docs(plans): decide nacelle tilt architecture; close hover ground strike - #194
Merged
Stab-Rabbit-coding merged 10 commits intoAug 29, 2026
Merged
Conversation
The four 10 AWG ESC feeds have no viable path under the Rev R2 architecture.
They do not fit the two O7 mm wing conduits (two O5.5 mm wires side by side
need >= 11.0 mm in one round bore), and those conduits sit 17.65 mm forward of
the tilt axis, sweeping a 44.7 mm arc every transition. Fixing this forces a
larger fixed hollow spar, a thicker airfoil to carry it, a pivot that no longer
crosses the thrust duct, and a tilt drive independent of spar rotation.
Implements the owner-directed architecture from the 2026-08-27 wiring plan,
with three numeric corrections to the external source conversation, measured
against the current SCAD via tools/wing_spar_station_fit.py:
- The bundle circumscribes 13.28 mm (exact 4-circle packing, R/r = 1+sqrt2).
The source's SPAR_BORE_D = 11.0 does not fit it at all, and 16 mm OD fits
only as 16x14 (1 mm wall, 0.72 mm total clearance). Free twist needs
~16.3 mm bore, so the spar is sized 20 x 16.3, not 16 mm.
- A 20 mm spar breaks out both skins at both stations at 45.15 mm and needs
the station moved to ~25 mm (root t_scale 1.453, tip 2.098). Holding
45.15 mm would cost a 40% t/c tip. This re-opens the Rev S1b decision.
- AS5600 stays rejected: the spar bore is occupied by the bundle, so no
on-axis shaft end exists and AK7455 off-axis remains correct.
Key geometric finding enabling the trunnion: a ring centred on the tilt axis at
the nacelle inboard face cannot intersect the duct at any diameter, since every
point is at distance >= X from the duct axis. Ring plane X = 28 mm gives 3 mm
duct margin in a ~51 mm envelope, removing the through-duct spar entirely and
restoring the canonical 11-fin stator.
Notable: the 20 mm CF spar is LIGHTER than the 8 mm steel it replaces
(67.5 vs 96.2 g/pair, -0.063 lbm), and belt sizing is bound by travel, not
torque (47T/25T at 270 deg gives 4.47 N.m against a 0.177 N.m requirement).
Supersedes the 2026-08-29-001 nav-wire/AK7455 plan, which assumed the rotating
spar would survive; its AK7455 pocket and nav-separation findings carry into
U5, but its "route the nav wire into the spar bore" fix is void under a fixed
spar. Adds WBS SS1.1.4 with SPAR-20-1..9 plus four standing flags, including
that the re-lofted section is no longer S1223 and every aero figure citing this
wing becomes requires-verification.
Source conversation committed as docs/Tilt-Rotor 10AWG Wire Clearance
Analysis.md so the plan's citations resolve in-repo.
Analysis and plan drafted by Claude (Claude Sonnet 5) under the author's
direction per AGENTS.md AI attribution; external source (Google Gemini) cited
and corrected where its figures did not survive checking.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…easured
Three open questions closed with measurements rather than judgement calls.
OQ1 -> station 22.0 mm. Restores the Rev R1a chord line exactly, so the wing
has been built on it before. Also effectively Pareto-optimal for a O20.4 bore:
tip t_scale 2.039 is within 1% of the achievable minimum (2.019 at 20 mm) and
root 1.464 within 0.8% of its own minimum (1.453 at 25 mm), so nothing
meaningful is bought by landing between them.
OQ5 -> the nacelle does not move at all. The canonical baked pivot is hull
Y +47.5 (port_tilt_spar_assembly.scad L108) and the nacelle is already slid
-9.35 mm forward of it. Keeping pivot=CG and sliding the nacelle to reach the
new station would put it -32.5 mm off canonical, a 3.5x worsening and ~18% of
the nacelle's own length. Instead PIVOT_Z moves 111.5 -> 79.0 inside the
nacelle, buying back full canonical placement for a 0.125 N.m gravity moment —
2.8% of the 4.47 N.m the belt delivers, and 0.71x the grounded requirement that
sized the old servo. The pivot-at-CG rule was written when the servo was 98%
consumed by its requirement; the belt's 1.86x multiplication retires that
constraint.
