Click for homepage
SIGSALY   Ciphony 1
Digital voice encryption with OTP - not in collection

SIGSALY was a digital speech encryption system, developed by Bell Telephone Laboratories (BTL) 1 in the US in 1941/1942, and built by Western Electric in New York (US) in 1943. The system went into service in April 1943, just two months before the invasion of Italy, and was used until at least 1946. SIGSALY was used heavily during WWII, in particular for confidential talks between British Prime Minister Winston Churchill and US President Roosevelt. The system used the highly-secure One-Time Pad (OTP) encryption and is known under various names, including The Green Hornet.

SIGSALY featured a number of innovative digital communications concepts, including the first transmission of pulse-code modulation (PCM). 2

SIGSALY was completely built with vacuum tubes (valves). A single system consisted of more than 30 full-height 19" racks, plus 4 synchronisable turn­tables. It weighted 50,000 kg, consumed 30 kW of power, and had special air conditioning requirements. A single SIGSALY terminal had a price tage of US$ 1 million in 1943. In total, 12 SIGSALY terminals were set up around the world, the first of which was installed at the Pentagon.

The second one was installed in London in a basement of Selfridges department store on Oxford Street. Another one was installed on a ship that followed General Douglas MacArthur during his South Pacific campaigns. The systems were installed and maintained by special vetted members of the 805th Signal Service Company of the US Army Signal Corps. Between 1943 and 1946, the 12 SIGSALY terminals conveyed more than 3000 high-level telephone conversations worldwide.

SIGSALY provided a full-duplex voice link via narrow-band HF radio channels in the Short Wave (SW) radio bands. Each half of the link used 12 individual data channels, or carriers, through which the data was sent by means of digital 6-level Multiple Frequency-Shift Keying (MFSK).

The human speech was analysed just 50 times per second (at 20 ms intervals), broken down into its characteristic parts, and then coded and sent across the atlantic. At the receiving end, the data was decoded and used to reconstruct, or synthesize, the original human speech again.

This resulted in a low data rate (comparable to 1500 baud today) but made it very difficult to recognise the person at the other end. Ultimately, this technology evolved into more advanced speech-coding techniques, such as LPC-10, CELP and MRELP in equipment like the STU-I.

Once the system was ready, the BTL developers spent the rest of the war working on SIGSALY's successor, known as Junior X or AN/GSQ-3, that occupied 'just' six 19" racks and could be fitted inside a movable van. Unfortunately, it was not completed in time to see any active service [9]. After the war, SIGSALY was taken out of service as part of the demobilisation process and most of it was destroyed, including some of its documentation. Rumour hast it that the large racks were dumped into the sea after SIGSALY had been decommissioned in 1946.

  1. Bell is also known as AT&T Bell Laboratories, Bell Telephone Laboratories (BTL), Bell Labs and just Bell. It is currently owned by the Finnish company Nokia and is known as Nokia Bell Labs.  More
  2. In this context, the term Pulse-Code Modulation (PCM) is used to describe the process of converting analogue signals into time-related numerical values, also known as sampling or quantizing.  More

SIGSALY used an unbreakable encryption scheme that was based on the so-called One-Time Pad. The principle of this scheme is that the human voice is first digitized and then mixed with an element from a previously generated random key-stream. When correctly applied, this system is unbreakable. The downside is that both sides need to have sufficient supply of key material.

With SIGSALY this was solved by recording the random key stream onto phonogram records or discs. Only two copies of the records were made, and one copy was sent to each end of the link by special courier. As each record could only hold 12 minutes of key material, a large quantity of key discs had to be distributed, especially if the system was used for very long conversations.

In addition to the key distribution problem, there was the problem of synchronisation. The fact that digitally sampled and coded data was sent over fading narrowband Short Wave (SW) radio channels, implied the use of high-precision frequency standards at both ends of the link, which had to remain synchronised during the conversation, even if the radio signal was lost temporarily. For the same reason, the turntables also had to be very accurate and very stable.

As each record could only hold 12 minutes of key material, two coupled turntables were used, allowing the next disc to be cued up whilst the current one was playing. The image above shows two synchronised turntables in an early SIGSALY setup. As SIGSALY was a full duplex system, four turntables were present at each terminal: two for the transmitter and two for the receiver. Initially vinyl records were used in 1943. They were identified by the codename SIGGRUV. This was later changed to acetate-coated aluminium records, which were identified by the codename SIGJINGS.

In today's terminology, the SIGGRUV and SIGJINGS records would probably be called OTP records or One-Time Records (OTR). Note that the records were destroyed at both sides immediately after use, so that a malicious party would never be able to decipher any part of the secret conversation.

 Technical description

The Washington terminal
The first SIGSALY terminal to be completed at the Western Electric facilities in New York City, was installed in Washington at the Pentagon, the new headquarters of the US Department of Defense that had just been dedicated in January 1943. The White House had briefly been considered as a possible location, but the installation was simply too large to be fitted anywhere in the building.

Instead, it was installed at the Pentagon and was connected to the White House via an extension line. After installation by BTL and WE personnel, maintenance was taken over by the 805th Signal Service Company of the US Army Signal Corps.

All personnel of the 805th was hand-picked and was trained by BTL staff in New York City, in a special school that had been established for the occasion. In 1944 both the 805th and the school were moved to the Pentagon in Washington. By that time, 193 officers and men had been trained on the use of SIGSALY, and more were to follow.

