Click for homepage
Crypto
Rotor
Crypto AG
Hagelin
  
H-460 →
← CX-52
  
HX-63   HX-61
Offline rotor cipher machine - under construction

HX-63 was an electromechanical rotor-based cipher machine, introduced in 1963 by Crypto AG in Zug (Switzerland). It features nine electrically wired permutations wheels, or rotors, that have more contacts than the 26 letters of the alphabet. It uses a principle – patented by Boris Hagelin – that is very similar to the – also patented – re-entry principle of the American AFSAM-7 (KL-7). It is arguably the most sophisticated and secure rotor machine ever to be taken into production.

The machine has nine cipher rotors (from a set of 12), 1 mounted on a common spindle. Each rotor has 41 contacts, and moves in an irregular stepping pattern, under control of configurable pins and a set of user configurable modifiers 2 that are part of the internal cryptographic KEY.

Text should be entered on the keyboard, which has a teleprinter layout. The wiring between the keyboard and the rotor stack can be altered with 41 modifiers that are also part of the internal KEY. The current position of the rotors can be altered quickly as part of the external KEY.
  
Hagelin HX-63

The output is available on a double paper strip printer, that prints plaintext and ciphertext simultaneously. The device can optionally be expanded with the PEH-61 tape reader/puncher. The HX-63 uses a novel principle known as re-entry, or re-injection. As each wheel has 41 contacts, whilst only 26 are needed for encryption of the alphabet, the remaining 15 lines are looped around. Hagelin probably picked up the idea for re-entry in 1952 during a visit to Bonn (Germany) where he discussed the American KL-7 cipher machine with Dr. Erich Hüttenhain. 3

Development of the machine started in late 1952 and took more than 10 years, during which time it was renamed several times (HX-54, HX-55, HX-56, etc.). An intermediate design, with fewer features and hand-crank operation, was released in 1960 as the HX-61. When the final design was ready, in 1963, it received the name HX-63. The first 12 machines were sold to a French government agency in 1964 [7]. The HX-63 was in production until at least 1967. 4 Based on surviving serial numbers, it is likely that between 50 and 200 units were made. It is the only rotor machine ever made by Crypto AG, 5 and was succeeded in 1970 6 by the fully electronic HC-460.

  1. The machine came with 9 rotors, but according to the manual 12 rotors were optionally available [A].
  2. In his memoirs [7], Hagelin uses the word Modificator for this, but as this is not a proper English word — it does not appear in the Oxford English Dictionary — we have replaced it with Modifier.
  3. During WWII, Dr. Erich Hüttenhain had been the chief cryptographer of OKW/Chi; the cipher bureau of the Third Reich (Nazi Germany). After the war, he worked for the Society of Scientific Work, which later became the Zentralstelle für das Chiffrierwesen (ZfCh) (now: BSI) of which he was the director until 1970.
  4. Deduced from the date codes on various electronic components inside the HX-63 in our collection.
  5. Boris Hagelin had previously worked on the TKG-35, a rotor machine that he co-developed in 1951 with Dr. Edgar Gretener from the Swiss company Gretag, but that was before Crypto AG had been established.
  6. The H-460 was withdrawn immediately after the initial release in 1970, because of electrical and mechanical reliability issues, and was re-released in 1972.  More

PLEASE HELP — We are still looking for original documentation of the HX-63, such as user manuals, service documentation and circuit diagrams. All we have at the moment is a 1964 draft description. You can help us to improve and expand this page by providing any additional information on this topic.  Contact us
Hagelin HX-63
HX-63 front view
HX-63 rear view
Interior seen from the front
Interior seen from the rear
Rotor set with 9 rotors on a spindle
Serial number tag
Mains power cable
A
×
A
1 / 8
Hagelin HX-63
A
2 / 8
HX-63 front view
A
3 / 8
HX-63 rear view
A
4 / 8
Interior seen from the front
A
5 / 8
Interior seen from the rear
A
6 / 8
Rotor set with 9 rotors on a spindle
A
7 / 8
Serial number tag
A
8 / 8
Mains power cable

Features
Front
The machine is housed in a grey/greenish enforced polyester enclosure that consists of two case shells: a bottom part and a top part. The upper part can be removed by pressing a hidden button under the front left corner of the case. This is necessary when accessing the internal KEY settings. At the front left is a large black five-position knob that is used to select the desired mode of operation. To its right is the keyboard, which has a layout that is similar to that of a teleprinter. 1

In plaintext mode (K) the device supports most (but not all) characters of the ITA-2 teleprinter alphabet. In cipher mode (C) the letters J and X are given up to provide letter-shift, figure-shift and SPACE. J and X are only available in figure-shift mode, but may also be substituded by other letters. 2 In decipher mode (D), only the 26 letters of the Latin alphabet (A-Z) can be entered.

Click to see more

At the top right is a narrow window through which the current positions of the rotors are visible. The rotor positions are numbered 1-9 from right to left. At the top left is the outlet of the built-in double printer, which prints plaintext and ciphertext simultaneously. At the centre is a 5-digit counter that shows the number of characters that have been entered on the keyboard.

