http://mathworld.wolfram.com/WalshFunction.html
From “transpositions of wires” don’t we naturally link to cipher boxes transposing bit positions?
http://frode.home.cern.ch/frode/ulfving/node9.html
It seems more useful to look at the core coding theory of the T.52, where one sees the gray code flip (between O and U, classes 0,12 and 3,4,5 – in each Horizontal impulse stream. (I cannot find bit pattern for char ‘+’, note, in the tunny table)
from http://www.rutherfordjournal.org/article010106.html (upper) and General Report on Tunny (lower)
let me correct that: the + char code is at 5, ringed green.
The Rutherford document goes on to focus on the logic of “wheel” combining, given use of the SR relays specifically. lets have have a look how (bit) weights, weighting functions (generally), graphs and their matrix form relate to each other:
via and from http://www.rutherfordjournal.org/article010106.html
so, we see from the first table above that wheels DCBA are controlled (collectively) by the two wheels above the group (EF). Then, each other wheel (e.g. E) is controlled by the two wheels above it. The control functions are split between dots and crosses, with no obvious pattern as to which dots and which crosses from a given wheel (pair) influence the successor wheel.
http://www.codesandciphers.org.uk/documents/wffishnotes/wf021.HTM describes this more fully:
The pentagram “encoding” reminds me a little of the DES cycles. Perhaps now we can _semantically_ distinguish the arcs that are on the surface of the sphere” from the arcs that cross the sphere. Can we learn from the semantics of the pentgram model, and its relation to “Baudot classes” (and the weights that define those classes)? Do we recall the 32*32*32 quantum model, in which we had 2 probability sets represented along 2 edges, and 1 set cross the interior of the cube (which we then sliced)?
lets look at the semantics, as described by the Rutherford reporter:
The graph in Fig. 10 is constructed from the wheel combining logic in Fig. 3. In the graph in Fig. 10 the SR relays are represented by the vertices and the controlling wheel output channels by the edges which join in a given vertex. The graph quickly reveals the relationship between the different SR relays; it shows the topology of the circuit clearly.
[…]In the BP Pentagon the vertices are numbered from 1 to 5 in a clockwise manner, while the sides and diagonals are labeled by the lower case letters a-j.
Well let’s do that, for some arbitrary assignment of the lower case letters (then figure why the modern reporter re-labeled things).
so what does this mean? SR1 (1) has 4 output channels (e,a,I,f)? does it mean wheel a is stepped when relays 1 and 2 are both in state a (assuming each relay has 2**4 states, from 4 bits)? Anyways, the idea is clear. The graph encodes the desired logic.
ok. lets break down and read the solution. First we notice something. the distribution of crosses in output channels I, II… V shows some symmetry in the row space, and reminds me of our tunny chart in the column space. There are 3 columns of all crosses, and 2 columns of two crosses, that are column-wise complementary.
Now, earlier text is pretty clear, and we see how SR4 node models the table:
Its not a 2**4 function however. It’s a odd-parity sets relay mode, for 4 inputs. (This helps explain ECM’s method for creating/constraining its probability stream generator.)
Let’s look now at case of SR5 xor SR9. SR5 sets on the odd-parity of 1,9,II, V as the table says. Similarly, SR9 sets on odd-parity of 1,9, I, and III. Clearly, 1 and 9 are common (meaning any odd parity from one set of 1,9 would be cancelled out by the other), leaving I, III, II, V – which matches the earlier table – that xors SR9 with SR5.
I can see how, similarly, the pairs of relays exploit the fact that the 13579 output channels are arranged so they cancel out, leaving only some 4 of the roman wheels to contribute.
The pentagon for the wheels 13579 is an inverse of the permutor’s coding wheel in some sense, much like the 2 outer hamiltonians (b0,b1 and b4,b5, in the 6 bits of input to an sbox) controlling the data flow in the DES engine are similarly complementary.
Hmm. They are labeled CD, EF – not AB, EF. But regardless (of that minor misconception on my part) can we take the union of those two graphs and by analogy say that “relay 1” is set as a function of output channels 5, 2, 8 and 6? – where an sbox is an “output channel??
And, then we add in 7 and 4, then, giving what is required – 6 distinct sbox contribute to the 6 bit input of an sbox? Well, given these are just expressions of the networks, its not surprising this all works out! But, its nice to trace it all from T.52 symmetry, for CDEF.
(DES diagrams from Davis and Price, by the way.)
You must be logged in to post a comment.