The 805th would eventually consist of 356 people: 81 officers and 275 men, who were divided over twelve detachments; one for each SIGSALY terminal. A single detachment consisted of 5 officers and 10 enlisted men, and was expected to operate the terminal on a 24-hour basis [6].

The London terminal
In the UK, the SIGSALY terminal was placed in the SWOD basement 1 of the Selfridges department store on Oxford Street (London), 60 metres below street level. The first conference took place on 15 July 1943, and the British Prime Minister – Winston Churchill – was one of its many users [4].

The image on the right shows the main building of Selfridges department store on Oxford Street in 1929, ten years before the outbreak of World War II. At the time, Somerset Street ran directly behind the main Selfridges building. At the other side of Somerset Street was an annex building of Selfridges, known as the SWOD, named after the four roads that enclosed it: Somerset, Wigmore, Orchard and Duke (the latter three still exist).

Below the SWOD annex was a basement, known as the sub, and below that a sub-sub basement that descends 60 metres below street level [4].

In 1942, after the US had entered WWII, the dry-sub-sub SWOD basement was used by the United States Army, as the site was safe from bombing. Here they installed one of the few secure telex lines and in April 1943 also the SIGSALY terminal, which was inaugurated on 15 July of that year. The discussions between the two nations were probably about the Allied Invasion of Sicily earlier that week 2 and the fortcoming invasion of Italy that was about to take place a few months later.

The map above shows that Selfridges' basement was very close to the US Embassy. 3 Even today there are persistent rumours of a tunnel between the two locations [2]. Initially, SIGSALY users had to come to Selfridges for a secure conversation with the US president, but the terminal was later connected via extension lines to the US Embassy on 1 Grosvenor Square, and to the offices of the British Prime Minister Winston Churchill at 10 Downing Street and the Cabinet War Rooms. Churchill wanted to be able to contact the US President any time of the day or night if necessary.

  1. SWOD is the abbreviation of Somerset, Wigmore, Orchard and Duke; the four streets that once enclosed the rear building of Selfridges department store, under which the sub-sub basement is located.  Wikipedia Somerset Street no longer exists and has been 'swallowed up' by the expanded Selfridges store, whilst a new road, Edwards Mews, has since been established further north, parallel to Wigmore Street. The SWOD building is now embedded in the main store building.
  2. The Allied Invasion of Sicily took place on 9/10 July 1943 and ended on 17 August [14]. It was followed by the Allied Invasion of mainland Italy on 3 September 1943 [15].  Wikipedia
  3. Between 1938 and 1960 the US Embassy was located at 1 Grosvenor Square, opposite the later US Embassy building at 24 Grosvenor Square. In 2017 the US Embassy was relocated to the Nine Elms area [5].

Extension lines   OPEPS
Because of the size of a SIGSALY terminal, the system was generally not located at the office where it was needed, but rather in a larger facility somewhere nearby. In such cases, the offices were connected to the SIGSALY terminal via special protected extension lines, known as OPEPS.

OPEPS [17] is short for Off-Premises Extension Privacy System. In Washington, two OPEPS lines were installed from the Pentagon to the White House and to the Navy Department building on Constitution Avenue. London even had three such extensions: one to the US Embassy, one to the Prime Minister's house at 10 Downing Street, and the third one to the Cabinet War Rooms.

As these extension lines carried high-level ultra secret traffic, they had to be protected against tapping in the best possible ways, which was done by means of a variety of safety measures.

The cables between SIGSALY and its extensions were piped, and were protected by gas pressure and microswitches. Any tampering would cause the gas pressure to drop and raise an alarm. In addition, a strong noise source was superimposed on the twisted telephone wire pairs by means of a bridge transformer. In normal operation, the signals on the two wires would balance out.

As soon as the wires were tapped, even within the building, this would cause an unbalance on the line and set off the alarm immediately. At the same time the user would hear a strong noise through the handset, indicating an unsafe line.

The OPEPS extension to the SIGSALY terminal should not be confused with the UK's Frequency Changer, commonly called scrambler phone by Churchill, that connected the various locations of the War Department in London during the war. Although the two phones look similar and served similar purposes, the scrambler was not secure.

It was based on the inversion of the voice frequency spectrum and was merely used as protection against an occasional eavesdropper – such as an operator in a telephone exchange – rather than against a professional interceptor. It was therefore called a privacy set rather than a secrecy set.

 More about the UK's scrambler phone

At the start of WWII, the US Military relied on Western Electric's A-3 Voice Scrambler for its high-level sensitive communications. Voice Scramblers were used by both sides during the war. It was known however, that they were by no means safe, and each party is known to have intercepted and broken the other party's communications 1 with simple means like an ordinary oscilloscope.

The problem was recognised by Bell Telephone Laboratories (BTL) as early as October 1940, where it was decided to start the development of a high-level unbreakable transatlantic voice telephony system under the name Project X.

At Bell, the research was carried out under the direction of A.B. Clark 2 by two teams: one to do the basic research and the other to handle the practical problems of design, construction and instruction. The basic research was handled by the Transmission Research group under R.C. Mathes, which included earlier work by Eugene Peterson and Homer Dudley on the Vocoder [10].

Dudley's Vocoder, named Voder 3 , had been demonstrated by BTL at the New York World's Fair in 1939 and proved that human voice could be synthesized by immitating the effects of the human vocal tract. The Vocoder offered a compression ratio of 10:1 and delivered a digital signal in the telegraph range that could be transmitted over short-wave (SW) radio channels. Furthermore, it allowed digital encryption with a system that was similar to the one developed by Gilbert Vernam.