 Look inside the machine

  1. Different layouts were available for different countries. In this case, the keyboard has a German QWERTZ layout, but there are also machines with the more common QWERTY layout [13].
  2. The letter J can commonly be replaced by I without affecting the readability, whilst the letter X can be replaced by a 2-letter combination such as KS.

Mode of operation
The mode of operation is selected with the large knob at the left. It has 5 possible settings:

  • O
    Off
  • E
    Ein (on) — set external key
  • K
    Klar (plain)
  • C
    Chiffrieren (cipher)
  • D
    Dechiffrieren (decipher)
Connections at the rear
All connections are at the rear. At the left is a 2-pin socket for connection of an external 12V DC power supply, such as the battery of a car. To its right is the 110 or 220V AC power inlet that is used when the device is powered from the mains (default). At the right are two multi-pin sockets for connection of an external (optional) PEH-61 five level paper tape reader/puncher.

Click to see more

Keyboard
Current rotor positions
Mode selector
Character counter (here shown at position 00000)
HX-63 rear view
Connections at the rear
B
×
B
1 / 6
Keyboard
B
2 / 6
Current rotor positions
B
3 / 6
Mode selector
B
4 / 6
Character counter (here shown at position 00000)
B
5 / 6
HX-63 rear view
B
6 / 6
Connections at the rear

Tape reader/puncher   PEH-61
The HX-63 could optionally be expanded with the PEH-61 paper tape reader/puncher that had been developed in 1961. It was connected to the two wide male/female connectors at the rear, and translated the parallel data of the HX-63 to and from the 5-bit ITA-2 telegraphy standard. It is very similar to the PEB-61 tape unit that was available as an option for the mechanical CX-52. The two devices share many components and were probably developed in parallel in 1961.

Photograph of a PEH-61 next to a HX-63. Extracted from [6 p.50]

Block diagram
Below is the block diagram of the HX-63, as printed in the manual [A]. There are 9 rotors, each of which has 41 contacts at either side. The contacts at the left side of a rotor are connected to the contacts at the right in a randomised (scrambled) order. The rotor set, or scrambler, is terminated at both ends with a circular contact plate with 41 contacts, that acts as the entry or exit disc.

The scrambler of the HX-63 — 9 rotors on a spindle with a contact disc at either end

At the contact discs, 26 contacts are used for enciphering the 26 letters of the Latin alphabet. The remaining 15 contacts are looped back to the other side of the drum. This is known as re-entry or feed back or re-injection. 1 It was patented by Boris Hagelin in 1953 and before him in 1944 by Albert Small, but that was unknown to Hagelin as it was registered as a secret patent [3].

Simplified block diagram of the HX-63, shown in Cipher mode (C)

Apart from the wiring inside the rotors and the order in which the rotors can be placed on the spindle, there are other modifiers 1 that have an effect on the encryption algorithm. The 26 wires of the input matrix can be altered in 26! ways (4 · 1026), the 15 re-entry wires can be swapped in 15! ways (1.3 · 1012) and the rotor drive unit can be altered in 9! ways (362880). In addition, there are two rotor stepping methods that can be selected with the M/MM switch.

  1. The re-entry principle was also used in the American KL-7 and in the Russian M-125-3 (Fialka). It is likely that Hagelin came up with the idea after discussing the KL-7 with Erich Hüttenhain in Bonn in 1952.

Key setting
The HX-63 has a wide variety of key settings, some of which are changed more frequently than others. In the draft manual [A], the key variables are split into two groups: the internal key and the external key. The external key can be set by the operator and is therefore also known as the message key. The internal key on the other hand, requires the case to be opened. Some of the internal key settings however, require components to be taken apart, and may in reality only be changed, say, once a year (if they are changed at all). We will call these the basic key settings.

1. Basic key
Typically changed once a year, or just once in the machine's lifetime:

  • Exit disc wiring (41!)
  • Wiring order of each rotor (9 × 41!)
2. Internal key
Typically changed once a day:

  • Rotor order (9!) (or 12!/3!) 1
  • Pin configuration of each rotor (41!9)
  • Input wiring modifiers (26!)
  • Re-entry modifiers (15!)
  • Stepping modifiers (9!)
  • Rotor stepping method M or MM (2)
3. External key
Also known as the message key. Typically changed before each new message:

  • Start position of each rotor (419)
  1. Depending on the number of rotors that came with the machine (9 or 12).


Rotors
The machine uses nine electrically wired cipher rotors that are placed in a variable order on a common axle, or spindle, that should be installed in the rotor bay. When lowering the locking bar, the rotor stack will be embraced by two fixed contact discs: the entry disc at the right, and the exit disc at the left. The wiring of the entry disc can be changed with the input modifiers, whilst the wiring of the exit disc can only be changed by opening the disc and swapping its wires.

Click to see more

The rotor positions are numbered 1-9 from right to left, as shown in the image above. Each rotor is identified by a red letter that is engraved in its left side. The first nine rotors are identified with the letters A B C D E F G H J (note that the 'I' is omitted). The optional rotors had the letters K L M.

