The Bombe A Remarkable Logic Machine: Cryptologia

Taylor &Francis
Taylor & Francis Croup
Cryptologia
ISSN: 0161-1194 (Print) 1558-1586 (Online) Journal homepage: http://www.tandfonline.com/loi/ucry20
THE BOMBE A REMARKABLE LOGIC MACHINE
Donald W. Davies
To cite this article: Donald W. Davies (1999) THE BOMBE A REMARKABLE LOGIC MACHINE,
Cryptologia, 23:2,108-138, DOI: 10.1080/0161-119991887793
To link to this article: http://dx.doi.Org/10.1080/0161 -119991887793
Published online: 04Jun 2010.
Submit your article to this journal B
Article vievvs: 125
View related articles B
a Citing articles: 3 Vievv citing articles B
Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journallnformation?journalCode=ucry20
( Dovvnload by: [University of Nebraska, Lincoln] Date: 10 September 2015, At: 09:4(ft
C^VPiOLOGiA April 1999 Volume XXIII Number 2
THE BOMBE
A REMARKABLE LOGIC MACHINE*
Donald W. Davies
Downloaded by [University of Nebraska, Lincoln] at 09:40 10 September 2015
ADDRESS: Department of Computer Science, Royal Holloway College, Egham, Surrey TW20
0EX ENGLAND.
ABSTRACT: The bombe was an electromechanical machine devised by Alan Turing and
Gordon Welchman for breaking the German Enigma cipher in World War II. The way in
which it used a redudio ad absurdum logic to reduce 263 possibilities to a few is a unique
example of ingenious circuit design, which is described in detail. Its relationship to the
Polish version of the ‘bombe’ is explored. The importance of the diagonal board is shown
as is the threat that this device faced from the German’s use of the ‘Enigma-Uhr’.
KEYW0RDS: Enigma machine, Bletchley Park, Alan Turing, Gordon Welchman, Bombe,
diagonal board, Ehigma Uhr, cryptanalysis.
INTRODUCTION
Today we perform cryptographic algorithms using state-of-the-art semiconductor
devices, so that long keys and heavy computation can be used. We are accus-
tomed to the idea that cryptanalysis needs stronger computing resources than
those employed by the crypto user. But this is a recent phenomenon. For nearlv
ali the long history of cryptology, paper and pencil was the only technology for
both cryptography and cryptanalysis. In this period, ingenuity was the main
tool of the cryptanalyst. Success depended on the relative simplicity of pencil
and paper algorithms.
The change from pencil and paper to electronics went in three stages. First
there was the introduction of simple mechanical devices during the 19th Cen-
tury, but cryptanalysis did not gain much from this change. Then the period
from about 1930 to 1950 saw a transition from simple mechanisms to electron
ics, by way of electro-mechanical techniques, pioneered in telephone switching.
There were not many widely used electro-mechanical designs, but they form a
*Thie paper was presented on the 30th anniversary of the Department of Computer Science, Royal Holloway
College, London University.
108
Davies The Bombe - A Remarkable Logic Machine
fascinating subject of study because of their short life and their combination of
mechanical and electrical ingenuity. Our subject in this paper is a remarkable
electromechanical machine for cryptanalysis, the “bombe”.
Because the cryptogra,phic machine of World War II have mostly been kept
secret by Governments, there are few for study and those of the German forces
štand out. There were two on-line cipher machines in wide use, the Siemens
and Halske T52, [1-4], and the Lorenz SZ40/42, [5], both appearing in a number
of variants as the war progressed. The largest quantity of cipher material was
produced by an off-line machine called Enigma.

British cryptanalysts took an interest in ali three machines, but the T52 was
never routinely broken. The Lorenz machines were used to send long composite
messages containing intelligence of the highest value between parts of the high
command. To break this cipher, the Colossus series of machines was developed,
the first electronic digital computers, though of a special kind.
Enigma began as a commercial machine and was developed in many ways,
creating a family of cipher machines[6]. We are only concerned here with one
version that provided copious intelligence to Bletchley Park, but at least three
other versions also received their attention.
The breaking of the Enigma cipher, first by a Polish team and then, on an
industrial scale, by the British at Bletchley Park is well known. After a period
during which the German users improved their procedures and put a stop to
early breaking techniques, the main tool for breaking Enigma was the ‘bombe’.
This used electrical connections (not voltages or currents) to represent logical
relations and it selected hypotheses by the method of reductio ad absurdum,
which was strikingly original and probably unique in machine design.
In this paper, the design of the British bombe and its method of operation
will be described. It is known that the Polish team designed and named the first
bombe, but its design and even its name remain obscure. At the end of this
paper we try to understand this early machine and see how much it anticipated
the British one.
It seems clear that the British bombe was a very different and altogether
more complex machine. Its conception is credited to Alan Turing, while Gordon
Welchman augmented the design in a critical way by adding the ‘diagonal board’,
which is described below.
Working at high speed in a highly parallel search, the bombe used an advanced
form of the technology of the Enigma machine itself. This typifies the way
in which a technologv used for cryptography also serves the cryptanalyst, by
pushing it to the limit. In no sense is the bombe’s principle a brute force attack
on the Enigma’s key space, for that would have been beyond any technology.
109
April 1999 Volume XXIII Number 2
Just how the difficult problem of breaking the Enigma cipher was solved is the
theme of this paper.
The source of the information is chiefly ‘The Hut Six Story’ by Gordon Welch-
man, [7]. For the Polish bombe we rely on a paper by Marian Rejewski, translated
from the Polish and published in Annals of the History of Computing, [8]. The
help of Frode Weierud (CERN) and Ralph Erskine is gratefully acknowledged.
Figure 1. Layout of rotors (schematic).
ENIGMA’S MECHANISM
Pressing down a key on the alphabetic keyboard of the Enigma machine lights an
alphabetic symbol which is the corresponding ciphertext letter. The relationship,
at that moment, between the 26 keys and the 26 lamps is a permutation (or
‘substitution) of the alphabet. Pressing and releasing the key also moves at least
one ‘rotor’ in the mechanism, so that the permutation for each letter is different.
There were many versions of the Enigma machine. In this paper we shall
keep to one version, used by the army and air force. Its mechanical parts are
outlined in Figure 1, showing on the right the fixed ‘entry plate’, three rotors
and a fixed ‘reversal wheel’ or umkehrwalze on the left. The German word for
this component has been used by those with first hand knowledge of Enigma,
and will be adopted here.
110
Level 1
Figure 2. Wiring scheme of an Enigma.
Each rotor turns round a common shaft and has 26 sprung pins on its right
side and 26 contacts on the left side. The entry plate has corresponding contacts
and the umkehrwalze has pins to match. As the rotors turn on the shaft, the
pins move in a cyclic fashion over the contacts, so that there are 26 connections
at each of the four sliding surfaces. Within each rotor the pins and contacts
are connected by a fixed permutation (fixed by the wiring within the rotor, but
turning relative to the machine itself.) So, as the rotors turn, the end-to-end
permutation of the 26 wires changes. The three rotors labelled L, M and R are
the left, middle and right rotors.
When a key is depressed and released, two thing happen. The rotors move to
their new position, then a circuit from the key, via the rotors etc., lights a lamp
to give the output letter. Roughly speaking, the three rotors move like a counter
in the scale of 26. The right rotor R moves every time. Once in its revolution
111
C3VPlOLOGA April 1999 Volume XXIII Number 2
(every 26 letters) the middle rotor M moves and once in M’s revolution the left
rotor L moves.
In the interest of accuracy I must mention that the ‘carry’ mechanism has
been simplified in this version of Enigma and does not quite do this. When the
middle rotor reaches its turnover position and is ready to step the left rotor, it
steps twice, so the full cycle takes not 26 x 26 x 26 steps but 26 x 25 x 26. This
quirk does not affect what we shall be explaining and the bombe’s operation
can mostly ignore it. The essential point is that almost ali of the 263 = 17, 576
joint positions of the three rotors will occur and each of them gives a different
permutation of the alphabet from keys to lamps. It is also important that only
one rotor, the R rotor, moves constantly; the others more rarely.
Figure 2 shows a simplified example of the whole wiring including the umkehr-
walze, where, for simplicity, we have shown only 6 wires instead of 26 and the
contacts have been laid out straight for ease of drawing. In the umkehrwalze U,
there is a fixed permutation, formed by connecting the pins in pairs.
The overall result in the full, 26 letter machine is to form 13 connections, so
that each wire at the entry plate connects to some other wire at that plate. This
pattern of pairing is set by the umkehrwalze where pins are connected in pairs to
reflect back the electrical connections. The rotors simply permute the 26 wires
which have this reflection property.
Between the entry plate and the lamps and keys there is a plugboard called
the 'stecker board’ (following custom we shall generally use the German word).
When the commercial Enigma machine was adopted by the German armed forces,
the plugboard was added to make it more secure. In the absence of plugs, the
connections go straight through, ignoring the plugboard, as shown in Figure 3
for wires C and D. A cable with a plug at each end carries two connection and
when it is plugged into two jacks it interrupts the straight though connection,
crossing the two connections over, as shown for wires A and B. As a result,
the stecker board performs a permutation which is combined with those of the
rotors. In principle, any permutation could have been used here by employing
single connections and plugging ali 26 wires, but this would have been tedious
to do and subject to error. The arrangement actually used does not need ali
jacks to be occupied. At first, only two cables were used (4 jacks) but the
number increased as the war progressed, presumably to inc.rease security. It is
an interesting fact that the largest contribution to the key space is not obtained
with the full number of 13 cables, filling ali the jacks.
When n cables are used, the number of possible pluggings (its contribution to
the total key siže) is 26!/(26 — 2n)!n!2”. This has a maximum at n = 11, leaving
4 jacks unplugged. Starting with only two cables, the number was increased until
112
10 cables was the standard, with six letters unplugged. The key siže for 10 cables
was 1.507 X IO14, compared with 2.056 X 1014 for the optimal choice of 11 cables,
only a small reduction from the maximum.
Figure 2 shows in simplified form the circuit diagram of an Enigma machine.
Pressing a key (such as F in the figure) puts the battery voltage onto one of the
26 wires which makes a connection, via the stecker board, entry plate, rotors R,
M and L to the umkehrwalze then returns this connection on a different wire
via L, M and R, then the stecker board, to reach a lamp (such as A in the
figure). An immediate consequence of this plan is that a letter can never be