But before a completely secure transatlantic telephone system would see the light of day, a large number of technical and mathematical problems had to be solved. After a lot of testing and experimenting, it was decided to divide the audio spectrum (150 Hz - 2950 Hz) into 10 spectral bands or channels, each of which was digitized or quantized into six discrete values.

In the process of developing the system and solving its problems, many new inventions were made and many (secret) patents were filed, some of which were not disclosed until 35 years later.

By late 1941, the designs of the individual parts were more or less ready and breadboard testing had been carried out. The next stage was to build up a prototype, which was done by the research group under A. M. Curtis. As the individual parts were finished, the complete system was assembled on the 12th floor of the Graybar-Varick building in New York City (shown below).

The actual manufacturing of the production machines was carried out by Western Electric (WE). As part of the process, BTL engineers converted much of the original design into standard WE parts that were readily available.

Although this conversion added to the overall size of the system, it significantly reduced the manufacturing time. The machines were built at WE's vacuum tube plant on Hudson Street in New York City, which was conveniently close to West Street and the Graybar-Varick building, which is shown in the image on the right. On the 12th floor the first X System was assembled in 1942.

The X System was made up of 12 parallel data channels, 10 for the spectrum data and 2 for the pitch, with were nearly identical, except for their position in the FSK arrangement. In March 1942, one complete channel was ready for review. It was tested with an artificial fader that simulated the fading of transatlantic radio. After successfully completing the tests, production of the other channels was started in April 1942. The experimental model was completed by late August 1942.

In November of that year the system was first tested over transatlantic radio, against a set of synthetic signals from a generator that had previously been sent to England. This allowed the system to be fine-tuned and improved.

The prototype had meanwhile been nicknamed The Green Hornet, after a popular radio show 4 of the 1930s with the same name. The show's tune resembled the buzzing sound that would be heard by someone eavesdropping on the system's coded signals. It was also heard by the exchange personnel that had to patch the lines.

Up to this point, all development work had been carried out on the initiative of BTL, although the National Defense Research Committee (NDRC) and the Signal Corps were aware of the work and had expressed an interest in the program as early as 1942. Around the time of completion of the experimental model, the Signal Corps decided to sponsor the building of several terminals, which started in September 1942. From this moment onwards, the project was codenamed SIGSALY.

Once the individual parts had been completed and tested at Western Electric, they were sent to Room L30 at West Street, the former sound movie laboratory, where they were assembled into a complete system. The first system was completed on 1 April 1943, and a second one shortly thereafter, allowing the first real test.

By the end of April 1943, just two months before the invasion of Italy [15], the first terminals were installed in Washington (Pentagon), London and North Africa. In June/July 1943 more terminals followed for the remainder of the war theatre.

SIGSALY was officially inaugurated on 15 July 1943, shortly after the Invasion of Sicily [14]. The photograph above was taken at the Pentagon terminal on that day. Present at the meeting were Lt. Gen. T. McNarney, deputy chief of staff USA (sitting at the head of the table at the right), Dr. O. E. Buckly, president of BTL (at the head of the table on the left) and Lt. Gen. Brehon Somervell, commanding general, operating the phone at the right, whilst the others listen on headsets.

At the London end, the conference was attended by Lt. Gen. J.L. Devers, US commanding general of the European Theatre of Operations, Maj. Gen. I.H. Edwards, US chief of staff of the European Theatre of Operations, and others. As part of the inauguration, Dr. Oliver Buckley, the president of BTL, held a brief speech for the participants, in which he predicted far-reaching effects.

  1. In the fall of 1941, the Deutsche Reichspost, tasked with handling telephone and telegraphic traffic, had broken the American A-3 Voice Scrambler, that was based on 1920s technology [6].
  2. A. B. Clark would later become research leader at the NSA.
  3. VODER was the abbreviation of Voice Operating Demonstrator.
  4. The Green Hornet was a fictional crime-fighter on US radio in the 1930s and in films from the 1940s onwards. It's tune resembled the buzzing sound made by SIGSALY on the radio channel.  Wikipedia

A complete SIGSALY terminal consisted of more than 30 full-height 19" racks. Approximately half of these racks were used for the receiver, whilst the other half contained the transmitting parts. The diagram below shows the receiving half of a SIGSALY terminal, with part of the transmitting side visible at the front. At the right are the four turntables that provide the random KEY streams.

Technical description
The simplified block diagram below roughly shows how SIGSALY works. At the left is the analyzer that converts the analogue speech into a digital signal that is encrypted and modulated onto the transmitter. The encryption key is produced by a one-time key phonogram disc on a turntable. A national timing standard (broadcasting on the LW band) is used to calibrate the internal timings.

At the receiving end, the internal timings are calibrated by the same national timing standard. After demodulating the incoming signal, it is decrypted and fed to a synthesizer that converts it to audible analogue speech again. In reality, the system had two turntables at either side, in order to allow a new OTP key record being cued up during conversations longer than 12 minutes.

A vocoder is a compression device for human speech, based on the principle that the properties of speech vary relatively slowly due to the way the sound is generated in a person's mouth. The frequency spectrum (150 Hz - 2950 Hz) is first divided into 10 nearly equal bands, of which the amplitude is measured by means of a linear rectifier and smoothed by a 25 Hz low-pass filter.

A separate system is used to determine whether the sound is voiced (e.g. a, e, o, z) or unvoiced (e.g. s, f, k). If the sound is voiced, its pitch or fundamental frequency is determined. Like the amplitude, this information varies slowly and can be limited by a low-pass filter to 25 Hz. For technical reasons, that will be explained later, this information was sent in 2 separate channels.