OQ2 -> ring envelope measured from the canonical shell STL rather than the
bounding-box ellipse: 52.0 mm max OD at Z 79 / X 28 (X 26 gives 54.9 mm at 1 mm
duct margin). The estimate was 1 mm conservative; the X=28 working point holds
with 3 mm duct margin, and moving the pivot forward costs only 1.4 mm of
envelope.
Consequences:
- RISK-4 closed (envelope measured, not estimated).
- RISK-5 dissolved: the CG feedback loop cannot occur once PIVOT_Z is no
longer defined by the CG. U9 now reports a gravity moment instead of
relocating the pivot, so U4's ring plane is stable.
- U4 gains a new check: Z 79 is inside EDF1's axial span (27.5-90), so the
ring's load-spreading web must clear the EDF1 casing and ESC1 seat. The
ring itself is outside r=25 and cannot foul the duct.
- The SCAD header's definition of PIVOT_Z as "the CG station" is retired and
must be rewritten, not left to mislead the next reader.
WBS SS1.1.4 synced to the resolved values.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…d in default assembly
Owner direction: the nacelles must rotate about their CG so the centre of thrust
does not depend on tilt. Verifying that against hover ground clearance killed the
off-CG pivot twice over and turned up a pre-existing bug.
KTD7 WITHDRAWN. The earlier draft moved PIVOT_Z 111.5 -> 79.0 to keep the nacelle
canonical. Two independent constraints rule it out:
1. Thrust centre / CG shift. With the pivot off the CG by d, the nacelle CG
swings on radius d through transition. At d = 32.5 mm the two nacelles
(786.8 g of a 2768 g AUW, 28.4%) move the AIRCRAFT CG ~9.2 mm longitudinally
and ~9.2 mm vertically between cruise and hover — trim that has to be flown
out every transition.
2. Ground clearance. Moving the pivot forward lengthens the pivot-to-nozzle arm,
and the nozzle is what swings down in hover. At PIVOT_Z 79.0 the tip reaches
hull Z -73.9, leaving +6.1 mm over even the extended gear.
The 0.125 N.m gravity moment that trade was buying is real but does not outweigh
either. PIVOT_Z stays 111.5 and the SCAD header's "pivot = CG station" definition
stands.
LG-HOVER-01 (new, pre-existing defect, not caused by this work). The rotating
assembly reaches nacelle-local Z 221.3 — the iris seats at NOZZLE_RING_Z 166.25
and its own STL runs 55.1 beyond, so the nozzle stack is 36.1 mm proud of the
185.2 mm shell. In hover that tip swings to hull Z -41.39. serenity_assembly.py
L505-518 wires lg_r6_1_5in_hull_legs.stl as "the ACTIVE (compact, default)
variant", ground plane -38.1. That is a 3.3 mm strike on every vertical takeoff
and landing. The 3.0 in gear clears by +38.6 mm but is "kept for rough-field
missions but not wired into this default assembly". Either the 3.0 in gear
becomes mandatory or the nozzle stack shortens. Worst case is exactly 90 deg; by
140 deg the tip has risen to 0.643x the arm.
OQ5 reopened as the plan's main unresolved trade: pivot-at-CG plus station 22
puts the nacelle -32.5 mm off canonical (vs -9.35 today). Either accept the
silhouette change, or decouple the tilt axis from the spar axis — spar at station
22 for the airfoil, trunnion at the canonical station on a tip-rib bracket, with
the wire bundle jogging inside the FIXED garage where nothing flexes. That costs
~0.5 N.m about the spar axis at limit, reacted by the tip rib and pinch clamp.
Also: OQ2 re-measured at the retained pivot station (53.4 mm ring OD at
Z 111.5 / X 28, vs 52.0 at the withdrawn station), and the ellipse-approximation
table in the problem frame replaced with measured shell values.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
… shaft+gear)
Plan 003, child of 002. Builds the mechanism hanging off the fixed spar: how the
nacelle pivots on it, how tilt is driven from a fuselage servo, and how the
nozzle drive survives losing its rotating-spar datum.