enciphered into itself, which is a clue for the cryptanalyst. Also useful is the
propertv that the key to lamp permutations are involutions - if A enciphers to
B, then B enciphers to A. For the two keys A and F shown in the figure, either
will light the other’s lamp. Consequently there is no need to provide a lever to
change from encipher to decipher. The machine, given correct settings and rotor
movement, will decipher a ciphertext in the same way as it was enciphered.
Note especially that the stecker board permutation itself is an involution,
resulting from the way it is produced by swapping connections with each wire.
By this we mean that if wire X is plugged to Y then at the same time wire Y
is plugged to X. It is important to remember that this is quite different from
the inherent involution property of the whole machine, which still would be
preserved if the stecker board was any permutation (like a rotor). This plug
board involution was an important source of weakness, exploited in the bombe,
as we shall see.
Figure 2 shows ‘level 1’ where the wires enter the stecker board and ‘level
2’ where they leave it to enter the scrambler. The permutation from keys to
lamps is expressed at level 1. On the other hand, if we chose to ignore the
stecker board, the permutation due to the rotors and umkehrwalze alone (the
scrambler) is expressed at level 2.
S~1C~RR~iCR~M M~xCm~lL~1ClUC~lLCl~m MCm~rRCrS

S - stecker board R, Af, L - right middle and left rotors
U - umkehrwalze R, M, L - positions of R, Af, L rotors
Figure 4. The Enigma permutation.
The cipher/decipher permutation of the alphabet, from key to lamp, can be

expressed as a product of 17 elementary permutations shown in Figure 4, where
the stecker board, rotor and U permutations are S, R, M, L and U respectively
and C is a cyclic shift due to rotor movement. For example, the term CR~M is a
cyclic shift by R — M places where these letters in the index position refer to the
113
CTVKOLOGiA April 1999 Volume XXIII Number 2
V—2C------ 6
---
r R\NClPLE
Figure 3. Principle of the stecker board.
rotational positions of the respective rotors. Some of the Polish work depended
on the mathematical properties of this expression. The bombe handled these
permutations in a purely mechanical way, so we shall not need the refer again to
the expression in Figure 4.
KEY SPACE OF ENIGMA

A ‘basket’ of 5 rotors was provided. Any three could be assembled in any order
on the shaft shown in Figure 1, giving a choice of 60 rotor sets. Fortunately,
the umkehrwalze was usually a fixed permutation, though a different model did
appear. It was nicknamed ‘Uncle Walter’ at Bletchley Park and, because it was
used unwisely, its wiring was deduced that same night, but it was never used
again. In mid 1944, an umkehrwalze named ‘D’ was introduced in which the
connections were pluggable. One of these is in the Oslo Military Museum, and
setting its connections appears to need some dexterity. This was one of the two
developments that threatened the cryptanalytic effort late in the war; the other
(Enigma-Uhr) will be described at the end of this paper. But for the most part
a single, known umkehrwalze was used.
The permutations used in the wiring of the rotors are not regarded as part
of the key because there were a limited number of them. The wirings in the
early rotors were discovered by the Polish team. New rotor wirings could then
be deduced or found by capture, so these were known at ali times.
A ring on each rotor could set the position of the ‘carry’ which triggered the
114
movement of the next rotor at one of its 26 positions, giving 262 possibilities, and
this ring also carried the letter markings. When the process of setting the wheel
positions, using these markings, was an essential feature of the cryptanalysis, the
ring settings were among the unknowns. This was important to cryptanalysis
until May 1940, but afterwards the initial wheel positions were determined by
using bombes, and the ring’s contribution to the key spare could be ignored.
The stecker board plugging, using 10 cables, presents 1.507 x 1014 possibilities,
a very large contribution to the key space. This permutation is a main concern
of the bombe.
When a message starts to be sent, the initial positions of the rotors form
the message key, chosen ‘at random’ for each message. The message key is
transmitted at the beginning of the session in an enciphered form. At the outset
of Bletchley Park’s work, the method of enciphering the rotor positions was a
weakness that could be exploited, but new techniques closed off this avenue in
May 1940, so that the 263 initial rotor positions became part of the key to be
discovered.
The key space was then 60 x 262 X 1.507 x 1014 x 263 = 1.074 X IO23. The
number is large enough to prevent anv possibility of a search over the whole
space. Nothing could be done with the choice of rotor set, so in the absence of
special information ali 60 rotor sets must be tried. The 262 ring settings were
made largely irrelevant by the way the bombe was used. Its master stroke was
to remove the enormous contribution from the stecker board, leaving it to search
over the 263 initial rotor settings as well as the 60 rotor sets, a heavy task, but
feasible.
Other versions of Enigma variously used a moving umkehrwalze, a basket of
8 rotors, 4 working rotors and rotors with multiple places to turn over the next
one, but not ali these in the same machines! We shall stay with the army and air
force Enigma that has 3 moving rotors, a fixed umkehrwalze and one turn over
position on each rotor.
The British used a machine similar to Enigma, called ‘Typex’ with multiple
turnovers. When modified, it could run the decipherments of Enigma messages
after the keys had been found, or test possible key settings.
THE BASIC BOMBE

