Defense Mechanism — LSDj Noise Channel Tutorial (v9.2.L) (video, 2022)

Written because the noise channel was rebuilt in 9.1.0 and the standard reference no longer applied: "there's a tutorial by Bowie Meets Robot about the noise channel for version four and it's really good … but it doesn't really apply to the new noise channel because it's been just totally revamped."

Key claims#

• The notes. Long-loop (noisy) shapes run from 00 to about 3B; past either end you cross into short-loop, pitched noise, which the screen labels with note names — "it's not exact pitch, it just is the closest note in pitch to what the noise sound is". Up/down moves by octave. The shapes are now ordered low to high, which "didn't used to be the case … makes choosing a descending progression a lot easier". The bottom of the range is "literally noisy, almost like static"; it turns crunchy around 20.
• Commands that now work on noise: C chords (real arpeggios — he builds an open fifth), P pitch bends ("smooth pitch sweeps … higher or lower values cause even crazier sounds"), V vibrato with speed and depth, combinable with P, and S — which he now rarely needs, "but it is kind of cool to chain it to the Z command" for randomised shapes (randomization-and-probability).
• Why PITCH SAFE exists, in hardware terms. The noise is generated by "a linear feedback shift register, and sometimes … there's a bug in it essentially that causes it to create zeros, and that results in silence". On hardware — "especially like the original brick DMG" — a pitch slide can therefore simply cut out with no K command. SAFE re-initialises the channel on every pitch change so the sound "will never cut out", at the cost of smoothness: "it creates a harsher sound, it's just a totally different texture", recoverable in part by slowing the rate of change. Affected models include DMG, old CGB and some Game Boy Advance units — "the GB Player for GameCube definitely has it".
• Emulators have caught up. "I'm using SameBoy — SameBoy emulates this random cutout, and I think even BGB does too. Older versions of SameBoy and BGB don't emulate that; the noise channel will basically never cut out, as if it were designed perfectly." So the mute is now reproducible on the desktop.
• COMMAND RATE (with PITCH set to TICK) changes the rate of P and V in the noise channel — "another way to slow down your pitch slides … in case PFF is too fast", and a way to keep some smoothness while staying on SAFE.
• Hardware LENGTH vs the envelope. He prefers the envelope for shaping, and names the one place LENGTH still wins: it overrides an E command, "so it kind of puts a cap on how long this envelope can actually be" — useful with a rising envelope, and for carrying an old patch forward unchanged.
• R retrig as a dynamics tool. Hardware envelopes are single smooth slopes, so a decaying repeat cannot be written with E; R with a volume digit "will reduce this by one level and keep these numbers", giving ghost snares, quieter repeats and, with a four-tick retrig, triplet fills. A variant: leave the instrument untriggered and change only the shape.
• Snare design: set the envelope rather than chaining E commands ("a much faster and easier way … and then I can just work on setting transpose commands"), add a couple of rising transposes, and stop with A20 (stop table) or K. He keeps SAFE on for "a little bit more crunchiness in the transient". Pitched noise plus a P command turns the same instrument into a tonal snare — and P keeps running after the note, so stop it deliberately.
• Hi-hats: volume falls naturally as the shape rises, "so this can be a way to add dynamics without even doing any kind of E commands"; a longer envelope decay gives the open hat.

Practical takeaways#

• Treat the post-9.1.0 noise channel as a pitched channel: C, P and V all work, and the shapes are ordered, so descending fills are just descending numbers.
• Use SAFE when the part must never drop out and you can accept the crunch; slow the pitch rate or lower COMMAND RATE to buy some smoothness back.
• R with a volume digit is the noise channel's dynamics control; the envelope is its shape control; LENGTH is a hard ceiling on both.

Notable quotes#

"Sometimes the randomization … creates zeros, and that results in silence."

"At the expense of the smoothness, you can now basically have reassurance that your snare will never cut out."

Relevance to the wiki#

game-boy-apu records an open contradiction: the within-mode noise mutes reported in 2021 "have no hardware explanation in the sources", and BGB/SameBoy were said not to show them. This source addresses both halves — an LFSR-produces-zeros explanation, and current emulators reproducing the mute — without fully matching Pan Docs' mechanism, so both accounts are now stated side by side.

To verify#

• His LFSR account (the register "randomly produces zeros") is not the mechanism Pan Docs describes (an all-ones lock-up when switching 15-bit → 7-bit shortly after a trigger, source: s-pandocs-audio-details). Two descriptions of the same symptom; neither source reconciles them (unverified).
• The claim that current SameBoy and BGB emulate the cutout is his, from 2022, and the wiki's earlier source says the opposite for its own time (source: s-intense-tech-19-new-noise-and-910-news) — treat as a change over time, not a correction (unverified).
• The long-loop range "00 to about 3B" is read off the screen while scrolling (unverified).

Pages touched#

instrument-design · game-boy-apu · retrigger · randomization-and-probability · pitch-slide-and-portamento · vibrato · lsdj · defense-mechanism