# Octoprints — A public archive of prototypes and improvements

[日本語](README.md) / [English](README.en.md)

The character's correct name is **Rubber Ducky**. Display labels use that name; existing `ducky` URLs, part IDs and distributed files remain unchanged.

**All five Copilot variants have been corrected in actual bilateral geometry:1.2x,1.5x,2x,3x and4x.**
Actual counts are **21,435 / 26,734 / 35,692 / 53,575 / 71,049**, with zero external or temporary assembly aids.
Eyes, goggles, face, sides and base share the same native geometry, whole-part colors and placements across CAD, CG, movie and guide.
[Compare actual images and physical sizes](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/density-matrix.html?revision=bilateral-symmetry-v3&character=copilot) /
[15-row CSV](artifacts/studies/part-count-matrix-20260921/revisions/bilateral-symmetry-v3/comparison.csv) /
[Public verification record](archive/copilot-symmetry-revision.json).
The17,873-part1x reference remains the original with known asymmetry and is not counted among the five corrections. Earlier images, models and URLs remain available.
Actual CAD symmetry and finite raster-boundary differences are reported separately. The2x one-sided eye comparison differs by1pixel (2pixels in the full eye mask);
all measured silhouette/material boundary distances remain within1pixel, rather than being rounded to zero. Physical retention and strength remain untested.

**[Public site / English](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/)** ·
**[Actual 3D assembly guide](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/density-guide.html)** ·
**[All downloads](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/downloads.html)** ·
**[Physical feedback](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/feedback.html)** ·
**[Design journey](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/history.html)** ·
**[Choose a model](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/models.html)**

Start with the homepage **print-data catalogue**: choose one of five variants for each character by part count, assembled size and assembly aids.
Use **“Rotate in 360°”** on a card to drag/swipe and zoom the selected actual model. Its 3D data loads only when opened.
Copilot's earlier1.2x/1.5x/2x/3x [support-free structural revisions](archive/copilot-support-free-revision.json) remain as historical versions:
**21,382 / 26,673 / 35,630 / 53,287 parts**, each with zero external or temporary assembly aids.
[Revised comparison with the actual 1× reference](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/density-matrix.html?revision=body-support-v2&character=copilot) /
[revised CSV](artifacts/studies/part-count-matrix-20260921/revisions/body-support-v2/comparison.csv).
The earlier zero-aid4x case with71,261 parts stays unchanged. These are separate historical versions, not the five bilateral revisions above.
The [four-revision publication verification record](artifacts/studies/part-count-matrix-20260921/revisions/body-support-v2/publication-verification.json)
records actual CAD/IDs, nominal loads, public browser/video checks, anonymous downloads, immutable history and storage limits.
The low-part r3 designs (Mona 519 / Copilot 695 / Rubber Ducky 413) are excluded from normal choices; trials, failures and refinements have a separate history page.
“STL available” does not mean physically validated printing. All cases remain unsliced and untested, with full-kit printing on hold. The sections below preserve revision-specific records.

### Upstream and reference projects

The original models come from **[martinwoodward/octoprints](https://github.com/martinwoodward/octoprints)**,
a project publishing GitHub-themed 3D-printable mascots. We reference pinned revision `981a85f0bec9d1a9c280fd5719d0d797852229c4`.
This archive adds brick decomposition/grouping, native CAD, BOMs, assembly sequences, comparisons and 360-degree viewing.
Model derivatives follow **CC BY-NC 4.0**. [Three.js](https://github.com/mrdoob/three.js) supplies browser rendering,
and [meshoptimizer](https://github.com/zeux/meshoptimizer) is used for offline display LOD and bit-exact native-buffer transport.
See [project introductions, modifications, interface references and rights](ATTRIBUTION.en.md)
and [third-party software licenses](viewer/assets/THIRD_PARTY_LICENSES.txt).

The [historical r3 proposed-sequence record](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/assembly.html)
reads IDs, types, colors, base positions, rotations and steps from the same revision's actual manifest. It is not physical assembly approval.

[Public repository](https://github.com/ktanino10/octoprints-brick-kit-downloads) ·
[日本語サイト](https://ktanino10.github.io/octoprints-brick-kit-downloads/ja/).
The legacy root URLs remain available:
[home](https://ktanino10.github.io/octoprints-brick-kit-downloads/),
[viewer](https://ktanino10.github.io/octoprints-brick-kit-downloads/viewer/),
[downloads](https://ktanino10.github.io/octoprints-brick-kit-downloads/downloads.html),
[feedback](https://ktanino10.github.io/octoprints-brick-kit-downloads/feedback.html), and
[history](https://ktanino10.github.io/octoprints-brick-kit-downloads/history.html).

