High Speed

Identity and evidence precedence

This is the Williams System 11 physical product released January 1986 - the machine that introduced the multiball jackpot and the police-chase light and siren sequence, and whose ruleset most of the industry copied for the rest of the decade. It covers the five-driver hs_* clone tree: hs_l4 (parent, production L-4 and the driver the retained table binds), hs_l3 (earlier U26 game ROM), hs_l1 (the earliest dumped revision, taken from a Unidesa / Cirsa licensed board and catalogued 1985), hs_p4g (German export, reported G-4, stickered PROTO4) and hs_l4c (a 2018 community Competition MOD patch of the L-4 ROMs). All five share one static hsGameData/init_hs pair through CORE_CLONEDEF, so every driver is physically identical; input_ports_hs aliases the shared input_ports_s11 table.

The 1985 date on hs_l1 is the licensee's build date, not a second physical machine. Pinned PinMAME's own comment says the Unidesa / Cirsa CPU board is "identical to the Williams board (also same P/N)". That cabinet did carry one extra board the Williams machine does not - a Credit Conversion PCB with an 8035, its own 2716 EPROM, a four-position DIP bank and a 6.144 MHz crystal, undumped - but it is coin-door pricing hardware outside the CPU board's address space and changes no playfield device, no switch and no lamp.

Do not conflate this machine with Williams The Getaway: High Speed II (1992), its WPC-Fliptronic sequel, which is a separate curated record in this repository at machines/partial/williams/the-getaway-high-speed-ii-1992.json.

Evidence precedence for this definition: the retained known-working VPX script is runtime and mechanism-causality ground truth; the Williams instruction manual controls physical construction, part numbers, wiring, quantities and device location; pinned PinMAME controls controller generation, public address topology and per-game bulb-type metadata; the retained VPX geometry supplies normalized coordinates. That last one carries an extra qualification on this machine, because the retained table gets two switch bindings geometrically wrong - see the jet bumpers below.

This manual has a genuine Acrobat Paper-Capture OCR text layer that is unusually good for a 1986 scan, but every table used here was still read from 300 dpi rendered pages and transcribed by hand. The reading log, including the verified printed-to-PDF page offset and the sheets this rebound copy has out of order, is external:pinmame-review-artifacts/high-speed-1986/manual-transcription.md.

Controller platform: System 11, reusing the profile but almost none of Whirlwind's per-game reasoning

High Speed is the second System 11 machine curated in this project and reuses controllers/pinmame/system-11.json unchanged, confirmed from hsGameData's own GEN_S11X field (0x100, the same bit as GEN_S11A and GEN_S11B, which gen.h defines as aliases of it). The platform rules that do carry over are the ones the profile exists for: switches and lamps share one sequential column-major 1-64 address space registered through core_swSeq2m rather than WPC's column-times-ten wpc_sw2m; matrix column 1 is ordinary matrix space rather than a WPC-style dedicated namespace; general illumination is an ordinary solenoid address rather than a separate channel; and the flipper coils are not CPU-controlled, so public 45-48 are synthetic.

What does not carry over from Whirlwind is most of the per-game address map, and this is worth stating explicitly because a reader who assumes the two System 11 machines share a layout will get four things wrong. Decoding INITGAMEFULL (src/wpc/s11games.c line 38) against core_tGameData's field order shows hsGameData expands to {GEN_S11X, s11_dispS11, {FLIP_SWNO(37,38),0,0,0,0,S11_BCDDIAG,0,0}, NULL, {{0}}, {0,{49,50,35,34,33,0}}}:

1. sxx.muxSol = 0: there is no A/C select relay. The fourth INITGAMEFULL argument is 0 where Whirlwind's INITGAME passes 12. updsol()'s mux-copy branch is gated on core_gameData->sxx.muxSol being non-zero, so it never runs and bits 24-31 of the solenoid word are never written; MACHINE_INIT(s11)'s hs_ block likewise declares no output type at or above address 25, unlike the many System 11B blocks that set eight muxed flasher outputs there. The manual agrees from the other side: its Solenoid Table has no A/C column, no nnA/nnC circuit pairs, and its only three relay rows are the police light, the GI and the coin lockout. Public addresses 25-32 are enumerated here as unpopulated platform slots, not devices.
2. hw.gameSpecific1 = 0: no sound-overlay board and no mux-feedback switch. S11_SNDOVERLAY is unset, so pia5cb2_w never diverts the sound byte into a solenoid pattern and the platform's 37-44 range has nothing in it. S11_MUXSW2 is unset too, so SWITCH_UPDATE(s11) does not overwrite public switch 2; on High Speed that address is a real device, the Ball Roll Tilt.
3. hw.display = S11_BCDDIAG only. Whirlwind passes S11_LOWALPHA|S11_DISPINV; High Speed passes just the 7-segment-diagnostic-LED flag, which is the CPU board's own numeric status LED - the manual's CPU LED Indicator Codes table (printed page 27) is that LED's code list. The display inventory itself comes from s11_dispS11 rather than s11_dispS11b2 and is six displays, not one: two seven-digit 16-segment alphanumeric player displays (segment starts 1 and 9), two seven-digit 7-segment-plus-comma player displays (21 and 29), and four small single 7-segment digits (0, 8, 20, 28) forming two two-digit displays. The manual names all six in its own Display Test text - "SPEEDER 1, 2, 3, 4, BALL IN PLAY/MATCH, Credits" - and only SPEEDER 1 and 2 are ever described as showing text messages, matching the two alphanumeric layout entries exactly. The Backbox Cables list in the Amendments sheet is the physical counterpart: Players 1 and 2 take cable 5795-10868-14 and Players 3 and 4 take 5795-09453-00. The two-digit displays are recorded with non-contiguous segment positions (0 and 8; 20 and 28) because that is what the layout says: each is built from the spare eighth digit of a player display's own group. Which pair is Credits and which is Ball In Play/Match comes from the retained script's own labelled Digits()/Leds() arrays, since the layout itself does not say.
4. One GI relay, not two. MACHINE_INIT(s11)'s hs_ block names a single CORE_MODOUT_BULB_44_6_3V_AC_REV address, 11, where Whirlwind's block names 11 and 16. The rest of the block declares #89-model bulb types for addresses 4, 5-6, 9-10, 12 and 22.

The special-solenoid switch map: the first game in this project to use sxx.ssSw

High Speed is the first curated machine whose driver populates core_tGameData.sxx.ssSw, the trailing six INITGAMEFULL arguments, here 49,50,35,34,33,0. Whirlwind used the plain INITGAME form and never exercised the field, and controllers/pinmame/system-11.json mentions it only in a source locator, so what it means at runtime had to be derived rather than looked up.

src/wpc/s11.c lines 191-200, inside the if (locals.ssEn) block of the VBLANK handler:

for (ii = 0; ii < 6; ii++) {
  if (core_gameData->sxx.ssSw[ii]) {
    int switchState = core_getSw(core_gameData->sxx.ssSw[ii]);
    if (switchState)
      locals.solenoids |= CORE_SOLBIT(CORE_FIRSTSSSOL+ii);
    core_write_pwm_output(CORE_MODOUT_SOL0 + CORE_FIRSTSSSOL + ii - 1, 1, switchState ? 1 : 0);
  }
}

The loop indexes CORE_FIRSTSSSOL + ii directly with ii, not through setSSSol's scrambled ssSolNo table, so the mapping is plainly sequential: ssSw[0] drives public solenoid 17, ssSw[1] drives 18, and so on to ssSw[5] and 22. A zero entry means "not switch-driven", and the loop skips it. Note the two paths coexist: a special solenoid can be asserted either by this loop (from its own switch) or by setSSSol (from a CPU PIA write), and the loop is skipped entirely under P-ROC or LISY support because there the switch drives the coil physically and the emulated copy would lock on.

This is PinMAME's model of real System 11 switched-solenoid hardware, where the actuating switch completes the coil circuit itself for zero-latency response and the CPU only gates the supply through the switched-solenoid enable. For High Speed the array reads as an exact match to the manual's Solenoid Table:

Six for six on the five that are paired, and the one slot left at zero is the one special solenoid that is a flasher circuit rather than a kicker or bumper - it has no actuating switch and can only be fired by a CPU write. The array therefore does double duty as independent confirmation that Special #1 through #6 are public 17 through 22 in order, which is the same answer setSSSol's ssSolNo[0] = {5,4,1,2,0,3} un-scrambles to and which the Amendments sheet's own transistor list (Q75 for SS1, Q71 SS2, Q73 SS3, Q69 SS4, Q77 SS5, Q79 SS6) confirms a third time. The definition records these five switch-to-coil pairings as direct relationships.

system-11.json did not need changing for this. Its solenoid-group note already describes addresses 17-22 as "six 'special' solenoids, each driven by one PIA CA2/CB2 line and gated by a software enable" and already lists sxx.ssSw among the core.h fields the profile is derived from; the switch-driven path is a per-game population of a field the profile correctly declines to fix a value for, exactly like the per-game GI addresses. The finding is recorded here rather than pushed into the profile because the mapping is game-specific data, not a platform address rule.