Alan Turing is credited with the initial concept, but the extent of its derivation
from the earlier Polish bombe, from which it got its name, is uncertain. Gordon
Welchman implies that he hit on the basic concept at the same time. Many
others were involved in the engineering of the bombes and developing their use.
115
«V>>iOLOGi£ April 1999 Volume XXIII Number 2
In this section we describe an early concept for the bombe and later an important
improvement.
To use the bombe, a ‘crib’ was required. A crib was a piece of text known
to be contained at a certain region of a cipher message. Great familiarity with
message content, military conventions and terminology was acquired at Bletchley
Park. When breaking was regular, cribs were easier to obtain and refine. The
routine nature of many of the messages helped a great deal, and also the need to
spell out numerals. For example, a report might regularly contain the sequence
NULLNULLNULL, a powerful crib, as we shall see. It seems that routine weather

reports and forecasts were a good source of cribs. The words used to get a very
early break on D-Day, 1944 were WETTERVORHERSAGE BISKAYA - weather
forecast Biscay. During one period in 1943, the best cribs for a naval network
were warnings of mines in positions stated on a grid. To improve the cribs and
keep them coming the RAF laid new mines, avoiding grid positions with a 0 or
5 in them, because these could have more than one spelling!
To be useful, the crib must be aligned with the ciphertext. Knowing the risk
of starting messages with standard sequences, the Germans began with a variable
amount of rubbish or quatsch. By sliding the crib around its expected position,
its real place was sought. The rule that a Ietter differs from its ciphertext value
could eliminate some choices, hopefully leaving few alternatives. Long cribs
would give precision of location. An example of a long crib at the beginning of
a message is
GESAMTEINSATZXSULTANXAMXEINSXEINSYYEINGESETZTYY
Typically the rotor choice, ring setting and stecker plugging were changed
daily. A rule was observed that the same rotor did not appear in a given position
on successive days, reducing rotor choice to 32 on the day following a known
choice. Also, the rotor set was not repeated during a given month, making the
choice even more restricted when it was late in the month. Lazy compilers of key
tables gave extra clues occasionally, but beyond this it was necessary to repeat
bombe runs for ali remaining rotor choices.
To illustrate the use of the bombe I shall assume that a valid crib has been
obtained and its location in the ciphertext is known, as well as the choice of the
rotor set. What remain to be found are the stecker board plugging, the ring
settings and the initial rotor positions.
As an example I have taken the sequence ROYALHOLLOWAYCOLLEGE as
the crib and used an arbitrary sequence for the ciphertext. The correspondence
of plain and cipher letters is shown in Figure 5.
116
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
R O Y A L H O L L 0 w AYCOLLEGE
C Q N Z P V L I E u I KTEDCGOVW
Figure 5. A message and its ciphertext equivalent.
The numerals give the sequence of rotor positions from an unknown starting
point. For example, the correspondence between L and G which occurs at letter
17 is for rotors advanced by 10 steps from the correspondence between L and

O at letter 7. When we speak of the 263 initial rotor settings we mean their
positions at the start of our crib - or the part of the crib we shall actually use.
If we had been dealing with a simple substitution cipher, the repeated L’s
and O’s in the crib would, as in ‘pattern words’ reveal its presence very strongly.
We can do something similar with the changing substitution, noting that this
type of analysis need not depend on the stecker board permutation if it deals in
equalities of letters rather than their identities.
Looking now at the relationships between letter pairs in our crib, each rela-
tionship is bi-directional. For example, the first letter pair means that at this
rotor position R enciphers to C but equally it means that C enciphers to R. The
complete set of relationships can be shown as a undirected graph with the letters
concerned as vertices and a cipher relationships as edges. The graph is shown in
Figure 6, where the numbers attached to each edge show where the relationship
of 2 letters lies in the sequence of permutations generated by the rotors, counting
from the first letter of our crib as T.
Figure 6. Graph of plain to ciphertext correspondence.
This graph has three kind of features that will be significant in what follows.
Firstly there are loops such as EWLI. Then there are ‘tails’ attached to the loops,
such as the long tail LGVH. Finally there are isolated pieces such as KAZ which
do not have loops. We shall use the term ‘closure’ for such a loop, following
117
CTVKOLOGA April 1999 Volume XXIII Number 2
Bletchley Park practice. The graph itself was called a ‘menu’.

There are four factors which make a crib effective:
1. It should be an accurate copv of a part of the plaintext