This is a noncommercial prototype record for approximately 180 mm-tall Mona, Copilot, and Rubber Ducky models assembled from independent bricks.
**On 2026-09-20, implementation of an 8 mm common-block redesign was authorized.**
The revised approach prioritizes parts such as 2×2 and 2×4 blocks, 9.6 mm brick bodies and 3.2 mm plates,
with open undersides and wall/tube/rib retention. Earlier 4 mm Fine and 1.10 mm pin constraints are superseded where they conflict with this direction.

The current revision is **`r3-8mm-20260920`**. The [revision registry](archive/revisions.json) and
[current-model page](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/models.html) use actual, checked deliverables from that same revision.
No earlier image or geometry is substituted for the redesigned models.
See the [common-block design/publication guide](docs/COMMON-BLOCKS.en.md) / [日本語](docs/COMMON-BLOCKS.ja.md).

| Actual r3 model | Individual parts | Types used | Width × depth × height / mm | Small 1×1 exceptions |
|---|---:|---:|---|---:|
| Mona | 519 | 68 | 199.8 × 167.8 × 181.0 | 5 |
| Copilot | 695 | 86 | 247.8 × 199.8 × 181.0 | 10 |
| Rubber Ducky | 413 | 56 | 151.8 × 215.8 × 181.0 | 0 |

There are **1,627 instances sharing 133 master types**. Seven new FreeCAD documents cover two shared libraries,
three assemblies, the trial and native section checks. The release includes complete/exploded PNGs, three Blender scenes,
three 6-second 720×720 turntables, per-type STL/STEP, **33 color-grouped geometry 3MF plates** (11 / 13 / 9),
individual/type-color BOMs and bilingual guides.

- **[New release and ZIP packages](https://github.com/ktanino10/octoprints-brick-kit-downloads/releases/tag/archive-2026-09-20-r3)**
- [Nine-piece trial 3MF](artifacts/revisions/r3-8mm-20260920/cad/trial/plate.3mf) /
  [equivalent trial STL](artifacts/revisions/r3-8mm-20260920/cad/trial/plate.stl) /
  [English trial instructions](artifacts/revisions/r3-8mm-20260920/TRIAL-en.md).
- [Same-scale old/new dimensions](artifacts/revisions/r3-8mm-20260920/size-comparison.png) /
  [actual BRep open undersides and sections](artifacts/revisions/r3-8mm-20260920/open-underside-sections.png).
- [Relocated native reopening and geometry-preservation record](archive/portability/r3-8mm-20260920.json).

The new trial has **nine pieces**: three male 2×2 variants, five female 2×4 variants, and one thin 2×4 plate.
Male values are **stud diameter corrections −0.10 / 0 / +0.10 mm**; female values are
**radial clearances −0.04 / 0 / +0.04 / +0.08 / +0.12 mm**, not the earlier r2 diameter differences.
The 0.4 mm nozzle is a comparison candidate, not claimed installed. The last reported installed nozzle was 0.2 mm;
confirm the current hardware.

**The new revision has not been physically tested. Fit, retention and full assembly are unvalidated; full-kit printing is ON HOLD.**
The small-part handling and blocked-hole feedback belongs to the earlier r2 4 mm trial on 2026-09-19, not the new revision.
Those holes were blocked **before** ultrasonic cleaning. The cause, actual slicer settings and condition-specific retention remain unresolved.
Geometry-only 3MF files remain **NOT_SLICED**, not configured Bambu Studio projects or G-code.

## All fifteen actual count-multiplier cases and native 3D assembly guides

**[Bilingual count matrix](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/density-matrix.html)** /
**[Copilot 1.2× actual 3D guide](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/density-guide.html?case=copilot-p120)** /
**[Copilot CG, Blender, animation and native CAD package](https://github.com/ktanino10/octoprints-brick-kit-downloads/releases/tag/part-count-matrix-20260921-copilot-p120)**.

