Attack From Mars

Identity and evidence precedence

WPC-95 machine, IPDB 3781. The PinMAME family roots at afm_113 with clones including afm_113b (1.13b, the common home-ROM revision), earlier afm_03/afm_10/afm_11, the Pinball FX and Ultrapin derivatives, and three FreeWPC community firmware revisions. The IPDB identity comes from the pinned OPDB snapshot and ROM crosswalk recorded in reports/opdb-identity.json; no model number is asserted.

Physical inventory authority is the Bally/Midway Operators Handbook 16-10206, retained in this repository's evidence roots at SHA-256 5900c779f3bfb14251ada18d25a4dde84dd608b6ddd988ece8c92ea20fd0114b: lamp matrix and lamp locations on printed pages 2-3, switch matrix and switch locations on 4-5, solenoid/flasher table and locations on 6-7, upper playfield parts on 8-9. Every table was read from rendered pages, not OCR, and transcribed into the retained review artifact.

Two further sources shaped the earlier partial revision of this note but are not retained in this repository's evidence roots and are therefore not authority for any promoted claim: a 176-page Bally/Midway operations manual scan (contributor-held; its page numbering 2-42 through 2-47 is cited in the sections below only to show where the contributor read a value) and a community altsound package used for the DCS opcode table. Where a claim rests only on one of those, it is prose reference material rather than a validated assertion in the definition.

Runtime authority is the pinned LibPinMAME harness recorded in evidence/runtime/wpc-95/attack-from-mars-boot-attract-and-ball-start.json. The earlier contributor switch/solenoid/lamp/sound-command logging on afm_113b under VPinMAME 3.6 retained no content-addressed traces, so claims backed only by that source remain candidate-grade prose here and were not promoted to validated assertions.

Corrections to the legacy import

The legacy VPE-derived definition carried several verifiable errors, all corrected against the manual tables:

• Solenoid 19 is 'Right Side High (2)', the two-bulb right-side flasher circuit (J133-6/J134-5, transistor Q27, #906 bulbs). The legacy 'ROM Started'/c_game_on label is wrong for this machine; the definition's open conflict on item 19 is resolved by the manual and the g5k script, which drives 19 as an ordinary flasher.
• Solenoids 11/12/13 are Left Jet / Bottom Jet / Right Jet (all A-9415-2). The legacy import duplicated 'Right Slingshot' onto 11 and shifted the jets up by one, losing the Right Jet entirely.
• Lamp rows 6-8 (61-88) were largely Medieval Madness names ('Castle Lock', 'Howard Hurtz', 'Sir Psycho', 'Duke of Bourbon', etc.). The manual matrix gives the true AFM names: 61/62/68/76/77/78 are the remaining MARTIAN letter-target lamps, 63-65 Atomic Blaster 1-3, 66/74 Right/Left Loop Jackpot, 67 Extra Ball, 71-73 Capture 1-3, 75 Left Loop Arrow, 81 Shoot Again, 82-85 outlane/return lamps, 86 Launch Button, 87 not used, 88 Start Button.
• Solenoids 25-28 are flasher circuits ('Left Ramp Left (2)', 'Left Ramp Right (2)', 'Left Side High (2)', 'Left Side Low'; general-purpose drives into #906/#89 flashlamps), not coils.
• Output 24 'Motor Bank' is the bank motor (part 14-8023) on a flasher-class drive circuit, not a flasher.
• Switch label fixes: 14 'Plumb Bob Tilt', 16 'Left Outlane', 71/72 'Right Loop High'/'Right Loop Low' (legacy said 'Right Bank').
• Solenoid 43 is PinMAME's mirror of LPDC output 39, not a separate circuit. core_getSol duplicates the WPC-95 LPDC outputs 37-40 at public addresses 41-44, so 43 reports the same physical output as 39 (Strobe Light). That is why 43 is absent from the manual table and why the contributor's runtime logging saw it toggle. The known-working script binds SolModCallback(39) and SolModCallback(43) to two different visual lamps, which is a table convenience rather than two physical circuits. The definition records 43 as a virtual duplicate of 39.

Addresses that duplicate an already-declared circuit

Attack From Mars publishes the same physical outputs on more than one PinMAME address, so a recreation must bind each device once and accept either address:

• Public solenoid 43 mirrors LPDC output 39 (WPC-95 duplicates 37-40 at 41-44). Output 39 is the printed Strobe Light, assembly A-20718 on J110-4, cabled through H-20705 - not a second saucer-dome circuit. The retained JPSalas script settles it by binding the strobe on the mirror address with SolCallback(43) = "SetLamp 130," where lamp 130 is the table's Strobe object, so a recreation must accept the strobe on either 39 or 43. Outputs 37 and 38 are mirrored the same way at 41 and 42.
• afmGameData declares custSol=3, so PinMAME also publishes game-specific solenoids 51-53. afm_getSol returns WPC_FLIPPERCOIL95 bit 0x20 for 51, bit 0x10 for 52, and bits 0xc0 for 53 - the same bits core_getSol reports on public solenoids 34, 33, and 35/36 respectively.
• The two ball-gate addresses carry a naming disagreement, now resolved. Every physical source agrees that public solenoid 33 is the right gate and 34 the left: the printed solenoid/flasher table (33 RIGHT GATE High Power Q84 J120-6, 34 LEFT GATE Low Power Q86 J120-4), the upper-playfield parts list (actuator coil A-17796 with A-17797-2 Right Ball Gate on item 24 and A-17797-1 Left Ball Gate on item 25), and both retained known-working scripts. Pinned PinMAME disagrees and is also internally inconsistent: wpc.c writes WPC_FLIPPERCOIL95 bits 4-7 to public solenoids 33-36, so bit 0x10 is public 33 and bit 0x20 is public 34, yet afm.c reads bit 0x10 for sLGate and bit 0x20 for sRGate while its own comments annotate sRGate as /* 33 */ and sLGate as /* 34 */. The manual owns physical identity and the known-working script owns runtime binding, so both entries are validated on that basis, the emulator defect is recorded in the device notes, and no conflict record is carried because the disagreement is between sources of unequal authority rather than unresolved.
• afmGameData also declares lampCol=2, so PinMAME exposes two auxiliary lamp columns above the standard 8x8 matrix for the 16-LED saucer board. They are now enumerated: the sixteen L.E.D.s appear at public lamp addresses 91-98 and 101-108. afm_wpc_w shifts a sixteen-bit word into tmpLampMatrix[8] and [9] using WPC_SOLENOID1 bit 4 as the clock and bit 5 as the data, and wpc.c documents that register's bits 4-7 as the J122/J123/J124 GPIO that "appear as LPDC outputs 37..40 in manuals", so the clock is public solenoid 37 and the data public solenoid 38 - exactly the printed L.E.D. Clock and L.E.D. Data entries on saucer board A-20670. The pinned harness confirmed it: in attract mode exactly eight of the sixteen addresses are lit at a time and the set rotates through four phases that between them cover all sixteen, while solenoids 37, 38 and their LPDC mirrors 41 and 42 are active; at ball start the ring and all four solenoids go quiet together.

Displays

One dot-matrix display, 128x32, WPC-95 standard. No segment or alphanumeric displays.

Custom mechanisms

• UFO saucer with subway return. Every saucer entry closes switch 78 'Center Trough' - attack-wave hits and the final destruction alike. The ball exits through a subway to the left popper (switch 36) and is ejected up a vertical tube by solenoid 3. The saucer-shake motor (solenoid 15) wobbles the UFO in short pulse bursts on hits; the destruction sequence runs roughly 5-7 pulses about 0.25 s apart, and the ROM raises the 3-bank wall at roughly pulse 5 - before the burst ends - so a recreation ejecting the left popper during this window must mind the moving bank geometry.
• Three-target motor bank. Standups 45-47 ride a motorized wall (output 24) between limit switches 66 (down) and 67 (up).
• Pop-up aliens. Four alien mech assemblies cover the seven MARTIAN standups, one per target group, each an A-20579 (item 15 is printed A-20479-2, see below) with figurine 23-6768 on its own support bracket. Reading the printed switch names letter by letter, MARTIAN is spelt 56 "M", 57 M"A", 58 MA"R", 43 MAR"T", 44 MART"I", 41 MARTI"A", 42 MARTIA"N". Solenoid 5 (Left Alien Low) covers the left bank M-A-R on switches 56-58, solenoid 6 (Left Alien High) covers the T on switch 43, solenoid 8 (Right Alien High) covers the I on switch 44, and solenoid 14 (Right Alien Low) covers the right bank A-N on switches 41-42. The retained table corroborates the grouping geometrically: alien primitives sit at (0.137, 0.535), (0.369, 0.271), (0.622, 0.270) and (0.826, 0.541), each beside its target group, and each group's lamp inserts sit directly below its targets.
• Right popper scoop (Stroke of Luck): capture switch 37, eject solenoid 4.
• Single center drop target: switch 77, reset solenoid 16.
• Loop gates (33 right, high power; 34 left, low power) and the center-ramp diverter on a power/hold pair (35/36), opened when a lock is lit.
• Strobe light: backpanel strobe tube (output 39) plus a 16-LED chase chain clocked serially via outputs 37 'L.E.D. Clock' and 38 'L.E.D. Data'; featured in Strobe Multiball.
• Standard complement: two flippers (outputs 29-32), two slingshots (9/10, switches 51/52), three jet bumpers (11-13, switches 53-55), four-ball trough (switches 31-35, eject solenoid 2, auto plunger solenoid 1), knocker (7).

