James Zern authored 12 years ago and Gregory Maxwell committed 12 years ago

Quiets:
warning: this decimal constant is unsigned only in ISO C90

when building with e.g., gcc -m32 -std=gnu90

opus_decode() and opus_multistream_decode() now apply soft clipping
before converting to 16-bit int. This should produce better a higher
quality result than hard clipping like we were doing before. The _float()
API isn't affected, but the clipping function is exported so users can
manually apply the soft clipping.

1) In cases where the SILK desired bandwidth went down, then quickly up, we
count get stuck in a mode with the LP variation going the wrong way.
2) Bandwidth detection can no longer force SILK to go below wideband to
avoid switches that require redundancy.

Detects a transient when the average band energy jumps by more than 6 dB
between two consecutive frames. This fixes some undetected transients on
Gainless' Muse_Breaks_Rmx sample.

By not reinitializing the stereo state during a switch, the old stereo
memory would create a false impulse (i.e. single sample) in the decoded audio.
This change affects the normative part of the decoder. Fortunately,
the modified decoder is still compliant with the specification because
it still easily passes the testvectors. For example, for the float decoder
at 48 kHz, the opus_compare (arbitrary) "quality score" changes from
from 99.9333% to 99.925%.

This improves quality of CELT->SILK switches by not having the SILK encoder
attempt to encode a discontinuity. Overall, it's a small PQ improvement,
confirmed by some listening.

This was causing periodic glitches in the right channel in hybrid mode.

The channel mapping family nested lists were hard to read.
Promoting each family description to a subsection makes it
more clear where one is in the document.

Also add suggested downmix matricies. These are what we're
currently using in opus-tools, opusfile, and Firefox.
Ascii-art matrices in 69 columns is hard.

With xml2rfc 2.3.9, I was getting the 'audio/ogg; codecs=opus'
figure hoisted out of the the containing paragraph, which was
very confusing.

I've tried to improve this by moving the surrounding paragraph
into the figure using the <preamble> and <postamble> tags, as
in the example in draft-mrose-writting-rfcs.

Unfortunately this still isn't perfect. Since the surrounding
paragraph is part of the figure we can't set align="center" on
the figure itself, and the processor seems to ignore it on the
artwork element. I've compensated by adding some leading
whitespace.

Mostly just being more verbose.

This clarifies the meaning in surround configurations and better
distinguishes it from 'rear center'.

Add sections on VLC, foobar200, and Rockbox.
Be more complete about listing which features of the format are
 actually supported by each implementation.
Still need to do more testing for Chrome, FFmpeg, and libav.

We mean to make .opus only for degenerate--not concurrently
multiplexed or grouped--but optionally sequentially multiplexed
or chained physical Ogg streams.

However, we don't mean to say you can't use this draft with
such streams.

Also, replace the reference to RFC3534 with RFC5334, which
obsoletes the previous reference by defining video/ogg and
audio/ogg.

This is based on the experiment described in
http://tools.ietf.org/html/draft-sheffer-running-code-01

I've listed several major implementations. Their formatting
could probably be better. More obviously, this really demonstrates
the need for a good test suite to evaluate converage and compliance
with all these different implementations.

This is true, but not important to understanding the limitation.

Would cause the preemphasis and deemphasis to fail, but only with
custom modes enabled.

Was causing the testvector to pass with lower quality, though
practically there was no quality degradation.

Should slightly improve coverage and reduce code size

This avoids a warning about obsolete AM_CONFIG_HEADERS() on more
recent autoconf. The new macro has been around at least since
autoconf 2.60. So this should be safe.

quant_band() now only handles the level0 case.

Fixes an MSVC double truncation warning.

This fixes a truncation warning in MSVC:

warning C4305: 'initializing' : truncation from 'double' to
'const opus_val16'.

opus_val16 is a float in the floating point build and an
int16 in the fixed-point build. The table isn't used in
the fixed-point build, but the compiler would presumedly
warn in that case as well.