In this document I'll try to describe audio file formats used in many (older) Electronic Arts games. Described are formats for music, sound effects, speech and movie soundtracks.
The games using these formats include: NBA Live'96, NHL'96, FIFA'96, The Need For Speed, NHL'97. Maybe many more, e.g.: NHL'95.
The files this document deals with have extensions: .ASF, .AS4, .KSF, .EAS, .SPH, .BNK, .CRD, .TGV. Note that the files described here may have other extensions (and the same structure!): Electronic Arts tends to change extensions from game to game.
Throughout this document I use C-like notation.
All numbers in all structures described in this document are stored in files using little-endian (Intel) byte order.
1. ASF/AS4 Music Files
The music in many Electronic Arts games is in .ASF/.AS4 stand-alone files. These files have the block structure analoguous to RIFF. Namely, these files are divided into blocks (without any global file header like RIFFs have). Each block has the following header:
dwSize -- size of the block (in bytes) INCLUDING this header.
Further I'll describe the contents of blocks of all block types in ASF/AS4 files. When I say "block begins with..." that means "the contents of that block (which begin just after ASFBlockHeader) begin with...". Quoted strings are block IDs.
"1SNh": header block. This is the first block in ASF/AS4. This block begins with the structure describing the audio stream:
bBits -- if multiplied by 8 gives the resolution of (decompressed) sound data, that is 1 means 8-bit and 2 means 16-bit.
bChannels -- channels number: 1 for mono, 2 for stereo.
bCompression -- if 0x00, the data in the file is not compressed: signed 8-bit PCM or signed 16-bit PCM. If this byte is 0x02, the audio data is compressed with IMA ADPCM. Note that non-compressed 8-bit files use SIGNED format! Signed 16-bit data may be sent to the wave output without any additional conversions, while signed 8-bit data should be converted to unsigned format. For example you can do that so: unsigned8Bit=signed8Bit+0x80 or, just the same: unsigned8Bit=signed8Bit^0x80 (this's a bit faster).
bType -- type of file: always 0x00 for ASF/AS4 (multi-block) files.
dwNumSamples -- number of samples in the file. May be used for song length (in seconds) calculation.
dwLoopStart -- beginning of the repeat loop (in samples). 0xFFFFFFFF means no loop.
dwLoopLength -- length of the repeat loop (in samples). Zero for no loop.
dwDataStart -- in ASF/AS4 files this is not used (equal to 0).
After the EACSHeader the first chunk of sound data comes. If the data isn't compressed, it's just signed 8/16-bit PCM. If the data is compressed, it starts with a small chunk header:
struct ASFChunkHeader
{
DWORD dwOutSize;
LONG lIndexLeft;
LONG lIndexRight;
LONG lCurSampleLeft;
LONG lCurSampleRight;
};
dwOutSize -- size of uncompressed audio data in this chunk (in samples).
lIndexLeft, lIndexRight, lCurSampleLeft, lCurSampleRight are initial values for IMA ADPCM decompression routine for this chunk (for left and right channels respectively). I'll describe the usage of these further when I get to IMA ADPCM decompression scheme.
Note that the structure above is ONLY for stereo files. For mono there're just no lIndexRight and lCurSampleRight fields.
After this chunk header the compressed data comes. You may find IMA ADPCM decompression scheme description further in this document.
Hereafter by "chunk" I mean the audio data in the "1SNd" data block, that is, compressed data which starts after ASFChunkHeader.
"1SNd": data block. If no compression is used these blocks contain just signed 8/16-bit PCM audio data. Otherwise the data in each of these blocks begins with the same ASFChunkHeader described above and after that comes compressed data.
Note that the first chunk of audio data is in "1SNh" block, along with the global EACS header!
"1SNl": loop block. This block defines looping point for the song. It contains only DWORD value, which is the looping jump position (in samples) relative to the start of the song. Note that you should make the jump NOT when you encounter this block but when you come across the "1SNe" block which may appear some "1SNd" data blocks after this block!
"1SNe": end block. This block indicate the end of audio stream. Make looping jump when you encounter it. It contains no data and its size is 8 bytes that is the size of ASFBlockHeader.
Interesting that some AS4 files contain audio data beyond this block. This should be considered as non-standard feature not worth to support.