2. It should be long, so that it can be accurately located, within very few
(preferably just one) possible position in the ciphertext.
3. Some part of the crib (the menu) can be selected to make a graph like the
one above with many closures.
4. If there are less than four closures, the number of letters in the menu should
be as large as possible. The reason for this is connected with the diagonal
board and will be explained later.
We have been very fortunate in our example in getting three closures, namely
CEL, OEL and EWIL. In the basic bombe we shall find that only closures are
useful. The repeated Ls and Os in our crib were the cause of our luck. Fortu-
nately, German messages often had repeated letters. Cipher letters as well as
plain text letters are used in the graph.
In the bombe, ali 263 starting positions are tried, and the sequence of permu-
tations generated below level 2 (see. Figure 2), which we called the ‘scrambler’ in
that figure, is mechanised.
A typical bombe set up has at least 12 separate sets of rotors, to simulate 12
different stages of an encipherment. Instead of stacking the three rotors together
as in Enigma itself, each rotor fits flat on the front surface of the bombe’s cabinet,
to make rotor changing easier and improve the mechanics. Figure 7 shows how
the 12 sets of three rotors might be laid out on the bombe’s front panel, with
the fast moving ‘right’ rotor at the top. Connections behind the panel join three
rotors together to make them function like an Enigma’s three rotors and its
umkehrwalze. In practice, bombes with 36 sets of rotors were used, and these
could have been running three separate searches at the same time. For an urgent
job, 20 bombes of this siže could search for ali 60 rotor choices at the same time.
The rotor’s connections to contacts on the panel are in the form of wire
brushes, as used for reading punched cards, which slide over contacts on the
panel surface. Four rings of 26 contacts are needed for the ins and outs of
the connections toward the umkehrwalze and back from it to the ‘entry plate.
Each set of three rotors with its back panel wiring produces the full 26-wire
permutation from entry plate to the umkehrwalze and back to entry plate of an
Enigma machine. In this way it reconstructs the permutation below level 2 of
an Enigma with a specific rotor set and rotor positions.
118
In the lower part of Figure 7, the connections between the bombe’s rotors
are shown, in outline. If we start with 26 wires on the left, these enter on the
third ring of contacts and leave, permuted by the R rotor, on the inner, fourth
ring. From here they go to a similar arrangement on the M rotor, then the
L rotor, From here, the 26 wires are permuted according to the wiring of the
umkehrwalze and enter the L rotor again for the return journey. The outer
two rings of contacts are wired in each rotor in the same wav as the inner two,
and perform the inverse permutation on the returning circuit. Finally the 26
wires leave from the outer ring of the R rotor. Note that the whole device is
symmetrical in operation. Entering from the right and leaving from the left
would have exactly the same effect, due to the inherent involution property of
the rotor set’s total permutation - a reflection of the involution property which
is inherent in the umkehrwalze wiring. Note that these ‘Enigma substitutes’ give
straight-through wires representing the layer 2 permutation by connecting 26
wires to 26 other wires, whereas the Enigma itself was single ended, representing
the (involution) permutation on the 26 wires at its entry plate.
oo oooo
oo oooo
oo oooo
oo oooo
0o oooo
oo oo0o
Figure 7. Wiring of bombe rotors.
The bombe will run through the 263 possible initial rotor positions to find
one consistent with the crib. (It does not recognise the slight oddity of the carrv
mechanism which removes some of these from consideration.) These rotors have
119
to rotate at the highest rate possible, to get through the search quickly. The top,
or ‘right ’ rotors move as fast as possible while the others step on at suitable times.
Since the top, or R rotors move continuously and cannot stop, probably some
of its positions are lost when the other rotors step on, but this does not matter
provided it scans at least 26 positions for each of the 262 states of the lower
rotors. The entire run for one rotor set took about 15 minutes, so the runs for
ali 60 rotor sets took rather more than 15 hours, allowing for the rotor changes.
In some bombes, three searches could be done in parallel and 60 rotor choices
took somewhat more than 5 hours. By careful planning of a sequential search,

the 60 sets could be reached with only one rotor in each set changing between
successive runs - like the schemes used in change-ringing bells. In practice there
were many problems and bombe time for breaking a key was more than these
estimates suggest.
How does the bombe use the relationships shown in the graph of Figure 6 to
discover which initial positions can be valid? This is the essence of the bombe’s
ingenuity, which overcame the stecker board and determined its plugging at the
same time as it found the initial setting of the rotors.
Figure 8. Plugging vvith three loops (basic bombe).
Figure 8 shows part of the same graph as Figure 6 in a different notation which
relates to bombe hardware. When this graph was given to bombe operators for
them to set it up it was called a ‘menu’ Here the squares indicate rotor sets in the
120
relative positions shown numerically, ranging from 7, for the O-L relationship, to
20 for E-W. The connections shown are cables of 26 wires which connect the entry
plates of the R rotor sets. The way in which cable connections are looped via
pairs of connectors on the entry plates ( a kind of double 26-pin socket) is simply
a convenient way to deal with an indefinite possibility of multiple connection. For
example, letter E has 3 cables which join together the entry plates in the rotor
sets at positions 9, 14, 18 and 20. Ganged sockets were also used to facilitate
the plugging of the menu.
The part of Figure 6 which is represented here is only the part which includes
closures, namely the letters C, E, W, O, L and I together with their ‘rotor’
relationships at relative rotor positions 7, 8, 9, 11, 14, 16, 18 and 20. In the
bombe, when it is plugged up with this ‘menu’, the 8 rotor sets are initialised
in these relative positions and, during the running over 263 rotor states, we
hope that they will maintain these relative positions so that thev represent the
actual connections that would have been present in an Enigma. But we do not
know when the middle rotor (M) moved on, so we have to guess that it did not
intervene (in our target Enigma) between these movements of the right rotor
(R). The probability that we are lucky in this respect is 1/2, since we are hoping
that the 13 transitions between positions 7 and 20 do not include a movement
of M. By making this guess we avoid having to know the ring settings - a factor
of 262 in the key space.
The total špan of the part of the crib used in the menu should be short, to
avoid turnovers. Long cribs have an advantage in determining the location of
the crib in the message with very few alternatives, but the chance that there is
a wrong letter in the ciphertext increases. If long cribs were possible, distant
letters might be used in the menu. For example, suppose we are guessing that
there is no turnover in the špan 7 to 20, then it is safe to assume there has been
just one turnover of the middle rotor in places 26 + 7 to 26 + 20, with only a
small probability of the left rotor moving.
The connections shown have a complex result, which can be interpreted in
terms of consistency of hypotheses. Consider the letter E in the crib and the 26
wire cable which represents this letter. The letter X in this cable represents the
hypothesis that, in the stecker board, E is plugged to X.
In Figure 8, the letter E is represented by 3 cables of 26 wires which join
entry plates on the bombe rotor sets at relative positions 9, 14, 18 and 20, since
letterE appears at these positions in the crib and its cipher equivalent. We are
looking for a letter which satisfies the rotor permutations at these positions.
It will be one of the 26 wires in the cable and will transform into letter C at
position 14, letter W at position 20, letter O at position 18, and letter L at
121
CTVPiOLOGA April 1999 Volume XXIII Number 2
position 9, as the crib and its cipher equivalent teli us. Therefore, this single
wire, if it satisfies ali the permutations set up by the bombe’s rotors, will join
up with its corresponding wires in ali the letters represented here. These are
wires in the cables and therefore represent letters at level 2, satisfying the crib’s
relationships in the ‘scrambler’. Cable E is letter E in plain or cipher (level 1)
and wire X which satisfies the scrambler relationships is letter X on the other
side of the plugboard. When the bombe reaches the rotor setting at which ali
the relationships of the menu are satisfied, one wire in each cable (the correct
plugging hypotheses) will interconnect with ali the other correct wires and with
nothing else. But we can only detect this result if we know what happens to ali
the other wires. This is where the reductio ad absurdum comes into play.
Only closures are of interest in this basic bombe. A graph which was a tree,
with no closures, would generate a tree connection for each of the 26 wires and
teli us nothing. But each closure brings a set of 26 wires back to connect with
itself in a permutation, for example the ECLE closure which runs through rotor
sets at positions 14, 16 and 9 in succession.
Each wire of one of these cables represents an hypothesis that the letter this
wire represents is part of the solution of the cipher. The individual closure in the
plugging of Figure 8 represents by its connections through the rotors a test of the
consistency of the various hypotheses. So a wire which goes through the ECLE
closure and returns to the same point is consistent with this part of the menu.
If it is consistent with the other closures EOLE and EWILE this single wire
joined up with its fellows in ali the six letters has no connections to other wires.
So total consistency is represented by an isolated wire, evidently an hypothesis
worth following up!
This is how the bombe is wired up to search for the initial rotor positions
consistent with this crib. When one is found, there is just one such connection
as the one we have described, linking one wire in each of the cables employed and
joined to no other. Call it the ‘solution’ wire. The rotors begin the search in the
relative positions shown by the numbers, and maintain these relative positions
during the search, so that ali possible positions at the start of the crib are tried.
THE REDUCTIO AD ABSURDUM