This results in 12 low-frequency data channels that are sent to the receiving terminal, where they are used to reconstruct (synthesize) the human speech again. If the original sound was unvoiced, the synthesizer generates white noise. If the original sound was voiced, it generates a source of harmonics of the fundamental frequency, that is used as the input to a series of bandpass filters, similar to the ones used for analysis. The output of each bandpass filter is then amplitude modulated with the value from the corresponding data channel (1-10). After filtering the output of each filter once more, the signals are finally combined to produced the reconstructed speech.

In the above block diagram, the 11 data channels, identified by the red dashed arrows, contain analogue information, which can not be sent over a narrow band radio channel, and can not be encrypted easily. This is solved by converting the information into a finite number of steps. This process is known as quantization. For SIGSALY the data was converted into 6 non-linear steps.

The example above shows how this works. The original (analogue) signal is represented by the black line. This signal is sampled at fixed intervals into one out of six possible levels, represented by the red line. This way, each of the 10 frequency/amplitude signals is quantized separately. At BTL, quantization was known as STEPPING and the sampling circuit itself was called a STEPPER. In today's terminology, it is known as companded digital pulse-code modulation (PCM) [17].

Note that in today's terminology, a more appropriate expression to describe the method of digitizing the individual parameters of human speech, would be parametric coding or vocoding, whereas PCM is commonly used to describe direct sampling of the raw (human) sound.

Pitch extraction
Sending the pitch information (i.e. the fundamental frequency of the human voice) is slighly more complex. Early experiments at BTL had shown that quantizing the pitch frequency to 6 discrete levels was not sufficient and that 30 steps or more were needed for a good reproduction of the original pitch. This was solved by using two 6-level channels for transmission of the pitch.

The first channel carries the pitch information quantized to the nearest 6-level value. This value is then subtracted from the original pitch signal in order to obtain the error, or delta (Δ). This Δ-value is always less than 1/5th of the original value. After multiplying it by five, it is quantized by a similar circuit into 6 discrete levels and sent in the second channel. This way, the pitch data is quantized in 36 levels (6 x 6), which is more than the minimum of 30 levels that were needed.

The above block diagram shows how the two pitch values were obtained. They were each sent in a separate 6-level channel. The sampling interval had to be much longer than the typical 2-3 ms path delay differences caused by selective fading on a transatlantic radio link. After trying many combinations, a 6-level arrangement with a 20 ms (50 Hz) sampling rate appeared to work best. The technique of sending the pitch information as two separate values (coarse and fine) is also known as two-step vernier quantization or residue quantization. It was first used on SIGSALY.

In a digital system like SIGSALY, timing is paramount. For correct operation it is mandatory that the systems at both ends are properly synchronised and stay synchronised for the duration of the conversation, even over a fading medium, like a (14,000+ km) transatlantic short-wave radio link which some­times dropped out completely. Although this initially seemed like a difficult-to-solve problem, it turned out to be relatively simple, mainly due to the choice of system parameters, such as the low sampling rate of 20 ms (50 Hz).

All that was needed at both ends, was a stable high-precision frequency standard or reference oscillator. Once both systems were 'in-sync', they would stay synchronised for several hours without any considerable drift. Short-term drifting was never large enough to cause trouble with the sampling rate, and long-term drifting could easily be compensated for. A BTL engineer 1 even developed a system for Automatic Frequency Correction (AFC) to correct for frequency shifts which might occur in carrier radio systems. This greatly improved the FSK transmission system.

In practice, the synchronisation never posed any serious problems. Although the signal was sometimes lost completely for considerable periods, due to the atmospheric conditions of the HF radio channel, the system would still operate correctly once transmission was restored. The only part of the system for which timing was really critical, was the keyer with its phonogram records.

  1. Harold L. Barney.

For encryption of the twelve 6-level signals, a mixing system is used in which a 6-level signal from a random noise source is added to the original quantized 6-level speech data, by means of modulo-6 addition, similar to the way in which the Vernam Cipher works for 2-level digital telegraphy signals in an OTT cipher system. At BTL, this technique was known as REENTRY.

The simplified block diagram above shows how the data from the KEY stream is added to the data from the digitized speech channel. Both values are in the range 0-5. If the sum of the two values is higher than 5, it gets 'wrapped around' by the REENTRY circuit, so that the result remains in the 0-5 range. The diagram below illustrates how this works for SIGSALY's 6-level numerical data.

The diagram at the top shows the original speech signal that is quantized into 6 discrete levels, as shown by the red line. The middle diagram shows the key that will be added to the original signal. The key is generated by a noise source and is recorded onto a disc. When playing back the disc, its signal is sampled at the same intervals and its quantized 6-level values (the blue line) are added to those of the original signal by means of a mathematical modulo-6 addition (at the time known as re-entry). The result is represented by the green line in the diagram at the bottom.

Modulo-6 addition can best be explained by following the hands of a clock, as illustrated by the diagram above. In this example, the initial value is 4 and the KEY value is 3. Adding these two values together produces 4 + 3 = 7. Which gives us the result (output) of 7 modulo 6 = 1. In other words: the KEY value determines the number of steps that have to be added to the INPUT.

At the receiving end, a copy of the key disc is used to retrieve the original data again, simply by subtracting the KEY value from the previous result. In the above example, the input value is 1 and the KEY value is 3, which produces the result 1 - 3 = -2, and -2 modulo 6 = 4. In other words: the value has to be rotated 3 steps counterclockwise in order to reproduce the original value.