Each wheel has 41 divisions that are numbered 01 to 41. For each division there is a pin that can be set to the left (inactive) or right (active), in the same way as on the earlier CX-52. The pins are sensed by a micro-switch at a certain position on the circumference of the rotor, and control the stepping of one of the other rotors, subject to the current settings of the stepping modifiers.

Click to see more

When the rotor stack is removed from the machine, the rotors can be removed from both sides of the spindle, as shown the image above. They are held in place by a series of springs that are embedded in the spindle. This way, the order of the rotors can be changed in less than a minute.

Changing the wiring order inside a rotor is more difficult. It involves the removal of the retaining clip and the pertinax cover plate. The numbered contacts at the right side of the rotor can then be swapped as required, after which the rotor can be reassembled. It is likely that a special tool was supplied to make contact swapping easier, but this tool is currently missing. It is unlikely that the rotor wiring was changed very often, as the wires are fragile and could easily break. The wiring of the exit disc – at the left side of the rotor bay – can be changed in the same manner.

Rotor set with 9 rotors on a spindle
Rotor set with 9 rotors on a spindle
Rotor set with 9 rotors on a spindle
Rotors close-up
Rotor identification 'J'
Left hand side rotor contacts
Right hand side rotor contacts - close-up
Right hand side rotor contacts
Rotor bay seen from the front
Rotor bay
Entry disc
Exit disc
Rotors taken from the spindle; one rotor opened
Configurable internal wiring of a rotor
Rotor wiring and contacts
Rotor wiring and contacts
C
×
C
1 / 16
Rotor set with 9 rotors on a spindle
C
2 / 16
Rotor set with 9 rotors on a spindle
C
3 / 16
Rotor set with 9 rotors on a spindle
C
4 / 16
Rotors close-up
C
5 / 16
Rotor identification 'J'
C
6 / 16
Left hand side rotor contacts
C
7 / 16
Right hand side rotor contacts - close-up
C
8 / 16
Right hand side rotor contacts
C
9 / 16
Rotor bay seen from the front
C
10 / 16
Rotor bay
C
11 / 16
Entry disc
C
12 / 16
Exit disc
C
13 / 16
Rotors taken from the spindle; one rotor opened
C
14 / 16
Configurable internal wiring of a rotor
C
15 / 16
Rotor wiring and contacts
C
16 / 16
Rotor wiring and contacts

Modifiers
The modifiers — Hagelin calls them modificators — are located at the rear of the machine. To alter their settings, the upper case shell must be removed and the machine must be reversed. The modifier block is removable, to allow a second (optional) unit to be prepaired in advance with the KEY settings for the next day, so that changing the daily KEY would only be a matter of seconds.

At the bottom are the input wiring modifiers. It is bascially a contact matrix that controls how the 26 lines from the keyboard are connected to the entry disc. For each wire, a 26-position thumb-wheel is available, marked A-Z. The top row shows the current setting. To avoid short circuits and open lines, and each letter must occur precisely once. This is a typical permutation feature.

Click to see more

At the top left are the re-entry modifiers, numbered 27 to 41. They are similar to the input wiring modifiers, but control how the 15 surplus wires from the exit disc are looped back to the entry disc. Although each selector has 26 positions, only the first 15 (A-O) are used. The top row shows the current setting. As it is also a matrix, each letter (A-O) must occur precisely once.

At the top right are the nine stepping modifiers. Each one is connected to a rotor-pin sensing switch, and controls which rotor will advance when an active pin is encountered. Each selector can be set in 9 different positions (1-9). Like the other modifiers, it is an electric matrix, which means that each number (1-9) must occur precisely once.

Modifier block
Permutation matrix (41 thumbwheels) and stepping modifier (9 thumbwheels)
Modifier block removed
Modifier block - external side
Modifier block - internal side
Wiring detail
Modifier block detail
Wiring detail
D
×
D
1 / 8
Modifier block
D
2 / 8
Permutation matrix (41 thumbwheels) and stepping modifier (9 thumbwheels)
D
3 / 8
Modifier block removed
D
4 / 8
Modifier block - external side
D
5 / 8
Modifier block - internal side
D
6 / 8
Wiring detail
D
7 / 8
Modifier block detail
D
8 / 8
Wiring detail

Key space
The table below provides a calculation of the maximum number of possible key settings of the machine. As not all settings will be equally useful — there may be doubles, conflicting settings and settings that produce a weak key — the number of useful settings is also given, based on the description from Hagelin's chief developer Oskar Stürzinger in the 1964 draft description [A].

