The MADGRAPH system (1977-1979)

In 1977 John McAlister and I were graduate students in the lab of M Sundaralingam at the University of Wisconsin, Madison. Sunda could not find money to buy a state-of-the-art graphics system from Evans and Sutherland, so instead he had purchased a Vector General. It was a rather fancy VG, however, and shared memory with the host computer, a PDP-11/35. We were very proud of the fact that our PDP could do floating point math in hardware, and that it had 5 MBytes of disk space! Unfortunately, at that time there was not a general purpose molecular graphics package available for this hardware, so John and I spent a year or two writing one. These screen shots were taken for a talk I gave at the 1979 ACA meeting in Boston describing the MADGRAPH molecular graphics system.

The pictures were shot from the VG screen using a 35mm camera loaded with black-and-white film. The negatives were mounted for direct projection, yielding a black on white image as shown above in the title slide for the MADGRAPH talk. I have inverted the colors of the other images after scanning them in, so that they give a better idea of what the VG screen itself looked like. These negatives were scanned in 2002 after sitting for 25 years in a binder on my bookshelf, and they are somewhat the worse for it. The distortion you see in the shots below is due mostly to the camera lens, but also due in part to the nature of the vector display tube.

The Vector General had a calligraphic display, which means that it drew each line segment of the figure one by one, deflecting the electron gun to trace out each segment in succession. The more line segments to be displayed, the longer it took to refresh the display screen - so complicated images had a tendency to flicker. You had to build up the image to be displayed from hardware-supported primitive objects consisting of lines, circles, arcs, and alphanumerics. The MADGRAPH machine was a VG series 3 (DEC3R), which had a special bank of memory called a refresh buffer that was shared with the host computer. The VG had hardware to apply a rotation/translation matrix to objects as they were displayed, but in order to nest levels of rotation (as in torsional rotation about a bond) or to change the overall viewing angle required intervention by the host CPU. That meant that the bulk of the graphics manipulation was done by a driver layer running on the host computer, analogous to the operation of a cheap "winmodem" or "winprinter" in today's PC world in which processor cycles are stolen from the host CPU to supplement a minimal hardware implementation.

The MADGRAPH software could display crystallographic models and electron density maps. It allowed for atom selection and torsion angle rotation. It was never fully developed for work with protein structures, however, because Sunda's laboratory was at the time focused on tRNA and on small molecule work with analogues and complexes of nucleic acids.

Vector General hardware sank into history without a trace, the 8-track tape of molecular graphics. MADGRAPH lost out to Alwyn Jones' program Frodo, which was developed on similar hardware at roughly the same time but was soon ported to the more capable display systems sold by Evans and Sutherland. MADGRAPH's steamboat logo lived on for a while in some of the programs John McAlister later developed for TRIPOS.