Key production
Generation and duplication of the cryptographic KEY material was one of the most critical parts of the entire system, as it directly determined the security of the cipher. Each of the KEY characters, or steps had to be completely random, and the random sequence of characters should never be repeated. Furthermore, a separate random character stream was needed for each of the 12 data channels. Using the same key for all channels would have been a cryptographic weakness.

Creating the random stream of characters was solved by using a white noise source, symbolised in the above block diagram by a diode. In reality, the noise generator was based on a hot cathode gas discharge valve (tube). After amplification, the noise signal is sampled by a so-called stepper circuit into 6 discrete levels (0-5) similar to the amplitude sampling steppers in the Vocoder.

After much consideration at BTL, it was decided to use phonogram records or discs for the actual distribution of the KEY material, as this was one of the most reliable and stable reproduction processes as the time. Nevertheless, it required very special precisely-driven turntables for both the production and play-back process, in order to meet the strict system timing requirements.

The quantized 6-level noise signal is used to control the amplitude of a fixed-frequency tone generator as illustrated above. This results in an Amplitude Modulated (AM) tone with 6 discrete levels. After filtering and amplification, this signal is then recorded onto a phonogram disc.

By using 12 nearly identical key production circuits, with a different tone for each of the 12 channels, the KEY stream for all 12 channels is recorded simultaneously onto a single disc, as illustrated in the block diagram above. The audio tones were selected carefully, as they were not allowed to have any relation between them. This process is described in US Patent 3,373,245.

As each phonogram record could only hold about 12 minutes of KEY material, several such discs were needed for a single conversation. In order to guarantee uninterrupted operation, two turn­tables were used, allowing the next disc to be 'cued up' whilst the current one was running.

For this to work properly, both discs had to be identical and had to be started at exactly the same position. This was solved by creating the discs simultaneously on two electrically coupled phono­gram recorders or cutters, and placing index marks on the discs automatically. All the operators at both sides had to do, was to place the needle in the correct groove at the index mark. In order to guarantee quick synchronisation after start-up, the motors of both turntables were kept running all the time, synchronised to the general system timing (reference oscillator).

The timing of a play-back turntable, is referenced to a contact that is briefly engaged at each full revolution. This producs a so-called tacho-pulse that can be used for synchronisation. This way, not only the rotational speed, but also the exact angle (position) of the disc could be controlled.

Towards the end of each KEY record, a control tone or pilot tone is recorded (with the existing key tones) to mark the start of the other turntable. At the beginning of the tone, an electric latching mechanism is engaged to couple the table instantly to the motor axle on the start of the next revolution of the table. At the end of the tone, the second turntable is assumed to be stable enough, and the key stream is switched to that turntable. The first one can then be reloaded.

The simplified block diagram above shows how the rising and falling edges of the control tones are used to start and stop the motors of the two turntables, and to select which record is used as the source of the key stream. The three units at the right are so-called set-reset flip/flops.

The diagram above illustrates how a solenoid-operated latch was used to couple the turntable to the continuously running (synchronised) motor. The motor drives a large heavy flywheel 1 that is kept running all the time. In the drawing, the leftmost latch is engaged, whilst the one at the right is waiting to be activated by the control tone. The operation of control tone-operated disc cueing and disc indexing is described in more detail by Kingsbury Davis in US Patent 3,024,321.

  1. Note that in reality the flywheel was probably much bigger than it is shown in the simplified drawing above.

Alternate key   SIGBUSE
Although the use of One-Time phonogram records with truly random keys was the only way to guarantee absolute cipher security, they were not very practical for testing and maintaining the machines. Their production was very expensive and each record only lasted 12 minutes.

BTL Engineers therefore developed a mechanical device with many electrical relays, that produced a pseudo random key that was good enough for lineup and testing of the installation. This device was, of course, never used for high-level calls.

Because of the noise it made in operation, the device was affectionally called the 'trashing machine' [9]. Its operation is described in great detail by Oscar Myers in US Patent 3,937,888, that was filed in 1943 and was kept secret for more than 31 years before it was declassified in 1975. The image on the right shows part of it.

In an operational context it the trashing machine was known as the Alternate Key, or AK. It was later assigned the official name SIGBUSE. The AK was used for daily maintenance and for less important conversations, but was known to be unreliable. If phonogram records were used and the system lost its synchronisation, the call would be interrupted immediately. When the AK failed, e.g. due to a failing relay contact, it would start with an intermittend sound that resembled a galloping horse, which gradually became worse as the error propagated through the system [1].

Transmitting a 6-level data signal was another challenge. Amplitude Modulation (AM) was tried, but was found inadequate as a result of selective fades that could at times be as much as 20 dB. In order to accurately reproduce the 6-level signal, the amplitues had to be produced with a 1% accuracy (1 dB). The solution was to use a technology known as: Frequency-Shift Keying (FSK).

FSK was already used successfully in digital telegraphy (telex). As a telex character is represented by a 5-bit digital value, each bit is represented by one of two levels: a 0 or a 1, or in telegraphy speak: a mark and space. This means that only two frequency positions were needed to send the state of a single bit, hence the name two-level FSK, or just FSK [11]. To send six-level data, six frequency positions are needed, hence the name Multiple Frequency-Shift Keying or MFSK [12]. The implementation posed many new problems for which new filter techniques were developed.

