*** TAP (raw C64 cassette TAPE images)
*** Document revision: 1.1
*** Last updated: March 11, 2004
*** Contributors/sources: Per Hakan Sundell,
                          Markus Brenner

  Designed by Per Hakan Sundell (author of the CCS64 C64 emulator) in 1997,
this format attempts to duplicate the data stored on a C64  cassette  tape,
bit for bit. Since it is simply a representation of  the  raw  serial  data
from a tape, it should handle *any* custom tape loaders that exist.

  The TAP images are generally very large, being a minimum of eight  times,
and up to sixteen times as large as what a raw PRG file would be.  This  is
due to the way the data is stored, with each bit of the original  file  now
being one byte large in the TAP file. The layout is fairly simple,  with  a
small 14-byte header followed by file data.

      00 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F        ASCII
      -----------------------------------------------   ----------------
0000: 43 36 34 2D 54 41 50 45 2D 52 41 57 00 00 00 00   C64-TAPE-RAW????
0010: 51 21 08 00 2F 0F 0D 31 64 1D 26 0D 07 21 0A 12   Q!??/??1d?&??!??
0020: 4A 2F 2C 34 07 18 0D 31 07 04 23 04 0D 42 0D 1E   J/,4???1??#??B??
0030: 34 04 42 0D 20 15 5E 04 0D 18 61 0D 26 29 34 0D   4?B???^???a?&)4?
0040: 23 0D 07 0A 3F 55 04 0A 13 3F 07 0D 12 2B 18 0A   #????U???????+??

    Bytes: $0000-000B: File signature "C64-TAPE-RAW"
                 000C: TAP version (see below for description)
                        $00 - Original layout
                         01 - Updated
            000D-000F: Future expansion
            0010-0013: File  data  size  (not  including  this  header,  in
                       LOW/HIGH format) i.e. This image is $00082151  bytes
            0014-xxxx: File data

  In TAP version $00 files, each data byte in the file data area represents
the length of a pulse, when the C64's hardware  will  trigger  again.  This
pulse length is determined by the following formula:

    pulse length (in seconds) = (8 * data byte) / (clock cycles)

  Therefore, a data value of $2F (47 in decimal) would be:

    (47 * 8) / 985248 = .00038975 seconds.

  A data value of $00 represents an "overflow" condition, any pulse  length
which is more that 255 * 8 in length.

  The value of "clock cylces" from above  (985248)  is  based  on  the  PAL
value.  Since  this  file  format  was  developed  in  Europe,   which   is
predominantly PAL video, this is only logical.  The  NTSC  value  would  be
1022730, which is very close to  the  PAL,  and  therefore  won't  cause  a
compatability problem converting European and NTSC tapes. I would stick  to
using the PAL value just in case.

  In TAP version $01 files, the data value of  $00  has  been  re-coded  to
represent values greater than 255 * 8. When a  $00  is  encountered,  three
bytes will follow which are the actual time (in cycles) of a pulse, and the
above formula does not apply.  The  three  bytes  are  stored  in  LOW/HIGH

  The actual interpretation of the serial data takes a little more work  to
explain.  The  typical  ROM  tape  loader  (and  the  turbo  loaders)  will
initialize a timer with a specified value and start it  counting  down.  If
either the tape data changes or the timer runs out, an IRQ will occur.  The
loader will determine which condition caused the  IRQ.  If  the  tape  data
changed before the timer ran out, we have a short pulse, or a "0"  bit.  If
the timer ran out first, we have a long pulse, or a  "1"  bit.  Doing  this
continuously and we decode the entire file.