The flippers: no Sol. No. anywhere, and a Lane Change switch the matrix page miscalls an EOS

hsGameData's hw.flippers is FLIP_SWNO(37,38), which expands to ((37)<<8)|(38)|FLIP_SW(FLIP_L) - a left-flipper switch of 37, a right of 38, and no FLIP_SOL bit. Whirlwind established what the missing FLIP_SOL means and it holds here: core_updateSw's "fake solenoids if not CPU controlled" branch fabricates public 45-48 from live button state, gated by the flipEn argument the driver passes as locals.ssEn, and no driver-board output exists behind them.

The manual confirms it from the hardware side more explicitly than Whirlwind's did. Its Solenoid Table has three flipper rows - Upper, Right and Left - and all three print — in both the Sol. No. and Driver Trans. columns. The two CPU-board connections that do appear (Orn-Vio at 1P19-1, Orn-Gry at 1P19-2) are on the special-solenoid connector, and note 1 says those two wires "connect from CPU Board to flipper switch" while note 3 says the parenthesised connections are "from flipper switch to flipper coil". So the CPU supplies the switched-solenoid feed, the cabinet button closes the circuit, and the coil fires - which is also why PinMAME gating its synthetic flipper outputs on ssEn is not an approximation but the actual topology.

What the CPU does read is a Lane Change switch on each lower flipper's own base assembly. This took resolving because the manual disagrees with itself. The Switch-Matrix Table prints addresses 37 and 38 as "Left/Right Flipper Engine Revving (EOS)". The Switches parts list prints the same two as "Left/Right Flipper Lane Change (Engine Revs)" with parts SW-1A-150-1 and SW-1A-150, and the C-9952-R Flipper Assemblies parts list identifies SW-1A-150 as item 2b, Lane Change Switch - a different part from that same assembly's item 2a, End of Stroke (EOS) Switch, 03-7811, which carries no matrix address anywhere in the manual. Three things settle it for the parts list: the part number itself, the fact that a real EOS switch exists as a separate part in the coil circuit, and PinMAME's own construction - locals.flipMask only receives an EOS bit when hw.flippers carries a FLIP_EOS bit (which FLIP_SOL implies), so a game with no FLIP_SOL models no EOS switch at all and core_setSw(FLIP_SWL(flip), swFlip & CORE_SWLLFLIPBUTBIT) writes button state into address 37 directly. Prefer the parts list, as this project's runbook already says. Worth carrying forward: the flipper-assembly page's own adjustment note also tells you the EOS contacts are normally closed and open at end of stroke ("The tip of the EOS Switch must travel .015 inch before the contacts fully open with the flipper in the actuated position"), which is why that switch belongs in the coil circuit and not in the matrix.

The retained table binds two jet-bumper switches to the wrong objects

The three jet bumpers are Upper Left (switch 33, solenoid 21), Lower Left (switch 34, solenoid 20) and Right (switch 35, solenoid 19). Both of the manual's numbered locations drawings - the switch drawing on printed page 32 and the solenoid drawing on printed page 34 - lay the triangle out the same way: the Upper Left bumper highest and leftmost, the Lower Left bumper directly below it, and the Right bumper to the right of both and between them in height. The retained table's three Bumper objects sit at raw (443.5, 202.1), (623.7, 274.3) and (472.1, 398.1), so by the drawings the first is Upper Left, the third is Lower Left and the second is Right.

The retained script binds Bumper2_Hit (the rightmost object) to PulseSw 34 and Bumper3_Hit (the lowest object) to PulseSw 35, i.e. it has 34 and 35 swapped. Two independent manual drawings against one community table is enough to resolve this outright, the same way Bally Centaur's reversed auxiliary lamps and Williams Indiana Jones's transposed slingshot row were resolved; the coordinates in this definition follow the manual and all three bumper switches disclose the disagreement. This is exactly the class of error the cheap geometric ordering test catches - tests/test_high_speed.py asserts that switch 35 sits right of switch 34 and that switch 34 sits nearer the player than switch 33 - and it is why the retained table is treated as a geometry source only where the manual's own drawings agree with it.

