It is a common misconception that floating point gives greater accuracy than fixed point. This is not neccesarily so. While FP will certainly give greater range, the resolution is limited.
The 32-bit IEEE-754 FP representation (the 'standard' for most 8051 C compilers) has a 23-bit fraction, an 8-bit exponent an a sign bit. This gives a range of ±1.2e-38 to ±3.4e+38. However, the resolution of any given number is only 1 in 2^23. Compare that to a 32-bit integer, which has a range of -2e9 to 2e+9 (or 0 to 4e+9), but a resolution of 1 in 2^32.


Rob.