PDP-11

Floating point numbers for PDP computers are stored in either 2 words (totaling 32 bits) or 4 words (totaling 64 bits). 

The normalized bit is assumed to be 1 unless the exponent is 0 (corresponding to 2 to the 128) in which case that bit is assumed to be 0.  The value 0 is represented by two (or four) words of all zeros.

The following is the representation of a 2-word floating point number:

 
PDP-11 Floating Point Representations: (Word 1 of 2)
15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
Sign Bit Exponent High-order fraction

 
PDP-11 Floating Point Representations: (Word 2 of 2)
15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
Low-order fraction

NOTE: The exponent is biased by 128.

For instance, 2.0 is 40200 in octal or 0100000010000000 in binary (remember the 128 bias and the hidden bit), plus a following 16-bit word of all zeros.

Four word floating point numbers add two more words for the fraction of decreasing significance and increasing precision.