The drive trade has a kinematic answer, not a preference one. The tilt axis is
the spar axis, running spanwise along hull X (port_tilt_spar_assembly.scad draws
it with x_cyl from SPAR_X0 -96 to SPAR_X1 88). A belt spanning fuselage to
wingtip must have its pulleys separated along X, so its pulley axes are
PERPENDICULAR to X — but the driven element must rotate ABOUT X. A spanwise
toothed belt therefore cannot reach this pivot without adding a bevel or worm at
the tip; it does not remove gears, it adds a belt in front of them. Timing belts
cannot absorb the twist themselves. A shaft parallel to X has no mismatch and
meshes the ring with a plain spur pair.
Options evaluated:
A spanwise O4 shaft + spur pair SELECTED - no right angle, no airfoil cost,
0.78 deg wind-up over 190 mm at load
B spanwise belt + bevel at tip rejected - strictly dominated by A, since
A's single spur pair replaces both the belt
and the bevel, and the tip is already the
most congested region on the airframe
C concentric torque sleeve rejected on airfoil - kinematically the
cleanest, but a realistic 26 mm sleeve
pushes tip t/c 24.8 -> 31.3% on a section
that is already the parent plan's top risk
D shaft + belt final stage fallback if spur backlash bites; the one
topology where a belt is legal here, both
pulley axes being X-parallel at the tip
Second finding: the drive is bound by travel, not torque. 145 deg from a 180 deg
servo needs a 1.24x step-UP, which divides torque rather than multiplying it, so
the shaft carries 0.143 N.m against the DS3225's 2.402 N.m rating — ~17x
oversized. Filed as a servo down-select, not left as silent margin.
Nozzle drive improves rather than degrades: a fixed spar is a true ground
reference, so the sync gear bolts to the non-rotating trunnion coaxial with the
tilt ring, and the orbiting pinion reads exactly the tilt angle.
Adds WBS SS1.1.5 (SPAR-25-1..6) and flags the blocking cross-item: SS1.1.4's
reopened OQ5 changes the trunnion geometry, so it must be settled before U2.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…irements)
Requirements-only plan grouping three deviations from the canonical shell that
turned out to share one subject and one authority. Written after investigating
whether the 36 mm nozzle overhang was an unintended drift that could be tuned
away. It is not, and that negative result is the plan's central finding.
The tandem stack fills the canonical nacelle: intake bell 27.5, EDF1 62.5,
stator 32.5, EDF2 56.3 — leaving 6.4 mm aft of EDF2 for an iris that measures
58.1 mm. A variable-area nozzle behind a tandem EDF stack cannot fit inside a
canonical-length Serenity nacelle. The overhang is structural to the propulsion
architecture, so this plan reduces it and documents the residual rather than
pretending it can be removed.
What IS owed work, and is now measured:
- Nozzle housing radial proudness. HOUSING_OUTER_R 35.6 against a canonical
shell max radius of 32.4 at the pocket start (3.2 mm proud), growing to
3.9 mm at the shell's aft end where the canonical profile narrows to 29.6.
The SCAD says this outright — "Fully ovalising the housing to the cowl mould
line ... deferred VERIFY". The WBS [x] for "Housing aft taper (Stage 2)"
covers only the cylindrical taper 35.6 -> 33.5, not the ovalising, so the
checkbox and the source disagreed; recorded.
- Intake fairing proudness. The additive blend rises monotonically while the
canonical dome does not, so they cross: +5.3 mm proud at Z 0, +0.4 at Z 5,
buried from Z 10 aft. That single crossing is the wavy flange already logged
as U5 in the 2026-08-26 plan.
Selected fix (owner-directed): flaps 40 -> 30 mm, chosen over 20 mm which would
give +13.7 mm but double the swing arc back to 3.58-25.94 deg, undoing the
Rev T2 change made to ease the linkage. Stator compression offered as a further
~7 mm lever and not selected; recorded as deferred with its aero justification
noted as the reason.