Ball paths and the two hideouts

A ball draining lands in the outhole (switch 9, part 17-1067) and solenoid 1 kicks it into the trough tube, where up to three balls queue on switches 10 (Left), 11 (Centre) and 12 (Right). Solenoid 2 (Ball Release) ejects the lead ball into the shooter lane. The manual's switch drawing runs the tube from callout 9 at the left up to callout 12 nearest the release, and the retained script's bsTrough.InitSw 9,12,11,10 independently declares the same order, so 12 is the position closest to release. None of the four positions has its own object in the table; all four are documented projections.

The machine's ball-locking feature is the pair of hideouts, one per side, each a two-switch lane built from an upper and a lower ball chute (D-10863/D-11050 left, D-10862/D-11049 right). A ball entering passes the Upper switch (39 left, 47 right, assembly A-11047) and comes to rest on the Lower switch (40 left, 48 right, part 17-1085); solenoids 7 and 8 kick it back out. Locking a ball in each hideout is what stages multiball, and lamps 39 and 40 ("Ramp Earns Hideout", "Ramp Earns Hideout Jackpot") are the inserts that advertise it.

Both manual pages print those two coils as "Left/Right Hideout Relay", and the manufacturer's own Amendments sheet corrects the wording: "change the description, LEFT HIDEOUT RELAY, to LEFT HIDEOUT COIL. (The same change applies to the Right Hideout Relay designator; it should be 'coil'.)" The part AE-24-900-02 is a coil in any case. There is a relay in each circuit, and the same amendment replaces it with a snubbed 2-pole 24 VDC assembly (Relay Snubber Assembly B-11160) in both hideout circuits and in the outlane kickback circuit. The definition uses the corrected name and cites the amendment on both devices.

The retained script models only each hideout's resting position (bsLeftLock.InitSw 0,0,40, bsRightLock.InitSw 0,0,48) and never drives switch 39 or 47, so those two addresses have no bound object. Their coordinates come from two unbound Trigger objects the table carries with no _Hit handler anywhere: sw39 at raw (50.3, 769.0) on the left and sw37 at (889.3, 765.3) on the right, each sitting directly above the kicker the script does bind to the lane's lower switch. The right-hand object's name is a misnomer - switch 37 is the left flipper's Lane Change switch at the bottom of the playfield - so the assignment is made from the mirror geometry of the two lanes plus the manual's own upper-above-lower ordering rather than from the object name.

No drop targets, and no optos

The task brief that seeded this curation described the machine's headline hardware as including a drop-target bank. It does not have one, and this is worth recording as a correction rather than reproducing. The three red/yellow/green groups the switch matrix calls "Stoplight Bank" targets are standup targets: each is its own assembly by colour (A-11022 red, A-11054 yellow, A-11055 green), no solenoid in the Solenoid Table resets a target bank, neither locations drawing shows a drop unit, and the retained table models all nine as HitTarget objects. The three banks plus the six A-8253 arrow standups make fifteen standup targets and zero drop targets. Completing the stoplights is what lights Escape at lamp 41; the eject hole at switch 16 is where the escape is collected, which is why lamps 11 and 12 in that hole read "Extra Ball" and "Escape".

Nor does the machine have any opto. Sweeping every row of the Switches parts list: all 52 used positions carry a part number, none is blank, none carries an opto or photo-transistor part, and none carries a second construction line. Neither copy of the Switch-Matrix Table has an opto legend or a single shaded cell - a real difference from the late-Williams WPC manuals this project has mostly been reading, and from Bally Scared Stiff's, which prints an opto legend it never uses. Pinned PinMAME agrees structurally: hsGameData's whole wpc struct is written by the positional initializer {{0}}, so wpc.invSw is entirely zero and core.c copies those zeros into the live inversion mask, normalizing nothing. grep -n invSw src/wpc/s11games.c returns no assignment anywhere in the file. Polarity is therefore not a gap on this machine: there is no printed normally-closed or opto matrix switch to have an unconfirmed rest state, and no emulator normalization to reapply. That is a different situation from Whirlwind, where six confirmed optos had no mask to compare against, and it is the reason polarity is absent from coverage.missing here. The one normally-closed contact the manual does document, the flipper EOS switch, has no matrix address at all.

Where the manual gives a part number but no construction - the 17-* parts at switches 9, 16, 40 and 48, and the A-11047 hideout assemblies at 39 and 47 - no switch_type is asserted rather than guessed.

