My first computer.

In 1965-1966, computer access was hard to come by. So I built my own.

	Front view Tape reader, version 2, at top. Control unit. Memory (8 words of 7 bits each) Arithmetic unit Rear view Early power supply (inadequate) at top. Tape reader Control unit Memory, with cover on Strowger switch. Arithmetic unit.
	Memory 8 words of 7 bits. Those are long-coil telephone relays. One of four banks are uncovered. Behind the relays is a Strowger two-dimensional stepping switch, the memory accessing device.
	Arithmetic unit Add, subtract, and shift. Data transfer to and from other units is bit-serial, through a stepping switch hidden by the support brackets.
	Paper tape reader and control unit At top, the control logic for the paper tape reader. At bottom, the control unit. The large trapezoidial object on the control unit is an up-down counter removed from a junked jukebox coin mechanism. It was used as the relative branch controller, indicating how many instructions on the paper tape were to be skipped. The filter capactitor and tall relay were the system clock for data transfers between units. About 1.5 Hz.
	Tape reader early version. Roller skate wheels above contacts. Wasn't reliable. The tape was 4 data bits and a clock track. Tape input could be simulated manually using the buttons at the left. Data read from tape appeared in the lamps at left. At the bottom is the front of the control unit, with big Start. Stop, and Reset buttons.
Not shown: output typewriter (an old manual typewriter, equipped with solenoids for zero, one, space, and CR only) and power supply.

The machine had, as I recall, 14 instructions. These were expressed as four bits on the tape, plus a clock track, which appeared between the rows of data bits. The tape was the program, and was cranked back and forth by two Erector set motors with solenoids on the gearshifts.

The tape would be threaded and attached firmly to the reels, and the RESET and START buttons pushed. This started the tape forward. The four data tracks would be read and latched into a 4-bit instruction register built out of telephone wire relays. When the clock track was read, the control unit executed the instruction specified by the four bits and cleared the instruction register.

For simple instructions, the tape did not stop. But for long instructions, the control unit stopped the tape drive by shifting the takeup reel motor into neutral, and started the 1.5Hz data transfer clock. Long instructions included add, subtract, store, and print. Bits were transferred serially between the arithmetic unit and the memory or printer, at 1.5Hz.

Branching was provided. Branching was conditional or unconditional, and forwards or backwards. Branching was performed by specifying the number of "mark" instructions to be skipped in the forward or backwards direction. The "branch" instruction itself caused the control unit to ignore all instructions other than "mark", while winding the tape forwards or backwards as required, counting down the number of marks to be ignored. When the number of marks to ignore reached zero, normal operation was resumed. The large trapezoidial object in the control unit, an up-down counter salvaged from a jukebox, controlled branching. Conditional branching was limited to "branch if accumulator negative", which is sufficient.

The big limitation was the paper tape reader, built to use adding machine tape punched with a hand punch. Three versions were built. The first was photoelectric, using cadmium sulfide photocells and "Little Jewel" high-sensitivity R/C radio relays. The problem was that the pull-in and release currents for the relays were too far apart, and while they'd detect a hole, they wouldn't drop out after the hole passed. Three decades later, it occurs to me that all I needed to do was make the tape black using black ink and a roller.

The second tape reader was mechanical, and is shown above. Reliable contact couldn't be achieved with metal strips against roller skate wheels. Finally, with time running out for the science fair, a brute-force solution was tried - vinyl seat cover material in place of paper tape, and staples to store the information. This worked, but limited program size severely, because the tape was so thick and filled the reels quickly.

Most of the parts came from surplus stores in the Washington, D.C. area. A few parts, including the big Strowger stepping switch, came from national surplus houses, such as Fair Radio Sales or Herbach and Rademan, both of which are still in business.

In retrospect, this was far too much work for what it did.

J. Nagle
Revised September 1, 2002