ADR-0020: A chassis and a lid: every board in the bottom shell, one plug to the top

Status: accepted (2026-10-14)
Date: 2026-10-14
Requirements touched: the kit's buildability and repairability
Builds on: ADR-0018 (every board screwed down); supersedes ADR-0015's board posts hanging from the roof

Context

Rafa's words, verbatim:

"Is the case openable with the wiring? Maybe we need to rethink how the pieces are screwed on so one of the cases (top or bottom) opens easily?"

"yes go ahead"

When asked what to do about the charger, Rafa chose "Grow the case".

With ADR-0018's layout, 18 of 27 wires joined a top-shell part to a bottom-shell part. The opened shells could lie side by side on 60 mm of slack, but separating them meant desoldering 18 wires.

Top shell: the Radxa, charger, secure element and Pololu switch hung from its roof.
Bottom shell: the cell, USB-C port, boost and light strip.
The result: power ran back and forth across the seam.

Decision

The bottom shell is the chassis. Every board is in it, so it powers up and runs on the bench with the lid off.
- The Radxa stands on four captive-nut bosses from the battery deck, one at each mounting hole; all four now, since nothing hangs near the eye.
- Its bosses are slimmer, r 3.8 mm (0.8 mm walls round the nut). The Radxa's vendor model puts underside parts 4.0 mm from the hole centres. Top-side parts are 3.0 mm away, outside the screw head's 2.25 mm radius.
- The other boards: the charger, secure element, clock, motor, boost, Pololu switch and ideal diode stand on the deck and the floor.
The top shell is a lid. It holds only what has to touch its skin: the eye (button and light), the two mics behind their pin-holes, and the NFC tag with its tap-icon LED.

One plug joins them: a latched 12-way connector. The conductors:

Function	Conductors
Shared	ground
Mics and eye light	the mic rail
Mics' audio bus (I²S)	BCLK, LRCL, DOUT
Eye switch	its line
Tap-icon LED	5 V, data in, data out
NFC tag (I²C)	3.3 V, SDA, SCL

The part: Harwin Gecko G125, 2 × 6 at 1.25 mm, as two ready-made 150 mm pigtails (DigiKey.de). Harwin rates the contacts "1,000 operations" and the latches "100 operations". JST GH, the cheap alternative, only mates wire to board, with surface-mount headers, which the kit avoids.
Where it sits: strapped under the lid's roof. The bottom harness plugs up into it, so unplugging frees the lid completely.
The router sends every lid wire through it. A STEMMA QT cable to the tag is cut and soldered to the pigtails.

The case grows to 182 × 79 × 37 mm as built (L 184 in the spec; was 158 × 79 × 37), and about 259 g.
- Why: with every board in the bottom shell, the 41 × 33 mm charger only fit once the body was about 184 mm long.
- Ruled out: a deck running past the cell's nose end. Near the nose the deck, which sits at the seam, would cut into the light strip's zone.
- Not chosen: two cheaper alternatives were offered, a smaller charger or keeping the charger in the lid on a second plug. Rafa chose growth.
- Print bed: each shell is now 182 mm long, so it needs one at least 190 × 85 mm. The Prusa MK4 and the Bambu A1 have that; the A1 mini doesn't.
The heat stack is closed by the lid.
- The contact: the stack (pad, heatsink, pad) stands on the Radxa in the chassis, and the lid's roof patch meets it as the lid closes.
- Pad stack: the gap is taller, so the top pad is three 1.5 mm layers. The roof boss that sets the 0.5 mm squeeze stays within ±1 mm of PETG.
- The washers go: the TPU board washers, ADR-0015's shock mount for the hanging posts, have nothing left to cushion.

Consequences

The seam: 7 wire runs cross it, every one ending at the lid connector. A new buildability check, "Only the lid connector crosses the seam", holds that.
Fasteners: 36 captive nuts and 40 screws: 36 into nuts, 4 into the deck's inserts. Four heat-set inserts in all, in the deck's floor posts; the board posts are gone.
The connector is the kit's most expensive part. About €77 the pair ex VAT, and DigiKey lists a 15-week standard lead time. The plug pigtail had 16 in stock on 2026-10-14.
Two checks no longer apply as written:
- "Board posts clear of the eye" reports that no posts hang from the roof.
- "Post roots flared" and "Heat-set insert bosses" now test the deck's floor posts, which carry the inserts.