CG, animation, Blender, native CAD, BOM and actual 3D guides are distributed for **all fifteen cases: five support-free Mona revisions, five Copilot cases and five Rubber Ducky cases**.
These are actual figure-part counts. The [fifteen-row comparison CSV](artifacts/studies/part-count-matrix-20260921/comparison.csv) records fixed targets, actual ratios, differences, dimensions and types.

| Actual pieces | 1.2× | 1.5× | 2× | 3× | 4× |
|---|---:|---:|---:|---:|---:|
| Mona, support-free revision | 14,913 | 18,649 | 24,811 | 37,255 | 49,483 |
| Copilot | 21,388 | 26,690 | 35,648 | 53,350 | 71,261 |
| Rubber Ducky | 11,586 | 14,460 | 19,309 | 28,932 | 38,473 |

Copilot's 1.2× case contains **21,388 parts versus target 21,448 (−60; actual ratio 1.196665×), 155 types,
519.8 × 439.8 × 397.0 mm**. The 1.5× case contains **26,690 parts versus target 26,810 (−120; actual ratio 1.493314×),
162 types, 567.8 × 487.8 × 425.8 mm**. The 2× case has **35,648 parts versus target 35,746 (−98; actual ratio 1.994517×),
182 types, 631.8 × 535.8 × 469.0 mm**. The 3× case has **53,350 parts versus target 53,619 (−269),
195 types, 727.8 × 615.8 × 541.0 mm**. The 4× case has **71,261 parts versus target 71,492 (−231),
214 types, 791.8 × 679.8 × 598.6 mm**. Separate temporary aids number one for 1.2×/1.5×/2×, two for 3× and zero for the actual 4× case.
The Mona-specific no-stand condition is not extended to Copilot. Crown/cheek steps and asymmetric lime regions remain;
neither aesthetic nor manufacturing approval is implied.

**New requirement: Mona whiskers must need neither external supports nor temporary assembly pedestals.**
The earlier Mona 1.2× and 1.5× data are preserved as historical support-dependent designs.
They are separate from the new support-free revisions and do not count toward the requested fifteen cases.
The two old Mona cases remain separate historical records. Simply hiding supports in the CG does not qualify.
[Mona's new 1.2× revision](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/density-guide.html?case=mona-p120-root-v2)
has **14,913 parts (target 14,922; −9), 118 types and zero external/temporary assembly aids**.
Two actual one-piece modules connect the whiskers into the cheek body. Evidence covers native contacts, 49 material sections,
support-stage sequencing and all 28/30 nominal CAD load-prefix checks. Printed mass, PLA retention and strength remain unmeasured.
Actual tip-bottom heights of 83.2/88 mm are preserved while insertion uses the 92.8 mm support stage.
Old URLs are not silently replaced.
[Mona's new 1.5× revision](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/density-guide.html?case=mona-p150-root-v2)
has **18,649 parts (target 18,653; −4), 128 types and zero aids**, separate from the historical 18,675-part design.
Its flatter uneven mouth and eye/nose color-transfer differences remain. No aesthetic or physical-retention approval is implied.
Mona's new 3× case has **37,255 parts (target 37,305; −50), 160 types, 615.8 × 503.8 × 545.8 mm and zero aids**.
Mona's new 2× case has **24,811 parts (target 24,870; −59), 135 types, 535.8 × 439.8 × 473.8 mm and zero aids**.
All five native contact/insertion/balance targets are checked: four WR roots and one actual internal CT cheek backing.
The new 4× case has **49,483 parts (target 49,740; −257), 172 types, 679.8 × 551.8 × 598.6 mm and zero aids**.

The first Rubber Ducky multiplier is **1.2×: 11,586 parts (target 11,603; −17), 93 types,
311.8 × 455.8 × 397.0 mm and zero temporary aids**. It uses the frozen 9,669-piece baseline, not the old 413-piece r3.
Rubber Ducky's actual temporary-aid counts across its five cases are **0 / 0 / 3 / 0 / 1**.
The 2× case keeps three separate aids outside its 19,309-piece figure count; the 4× case keeps one outside its 38,473-piece count.
Adjacent aids are separate IDs even when they look like one object. Guides show actual supported IDs, assembly steps and heights;
do not remove aids before physical retention/stability checks.
Slicer-generated print supports are a separate unvalidated condition.
All three 1× baselines now have separate public CG, native CAD, Blender and three-chapter animation packages linked from their matrix reference cards.
**The three baselines do not count toward completion of the fifteen multiplier cases.**
Mona's additional support-free reference is explicitly shown as **12,411 actual parts (−24 after root integration)**.
The **frozen multiplier denominator remains 12,435**, and all five targets are unchanged. Historical supported 1× assets remain downloadable.
The additional reference is not counted among the fifteen multiplier cases.
Baseline counts are Mona 12,435 / Copilot 17,873 / Rubber Ducky 9,669, using a consistent initial-Fine-C-adapted 8 mm approach.
This is a **comparison assumption, not user adoption**; earlier r3 counts of 695/413 are not mixed in.
Multipliers refer to individual physical parts, not linear dimensions.