Suppose that such a solution has been found, what happens to the other wires?
Figure 9 shows an example for the closure comprising C, E and L, simplified as
for a six letter alphabet. The permutations of the scrambler have been shown
(arbitrarily in this example) for the relative positions 9, 14 and 16 of our crib.
The hypothesis that C plugs to A (C" = A) is the top left hand wire. Tracing
122
the implications of this hypothesis round the closure gives E' = E, L' = B and
C' = C, clearly an absurdity, since the solutions for cable C cannot be both A
and C. ri +14-
The paths taken by wires in this diagram can be summarised by tracing

through the various rotor sets, beginning at cable C, thus:
Cable CE L C E L C E L C E L C
Wire A E B C D A F B E D C F A
B F C B -
E A D E -
As a result of the reductio ad absurdum, the wires identified in the C cable as

A, C, F and D are connected together. For the other wires, wire B in cable C is
a possible solution since the connection after one circuit of the closure returns to
B. Similarly, wire E returns to E. In our example we therefore have C' = B and
C' = E as possible pluggings, and for each case we could deduce the pluggings of
123
CTVKOLOGA April 1999 Volume XXIII Number 2
E and L by following the connection round the closure. Solutions can be detected
by their lack of connections to any other wire.
For just one closure of the graph, the non-solution wires will be joined in
groups corresponding to the way in which the overall permutation of the closure
breaks up into cvcles. But a randomly chosen permutation has on average just
one cycle of length one, so with one closure there would, on average, be a false
solution wire at every step of the bombe!
The probability of false stops can be reduced by having more than one closure.
With enough closures, the result will almost always be a solid connection together
of ali 25 wrong hypotheses. What we have is a reductio ad absurdum machine,
with ali the false hypotheses implying each other. From a false assumption ali
the other false conclusions can be deduced.
This is what makes the bombe such an extraordinary machine. Whereas other
electro-mechanical (and electronic) logic machines manipulate truth values of
known expressions, this one checks the consistency of a set of logical relationships.
There is no certainty that a solution is the correct one, but we hope to keep
the number of false stops small and this requires sufficient closures in the graph
which underlies the connection pattern of the bombe.
With one closure, we are looking for a cycle of length 1 in the overall per
mutation and we know that, on average, one such cycle (a fixed point of the
permutation) will occur in a randomly chosen permutation. So in a search over
263 starting positions there will be 263 wrong results in addition to the correct
on we are looking for. Not a good start! Each extra closure reduces this number
by a factor of 26. Only if we have four closures in our graph does the number of
wrong results approach unity, giving us a reasonable hope that the correct result
will be found. The operators of the bombes could live with some wrong results
because it was easy to test the rest of the message and verify whether a good one
had been found, but too many would make progress very slow. Unfortunately,
four closures are more than we can confidently expect. It is difficult to be certain,
but a few trials seemed to show that two or three closures could often be found,
but four closures were rare.
The solution, when found, manifests itself as a connection pattern with a
single isolated wire. Any of the 26 wires can be sampled anywhere in the circuits
of Figure 8, but well-connected letters, such as E and L in our example, will prove
best for reasons that appear later. At each instant of the search period, either
there is no solution and ali 26 wires are connected together, the commonest case,
or 25 are connected and one is disconnected, which means a candidate solution.
But we need a mechanism to find what connection pattern exists. A voltage
is applied to just one wire of the chosen cable. When there is no solution, the
124
voltage usually appears on ali the other wires. When there is a solution, in nearly
ali cases we have not sampled the chosen wire, so ali wires have a voltage except
the chosen one. This is the essence of reductio ad absurdum, finding the correct
hypothesis by disproving ali the alternatives. Any wire which joins round the
closure to a different wire means that one assumption about the stecker board
plugging implies a different plugging, which must be false. If, bv chance, the
voltage happens to be given to the solution wire, it will be on none of the others.
The tests are a little tricky, but they can be done. When an apparent solution
is found, the bombe stops, but the fast moving rotor does not stop at once.
However, the positions of the two lower rotors can be read off, and the identity
of the solution wire noted, as well as the instantaneous position of the fast rotor
at the moment when the solution has appeared. As the bombe runs through its
263 positions, it tests 26 alternative hypotheses for each letter of the crib that
has been used, at each position.
Detection of the solution at first used fast (telegraph) relavs, but when the
bombes were run at a higher speed to handle the four-rotor Naval Enigma, elec-
tronic detectors were introduced.
We have glossed over one difficulty by assuming that the rotors keep their
relative positions during the search. If the middle rotor movement occurs within
the range covered by the menu, this fails. So the range used in the menu should
be as small as possible, yet we want to find three or four closures if possible, or
the number of false stops will be troublesome.
When a solution is found, it conflrms the crib and the choice of rotor set and
gives the start position of the rotors, and the identity of the solution wire implies
the plugging of ali the letters in the connected part of the menu, which in our
example has six letters joined in three closures. The other ‘tail’ connections on
the graph within the period hypothesised to be free of middle rotor movement
(8-20 in the example) namely those to U, D, G, and V give the plugging of further
letters. These largely complete the information needed about the stecker board.
THE ENHANCED BOMBE
If the stecker board had been an unrestricted permutation, little more could be
obtained from cribs, but we know that it is an involution, that is - if A permutes
to B then B permutes to A. This feature is not inherent in the principle of
the machine, like the involution of the cipher as a whole. It is due to the way
the designers chose to implement the stecker board, with each two-wire cable
performing one transposition.
Gordon Welchman saw that the involution property of the stecker board could
be used to add extra connections to the bombe’s existing array, thus decreasing
125
the number of false stops. To see how it worked, we must recall the significance
of each wire in each cable.
Wire X in cable C represents the hypothesis that the stecker board plugs letter
C of plain or cipher text to wire X that enters the scrambler. If this wire is part
of the solution wire, the hvpothesis is not disproved by the relationships of the
menu.
So the wire E of the Enigma being modelled, which derives from a lamp or key,
at level 1 of Figure 2, connects through the stecker board to solution wire X at
the entry plate, level 2 of Figure 2. We now know one of the stecker connections,
and if there are 6 letters, as in this example, there are six known pluggings, for
letters C, E, I, L, O and W. And if we know these, we know up to six more, from
the involution property, for example if C was connected to Q, for example, then
Q was connected to C. How can this property be exploited?
a I? c d e
Figure 10. The diagonal board (5 letters only).
If the solution wire in the bombe is wire X in cable C, as we earlier supposed,

this is consistent with wire C in cable X, and the two wires can be connected
together as a further test of consistency. If it is a true solution, the wires are
still disconnected from ali else and the new connection has had no effect. If it is
wrong, which is most often the case, then it makes a new connection and spreads
126
the message of inconsistency wider, helping to ensure that inconsistent choices

