Files
hexapod-robot/2026-current/hardware/openscad/PLANNING.md
T

242 lines
14 KiB
Markdown
Raw Normal View History

# OpenSCAD body redesign: notes
`body.scad` rebuilds the body of `../fusion360/body-bottom/bottom v59.f3d`
(top + bottom plate, PCB, coxa servos; legs are placeholders).
## Source of truth
The `top` and `bottom` **solids** in `bottom v59.stp` (inside
`bottom v59.f3d_files.zip`). The `top.dxf` / `bottom.dxf` sketches in the
same zip are older and differ (front/back notch, side curve, one servo
pocket with an extra notch), so they are not used.
Coordinates: the plates are written in the original's top-view frame
(u, v) = (Fusion x, -Fusion z); `fusion_to_robot()` rotates that by -90°
into the robot frame of `robot/Config.h` (X = v forward, Y = -u left).
Fusion's up axis is Y; Fusion Y = 0 is the top face of the bottom plate,
which is also Z = 0 here.
`body.scad` slices were overlaid on the original solid edges at every
feature height (plate faces, standoff split, counterbore floor, nut wells,
PCB bosses, pivot bosses, battery bay): all match, apart from the fixes
below.
## What the original body is
- Two 3 mm plates, cavity 27 mm (Z 0..27, top plate Z 27..30).
- Outline: arm ends r 19.5 around each coxa axis, concave fillets r 15.5
at the mid arms, quadratic curves at the sides and a notch front/back.
Fully symmetric (the "asymmetry" seen earlier came from a stray line
near (128, 8) in `bottom.dxf` and construction points at x = 100).
- Coxa axes (u, v): corners (±50, ±100), mid (±65, 0) = robot
(±100, ±50) and (0, ±65), same as `Config.h`.
- Coxa servos (HS-485HB): ears rest on the top plate (Z 30), case hangs
in a 41 × 21 pocket down to Z 2. Shaft 10 mm from the outer case end.
Ear screw holes ⌀2.5, 48.25 × 10 mm. Corner servos turned 50°.
- Under each coxa axis a ⌀14 × 2 mm pivot boss on the bottom plate; the
coxa bracket (`hinge1` above / `hinge2` below, leg parts) wraps the
plate edge between the servo horn and this boss.
- 4 standoffs ⌀15 at (u, v) (±35, ±30), split at Z 13.5: M3 from the top
in a ⌀7 counterbore (floor Z 17.5), 8 mm grip, M3 nut in a hex well
from below (top of the well Z 9.5).
- PCB: long axis along the body, 4 bosses ⌀5 × 1 mm under the top plate
on the PCB hole grid (35.78 × 78.96), ⌀2.5 holes (self-tapping M3?),
⌀5 × 1 mm recess on top. PCB top face at Z 26.
- Teensy cut-out 23.78 × 66.96, centred.
- Slide switch at the front: slot 8 × 16 (v 57..73), ⌀2 screw holes 19 mm
apart, "ON"/"OFF" engraved 1 mm, readable from behind the robot.
- Battery bay on the bottom plate: rails 3 × 13.5 mm high at |u| 26.06..29.06
(52.1 mm between them), from v -74.21 (open, rear) to 31.79, end stop at
v 28.79..31.79.
## Bugs in the original, fixed here
- Left-front pivot boss at u = -50.894 instead of -50.
- Right-front corner has no coxa servo (only 5 of 6 placed).
- Only 2 of 6 legs (the mid legs), plus mirror-naming leftovers and the
empty `dontdeleteme` component (nothing to port).
- Hex nut wells rotated randomly; now one orientation.
## Leg dimensions (for robot/Config.h and the Unity sim)
Measured on the original solids (left mid leg, `bottom-leg v31`):
| Item | Value | Config.h today |
|------|-------|----------------|
| Coxa: coxa axis to femur axis, horizontal | 55.0 mm | `COXA_LEN_MM` 55 (matches) |
| Femur: femur axis to knee axis | 100.0 mm | `FEMUR_LEN_MM` 100 (matches) |
| Tibia: knee axis to foot ball centre | 150.0 mm | `TIBIA_LEN_MM` 150 (matches) |
| Foot ball radius | 5.0 mm + 0.5 mm TPU sock wall (same on the real robot) | not modelled |
| Femur/knee axes height | in the coxa's joint plane, 13.5 mm above the bottom plate's top face | coxa pivot at Z 0 (matches if Z 0 = that plane) |
| Leg mounts (coxa axes) | (±100, ±50), (0, ±65) | matches |
| Original pose | femur 0°, tibia -90° (foot straight under the knee) | |
So the link lengths need no change. What the firmware and Unity don't know
yet: the ground contact is the foot ball's surface (5 mm + sock from the
point the IK places), and the body's underside (coxa bracket lower arm) is
21.5 mm below the joint plane.