The police beacon is backbox hardware

The machine's signature device is a genuine rotating beacon: Police Light Assembly C-10933 is a 100 rpm 24 VAC motor (14-7939) turning a reflector assembly (B-10917) in front of a single #1683 28 V bulb (24-8771) behind a red lens, held on with a lens clip, washer and wing nut. Solenoid 4 energises the relay (5580-10883-00) that powers it, so the CPU controls only on and off - not rotation speed or phase, which a recreation must therefore drive itself.

It sits in the backbox, not on the playfield and not on the cabinet: the Solenoid Table's Playfield/Cabinet column reads "Backbox" for this address, the Solenoids/Flashers parts list repeats "(Backbox)" in the item description, and neither numbered locations drawing has a callout for it. It takes a controlled not_applicable / cabinet_or_service spatial record. Pinned PinMAME's own comment for the address says the same thing with the bulb number transposed - "In fact, this is a relay controlling police light which is a #1628 28V bulb" against the parts list's #1683 - which is an emulator comment typo, not a disagreement worth recording as a conflict.

The retained table has no SolCallback for address 4 at all. It animates its police-beacon flashers off lamp 41's own light state instead (if L41.state = 1 Then ... PoPoLiTimer.enabled = 1). That is a table shortcut, not evidence about the machine, and a recreation should drive the beacon from solenoid 4.

General illumination: one relay, three destinations

Solenoid 11 is the game's only GI address. The Solenoids/Flashers list describes it as a "Pwr Sup Bd Relay"; the Solenoid Table gives it part 5580-09555-00 and connection 3P7-1 on the Power Supply Board D-8345-541.

Two sources describe its scope more narrowly than it is, and neither is authoritative for physical wiring. Pinned PinMAME's hs_ block comments the address "Backbox GI output". The retained script's handler is named PFGI and drives a 68-member playfield-only GI light collection, inverted - energising the relay turns that collection off. The manual settles it: the Power Wiring Diagram (printed page 47) draws the relay's two contact sets in series with the supply legs feeding 3J6/3P8, whose eight fused lines fan out to the backbox Insert Board (9J2 pins 1-4), the cabinet (7J6-2/7J6-3 through 7J4-1/7J4-2) and the playfield (8J4 pins 3-6). Each lamp string is drawn across one pin from the 1-4 group and one from the 6-9 group, so every string has at least one switched leg. One relay switches playfield, cabinet and backbox general illumination together. The supply is nominally 6.3 VAC, amended to 5.9 VAC measured at the board input.

What no page of the manual gives is a GI bulb count, bulb type by number, or any bulb position - there is no GI inventory and no GI locations drawing. So the device is recorded with its correct scope and no spatial record. The retained table's 68-object collection is deliberately not adopted as a placement set: it cannot be checked against the manual, it includes three jet-bumper cap lights that belong to those assemblies (each B-9414 Jet Bumper Assembly carries its own #44 bulb, part 24-6549, and no lamp-matrix address names a jet bumper, so those caps are GI), it duplicates one member, and one of its lights sits above the top playfield edge.

Flashers: how many bulbs each circuit has, read off the drawing

Five solenoid addresses drive #63 flashlamps. The manual gives no bulb count in any column, so the counts here come from its own numbered solenoid-locations drawing, cross-checked against the retained table:

• Solenoid 9 (Left Red) and solenoid 12 (Right Red) each have two leader lines on the drawing, to two positions on their own side of the playfield, one above the other. The retained table independently models two Flupper flasher assemblies for each, whose dome base primitives are positioned and land where the drawing's leaders point. Two placements each.
• Solenoid 5 (Left Blue) and solenoid 6 (Right Blue) each have a single leader to a single elongated rounded lens, 5 to the left of the freeway insert fan and 6 to its right at the same height. The retained table models each lens with three lights - a co-located bulb-and-glow pair at the centre plus two more offset either side by about (43.5, -28.5), i.e. the centre light sits close to the arithmetic midpoint of the other two. One drawn lens with one leader means one device location, so one placement each at the lens centre; the end lights are excluded as glow companions of the same lens. How many #63 bulbs sit under each lens is not stated anywhere, so no quantity is asserted.
• Solenoid 22 (Top Playfield Flashers) is special solenoid #6, the one with no actuating switch. Its two Flupper dome primitives are parked off-table in the retained build, so only their Light members carry usable coordinates; both sit along the top edge of the playfield, where the drawing's item-22 callout points. Two placements.
• Solenoid 10 (Insert Board Flashers) is a backbox device. Its only connection is 9P1-7, and connector prefix 9 is the backbox Insert Board per the manual's own circuit-board number list, whose introduction says all High Speed circuit boards are in the backbox and are reached by unlatching that insert board. Controlled not_applicable record; the retained script has no callback for it.