In SIGSALY, 12 such 6-level MFSK channels had to be transmitted to the other end. This was done by Frequency Modulating (FM) the 6-level stepped data onto a Low Frequency (LF) carrier in the 1000-3000 Hz range. This part was identical for all 12 channels. The 12 FM signals were then each Amplitude Modulated (AM) onto LF carriers with incrementing frequencies. After removing one of the sidebands, the results were mixed and sent on a High Frequency (HF) AM carrier.

This resulted in another first for SIGSALY: Multi-Carrier Transmission. A good description of the modulation techniques used for SIGSALY are provided by Robert Mathes in US Patent 3,991,273 that was filed on 4 October 1943 and was kept secret for 33 years. It was disclosed in 1976.

Alan Turing
British mathematician Alan Turing, known for his work on breaking the German Enigma cipher, the development of the Bombe codebreaking machine and for his pioneering work on computers, was briefly involved in SIGSALY's development. In November 1942 he went on a two month top-level liaison mission to the United States, to oversee their codebreaking work on Naval Enigma.

In his first week he visited Benjamin deForest Bayly in New York, with whom he discussed the security of Telekrypton, an old Western Electric cipher machine that Bayly was converting into a One-Time Tape device. Bayly's machine would later evolve into the unbreakable Rockex.

On his trip, Turing also visited BTL where, after the initial security clearance problems, he was allowed to see and discuss the progress on their developments of the voice encryption system. It required personal intervention by Field Marshal Sir John Dill and General George C. Marshall, but utimately Turing was allowed to see SIGSALY.

Turing recognized the quality and the importance of the work that was done at BTL [16]. He was convinced about the security of the system, but had his reservations about the fact that SIGSALY would be operated exclusively by US personnel, which gave them the ability to listen in 'if they so desired' [17]. Nevertheless, he gave his approval for use and installation of the device in London.

Turing's consent was a very important one, as it probably contributed to the development of the US-British inter-relations, that were later formalised by the signing of the British-United States Agreement (BRUSA) in May 1943. On his return to the UK, Turing started the development of Delilah, a similar system that was based on the work he had seen at BTL. Although Delilah was never taken into production, Turing fed some of his ideas back to BTL for use in SIGSALY.

The name SIGSALY was not an acronym, but a cover name or codename, starting with the letters SIG, just like in SIGABA and SIGTOT. The first prototype was named The Green Hornet, after the buzzing sound it produced on the communication channel. The following names were used:

  • X System
  • Project X
  • Ciphony I
  • Green Hornet
In total 12 complete SIGSALY terminals were built and installed around the world. The following SIGSALY locations have been confirmed [8]:

  1. Washington (US, Pentagon)
  2. London (UK, Selfridges)
  3. Algiers
  4. Brisbane (Australia)
  5. Fort Shafter (Hawaii)
  6. Washington (US, for Pacific)
  7. Oakland (US, California)
  8. Paris (France, after liberation)
  9. Guam (maritime installation) 1
  10. Frankfurt (Germany, post-war)
  11. Berlin (Germany, post-war)
  12. Tokyo (Japan, post-war)
  13. Manilla (Philippines) 2
Each terminal was able to contact every other terminal, and could also be used for relaying a conversation. There are examples of conversations between London (UK) and Brisbane (Australia), that were relayed via Washington (USA), as described by David Kahn in September 1984 [17].

  1. This system was installed on a 250-ton lighter that followed General MacArthur during his South Pacific campaigns. As an example, the image shows a 250-ton Australian AV-2050 120ft Motor Lighter, which might have been a candidate for the maritime installation [21].  Wikipedia
  2. One or more installations may have been relocated after the war. It is also possible that the maritime installation (9) showed up at this location. There were never more than 12 terminals in total.

New technologies
The following technologies are claimed as SIGSALY 'firsts' [8]:

  • Encrypted telephony (as opposed to voice scrambling)
  • Quantized speech transmission
  • Transmission with Pulse Code Modulation (PCM)
  • Companded PCM
  • Multiple Frequency-Shift Keying (MFSK)
  • Speech bandwidth compression
  • Frequency Shift Keying-Frequency Division Multiplex (FSK-FDM)
  • Multilevel 'eye pattern' to adjust sampling intervals
  • Two-step vernier (residual) quantization
  • Multi-carrier transmission 1
  1. In some literature, this feature is described as Spread Spectrum transmission, which was probably correct at the time it was invented. The term 'multi-carrier transmission' describes it more accurately however.

Amoung others, the following people were involved in the development of SIGSALY (ordered alphabetically by surname):

