Talk:Delay-line memory

Delay Line Computers
Removed the sentence:
 * "The only commercial delay line computer ever built (the IBM 701), used steel wire, formed into a shapes resembling a long bed-spring."

The IBM 701 used Williams tube memory (and later core memory), not delay lines. Also the UNIVAC I, a commercial computer used mercury delay line memory. -- RTC 23:20 Mar 18, 2003 (UTC)

If that was inaccurate you should have removed the IBM 701 part, not the mention of wire delay lines; those really existed. PML.

I removed the sentence because it was contrasting "experimental" computers to "commercial" computers and claimed that there was only one delay line based commercial machine. Initially I simply removed the 701 part, but decided that the "only one" part had to be wrong too. I am sure there were other delay line based commercial computers! Some of them probably did use wire. -- RTC 23:30 Mar 18, 2003 (UTC)


 * BTW, thanks for blending the "wire" part into the rest of the article! -- RTC 23:31 Mar 18, 2003 (UTC)

Will someone please explain what the picture is in the first link at the bottom? What are all the parts, etc? Would help to visalize what is going on...

__ According to the EDSAC website, the first machine had 512 words of 17-bits each (the 18th bit didn't work due to timing problems.) Two years later the bank was doubled. Where did the information on this page come from ? --Simsong 20:58, 30 July 2006 (UTC)

Moon Reflector Delay Line
I cannot find any source regarding using the moon reflector for delay line memory, and it's been uncited for nearly a year. I've also had no luck tracking down any former Prof. William Wolf at the University of Colorado, and it also seems as though the first THz pulse rate lasers were not around until the 90s. As such I'm removing the section.


 * Maybe you should dig deeper? William L. Wolfe, Optics made clear: the nature of light and how we use it --Avesus (talk) 10:37, 9 July 2012 (UTC)

Removed this section again. Basic math says the claimed data storage is impossible. The fastest transmission rate for a single laser as of 2012 was ~3Gigabit/s. Even without counting the phenominal packet loss due to laser dispersion that we see even in the most advanced lunar lasers, that's < 1GB data storage. — Preceding unsigned comment added by 89.26.19.1 (talk) 22:15, 21 August 2012 (UTC)

Magnetostrictive delay lines
The text currently explains magnetostrictive delay lines as follows:
 * ... small pieces of a magnetostrictive material, typically nickel, were attached to either side of the end of the wire ...

This is not how I understand magnetostrictive delay lines. The ones I remember helping to build in the early 1970s used nickel wire (actually, flat nickel ribbon, about 1mm wide) for the full length of the delay line. The transducers at the ends were coils around the wire. One coil as a "write head" and the other a "read head". A reversal of the current in the write head strong enough to reverse the magnetization of the wire would induce a magnetostrictive pulse that traveled down the wire at the speed of sound. This pulse was a domain wall, and as the pulse passed the read head, it would induce an electrical pulse. No mechanical motion of the wire was involved. The fact that the speed of sound in the wire depended on temperature was a problem -- fixing the delay required careful temperature regulation, and when you first turned on the delay line, it would function incorrectly until it warmed up to operating temperature. I strongly suspect that the loose coil of wire shown in the illustration worked the same way -- although it could be a purely mechanical delay line with mechanical transducers at each end. In that case, piezo-electric or voice-coil transducers would be very likely to work better than some kind of electromechanical magnetostrictive actuators. Douglas W. Jones (talk) 18:53, 8 July 2013 (UTC)

Mechanical delay lines
I suspect the illustration of a torsion delay line is not magnetostrictive but purely mechanical. This article should be edited to include mechanical delay lines, a topic that goes back long before the earliest radar-related delay lines discussed here. Henry C. Harrison received two patents on mechanical delay lines: U.S. Patent 1,678,116, granted in 1928, and U.S. Patent 1,788,519, granted in 1931. He documented a variety of uses for these in telephony and sound recording. Then, there is Hammond's spring reverberator, U.S. Patent 2,230,836, an application of delay lines that Harrison didn't seem to anticipate (he came close). Douglas W. Jones (talk) 19:41, 8 July 2013 (UTC)

