function y = round_float(x, n)
% ROUND_FLOAT  Round off the fractional part of a floating-point number
%
%             y=round_float(x,n);
%   
%  rounds the input vector of floating-point values `x' by approximating 
%  the "fractional part of x" using an `n' bit representation.
%
%  That is, if 
%      x = f*2^m       0.5<f<1
%  then y is obtained using an n-bit representation for f.
%
%  (since the input vectors are generally of type 'double', this function
%   really only makes sense for 1<n<52 )

% 

[f,e] = log2(x);
f = round(f*2^n)/2^n;
y = pow2(f,e);
return;