Amended: the lid also opens like a book

Rafa asked: "So the wires to the items fastened to the top need to be long enough to fully open like a clam right?"

The lid unplugs, so it doesn't strictly need to. But bring-up wants the lid open beside the chassis, still plugged in. The flat "+ 60 mm across the seam" cut rule didn't hold up when checked against that:

Hinged on the +Y edge (away from the slide switches), 3 of the 10 runs were 3–7 mm short.
Hinged on the other edge, 6 of the 10 were short.

Each run is now cut for the lid turned 180° about its +Y edge: the straight line, + 15 % for bends, + 10 % to cut. The longest is 140 mm, inside the 150 mm pigtails. A buildability check holds both limits.

Amended: shared lines cross once

Rafa asked: "Did you check the right wires cross over? I feel like I saw a bunch of wires not relevant but maybe I'm wrong".

The split was right. No power, USB, battery or motor wire crosses; every crossing is a lid signal ending at the plug.
But two shared lines were drawn more than once:
- the mic rail crossed three times (head mic, tail mic, eye light);
- the I²S bus crossed twice, once per mic.
Now each line crosses once and branches on the lid side of the plug. That's 7 runs across the seam, for the same 12 conductors.
Cut lengths: with the clamshell rule cutting every crossing run, the old "+ 60 mm across the seam" only added loops in the closed case. Crossing runs are now cut to the longer of the routed length + 10 % + 10 mm and the clamshell length + 10 %.
Total wire: 1,437 mm.

Found on the way

The packer had placed the Radxa's bosses after the deck boards, so they overlapped the clock's. They now go in first.
Union and cut guards: the shells' volume integration now errs by hundreds of mm³, so the union guard rejected a good USB-C boss. Unions are now judged on the piece, as cuts already were: is any of it left outside the result?
The band: its mesh had a 6-edge tessellation crack where the USB-C cut meets it. The export closes cracks of up to 24 edges; a bigger opening still fails "Printable meshes".

