You have to understand that float are negative powers of 2 and that conversion from decimal is not as straightforward as for integer: conversion from decimal often leads to loss of precision contrary to what happens for integers.