The Lorenz SZ42
Schlüsselzusatz 42 — the cipher attachment Allied codebreakers called Tunny.
The traffic served the army high command — OKW to senior field commanders across occupied Europe. At Bletchley Park the non-Morse stream was named Tunny. It is a different machine from Enigma, a different traffic, and a different literature. It belongs in this collection next to the field sets, not inside them.
SZ42 from above, hood off — motor, EMO / EMF receivers, UR cover
Figure placeholder. On the Signal article this view shows the 50 W / 1,500 rpm motor, the two receiver magnets (EMO local, EMF remote) and the automatic transmit/receive relays. Replace this box with your own photograph or a licensed museum image.
What it was for
Gilbert Vernam’s 1917–19 idea was simple and strong: take each bit of a five-unit teleprinter character and combine it, modulo 2 (XOR), with a secret keystream bit. Do the same operation at the far end with the same keystream and the plaintext comes back. The method is only as good as the key. If the keystream is random, as long as the message, and never reused, the cipher is a one-time pad. If the keystream is generated by a machine and the starting point is reused, the traffic develops “depth” and can be attacked.
In wartime it was not practical to print matching random tapes for every high-command circuit and get a copy to each end. Lorenz’s answer was a machine that generated a long pseudo-random five-bit stream from twelve pinwheels, added it to the message electrically, and sent the result down the existing teleprinter path. The distant SZ42, set to the same wheel start, added the same stream again and printed the clear text.
The attachment ran from 220 V, 50 Hz mains. Inside the base, two rectifier chains fed the telegraph magnets (about 100 V DC) and the relays (about 70 V DC). A governed motor held speed to within half a percent so the two ends stayed in step. There was no manual send/receive switch. Whichever receiver saw the first character after a pause — local teleprinter or remote line — set the direction.
From Vernam to Lorenz
Vernam’s US patent 1,310,719 (issued 1919) and his 1926 AIEE paper set out the tape-and-relay form of the idea. C. Lorenz AG of Berlin, a major teleprinter and radio firm, had been owned since 1930 by ITT of the United States while remaining an operating German company. Whether that corporate path carried American printing-telegraph cipher patents into Lorenz’s drawing office is a question the General Report on Tunny raised after the war; it is noted here, not settled.
What the German patents do show is that Lorenz was filing the pieces of an on-line multi-alphabet teleprinter cipher from 1938. Siemens already had the T52. Lorenz’s machine had to work with any standard five-unit teleprinter. That choice — stay inside International Telegraph Alphabet No. 2 — kept the attachment easy to install and, later, gave Bletchley a known alphabet to work on.
Dr Gerhard Grimsen
The designer at the centre of this page is Dr Gerhard Grimsen (1899–1963). Born in Braunschweig, he served in Field Artillery Regiment 46 in 1917–18, studied mathematics and physics at Braunschweig, Göttingen and Berlin, and took his doctorate at Halle. From 1922 he worked on long-distance communications at the Telegraph Technical Reich Office. On 1 October 1926 he joined Lorenz. From 1929 his work was telegraphy, including message encryption at Zehlendorf and with the Mühlhausen development group. British intelligence interviewed him in May 1945. He was still at Lorenz in the late 1950s.
His May 1938 patent DE767350, “Method for generating and sending secret multi-alphabet telegrams”, is the cryptographic foundation later visible in the SZ42. Later Grimsen filings cover key-signal storage, teleprinter memory, the drive of encrypted multi-alphabet pulses, and a fully automatic cipher converter (DE928773, filed July 1943). Karl Weber filed important mechanical details, including adjustable cam wheels.
After the war Grimsen designed the MI-544 (1956), a tape-key teleprinter cipher used in banks and by some NATO users. It keeps the same XOR principle and drops the wartime rotor nest — the form Vernam had wanted when tapes could be exchanged in peacetime. Grimsen’s own 1956 note on the MI-544 points back to the American work of 1915–18.
How the twelve wheels make a key
Twelve wheels stand in a row, each with a thumb-wheel and a number ring. The number of positions, left to right, is 43, 47, 51, 53, 59, 37, 61, 41, 31, 29, 26 and 23. Most of those lengths share no common factor, which stretches the period before the combined pattern repeats.
- Psi / A / Spri (wheels 1–5) — 43, 47, 51, 53, 59. They move only some of the time.
- Motor / Mu (wheels 6 and 7) — 37 and 61. They decide whether the psi wheels step.
- Chi / B / Spa (wheels 8–12) — 41, 31, 29, 26, 23. They step once per character.
Each wheel carries camlets — small cams that can be pushed into an active or an idle position. The pattern of camlets is the basic key and was changed by an officer, at first monthly, later more often. The start position of every wheel was meant to change with each message. The operator set the rings, then sent an unencrypted indicator (QEP) plus the start figures so the far end could match them.
For each of the five bits of a character the machine adds, modulo 2, the plaintext bit, the chi bit and the psi bit. Enciphering and deciphering are the same operation if both attachments begin on the same settings. The mechanical story — clutches, five-bit sword memories, locking bars, the KM and UM magnets — is set out in detail by D. W. Davies in Cryptologia 19(1), 1995, and in the 1943/44 German operating instructions. This page does not repeat that paper. The patent index next door maps Lorenz filings onto those mechanisms.
Cryptographic unit (Schlüsselteil) — twelve wheels, Zeiss-Ikon lock
Figure placeholder. A clear photograph of the removable crypto unit belongs here. The Deutsches Museum holds a well-known example; use your own plate or a licensed image and caption the source.
The limitations, and why the cipher broke
From 1943 Lorenz added “limitations” so that psi stepping depended not only on the motor wheels but on other parts of the machine — chi-2 from February 1943, the fifth plaintext bit (P5) from late 1943, psi-1 from June 1944. The aim was to spoil the clean statistics Bletchley was using. P5 also meant a single line error could throw the two ends out of step until an operator reset.
Two operational facts mattered more than any extra magnet. The machine stayed inside the standard five-unit alphabet. And the keystream could be reused: the same wheel start, or the same cam pattern, appearing on more than one message. Depth of that kind, not a failure of the XOR idea itself, is what opened Tunny. Colossus was built to grind the statistics at electronic speed. That computing story is well published elsewhere; this site’s job is the attachment, the patents and the documents in the collection.
In this collection
This page is written from original papers in the collection (including material that passed through TICOM after 1945), from the 1943 circuit (Werner 101b / N 46 292, 4 October 1943), from the Vorläufige Betriebsvorschrift, from Lorenz’s own company history, and from Grimsen’s postwar MI-544 notes. The mechanical analysis follows Davies (1995) and is cited, not reproduced.
- Complete patent cross-reference — 21 Lorenz AG patents, Grimsen’s filings marked.
- German landing page — Field power, SZ42, Enigma, air as four doors.
- Photographs and any Lo15 or line equipment in the store will be attached to this page as they are catalogued. Until a set is on the shelf, the record is documentary.
I have not taken an SZ42 to pieces. Where a mechanical detail matters, the citation is Davies or the Betriebsvorschrift. Corrections from people who have are welcome.