PinMAME models all of these as #89 25 V bulbs where the manual specifies #63, which is emulator brightness metadata rather than a device fact.

The ramp, its gate, and the traffic light

The D-10905 HIGH SPEED Ramp Assembly is the main scoring shot and carries two SW-1A-160 rollover switches, numbered 42 (Left Ramp) and 43 (Right Ramp), both across the top of the playfield with 42 left of 43. Lamps 36-40 are the "Ramp Earns" ladder - Bonus X, Ramp Bonus, Getaway, Hideout, Hideout Jackpot - and lamps 42-44 the Ramp Stoplight trio showing which award is live.

Solenoid 13 drives the ramp gate. The D-10884 Ramp Gate Assembly parts list contains exactly one Coil Assembly (AL-23-800-01), one Drive Arm Assembly, one Drive Link, one Gate Mech. Subassembly and one Gate (C-10888), and the assembly appears once as item 13 of the Playfield Parts list - so the Solenoid Table's plural "Ramp Gates" is not matched by a second gate part. Critically the parts list contains no switch, opto or sensor part of any kind, and no matrix address names the gate, so the mechanism has no position feedback at all: a recreation must track its state from the drive signal. The retained script's Divert handler drops and raises two gate objects together and moves two companion primitives in z, splitting the one physical gate in two; the left object is used as the documented projection anchor.

The traffic light itself (B-10921, item 9 of the Playfield Parts list) is a three-bulb playfield fixture built from a light housing plate, a housing subassembly, a PCB standoff and one C-10915 Light Socket & Cable Assembly covering all three sockets. Its bulbs are lamp addresses 42, 43 and 44. It has no coil and no switch.

The three freeway spin targets

Switches 44, 45 and 46 are three rotating decal-screened spin targets - SW-1A-118 switches on shafts with 12-6620 switch actuator wires - which the Playfield Parts list names as Spin Target Assemblies B-11019-2 "Bayshore Freeway", B-11019-1 "Santa Monica Freeway" and B-11019-3 "San Diego Freeway" (items 8, 20 and 21). Which decal is fitted at which of the left, centre and right positions is not recoverable: the item numbers key to a playfield illustration whose callouts are not legible at this scan quality, and neither the Switches parts list nor the Lamp list uses the freeway names. Lamps 7, 10 and 8 are the matching 1000-point arrow inserts and lamps 31-35 the Freeway score ladder (25,000 / 50,000 / 75,000 / 100,000 / Lights Extra Ball). The definition uses the manual's own left/centre/right switch labels and does not attach a decal name to any of them.

Lamps: the two-lamps-in-circuit legend, and where it leaves a gap

The Lamp-Matrix Table carries a boxed [2] Two lamps in circuit legend. Swept across all 64 cells of the rendered page, it appears on four addresses and no others: 1, 3, 9 and 40. The Lamps list then makes three of them concrete:

• 1 (Game Over) and 2, 6 are backglass-only per their (Backglass) locations, and the retained script drives them with FadeReel against Reel objects rather than Lights. Controlled not_applicable records; lamp 1 carries quantity 2 with no placement, which is correct for a two-bulb backglass indicator.
• 3 is split across the two: "Shoot Again (Backglass) / Drive Again (Playfield)". Quantity 2, one placement, and the script confirms both halves - FadeReel 3, L3a for the backglass bulb and NFadeLm 3, l3 for the playfield insert.
• 9 has both bulbs on the playfield, one in each flipper return lane ("20,000 Light Kickback (Left & Right Flipper Lanes)"), and the script drives two Light objects. Quantity 2, two placements.
• 40 (Ramp Earns Hideout Jackpot) is the gap. The manual states two bulbs and never says where the second one is - neither the matrix nor the Lamps list breaks the location down, and there is no third lamp page - and the retained table models one Light object. Quantity 2, one placement.

