That Time When Computer Memory Was Handwoven by Women

Imatge
Àmbits Temàtics

Arti­cle by by KAUS­HIK PATO­WARY

Origi­nal post avai­la­ble here

Magnetic core memory

Compu­ter tech­no­logy from yesterye­ars look comi­cally primi­tive and bulky. One popu­lar image frequently shared in social media sites show a large cupbo­ard-sized box lifted on to the cargo bay of a Pan Ameri­can Airways flight. The caption accom­pa­nying the image iden­ti­fies the box as the IBM 305 RAMAC, the world’s first commer­cial hard disk deve­lo­ped in 1957. It had a whoo­ping capa­city of only 5 megabytes.

In the early days of compu­ting, memory tech­no­logy permit­ted a capa­city of very few bytes. The first elec­tro­nic compu­ter deve­lo­ped during the Second World War to help the mili­tary calcu­late arti­llery firing tables used vacuum tubes to store data. John Pres­per Eckert then inven­ted a compli­ca­ted device using mercury-filled glass tubes and quartz crys­tals that could store up to a few hundred thou­sand bits—a vast impro­ve­ment from early memory tech­no­lo­gies.

In the late 1940s, Frede­rick W. Viehe, an amateur inven­tor from Los Angles, filed patent for a new kind of memory that used tiny trans­for­mers to store data. This was impro­ved subs­tan­ti­ally by Harvard physi­cist An Wang, and later by Jay Forres­ter and Jan A. Rajch­man in the early 1950s, leading to the deve­lop­ment of the magne­tic core memory. This new memory tech­no­logy was the first non-vola­tile memory—a memory that doesn’t lose data when it loses power— to be deve­lo­ped. It was used exten­si­vely in the US Navy’s Whirl­wind compu­ters for real-time aircraft trac­king.

Magne­tic core memory consists of tiny donuts of ferrite mate­rial strung on wires into arrays. Each donut could store a bit, and the value of the bit (zero or one) was iden­ti­fied by the direc­tion of their magne­tic flux. The wires that ran through the holes in the donuts could both detect (that is, read) and change (that is, write) the magne­ti­za­tion of the cores. Core memory became the domi­nant memory tech­no­logy during the first two deca­des of the Cold War. But manu­fac­tu­ring it was a deli­cate job. The cores were tiny and had to be thre­a­ded by steady hands using magnifying glas­ses. As cores got smaller, engi­ne­ers joked that new cores were made from the holes punched out of the previ­ous gene­ra­tion of cores.

Magnetic core memory

A core memory module with a capa­city of 128 bytes. Photo: Kons­tan­tin Lanzet/Wiki­me­dia Commons

Magnetic core memory

Close-up of a core memory module. Photo: Kons­tan­tin Lanzet/Wiki­me­dia Commons

Magnetic core memory

Diagram of a 4×4 plane of magne­tic core memory in an X/Y line coin­ci­dent-current setup. X and Y are drive lines, S is sense, Z is inhi­bit. Arrows indi­cate the direc­tion of current for writing. Diagram by Tetro­mino/Wiki­me­dia Commons

Like all crafts invol­ving weaving, sewing and other forms of textile making that has been histo­ri­cally asso­ci­a­ted with women, the job of weaving core memory was also entrus­ted to women. During the first Apollo missi­ons, the soft­ware of the Apollo Guidance Compu­ter was physi­cally weaved into a high-density storage called “core rope memory”, which was simi­lar to magne­tic core memo­ries. To build the memo­ries, NASA hired skilled women from the local textile industry as well as from the Walt­ham Watch Company, because of the preci­sion needed to work around the cores with a needle. Sitting across each other at long desks, these women passed wires back and forth through a matrix of eyelet holes, each compri­sing a magne­tic core bead. Passing a wire through the core crea­ted a “one, ” while bypas­sing the core crea­ted a “zero”.

The core rope memory was nick­na­med “LOL memory”, where LOL stood for the «Little Old Ladies» who assem­bled it. They were super­vi­sed by “rope mothers”, who were often males. But the rope mother’s boss was a woman named Marga­ret Hamil­ton.

Magnetic core memory

Marga­ret Hamil­ton was the direc­tor of the Soft­ware Engi­ne­e­ring Divi­sion of the MIT Instru­men­ta­tion Labo­ra­tory, which deve­lo­ped on-board flight soft­ware for NASA’s Apollo space program. One of Hamil­ton’s chief contri­bu­ti­ons to Apollo mission was to devise a way to deal with compu­ter errors. In the 1960s there were few forma­li­zed guide­li­nes about how to write, docu­ment, and test complex soft­ware. But the Apollo soft­ware was remar­kably error-free. What weren’t were humans.

During the landing sequence of Apollo 11, the astro­nauts inad­ver­tently left the rendez­vous radar switch on, over­lo­a­ding the compu­ter. Hamil­ton had fore­seen such an emer­gency and incor­po­ra­ted an error-detec­ting-and-correc­ting mecha­nism which allo­wed the over­lo­a­ded Lunar Module compu­ter to shed unim­por­tant tasks and focus on stee­ring the descent engine.

“If the compu­ter hadn’t recog­ni­zed this problem and taken reco­very action, ” Hamil­ton later wrote to the Direc­tor of Apollo Flight Compu­ter Program­ming, “I doubt if Apollo 11 would have been the success­ful moon landing it was.”

If you want to learn more about how magne­tic core memory works, the Nati­o­nal Science Foun­da­tion of Florida has an inter­ac­tive tuto­rial on this page.

Margaret Hamilton

Marga­ret Hamil­ton stan­ding next to the navi­ga­tion soft­ware that she and her MIT team produ­ced for the Apollo Project.

Magnetic core memory

An unknown woman strin­ging the wiring compo­nents of the Apollo Guidance Compu­ter memory.

Magnetic core memory

An unknown woman strin­ging the wiring compo­nents of the Apollo Guidance Compu­ter memory.

Magnetic core memory

A tech­ni­cian weaving the core ropes at the Rayt­heon plant in subur­ban Boston.

Magnetic core memory

Magnetic core memory

A fully wired tray A of the Apollo Guidance Compu­ter.

Magnetic core memory

A tech­ni­cian assem­bling the micro­lo­gic and core memory panels that make up the Apollo Guidance Compu­ter into their housing.

Magnetic core memory

An 8-GB microSD card on top of 8-Bytes of magne­tic-core memory. Photo: Daniel Sancho/Wiki­me­dia Commons

Refe­ren­ces: 

# Compu­ter History Museum, https://www.compu­ter­his­tory.org/revo­lu­tion/memory-storage/8/253 

# Dani­ela K. Rosner et al. Making Core Memory, https://faculty.washing­ton.edu/dkros­ner/files/CHI-2018-Core-Memory.pdf 

# Ken Shir­riff’s blog, http://www.righto.com/2019/07/soft­ware-woven-into-wire-core-rope-and.html 

# Nati­o­nal Air And Space Museum, https://airands­pace.si.edu/stories/edito­rial/rope-mother-marga­ret-hamil­ton 

# Wiki­pe­dia, https://en.wiki­pe­dia.org/wiki/Magne­tic-core_memory 

# IEEE Spec­turm, https://spec­trum.ieee.org/tech-history/space-age/soft­ware-as-hard­ware-apollos-rope-memory 

# Wiki­pe­dia, https://en.wiki­pe­dia.org/wiki/Marga­ret_Hamil­ton_(soft­ware_engi­neer)