Make triangle wave in an OscillatorNode match WebAudio spec.

Make triangle wave in an OscillatorNode match WebAudio spec.

The triangle wave is specified to be an odd function that is 0 at time
0 and has a positive slope at time 0. (The current version was an even
function.)

Also took this opportunity to add comments to the code to make clearer
what the transforms are and also made the formulas match. (Some were
off by a fixed factor, which didn't matter since all the waveforms
were normalized later in createBandLimitedTables()). This makes it
easier to verify against known results on the web.

Layout tests are also updated for the change to the triangle wave, of course. The sawtooth and square wave tests were also updated for the slight formula change. (Differences are 1 lsb for these tests.)

Also rebased the Android tests for the custom and sine tests. Apparently something has changed on Android. The changes are small with differences of 1-2 bits at most.

BUG=316666
R=kbr@chromium.org

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=162139

[Raymond Toy (Google), 2013-11-13 21:14:46]

PTAL.

[Ken Russell (switch to Gerrit), 2013-11-15 20:30:04]

I'm not qualified to review the math -- at least not without taking a long time
to come up to speed on it -- but the code LGTM, especially given the thorough
test description.

Nit: move BUG= to the bottom of the CL description before checking the CQ flag.

[Raymond Toy (Google), 2013-11-15 21:14:32]

On 2013/11/15 20:30:04, Ken Russell wrote:
> I'm not qualified to review the math -- at least not without taking a long
time
> to come up to speed on it -- but the code LGTM, especially given the thorough
> test description.

Would it help if I added a brief derivation?  Or maybe I can just add the
following links:

http://mathworld.wolfram.com/FourierSeriesTriangleWave.html
http://mathworld.wolfram.com/FourierSeriesSquareWave.html

I can't find any web reference that matches our sawtooth wave exactly.

> 
> Nit: move BUG= to the bottom of the CL description before checking the CQ
flag.

Will do.

[Ken Russell (switch to Gerrit), 2013-11-15 21:41:52]

On 2013/11/15 21:14:32, rtoy wrote:
> On 2013/11/15 20:30:04, Ken Russell wrote:
> > I'm not qualified to review the math -- at least not without taking a long
> time
> > to come up to speed on it -- but the code LGTM, especially given the
thorough
> > test description.
> 
> Would it help if I added a brief derivation?  Or maybe I can just add the
> following links:
> 
> http://mathworld.wolfram.com/FourierSeriesTriangleWave.html
> http://mathworld.wolfram.com/FourierSeriesSquareWave.html

Sure, those links would be helpful.

[Raymond Toy (Google), 2013-11-15 21:54:36]

On 2013/11/15 21:41:52, Ken Russell wrote:
> On 2013/11/15 21:14:32, rtoy wrote:
> > On 2013/11/15 20:30:04, Ken Russell wrote:
> > > I'm not qualified to review the math -- at least not without taking a long
> > time
> > > to come up to speed on it -- but the code LGTM, especially given the
> thorough
> > > test description.
> > 
> > Would it help if I added a brief derivation?  Or maybe I can just add the
> > following links:
> > 
> > http://mathworld.wolfram.com/FourierSeriesTriangleWave.html
> > http://mathworld.wolfram.com/FourierSeriesSquareWave.html
> 
> Sure, those links would be helpful.

Done.

Thanks for the review. I'll have to commit this by hand since the CQ still
doesn't handle binary files.

[Ken Russell (switch to Gerrit), 2013-11-15 21:55:09]

Sounds fine.

[Raymond Toy (Google), 2013-11-15 21:57:06]

Committed patchset #2 manually as r162139 (presubmit successful).