Home›Articles›SIGSALY: How a 55-ton secret phone made the vocoder bef…
Articles

SIGSALY: How a 55-ton secret phone made the vocoder before music got it

SIGSALY: The vocoder is a staple of pop, but it began as a military secret that filled a basement. SIGSALY was a Second World War system for scrambling…

Bernardo de Azevedo
Bernardo de Azevedo
· Updated
SIGSALY - historic photo

The vocoder is a staple of pop, but it began as a military secret that filled a basement. SIGSALY was a Second World War system for scrambling telephone calls between Allied leaders, and its core idea was to turn the human voice into numbers and put it back together at the other end.

It is often remembered through the image of Churchill and Roosevelt chatting on a hidden line. That version is only partly true, and the corrected account is more interesting, because it shows how much of the system’s life was about engineering rather than famous voices.

Table of Contents

  • SIGSALY: a secure voice built from a voice coder
  • The scrambler that failed
  • Homer Dudley’s idea
  • Vinyl records, Muzak and a one-time key
  • A machine that needed a basement
  • Churchill, Roosevelt and what the legend gets wrong
  • From secrecy to the synthesizer
  • Final perspective

SIGSALY: a secure voice built from a voice coder

SIGSALY was a secure speech system in service from 1943 to 1946 for the highest-level Allied communications, according to Wikipedia. It was also called Project X, Ciphony I and the Green Hornet. The name was a cover name rather than an acronym. Bell Labs developed the system under A. B. Clark, assisted by Alan Turing, and the US Army awarded a contract for two systems in 1942.

Its job was to protect conversation, not music, but the way it did so is the same technique musicians later borrowed. A vocoder analysed speech every 20 milliseconds, extracting band energy levels, whether sounds were voiced or unvoiced, and pitch. Those values were encrypted with a random key and sent as tones, and the receiver reversed the process to resynthesize the voice. The result was a voice that had been taken apart, scrambled and rebuilt.

The scrambler that failed

SIGSALY replaced an earlier device. Wikipedia says the A-3 voice scrambler could be intercepted and broken by the Germans, and 99% Invisible adds that German codebreakers could easily decode Allied leaders’ conversations on it. The Army wanted what the Quietus calls “indestructible speech”: a system that would resist code-breaking.

Workers reviewed more than 80 patents, according to the same article, and chose Homer Dudley’s vocoder as the system’s basis. That selection mattered, since the vocoder was already a decade old when the war started. Dudley, who worked at Bell Labs for about 40 years, began experimenting with speech-producing devices in 1928, and his work on parallel band-pass filters led to the vocoder patent, according to Wikipedia.

Homer Dudley’s idea

The vocoder, or voice encoder, did not transmit sound as such. As Red Bull Music Academy puts it, it analysed speech and sent coded parameters instead of the sound itself, to save bandwidth. In its early form the result sounded robotic and had pitch problems.

Dudley also built a showpiece, the Voder, with Robert Riesz. It was a console played by trained operators that synthesized speech live, and Wikipedia says it was shown at the 1939 New York World’s Fair and the 1939 Golden Gate International Exposition. 99% Invisible notes that its operators, called “Voderettes”, trained for up to a year. The Quietus describes SIGSALY’s own vocoder as a twelve-channel design, with ten channels measuring voice power across the speech band and two tracking pitch and unvoiced hiss.

Dudley was born in Virginia in 1896 and died in 1980, and was awarded the Franklin Institute’s Stuart Ballantine Medal in 1965, according to Wikipedia. Later in life he designed a hobbyist kit, “Speech Synthesis: an Experiment in Electronic Speech Production”, which entered production in 1963. The war system itself is credited with several engineering firsts, including enciphered telephony, speech transmitted by pulse-code modulation and multilevel frequency-shift keying.

Vinyl records, Muzak and a one-time key

The encryption was the unusual part. According to Wikipedia, the key noise came from mercury-vapor rectifier tubes and was stored on phonograph records. The Quietus says the records were produced by the Muzak Corporation and that the noise was sampled every 20 milliseconds, quantized into six levels and recorded as frequency-shift tones on a master. Each call used a pair of identical vinyl records, one at each end, and 99% Invisible adds that records were destroyed after each call.