2. KSF Music Files
Some EA games use other format for music/speech files: .KSF. These files begin with "KWK`" ID string. Following this ID, comes PATl header. It begins with "PATl" ID string and its size is 56 bytes (always?) including its ID string. After PATl header comes TMpl header:
bBits -- resolution of sound data (0x10 for 16-bit, 0x8 for 8-bit).
bChannels -- channels number: 1 for mono, 2 for stereo.
bCompression -- if 0x00, the data in the file is not compressed: signed 8-bit PCM or signed 16-bit PCM. If this byte is 0x02, the audio data is compressed with IMA ADPCM. See the note for EACS header above.
wSampleRate -- sample rate for the file.
dwNumSamples -- number of samples in the file. May be used for song length (in seconds) calculation. Should be divided by 2 for mono sound.
After TMpl header comes sound data. For compressed files, IMA ADPCM compression
is used (see below).
3. IMA ADPCM Decompression Algorithm
During the decompression four LONG variables must be maintained for stereo stream: lIndexLeft, lIndexRight, lCurSampleLeft, lCurSampleRight and two -- for mono stream: lIndex, lCurSample. At the beginning of each "1SNd" data block and at the beginning of the file -- when processing "1SNh" block -- you must initialize these variables using the values in ASFChunkHeader.
Note that LONG here is signed.
Here's the code which decompresses one byte of IMA ADPCM compressed stereo stream. Other bytes are processed in the same way.
BYTE Input; // current byte of compressed data
BYTE Code;
LONG Delta;
Code=HINIBBLE(Input); // get HIGHER 4-bit nibble
Delta=StepTable[lIndexLeft]>>3;
if (Code & 4)
Delta+=StepTable[lIndexLeft];
if (Code & 2)
Delta+=StepTable[lIndexLeft]>>1;
if (Code & 1)
Delta+=StepTable[lIndexLeft]>>2;
if (Code & 8) // sign bit
lCurSampleLeft-=Delta;
else
lCurSampleLeft+=Delta;
// clip sample
if (lCurSampleLeft>32767)
lCurSampleLeft=32767;
else if (lCurSampleLeft<-32768)
lCurSampleLeft=-32768;
lIndexLeft+=IndexAdjust[Code]; // adjust index
// clip index
if (lIndexLeft<0)
lIndexLeft=0;
else if (lIndexLeft>88)
lIndexLeft=88;
Code=LONIBBLE(Input); // get LOWER 4-bit nibble
Delta=StepTable[lIndexRight]>>3;
if (Code & 4)
Delta+=StepTable[lIndexRight];
if (Code & 2)
Delta+=StepTable[lIndexRight]>>1;
if (Code & 1)
Delta+=StepTable[lIndexRight]>>2;
if (Code & 8) // sign bit
lCurSampleRight-=Delta;
else
lCurSampleRight+=Delta;
// clip sample
if (lCurSampleRight>32767)
lCurSampleRight=32767;
else if (lCurSampleRight<-32768)
lCurSampleRight=-32768;
lIndexRight+=IndexAdjust[Code]; // adjust index
// clip index
if (lIndexRight<0)
lIndexRight=0;
else if (lIndexRight>88)
lIndexRight=88;
// Now we've got lCurSampleLeft and lCurSampleRight which form one stereo
// sample and all is set for the next input byte...
Output((SHORT)lCurSampleLeft,(SHORT)lCurSampleRight); // send the sample to output
HINIBBLE and LONIBBLE are higher and lower 4-bit nibbles:
Note that depending on your compiler you may need to use additional nibble separation in these defines, e.g. (((byte) >> 4) & 0x0F).
StepTable and IndexAdjust are the tables given in the next section of this document.
Output() is just a placeholder for any action you would like to perform for decompressed sample value.
Of course, this decompression routine may be greatly optimized.
