A balsa model block is a foam brick with one accurate feature: a 6 × 6 mm tether groove moulded down its whole length. Every fixture here grips that and nothing else — no outer face, no corner, no assumption that a block is square. Four designs, from 5 printed parts to a rotary carrier that indexes the block through 180° without ever taking it off the rail.
The F1 in Schools model block is 223 × 65 × 50 mm of balsa with a 6 × 6 mm slot down the underside, a ⌀18 × 52 mm CO₂ chamber and a no-go zone that must survive in one piece. To machine it you have to hold it — and every commercial jig I could find clamps the outer faces, which are the two surfaces the rules do not control. A block that measures 65.4 mm across is legal and will sit in a 65.0 mm vice either sprung or crooked.
The groove is different. It is moulded, it runs the full 223 mm, and it is 6.00 mm on every block. So the fixture is built around a 6 mm rail: the block hangs on the rail, the rail's top face is the Z datum, and the two operations are the same setup with the block flipped over. Everything else in four designs is a different answer to “how do I flip it, and how do I stop it sliding”.
| feature | value (mm) | what hangs off it | where it is proven |
|---|---|---|---|
| groove ceiling — the seat | 74.000 | the block's bottom face in OP 1 | socket_floor_z / seat_z, identical in all four reports |
| rail top face | 80.000 | Z0 for the cutter, in both operations | rail_top_z_actual = 80.0 in all four |
| mirror plane | 77.000 | where OP 2's machined face lands | mirror_plane_z = 77.0 in all four |
| X stop | 223.000 | the block's front face | stop_x = 223.0; jaw reaches 203 / 213 / 223 |
| groove width | 6.000 | the rail: 5.900 printed, 6.000 bought | slot_w 6.0 / 6.0 in the compliance check |
These five numbers are identical in the printed, machined, zero-hardware and rotary reports. That is the point of building four: the fixture is disposable, the datum is not. A toolpath written against this chain runs on any of them, which is why the CAM formula in the guide is one formula and not four.
| SLOTRAIL-223 | −M machined | −ZH zero hardware | −X rotary carrier | |
|---|---|---|---|---|
| parts | 5 printed | 7 machined + 1 bought bar | 5 printed | 10 printed / 11 pieces |
| hardware | 11 screws + 8 T-nuts | 18 fasteners + 6 × 6 bar | none | 9 fasteners |
| the flip | lift off the rail, put it back | lift off the bar | lift off, 11.5 mm clear | lever through 180°, indexed |
| block length | 223 only | 223 only | 223 only | 203 – 223 (jaw travels 20) |
| mass | 298 g PLA | 2.0 kg 6061 | 439 g PLA | 448 g PLA |
| largest part | 153 mm | 285 mm base | 188 mm | 179 mm rail / 175 mm base |
| support in the slicer | 94 mm² | none — it is metal | 49 mm² | 429 mm² |
| print time | 12 – 20 h | — | 18 – 24 h | 18 – 24 h |
No Fusion 360, no sketching by hand: parametric code (build123d), geometry as boolean
algebra over boxes and prisms in absolute coordinates. The block is modelled first, from the official
dimensions, and then every fixture is built around it — so the block's slot is the fixture's rail, and
a fixture that does not clear the block is a boolean intersection that returns a number instead of a
surprise on the machine.
official dimensions → block solid (703,489.6 mm³, 9/9 exact)
│
├─ groove = datum ──→ rail 6.000 ──→ four fixture designs
│ │
│ ├─ printed 5 parts, 298 g
│ ├─ machined 7 parts, 2.0 kg
│ ├─ zero-hardware 5 parts, 439 g
│ └─ rotary 10 parts, 448 g
│
└─ every design writes report.json
│
├─ interference · tool envelope · rail contact · hex-key corridors
├─ print audit: support mm², bridge mm², overhang angles, tallest part
└─ guide_extra.py reads those reports → the guide's tables
no number in the guide is typed by hand
The compliance check is nine official features against the model: block length, width, height, slot width, slot depth, chamber diameter, chamber depth, chamber height, and the solid's bounding box. Nine out of nine exact, and the volume agrees to −0.0 %. That is not decoration — it is the reason a groove depth read off a caliper can be trusted as a Z offset, and the guide spends a whole section on measuring it twice instead of assuming 6.00.
On the first three fixtures you lift the block off the rail, roll it, and put it back. That works, and it is why they all clamp the rail: the rail's height is the only thing that has to repeat. The rotary variant refuses to lift at all — the rail becomes an axle, the block stays on it, and a lever throws it through 180° into a detent. A pin of 5.00 mm in a 5.05 mm hole gives 0.169° of lever slop, which is 0.191 mm across a 65 mm face, against a 1.0 mm rulebook tolerance.
Because the part rotates, it could not be checked at two poses. The axle group and the block are rolled against every static part at 37 angles from 0° to 180°, and the worst intersection is 0.000 mm³ — at 0°, where there is the most material on both sides. A fixture that clears at two endpoints and hits at 90° is exactly the failure a two-pose check would have shipped.
