The tape format of Vic Rabbit
reverse engineered and documented by Marko Mäkelä, September 22, 2002

The Vic Rabbit format appears to use three different pulse widths.
The pulse widths vary surprisingly much, and there doesn't seem to be
too much error tolerance.  The difference of the widest short pulse
and the narrowest medium pulse I observed is only 24 microseconds.
The difference between the widest medium pulse and the narrowest long
pulse is 32 microseconds.  I wonder if this format tolerates any tape
speed variations.

The initial sync is CDBBBDBDDBBBBBBBBBBBBBBBBB repeated 1200
times. (B=short pulse, C=medium, D=long).  CD appears to mark the
start of a byte.  The tape header is 26 times CDxxxxxxxx where each x
is either B=0 or D=1.  This header is followed by a C (medium) pulse
and 80 times CDBBBDBDDBBBBBBBBBBBBBBBBB again, followed by one copy of
the program, again followed by a C pulse.  Unlike in the normal
format, the data bytes are recorded most significant bit first.

Each block seems to start with a $0f byte.  The last two bytes of a
block are a 16-bit sum of the data bytes, excluding the $0f byte and
the checksum bytes themselves.

In the tape header, the file name (16 bytes) comes first.  Then are
two or three unknown bytes followed by the start and end address of
the program block.  The program block contains just the raw program
data between the $0f byte and the checksum.