[Historical Mona 1.2×](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/density-guide.html?case=mona-p120)
targets 14,922 and contains **14,936 parts (+14; actual ratio 1.201126×), 118 types,
439.8 × 359.8 × 397.0 mm**. Three actual temporary supports are separate from the figure count.
The additional Mona 1.5× targets 18,653 and contains **18,675 parts (+22; actual ratio 1.501809×), 129 types,
487.8 × 391.8 × 425.8 mm**, with four separate temporary supports.
Its mouth appears flatter and less even than initial C/the 1× smile, with inner-eye and nose color-transfer shifts.
This visual regression is disclosed; a larger count is not an appearance pass.
The guide uses native meshes for all-direction radial separation, empty-to-bottom-up part/course/stage playback,
type/color BOMs, all matching-part destinations and selected-part underside inspection. Motion is not a collision-free disassembly simulation.

Heavy cases offer an explicitly labeled **lightweight display model** with an original-native switch.
Only ordinary rectangular BR/PL meshes lose triangles; retained vertices, root/special geometry and selected-part previews preserve their native originals.
CAD/STL/Blender/final CG, counts, IDs and support evidence are unchanged. CP2 drops from about 290M to 62M rendered triangles.
On the measured macOS WebKit host, completed drag frames fell from about 402/408 ms to 117/194 ms at 1440/390px.
The approximate 0.04 mm simplifier error is not a manufacturing tolerance or a Hausdorff guarantee, and smoothness is not guaranteed on every device.
[Quality limits, environment and actual timings](artifacts/studies/part-count-matrix-20260921/display/performance.json) are preserved.

The public ZIP includes a relatively linked FreeCAD assembly, shared masters and original parametric authoring documents,
STL/STEP, Blender, assembled/exploded CG, BOM/motion and a 12-second turntable/radial/bottom-up movie.
Extract everything and preserve relative directories. Print plates/slots, slicing and G-code are not generated.
Local eye-rim, nose and smile color-transfer differences remain; more parts do not guarantee better likeness.
**NOT_SELECTED / NOT_SLICED; physical fit, retention, loads and tipping UNKNOWN; full-kit printing on hold.**
The [machine-readable receipt](archive/block-budget-matrix.json) becomes `READY` only after all fifteen cases,
comparison assets, anonymous downloads and the real public browser pass. Publication completion is not physical approval.
[The final public verification record](artifacts/studies/part-count-matrix-20260921/publication-verification.json) records all fifteen actual counts/native checks,
the public browser, 57 actual movie-chapter playbacks, anonymous hash verification and environment limitations.
The matrix includes **nine final comparison images**, each combining the actual reference with five multiplier cases.
Front/three-quarter sheets use matched projected height; separate physical-size sheets use a uniform 0.9 px/mm in the source image and shared ground.
Responsive display is not a physical ruler. All eighteen source manifests/BOMs, original CG pixels and cameras are hash-bound;
no proxy image or geometry regeneration is used.

## Further Mona refinement toward initial Fine C

The **[actual refinement image and count comparison](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/mona-refinement.html?comparison=face-front)**
compares initial Fine C, the previous 36 cm pilot and the refinement in matched face, front and three-quarter views,
with additional forehead, eye-rim and mouth close-ups.

The refinement has **12,435 parts using 121 types, at 423.8 × 343.8 × 368.2 mm**:
**+1,527** against the previous 10,908 parts and **−1,402** against initial C's 13,837.
Neither reference is a ceiling, and neither a higher nor a lower count is a success criterion by itself.

The 8 mm XY interface is retained while C's actual cells, single-color boundaries and course pattern guide a new
**4.8 mm half-height body family**. Actual families are 3.2 mm: 50 parts/6 types; 4.8 mm: 12,359 parts/110 types;
and 9.6 mm: 26 parts/5 types. There are 77 placement layers. The 1.6 mm bookkeeping unit is not a thin part or an FDM layer height.
A real native 0.20 mm upper-perimeter chamfer affects only 3,763 exterior parts; this is not a rendering-texture change.

