In order to conduct behavioral reverse eng of the historical TRAU we have in
our lab (Nokia TCSM2), we will need to feed TRAU-UL frames to its Ater interface
(first GSM 08.60 FR and EFR, then later GSM 08.61 HR), exercising various
combinations of speech parameters and out-of-band control bits.  In order to
conduct these exercises, we need some test file formats of our own invention:

* The source format for human editing is ASCII-based, friendly toward arbitrary
  manipulation of speech parameters and flags.  We call this ad hoc ASCII source
  format *.tsrc, for TRAU-UL source.

* trau-ul-compile program in the present code repository will compile these
  *.tsrc source files into *.tul (TRAU-UL) binary format;

* The Ater test program in ice1-trau-tester repository will read binary frames
  from our ad hoc *.tul file format and turn them into actual TRAU-UL bits
  going to E1.

*.tul binary format
===================

Each frame (FR or EFR) is stored as a fixed record of 34 bytes.  The format for
FR is:

* 33 bytes of standard RTP format representing 260 bits of payload;
* 1 byte of BFI+SID flags.

The format for EFR is:

* 31 bytes of standard RTP format representing 244 bits of payload;
* 1 byte controlling CRC inversion;
* 1 byte spare;
* 1 byte of BFI+SID flags.

Because each *.tul record begins with an FR or EFR frame in standard RTP
encoding, and the latter includes a signature (0xD or 0xC) in the upper nibble
of the first byte, each *.tul record can be trivially checked if it is for FR
or EFR testing.

Byte of BFI+SID flags
---------------------

Bit 0x80 is BFI, bits 0x03 are SID, the rest are reserved.

EFR CRC inversion control
-------------------------

We can generate EFR TRAU-UL frames with either correct or inverted CRC.
Furthermore, GSM 08.60 frame format for EFR has 5 separate CRC fields covering
different parts of the frame, and we can independently control the correct-or-
inverted choice for each.

The CRC manipulation byte included in *.tul frames for EFR has 5 used and 3
reserved bits.  The 5 most significant bits correspond to the 5 CRC fields in
GSM 08.60 frame format; control bit equal to 0 means send correct CRC, otherwise
send inverted CRC.

ASCII source format
===================

Example for FR:

Frame_FR {
	LARc 9 23 15 8 7 3 3 2
	sf 40 0 0 0 4 4 4 4 4 4 4 4 4 4 4 4 4
	sf 40 0 0 0 4 4 4 4 4 4 4 4 4 4 4 4 4
	sf 40 0 0 0 4 4 4 4 4 4 4 4 4 4 4 4 4
	sf 40 0 0 0 4 4 4 4 3 4 4 4 4 4 4 4 4
}

Example for EFR:

Frame_EFR {
	BFI 0
	LPC 4 47 180 144 62
	sf 342 11 0 1 15 1 13 0 3 0 3 0 3
	sf 54 1 8 8 5 8 1 0 0 1 1 0 0
	sf 342 0 0 0 0 0 0 0 0 0 0 0 0
	sf 54 11 0 0 0 0 0 0 0 0 0 0 0
	SID 0
}

Each frame to be encoded must be given as a block beginning with "Frame_FR {"
or "Frame_EFR {" opening line and closing with "}".  The opening brace must be
on the same line as the frame type keyword, and that line must have no other
fields; the closing brace must be on its own line.  (The actual parser is line-
based, despite the appearance of braces.)

Each interior line shall consist of a keyword followed by numbers; the keyword
indicates which part of the frame is being set.  LARc for FR or LPC for EFR sets
the frame-global speech parameters (8 for FR and 5 for EFR); sf sets subframe
parameters.  Every frame must have one LARc/LPC line and exactly 4 sf lines.

BFI and SID settings are optional.  If BFI is not set, it defaults to 0; if SID
is not set, it is computed from the speech parameters per standard bit counting
rules of GSM 06.31 or 06.81 section 6.1.1.

For EFR only, there will also be an optional crc-inv setting that sets the CRC
manipulation byte, defaulting to 0 (correct CRC).

Aside from the required line structure, white space (' ' and '\t') can appear
anywhere and is ignored.  '#' character introduces a comment, continuing to the
end of the line.

All keywords are case-insensitive.