The Fibonacci code can be viewed as a self-clocking
serial code for data storage or transmission. For storage
use on a medium such as a single-track magnetic tape, or for use as
a serial data communications code, Fibonacci code (approaching
0.6667 bits per cell) is 33% denser than Manchester Encoding or
Phase Encoding (0.5 bits per cell).
When considering serial codes,
a "cell" should be interpreted as the shortest time
between transitions. Note that Manchester Encoding is used at
the lowest (physical) level for Ethernet communications. A change
to Fibonacci coding, implemented in hardware, could increase the
speed of any Ethernet link by 33%, with no increase in bandwidth
on the link.
"Self-clocking" refers to a code which
allows clock and data signals to be easily separated from a single
signal. In the Fibonacci code, the time between signal
transitions is either 1 or 2 cells. Manchester Encoding and Phase
Encoding also have this property. Such codes are most useful for
storage on media with a large amount of jitter, or poor speed
control, such as magnetic tape. On media with accurate speed control,
denser codes, such as group codes, can be used. Group codes achieve
greater density by allowing a larger number of cells to pass without
a transition, and are commonly used on such media as hard disk
drives.
Codes like Manchester Encoding and Phase Encoding convey binary information directly, bit by bit. The Fibonacci code is sent in blocks of a pre-determined length, which can then be decoded to yield a binary number. For instance, since a 32-cell long code maps into an integer between 0 and 2,178,308, it can certainly transmit 21 binary bits (0 - 2,097,151), and hence can replace 42-cells that are Phase Encoded (not counting any prefix or suffix required by either code). This actually leaves 81,157 unused Fibonacci codes, that can be ignored, or used for control or other such purposes.
Consider a serial code represented by two voltage
levels, V+ and V-. Let's call the length of the shortest time
between voltage transitions a "cell". Further, consider
a code in which a single cell between transitions represents
a "1" element, and two cells between transitions
represents a "0" element. Thus, two "1" elements
take exactly the same time, two cells, as one "0" element.
Note that the time between transitions alone determines
whether an element is a "0" or a "1", not
voltage level during the cell time. Elements must alternate
between V+ and V- to define the code. (Interested readers might
like to note an alternative definition).
It turns out that the number of different codes that
are exactly N cells long is equal to FN+1, where FN
represents the Nth Fibonacci number. See a proof of this.
You may recall that the Fibonacci series is
a series of numbers beginning with F1 = 1
and F2 = 1, and in which each following number
is obtained by adding the previous two:
Fibonacci Number: | F1 | F2 | F3 | F4 | F5 | F6 | F7 | F8 | ... |
Value: | 1 | 1 | 2 | 3 | 5 | 8 | 13 | 21 | ... |
Note that the traditional definition just above can
easily be extended by defining F0 to be zero; the recurrence
relationship (each number is the sum of the previous two) is maintained.
We will use the notation F0 (equaling zero) in the
rest of this document:
Fibonacci Number: | F0 | F1 | F2 | F3 | F4 | F5 | F6 | F7 | ... |
Value: | 0 | 1 | 1 | 2 | 3 | 5 | 8 | 13 | ... |
The next step:
individual codewords in a Fibonacci code of length exactly N
cells can be assigned numerical values, assigning each
codeword a unique value between zero and FN+1-1.
See a description of how to do this .The
result is a weighted code, just like decimal or hexadecimal or
octal number representation, except with "column" weights
which are the Fibonacci numbers instead of powers of the radix
(base), and with the additional property that a single "0"
code element uses up two columns. Conversion between the Fibonacci
code and binary can then be easily done, implemented either in
hardware or software.
Assuming the idle value of the code is, say, V-,
it obviously must begin with a transition to V+. The remaining
elements then alternate, V-, V+, V-, V+, etc. This produces a
problem, however at the end of the code. Although the code has
been defined such that each codeword contains the same number
of cells, some codes have an even number of elements,
and some an odd number. As a result, some codewords end up at
V-, and others at V+, after the final transition. The latter cannot
be allowed to happen, since the line needs to return to its idle
value. This problem can be solved by adding another 1-cell element
at the end of the code as a suffix, if the code ends up at V+.
Alternatively, if the overall length of the code,
including the suffix, must be fixed, then add a 2-cell element
if the code ends up at V+. If the code ends up at V-, add 2 single-cell
elements. Thus, the suffix will always be the same length, two
cells. Serial codes often also need some sort of prefix, to define
the first transition, or to properly synchronize the receiver.
The figure below shows the number "12345678"
encoded in Fibonacci code, using a 39-cell code, which is sufficiently
long to encode numbers between zero and 102,334,156 (F40-1).
Thus, a code this length could be used to encode any arbitrary
eight decimal digits. The corresponding number range using
Phase Encoding requires 27 bits, or 54 cells, and is also shown.
No prefix or suffix is shown on either code.
In phase-encoding, one data bit is transmitted by every pair of cells. In the example shown, the data bit is always the second cell of each pair, in which an up-going transition represents a zero, and a down-going transition represents a one. A transition is placed in the first cell of each pair if and only if it is needed to set up the second cell to transition in the proper direction. Thus, there is always a transition in the second, data cell, and only sometimes in the first cell.
Click here to return to Larry Krakauer's home page