chip masks. The same specification already describes the photographic use. The barb: the hand work was not ended at once by the computer but carried on alongside it - Calma was founded in 1964, brought out the Graphic Data Station in 1971 and GDS II in 1978, and Steven Leibson relays Dan Clein's recollection that in 1984 Calma systems at Motorola were still digitizing layers drawn in coloured pencil on Mylar.
What the sources leave open: The time taken and the size of a hand-cut layout are not documented; the anecdotes about the 6502 appear only in secondary sources. Whether the Ulano specification is the Rubylith specification is not documented: it does not name the trade mark, and its carrier is vinyl resin while the advertisements name polyester. The documented examples are General Microelectronics in 1964 and Intel around 1970; no error rate for hand-cut masks was found.