will show up.
Figure 11. Connection of three loops to the diagonal board.
Plugging individual wires like this would be slow and tedious, so Welchman
invented the ‘diagonal board’ to make it routine. Figure 10 shows a simplified
example for just 5 letters, whereas the full version had 26. There are sockets for
each letter (A to E) and each cable has a wire for each letter. Cable A, letter
‘B’ is connected to cable B, letter ‘A’ and so forth. With 26 letters there are
127
»VPiOLOGA April 1999 Volume XXIII Number 2
325 connections since each cable makes 25 connections, the diagonal letter not
being connected. But only as many connections are made as are provided for the
cables actuallv used, and not ali 26 letters appear in the crib. Figure 11 shows
how the diagonal board would be connected for the 6 letters of our example that
make up the closures of our graph. Each letter used in the former plugging is
now also connected to its socket on the diagonal board.
With 6 letters in the crib of our example, the board only makes 15 connections,
which is about half of the number of connections in a single closure of the basic
bombe, so it seems that not much has been gained. But now we can extend
the graph of the crib, because it is no longer true that only closures provide
connection paths. We have used 8 of the 12 rotor sets that the bombe provides,
so we can add 4 more within the compass of a typical bombe. From the graph
of Figure 6, new letters can be added, keeping the length of the crib section that
is employed as short as possible. ‘Tails’ to join the letters U, D, G and V are
an obvious choice, keeping the špan from positions 7 to 20. If there were more
rotor sets in the bombe, the disconnected relations A-K and T-Y could be added
- but are they useful? The new connection pattern is shown in Figure 12. The
choice of additional letters has been made in a way which avoids spreading over
a longer set of positions since, with positions from 7 to 20, there is alreadv a 50%
risk that a turnover of rotor M will prevent a solution being found.
If the extra two rotor sets had not been used for the disconnected portions,
we would have 10 letters in the menu and the diagonal board would give 45 extra
connections, more than would be given by an extra closure in the basic bombe.
An analysis of the actual effectiveness of the diagonal board is given in a
separate note [9]. For example, with a 10 letter (connected) menu the diagonal
board reduces the number of false stops by a factor of 60, the equivalent of 1.26
extra closures. This is less than simplv counting the number of extra connections
would indicate, but still a very useful improvement. With 13 menu letters the
diagonal board is more effective than two extra closures and with 15 it is more
effective then three.
It seems that the disconnected parts of the graph of Figure 6 do not contribute
much to its usefulness. For example the K-A connection will not have an effect
unless both K and A are implied pluggings for letters of the menu. But it is
also clear that the connected parts of the graph outside the closures, used in
conjunction with the diagonal board can greatly reduce the frequency of false
stops of the bombe.
As the war progressed and the number of keys and frequency of key change
increased, more and more bombes were manufactured and put to work. Even-
tually there were about 200 bombes in operation, 24 hours per day, with 2,000
128
Davies The Bombe - A Remarkable Logic Machjne
Figure 12. Enhanced use of the diagonal board.

CTVPiOLOGA April 1999 Volume XXIII Number 2
members of the VVomen’s Royal Naval Service (‘Wrens’) as their operators. First
they were in outposts of Bletchley Park in 3 neighbouring villages, then in large
buildings in Stanmore and Eastcote, in the Northern outskirts of London. The
Eastcote buildings remained occupied by GCHQ for many years after the war
ended. The scale of the manufacturing operation is suggested by the number
of rotor brushes in use, which was 15 million! A similar effort in bombe man-
ufacture and operation took place in USA, concentrating on the 4-rotor Naval
Enigma, where the search was over 264 rotor positions and the extra speed of
operation was an advantage.
AFTER THE FIRST BREAK
When the bombe stops and indicates a solution, it gives the stecker board wiring
and the initial rotor positions for a given crib and confirms the choice of rotor
set. To be sure that it is a valid solution, the whole message can be deciphered,
subjećt only to finding the turnover positions (ring settings) on the R and M
rotors, which is easily done by noting where the returned plaintext breaks up.
Now the stecker board wiring, rotor set choice and ring settings may be valid
for a whole day of transmission under the given key and there may be hundreds
of messages for which only the initial rotor positions are needed to reveal their
intelligence. This information was present, in enciphered form, in the indicator
at the start of the message. In many cases this cipher could be solved, but
when it could not be, ali subsequent messages would need cribs and bombe time,
which was in short supply. This is where the other machine ‘Baby’ čame into
play. For some time, it was an important tool for breaking naval traffic, where
the indicator was problematic.
A single scrap of text: ‘EINS’ was employed as a “mini-crib” for ali messages,
This would be the numeral ‘ONE’ in a spelled number. With the aid of the
known key material, only the initial rotor positions were unknown. ‘EINS’ was
enciphered under ali 263 initial positions and the resulting 4 letter ciphertext
punched out on standard punched cards, together with the initial position. I
guess that it was done in 4 runs with a bombe-like device continually enciphering
EINSEINSEINS... in one of its four phases.
Then the punched cards were sorted to alphabetic sequence of ciphertext. Ali
the messages to be broken were cut into 4 letter fragments in ali four phases and
these were punched up with message identity and fragment position, these cards
also being sorted into alphabetic sequence of ciphertext.
A standard punched card ‘collator’ could take the two streams of punched
cards in alphabetic sequence and determine anv occurrence of one of the 263
130
encipherments of EINS in the given cipher texts. The rotor positions for crib
and ciphertext being known, the initial position of the rotors is obtained.
Assuming that these 4-letter ciphers were random, each would occur only once
in 264 positions of random text, so it is reasonable to expect a match with any
of them at intervals of about 26 letters, or a few times in each short message. If
the use of EINS as a mini-crib is valid, the Babv is a substitute for a bombe in
discovering the initial rotor positions of numerous short messages. With a 100
card per minute punch, a run takes three hours, then the cards can be used until
the Enigma keys change.
PRACTICAL EFFECT OF THE DIAGONAL BOARD
In a separate paper, “Effectiveness of the diagonal board” [9], I have made an

estimate of the extent to which the diagonal board reduces the number of stops
of the bombe, ali but one of which must be unacceptable solutions (though they
comply with the logical relationships of the menu). The factor of improvement
is a function of the number of letters in the menu, assuming that it is based on
a connected graph.
A long crib has advantages. Matching a crib to its place in the ciphertext by
using the rule that plain and cipher letters must differ will work best for long
cribs. About 26 letters would have a probability of a false match in any one
position of about 0.36. A firm match has two advantages - it avoids the need
to repeat the whole process for several locations and it gives confidence that the
crib is there and it is worth the investment of bombe time. But a 26 letter crib
used as a menu would ensure that the middle rotor moved somewhere during its
length, negating the whole procedure. However, a long throughout to makecrib
can have a short part of it used for the menu, which was usually the case.
An ideal outcome is a 27 letter crib from which two menus can be derived, one
from letters 1-14 and one from letters 14-27. In a 27 letter crib, only one turnover
can occur. One of the two derived menus is certain not to have a middle or left
rotor movement. Unfortunately there may be only a poor chance of 3 closures in
the graph derived from 14 letters and their cipher equivalents. But, with just 2
closures in a connected graph and 11 menu letters, the expectation of 676 false
stops without a diagonal board is reduced to 4 stops (see [9]), which is acceptable.
It is hard to estimate the relationship between menu length and number of
closures, but my guess is that the bombe would have had little success without
the diagonal board, whereas the enhanced bombe had a good chance of finding
workable menus. Short portions of cribs, taking a chance on middle rotor move
ment, or longer ones split in two, brought the chance of success into the useful
range.
131
THE ENIGMA-UHR
In the Enigma machine there are two distinct kinds of involution. A permutation
is an involution if whenever A permutes to B then B permutes to A. The basic
involution of Enigma is the permuted alphabet which it employs at each setting
of its rotors. This is inherent in the design of the machine, which reflects circuits
back at the umkehrwalze (which produces an involution) over the same 26 wires.
To avoid this property would require a complete redesign, and it was extremely
unlikely that it would happen.