The guide's fourteen cross-variant tables are generated by a module that reads the four
report.json files at build time. It exists because hand-typed tables had already drifted:
the rotary BOM said its base was 170 mm wide while the solid measured 174, and a mass row read
0.0 kg because a cm³ volume met a g/mm³ density. Both are now impossible by construction,
which is worth more than the two fixes.
| variant | fastener | qty | into what |
|---|---|---|---|
| −223 | M4 × 8 socket cap | 2 | the rail clamps — self-tapping into a ⌀3.3 pilot |
| −223 | M4 × 16 socket cap | 1 | the collar |
| −223 | M6 × 20 + T-nut | 8 | the table's T-slot |
| −M | M6 × 45 T-bolt + T-nut | 8 | the table, through the base's eight T-slots |
| −M | M6 × 25 socket cap | 4 | up from under the base into the saddle's tapped M6 |
| −M | M5 × 12 socket cap | 4 | the clamp caps, two each |
| −M | M4 × 10 socket cap | 2 | the collar cap |
| −X | M4 × 12 socket cap | 4 | the cradle caps, two each |
| −X | M4 × 8 grub | 2 | the lever clamp on the left hub |
| −X | M6 × 60 grub | 2 | the jaw clamp through the left stand |
| −X | 5 mm dowel × 30 | 1 | the detent pin |
| −ZH | nothing | 0 | the dovetail, the wedge and 40 N of thumb |
| symptom | cause | what to do |
|---|---|---|
| the block rattles on the rail | your groove is deeper than the rail by more than 0.1 mm | one turn of masking tape on the rail — allowed to stay, it is a shim not a bodge |
| OP 2 leaves a step you can feel with a fingernail | you assumed the groove depth equals the rail height | measure both with calipers and use Z = (s − h) − zb; 0.2 mm of groove error is 0.4 mm of step |
| the first turn of an M4 is stiff and the second is easy | the pilot is ⌀3.3 because the screw cuts its own thread | keep turning. Do not drill it to 3.6 — that removes the clamp |
| the jaw binds and jumps instead of sliding | the two M6 grubs are at different depths and the jaw is being cocked | back both out two turns, push the jaw home by hand, bring them in a quarter turn at a time |
| the hub will not spin after the caps go on | the caps are torqued unevenly and have pinched the cradle | snug one, then the other, then torque in pairs — the test is a finger, not a wrench |
| the detent pin stops at the mouth of its hole | a burr in the ⌀5.05 hole | 5.0 mm drill through by hand, no press |
| the wedge creeps back under load | dust on the 10° incline | wipe both faces, rub pencil graphite on the ramp |
| the dovetail will not drop together | elephant's foot — a printed vertical wall is 0.1–0.2 mm fat at the bottom | 240 grit on the tongue's two flanks, four or five strokes |
None of these has been printed or cut. Everything above is measured on the solid model, which is a real and repeatable claim — it is not the same claim as “a part exists”. The first print is the first physical test, and the settling tests are named in the guide: rail in the real groove, straightness of the glued 277 mm rail, whether the 0.19 mm index is a measurement or arithmetic, and whether the fixture survives a season.
0.1 mm of print tolerance decides the fit. The rails are 5.9 mm in a nominally 6.0 mm groove; a printer running ±0.2 mm can make that a rattle or a press fit. The guide's answer is a turn of masking tape, not a redesign — and if you buy the 6.00 mm ground bar, the socket has to be widened to 6.2.
The rotary version indexes to 0° and 180° only. At 90° the block hangs sideways off the rail with 0.96° of droop, which is 0.83 mm across the face. It is a two-position fixture pretending to be a fourth axis, and the guide says so.
Four fixtures is three more than anyone needs. The printed one works; the others exist because each one deletes a specific failure — creep, hardware, the lift — and because comparing them is how you find out which failure actually bites.
| claim | value | how it was produced |
|---|---|---|
| the guide is one file | 478,737 bytes · sha256 7398372aaa8c5235… | make_guide.py inlines every figure as base64; opens with the Wi-Fi off |
| the published copy is the artifact | byte-identical | sha256 of guide.html equals sha256 of demos/f1-jig-mirror/index.html |
| the block is the official block | bbox 223 × 65 × 50 · 9/9 exact | solid volume 703,489.6 mm³ against 703,489.6 expected, error −0.0 % |
| no variant can hit the block | 0.000 mm³ | boolean intersection, fixture ∩ block, both operations, all four |
| the rotary variant clears itself through the whole flip | 0.000 mm³ at 37 poses, 0° → 180° | axle group + block rolled against every static part at 5° steps; worst case at 0° |
| the index is inside the rulebook | 0.191 mm across the face | 5.00 pin in a 5.05 hole → 0.169° of lever slop; the rule allows 1.0 mm |
| every part is a legal 3-axis part | 0 sloped faces | face-normal audit across all seven machined solids, 12–86 faces each |
| the guide's own layout is sound | missing [] · overflow [] · ragged [] · wide 0 · 45,521 px | verify_guide.py — anchors, column widths, table raggedness, horizontal overflow |
| it builds from scratch | 8/8 scripts · 3/3 print checks | full rebuild of all four fixtures plus three slicer audits, exit 0 |
~/Desktop/f1-jig — twenty-eight STL files (twenty printed, eight machined), four STEP
directories, four report files plus three slicer audits, and guide.html.
make_guide.py and guide_extra.py build the guide; verify_guide.py
checks it. Rebuild everything with one loop over the eight scripts — the guide is downstream of the
reports, so it cannot be built before them.
Read the official dimensions and the class rules first (Development Class, 2024-25), modelled the block, then designed against the groove as datum. Every fixture is parametric code; every fixture writes a report; the report is the only source for the guide's numbers. Checks are adversarial — interference between every pair of parts, the cutter's envelope, hex-key corridors, print overhang angles, and a full rotational sweep for the one part that moves.
The guide says “four fixtures” and means four designs, not four finished objects: one of them (the machined version) also needs a mill and a 6 × 6 mm ground bar you have to buy. Photographs are renders of the solid model, not of printed parts. Where the model and the hardware would disagree, the page says which one wins — the 5.9 mm rail is a deliberate 0.1 mm undersize, and the 0.4 mm gap at the rail's blind end is a designed clearance, not a mistake.
artifact 478,737 bytes · sha256 7398372aaa8c52350b26c95c9ed311091d5730a7cf39df724106d3f4e070a41f · mirror byte-identical