[Theora-dev] FPGA implementation

[Andrey Filippov]

> Also, strangely, libogg says there are 1151 (=894+257) bytes left in the
> buffer at this point, despite the fact that the page header+data only
> take up 894 bytes.

Yes, it was my bug - while manually manipualting data I missed 0x100 so
high frequency components were lost (nice feature of the format). Since
that I've noticed my FPGA bug - DC predictor supplied two's complement
data and the next module expected sign+magnitude.