The XM module format description for XM files version $0104. By Mr.H of Triton in 1994. - Be prepared! Are you sure you want to know? :-) << A LOT of comments, corrections and notes added by Guru and Alfred of Sahara Surfers in 1995. All additions are included in double angle brackets << like this >> and everything outside them is left intact. Note the HUGE amount of errors in the original text. Triton: Perhaps you'd like to include THIS in the next FT2 release? :-) Not copyrighted, released into public domain (at least the additions by us). Feel free to do what you wish. Credits please. Remember - the added information is based on REAL XM files. >> ============================================================================ ****************************** * The XM file structure: * ****************************** Offset Length Type 0 17 (char) ID text: 'Extended module: ' << Nope: 'Extended Module: ' >> 17 20 (char) Module name, padded with zeroes 37 1 (char) $1a 38 20 (char) Tracker name 58 2 (word) Version number, hi-byte major and low-byte minor The current format is version $0103 << Nope, it's $0104. Format versions below $0104 have a LOT of differences. Check this field! >> 60 4 (dword) Header size << Calculated FROM THIS OFFSET, not from the beginning of the file! >> +4 2 (word) Song length (in patten order table) +6 2 (word) Restart position +8 2 (word) Number of channels (2,4,6,8,10,...,32) +10 2 (word) Number of patterns (max 256) +12 2 (word) Number of instruments (max 128) +14 2 (word) Flags: bit 0: 0 = Amiga frequency table (see below); 1 = Linear frequency table +16 2 (word) Default tempo +18 2 (word) Default BPM +20 256 (byte) Pattern order table Patterns: --------- ? 4 (dword) Pattern header length +4 1 (byte) Packing type (always 0) +5 2 (word) Number of rows in pattern (1..256) +7 2 (word) Packed patterndata size << Note! This is zero if the pattern is completely empty and no pattern data follows! >> ? ? Packed pattern data Instruments: ------------ ? 4 (dword) Instrument size << header that is >> +4 22 (char) Instrument name +26 1 (byte) Instrument type (always 0) << This seems pretty random, don't assume it's zero >> +27 2 (word) Number of samples in instrument If the number of samples > 0, then the this will follow: ! +29 4 (dword) Sample header size ! +33 96 (byte) Sample number for all notes ! +129 48 (byte) Points for volume envelope ! +177 48 (byte) Points for panning envelope ! +225 1 (byte) Number of volume points ! +226 1 (byte) Number of panning points ! +227 1 (byte) Volume sustain point ! +228 1 (byte) Volume loop start point ! +229 1 (byte) Volume loop end point ! +230 1 (byte) Panning sustain point ! +231 1 (byte) Panning loop start point ! +232 1 (byte) Panning loop end point ! +233 1 (byte) Volume type: bit 0: On; 1: Sustain; 2: Loop ! +234 1 (byte) Panning type: bit 0: On; 1: Sustain; 2: Loop ! +235 1 (byte) Vibrato type ! +236 1 (byte) Vibrato sweep ! +237 1 (byte) Vibrato depth ! +238 1 (byte) Vibrato rate ! +239 2 (word) Volume fadeout ! +241 2 (word) Reserved << "Instrument Size" field tends to be more than the actual size of the structure documented here (it includes also the extended instrument sample header above). So remember to check it and skip the additional bytes before the first sample header >> ! ! Sample headers: ! --------------- ! ! ? 4 (dword) Sample length ! +4 4 (dword) Sample loop start ! +8 4 (dword) Sample loop length ! +12 1 (byte) Volume ! +13 1 (byte) Finetune (signed byte -16..+15) ! +14 1 (byte) Type: Bit 0-1: 0 = No loop, 1 = Forward loop, ! 2 = Ping-pong loop; ! 4: 16-bit sampledata ! +15 1 (byte) Panning (0-255) ! +16 1 (byte) Relative note number (signed byte) ! +17 1 (byte) Reserved ! +18 22 (char) Sample name << Note! After the instrument header the file contains ALL sample headers for the instrument followed by the sample data for all samples >> ! ! Sample data: ! ------------ ! ! ? ? Sample data (signed): The samples are stored ! as delta values. To convert to real data: ! ! old=0; ! for i=1 to len ! new=sample[i]+old; ! sample[i]=new; ! old=new; *********************** * Pattern format: * *********************** The patterns are stored as ordinary MOD patterns, except that each note is stored as 5 bytes: ? 1 (byte) Note (0-71, 0 = C-0) << Hah. Actually note numbers are 1-96, 97 being the key-off note (try to find that in the documentation!) >> +1 1 (byte) Instrument (0-128) << More like 1-128 >> +2 1 (byte) Volume column byte (see below) +3 1 (byte) Effect type +4 1 (byte) Effect parameter A simle packing scheme is also adopted, so that the patterns not become TOO large: Since the MSB in the note value is never used, if is used for the compression. If the bit is set, then the other bits are interpreted as follows: bit 0 set: Note follows 1 set: Instrument follows 2 set: Volume column byte follows 3 set: Effect follows 4 set: Guess what! It is very simple, but far from optimal. If you want a better, you can always repack the patterns in your loader. ****************************** * Volumes and envelopes: * ****************************** The volume formula: FinalVol=(FadeOutVol/65536)*(EnvelopeVol/64)*(GlobalVol/64)*(Vol/64)*Scale; << The FadeOutVol is originally 65535 and is decremented by instrument.fadeout each tick after note is released >> The panning formula: FinalPan=Pan+(EnvelopePan-32)*(128-Abs(Pan-128))/32; << Note that the panning envelope values range from 0-64, not -128 - +127 >> Envelope: --------- The envelopes are processed once per frame, instead of every frame where no new notes are read. This is also true for the instrument vibrato and the fadeout. Since I am so lazy and the tracker is rather self-explaining I am not going to write any more for the moment. << Anyone have ANY idea how instrument autovibrato works? >> ******************************** * Periods and frequencies: * ******************************** PatternNote = 0..95 (0 = C-0, 95 = B-7) << Blah: Note = 1..96, 1 = C-0, 97 = key off >> FineTune = -128..