How to represent a floating point number into a fixed point number

[snoop911]

Im interfacing to a chip that uses a 16-bit register where bits 13:6 represent the integer portion, and 5:0 the fractional portion.

So, my problem is converting from my chip's native floating point representation into a fixed point.

64bit Floating point is represented by my TI chip as

s A 1-bit sign field (bit 31)
e An 11-bit biased exponent (bit 30:20)
f A 52-bit mantissa (bit 19:0 and 31:0)

for example, -1.0e25 is listed in memory as...
000000 C520
000001 8B2A
000002 2C28
000003 0291


Now, stripping off the integer portion is easy, doing something like...
ulSlopeInt = (uint16) (floatingPointVal);

works ok, but how can I represent the fractional portion to a fixed point???

[snoop911]

Actually, perhaps, all that I need to understand is how in general would you represent a fractional number into a fixed point data type?

Gut instinct would suggest that .5 would probably be represented somewhere along the lines of 0x8000, no? This is perhaps the real question!