Spatial gaps that keep this record partial

Five addresses. Three are lamps with no coordinate at all: 42, 43 and 44, the Ramp Stoplight trio. The retained table drives them against Light objects parked at raw x about -2237, far off-table, and the commented-out alternative in the same routine addressed Primitive.stoplight_prim, which sits at position (0,0,0) with an all-zero rot_and_tra and its geometry baked into mesh hslight.ob; deriving that mesh's centroid through the table's own position-plus-size convention - calibrated against Primitive.Apron and Primitive.batleft, whose positions are known - puts it at raw y about -499, also off-table. Two independent object families agree the coordinate is simply absent, so the spatial key is omitted rather than invented, the pattern Star Trek: The Next Generation's unresolved lamps 53/85/86 and White Water's lamps 17/55 established. The fourth is lamp 40, above. The fifth is solenoid 11, the GI relay, whose emitter set cannot be verified against any manual page.

Resolving them needs a second independent recreation of the machine, a photograph of an unrestored playfield showing the traffic light's position and the second Hideout Jackpot bulb, or a GI bulb inventory from a Williams parts catalogue.

Unresolved: which button fires the upper-right flipper (conflict.upper-flipper-driving-button)

The machine has a third flipper. The Solenoid Table and the Solenoids/Flashers list both name an "Upper Flipper" FL 23/600-30/2600-50VDC coil, and both give it no Sol. No., no driver transistor and no CPU-board connection - only its own playfield connector pins, Blk-Yel to 7J1-19 and 8P3-33, which sit between the left flipper's 8P3-32 and the right flipper's 8P3-34. Neither page names its side. Pinned PinMAME models no upper flipper at all here: hw.flippers carries no FLIP_SW or FLIP_SOL bit for either upper position, so public 33-36 read as always zero and no third synthetic output exists.

The side is well corroborated - the manual's own switch-locations drawing places a third flipper bat on the right side of the playfield above the right hideout lane, exactly where the retained table's RightFlipper1 sits at normalized (0.863, 0.369). What is not corroborated is which button feeds the coil. The only source that asserts anything is the retained script, whose SolRFlipper handler rotates RightFlipper and RightFlipper1 together, i.e. it assumes the upper-right flipper shares the lower-right flipper's button and coil circuit. That is very likely right - the Switches parts list has no third flipper-button row anywhere, so the coil must be paralleled onto one of the two existing button feeds, and an upper-right flipper on the right button is the near-universal arrangement - but it rests on one community source. Recorded as unresolved rather than promoted; the Section 3 flipper-circuit schematic sheet, not transcribed here, or a photograph of an unrestored 8P3 harness would settle which feed 8P3-33 joins.

Unresolved: the coin-lockout relay's part number (conflict.coin-lockout-relay-part-number)

The manual gives two different Coinco part numbers for the same device. The Solenoid Table prints 404603-22 with note 2, "Solenoid 16 has a Coinco part number"; the Solenoids/Flashers parts list prints 904218-696 with "Coin-Lockout Relay (Coinco p/n)". Both pages agree on the device, its address, its wiring (Brn-Gry, 1P12-9, 7P1-7 and 7P2-4, driver Q6) and that the part is Coinco's; nothing in the manual or the Amendments sheet reconciles them. The definition records the wiring table's number and the conflict preserves the other. Authoring impact is nil - it is a coin-door relay with no playfield presence - but a part number two pages of one manual disagree about should not be presented as settled.

Smaller manual anomalies, resolved

• The Solenoids/Flashers list prints "Upper Left Jett Bumper" at item 21; items 19 and 20 and all three jet-bumper rows of the Switches parts list read "Jet".
• The Lamps list prints "Right Spinner 1000 Arow" at lamp 8; its partners at 7 and 10 read "Arrow".
• The Amendments sheet corrects the SWITCH TESTS jumper example on printed page 27: the test jumper for a switch 09 indication goes between 1J10-9 and 1J8-2, not 1J10-1, "(A jumper between 1J10-1 and 1J8-2 actually gives an indication of switch 16.)" Both figures agree with the matrix as transcribed - 1J8-2 is column 2, 1J10-9 is row 1 (switch 9) and 1J10-1 is row 8 (switch 16) - which is a useful independent check on the column-major address formula.
• The Solenoid Table's own SOLENOID TEST text cycles steps 01 through 22 with no gaps and shows each solenoid's name on the displays as it fires it. That is the manual's own proof that its Sol. No. column is a contiguous public numbering rather than a Bally-style self-test order, which is the trap Bally Kiss and Centaur both sprang.