The other involution is in the plugging of the stecker board, enforced by the
two-wire cables it uses (see Figure 3). This is not inherent in the design. In fact
the stecker board is in the same position, in this respect, as the rotors, whose
permutations are not involutions. Neither affects the basic involution property
because permutations related by a transformation, such as X and T~'1XT have
the same cycle structure. The German cryptologists could have changed the
stecker board to prevent the diagonal board being used, which would have greatly
reduced the usefulness of the bombe. Of course they had no idea that this was
possible, though they might have reasoned that any involution property could
possibly help a cryptanalyst.
However, they almost did make this powerful change by accident. stecker
board re-plugging is a slow business and fraught with errors. To enable very fre-
quent changes, late in WWII they introduced the Enigma-Uhr (meaning ‘clock’
or ‘dial’).
The device was a box with 20 stecker board cables emerging from it. On its
top face was a large rotor which could take any one of 40 positions, though behind
it there were only 20 contact positions, so it operated as two independent rotors
each used in half of the positions. The cables were marked in two sets, with red
or silver dots on them, each set numbered 1-10. In one position of the rotor, red
and silver plugs were connected according to their numbers, so a specific stecker
board plugging could be set up which, in this position, was consistent with other
Enigmas which did not yet have an Enigma-Uhr.
In other positions, new stecker board pluggings would be obtained, but would
they be involutions? Because of the way the device was constructed, this would,
in principle, be possible. There are several examples of the Enigma-Uhr in the
Defence Museum in Oslo. Frode Weierud has traced the wiring of one of them
and found that 10 of its positions are involutions (including the neutral position),
and 30 are not.
It seems that the setting of the Uhr was part of the message key. In other
words, a setting was chosen at random, then enciphered in some way, like the
132
rotor positions, and transmitted in a preamble to the message.

Since those who designed the Enigma-Uhr did not try to keep the involution
property, they had stumbled on a countermeasure to the diagonal board. We do
not know how this affected the work of Bletchley Park and what new techniques
were devised to meet the challenge.
THE POLISH BOMBE
In the paper by Marian Rejewski [8] there is a rather brief description of the
Polish bombe and how it was used, which must be taken as definitive, as far
as it goes. But it raises many questions because of its brevity. If we could
understand this bombe fully it would be possible to see how far the Polish team
had anticipated the Bletchley Park bombe.
The message key of the Enigma consists of three letters which specify the
initial rotary positions of the three rotors. It is clear that the Polish bombe
attacked the transmission of the message key rather than the message itself. The
British bombe attacked the message itself and thus was immune to changes of
initial procedures, which the German forces could alter at low cost.
At this time, the message key was itself enciphered by Enigma and was du-
plicated before encipherment to make six letters. The rotor settings for this first
encipherment were sent in clear, followed by the six letters produced.
Rejewski gives the following example of the preambles of three messages on
which their bombe could work:
RTJ WAH WIK
DQW DW,J MWR
HPN RAW KTW.
The letters before the comma are the initial rotor settings that were used to
encipher the following six letters. Ali three messages will have the same choices
of rotors in the three positions, the same pluggings and the same ring settings,
but these things are unknown. However, there were at this period only three
rotors to chose from, giving six choices and ali these would be tried, so we can
assume that the chosen rotor in each of the three position is known.
In each of these preambles the two occurrences of the letter W are encipher-
ments of the same plaintext letter, namely one of the letters of the message key
chosen in this case. Also, each preamble has this letter Win a different place of
the message key. Of course, Wcould be any letter but it will be easier to stick to
W in our account.
Rejewski wrote “With enough cipher material it can happen that on a given
day three messages will be found with keys as in the following example.” The
133
C3VPCOLOGA April 1999 Volume XXIII Number 2
point about “a given day” is that rotor choices and ring settings and stecker
board were unchanged during a whole day at that time. I estimate that about
100 messages would be needed to have even odds of success.
The effect of the stecker board is ignored on the basis that, working with
between 5 and 8 cables plugged in, affecting 10-16 letters, the repeated letter,
Win this case, would be unchanged often enough for a reasonable success rate.
The reason that the same letter is used in ali three messages is to minimise the
chance that this encipherment is affected by the stecker board.
The initial settings, RTJ, HPN and DQY give only relative positions of the
rotors because of the unknown ring settings. The bombe enables 263 ring settings
to be explored, in effect. Six Enigmas are run through ali their 263 rotor positions,
“preserving the known mutual positions of the drums.” We assume that these
are the relative positions of the rotors for the six occurrences of the letter W of
our example, as given by the three initial letters and the positions of the W in
the enciphered sequence. RTJ etc. give the rotor positions at the start of the six
letters. They had to guess that a turnover of the middle rotor did not happen
in each preamble before the last of the W letters was enciphered. How the six
Enigrnas are connected was not stated in the paper.
The solution emerges from this bombe when “three pairs of lamps (the same
lamps in each pair) lighted.” This is far from clear, but for each message which is
handled by one pair of Enigmas, at the setting we are looking for, W transforms
in both Enigmas into the same letter, namely the message key letter in this
position. We assume that the W was entered into ali six Enigmas and the
output lamps of a pair were compared for equality. When ali three pairs showed
equality together, a possible match had been found.
In addition to the match of ali three pairs which is sought, there may be false
matches and these will occur in about 1/26 of settings for each pair, or with
probabilitv of 26-3 for the complete match. This is fortunate, because there will
be only about one false match per run.
When the true match is known, this gives the ring settings and enables a
message key to be found for each of the three messages.
QUESTIONS RAISED
Does the duplication in the six letters (the W in the example) have to be in three
different places? There seems nothing in the logic of the method that requires
this, so we have to assume that the example given in the paper had this form
simply to show that any of the three places was acceptable. In fact earlier places
in the triple have an advantage relating to turnover. If repetitions in any of. the
134
Davies The Bombe - A Re mar kabl e Logic Machine
three pairs are acceptable, the number of messages to collect the samples with
even odds of success fališ from about 100 to about 70.
p eri ti o k 0 3
Figure 13. Conjectured Polish bombe.
135
CTVPtOLOGiA April 1999 Volume XXIII Number 2
It does not seem very practicable for the pattern of three output equalities to
be detected by an operator looking at 6 rows of 26 lamps. To make the process
automatic, an obvious move would be to connect ali the wires of a pair except
the W wires (of our example) together and check continuity from one W wire to
another as shown in Figure 13. In fact, ali three such circuits could be put in
series, as shown in the figure so that a circuit is closed at the complete match.
This is a conjecture but it seems highlv probable that an arrangement of that
kind was used.
This looks very much like a menu for a British bombe, so perhaps it can
be seen as a simplified version of such a menu. The graph of the relations for
the duplicate W would be W-X-W, where X is the unknown letter of message
key. There are three such graphs with the letter W in common, making a three
looped graph, which, in the British bombe would mean about 26 false matches.
Assuming that W is unsteckered reduces the problem for the Polish Bombe.
The simplified scheme of the Polish bombe does the same job as the British
version in this special kind of menu and the single W wire, due to the W being
unsteckered, allows single ended Enigmas to be used.
COMPARISON OF THE TWO BOMBES
Attacking the transmission of the message key makes the cryptanalysis vulnerable
to changes in procedure, which the enemy can do with comparative ease. But
the cost of immunity to the initial procedure and attacking the message itself is
the need to find cribs.
The Polish bombe anticipated the British to the extent that it performed
a search over the 263 rotor positions, looking for a logical pattern among the
permutations created by a group of Enigmas spaced out in the “message space.”
In the British case, the origin of the logical patterns was a supposed crib. In the
Polish case, it was the pattern imposed by the repetition of three letters before
encipherment.
The Polish bombe represented the logical relations by electrical circuits with
a defined starting and end points. The British bombe used electrical connec-
tions which could carry currents in either direction, allowing the consequences
of hypotheses to be developed in a way that straightforward “relay logic” can-
not match. In this respect the electrical representation of logical relations by
the bombe was unique and has remained so, nothing similar having been used
elsewhere.
The British bombe made a fundamental step forward and was a remarkable
reductio ad absurdum machine. The Polish bombe did not anticipate those key
136
features which made the British bombe’s design unique and highly effective.
It is possible that the Polish machine was a starting point for the innovative
thinking that led to the British bombe, but we shall never know.
CONCLUSIONS
This paper has described a machine different from anything before or since,
which used electrical connections (not circuits) to represent hypothetical logical

