Illustrating concepts of FFT bin width and spectral leakage

Recent discussion of bin width got me thinking about that and the 
how/why of windowing data before performing an FFT.

As one who always did better in lab courses than theory, I decided to do 
some experiments. The attached program is my attempt to illustrate 
effects of window width and sampling frequency. I plan to add 
illustrations of effect of window type and numerical precision 
(emulating 8,10 or 16 bit precision on an 32 bit machine with automatic 
multiprecision floating point should be interesting ;)

I'd appreciate comments from:
  1. fellow newbies/novices/STUPIDENTs
     Did you find this helpful?
  2. the DSP experts
     1. First importance - any blatant errors?
     2. Have I done anything that, although correct, would mislead?
     3. Any suggestions?
        PS - I explicitly avoid "normalization" as much as possible.
             It has been a problem with DSP (and other) textbooks.


The following program is written for Scilab 4.1.2 (should be compatible 
with Scilab 5.x) and has been run under WinXP.

drawlater
for i=240:250; f(i-239)=i; end;

N=100;
t=(0:N-1)*.001;       // 1 kHz sample rate
w = window('hn',100); // .1 second window

subplot(3,2,1)
for i=1:11
plot2d( 1:N, abs(fft(sin(2*%pi*f(i)*t) .* w)), i);
end
title("100 SAMPLES -- NO PADDING")

N=1000;
t=(0:N-1)*.001;       // 1 kHz sample rate
w = window('hn',100); // .1 second window
w(101:1000)=0;        // zero pad

subplot(3,2,3)
for i=1:11
plot2d( 1:N, abs(fft(sin(2*%pi*f(i)*t) .* w)), i);
end
title("100 SAMPLES ZERO PADDED TO 1000")

w = window('hn',200); // .1 second window
w(201:1000)=0;        // zero pad

subplot(3,2,5)
for i=1:11
plot2d( 1:N, abs(fft(sin(2*%pi*f(i)*t) .* w)), i);
end
title("200 SAMPLES ZERO PADDED TO 1000")


// =========================================================

for i=240:250; f(i-239)=10*i; end;

N=1000;
t=(0:N-1)*.0001;       // 10 kHz sample rate
w = window('hn',1000); // .1 second window

subplot(3,2,2)
for i=1:11
plot2d( 1:N, abs(fft(sin(2*%pi*f(i)*t) .* w)), i);
end
title("1000 SAMPLES -- NO PADDING")

N=10000;
t=(0:N-1)*.0001;       // 10 kHz sample rate
w = window('hn',1000); // .1 second window
w(1001:10000)=0;        // zero pad

subplot(3,2,4)
for i=1:11
plot2d( 1:N, abs(fft(sin(2*%pi*f(i)*t) .* w)), i);
end
title("1000 SAMPLES ZERO PADDED TO 10000")

w = window('hn',2000); // .1 second window
w(2001:10000)=0;        // zero pad

subplot(3,2,6)
for i=1:11
plot2d( 1:N, abs(fft(sin(2*%pi*f(i)*t) .* w)), i);
end
title("2000 SAMPLES ZERO PADDED TO 10000")

drawnow