BLOCKING open question carried on the artifact: flaps at 30 mm leave +3.7 mm on
the 1.5 in gear. Positive, but not a margin. The plan cannot close without
either an accepted rationale for 3.7 mm or a longer gear (2.0 in -> +6.4,
2.25 in -> +12.7). Flagged rather than quietly treated as sufficient.
Adds WBS NAC-MOULD-01 beside LG-HOVER-01.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…C relocation
Answers three owner questions with measurements.
OQ4 (intake curve) RESOLVED, owner-directed: build a clean new curve following
the most convex line of the canonical profile at Z 0, blending tangentially into
the mould line before the dome departs from monotonic — rather than conforming
point-by-point to possibly-noisy voxel-repair geometry. Measuring that pins the
endpoint: the canonical dome's max radius rises monotonically only to Z ~30
(22.2 -> 28.0 -> 31.5 -> 34.3 -> 36.3 -> 38.1 -> 38.4) and falls after. The
existing fairing peaks at Z 60 with INTAKE_BLEND_L = 90 — past the dome's own
maximum, which is precisely why the curves cross. R5 retightened to require
termination by Z ~30, tangent.
Spar station: it is a THREE-way trade, not the two-way one plan 002 recorded.
The bore rides the camber midline, and that midline is lower forward, so the
station also moves hover clearance. Moving it AFT improves canonical offset AND
clearance, costing only thickness:
stn 22.00 tipT 2.039 (24.8%) canon -32.5 clr +2.6
stn 30.00 tipT 2.268 (27.5%) canon -24.5 clr +5.5
stn 45.15 tipT 3.317 (40.3%) canon -9.3 clr +5.6
Holding the spar at today's height instead of riding the midline is worse than
either option: at station 22 the bore sits 3.74 mm above the tip camber midline,
making the upper skin binding and pushing tip t_scale 2.039 -> 2.711 (+33%,
t/c 33%). Rejected.
Ballast question: yes, and there is a free version. PIVOT_Z and the station push
canonical offset in OPPOSITE directions (offset = station + 57 - PIVOT_Z), so
moving both aft together buys clearance at constant offset. Relocating ESC1 aft
alongside ESC2 shifts the CG +6.0 mm at ZERO added mass, and the pivot follows
the CG:
baseline stn 22.0 PIVOT 110.1 26.6%->24.8% canon -31.1 clr +2.6
+ ESC1 relocated (free) stn 28.0 PIVOT 116.1 26.6% canon -31.1 clr +9.8
+ ESC1 + 12mm ballast stn 34.0 PIVOT 122.1 29.9% canon -31.1 clr +16.6
So +7.2 mm of clearance for no mass and 1.8 points of tip t/c, canonical offset
unchanged. Ballast beyond that costs 17.7 g per nacelle per 4 mm of CG shift
(T/W 1.59 -> 1.55 at 12 mm); take the free move first.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
… 116.1, +9.8 mm accepted Owner accepted ~1 cm of hover ground clearance, which settles the last blocking question and freezes the geometry across all three nacelle plans. FROZEN: SPAR_BORE_STATION 28.0 (was 22.0) THICKNESS_SCALE 1.456 (root t/c 17.7%) THICKNESS_SCALE_TIP 2.190 (tip t/c 26.6%) SPAR_Z 66.85 (derived; do NOT carry 68.42 forward) PIVOT_Z 116.1 (= CG after ESC1 relocates aft, zero added mass) canonical offset -31.1 mm hover clearance +9.8 mm on the 1.5 in gear, owner-accepted Station 22.0 minimised airfoil thickness and was picked before the station's effect on hover clearance was measured. The station is a three-way trade — the bore rides the camber midline, which is lower forward — so moving aft improves canonical offset AND clearance while costing only thickness. 28.0 buys 7.2 mm of clearance for 1.8 points of tip t/c, paired with the pivot moving aft by the same amount so the canonical offset is unchanged. LG-HOVER-01 CLOSED. No landing-gear change is forced; the 3.0 in variant stays a rough-field option rather than becoming mandatory. R3 in plan 004 now states the accepted margin explicitly so downstream geometry changes cannot silently trade it away. Knock-on caught while propagating: plan 003's drive-shaft bore at station 40 OVERLAPS the relocated spar by 0.4 mm (spar O20.4 at station 28 occupies 17.8-38.2; a O4.4 shaft at 40 occupies 37.8-42.2). Minimum clear station is 41.6; moved to 43, where the tip section is 16.0 mm deep against 6.72 needed and the AK7455 SPI conduit at 54 stays clear. Also resolved OQ4 (intake curve) per owner direction and retightened plan 004's R5: the fairing must terminate tangent by Z ~30, the canonical dome's monotonic limit. The existing INTAKE_BLEND_L = 90 peaks at Z 60 — past the dome's own maximum, which is why the curves cross and the flange reads wavy. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
