[fitsbits] 16-bit floats {External}

[Richard J. Mathar]

There was a Michelson-interferometer at the KIT that (as usual)
has a much brighter intensity (white light peak) for
path differences near zero than elsewhere. That was actually
communicated within the instrument by reserving one
of the 16 bits to flag a "scale by factor x". Essentially
that type of coding convention enhanced the range while
storing data with a precision of 15 bits. 

This type of reserving some bits of the 16 is of course not as universal 
as using a IEEE convention.