+127 (-128 = -1 halftone, +127 = +127/128 halftones) RelativeTone = -96..95 (0 => C-4 = C-4) RealNote = PatternNote + RelativeTone; (0..118, 0 = C-0, 118 = A#9) Linear frequence table: ----------------------- Period = 10*12*16*4 - Note*16*4 - FineTune/2; Frequency = 8363*2^((6*12*16*4 - Period) / (12*16*4)); Amiga frequence table: ---------------------- Period = (PeriodTab[(Note MOD 12)*8 + FineTune/16]*(1-Frac(FineTune/16)) + PeriodTab[(Note MOD 12)*8 + FineTune/16]*(Frac(FineTune/16))) *16/2^(Note DIV 12); (The period is interpolated for finer finetune values) Frequency = 8363*1712/Period; << The interpolation code above doesn't work because of several reasons: 1. It does not interpolate (try adding 1 to the PeriodTab index in the second line) 2. It may go past the table beginning for negative finetune values Write your own interpolation routine instead - it's not that hard. >> PeriodTab = Array[0..12*8-1] of Word = ( 907,900,894,887,881,875,868,862,856,850,844,838,832,826,820,814, 808,802,796,791,785,779,774,768,762,757,752,746,741,736,730,725, 720,715,709,704,699,694,689,684,678,675,670,665,660,655,651,646, 640,636,632,628,623,619,614,610,604,601,597,592,588,584,580,575, 570,567,563,559,555,551,547,543,538,535,532,528,524,520,516,513, 508,505,502,498,494,491,487,484,480,477,474,470,467,463,460,457); << Note! The period table is made for 1-based note numbers, so in practise it contains the period values for B-3 to G#4. Fun. >> ************************* * Standard effects: * ************************* 0 Appregio << ;) >> 1 (*) Porta up 2 (*) Porta down 3 (*) Tone porta 4 (*) Vibrato 5 (*) Tone porta+Volume slide 6 (*) Vibrato+Volume slide 7 (*) Tremolo 8 Set panning 9 Sample offset A (*) Volume slide B Position jump C Set volume D Pattern break E1 (*) Fine porta up E2 (*) Fine porta down E3 Set gliss control E4 Set vibrato control E5 Set finetune E6 Set loop begin/loop E7 Set tremolo control E9 Retrig note EA (*) Fine volume slide up EB (*) Fine volume slide down EC Note cut ED Note delay EE Pattern delay F Set tempo/BPM G Set global volume H (*) Global volume slide K Key off << also note number 97 >> L Set envelope position P (*) Panning slide R (*) Multi retrig note T Tremor X1 (*) Extra fine porta up X2 (*) Extra fine porta down (*) = If the command byte is zero, the last nonzero byte for the command should be used. << The commands are reasonably Protracker compatible although not all PT "features" (some might call them replay routine bugs) are implemented. >> ********************************* * Effects in volume column: * ********************************* All effects in the volume column should work as the standard effects. The volume column is interpreted before the standard effects, so some standard effects may override volume column effects. Value Meaning 0 Do nothing $10-$50 Set volume Value-$10 : : : : : : $60-$6f Volume slide down $70-$7f Volume slide up $80-$8f Fine volume slide down $90-$9f Fine volume slide up $a0-$af Set vibrato speed $b0-$bf Vibrato $c0-$cf Set panning $d0-$df Panning slide left $e0-$ef Panning slide right $f0-$ff Tone porta << And here goes one more section that this document really should have had: ******************** * General layout * ******************** The layout of a typical XM is like this: XM header (up to and excluding header size field, now 60 bytes) Rest of the header (length including header size field) Pattern 0 header (length in header) Pattern 0 data (length in header) Pattern 1 header (length in header) Pattern 1 data (length in header) ... (* number of patterns) Instrument 0 header (length in size field is this and next one together) if (numSample > 0) Extra header Sample 0 header (size in instrument extra header) Sample 1 header (size in instrument extra header) ... (* number of samples in instrument) Sample 0 data (length in sample header) Sample 1 data (length in sample header) ... (* number of samples in instrument) Instrument 1 header (length in size field is this and next one together) ... (* number of instruments) >> ============================================================================ This should be just about everything (I hope?). You will probably need some information about the MOD format and maybe about S3M. Have fun! << And we sure had! >> Fredrik Huss / Mr.H of Triton << Jarno Paananen & Petteri Kangaslampi - Guru & Alfred / S2 If you have questions that you think we might be able to answer, have found out more mistakes in this document, or know how instrument autovibrato works, e-mail us at: Jarno: jpaana@kauhajoki.fi _or_ Petteri: pekangas@sci.fi >>