  • Badgley, Robert H.
    Vernier quantization
  • Barney, Harold L.
    Automatic Frequency Crrection (AFC)
  • Barstow, J. M.
    Vocoder manufacturing
  • Bennett, W. R.
    PCM, Transmission systems
  • Blye, P.W.
    Project engineer expanded project (September 1942)
  • Busch, Aloysius J.
    Pseudo Random Key Generator (US Patent 3,968,454)
  • Clark, A.B.
    Project leader Vocoder Research group
  • Cole, I.E.
    Precision phonogram recording
  • Curtis, A.M.
    Project leader Circuit Research group
  • Davis, Kingsbury H.
    Key system
  • Dow, J. L.
    Switching Development group
  • Dudley, Homer W.
  • Edson, J. O.
  • Gannett, Danforth K.
  • Gray, C. R.
    Manufacturing at Western Electric
  • Hartley, R.V.L.
    Transmission systems, Consultancy
  • Joel, A.E.
    Pseudo Random Key Generator
  • Llewellyn, F. B.
  • Lundstrom, Alexis A.
    Stepper, Reentry
  • Marrison, A. C.
    Precision phonogram recording
  • Mathes, Robert C.
    Transmission Research group, Random Noise
  • Meacham, L.A.
  • Melhose, Alfred E.
    Prototype building coordination and drawings
  • Vernier quantization
  • Mitchell, D.
    Message coding equipment manufacturing
  • Mohr, Milton E.
    Digital encoding, Quantizer (stepper)
  • Myers, Oscar
    Pseudo Random Key Generator (US Patent 3,937,888)
  • Newby, Neil D.
    Key system (early work)
  • Norwine, Andrew C.
    Key system
  • Nyquist, Harry
    Vocoder, PCM, Consultancy
  • Olcott, E. W.
    Manufacturing at BTL (also at Western Electric)
  • Peterson, Eugene
  • Potter, Ralph K.
  • Riesz, R. R.
    Vocoder, pitch extraction
  • Schimpf, Luther G.
    Limiter/detector, Stepper, Reentry
  • Shannon, Claude
    PCM, Reentry (modulo 6), cryptanalysis
  • Consultancy, Project oversight (UK)
  • Vaughan, Henry E.
    Key system (early work)
Related patents
Most of the SIGSALY-related patents were filed in or around 1942 by some of the people listed above, but were not disclosed until 1976 [9]. The technologies described in the patents are at the heart of modern communications, including the Global System for Mobile communication (GSM).

Below is an overview of the patents that were filed in relation to SIGSALY. The rightmost column, highlighted in red, shows how many years the patent was kept secret, before it was officially released to the public. Click the patent number in the first column to view the original patent.

Earlier patents
Patent Inventor Title Filed 1 Released 1 # 2
2,151,091 HW Dudley Vocoder 30-10-1935 21-03-1939 3
Patents related to Project X
Patent Inventor Title Filed 1 Released 1 # 2
3,024,321 KH Davis, AC Norwine Key record (Continuous recording system with indexing means) 29-12-1944 06-03-1962 17
3,076,146 ME Mohr Cathode beam tube circuit having means for converting current variations to stepped waveform 27-12-1945 29-01-1962 16
3,188,390 ME Mohr Signal transmission with secrecy 20-12-1943 08-06-1965 21
3,193,626 HL Barney Duplicate record indexing system 29-12-1944 06-07-1965 20
3,340,361 RK Potter Signaling system with cathode ray tube quantizer 09-07-1945 05-09-1967 22
3,373,245 ND Newby, HE Vaughan Production of current of random variation 27-08-1942 12-03-1968 25
3,394,314 LG Schimpf Circuit supplying impulses of regulated peak amplitude 17-07-0943 23-07-1968 25
3,405,362 RH Badgley, LG Schimpf Space discharge tube circuit 20-12-1943 08-10-1968 25
3,470,323 HW Dudley Signalling system 30-06-1944 30-09-1969 25
3,967,066 RC Mathes Secret telephony 24-09-1941 29-06-1976 35
3,967,067 RK Potter Secret telephony 24-09-1941 29-06-1976 35
3,985,958 HW Dudley Secret telephony 18-12-1941 12-10-1976 35
3,897,591 AA Lundstrom, LG Schimpf Secret transmission of intelligence 27-08-1942 29-07-1975 33
3,912,868 RH Badgley, RL Miller Telephone pricacy system 17-07-1943 14-10-1975 32
3,937,888 O Myers Signal transmission with secrecy 17-07-1943 10-02-1975 31
3,991,273 RC Mathes Speech component coded multiplex carrier wave transmission 04-10-1943 09-11-1976 33
3,979,558 E Peterson Signalling system 30-06-1944 07-09-1976 32
3,976,839 RL Miller Telephone privacy system 30-06-1944 24-08-1976 32
3,965,296 RL Miller Signaling system 30-06-1944 22-06-1976 32
3,887,772 RL Miller Signal privacy with safety feature 30-06-1944 03-06-1975 31
3,891,799 AE Melhose Code device with light responsive key generator 27-09-1944 24-06-1975 31
3,983,326 DK Gannett Key pulse generator for secrecy signalling circuit 27-09-1944 28-09-1976 32
3,968,454 AJ Busch Signaling circuit 27-09-1944 06-07-1976 32
3,944,744 DK Gannett Matrix coding secret signalling system 10-05-1945 16-03-1976 31
3,944,745 DK Gannett Secret signaling system with means for preventing key disclosure 10-05-1945 16-03-1976 31
3,953,677 DK Gannett Key signaling system with multiple pulse generators 10-05-1945 27-04-1976 31
3,953,678 DK Gannett Speech component key signaling system with code combinations 10-05-1945 27-04-1976 31
3,924,074 E Peterson Pulse position modulation key signaling system 19-05-1945 02-12-1975 30
3,983,327 DK Gannett, AC Norwine Electrical signaling 09-07-1945 28-09-1976 31
3,934,078 DK Gannett Key generating system 01-05-1946 20-01-1976 30
3,965,297 DK Gannett Secret communication signal generating system 01-05-1946 22-06-1976 30
3,924,075 DK Gannett Two-way privacy system terminal with single key pulse generator means 20-03-1947 02-12-1975 28
  1. All dates are in European notation DD-MM-YYYY.
  2. Number of years before the patent was declassified.