Synchronization was the hard part. Operators listened for a click at an agreed time and started the turntables together, using fine adjustments to hold them in step. If the systems drifted apart, only garbled speech came through. The Quietus notes that the two-turntable setup has been compared to DJing, an argument made in Dave Tompkins’s book How to Wreck a Nice Beach. Around 3,000 conferences required at least 6,000 pressed records, per 99% Invisible.

A machine that needed a basement

Each terminal was enormous. The Finest Hour article in the Churchill Centre’s journal describes roughly forty racks of vacuum-tube equipment weighing about 55 tons, occupying 2,500 square feet and using 30,000 watts, and notes that each terminal was air-conditioned. Wikipedia says the figure was over 50 tons. The 805th Signal Service Company installed and maintained the machines.

The London terminal arrived on HMT Queen Elizabeth in May 1943. Because it was too large for the Downing Street Annexe, Finest Hour says, it was installed in the basement of Selfridge’s Annexe on Oxford Street, about a mile away, and linked by an underground cable. The first terminal was in the Pentagon.

About a dozen terminals were eventually installed, and the system supported roughly 3,000 high-level conferences. Service began in mid-July 1943, and Wikipedia says that after the invasion of Germany, investigators found that the Germans had recorded its traffic but mistaken it for a complex telegraphic code.

Churchill, Roosevelt and what the legend gets wrong

The popular image is of Churchill and Roosevelt using SIGSALY to talk freely. The Finest Hour article, by Christopher Sterling, complicates it. Churchill apparently did not use SIGSALY until April 1944, and then only occasionally, since he preferred the older scrambler. Roosevelt chose not to install the main Washington terminal in the White House, perhaps because he knew Churchill’s working hours and did not want to be put on the spot without time to consider an issue.

The longest call came afterwards. Churchill’s first SIGSALY call after Roosevelt’s death in April 1945 was to Truman, and two weeks later they held a two-hour discussion about tentative German surrender offers, the longest SIGSALY call ever made. Most of the traffic was between military commanders rather than heads of government, and fewer than twenty-five senior British leaders were cleared to use the London terminal.

The first conference, as Wikipedia and the Quietus describe it, linked the Pentagon with London on 15 July 1943. Alan Turing, who had to be allowed to see the device after high-level negotiations, reported his evaluation to Churchill.

From secrecy to the synthesizer

The system stayed secret for roughly three more decades after the war. Finest Hour says it became public in news accounts in mid-1976, when more than thirty Bell Labs patents were granted, and 99% Invisible says most information was declassified in the mid-1970s after AT&T sued the US military under the Freedom of Information Act. The terminals themselves were scrapped after the war.

By then the technology had begun a second life in music. Red Bull Music Academy traces a path from Werner Meyer-Eppler, who promoted the vocoder for electronic music in Detmold in 1949, through Sennheiser and Siemens machines in the 1950s, to Wendy Carlos’s work on A Clockwork Orange and Kraftwerk’s Autobahn in 1975. It reports that by 1978 the Korg VC-10 cost about $500, compared with around $10,000 for earlier Sennheiser units, which made vocoders far more accessible.

Final perspective

SIGSALY teaches a lesson that the music industry sometimes forgets: a sound can be a by-product. The vocoder voice we now hear on pop records exists because engineers once needed to protect a conversation, and a robotic buzz was an acceptable price for secrecy. Nobody involved was trying to design an instrument.

The system also shows how little a good story needs in order to spread. The Churchill and Roosevelt image is vivid, but the surviving evidence points to a quieter reality of military conferences, two synchronized turntables and a maintenance routine so heavy that, per the Quietus, every eight hours of operation required sixteen hours of calibration. When the vocoder next appears on a chorus, it is worth remembering that it comes from a machine built to make speech impossible to understand for anyone who was not meant to hear it.

Related stories

Sources

Bernardo de Azevedo
Written by
Bernardo de AzevedoFounder & Editor

His connection with music began at age 6, in the 1980s, when his father introduced him to Jean-Michel Jarre's Rendez-Vous on vinyl. He works professionally in the legal field, while synthesizers became his space for abstraction and creative exploration. Founder of The Synth Source.