As to mono sound, it's just analoguous:
Code=HINIBBLE(Input); // get HIGHER 4-bit nibble
Delta=StepTable[lIndex]>>3;
if (Code & 4)
Delta+=StepTable[lIndex];
if (Code & 2)
Delta+=StepTable[lIndex]>>1;
if (Code & 1)
Delta+=StepTable[lIndex]>>2;
if (Code & 8) // sign bit
lCurSample-=Delta;
else
lCurSample+=Delta;
// clip sample
if (lCurSample>32767)
lCurSample=32767;
else if (lCurSample<-32768)
lCurSample=-32768;
lIndex+=IndexAdjust[Code]; // adjust index
// clip index
if (lIndex<0)
lIndex=0;
else if (lIndex>88)
lIndex=88;
Output((SHORT)lCurSample); // send the sample to output
Code=LONIBBLE(Input); // get LOWER 4-bit nibble
// ...just the same as above for lower nibble
Note that HIGHER nibble is processed first for mono sound and corresponds to LEFT channel for stereo.
.TGV movies have the block structure analoguous to that of ASF/AS4.
Video-related data is in "kVGT" and "fVGT" (or "TGVk" and "TGVf") blocks and sound-related data is just in the same blocks as in ASF/AS4: "1SNh", "1SNd", "1SNl", "1SNe".
So, to play TGV movie soundtrack, just walk blocks chain, skip video blocks
and process sound blocks.
6. Sound/Speech Files: .EAS, .SPH
Some sounds and all speech are usually in .EAS and .SPH files.
These files have the header which is just the same as EACSHeader structure described above with two additions: (bType) is always 0xFF for sound/speech files, (dwDataStart) is the starting position of audio data relative to the beginning of the file.
After the header, starting at (dwDataStart), comes audio data, up to the end of the file. The data is either non-compressed or IMA ADPCM compressed depending on the (bCompression) byte in the header. If it's IMA ADPCM compressed, there're no initial values for samples and indices at the beginning of the audio data. Just initialize them all to zeroes and start decompression at (dwDataStart).
7. Sound Effects in .BNK/.CRD Files
Most of sound effects are stored in .BNK and .CRD resource files. Those .BNKs and .CRDs may contain several sounds. They begin with some seemingly meaningless data, but after some junk of that data (typically starting at position 0x228, but not necessarily) come several EACS headers describing all sounds in .BNK/.CRD. Each EACS header has almost the same format as described above with some minor changes (some fields have different placement):
and with the same two additions just as for .EAS/.SPH speech/sound: (bType) is always 0xFF, (dwDataStart) is the starting position of sound data relative to the beginning of the .BNK/.CRD file containing that sound. So, what you need to do is just search in .BNK/.CRD for "EACS" ID string and read EACSHeader from the position where you found "EACS". And the same for all sounds contained within .BNK/.CRD. The sound data itself (for each EACS header describing it) starts at (dwDataStart) and its size may be computed using (dwNumSamples) EACSHeader field (for example) with the following formula:
Size=dwNumSamples*SampleSize/CompressionRatio, where:
CompressionRatio=1 for non-compressed sounds,
2 for 8-bit IMA ADPCM compressed sounds,
4 for 16-bit IMA ADPCM compressed sounds,
SampleSize=bChannels*bBits (1 for mono 8-bit, 2 for mono 16-bit, etc.).
So, starting at (dwDataStart) comes just either PCM audio data (as described above for .EAS/.SPH files) or IMA ADPCM compressed data (without initial sample/index values, just as in .EAS/.SPH). Set CurSample(Left/Right) and Index(Left/Right) to zeroes and start the decompression.
8. Credits
Vladan Bato ()
The IMA ADPCM decompression scheme above is based on that described in his AUD3.TXT document.
Peter Pawlowski (, )
Pointed out corrections in IMA ADPCM decoder.
Toni Wilen ()
Provided me with info on KSF files, PATl and TMpl headers. Toni Wilen is the author of SNDVIEW utility (available on his pages) which decompresses Electronic Arts audio files and compresses WAVs back into EA formats.
Denis AUROUX (MXK) ()
Additional info on EACS headers. The author of The Unofficial NFS File Format Specifications.
Valery V. Anisimovsky ()
On these sites you can find my GAP program which can search for audio files in .BNK/.CRD resources, and play back .ASF/.AS4 songs, .BNK/.CRD sounds, .EAS/.SPH speech and soundtracks of .TGV movies. There's also complete source code of GAP and all its plug-ins there, including ASF plug-in, which could be used for further details on how you can deal with these formats.