Setting Possible 1 Useful 2 9 rotors 12 rotors 3
Wiring modifiers 26!×15! ≈ 5 × 1038 1 × 1030 1 × 1030  1 × 1030 
Internal rotor wiring 41! ≈ 3.3 · 1049 1 × 1030 1 × 10270 1 × 10270
Rotor order 9! or 12!/3!     all    3.6 × 105   8 × 107  
Pin settings 241 ≈ 2.2 × 1012 10 1 × 1054  1 × 1054 
Stepping method M/MM 2 2 × 10 2 × 100   2 × 100  
Stepping modifiers 9! ≈ 3.6 × 10 3.6 × 10 3.6 × 105   3.6 × 105  
Exit disc wiring 41! ≈ 3.3 × 1049 1 × 1030 1 × 1030  1 × 1030 
Initial rotor position 419 ≈ 3.3 × 1014 3.3 × 1014 3.3 × 1014  3.3 × 1014 
  ↳  Possible 7.8 × 10640 1.7 × 10643
    Useful 8.5 × 10409 1.9 × 10412
    Realistic 8.5 × 10109 1.9 × 10112
In his memoires [7 p.49], where Hagelin says that 10600 settings were theoreticaly possible, he is probably referring to the maximum number of 10640 (when using 9 rotors) or 10643 (when using 9 rotors from a total set of 12). In reality, the number of useful settings was closer to 10409 (the equivalent of ~ 1359 bits), which is still quite an impressive number for the era.

However, as altering the wiring of the rotors and the exit disc is tedious and time consuming, it is likely that in practice they were never changed. This means that, when a machine was captured, the internal wiring would become known to the enemy. This reduces the number of possible permutations to a more realistic value of > 10109 (~ 366 bits), which is still much more than the Enigma (~ 72 bits or less), and the later Data Encryption Standard (DES) which had only 56 bits.

  1. Possible settings per rotor (when applicable).
  2. According to chief developer Oskar Stürzinger in the 1964 draft manual [A].
  3. 9 rotors taken from a total of 12.

Versions
The HX-63 was available in a number of versions with full or reduced fuctionality [A]:

  • HX-63
    This is the most complete version of the machine, that has all of the (optional) features present. It offers a speed of 8-10 cps and has a useful key space of > 10409 (1359 bits).

  • HX-63 without stepping modifiers
    In this version the stepping behaviour of the rotors was fixed and could not be modified by the user. The M/MM selector might also be omitted in this case. The irregular stepping pattern could still be changed with the pin-settings however (41 on each rotor).

  • HX-63 without modifier block
    In this version, the modifier block at the rear was not present and could be replaced by a fix-wired dummy. This significantly reduced the cost of the machine.

  • HX-63 without connection for the tape device
    The two wide centronics connectors at the rear – intended for connection of a PEH-61 paper tape reader/puncher – were omitted from this version. This reduced the complexity of the internal wiring and therefore also the cost of the machine.

  • HX-61
    This is an earlier version of the HX-63 that was probably introduced in 1960 [A][16]. It is a spin-off from the HX-63 development and allows manual operation by means of a hand-crank. It has a reduced speed of 5 cps (or 1 cps in manual mode). The crank was later dropped from the final HX-63 design.

  • HR-61
    Motor-driven variant (like HX-63) without M/MM feature, and with a single 41-point modifier. It is likely that this is a combination of the 26-point and the 15-point modifier. Delivered with 9 cipher rotors with straight through wiring (to be modified by the customer) [16].
History
By 1952, most of the developed countries were using rotor-based cipher machines, also known as rotor machines, whilst Crypto AG (Hagelin) was still selling mechanical cipher machines. The rotor machine was based on the German Enigma that had played an important role during WWII.

The UK used Typex and the USA had its SIGABA. Furthermore, the US had just introduced the KL-7, a machine that was also used by NATO. Rotor machines seemed to be the way forward, and Hagelin dreamed of making the best one ever: a super rotor machine that surpassed all others.

The idea for the machine was born in 1952, shortly after Hagelin had made a trip to Bonn, where he spoke with Dr. Erich Hüttenhain of the Society of Scientific Work. 1 It is likely that they discussed the recently introduced KL-7 cipher machine and the reported reliability issues. 2
  
The American SIGABA

On the way back to Zug (Switzerland), Hagelin prompted the idea of a super rotor machine to his chief engineer Oskar Stürzinger. It would have more rotors and more permutation possibilities than any of the high-end rotor machines that were in use at the time, such as Enigma, SIGABA, AFSAM-7 (KL-7) and Typex. In addition, the machine would feature the irregular stepping of his famous CX machines. He asked Stürzinger to find out whether the re-entry principle could work.

Development of the machine started in late 1952 and took more than a decade, during which time various improvements and enhancements were made. During the development of the HX, Hagelin kept his contact at the NSAWilliam Friedman – informed of the progress. At least twice – in 1955 and 1957 – Friedman mentioned the HX-machine in reports to his superiors [9][10].

When Friedman visited Hagelin in Switzerland in February 1955 [9], Hagelin told him that he had several prototypes in mind, such as HX-54, HX-55, HX-56 and THX-55. The number of rotors was not finalised at that point, but he had already shown him an early lab model with 5 rotors and mentioned that the HX-54 would have 10 rotors with 32 contacts each, which is more than the 26 contacts needed for the Latin alphabet. When Friedman asked him what he was planning to do with the extra 6 contacts, Hagelin replied that he would be using them for re-entry. 3

In April 1956, Hagelin wrote in a personal letter to Friedman that two prototypes of the machine – by then known as HX-56 – were about to be released to the French Government for evaluation. Apparently the French had promised Hagelin to demonstrate the machine to NATO later on. Although Hagelin was confident that the machine would survive the evaluation, he confirmed that there were still problems with the construction of the rotors, which he would address later [14].