relationships and the electrical isolation of a wire to show that one particular
hypothesis is consistent with ali the relationships. Bletchley Park had two re
markable machines, the bombe and Colossus, of which I believe the bombe to be
more clever in concept, while Colossus pushed technology to its limit.
The bombe was onlv one part of the brilliant attack on the Enigma cipher. The
success of Bletchley Park required intellectual power as much as clever machines.
The correlation and use of the recovered intelligence and the organisation of
cryptanalysis on a factory-like scale were very important factors. But among
ali the work at Bletchley Park, the logical principle of the bombe, due to Alan
Turing and Gordon Welchman, was outstanding.
REFERENCES
1. Davies, D. W. 1982. The Siemens and Halske T52 cipher machine. Cryp-
tologia. 6(4): 289-307.
2. Davies, D. W. 1983. The early models of the Siemens and Halske T52
cipher machine. Cryptologia. 7(3): 235-253.
3 Mache, W. W. 1986. Geheimschreiber. Crgptologia. 10(4): 230-242.
4. Davies, D. W. 1994. New information on the history of the Siemens and
Halske T52 cipher machines. Cryptologia. 18(2): 141-146.
5. Davies, D. W. 1995. The Lorenz cipher machine SZ42. Crgptologia. 19(1):
39-61.
6. Hamer, D. H., Sullivan, T. W. and Weierud, F. 1998. Enigma variations:
An extended family of machines. Crgptologia. 12(3): 211-229.
7. Welchman, G. 1982. The hut six story. London: AllenLane. (Also McGraw
Hill Book Co, US A.)
8. Rejewski, M. 1981. How Polish mathematicians deciphered the Enigma.
Annals of the historg of computing. 3(3): 213-234. (Translated by Joan Ste-
penske).
9. Davies, D. W. 1998. Effectiveness of the diagonal board. To appear in
Crgptlogia.
137
CTVPCOLOGA April 1999 Volume XXIII Number 2
BIOGRAPHICAL SKETCH
Formerly head of Computer Science at the UK National Physical Laboratory.
Helped to build the ACE Pilot moclel computer in 1947-50. In 1965-73 led a
team which pioneered packet switching for computer networks and built a local
network for the NPL campus, with a file server and e-mail. Subsequently applied
cryptography to computer security and after leaving NPL in 1984 has acted as
consultant to banks, pay-TV broadcasters and others and as expert witness in
civil and criminal proceedings. Has written in Cryptologia on the Siemens and
Halske T52 machine, the Lorenz SZ42 machine and other cipher devices. Fellow
of the Royal Society.
138

The Bombe A Remarkable Logic Machine: Cryptologia

Uploaded by

Copyright:

Available Formats

The Bombe A Remarkable Logic Machine: Cryptologia

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Bombe A Remarkable Logic Machine: Cryptologia

Uploaded by

Copyright:

Available Formats

Taylor &Francis

Taylor & Francis Croup

ISSN: 0161-1194 (Print) 1558-1586 (Online) Journal homepage: http://www.tandfonline.com/loi/ucry20

THE BOMBE A REMARKABLE LOGIC MACHINE

To link to this article: http://dx.doi.Org/10.1080/0161 -119991887793

Published online: 04Jun 2010.

Submit your article to this journal B

Article vievvs: 125

View related articles B

a Citing articles: 3 Vievv citing articles B

Full Terms & Conditions of access and use can be found at

produced by an off-line machine called Enigma.

Figure 1. Layout of rotors (schematic).

Figure 2. Wiring scheme of an Enigma.

figure). An immediate consequence of this plan is that a letter can never be

S~1C~RR~iCR~M M~xCm~lL~1ClUC~lLCl~m MCm~rRCrS

The cipher/decipher permutation of the alphabet, from key to lamp, can be

Figure 3. Principle of the stecker board.

KEY SPACE OF ENIGMA

THE BASIC BOMBE

NULLNULLNULL, a powerful crib, as we shall see. It seems that routine weather

17 is for rotors advanced by 10 steps from the correspondence between L and

Figure 6. Graph of plain to ciphertext correspondence.

Bletchley Park practice. The graph itself was called a ‘menu’.

1. It should be an accurate copv of a part of the plaintext

took somewhat more than 5 hours. By careful planning of a sequential search,

Figure 8. Plugging vvith three loops (basic bombe).

THE REDUCTIO AD ABSURDUM

The paths taken by wires in this diagram can be summarised by tracing

As a result of the reductio ad absurdum, the wires identified in the C cable as

Figure 10. The diagonal board (5 letters only).

If the solution wire in the bombe is wire X in cable C, as we earlier supposed,

the message of inconsistency wider, helping to ensure that inconsistent choices

Figure 11. Connection of three loops to the diagonal board.

Figure 12. Enhanced use of the diagonal board.

AFTER THE FIRST BREAK

the Enigma keys change.

PRACTICAL EFFECT OF THE DIAGONAL BOARD

In a separate paper, “Effectiveness of the diagonal board” [9], I have made an

unlikely that it would happen.

rotor positions, and transmitted in a preamble to the message.

Figure 13. Conjectured Polish bombe.

COMPARISON OF THE TWO BOMBES

which used electrical connections (not circuits) to represent hypothetical logical

You might also like