Add MA-7 SMAF/SEQU decoding support by psi29a · Pull Request #290 · Mindwerks/wildmidi

psi29a · 2026-07-27T09:24:44Z

Add MA-7 SMAF support (score + FM voices)

MA-7 .mmf files previously failed to play. This decodes both halves of the format and fixes the bank lookup that kept them silent.

SEQU score (`format_type 0x03`)

Not a compressed format as previously assumed — it's Mobile Standard with the status byte repacked for 32 channels:

bit 7 = channel bank (0-15 / 16-31), bits 6-4 = event type (MIDI nibble 0x8+n), bits 3-0 = channel low nibble
0xF0/0xFF literal; no running status (status bytes are < 0x80)
Reuses the Mobile decoder via a sequ flag

Explains the old "mystery 0x61 record": pitch bend on ch1 + the next duration byte.

MA-7 FM voices (`43 79 08 7F 21`)

Operators are stored in the chip's native 10-byte layout, not packed 7-byte VM35: MULTI/DT → byte 9, WS/FB → byte 6, bytes 5/7/8 reserved. Undoing the shuffle reuses the existing VM35 decoder.

Bank fallback (`src/patches.c`)

SMAF selects Yamaha voice banks (0x7c) that no GUS/SF2 set defines; _WM_get_patch_data had no fallback → silence
Now falls back to bank 0, as hardware does
Pre-existing bug: any SMAF file without custom FM voices was mute (MA-5 only escaped via MAFM)

Verification

Both decodes checked against Yamaha's MCP-MA7 middleware (Ghidra + Unicorn, clean-room; the DLL is 32-bit x86 so it emulates directly, no Wine).

Voice mapping round-trips through Yamaha's packer: 20/20 operator bytes exact, all 4 corpus voices
Decoded params are textbook 2-op FM (carriers TL=0/MULTI=1); earlier guesses gave MULTI=0 on both ops, which no real patch has
SEQU parse byte-exact on both MA-7 files; pitch classes cleanly diatonic
Corpus re-render: 30/32 bit-identical, only the two MA-7 files change
2 new tests (smaf_sequ, ma7_voice), 3/3 pass

Limitations

MA-7 FM voices: only body form 0x00 is decoded — 10 of the 58 voice
records across the 5 known MA-7 files. p[10] selects the form; 0x01
(plain VM35), 0x02, 0x03 and 0x07 are declined rather than guessed,
and those voices fall back to GM. Form 0x00 is not the common case —
form 0x03 is (36/58). Per file: AB00221GM7/AB03957GM7 get all their
voices, Blossom gets 2 of 25, AccidentCafe7 none, First Love has no
voice records (streaming Mtsp).
MA-7 native operator bytes 5/7/8 hold MA-7-only parameters with no VM35
equivalent; they're ignored, so those voices lose whatever those fields do.
Channels 16-31 fold onto MIDI's 16 — no known file plays on them.
Possible constant octave offset in SEQU notes unresolved; notes pass through
unchanged.
Pre-existing, unchanged: Huffman-compressed Mobile (format_type 0x01) is
not decompressed, and MA-3/MA-5 PCM (sampled) voices are parsed but not
played.
No other open-source implementation reads MA-7 correctly (vavi-sound and
smaf825 both route it to their HandyPhone decoder), so there was nothing to
cross-check against. docs/formats/SmafFileFormat.txt records the
derivation, the corpus sources, and what was ruled out.

From AB00221GM7.MMF and AB03957GM7.MMF
AB00221GM7.mp3
AB03957GM7.mp3

MA-7 stores operators in the chip's native 10-byte layout rather than the packed 7-byte VM35 one: MULTI/DT moves to byte 9, WS/FB to byte 6, and bytes 5/7/8 are reserved. Undoing that shuffle reuses the existing VM35 decoder, so MA-7 files now play on the FM engine instead of falling back to the GM wavetable.

psi29a · 2026-07-27T09:25:00Z

@CodeRabbit summary

coderabbitai · 2026-07-27T09:25:01Z

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 508ed3bd-3f31-4b3b-b835-207140c659ab

📥 Commits

Reviewing files that changed from the base of the PR and between 474b549 and 01d2ef4.

📒 Files selected for processing (1)

docs/formats/SmafFileFormat.txt

🚧 Files skipped from review as they are similar to previous changes (1)

docs/formats/SmafFileFormat.txt

📝 Walkthrough

Walkthrough

Adds MA-7 SEQU score-track decoding, MA-7 FM voice-exclusive decoding, format documentation, regression tests, and bank-0 fallback lookup for missing non-zero-bank patches.

