function y=zoomfft(x,n,m,s,d)
%ZOOMFFT    DFT evaluated over a linear frequency range Y=(X,N,M,S,D)
% Inputs:
%    x    vector (or matrix)
%    n    frequency increment is fs/n [default n=length(x)]
%    m    mumber of output points is floor(m) [default m=n]
%    s    starting frequency is s*fs/n [default s=0]
%    d    dimension along which to do fft [default d=first non-singleton]
%
% This routine allows the evaluation of the DFT over an arbitrary range of
% frequencies; as its name implies this lets you zoom into a narrow portion
% of the spectrum.
% The DFT of X will be evaluated along dimension D at the M frequencies
% f=fs*(s+(0:m-1))/n where fs is the sample frequency. Note that N and S
% need not be integers although M will be rounded down to an integer.
% Thus zoomfft(x,n,n,0,d) is equivalent to fft(x,n,d) for n>=length(x).

% [1] L.R.Rabiner,  R.W.Schafer and C.M.Rader, "The chirp z-transform algorithm"
%     IEEE Trans. Audio Electroacoustics 17 (2), 86