…collision Two sessions worked this branch concurrently and reached the same core findings independently. Merged rather than rebased so neither side's work is lost. The remote session (1ff2615) split power routing into its own federated plan and correctly identified what the first pass of 2026-08-29-001 got wrong: the O7 mm double-D conduit was never checked against real wire OD. This session reached the same conclusion by the same packing bound, then carried it forward into an adopted architecture. Collision: both sessions created a plan numbered 002, with different filenames. The remote's was already pushed, so this session's three plans are renumbered down and all cross-references repointed: 002-feat-unified-20mm-spar-trunnion-belt-drive -> 003 003-feat-nacelle-trunnion-pivot-tilt-drive -> 004 004-nacelle-mould-line-conformance -> 005 That also gives the set a clean narrative: 002 poses the architecture choice, 003 records the choice and freezes the geometry, 004 builds the pivot and drive, 005 handles mould-line conformance and ground clearance. Reconciled 2026-08-29-001, which git auto-merged into a contradiction — this session marked it fully superseded while the remote narrowed it to "still-valid scope". Absorbed is correct: both surviving findings (the AK7455 pocket still sized for the rejected MT6701, and nav-wire separation from the ESC harness) are carried into 003 U5, and its U1 fix is void because routing the nav wire into the spar bore depended on the spar rotating, which the adopted fixed spar does not. Both dispositions are now stated together rather than one overwriting the other. Annotated the remote's 002 with the outcome: both of its gates have closed. The owner selected the unified on-axis spar, and it was re-sized against the packing math that plan itself flags — 4 x O5.5 circumscribe 13.28 mm, so the adopted tube is 20 x 16.3, not 16 mm. PROJECT_INDEX.md and tools/index_tags.json conflicted, as they do on every merge; resolved by regenerating rather than picking a side. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
The nacelle tilt mechanism has to change: the four 10 AWG ESC feeds have no viable path to the nacelles, and fixing that cascades into the spar, the airfoil, the pivot, and the tilt drive. PR #192 (merged) posed that architecture choice. This PR records the choice, freezes the geometry, and closes a flight-safety defect found while verifying it. Implementation is follow-on — nothing here touches SCAD or STL yet.
The safety item is the one to read first. In the default assembly the nozzle tip sits 3.3 mm below the ground plane whenever the nacelles are vertical — a strike on every vertical takeoff and landing. It predates this work and was not previously recorded.
What was actually broken
Two independent problems with the same wiring:
CABLE_BORE_D = 7.0 mm, but two 10 AWG silicone wires side by side in one round bore need ≥ 2 × OD ≈ 11 mm. One bore holds one wire; four conductors are required.CABLE_BORE_STATION = 27.5againstSPAR_BORE_STATION = 45.15— 17.65 mm forward of the tilt axis, sweeping a ~45 mm arc every transition with no service loop, flex boot, or disconnect anywhere in the geometry.An earlier pass of
2026-08-29-001claimed the conduit was "already a dedicated, sized, working path." That was wrong — it never checked wire OD against bore area. Both sessions on this branch caught it independently.The decision, frozen
SPAR_BORE_STATIONTHICKNESS_SCALE/_TIPSPAR_ZPIVOT_ZNotable: the 20 mm CF spar is lighter than the 8 mm steel it replaces (67.5 vs 96.2 g/pair), and no landing-gear change is forced.