## Physical limits
Collision sweep of the model's real parts (2D, per joint, in 0.25° steps;
cross-checked with full CGAL intersections of body and legs). Joint
convention as `robot/Config.h`.
| Joint | Range | Stopped by |
|-------|-------|-----------|
| Coxa, mid legs 1/4 | -66.5° .. 66.5° | coxa bracket shell hits the plate waist |
| Coxa, legs 0 (front-right) / 3 (rear-left) | -95.75° .. 66.75° (sideways mount: -35.75 .. 126.75) | same |
| Coxa, legs 2 (rear-right) / 5 (front-left) | -66.75° .. 95.75° (sideways mount: -126.75 .. 35.75) | same |
| Femur | -101.25° .. 101.25° | femur link hits the coxa bracket |
| Tibia | -177° .. 12.75° | rod folds into the femur link / tibia shell hits it |
| Servo travel | coxa/femur ±90°, tibia -180..0° | from the `SERVOS` pulse ranges |
| Leg width at the joints | 33.7 mm to the horn side + 21.5 mm to the bearing side | neighbours touch at 55.2 mm |
Hexagon layout (owner's request): corner legs mounted 60° round from the mid
legs (mount angles ±30 / ±90 / ±150); this needs the four corner coxa
brackets turned 60° on their servo splines. In the firmware (`Config.h`):
per-leg `COXA_MIN_DEG`/`COXA_MAX_DEG`, all ±60 now (the sideways layout
needed -32..60 / -60..32 at the corners),
femur ±80 and tibia -150..-20 unchanged (inside the physical range), and
`LEG_CLEARANCE_MM` 56 (was 30). `sim_smoke` passes; turning on the spot
runs at ~77°/s with legs at least 97 mm apart (sideways layout: ~29°/s).
## Unity simulation
`export_unity.py` renders the parts into
`Hexapod_Simulation_Unity/Assets/HexapodSim/Resources/OpenScadModel/*.bytes`
(rerun it after changing the SCAD files). `OpenScadModel.cs` hides the old
scene's robot and builds a new rig with the same names at the exact
Config.h geometry; `HexapodSimulation.useOpenScadModel` switches it off.
`Editor/OpenScadModelSnapshot.cs` renders it in batch mode for checks.
## Remote (`remote.scad`)
Rebuilt from `../fusion360/remote/remote v40.f3d` the same way (slices
match the original). Bugs fixed: the lid was off by (0.406, 0.21, -0.365)
mm (sat in the ledge, posts overlapping the bosses); the display bezel's
mesh facets became a clean window; the stray Body12 (an ON-OFF bracket
line never cut in) is left out. The display is a Nextion NX4832T035
(3.5" Basic); its board's M3 holes are the lid's 94.2 x 48.5 post grid.
Reference parts in `remote.scad`: the joysticks are modelled as JH-D400X-R4 (owner's photos later showed JH-D300X-R1; 4-axis,
X/Y ±25°, twist ±45°, button; owner), mounted with the dust cover on the
lid's top face; their M2.5 square is 32.5 mm, so the lid's screw holes moved
from the original's 33 mm square to 32.5. The battery is an 18650 UPS/boost
board (89 x 22 mm, SMD-holder version, 22.4 mm tall; owner), on the bottom's
two d5 standoffs at (±27.9, -15) with the floor slots beside them for a
strap; its hole spacing (55.8 mm here) is not confirmed. No collisions.
Also from the owner: the bump on top of the middle post is a lanyard loop
(the original's tunnel through it is modelled); the square hole is the on/off
slide switch, mounted on top (same switch as the robot). Inside: an IMU on the
21 x 9 mm block (assumed GY-521), a USB-C breakout on the side-port block, and
a Teensy **4.0** (confirmed by the owner; `platformio.ini` now builds
the remote for it); it has no mounting holes and is placed on the
hole-less 8 x 24 mm block as a guess.
Open: what the 6 mm x 10 mm holes in the bottom bosses hold (magnets?), and
whether the upper two bosses are meant to be asymmetric (x -80.387 / +82.73).
## Remote v2 (`remote_v2.scad`), draft 4
A new remote for the same electronics; `remote.scad` stays as the record of
the printed one. How it's used (owner): the sticks are big twist grips that
take a whole hand each, so nothing holds the case. It hangs from a neck strap
or lies on a table. Earlier drafts: gamepad grips (dropped: nobody holds
them), strap tabs at the rear corners (it tipped forward), a hexapod emblem
inlay (owner preferred a honeycomb).
From the owner's photos and drawing: the sticks are JH-D300X-R1; the Teensy
sits socketed on a cut perfboard (56 x 80, measured) with the nRF24L01
(100 uF cap), an AMS1117, the ribbon connectors and now also the GY-521; the
switch's plate sits on top of the case, body through the slot, and the old
8 x 16 slot with d2 holes 19 mm apart fits it exactly; the UPS board's
charge/on LEDs show through two slots in the floor.
- Size 216 x 116 x 42 mm (top shell 10 mm taller than first printed, TOP_GROW: room behind the display) (old remote: 208 x 175 x 45). Sticks at u +-80,
display forward of them, strap eyelet behind the display, UPS board behind
that, switch at the rear centre. Main board under the display.
- JH-D300X-R1: dust cover d53.5 on the top face; the 40.4 mm control box
(25.5 tall) against the plate's underside, 4 x M2.5 x 14 through cover and
plate into the box (no nuts). Reaches 28.5 mm down. Pots (d16.5, 9.2 out)
towards the middle and towards the user.