When Friedman visited Hagelin again in September 1957, the HX-machine was still in the proto­type stage, but progress had been made with respect to the reliability of the rotor contacts. When Friedman argued that the new machine might be too expensive for the customers, Sture Nyberg – Hagelin's plant manager – commented that it would cost less than US$ 3000 [10]. After this, neither the HX-machine nor any sales restrictions were ever discussed again between Friedman and Hagelin. It is possible that Friedman regarded it a 'dead end', as electronic cipher machines were about to take over. Shortly afterwards, Friedman left the stage to enjoy his retirement. 4

In 1961, an intermediate version of the machine was released. It was known as the HX-61 and could be operated with a hand-crank as well as by an internal motor; driven from the AC mains or an external 12V DC source [A]. It is believed that some of these were released to the French.

By the time the HX-63 was ready for production, in 1963, the first electronic cipher machines were already appearing on the market, and rotor machines would soon be something of the past. Nevertheless, Hagelin managed to secure orders from a French Government agency, the German intelligence service BND and possibly several others, and produced a modest quantity of them. According to some sources 5 only 12 to 15 units were made, but judging from the build quality and the serial numbers of the surviving machines, a quantity of 50 - 200 seems more likely.

In his memoirs [7 p.49], Hagelin wrote:

A number of these rotor machines were supplied together with the tape punch device PEH-61 to a French government agency. [...] Due to the development of electronic machines, the manufacture of he HX-63 and the PEH-61 machines was discontinued. We manufactured for another firm a sizeable number of high-speed printers and tape readers originally developed for this machine.
For NSA and CIA this was a fortunate coincidence. The HX-63 was probably a cryptographically secure machine – arguably one of the most sophisticated rotor machines ever to be taken into production – which would have been very difficult to manipulate. Electronic cipher machines on the other hand, were based on shift registers, and could be manipulated much easier. A few years later, the NSA would develop the cryptologic 6 for Hagelin's all-new electronic H-460 machine.

  1. The Society of Scientific Work was the predecessor of the Zentralstelle für das Chiffrierwesen (ZfCh), today known as the BSI. During WWII, Hüttenhain had been the chief cryptologist of OKW/Chi.
  2. The contents of this meeting are still classified, but it is likely that they discussed the KL-7 (AFSAM-7), as shortly afterwards Hagelin came up with the idea of the re-entry principle. It is also likely that they discussed the reliability issues of the KL-7, as Hagelin later discussed these with Friedman [9][10 p.4].
  3. This was an interference with a secret American patent, but Hagelin wasn't aware of that, and Friedman didn't tell him either.
  4. Friedman had already retired in August 1955 after suffering a heart attack, but was recalled from retirement in the summer of 1957 to fix the broken relationship between Hagelin and the NSA [2]. After his second retirement, Friedman's health steeply declined and he finally died in November 1969.  More
  5. Various publications quote the book Machine Cryptography and Modern Analysis of Deavours and Kruh (1985) as source for the quantity of 12 units, but we believe this to be incorrect. The book only briefly mentions the HX-63 and does not discuss production quantities at all.
  6. In this context, cryptologic refers to the electronic circuit that implements the cryptographic algorithm. It is also known as the crypto heart.

Re-injection patent
During the 1950s, the US National Security Agency (NSA) had a secret deal with Boris Hagelin, the owner of Crypto AG, that controlled to which countries he could sell unreadable 1 machines. This was known as the Gentleman's Agreement. During the 1950s, NSA's chief cryptologist William Friedman visited Hagelin a number of times to negotiate and alter the terms of the agreement. One of these visits was on 22 September 1957 at Hagelin's home in Zug (Switzerland) [2][10].

The first thing to be discussed between Hagelin and Friedman, was patent 2,802,047 that Hagelin had filed in the US in October 1953 and that had been granted a month earlier. Although it was granted in the US, the patent had been declined in Japan, and Hagelin was wondering why.

The patent describes a cipher machine in which more contacts are used on the cipher rotors than are actually needed and where the extra contacts of the output of the drum are looped back to the input. This method can be described as re-entry or re-injection, and was going to be used in Hagelin's new HX-machine that was underway.

Friedman was shocked when he saw the patent, but didn't tell Hagelin. The re-injection principle had already been registered in the US in 1944 in a secret patent [15], so Hagelin's attempt should have raised a declaration of interference [10].
  
The American AFSAM-7 (later: KL-7) that also uses the re-injection principle

The re-entry principle was conceived during WWII by Albert Small, whilst working for the Army Signal Intelligence Service (SIS), when trying to solve the high-level Japanese diplomatic Purple cipher. It is covered by US Patent 2,984,700 and has since been used at the heart of the high-level American cipher machine AFSAM-7 (later: KL-7), that had also been adopted by NATO.

Although officially Hagelin should not have been aware of the existence of the AFSAM-7, in later talks he mentioned that many of its operators were experiencing contact problems with the rotors. 2 This proves that Hagelin was aware of the existence of the machine and that he might have been aware of its re-injection principle.

