drum kit levels on older amyboard world examples

[bwhitman]

a lot of the older amyboardworld examples, like my beloved aphex_twin_polygon, now sound bad because the drum kits were leveled to be closer to other synths, and the older sketches boosted the levels of drums to make them sound good. now they're too loud!

[bwhitman]

Root cause: amy#910 turned an inert amp= on drum synths into a real 4–5× gain

Not a re-leveling regression — the old sketches' amp=4/amp=5 on the drum synth used to be completely ignored, and now it applies.

shorepine/amy@18ca7a7 ("Drum kits: keep per-drum gain in the note map's velocity scale, freeing amp CONST as the channel level", shorepine/amy#910, merged 2026-07-16).

Before — kit patches baked each drum's balancing gain into the note template's amp CONST, and the patch header carried a5:

/* 384: TR-808 */ "v0w7a5if3Z...io35,0,0,1,0,n60l%va4.02G0i%ip0ZZ..."
                       ^^                              ^^^^^

Every hit rewrote osc amp, so a user's amy.send(synth=10, amp=5) was clobbered by the next drum — inert.

After — the gain moved into the mapping's linear velocity scale, and amp CONST stays at its 1.0 reset default so it works as a real per-channel level (which was the point: MIDI CC 7 / the mixer Level slider previously had no effect on drums):

/* 384: TR-808 */ "v0w7if3Z...io35,0,0,4.02,0,n60l%vG0i%ip0ZZ..."
                                        ^^^^

Old sketches that set amp=4/amp=5 believing they were compensating for quiet drums now get exactly that multiplier on top of a correctly-leveled kit.

Measured

Built AMY's CPython module at the pre-change pin (145e3d8f) and at tulipcc origin/main's pin (eaf4890), rendering kick/snare/hat at patch 384 and 258:

build	amp=4	amp=5	peak
pre-#910	+0.00 dB (no-op)	+0.00 dB (no-op)	0.46 → 0.46
current pin	+12.02 dB	+13.81 dB	0.35 → 1.000 (hard clip)

So it isn't only "louder" — at amp=5 the drum bus clips flat against full scale, which is most of why it sounds bad.

Scope: 66 World sketches

Of 524 AMYboard World sketches, 91 use a drum patch (258 or 384–390), and 66 set amp > 1 on the drum synth. Every one was created 2026-07-01 → 2026-07-17 — the window that closes right at amy#910. Nothing generated since does it, and the current refdocs/amy/synth.md drum example has no amp=, so the generator won't reproduce it.

Values: amp=5.0 ×51, 4.0 ×8, 6.0 ×3, 3.5 ×2, 4.5/2.5/1.0 ×1 each.

The two named examples:

• aphex_twin_polygon.py:9 — amy.send(synth=10, num_voices=6, synth_flags=3, amp=5, patch=258) → +13.8 dB
• dx7_groove_128_bpm.py:16 — amy.send(synth=10, patch=384, num_voices=1, synth_flags=3, amp=4) → +12.0 dB

Fix

Strip the amp= kwarg from drum-synth sends in those 66 sketches — don't set amp=1, just drop it. Because the old behavior was exactly inert (+0.00 dB, measured), dropping it reproduces the original mix precisely; there's no re-balancing judgment call involved.

Secondary note

The default (no-amp) kit level also shifted between those pins, as intended by #910/#798: patch 258 and 384 are now the same balanced Gamma kit, with per-drum peaks in a 0.23–0.48 band instead of the old 384 kit's 0.22–0.92 spread. Not a bug, but it does mean drums sit somewhat lower against melodic synths than in the old kit — worth a separate look if the balance still feels off after the amp= strip.