**The result is not identical to initial C; user visual approval remains pending.**
Fine bright edges, dark-eye surfaces and small nose/shadow/outline differences remain. The finish is not C's dark open vertical seam.
C-target render metrics improve, but the separate front IoU against the original source decreases from about 0.95474 to 0.94882,
which is also disclosed. These are not perceived-likeness percentages or printed accuracy/strength pass values.

The total includes 31 one-cell exceptions and 50 foundation parts, but excludes two temporary pedestals.
Retention, load, tipping stability and safe pedestal removal remain unverified.
This is **NOT_SELECTED / NOT_SLICED, with full-kit printing on hold**—not a new complete kit, plate set, trial or print job.
The first 36 cm URLs/images/counts, current r3, historical artifacts and all seven existing release ZIPs are preserved.

## First 36 cm pilot: Mona likeness, size and actual quantities

The **[Mona comparison against initial Fine C](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/mona-likeness.html)**
shows the original, initial C, published r3 and the new approximately 36 cm pilot.
Front and three-quarter views use **matched on-screen heights**, with a separate facial close-up.
True scale is shown separately; a larger display is not counted as better likeness.

| Mona configuration | Actual parts | Types | Width × depth × height / mm |
|---|---:|---:|---|
| Unsubdivided original | Not applicable | Not applicable | 69.126 × 56.769 × 60.776 |
| Initial Fine C | 13,837 | 7 | 211.82 × 171.82 × 180.71 |
| Published r3 | 519 | 68 | 199.8 × 167.8 × 181.0 |
| New Mona pilot — unselected | 10,908 | 137 | 407.8 × 343.8 × 360.2 |

The pilot retains the 8 mm interface and resamples the original geometry; it is not an enlarged coarse r3 arrangement.
It has 2,929 fewer parts than initial C and 10,389 more than r3. Different grouping reduces the physical count relative to C,
while occupied sampling cells increase from 82,357 to 115,033. A low part count is not an aesthetic success criterion.
There are 10,572 plates, 336 internal blocks and 112 placement layers. The 3.2 mm value is plate body height, not an FDM slicing layer height.

**User visual approval is still pending.** The broad same-color forehead, white eye rims and stepped smile are not identical to initial C.
The total includes 44 foundation parts and excludes two separate temporary whisker pedestals; these must not be removed before actual retention checks.
Weight, retention, tipping and printability remain unvalidated: NOT_SLICED, with full-kit printing on hold.
This publication provides a Mona image/count comparison only, not a new kit, trial, one-piece model or adoption of the other two characters.
Current r3, the earlier A/B comparison and formal releases remain unchanged.

## Earlier shape and actual part-count comparison

The **[bilingual image and quantity comparison](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/shape-options.html)**
shows adopted r3 alongside two unselected studies with matched camera, scale and palette.
All retain the 8 mm connection system and 181 mm overall height.
Counts were checked against actual manifest and BOM part IDs; they are individual instances, not type counts or estimates.

| Character | Adopted r3 | A: surface plate refinement | B: contour / slope replacement |
|---|---:|---:|---:|
| Mona | 519 | 670 (+151) | 519 (no change) |
| Copilot | 695 | 916 (+221) | 695 (no change) |
| Rubber Ducky | 413 | 522 (+109) | 413 (no change) |

A refines surface height steps with 3.2 mm plates, at the cost of more seams, exposed studs and assembly work.
B replaces 36 blocks per character with slopes one-for-one, keeping the part count unchanged but adding specialized types,
thicker roofs and shallower underside cavities. Coarse eyes and mouths from the 8 mm XY grid remain.
No new 1×1 chips, trial sets or one-piece print version were created.
**The alternatives are unselected and physically unvalidated. Adopted r3, existing downloads and formal releases are unchanged;
full-kit printing remains on hold.**

## Preserved earlier revisions and physical feedback

The quantities below describe Phase1/r2 history. They are not the new revision's counts or trial sets.