He explained how he conceived the idea after a trip to Bonn in 1952 were he was told something by Erich Huttenhain, 3 followed by discussions with Oskar Stürzinger– his chief developer – but insists that re-injection was his own idea [10].
  
A Russian M-125-3 (Fialka) cipher machine with 10 rotors and re-entry

The re-entry principle was also used in the Russian M-125-3 (Fialka). Introduced in 1965 4 it was used in the former USSR and in the countries of the former Warsaw Pact. It features 10 cipher rotors, irregular stepping, a card reader that replaces the modifiers, a printer, a tape puncher and a tape reader. Yet it was much smaller than the HX-63. Fialka was used until the early 1990s.

 More about the KL-7
 More about Fialka

  1. The term unreadable means that the algorithm could not be broken by NSA. Also known as unfriendly or secure. In contrast: algorithms that are breakable by NSA, are called friendly or readable or exploitable.
  2. This is actually correct; the KL-7 was known for its many contact problems if the maintenance instruction were not strictly followed. This proves that Hagelin had knowledge about the AFSAM-7 (KL-7) [10 p. 4].
  3. It is currently unclear as to what Hüttenhain told Hagelin on this occasion, or what Hagelin discussed with Stürzinger afterwards, as the original document is redacted at this point. It is entirely possible though, that Hüttenhain was talking about the AFSAM-7. The same principle was later also used in the Russian Fialka.
  4. Fialka was originally introduced in 1956 (before the HX-63) as the M-125, but that model did not yet feature re-entry. The M-125-3 version, released in 1965 (after the HX-63), introduced re-entry, but only when the device was placed in numbers-only mode.

Possible misunderstandings
Not much has been published about the HX-63 in the past, mainly because very few machines have survived and hardly any documentation is available in the public domain. As a result, some claims in other publications are incorrect, or at least do not apply to the final version of the machine. Here are some examples:

  1. Two wiring sets in each rotor
    In the book Codeknacker gegen Codemacher [12 p. 122], author Klaus Schmeh suggests that each rotor has two sets of wiring of which the selection is under control of the user. This is not the case. Each rotor has only one set of 41 electric circuits. The confusion is probably caused by the fact that Deavours and Kruh explain in [11] that the current takes one of two paths. However, they are referring to the re-entry principle.

  2. Hand crank operation
    In the IEEE publication The Scandalous History of the Last Rotor Machine [13], author Jon D. Paul shows an early prototype of the HX-63 on which a hand crank can be inserted at the right side of the machine; alledgedly for manual operation in case of a motor failure. This feature is not present on the final machine. It is likely that the machine featured in the article is an early prototype, or that it is the earlier HX-61.

  3. Withdrawn at request of NSA
    In the IEEE publication The Scandalous History of the Last Rotor Machine [13], author Jon D. Paul claims that the HX-63 was taken off the market at the request of NSA's William Friedman, but none of the quoted sources support this claim. By 1963/64, Friedman was no longer working for the NSA; he had retired in 1957 [2]. It is more likely that the HX-63 was withdrawn because electronic cipher machines were rapidly taken over from rotor machines, as confirmed by Boris Hagelin in [7 p.49]. In other words: it came too late.

  4. Only 12 units produced
    In many publications, such as [5] and [13] it is claimed that only 12 to 15 units were ever made. Although this might be correct, we have yet to find a source that confirms it. In any case, the book that is commonly referred to as the source – Machine Cryptography and Modern Cryptanalysis by Deavours & Kruh [11] – does not reveal the production quantity.

  5. Plugboard
    It is claimed in a number of publications, for example in [5], that the HX-63 had one or two plugboards, similar to the Steckerbrett of the Enigma cipher machine. This is not the case. Rather than plugboards, it has two thumbwheel-operated matrices: one for the 26 letters of the alphabet and one for the 15 re-entry lines. Although their function is similar to Enigma's plugboard, they are more advanced and offer many more permutations. 1
  1. In fact, the operation of the HX-63 input wiring modifiers can better be compared to the operation of the card reader of the Russian M-125 (Fialka), because of its single-ended nature. The double-ended plugboard of the Enigma always swapped letters in pairs, which greatly reduced its strength.  More



Interior
Front
The machine is housed in a polyester enclosure that consists of two case shells. To access the interior, the user only has to press a button at the bottom of the front left corner, after which the upper case shell can be taken off. This reveals the interior as shown in the image below.

Top view of the machine's interior after removing the upper case shell

The machine consists of a large heavy die-cast aluminium frame to which all parts and sub-circuits are mounted. The frame itself is bolted to the bottom case shell with four large bolts that are accessible from the bottom of the case. The five sub-systems are directly visible: the MODE selector, the keyboard, the scrambler (rotor set), the printer and the modifier block (at the rear).