Ball-state transitions

All observed on afm_113b:

• Serve: trough eject (solenoid 2) to shooter lane (switch 18), then player plunge or auto plunger (solenoid 1).
• Lock / saucer capture: lock shots and saucer balls end at switch 78, transit the subway, and land in the left popper (switch 36). For locks 1 and 2 the ROM keeps the ball in the popper and serves a replacement from the trough; lock 3 releases for multiball. Popper eject is solenoid 3.
• Ball saver: indicated by lamp 15 'Return To Battle (2)' flashing - not lamp 81, which is the extra-ball 'Shoot Again' indicator. The saver re-serve fires within roughly 100 ms of the drain.
• End of ball: with a trivial bonus count the next player's serve can follow the drain in as little as 1.6 s. Recreation logic that distinguishes ball-saver serves from next-player serves purely by timing will misclassify (measured failure at a 2 s threshold).
• Stroke of Luck: switch 37 capture, award, solenoid 4 eject.

Lamp semantics

Behavioral notes beyond the matrix names, observed at runtime:

Availability indicators blink in roughly 200-300 ms phases; solid versus flashing distinguishes owned versus offered throughout the rule set.

General illumination

Five GI strings per the manual solenoid/flasher table: 01 Bottom Playfield, 02 Middle Playfield, 03 Top Playfield (#44/#555 bulbs, dimmable) plus 04 Top Insert and 05 Bottom Insert (#555 bulbs; these two strings do not brighten or dim - they are always on). All five strings are modelled as machine outputs on pinmame.output.gi addresses 0-4. Strings 01-03 carry playfield placements taken from the retained table's aGiLLights, aGiMLights and aGiTLights arrays, deduplicated from 50 light objects to 29 physical sockets because the table pairs every bulb with a co-located halo. The manual publishes no per-string bulb count, so the quantity comes from those arrays and is not presented as a printed figure. Strings 04 and 05 are backbox insert-panel circuits and take a controlled cabinet_or_service spatial record with no asserted count. String 01 is the one string that spans both: the printed table gives it a playfield harness on J105-1/J105-7 with #44 bulbs and a separate backbox harness on J106-1/J106-7 with #555 bulbs, and only the playfield sockets are placed.

Dedicated and flipper switches

The manual's dedicated grounded switches include the four coin chutes (D1-D4), service credits/volume/begin-test controls (D5-D8), and the Fliptronic flipper block: F1 Lower Right Flipper EOS, F2 Lower Right Flipper cabinet button (opto), F3 Lower Left Flipper EOS, F4 Lower Left Flipper cabinet button (opto), F5-F8 not used (no upper flippers on this game). The manual's F-numbers map onto the emulator's dedicated switch addresses in order, F1 through F8 to public 111 through 118, which the definition now records with a manual.address alias on each. F1-F4 are used; F5-F8 print --- and "Not Used" in both the switch matrix and the switch-locations parts list, and the switch-locations playfield map draws F6 and F8 floating outside the playfield outline with no leader line, so public switches 115-118 are marked unused. Note that afmGameData still declares FLIP_SW(FLIP_L|FLIP_U); that is emulator scaffolding for the generation and is not evidence of upper flippers on this machine. The matching upper Fliptronic outputs are genuinely in use - they drive the two loop gates and the ramp diverter.

Timing and tuning observations

• Saucer destruction: solenoid-15 shake burst of ~5-7 pulses at ~0.25 s spacing; the 3-bank rises at roughly pulse 5, before the burst ends.
• The ROM re-asserts the current background-music command about twice per second (relevant for alt-sound implementations).
• Ball-saver re-serve within ~100 ms of drain; next-player serve possible within ~1.6 s of drain.

Controller interactions - DCS sound commands

Commands to the DCS sound board arrive as 16-bit big-endian pairs in the PinMAME sound-command stream. Protocol notes, runtime-verified on afm_113b:

• Parse against the known-id list below; naive byte splitting desynchronizes on the 0x03D2/0x03D3 heartbeat pair.
• The current background-music command is re-asserted roughly twice per second during play.
• Rows marked (loop) are looping music tracks; everything else is one-shot.
• Many voice/SFX commands exist as even/odd duplicate-id pairs; both trigger the same sample.

Selected stable mappings, each observed dozens of times: 0x01A5 ball saved ('Return to battle!'), 0x0128/0x0129 skill-shot award (held-flipper super skill shot and top-lane skill shot both fire these), 0x0346/0x0347 saucer destroyed (wave completion), 0x0122/0x0123 extra ball awarded, 0x000F game-over music.

Opcode reference (586 ids)

Sample names are from the contributor-held community altsound package for afm_113b, which is not retained in this repository's evidence roots, (altsound.afm-113b.community-package, SHA-256 bc139099b58ed94240d86f4fcc8e06ba37dafa5a14d9d5417917b5f225eac830); annotations are contributor runtime observations. Note: the package itself writes 0x006A on two rows (martian_target_mid and martian_target_mid2, both mapping 0x006a-laser_shot.ogg) and has no 0x006B row; the table below lists the second row as 0x006B as a contributor-inferred correction per the even/odd pairing convention, flagged in its annotation.