Before trying this you should read the introduction to Virtual Colossus.
This scenario is an intercept of 5668 characters accessed in
Virtual Colossus as "KHcipher" and "KHpatterns". It needs the K2 one-back
1. Getting started.
Click on the "cipher?" drop down menu on the far left hand side. Select
"KHcipher". Ignore "starts?" for the moment and click on "pattern?". Select
"KHpattern". This takes anything up to 15 seconds to load. When it has loaded
zeros appear in the Counts box and the vertical sliders can be moved.
For starters check the cipher text length. On the left hand big switch panel click
underneath the first counter switch on the top row to set it down. Now scroll left
the big right hand panel to show the three big black ganged switches. Click above
the top ganged switch to set it up and to switch the cipher characters, Z, to the Q
bus on the big switch panel. Now continue scrolling the right hand panel to reveal
the master switch panel. Click underneath the black with "LC" (letter count) above
it. This should result in 2042 appearing as the left hand (No 1 counter) count in the
Counts box. Now click the counter key back up to horizontal.
2. Setting X1 and X2.
We are going to first set X1 and X2 using the "double delta" algorithm. This
requires DeltaZ to be added to DeltaX and put on the Q bus. So scroll back to the
three big ganged switches and set the top and middle keys down to get DeltaZ
and DeltaX added onto the Q bus.
Now turn to the left hand big switch panel and scroll it up tp show the rows of red
keys. We are going to run the "1+2=." algorithm so on the top left hand row of red
keys put down 1 and 2 (reading from the left). This "adds" track 1 and 2 coming
down the Q bus and sends the result right towards the counter switches, the red
switches on the left. But we have to set the right hand yellow switch up to "dot" to
enforce only counting when the result = dot. Now set down the first of the red
counter switches to put the result into counter 1.
Now we need to try all possible pattern start postions for X1 and X2 to find the
highest score of the algorithm which should reveal the correct start positions fo
this cipher text. To do this, scroll the right hand display panel to show the master
switch panel. On the lower row of switches at the left hand side are the stepping
control switches. Set switch 1 down and switch 2 up. This tells Colossus to step
the X1 start position forward after every pass through the cipher text, and to step
X2 when X1 has done a complete cycle of its start postions.
3. Set Totals.
The next thig to set is the "Set Total" switches. By setting these only counts
greater than, or less than, the Set Total will be printed out, thus significantly reducing printout.
But first the Set Total must be calculated. This requires a bit of maths.
We found that the cipher length, n = 2042. We are using two elements of the 5 in
the Baudot characters, X1 and X2, so the expected random score is half n, i.e.
1021. The standard deviation(sigma) of this is 1/2(sqroot(n)) = 22.6
For a 1+2=. run the Set Total was usually set to 2.5 times sigma up on the random
score. This is 1021+56 = 1077. To set this in, scroll up the right hand panel to
reveal the verticle column of rotary switches for Thousands, Hundreds, Tens and
Units. By clicking on the arrows to the right of the switches, rotate them until 2
shows on the Thousands, 9 on the Hundreds etc.
(Health Warning!! the hit boxes on the rotate arrows don't always work properly
causing the right hand panel to revert to zero offset. If it does this, when you scroll
back to the Set Total switches check that the "revert to zero" has not caused any
of them to change from their set postion!! Make absolutely sure that the
dials show the correct Set Total before leaving!!)
Click underneath the "M" key, top row right, on the master swtich panel to start the run.
The printout window will appear with K1 K2 count as the heading.
Say "no" to the slow running abort Alert panel when it appears.
(The run takes about 20 minutes on an 800Mhz PC)
What comes out is:
K1 . K2 . count
31 .. 1 . 1087
.5 .. 3 . 1095
20 .. 6 . 1080
.8 .. 7 . 1079
20 .. 9 . 1082
34 . 17 . 1083
top counts at K1=31, K2=1 = 1087 = 3xsigma
K1=5, K2=3 = 1095 = 3.4xsigma.
so more runs would be needed to prove or disprove which one is correct.
The run to set K4 and K5 given K1 and K2 is 4=5=/1=2 this needs to be run
for both the two contending K1, K2 settings
4. The X4 and X5 run.
So now scroll the right hand panel to
show the "starts" plug panel. First look at the X1=31, X2=1 contender.
At present all starts are set to 1 so now move the
start of X1 to 31. The jack sockets are not exact, the important thig
is to have the numbers 31 and 1 in the starts box in the top panel display.
The next thing is to set wheels X4 and X5, given X1 and X2. The BP algorithm for
this was: 4=5=/1=2. The "/" means everything to the right is "given".
First clear all the red and yellow keys in the bottom half of the big switch panel.