Interior seen from the front
Drum seen from the front and M/MM selector
M/MM selector
Interior - mode selector and printer
Double printer with two commutators
Printer commutator
Interior seen from the rear right
Gears
Clutch
Printer output path
Counter
Detection switch at the right side
E
×
E
1 / 12
Interior seen from the front
E
2 / 12
Drum seen from the front and M/MM selector
E
3 / 12
M/MM selector
E
4 / 12
Interior - mode selector and printer
E
5 / 12
Double printer with two commutators
E
6 / 12
Printer commutator
E
7 / 12
Interior seen from the rear right
E
8 / 12
Gears
E
9 / 12
Clutch
E
10 / 12
Printer output path
E
11 / 12
Counter
E
12 / 12
Detection switch at the right side

Rear
In order to alter the settings of the modifiers, the open machine must be turned around as shown in the image below. At the top left is a removable block that holds three sets of thumb-operated switches: 26 for the altering the wiring of the entry disc, 15 for altering the order of the re-entry wires, and 9 for altering the wheel stepping pattern. Note that not all settings are useful.  More

Bottom view of the interor after removing the frame from the bottom case shell

Interior seen from the rear
Modifiers (rear side)
Modifiers close-up
Printer seen from the rear left (paper tray closed)
Printer with open paper tray (seen from the rear)
Four fuses
Connections at the rear
F
×
F
1 / 7
Interior seen from the rear
F
2 / 7
Modifiers (rear side)
F
3 / 7
Modifiers close-up
F
4 / 7
Printer seen from the rear left (paper tray closed)
F
5 / 7
Printer with open paper tray (seen from the rear)
F
6 / 7
Four fuses
F
7 / 7
Connections at the rear

Chassis
The image below shows the bottom side of the machine after the frame has been removed from the bottom case shell. This section holds the mains transformer, the DC motor that drives the gears, and most of the wiring. At the left is a 4-deck rotary switch that is operated by the MODE selector. It moves in tandem with the plain/cipher/decipher selector below the keyboard contacts.

Bottom view of the interor after removing the frame from the bottom case shell

The 12V DC motor that drives the machine is housed in the die-cast aluminium compartment at the centre, together with appropriate filtering parts. Bolted to the compartment is an aluminium panel that holds two large 1000µF/35V capacitors. This was probably a last-minute upgrade.

Interior seen from the bottom
Mains transformer
PSU components
Two relays
Printer drive electronics
Printer electronics close-up
12V DC motor
G
×
G
1 / 7
Interior seen from the bottom
G
2 / 7
Mains transformer
G
3 / 7
PSU components
G
4 / 7
Two relays
G
5 / 7
Printer drive electronics
G
6 / 7
Printer electronics close-up
G
7 / 7
12V DC motor

Disassembly
When repairing or restoring an HX-63, it might be necessary to take the machine (partially) apart. Whilst most of the items at the bottom side are fixed, the two major building blocks at the top – the scrambler and the printer – can easily be removed.

Main chassis with scrambler and printer removed - seen from the front right
Main chassis with scrambler and printer removed - seen from the front left
Main chassis with scrambler and printer removed - seen from the rear
Scrambler, seen from the front right
Scrambler, seen from the front left
Scrambler, seen from the rear
Printer assembly
Printer assembly
H
×
H
1 / 8
Main chassis with scrambler and printer removed - seen from the front right
H
2 / 8
Main chassis with scrambler and printer removed - seen from the front left
H
3 / 8
Main chassis with scrambler and printer removed - seen from the rear
H
4 / 8
Scrambler, seen from the front right
H
5 / 8
Scrambler, seen from the front left
H
6 / 8
Scrambler, seen from the rear
H
7 / 8
Printer assembly
H
8 / 8
Printer assembly

Restoration
When we received the machine featured here in September 2021, it was in unknown condition. It had not been switched on since it was decommissioned in the 1970s or 80s. With a machine of this vintage, it is always important to carry out a thorough inspection before connecting it to the mains power. Dried out capacitors and binding mechanical parts may cause serious problems.

After opening the case, we noticed that some parts of the 4-deck MODE-selector were floating around in the bottom case shell. Apparently, one of its swiper contacts had been blocking the selector, after which a previous owner had used excessive force to get the selector going again.

In doing so, one of the flat-faced contacts on deck #3 had broken off, as shown n the image on the right. The swiper contact and its spring had fallen out, but the (damaged) plastic holder was still in place. Ceramic four-deck switches of this type are hard to find and difficult to repair.
  
Broken switch deck

Removing the switch and taking it apart was not a tempting option — it is fully integrated with the chassis and the wiring — so we had to come up with a different solution. In the end we fixed it by wiring the affected segment in parallel to an unused segment on the adjacent deck #4.

For this we needed access to the rear side of the switch, which involved removing the scrambler and the printer from the top side of the frame.

Next, the two 1000µF/35V electrolytic capacitors that are bolted to the motor compartment were checked. They were both gone and had to be replaced. The problem however, was that capacitors of this type are no longer available.

This was solved invisibly by emptying the existing aluminium enclosures and placing modern high-quality capacitors inside them.
  
Replacement capacitors in old enclosure

At the moment, the restoration of this machine is still in progress. Additional information about this process will be added here in due course.