Changes

MA-7 SMAF support

Layer / File(s)	Summary
SEQU format contract `docs/formats/SmafFileFormat.txt`	Documents MA-7 container mappings, required `Mtsq` tracks, format `0x03`, packed channel status bytes, and SEQU event mappings.
SEQU score conversion `src/smaf2mid.c`, `test/CMakeLists.txt`, `test/test_smaf_sequ.c`	Decodes SEQU events, remaps 32 logical channels onto 16 MIDI channels, and verifies generated MIDI events.
MA-7 FM voice decoding `src/mafm/smaf_voice.c`, `test/CMakeLists.txt`, `test/test_ma7_voice.c`, `docs/formats/SmafFileFormat.txt`	Decodes supported MA-7 FM forms by reconstructing VM35-style operator data and tests sampled-form rejection.

Bank fallback lookup

Layer / File(s)	Summary
Bank-0 fallback lookup `src/patches.c`	Retries bank-0 patch lookup when a requested non-zero bank has no matching patch.

Estimated code review effort: 3 (Moderate) | ~25 minutes

Sequence Diagram(s)

sequenceDiagram
  participant MMF
  participant _WM_smaf2midi
  participant decode_mobile
  participant MIDI
  MMF->>_WM_smaf2midi: MA-7 format_type 0x03 track
  _WM_smaf2mid->>decode_mobile: Pass SEQU decoding flag
  decode_mobile->>MIDI: Emit repacked MIDI events

Possibly related PRs

Mindwerks/wildmidi#286: Extends the SMAF-to-MIDI and MA-FM implementations introduced by that change.

Suggested reviewers: sezero

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly matches the main change: adding MA-7 SMAF/SEQU decoding support.
Description check	✅ Passed	The description is directly related to the PR and summarizes the SEQU, FM voice, and bank-fallback changes.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

coderabbitai · 2026-07-27T09:25:08Z

✅ Action performed

Summary regeneration triggered.

p[10] selects the MA-7 voice body layout; only form 0x00 (3 global + 10 bytes/operator) is established, so require it and an exact body length instead of assuming every 7F 21 record uses that shape. Also drop the bogus voice-type reading of p[9], which holds split points.

sezero

Built and played cleanly. My valgrind was silent.

psi29a · 2026-07-27T10:16:53Z

Built and played cleanly. My valgrind was silent.

Awesome, I just found a stash of 1000+ MMF files; I'm now iterating over all of them to make sure they play and nothing blows up (smoke test). Might be some more missing features.

The cool news is that no other FOSS player on the planet supports MA-7 based files. :)

p[10] is effectively the voice type: forms 0x00 and 0x02 are FM and share the 3-global + 10-byte-operator layout (0x02 adds a 16-byte trailer), while 0x01/0x03/0x07 are sampled voices whose bodies start with a PCM sample rate. Decoding 0x02 covers every FM voice in every known MA-7 file. Also document the MA-1..MA-6 gaps in SMAF_TODO.md.

psi29a added 4 commits July 27, 2026 10:04

initial work on ma7

29d5c39

add test

62609ef

update smaf file format

baa24bb

psi29a changed the title ~~Ma7~~ Add MA-7 SMAF/SEQU decoding support Jul 27, 2026

psi29a requested a review from sezero July 27, 2026 09:36

sezero approved these changes Jul 27, 2026

View reviewed changes

psi29a added 2 commits July 27, 2026 12:34

no octave offset note

01d2ef4

psi29a merged commit fbe6482 into master Jul 27, 2026
23 checks passed

sezero deleted the ma7 branch July 27, 2026 11:11

coderabbitai Bot mentioned this pull request Jul 29, 2026

Smaf sampled voices #291

Merged

coderabbitai Bot mentioned this pull request Aug 14, 2026

Fix sf2 volume control, note release, and sparse-bank patch fallback (#295) #296

Merged

Navigation Menu

Uh oh!

Add MA-7 SMAF/SEQU decoding support - #290

Add MA-7 SMAF/SEQU decoding support#290
psi29a merged 7 commits into
masterfrom
ma7

psi29a commented Jul 27, 2026 •

edited

Loading

Uh oh!

psi29a commented Jul 27, 2026

Uh oh!

coderabbitai Bot commented Jul 27, 2026 •

edited

Loading

Walkthrough

Changes

Sequence Diagram(s)

Uh oh!

coderabbitai Bot commented Jul 27, 2026

Uh oh!

sezero left a comment

Uh oh!

psi29a commented Jul 27, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

psi29a commented Jul 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Add MA-7 SMAF support (score + FM voices)

SEQU score (format_type 0x03)

MA-7 FM voices (43 79 08 7F 21)

Bank fallback (src/patches.c)

Verification

Limitations

Uh oh!

psi29a commented Jul 27, 2026

Uh oh!

coderabbitai Bot commented Jul 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

Changes

Sequence Diagram(s)

Uh oh!

coderabbitai Bot commented Jul 27, 2026

Uh oh!

sezero left a comment

Choose a reason for hiding this comment

Uh oh!

psi29a commented Jul 27, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

psi29a commented Jul 27, 2026 •

edited

Loading

SEQU score (`format_type 0x03`)

MA-7 FM voices (`43 79 08 7F 21`)

Bank fallback (`src/patches.c`)

coderabbitai Bot commented Jul 27, 2026 •

edited

Loading