Jim S. wrote: The delay line in the 130 and 132 store 480 'bits', but the representation is somewhat unusual. Each digit is allocated a slot of time on the delay line, and the digit is represented by zero to nine pulses within the timeslot. The 'five bit' representation is used in the four (or three on later versions of the 130/132) counters that make up the arithmetic unit of these machines. The delay line (as previously posted) is 50 ft. long (coiled up), and has a delay of approximately 5 milliseconds. Exhibit on the Friden 130: http://oldcalculatormuseum.com/friden130.html . In this exhibit there is a block diagram of the 130 that clearly shows the unique architecture of the machine. It is really quite an elegant design, the brainchild of Robert Ragen, Friden's chief electronic calculator architect. The 130 actually has a total of six "registers" that are stored on the stack. There are the four stack registers, one store/recall memory register (not displayed), and another not-displayed temporary register. I'll be posting a photo of the delay line in the Friden 130 exhibit shortly. I just got my hands on a "display" delay line that has been opened up. Just need to take a photo and post it. of I'm on the list, and an avid reader, and sometimes poster. Rick Bensene The Old Calculator Museum http://oldcalculatormuseum.com

Delay lines were used a lot in early electronic calculators from the around 1963 through the early 1970's, by which time MOS integrated circuits (first shift registers, then RAM) took over. Delay lines were used extensively in early electronic calculators by Olivetti, Canon, Friden, and Sony, among others. Typically the delay line was used to store the working registers of the calculator, with the data continuously circulating through the delay line and the arithmetic/display logic. Most of the delay lines used in calculators were of the magnetostrictive type, which used thin tapes of nickel (which contract when subjected to a magnetic field), that impart torque twists into a special wire that served as the delay medium. This wire was strung in a spiral, with lengths ranging up to around 50 feet. Magnetostrictive transducers were affixed to each end of the wire, with one end being the transmitter, and the other the receiver. Capacity of the delay line was a function of the length of the loops of wire, and the data rate pushed through the delay line. In the Friden 130, the first production calculator to use a delay line, the delay line wire is about 50 feet long, and the delay is about 5ms, with 480 "bits" of information stored in the delay line. Typically, delay line memories in calculator implementations max out at around 1500 bits of storage. Programmable calculators, like the Monroe EPIC 2000/3000, the Olivetti Programma 101 and follow-on machines, and some Canon programmables had the highest capacity delay lines since the line needed to store both the working registers of the calculator, as well as the user's stored program. In the case of the Monroe EPIC 2000/3000, two delay lines were used (housed in the same enclosure), one for working register storage, and the other for program storage. Rick Bensene The Old Calculator Museum http://oldcalculatormuseum.com

The delay line wires are well-isolated via silicone pads that damp out external vibration, and isolate the coils of wire from touching each other. Heavy-handed typing isn't going to cause any troubles, although a shock like lifting and dropping the machine (generally not a good thing for glass-based display elements like Nixie tubes or CRT's) could definitely induce errors. Ambient vibration typically wouldn't cause problems, such as use aboard a ship or vehicle, except in extreme conditions. For extreme environments, a calculator using magnetic core memory or flip-flops would have been a better choice. Rick Bensene The Old Calculator Museum http://oldcalculatormuseum.com

Rick Bensene wrote: I've done a fairly detailed RE of the memory organisation of a couple of delay-line-based calculators. To throw in some more numbers to Rick's explanation, using the Monroe 925 NIXIE desktop calculator (torsion-mode delay-line) from 1970 as an example: The delay line provides for 317.5 bits of storage, across a delay of 635 uS. It is 5 ft, 14 inches (1.88 m) in length. Bits are injected into the line as 0.5 uS pulses within a 'bit slot' of 2 uS (500 KHz bit rate). Translating these values to length shows the bit pulse to occupy 0.058 inches (1.5 mm) in a slot 0.23 inches (6 mm) long, or a bit density of 51 bits/ft (169 bits/m). Calculating the propagation velocity gives 9700 ft/S (2960 m/S) or 6600 mph (10,700 Km/H). Another point, I expect, about shock sensitivity (particularly for the torsion-mode lines, not as applicable to compression-mode), is that external shocks are generally going to produce 'common-mode' vibrations in the delay elements, and so are not going to be picked up by the differential-mode transducers.