First generation secure telephone unit
STU-I was the first generation digital
Secure Telephone Unit (STU) introduced by the
NSA in the 1970s as a replacement for the
It was the first large-scale deployment of a
secure telephone system
that introduced Linear Predictive Coding (LPC) as well as the use of a
central key facility known as a Key Distribution Center (KDC).
For encryption it used the secret SAVILLE algorithm
that was jointly developed by GCHQ
and the NSA.
STU-I is also known as KY-70
Although the initial design goal was to create a system that would fit
on a desktop, the STU-I became a bulky two-unit setup, consisting of a
half-height 19" rack with the electronics, and a
converted telephone set as the voice terminal.
The image on the right shows the most visible part of the STU-I setup,
the voice terminal, which is actually a converted 5-line Western Electric
telephone set, of which the line buttons were used for switching between HOLD,
CLEAR and SECURE modes. The red buttons on the right are for priority
override on military networks.
The actual electronics were housed in a large half-height 19" rack that
was placed elsewhere in the office, or in an adjacent room, with a thick
cable leading to the voice terminal on the operator's desk. In some cases
the entire setup was placed in the corner of the office, with the voice
terminal (i.e. the telephone set) on top. The image below shows a complete
The STU-I system was jointly developed by the
National Security Agency (NSA),
the Defence Communications Agency (DCA) and GTE Sylvania (then: Philco-Ford).
It was based on one of the first generations of Sylvania's Programmable
Signal Processors (PSP), that were used for the implementation of LPC-10
The rather low-quality photograph on the right is probably one of the few
remaining images of a complete STU-I system.
The front door has been removed here, so that we get a clear picture of what
is inside the cabinet. There are two units: a PSU at the bottom, and the
actual digital unit with the controls and connections at the top.
At the center is a large MODE selector and at the top right a two-digit
LED display. Furthermore the
Crypto Ignition Key (CIK) is visible
to the left of the MODE selector, just below the FILL socket.
Although the STU-I was designed as a full-duplex system (probably using
a data transfer speed of 2400 or 4800 bps) it could also be operated in
half-duplex mode in case the line-quality was insufficient.
In that case the system would degrade gracefully and the PTT switch on the handset was used to switch between transmit and receive.
This mode was also used over radio links.
Technically speaking, the STU-I was a big success,
as it solved the COMSEC problems of the era.
Nevertheless it was too bulky and far too expensive.
The price of a single init was US$ 35,000 as opposed to the projected
US$ 5000, and it did not fit onto a desktop.
So, immediately after the introduction of the STU-I, in 1977,
the development of its successor, the STU-II,
was initiated. Although
STU-II was much smaller than the STU-I, it still was a two-piece installation,
and it was not until the arrival of the STU-III
in the late 1980s,
that the original design goals were reached.
The diagram below gives an overview of the various features of
the STU-I voice terminal. The unit has the size of a regular
telephone set of the 1970s, that has been extended with multi-line
buttons along the front edge. For the STU-I, a standard Western
Electric telephone set was modified and the line-buttons were used
to control the various features of the STU-I system.
The four red buttons on the keypad are for compatibility with the
military AUTOVON and IVSN networks.
They are used for priority override.
Note that for half-duplex calls, a
Push-To-Talk (PTT) switch
is present inside the grip of the handset.
used when the line quality was too bad for full-duplex calls, but
also when the remote party was connected via a radio link.
During the 1960s, it became clear to the US Government
that clear voice communication, via
radio or telephone, was among the most serious security threats.
An attempt to introduce speech protection to the AUTOVON
Defence Communication System (DCS) was not very successful.
The so-called AUTOSEVOCOM
system turned out to be too expensive and too cumbersome to use, and was
abandoned in the late 1960s after no less than
1850 terminals had been installed .
Other attempts and co-operations with other US departments
failed and by the mid-1960s, two problems had become clear. The first
problem was the voice quality: most narrowband systems produced
a Donald Duck-like sound, and voices were hardly intelligible.
The second problem was the complex key distribution, which had caused most users
to return to non-secure systems.
In the late 1960s, the NSA
set out to develop a secure digital telephone
system that would solve the two problems described above. The system
would be called STU-I.
The problem of voice quality was solved by applying a revolutionary technique
called Linear Predictive Coding (LPC), which greatly improved voice quality
in narrowband systems. The Adaptive Prediction Coding (APC) scheme used
the Levinson method .
For voice data signal processing, one of the first
Programmable Array Processors (PAP) was used, which
offered great advantages over the existing Programmable Signal Processors (PSP)
that were avaliable at the time. It was able to perform a so-called
multiply-and-accumulate instruction in a fraction of the time,
opening the door to the implementation of real-time LPC.
It was thought at the time that the
equivalent of 1000 Intel 8008 processors was required to do the same .
With the PAP, it was possible to achieve half-duplex LPC speech
over analogue phone lines.
In 1971, this resulted in the so-called LONGBREAK II that is shown above.
It was developed at Philco-Ford in Pennsylvania and weighed approx. 113 kg.