Now scroll up to the top half, the black keys on the left hand side. On the first row
push down keys 1, 2, 4 and 5 leaving key 3 horizontal. This will give a "true" result
when bits 1, 2, 4 and 5 on the Q bus are = cross with bit 3 ignored and this result
can be switched into counter 1 by putting down the left hand grey counter key. But
what we actually want is a count when either the selected Q bus bits are all dot
OR all crosses. To get this final result, set the second row of black keys 1, 2, 4
and 5 up to match on dots on the Q bus. Now a cleaver bit of Boolean logic:
A OR B = NOT( NOTA + NOTB), so press down the yellow negate key just to the
left of the counter key on both the two rows and finally at the bottom of the set of
counter keys is a rowof yelliw keys which negate the whole counter output, set
down the left hand yellow key to negate the output of counter 1. The result of all
this is that counter 1 will count every instance of 4=5=/1=2 down the whole length
of the cipher text.
Now we need to set which wheels are to have their starts incremented. Scroll the
right hand panel to show the master switch panel. Reset lower row black keys 1
and 2 and then put down key 4 and up key 5. This means X4 will step after every
travers of the cipher text but X5 will only step for every completion of the X4 starts.
(26 x 23 = 598 runs through the cipher text).
Next what should the Set Total be? The expected random score for 4 elements
out of 5 is n/8 which gives 255. The sigma is 1/4 sqrt(3 x n) = 20 so Set Total
should be random + 2.5 sigma = 305. So
put this onto the Set Total rotary switches, then scroll the right hand panel to
show the right hand side of the master
switch panel and click down the "M" key to start the run. The print out window
should now show K4 K5 count on its top line and print out the scores.
Again say "no" to the abort Alert panel.
These results re-inforce K1=31, K2=1 as the correct settings, but
leave three contenders for K5, K4. The setting runs for K3 resolve
5. Now set X3.
So finally set X3 on a count of /// given 1 2 4 and 5.
Scroll to the top of the big switch panel. Put up key 3 on the top row, centre the
negate key, leave the counter 1 key down and centre all other keys. the top row
will now detect a "/" character (all dots) and if found count it into counter 1. Scroll
down and centre the yellow counter negate key. Now get the master switch panel
in the right hand window. On the bottom row centre keys 4 and 5 then put down
Now we need to Set Total. The expected random score is n/32 = 64.
sigma = 1/8 sqrt(3 x n) = 10 so put Set Total to random + 2.5 sigma = 89.
Now press down the "M" key and do the
run. The output window appears with K3 count on the first line and counts come
out for ///.
So this confirms K4=15, K5=1 and leaves two contenders for K3
6. A further K3 run.
3x/1x4x is another K3 run. This is set up on the red + keys, lower left
on the big switch panel. On the first row put down key 3, down the yellow key
to test for x and switch into counter 1.
On the next ro put down key 1 and down the yellow key to test bit 1 for x,
putting the result also into counter 1. Lastly put down key 4 on the third row,
put its yellow key up to test for dot and put the result into counter 1 as well.
Now make sure that the black K3 key is down to make K3 go through all its
starts and press down "M" to do the run.
This gives a maximum of 289 at K3 = 1, confirming this as the correct setting.
This confirms K1=31, K2=1, K3=1, K4=15, K5=1 as the settings for this
7. Now for the Motor wheels.
Next comes the setting of the Motor wheels M1 and M2.
This can usually be achieved by counting the number of "/" characters in the Delta
deChi at all places where Total Motor = x. (remember total Motor is the basic
motor, BM, caused by a "x" in M2 plus any limitation such as in this case, X2 one
To set the motor wheels first put a "/" character, all dots (0) onto the top row of
black switches, first row on the Q bus on the big switch panel. Set down the first
counter switch on this row to add the result of finding a "/" into counter 1.
Now scroll to the bottom of the big switch panel and set down to "x" the yellow TM
switch. This causes the Total Motor signal to only allow counting when Total Motor
is an "x".
Moving to the right hand panel, scroll across to the big ganged switch panel and
put down the Z and X switches to add Delta Z and Delta X. ie to get Delta deChi
onto the Q bus. On the row of Limit control switches uder the ganged switches,
put down the centre key to enable the X2 one-back limitation.
Scroll to the master switch panel and put down key M1 on the bottom row, just to
the right of the red X4 key. The next black key is M2 and put this up. This means
that M1 will step after every run through the cipjher text but M2 will only step after
M1 has gone through all its start positions.
Make sure that the M1 and M2 start positions are at 1 and scroll to the Set Total
8. Set Total for motor runs.
To caculate the Set Total, first scroll the switch panel up and set up the top
counter 1 switch, put down the counter 1 switch on the next row down. Since no
keys are set on the Q bus, every character of the cipher text is counted. With the
TM switch centred, press the LC key. Counter 1 shows 2042 the total length of the
cipher text. Now put down TM and do LC again. Now counter 1 shows 644, the
number of places down the cipher text where Total Motor = "x". The random score
is 644/32 approximately 20. The sigma is 1/8(sqrt(7 x 644) = 8. The motor
wheels usually set at about 2 sigma, so a Set Total of 36 should be set.
