SIDin #2 — David Whittaker's Lazy Jones music routine (Stefano Tognon, 2002)

Tognon's reverse-engineering of the hvsc rip of Lazy Jones (Terminal Software, 1984) — david-whittaker's earliest routine in the wiki and, in his words, "the most simple I had seen until now": two voices of raw frequency pairs, one step every twelve frames, one fixed sound and no effects at all. "You probably will be surprise to see a so simple engine gives this magic music: maybe it is all David talent!" The listing is reproduced in full; this page keeps the mechanics and the numbers. The Panther driver of 1986 on whittaker-player is a different, far larger routine.

Key claims#

• Basis: "the Sid rip present in HVSC. How you see it is not complete with the sound SX [effects]: however it is just what we need." The rip "was ripped from a cracked version of the game, as it contains a faulty track (not used in the rip, but present)" — the bytes after tune 21 under the label wrong, which "you can listen to … if you change the source and you active it again". The copyright note: "contact the copyright owner author for making any business use of it."
• Data: "The 20 music peaces are with the same length (the first is a case we see after): 80h bytes of data. Each peaces of music contains the low/high frequency to set in the sid registers for the 2 voices used (so 4 bytes)." So a piece is 32 steps; the index is incremented by 4 per step and compared with $80 (COMP) to loop the piece. "Tune 1 is an exception: it is the running of tune 10 and 11 in sequences" — for song 0 init patches the compare value to $00, so the index wraps after 256 bytes, and Tune10 and Tune11 are consecutive in memory (Tune1: is the same label as Tune10:).
• Timing: "There's not a note duration: each notes have the same length." DELAY is reloaded with $0C and decremented once per call; only when it reaches zero is a step played.
• One sound: "There's not special SID effect: all the notes use the same setting: Attack/Decay of 66h and 21h wave form. As the Decay/Released is not set (so 0), there's a simple 20h to 21h setting in the wave form when the note changes." setAD writes $1F to $D418 and $66 to $D405 and $D40C; initEngine writes $FF to $D416 and $03 to $D417, sets MUSIC = 1, INDEX = 0, DELAY = $0C and the pattern pointer. Sustain/release is never written (adsr-envelope).
• Per step (IRQ): read the low byte at ($B4),y; if it is zero skip the voice entirely (beq skip1), otherwise write $D400, $D401, then $20 and $21 to $D404; advance the pointer by 2; the same for voice 2 ($D407, $D408, $D40B); INDEX += 4; at $80 reset the index and reload the pointer from PATTL/PATTH; jmp $EA31. Voice 3 is never touched. Rests are $00,$00,$00,$00 rows (Tune17).
• Variables: MUSIC = $03E0 (1 = play), INDEX = $03E1, DELAY = $03E3, PATTL = $03E4, PATTH = $03E5; the running pointer in $B4/$B5; tableLo/tableHi hold the 21 tune addresses and init copies the chosen one into the LOP/HIP immediates (self-modifying code).
• PSID header as typed: org 1921, .byte "PSID", .word $0200 (v2), .word $7C00 (data offset $7C), load and init = init, play = IRQ, .word $1500 = 21 songs, .word $0100 = start song 1, the strings "lazy Jones", "David Whittaker", "1984 Terminal Software" (sid-format).
• What is in the notes: Tognon annotates Tune10 with note names — $26,$04,$3A,$04 ; . H1 C2, $4D,$08,$75,$08 ; . H2 C3, $30,$04,$8F,$0C ; ^ H1 F#3, $61,$08,$8F,$0A ; ^ H2 D#3 … — voice 1 alternating H1/H2 (B in two octaves, German naming) on every step while voice 2 carries the tune, with ./^ marking each row "low"/"hi" (octave-bass). His guess at the magic: "I think from the music notes themselves! Maybe David had used some frequencies that give a better result using a little less or high values from the expected."
• Rastertime: "this engine use a maximum of 4 rasterlines when it is piking a new note, and only 1 rasterline in the other case" (sid-player-routine). "Don't forget that the game is from 1984, so it was develop at the beginning of the Commodore time."

Practical takeaways#

• A complete SID tune in 32 rows × 2 voices with a single sawtooth blip (AD $66, SR $00, gate retriggered by $20→$21) — the smallest possible "instrument" and a lesson in letting the note choice do the work (instrument-design).
• The retrigger is the whole articulation: dropping and raising the gate in the same call restarts the attack; with sustain 0 each note is a fixed-length blip whatever the step length (hard-restart is the later refinement of the same idea).
• Chaining two pieces by widening the loop compare ($80 → wrap at $00) is a two-byte song structure (orderlist).

Notable quotes#

"Now I simply show you how David made this music. You probably will be surprise to see a so simple engine gives this magic music: maybe it is all David talent!"

"this is a very simple engine music (I can say that it is the most simple I had seen until now)."

Relevance to the wiki#

The documented Lazy Jones routine for whittaker-player and david-whittaker (until now cited there only through s-sidin02-matt-gray-driller); a hand-written PSID header and a rip from a cracked source for sid-format and hvsc; a 1984 octave bass for octave-bass; the smallest driver on sid-player-routine.

Pages touched#

whittaker-player · david-whittaker · sid-format · hvsc · sid-player-routine · octave-bass · adsr-envelope · sidin

To verify#

• (unverified) Read with the datasheet: $D417 = $03 routes voices 1 and 2 through the filter with resonance 0, $D418 = $1F is low-pass on at volume 15, $D416 = $FF puts the cutoff at its maximum — so both voices pass through a wide-open low-pass filter; Tognon prints the register comments only.
• (unverified) The wrong block ($3F,$3F,$37,$7F … $18,$18,$18,$18 …) looks like character or sprite graphics that the cracked image left in place of music data.
• (unverified) Tognon's note names use German H for B; his octave numbers are not mapped to tracker octaves here.