Michael J. Wilson
Music 256a Fall 2010
Homework #3

----------------------------------------------------------------------

IDEAS / COMMENTS:

I was able to auto-detect sampling rate this time.  I hard-capped it
at 48000 since if Alsa is used it can pick a very high sampling rate
if we just look for the highest.

The audio transformations are handled by a separate class.  It
contains a mutex to control access to data shared between the audio
and display threads.  It can fill up a buffer for transformation
independently of the audio buffer size, which lets us pick whatever
frame size we want when starting Jack or similar systems.

I implemented a class to handle the waterfall plot data.  It's
basically a circular list of vectors holding waterfall data with all
the indexing hidden from the outside.

I also implemented a few options to change the spectral view in
realtime, such as window changes, scaling changes, zooming in on the
lower part of the frequency band, and switching between a linear and
dB scale.

Finally, I implemented some rudimentary pitch estimation.  I used a
thresholded unbiased autocorrelation to find periodic structures in
the sound.  It doesn't work very well, unfortunately.

I really enjoyed looking at different sound files encoded in different
formats and seeing how their spectra were changed.  I found the
concept of this assignment very satisfying.

----------------------------------------------------------------------

DIFFICULTIES:

Due to external factors I didn't have as much time to work on this
assignment as I would have liked to.

I initially had a hard time getting the magnitude spectrum to display
correctly.  Once I worked out the proper scaling and translation
parameters it started to look good.

I am disappointed in how poor the pitch detection turned out, but
given how I am doing it I don't think that was necessarily unexpected.
I am planning to use pitch detection as a key part of my final project
so I will be studying this in more detail in the coming weeks.

The code is also not very optimized; there are occasional xruns noted
by qjackctl.

----------------------------------------------------------------------

COLLABORATORS:

I didn't have any direct collaborators on this project.  I used the
chuck_fft code and referred to the source code of sndpeek as I was
writing this program.