Video footage
Below is a selection of video clips that are related to SIGSALY or some of its features. Please note that Crypto Museum is in no way affiliated with or involved in any of these video productions or its makers, and can not accept any responsibility for the correctness of the provided information.

Encryption, Episode 1 - SIGSALY: AT&T Labs
The clip below is the first episode of a series of video presentations by AT&T. It discusses the invention of SIGSALY, the worlds first unbreakable telephony device that was developed by Bell Telephone Labs (BTL). At the time, BTL was a full subsidary of AT&T.

 Watch the clip on the AT&T website

The Voder - Homer Dudley (Bell Labs) 1939
Voder was the first device that could produce an electronically synthesized human voice. It was developed by Homer Dudley at Bell Telephone Laboratories (BTL) in Murray Hill (New Jersay, USA). The video clip features the full audio of a demonstration at the New York World's Fair in 1939.

VODER (1939) - Early Speech Synthesizer
Below is a short video clip that shows how VODER was operated manually by Helen Harper to imitate the human voice. It gives a good impression of how SIGSALY must have sounded.

AFC   Automatic Frequency Correction
Method for automatically correcting the frequency of an oscillator in order to keep a receiver tuned to the center of a signal's carrier.
AK   Alternate Key
Electromechanically produced pseudo-random key that could be used with SIGSALY instead of the (expensive) phonogram records. The AK was used for testing and for low-level calls. Codenamed SIGBUSE.  More
FSK   Frequency-Shift Keying
Method for transmitting a 2-level (binary) digital signal by means of shifting the carrier frequency. In Frequency Modulation (FM) this is done by sending the two levels as audio tones (AFSK).  Wikipedia
MFSK   Multiple Frequency-Shift Keying
Similar to FSK, but suitable for multi-level digital signals. In the case of SIGSALY, 6-level data was transmitted this way.  Wikipedia
OPEPS   Off-premises extension privacy system
Secure (local) extension lines to SIGSALY, protected by means of gas pressure on the cables, microswitches on the connection boxes, and balanced noise on the twisted wires of the lines.
PCM   Pulse-code modulation
Digital representation of a sampled analogue signal [17].  Wikipedia
SIGBUSE   codeword
Codeword for the Alternate Key (AK), an electromechanical device that generated a pseudo-random key stream, used for testing. Also known as the Trashing Machine.  More
SIGGRUV   codeword
Codeword for the vinyl-based phonograph records with 12 minutues of OTP key material for SIGSALY, that were initially used.  More
SIGJINGS   codeword
Codeword for the acetate-coated aluminium phonogram records that replaced the SIGGRUVS vinyl records.  More
SIGSALY   codeword
Codeword for the complete full-duplex secrecy telephony system described on this page. Initially known as X System or Project X.  Other names
  1. J.V. Boone and R.R. Peterson, Sigsaly - The Start of the Digital Revolution
    NSA Website. Retrieved September 2015.

  2. Wikipedia, SIGSALY
    Retrieved October 2016.

  3. Jerry Proc and contributors, SIGSALY
    Retrieved November 2015.

  4. Wikipedia, Selfridges, Oxford Street
    Retrieved October 2016.

  5. Wikipedia, Embassy of the United States, London
    Retrieved October 2016.

  6. Patrick D. Weadon, Sigsaly Story
    Retrieved October 2016.

  7. Wikipedia, Pulse-code modulation
    Retrieved October 2016.

  8. James V. Boone, The WWII Cryptologic Heritage of the United States'
    Computer and Communications Industries. Date unknown.
    Page 6 refers to an 1983 IEEE review.

  9. Project X - A True Secrecy System for Speech - Section 4.3 (extract)
    Author: RL Miller (Section 4.3), Engineering and Science in the Bell System
    Bell Telephone Laboratories, Inc. 1978. pp. 296-317. ISBN 0-932764-00-2. 1
    Ralph Miller was one of the SIGSALY developers  More

  10. Wikipedia, Vocoder
    Retrieved October 2016  Homer Dudley's Voder

  11. Wikipedia, Frequency-shift keying
    Retrieved October 2016.

  12. Wikipedia, Multiple frequency-shift keying
    Retrieved October 2016.

  13. Donald Mehl, Personal correspondence
    Various SIGSALY images. October 2016. 2

  14. Wikipedia, Allied invasion of Sicily
    Retrieved October 2016.

  15. Wikipedia, Allied invasion of Italy
    Retrieved October 2016.

  16. Alan Turing, Report from Washington
    Report on Cryptographic Machinery available at the Navy Department Washington.
    28 November 1942. Crown Copyright.

  17. David Kahn, Cryptology and the origins of spread spectrum
    IEEE Spectrum, September 1943. pp. 70-80.

  18. Donald E. Mehl, The Green Hornet: America's unbreakable code for secret Telephony
    1997-2002. Self-published.

  19. The National WWII Museum, Pentagon aerial view 1943
    New Orleans. Source unknown. Retrieved October 2013.

  20. Sydney W. Newbury, Image of Selfridges, Oxford Street, 1929
    RIBA Library Photographs Collection. Obtained via Victoria and Albert Museum.
    October 2016.

  21. Wikipedia, 120ft Motor Lighter
    Retrieved November 2016.
  1. Restored by means of OCR from a scanned document by David Allen. October 2005.
  2. Reproduced here by kind permission from the author.

Further information
Any links shown in red are currently unavailable. If you like the information on this website, why not make a donation?
Crypto Museum. Created: Sunday 30 October 2016. Last changed: Tuesday, 19 December 2023 - 23:36 CET.
Click for homepage