Fixes a potential overflow in high-passed signal for transient detection
and ensures that the shift can never go negative

We can't use the attenuation since it doesn't apply to the memory

extrapolation_len can be up to 960+120, so we we need a shift of 10
if the values are very large.

Gives us a tighter bound on the pitch postfilter input to avoid overflows

This could cause overflows when processing non-saturated TDAC values.

Otherwise, the output can grow as large as SIG_SAT/(1-0.75), which can cause
an overflow when adding two values (before multiplying by the filter tap).

Reduces the XX values to avoid an overflow in silk_VQ_WMat_EC(), even in
floating-point.

Not observed yet (except when caused by a different bug), but "just in case"

Felicia Lim authored 8 years ago and Timothy B. Terriberry committed 8 years ago

Signed-off-by: Timothy B. Terriberry <tterribe@xiph.org>

When the energy is stable, we slightly bias energy quantization towards
the previous error to make the gain more stable (a constant offset is
better than fluctuations).

We reduce the bitrate by about 0.2% to 1% at low bitrate for the same quality.

Increasing the value at low rate seems to help a bit.