Problems
  • Mains cable missing
  • One contact of 4-deck rotary switch broken off
  • Two 1000µF capacitors in DC circuit gone
  • Rubber rollers inside printer desintegrated
  • Clutch not always released
  • Mains earth leakage circuit breaker occasionally trips
  • Clear/Cipher lamps do not come on
Fixed
  • New power cable made
  • Broken segment of 4-deck rotary switch routed to unused segment
  • Modern capacitors placed inside old 1000µF enclosures
Mode selector and printer wiring
Broken switch deck
Defective capacitors
Replacement capacitors in old enclosure
I
×
I
1 / 4
Mode selector and printer wiring
I
2 / 4
Broken switch deck
I
3 / 4
Defective capacitors
I
4 / 4
Replacement capacitors in old enclosure

Connections
Mains power
The device can be powered directly from the AC mains, but you should always check the voltage selector first, to ensure that it matches the local mains voltage. Below is the pinout of the mains receptacle at the rear side of the device, when looking into the receptacle.

  1. 220V AC (1)
  2. 220V AC (2)
  3. Ground
12V DC
The device can also be powered by an external 12V DC source, such as the battery of a car. Below is the pinout of the 12V receptacle at the rear of the machine, when looking into the receptacle.

  1. +12V (+ terminal)
  2. 0V (- terminal)
Specifications
  • Device
    Cipher machine
  • Class
    Rotor-based cipher machine
  • Manufacturer
    Crypto AG (Hagelin)
  • Model
    HX-63
  • Years
    1963-1967
  • Rotors
    9 (selected from an optional set of 12)
  • Stepping
    Irregular
  • Keys
    Basic, internal, external
  • Settings
    > 10640 (2127 bits)
  • > 10409 (1359 bits)
  • Period
    419 ≈ 3.3 · 1014
  • Printer
    8 - 10 characters per second (2x)
  • Paper
    17.4 mm wide
  • Mains
    110 or 220V AC
  • Battery
    12V DC (external)
  • Dimensions
    400 × 400 × 230 mm
  • Weight
    16 kg
Known serial numbers
  • 58315
    Private collector (Germany)
  • 58316
    Crypto Musem (Netherlands)
  • 58317
    National Cryptologic Museum (NCM, NSA) (USA)
Documentation
  1. Schlüsselgerät Typ HX-63, Vorläufige Beschreibung
    A-1130-c. Oskar Stürzinger, March 1964.
References
  1. Oskar Stürzinger, Chiffriertechnik Heute
    Vorlesung 'Krieg im Aether' 1976/1977.
    ETH Zürich (German). pp.15-16. 1

  2. Paul Reuvers and Marc Simons, The Gentleman's Agreement
    Crypto Museum, 30 July 2015.

  3. Boris Hagelin, US Patent 2,802,047
    Electric switching device for ciphering apparatus.
    Filed 16 October 1953.

  4. Albert W. Small, US Patent 2,984,700
    Method and Aparatus for Cryptography.
    Filed 22 September 1944. Secret until 16 May 1961.

  5. Wikipedia, HX-63
    Visited January 2020.

  6. Boris Hagelin, Die Geschichte der Hagelin-Cryptos
    Original manuscript by Boris Hagelin in German language. Zug, Fall 1979.

  7. Boris Hagelin, The Story of Hagelin Cryptos 2
    English translation of the above. BCW Hagelin, Zug, Spring 1981. Later edited by David Kahn and published in Cryptologia, Volume 18, Issue 3, July 1994, pp 204-242.

  8. Paul Reuvers and Marc Simons, Operation RUBICON
    Crypto Museum, 19 March 2020.

  9. William Friedman, Report of Visit to Crypto AG (Hagelin) 21-28 February 1955
    28 March 1955. TOP SECRET. Final version.
    A2436243. Partly declassified by NSA on 22 July 2014 (EO 13526).

  10. William Friedman, Hagelin Negotiations (draft)
    Report about Friedman's visit to Zug, 18 December 1957.
    A60669. Partly declassified by NSA on 20 July 2014 (EO 13526).

  11. Cipher A. Deavours and Louis Kruh, Machine Cryptography and Modern Cryptanalysis
    1985. ISBN 0-89006-161-0.

  12. Klaus Schmeh, Codeknacker gegen Codemacher
    2008. ISBN 978-3-937137-89-6. p. 122.

  13. Jon D. Paul, The Scandalous History of the Last Rotor Machine
    IEEE Spectrum, 31 August 2021.

  14. Boris Hagelin, Letter to William Friedman about HX-56
    Zug, 9 April 1956

  15. Signal Corps Patent Board, Meeting no. 30
    22 November 1940.

  16. Frode Weierud, Production details of Hagelin HX and HR machines
    Personal correspondence, 17 September 2023.
  1. Retrieved from HAMFU History, December 2018.
  2. The English version of Hagelin's memoirs [7] is less accurate than the German version [6]. In particular in the description of the HX-63 [7 p.49] he made a number of mistakes (e.g. saying that a rotor has 241 possible wirings) that are not present in the German version [6 p.48].

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: Wednesday 01 January 2020. Last changed: Monday, 18 September 2023 - 06:08 CET.
Click for homepage