| Category | Preserved work |
|---|---|
| Phase1 candidates | Three characters at three detail levels: **9 candidates**, 9 images, **9 Blender scenes**, **3 Balanced videos**, individual/aggregate bills of materials (BOMs), and placements. |
| Selected r2 | Mona Fine 4 mm / Copilot Chunky 8 mm / Rubber Ducky Fine 4 mm: 3 images, 3 Blender scenes, 3 videos, BOMs, proposed assembly sequences, and part IDs. |
| Native CAD | The r2 assemblies have **26,766 individual instances sharing 236 native types**: 2 shared FreeCAD libraries, 3 assemblies, and STEP/STL files by type. The legacy version preserves **21 basic types** and **5 historical 6 mm coupons**. |
| Trials and aids | Separate 4 mm and 8 mm trial sets, **11 parts each**, with 3MF/STL/STEP, FreeCAD, layouts, and a results CSV; left/right temporary supports for Mona's whiskers. The two legacy root-level 4 mm downloads are unchanged. |
| Physical feedback | 10 metadata-sanitized JPEGs, 2 silent MP4s plus 1 combined MP4 in the corrected order, and records of the issues and redesign discussion. |

The selected r2 part counts are **Mona Fine 4 mm: 13,434 / Copilot Chunky 8 mm: 3,021 / Rubber Ducky Fine 4 mm: 10,311**.
Selection for appearance is separate from manufacturing approval. Changes to grouping changed the seams; pixel-for-pixel identity with Phase1 is not claimed.

- [Static gallery of all 9 candidates](artifacts/phase1/gallery.html) / [Selected r2 images and videos](artifacts/selected/r2-20260919/gallery.html).
- [Physical photos, videos, and reports](feedback/2026-09-19/README.en.md) / [Feedback page](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/feedback.html).
- [Design history and unresolved conditions](https://ktanino10.github.io/octoprints-brick-kit-downloads/en/history.html).
- [Complete file inventory, sizes, and SHA-256 hashes](archive/inventory.json) / [Checksums](archive/SHA256SUMS.txt).
- [Public-copy metadata cleanup and native-geometry preservation record](archive/portability.json).
- [English archive and ZIP guide](docs/ARCHIVE.en.md) / [English r2 trial guide](docs/TRIAL-GUIDE.en.md).

The four existing ZIP packages at **`archive-2026-09-19-r2` are immutable archives**.
They retain their original documents; these English companions are provided separately, outside the ZIPs.

To open a FreeCAD assembly, extract the entire ZIP and preserve the relative layout of `cad/assemblies/` and `cad/libraries/`.
The preserved **Phase1/r2 history contains 11 FreeCAD documents and 12 Blender scenes**.
The 26,766 instances are not 26,766 separately supplied STL files.
The viewer supports the assembled model, exploded view, layers, proposed assembly sequence, part IDs, BOM, viewpoints, and videos.
It does not run CAD, slice geometry, or operate a printer from the browser.

The originating development repository and history, runtime environment, conversations, and internal logs are not included.
Where local paths or other metadata were cleaned up, source hashes and public-file hashes are distinguished.
Use the public inventory to verify current downloads. Historical numerical checks must not be read as physical approval.

For the selected r2 4 mm joint, the inlet wall was revised from the old **0.877 mm** to **1.21 mm**.
The stud has a **1.10 mm stem diameter, 0.98 mm tip diameter, and 0.60 mm height**.
These are CAD dimensions; strength and retention have not been validated.

## Desired output modes — pending clarification

The requested direction is **printing separate parts and assembling them afterward**, plus a proposed **“print everything at once”** alternative.
The second option is **not yet defined**: it could mean printing the assembled model as one piece, or batch-printing still-separate parts on one plate.
A one-piece model has not been selected. The new authorization concerns redesigning the separate-part assembly system.
Color-grouped geometry-only plate layouts do not confirm a preference for a monolithic model or approve manufacturing.
The assembled viewer is not a one-piece printable file. No slicing, printing or printer transmission is performed.
**Full-kit printing remains ON HOLD.** This intention is separate from the existing trial record.

## Legacy 4 mm joint and small trial set

The following preserves the historical preparation and observation procedure; it is **not a current instruction to reprint**.

> **2026-09-19 physical feedback: redesign under discussion**
>
> The trial parts were reported to be “small and difficult to make,” with “some holes filled with plastic.”
> The current design will not be carried directly into full-model printing. The files below remain as a record of the earlier trial.
> Read the [photos, videos, and redesign discussion](feedback/2026-09-19/README.en.md).
> The rights for these photos and videos are explained in that record's “Photo and video rights” section.