9. A cheat way using "10"!
The next problem is the length of run needed to examine all M1 and M2 start
postions, 61 x 37 = 2257. This takes about 30 minutes on an 800 Mhz PC and you
have to keep watching for IE slow running Alerts! A cheating method is to know
that the correct start positions for the KHcipher text are M1=59 and M2=26. This
is where the "10" switch on the master switch panel comes in. By putting this
down only 10 start positions will be explored. So set M1 to start at 54 and put M2
at 26. Now after returning the big switch panel top row to counter 1 and clearing
the second row, press down the "M" switch on the master switch panel. This
will bring up the print-out panel and show the counts + - 5 of the
If this has given you confidence that it will work you can now do the full run by
setting M1 and M2 starts back to 1, centering the "10" switch and pressing "M"
again. Watch out for Alerts!! This run is however inconclusive. There are lots of
high scores around 42 with no clear winner.
The Newmanry report (23L) suggests trying a minimum count rather than a maximum
count. The least frequent character in the DeChi is "B". To count this
set the "B" pattern (x..xx) into row of black keys and switch the result into counter 1.
We now need the Printout to only show counts LESS than the Set Total.
The Set Total should be 1.5 sigma less than the random score, ie. 8 and the
top black key on the right hand side of the Set Total panel needs to be set to the
left for less than.
This Motor setting run again takes about 30 minutes and shows 6 low counts of 6.
M1/M2: 50/3 47/9 55/9 2/25 59/26 and 34/32
These can be resolved by attempting to set the Psis on each of them in turn.
10. Setting the Psi wheels.
Setting the S wheels requires the direct Z, X and S signals switched onto the Q
bus, not the Deltas as used for X and M setting. Any limitations must be set, in this
case the X2 one-back limitation.
The most common character in the plain language text is usually "space", 00100
in Baudot code, thus checking for maximum count of 1+2=. 3=x and 4+5=. should
set the Psis.
S1 and S2 can be set first by doing 1+2=. on the red keys, just like for setting X1
On the master switch panel set down the lower black S1 key and set up the S2
key so that S1 start increments after each traverse of the cipher text and S2
moves on only after S1 has gone through all its starts.
Again Set Total needs to be set and since two elements are being investigated
this can be 1077 as for X1 and X2. So scroll the right hand Colossus panel to the
Set Total rotary switches and set them to 1077. (remember to clear the
"less than" key if you have used it)
Set the M1 and M2 starts to the first set to be tried.
Now press down the "M" key top right hand of the master switch panel.
The complete run will take about 30 minutes on a 800 Mhz PC.
The printout will only show a clear maximum count of over 1200 when the
correct M1 M2 starts are used. The maximum is at S1 = 32, S2 = 36.
Now set S4 and S5 by using 4+5=. Scroll the big switch
panel down to bottom left. Centre the top row 1 and 2 red Qbus keys and put
down 3 and 4. Scroll the right hand panel to show the master switch panel. Centre
the lower row S1 and S2 keys and put S4 down and S5 up to increment the S4
start position every traverse of the cipher text and S5 only when S4 has done all
Now press the "M" key to do the run. Now the printout
shows a clear maximum count at S4 = 9, S5 = 43.
Finally set S3 by counting 3=+ Centre red keys 4 and 5
on the top row of red Q bus keys and put down red key 3. Move right along this top
row to the yellow key which should have been set to "dot" for the previousruns.
Centre it then put it down to count "x". Change the S start increment keys on the
master switch by centering S4 and S5 and putting down S3. This will just take the
S3 ring through all its start positions. The same Set Total can be used and a
maximum count shows at S3 = 11.
Finally, set the S start positions to S1 = 32, S2 = 36,
S3 = 11, S4 = 9 and S5 = 43. M1 should be at 59 and M2 at 26.
11. A test deciphering.
With all the other ring starts set correctly, try deciphering the cipher text.
Check that you have all the big black keys set up to Direct not down to Delta. On
the master switch panel press down the "SU" key to set everything up. Now press
down the "T" for Trace key. The top test panel boxes will be filled up with Z, X, S
and Q bit patterns. Also shown are the Baudot characters corresponding to the bit
patterns. The first character read from the tape, (in the Z box) is "J". Also shown is
the deciphered character in the Q box, which is "X". This output was not available
on real Colossus. On real Colossus the operator set spanning to one character
and switched the Q bus bits to the five counters. The bit pattern of the deciphered
character could then be read from the counter outputs and the character looked
up in the teleprinter code chart.
So now by repeated pressing of the "T" key, successive characters of the decipher
should appear in the Qbus box as:
Read the introduction to Virtual Tunny, then after selecting the cipher,
KHcipher and the patterns, KHpatterns, set in the wheel start positions
you have just found and decipher the cipher text.