From Russia with Delay Love.
Some IT-history literature claims soviet-russians were highly influential in the development of acoustic delay lines for earliest computer use. In the immediate aftermath of Stalin's death in 1953, computer science was declared an "imperialist faux-science" for political purposes and supressed for many years within the eastern bloc. This is why their contributions are not well-known nowadays. But maybe something about them should be included in this article? 92.52.252.253 (talk) 18:07, 14 February 2015 (UTC)
 * No. This article is about delay line memory, not Soviet politics. --Wtshymanski (talk) 19:22, 16 February 2015 (UTC)

Analogue delay line
Should mention be made of reverb boxes that had a spring that had one end shaken and a pickup on the other end. The reveb was the primary goal but a delay would also have taken place

Idyllic press (talk) 10:12, 5 May 2015 (UTC)

How much memory did the EDSAC Have?
The booklet at Cambridge says that each mercury delay line stored just 35 bits, not 576 bits. http://www.cl.cam.ac.uk/events/EDSAC99/booklet.pdf — Preceding unsigned comment added by Simsong (talk • contribs) 16:55, 5 August 2017 (UTC)


 * In reality one delay line (for main memory) stored only 560 (usable) bits (see reference in article near EDSAC in article: Wilkes, M. V., and W. Renwick. "An Ultrasonic Memory Unit." Electronic Engineering 20 (1948), pages 209-210).
 * That booklet describes short delay line tube (picture on last page of referenced article above, or here): "Short tanks which can hold one number only are used for the accumulator and multiplier registers in the arithmetical unit, and for control purposes in various parts of the machine." (, p. 61). --MarMi wiki (talk) 03:04, 20 December 2018 (UTC)

Generalized paragraph description is under a specialized subheading
Under the 'Magnetostrictive delay lines' subheading, it states in the last paragraph:

"Delay line memory was far less expensive and far more reliable per bit than flip-flops made from tubes, and yet far faster than a latching relay. It was used right into the late 1960s, notably on commercial machines like the LEO I, Highgate Wood Telephone Exchange, various Ferranti machines, and the IBM 2848 Display Control. ..."

It seems this is a general description of Delay line memory-type systems, yet this paragraph is under a 'Magnetostrictive delay lines' heading, rather than directly under the main title or a more general heading. If the description only fits Magnetostrictive delay lines only, maybe the paragraph should start with "Magnetostrictive delay line memory", and/or contrast it's reliability and cost with other forms of delay line memory (if it indeed is noticeably cheaper or more reliable)? I don't have actual knowledge about this, so this placement confused me a little --Eankeen (talk) 21:44, 24 June 2021 (UTC)

Optical delay lines?
Hi, incidentally an invention not often mentioned is the "phosphorescence delay" aka persistence memory. This works by using a rotary cylinder or disc coated with a ZnS:Cu or similar phosphorescent powder, freely rotating within a pair of bearings or using pager or Zip drive motors for a miniature version. The delay is achieved by using a sequence of emitters and sensors with analogue switch ICs ie 4066 to select a given channel and shunt data from one to the next, and an infrared emitter to "wipe" the memory when writing new data, relying primarily on controlled 78.111.199.23 (talk) 11:08, 26 December 2021 (UTC)temperature to ensure data integrity. A passive version is possible but very noisy and prone to unwanted feedback loops. The delay in this configuration can be quite long (tens of seconds) and only really limited by the specific emitters and sensors used, so for blue LEDs the device capacitance becomes a factor. It appears that this was actually considered by GC&CS (now GCHQ) for cracking Lorenz back in the day but discarded in favour of flying capacitor or mercury acoustic delay based memory as not practical at the time due to the lack of a fast enough gas discharge lamp or signal post processing components. Hence now safe to discuss as >75 years post.