Last For Now

Last for now.  No padding vs. One Second Padding.

Paddington Goes To Town

No pad, one second pad, two second pad.   Note the already high granularity (goes from 1/190 Hz, to 1/191, to 1/192, ie. roughly 190 samples per HZ).  Note the sinc characteristics.

Zero Padding Examples

First, a 190 second window, sampled at 100 Hz, 1.9Hz and 5.2Hz sine.  Note the near-zero values in the spectrum

Now a ONE second zero padding addition to the same.  Note the sinc convultion in spectrum.  This is clearly not due to change in freq-domain granularity increasing from 1/190 sec to 1/191 sec.

 Now lets remove that one second pad by filling in the second w/ real data.  Poof, the sinc convolution is gone, yet we are at the same (1/191sec) granularity as the zero-padded.  I think that this spectrum differs qualitative from the first b/c we are no longer at exact multiples of the fundamental freq

Now the 191 sec window w/ a one second zero pad:  

Here's the same as above, w/ noise added.  Note the persistence of the sinc convolution.

Finally, let's add a bunch more padding and zoom in around one of the fundamentals.