Findings a reviewer should weigh
The belt drive was rejected on kinematics, not preference. The tilt axis runs spanwise along hull X. A belt spanning fuselage → wingtip must have its pulleys separated along X, so its pulley axes are perpendicular to X — but the driven element must rotate about X. A spanwise toothed belt cannot reach this pivot without adding a bevel at the tip; it does not remove gears, it adds a belt in front of them. A shaft parallel to X has no mismatch. This overturned both the external recommendation and the stated preference in PR #192.
Three numeric corrections to the source analysis.
SPAR_BORE_D = 11.0does not fit the bundle at all; 16 mm OD fits only as 16 × 14 with 0.72 mm total clearance; AS5600 was already datasheet-rejected as on-axis-only and stays rejected because the bore is occupied.The variable-area nozzle cannot fit inside the canonical shell. The tandem stack fills it to 178.8 of 185.2 mm, leaving 6.4 mm for an iris measuring 58.1 mm. The 36 mm overhang is structural to the propulsion architecture, not drift — so it is reduced (flaps 40 → 30 mm) and the residual documented, rather than pretended away.
An off-CG pivot was proposed and withdrawn. It would have kept the nacelle canonical, but moving the pivot 32.5 mm off the CG swings the aircraft CG ~9.2 mm through every transition, and it cost hover clearance. Rejected on both grounds.
Merge of a parallel session
Two sessions worked this branch concurrently and converged on the same core findings. Merged rather than rebased so neither side's work is lost. Both created a plan numbered
002, with different filenames; the remote's was already pushed, so this session's three plans were renumbered down and all cross-references repointed:That gives the set a clean arc: 002 poses the architecture choice, 003 records it and freezes the geometry, 004 builds the pivot and drive, 005 handles mould-line conformance and clearance.
2026-08-29-001was auto-merged by git into a contradiction (one session marked it superseded, the other narrowed it) and has been reconciled to state both dispositions.Validation
No geometry changed, so there is nothing to render. Every dimensional claim above was computed against the current SCAD sources and the canonical shell STL rather than asserted:
tools/wing_spar_station_fit.pyfor all section-depth, wall, and thickness-scale figures.tools/wing_spar_carrythrough.pyfor the wall couple used in the spar stress check.trimeshmeasurement ofeng_left_shell24_50mm_repaired.stl(bore-centred, as the SCAD imports it) for the trunnion ring envelope, the nozzle overhang, and the intake-fairing deviation; of the baked hull-frame gear STLs for the ground plane.Unverified inputs are flagged in-plan rather than silently assumed — notably the CF tube allowable (a 300 MPa cross-ply stand-in), the 10 AWG wire OD (the BOM records none), and the aero consequences of the re-lofted section, which is no longer S1223 at 17.7 %/26.6 % t/c.
New concepts
Equal-circle packing as a bore-sizing gate
Sizing a bore for a bundle of round wires is not
n × wire area ≤ bore area— area comparison passes cases that physically cannot assemble. The governing number is the circumscribed circle of the optimal packing: fornequal circles of diameterdinside one circle, the required bore diameter is a known constant timesd.2d2.155d(1 + √2)d ≈ 2.414dFor this PR's four Ø5.5 mm conductors that is 13.28 mm at literal zero clearance — before any allowance for the bundle to rotate inside the bore as the nacelle tilts.
This one check invalidated three separate proposals that had all passed casual inspection: an 11 mm lumen (doesn't fit at all), a 16 mm tube (fits only at 1 mm wall, 0.72 mm total clearance), and the claim that two Ø7 mm conduits already carried the load. Area math would have cleared the first two.
Use it whenever a harness passes through a bore, a grommet, or a rotating joint. It is unnecessary when conductors are individually routed or free to fan out — a flat ribbon in a racetrack slot is bound by width, not by this constant.