**NOT_SLICED — These trial parts are unsliced, and a physical-fit pass has not been confirmed. Issues were reported in the first physical trial.**

These files are for comparing joints under the design assumption of a **Bambu Lab P1S, 0.2 mm nozzle, and PLA**.
They are not a Bambu Studio project with printer settings, G-code, or print-ready data.
A 0.2 mm nozzle does not guarantee that the thin stem will resist breakage, or that fit and retention will be satisfactory.

### Downloads

- [Geometry-only 3MF with 11 parts](p4-trial-11-parts.3mf) — Layout data referencing shared types. This is normally the preferred input.
- [STL of the same 11 parts](p4-trial-11-parts.stl) — Units are millimeters. This is an alternative to the 3MF, not an additional set.

**Do not load the 3MF and STL together on the same plate: doing so duplicates the parts. Use only one.**
Download through the GitHub file page and confirm that the saved file still has its `.3mf` or `.stl` extension.

### Contents

| Part | Quantity |
|---|---:|
| Receivers identified by 1–4 notches | 1 of each; 4 total |
| Common male keys | 4 |
| Lower parts of the offset joint | 2 |
| Upper part of the offset joint | 1 |
| Total | **11** |

Receiver notch counts **1 / 2 / 3 / 4** correspond to diameter differences of **−0.05 / 0 / +0.10 / +0.20 mm**.
The difference is **the receiver's straight bore diameter minus the male stem diameter**, not a per-side clearance.
The −0.05 mm condition is intentional nominal interference.
Even keys printed from the same type can differ physically or wear, so use a separate key for each condition.

### Preparing in Bambu Studio on another PC

1. Open the geometry-only 3MF and select an **existing preset for the actual Bambu Lab P1S / 0.2 mm nozzle**.
   If that preset is not available, stop. Do not substitute another printer or a 0.4 mm nozzle preset.
2. Select the plate type and PLA that will actually be used. This pack does not specify a PLA brand, color, or plate type.
   An existing, compatible Generic PLA preset may be a provisional candidate, but it must be checked against the actual material.
   Use compatible existing profiles for layer height, temperature, speed, and other settings; record the profile names and values used.
3. Confirm that there are **11 parts**. **Receiver holes face down, male stems face up, and the bottoms are at z=0.**
   Do not change orientation or size by automatic rotation or scaling.
   The layout is within a nominal **256 × 256 mm** area, but must be checked again against the selected printer and plate.
4. Slice locally and inspect the layer preview.
   Check that receiver holes are not unintentionally filled at their entrances, that paths for the thin male stems have not disappeared,
   that the hole roofs have bridging and top-surface paths, and that unwanted supports do not block the holes.
   If there are warnings or anything you cannot judge, stop for review rather than proceeding to printing.
5. Record the selected printer, nozzle, material and process profiles, layer height, warnings, and preview.
   Save a Bambu Studio project locally under a **different name from the original file**.
   Record print time and PLA quantity only from an actual slice using those settings; do not substitute estimates based on geometry alone.

This pack has not validated profile compatibility, layer previews, print time, PLA quantity, or supports.
Confirming that geometry exists and fits in a layout is not a slicing pass or a physical manufacturing guarantee.
There is no automatic printer connection or Print/Send function.
**Do not connect to a printer or use Send/Print as part of this preparation.**

### Minimum observations for a physical trial

Begin with the loosest condition, **+0.20 mm (4 notches)**. If it is tight, stop rather than forcing it.

- Actual nozzle diameter, PLA, plate, profiles used, and layer height.
- Whether the part seats without excessive force, with any incomplete seating, play, or bottoming-out noted.
- Whether the key falls out under its own weight when the receiver is gently turned upside down.
- Cracks, stress whitening, or looseness after **5 insertion/removal cycles**, and any damage to the receiver roof. If possible, include photographs and the male stem's outside diameter.

There is no need to force calipers into the small holes. Record unmeasured values as unknown.
Only try the **3-part offset joint** after basic fit has been checked.
This small trial does not validate the loads, impacts, or assembly feasibility of the large model.

**Physical fit, retention, and full physical assembly remain unvalidated. This is not a final manufacturing release.**

## License and attribution

The model-derived data is for noncommercial experimentation under [CC BY-NC 4.0](LICENSE).
See [ATTRIBUTION.en.md](ATTRIBUTION.en.md) for upstream credit, modifications, trademark notices, and the separate rights for contributor-owned physical-feedback photos and videos.