Breadboard Bench / build hub

Storyboard

1

2

3

4

5a

5b

5c

5d

6

7

8

9

10

11

12

13

17

18

19

20

21

15

16

22

23

24

25

26

27

14

Catalogue

Every printed piece of the bench with its exact size. Sizes are the part as modelled, length × width × height in millimetres; “as printed” is the same part turned the way it sits on the plate. Weights are solid-equivalent at the filament's density — sliced parts weigh 50–65% of that. The whole bench first:

Test pieces

The gauges: small parts printed to settle a fit before the big part that depends on it. None of them ends up in the bench.

Hardware and electronics

Everything that goes into the bench that is not printed. Where the assembly carries a model of the item, its size is measured off that model and its quantity is the number of copies in the assembly; items the assembly does not model are drawn to their standard's nominal size and marked so.

Schematics

The assembled bench from the front, from above and from the right. The views are rendered from the seated assembly at exactly 5 pixels per millimetre and the dimension lines are drawn over them in the same millimetres, so each number sits on the geometry it measures. Heights are taken from the desk: the bench stands on TPU pads 1.0 mm proud of the printed parts. Orange parts are PETG, red is TPU, the dark bar is the DIN rail, grey-blue is the display and breadboards.

Frame — skeletal, open back

The newer design: the panel is deleted, the DIN rail is the structure, the rear tray carries storage and stiffness. Everything below the modules is open on purpose.

Console — moulded, solid back

The earlier philosophy: one big print with wells and bins moulded in, and a solid raked rear panel the rail screws to — the white wall in the rear view.

Display mount — v1 H-frame vs v2 posts

v2 is now what the frame ships. v1 was spine + two uprights + two crossbars: 5 parts, 63.4 cm³, ~81 g, plus eight M3 × 12 that existed only to bolt the mount to itself. v2 is one part printed twice — a T-slot post with its own clip — spaced on the rail to the display’s holes: 2 parts, 25.1 cm³, ~32 g. 60% less plastic. The uprights only ever needed a spine because their spacing was fixed; the rail already gives X and each post’s slot gives Y, so nothing is left that exists just to hold other parts. v1 is still selectable in the 3D viewer as Mount v1.

Breadboard retention — five concepts

Nothing stops a board sliding along the 288 mm cradle today. Each concept below is staged on the real deck at the 830/400 boundary. Stops in orange.

Side-bin form studies

Motion

Build sheet

python cad/bench.py        # console + shared modules
python cad/frame.py        # frame variant + side bin
python cad/coupon.py       # console fit gauges
python cad/coupon_frame.py # frame fit gauges
python cad/check.py        # printability + freshness
python cad/screws.py       # every fastener, measured
python cad/build_viewer.py # the 3D viewer tab

Checks

Every printed gauge exists to settle a number, and a number that stays in your head is a number the CAD never gets. On the workshop LAN these are tap-through sheets on an iPad that save to disk as you go and work the constant out for you. Here they are the results: what was answered, and the constants that went back into the CAD.

Build plan

Never print a part until every fit it depends on has passed on something smaller. Click a stage for its files and checks. Ticks persist here.

Every stage below is already a Bambu Studio project. out/plates/ holds one .3mf per stage — right machine, right filament, parts in their gated print pose, per-object wall / infill / brim overrides applied. Open 01_tpu_regime.3mf and print. The per-plate card, including the two things a .3mf cannot carry (painted support dabs, and the Z = 3.2 magnet pause on stage 9), is out/plates/PLATES.md. Regenerate with python cad/plates.py.

Joint study

The question under review: the end frame’s rail joint reads as a thick block patched behind the mortise, and its gasket pocket (the indent) may be redundant now the shroud captures the rail by shape. Each candidate below was built as real geometry by an independent agent, then attacked by two more (structural + printability). Angles: front three-quarter, rear-top (boss depth and gasket land), inner-face silhouette. Click any image to zoom.

Where the tray meets the rear tie

The first pass asked “what bracket holds the rod?” and got five trapezoids for an answer. Wrong question. The actual offence is that TRAY_Y1 sits at TIE_Y − TIE_R − 2, so there is a 2 mm slot running the whole 288 mm between the tray's rear wall and the rod. That slot is the gap — and the tails were lumps stuck on to bridge it.

So: delete both tails and let the rear wall do the job along its entire length. The wall comes down and lands on the rod, and the rod's own cutter carves the mating surface, so the fit is conforming by construction and there is nothing to align. Contact goes from 21 mm² to ~1400 mm² — and the part count goes down, not up.

How far back the wall may come is not a style choice. The tray prints upright, so the wall's underside may not exceed 45°. That lands it exactly on the rod's 45° tangent point, y = 144.65 — the geometry hands over the design line for free. The three options differ only in how the wall face gets there.

Two seconds a layer, 190 layers running

The tray is 288 × 53 × 53 and 61 g. Below z = 14 it prints 1 300–14 000 mm² layers. Above the rim it collapses to two isolated 6 × 25 mm fins and stays there for ~190 layers — 4.8 s at the rim, 2.3 s at the rail land. Nothing cools before the next layer lands, and it happens on the one surface that has to be accurate.

Geometry can only do so much. Layer time = area × height ÷ flow, so 8 s at 0.2 mm and 12 mm³/s needs ~480 mm² — more than 3× what the fins give, and the H2S runs faster than 12, which makes every number here optimistic. All times below are measured off the real solids at 0.2 mm / 12 mm³/s.