# ADR-0020: A chassis and a lid: every board in the bottom shell, one plug to the top - **Status:** accepted (2026-10-14) - **Date:** 2026-10-14 - **Requirements touched:** the kit's buildability and repairability - **Builds on:** ADR-0018 (every board screwed down); supersedes ADR-0015's board posts hanging from the roof ## Context Rafa's words, verbatim: > "Is the case openable with the wiring? Maybe we need to rethink how the pieces are screwed on so one of the cases (top or bottom) opens easily?" > "yes go ahead" When asked what to do about the charger, Rafa chose "Grow the case". With ADR-0018's layout, 18 of 27 wires joined a top-shell part to a bottom-shell part. The opened shells could lie side by side on 60 mm of slack, but separating them meant desoldering 18 wires. - **Top shell:** the Radxa, charger, secure element and Pololu switch hung from its roof. - **Bottom shell:** the cell, USB-C port, boost and light strip. - **The result:** power ran back and forth across the seam. ## Decision 1. **The bottom shell is the chassis.** Every board is in it, so it powers up and runs on the bench with the lid off. - **The Radxa** stands on four captive-nut bosses from the battery deck, one at each mounting hole; all four now, since nothing hangs near the eye. - **Its bosses are slimmer, r 3.8 mm** (0.8 mm walls round the nut). The Radxa's vendor model puts underside parts 4.0 mm from the hole centres. Top-side parts are 3.0 mm away, outside the screw head's 2.25 mm radius. - **The other boards:** the charger, secure element, clock, motor, boost, Pololu switch and ideal diode stand on the deck and the floor. 2. **The top shell is a lid.** It holds only what has to touch its skin: the eye (button and light), the two mics behind their pin-holes, and the NFC tag with its tap-icon LED. 3. **One plug joins them: a latched 12-way connector.** The conductors: | Function | Conductors | |---|---| | Shared | ground | | Mics and eye light | the mic rail | | Mics' audio bus (I²S) | BCLK, LRCL, DOUT | | Eye switch | its line | | Tap-icon LED | 5 V, data in, data out | | NFC tag (I²C) | 3.3 V, SDA, SCL | - **The part:** Harwin Gecko G125, 2 × 6 at 1.25 mm, as two ready-made 150 mm pigtails (DigiKey.de). Harwin rates the contacts "1,000 operations" and the latches "100 operations". JST GH, the cheap alternative, only mates wire to board, with surface-mount headers, which the kit avoids. - **Where it sits:** strapped under the lid's roof. The bottom harness plugs up into it, so unplugging frees the lid completely. - **The router** sends every lid wire through it. A STEMMA QT cable to the tag is cut and soldered to the pigtails. 4. **The case grows to 182 × 79 × 37 mm as built** (L 184 in the spec; was 158 × 79 × 37), and about 259 g. - **Why:** with every board in the bottom shell, the 41 × 33 mm charger only fit once the body was about 184 mm long. - **Ruled out:** a deck running past the cell's nose end. Near the nose the deck, which sits at the seam, would cut into the light strip's zone. - **Not chosen:** two cheaper alternatives were offered, a smaller charger or keeping the charger in the lid on a second plug. Rafa chose growth. - **Print bed:** each shell is now 182 mm long, so it needs one at least 190 × 85 mm. The Prusa MK4 and the Bambu A1 have that; the A1 mini doesn't. 5. **The heat stack is closed by the lid.** - **The contact:** the stack (pad, heatsink, pad) stands on the Radxa in the chassis, and the lid's roof patch meets it as the lid closes. - **Pad stack:** the gap is taller, so the top pad is three 1.5 mm layers. The roof boss that sets the 0.5 mm squeeze stays within ±1 mm of PETG. - **The washers go:** the TPU board washers, ADR-0015's shock mount for the hanging posts, have nothing left to cushion. ## Consequences - **The seam:** 7 wire runs cross it, every one ending at the lid connector. A new buildability check, "Only the lid connector crosses the seam", holds that. - **Fasteners:** 36 captive nuts and 40 screws: 36 into nuts, 4 into the deck's inserts. Four heat-set inserts in all, in the deck's floor posts; the board posts are gone. - **The connector is the kit's most expensive part.** About €77 the pair ex VAT, and DigiKey lists a 15-week standard lead time. The plug pigtail had 16 in stock on 2026-10-14. - **Two checks no longer apply as written:** - "Board posts clear of the eye" reports that no posts hang from the roof. - "Post roots flared" and "Heat-set insert bosses" now test the deck's floor posts, which carry the inserts. ## Amended: the lid also opens like a book Rafa asked: "So the wires to the items fastened to the top need to be long enough to fully open like a clam right?" The lid unplugs, so it doesn't strictly need to. But bring-up wants the lid open beside the chassis, still plugged in. The flat "+ 60 mm across the seam" cut rule didn't hold up when checked against that: - **Hinged on the +Y edge** (away from the slide switches), 3 of the 10 runs were 3–7 mm short. - **Hinged on the other edge**, 6 of the 10 were short. Each run is now cut for the lid turned 180° about its +Y edge: the straight line, + 15 % for bends, + 10 % to cut. The longest is 140 mm, inside the 150 mm pigtails. A buildability check holds both limits. ## Amended: shared lines cross once Rafa asked: "Did you check the right wires cross over? I feel like I saw a bunch of wires not relevant but maybe I'm wrong". - **The split was right.** No power, USB, battery or motor wire crosses; every crossing is a lid signal ending at the plug. - **But two shared lines were drawn more than once:** - the mic rail crossed three times (head mic, tail mic, eye light); - the I²S bus crossed twice, once per mic. - **Now each line crosses once** and branches on the lid side of the plug. That's 7 runs across the seam, for the same 12 conductors. - **Cut lengths:** with the clamshell rule cutting every crossing run, the old "+ 60 mm across the seam" only added loops in the closed case. Crossing runs are now cut to the longer of the routed length + 10 % + 10 mm and the clamshell length + 10 %. - **Total wire:** 1,437 mm. ## Found on the way - **The packer** had placed the Radxa's bosses after the deck boards, so they overlapped the clock's. They now go in first. - **Union and cut guards:** the shells' volume integration now errs by hundreds of mm³, so the union guard rejected a good USB-C boss. Unions are now judged on the piece, as cuts already were: is any of it left outside the result? - **The band:** its mesh had a 6-edge tessellation crack where the USB-C cut meets it. The export closes cracks of up to 24 edges; a bigger opening still fails "Printable meshes".