Optical design has always been one of the technologies that have ridden the leading edge of computer technology. In 1951, James G. Baker, well known later in his career as, among other things, the optical designer of Polaroid’s innovative SX-70 camera, took on the role of principal investigator on what was to become a four-year program to attempt to implement optical lens optimization on the emerging set of first generation mainframe computers that came to include the Mark I, Mark II, and Mark IV generations of the Harvard Card Programmed Calculator computers that were developed by Aikens (Ref. 1). Dr. Baker summarized his goal as follows:
“…we have before us the problem of transforming the art of optical design as pursued heretofore by the author into a science of optical design, as performed automatically by calculating machines.” – James G. Baker, 1951
In the initial benchmark, the ray trace speed for the first aspheric surface was 36 minutes per ray surface (0.0005 ray surfaces per second). In 1998, that number was 1,000,000 ray surfaces per second. Even over the four years when Baker was writing, the ray trace speed increased one order of magnitude as the Mark I evolved to become the Mark IV.
According to the reports, the work was initiated in August of 1950 in Cambridge MA, at Harvard using the Harvard “Mark I Sequence Controlled Calculator.” Baker reports that as of December 24 (Christmas Eve), the group was using a “reasonably complete IBM set of machines,” including, as he notes, the supremely important Card Programmed Calculator (italics per the original report). Baker characterizes this device as a “modern tractor with ancient plowshares.” The actual computer is shown below. This computer physically resides in the Smithsonian.
This characterization is accurate in that the Harvard Mark I-IV series of computers were not in fact electronic digital computers at all; rather, they were electro-mechanical machines along the lines of Babbage’s 1833 devices, created by Howard Aiken at Harvard beginning in 1937 with funding from IBM, and in fact were driven by a 50-foot-long drive shaft. Aiken had come across the original hardware developed by Babbage in 1833 when he tried to develop (with only limited success) the first mechanical computer; the hardware was essentially lost in the archives at Harvard. In studying the machine, Aiken came upon the concept of “a switchboard on which are mounted various pieces of calculating machine apparatus.” This became the Mark I. It is worth noting that the Mark I was only 10X faster than the 1833 machine.
The final selected set of ray trace equations that were adopted based on optimal timing considerations is provided as a historical reference below. The run time for the spherical version of the equation, in 1953, was 7.6 seconds per surface.
Further Reading
1. “A Manual of Operation for the Automatic Sequence Controlled Calculator,” Volume 8 in the Charles Babbage Institute Reprint Series for the History of Computing, Tomash Publishers for MIT, a 1985 facsimile reprint of the 1946 original edition.
2. James G. Baker, “The Utilization of Automatic Calculating Machinery in the Field of Optical Design,” compiled and edited by James G. Baker, Volumes 2-12, 1952 – 1954.
If anyone is aware of a copy of this report, please respond. One copy, recently discarded by Perkin-Elmer, found its way to Dr. Baker just before he died. Does anyone know where his papers now reside? I heard that they are somewhere in Kentucky or Tennessee. One copy is known to exist at the Wright Patterson Air Force Base. No other copies are known. There were 38 copies originally printed.
For a perspective on ray trace speed and optical design, the plot below is an interesting metric proposed by Prof. Grievenkamp of Arizona. It is a plot of the number of rays surfaces per second per dollar for computers historically used in optical design. Additional data would be great to add.
Comments