LONGBREAK II provided full-duplex digital speech communication at bit rates of 2400,
3600 and 4800 bits per second. Four units were sold to the US Navy and the NSA .
The report on LONGBREAK was presented in 1974 .
At the NSA, speech coding research was led by
Tom Tremain . The NSA was also responsible for the
implementation of the SAVILLE cryptographic algorithm
that was ideally suited for speech encryption.
Development of the STU-I took several years and was finally completed
in December 1974.
The LPC/APC technology that was used,
would later evolve into the LPC-10 standard .
The key distribution problem was solved by implementing the so-called
Bellfield concept, proposed by Howard Rosenblum of NSA's research and
development department . In the Bellfield concept, each terminal
has its own unique key, whilst a common session key is issued
by an external Key Distribution Center (KDC) to which both parties
must connect first. The external KDC was operated by the NSA.
The image on the right shows UN Ambassador Andrew Young using a STU-I
unit in New York during the Israel-Egypt peace talks in 1978 .
The initial design goals were to produce a small and cost-effective
terminal, so that eventually it would be on every desk in the DoD.
It was hoped that the electronic circuits could be fitted inside a case
that was slightly larger than a standard desktop telephone.
The price target was set at US$ 5000 for the first units, aiming
at US$ 2500 when the unit was produced
in quantitities .
In reality, the STU-I turned out to be the size of a small
refrigerator, whilst the price tag was set at US$ 35,000 per unit. Nevertheless,
STU-I was a technical success as it solved the immediate COMSEC needs of the NSA.
Production of the STU-I started in 1977 and ended two years later in 1979 .
The final report on the development and implementation of Wideband and Narrowband Speech, LPC, CVSD and PCM, was presented by Philco-Ford (now: GTE Sylvania)
in 1979 .
Despite the high cost of a single unit, the NSA regarded the STU-I
as a proof of concept and initiated the development of its successor,
immediately after the STU-I introduction in 1977.
Although the STU-II
was somewhat smaller than the STU-I, it was still a two-piece solution.
Compared to a standard DTMF telephone set, the STU-I terminal has
an extra row of keys to the right of the usual keys. These keys
were provided for compatibility with the AUTOVON (and later: IVSN)
non-secure telephone network used world-wide by the Department
of Defence (DoD) .
AUTOVON (Automatic Voice Network) was a military phone system that was
designed in the US in 1963 to survive nuclear attacks.
It allowed non-secure voice calls with precedence (piority override).
In the late 1960s, the DoD started the roll-out of a secure version
of AUTOVON, called AUTOSEVOCOM, but this project was cancelled
a few years later due to problems and high cost.
The STU-I allowed secure
calls over the non-secure AUTOVON network.
Later crypto phones, like the STU-II,
and the Spendex-40,
featured the same AUTOVON compatibility.
IVSN (Initial Voice Switched Network) was developed by
in the mid-1970s for unclassified voice calls, replacing
the cumersome expensive AUTOSEVOCOM network. Starting with 4 switches in
Europe in 1980, the system grew to 24 switches at the peak of its use in the
➤ More about AUTOVON
- Wikipedia, Image of STU-I Secure Phone
US Government photograph, photographed by Austin Mills (via Wikipedia).
Retrieved November 2012.
- Thomas R. Johnson, American Cryptology during the Cold War, 1945-1989
NSA 1998. Series VI, Volume 5, Book III: Retrenchment and Reform, 1972-1980.
Chapter 17, The New Targets and Techniques. pp. 142-144. 1
- Robert M. Gray, Linear Predictive Coding and the Internet Protocol
A survey of LPC and a History of Realtime Digital Speech on Packet Networks
Stanford University, 2010. Retrieved November 2012.
- Crypto Museum, The SAVILLE Encryption Algorithm
Interview with a former cryptographer at Crypto Museum, December 2011.
- Wikipedia, LPC-10 Vocoder
FS-1015 standard. Retrieved July 2011.
- Robert M. Gray, California Coding: Early LPC Speech in Santa Barbara,
University of California, Santa Barbara, 9 August 2004. Retrieved July 2011.
- Department of Defence, DARPA, et al., Packet Speech Program Review Meeting
3 June 1982. Department of Defense, Defense Advanced Research Projects Agency.
Massachusets Institute of Technology, Lincoln Laboratory.
- John Welch, Longbrake II, Final Report
Philco-Ford Corporation. Willow Grove PA. 22 November 1974. 2
- RS Cheung and AJ Goldberg, Wideband Speech Multiple Rate Processor Study. Part I.
GTE Sylvania, 1 October 1979. 2
- RS Cheung and AJ Goldberg, Wideband Speech Multiple Rate Processor Study. Part II.
GTE Sylvania, 1 October 1979. 2
Declassified and approved for relase by NSA on 9 July 2007. Retrieved November 2012
Approved for public release. Distribution Unlimited.
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 11 November 2012. Last changed: Monday, 11 February 2019 - 07:55 CET.