ECE4253 Digital Communications
	Department of Electrical and Computer Engineering - University of New Brunswick, Fredericton, NB, Canada

MNP5 Data Compression

There is a fixed selection of codewords ranging from 4 to 10 bits in length. For most codewords, the first 3 bits (highlighted) indicate how many bits remain in the codeword. Exceptions are the first two 4-bit words (beginning in 000), and the last two words where an additional bit is added to allow a 'reset' codeword (number 256). While this approach limits the number of codewords available with a given length, it ensures that codewords can be extracted from a binary stream by first examining three bits to determine how many more bits to take.

For example, consider this MNP5 compressed binary message:

0100010110111000110001100011...

01000 10110111 0001 1000110 0011...

MNP5 Codeword Distribution

It is significant that most MNP5 codewords are more than 8 bits long. This implies that for random data, application of the MNP5 compression algorithm will lead to a file which is longer than the original file. To achieve effective compression, the most frequently occuring characters must be assigned to codewords having less than 8 bits.

The initial (default) assignments for codewords are shown in the table below. Note that in this default assignment, printable ASCII characters align with 9-bit codewords. Again, this suggests that without some adjustment, this table would be useless for compressing text messages. In practice, the table assignments change after the first character is transmitted, such that on the second occurence of a given character, a shorter codeword would be used. (See Examples)

A full description of the MNP5 algorithm may be found in The Complete Modem Reference by Gilbert Held

MNP5 Codeword Table (Initial State)