- Display mounted turned 180 deg so its J1 connector is on the right, clear
of the Teensy's USB plug: set display direction 180 in the Nextion Editor.
Glass 1.0 mm under the top face, board on 4 posts tapped M3 (3.66 mm of thread, 1.2 mm under the top face);
LCD 85.7 x 55 x 3.86 plus a 1.3 mm flex cable on one side, 11.21 mm glass to
back of J1 (owner measured).
- Main board clamped at 4 points on its long sides, no holes in the board
(owner): posts beside the edges carry it, M3 x 8 button heads with d7
washers in heat-set inserts clamp the edge. 2 mm above the kickstand
recess for the wires underneath; Teensy micro-USB to the left, nRF24 to
the right. The cable plug clears the display board by ~0.2 mm and the left
stick's pot narrowly: plug and board sizes are estimates.
- UPS board on 2 standoffs behind the eyelet, Type-C end to the right (room
for the cut cable's plug); LEDs through 2 x 6 mm floor slots.
- Strap eyelet (separate part, 2 x M3 into heat-set inserts): on the top face
behind the display (slots +-5 mm to balance it), or on the front face (the
edge away from you, 2 more inserts at Z -19). Folded forward, the bail
hits a front-mounted eyelet: then park the bail back or down.
- Brace bail: U pivoting on the side faces (v -25, Z -22), M3 x 16 into
heat-set inserts; 0 = folded forward, -150 = brace against chest or belly,
-180 = straight back. Swings down under the case.
- Kickstand: 30 x 35 mm flap in a recess in the underside, hinged at the front
on an M3 x 40 into a nut trap; lifts the front a little on a table.
- Shells split at Z -12, both print without supports (top face down, bottom
on its flat underside), 45 deg chamfers, 1.2 mm locating lip, 4 x M3 x 12
from below into heat-set inserts. USB-C breakout glued into a pocket at
the front left, receptacle 0.6 mm inside the front face.
- Inlay (second colour, first 3 layers of the face-down top shell): ON/OFF
plus a honeycomb of whole cells (4.5 mm, 1.2 mm walls) right of a diagonal
through the middle, clear of display, sticks, eyelet and switch; lone
cells dropped. Colours (owner, ASA): black case, red honeycomb (PART
"inlay"), white ON/OFF (PART "text"); load "top", "inlay" and "text" as one
multi-part object. Each extra colour adds a swap on the first 3 layers only.
- Printability (the STLs sliced in 0.2 mm layers, every area not carried by
the layer below listed): the brace pivot bosses stand on pillars and the
lip's tie band steps out at 45 deg (both were floating overhangs in the
first export); what's left are bridges: the 30 mm kickstand-recess roof,
the hinge trough, hole and counterbore tops, the USB-C opening.
- Checked: no part collides with the shells or with each other, the shells
and the inlay don't overlap, kickstand clear at 0 and 70 deg, bail clear
from 0 to -180 deg (CGAL intersections).
- Open: component positions on the main board, Teensy cable plug size, LCD thickness, UPS
hole spacing; hand room untested; set `IMU_AXIS_FROM_CHIP` /
`IMU_AXIS_SIGN` in `remote/Config.h` for the IMU on the main board.
## Notes for future designs
- Robot power (still to choose): a battery or supply that gives a **stable
6 V** with enough current for all 18 servos (12 x HS-485HB + 6 stronger
ones of the same size). Servos draw well over 1 A each when stalled or
starting, and many move at once while walking, so plan for a 6 V regulator
/ UBEC of 15-20 A peak (e.g. a 2S LiPo through a high-current 6 V BEC), or
a pack that stays near 6 V under that load. Check the servos' datasheet
for the stall current. The battery must also fit the bay (see open
question 3) and leave room for the servo plugs.
- Remote: connect the IMU to the normal pin headers instead of the Teensy
4.0's underside pads; see `remote/README.md`, "Planned" (Wire on 18/19,
the two stick axes there move to A7/A8).
- Remote: measure the 18650's voltage on the Teensy to show how much charge
is left (owner, later). Divider ~100k/100k from the cell's + (holder or the
UPS board's B+, not the 5 V output) to A6 (pin 20, free), 100 nF to GND;
full cell = ~2.1 V at the pin. Firmware: average, Li-ion curve to %, show
on the Nextion with a low-battery warning (optionally send it to the robot).
- Servo wires barely fit on the real robot: the six servo headers hang
above the battery bay, and the front pair sits right over the bay's end
stop (X 28.8..31.8, top at Z 13.5). Kept as-is for now; give the plugs and
wires more room next time (e.g. a lower or shorter end stop, or a battery
that leaves space under the headers).
- The nRF24 under the switch is fine on the real robot.
## Open questions
1. Front = switch end (inferred from the ON/OFF text orientation)?
2. PCB direction: `PCB_USB_FRONT` (Teensy USB, 7805 and LED at the front,
nRF24 and battery JST at the rear) or reversed?
3. Battery size. The six servo headers hang above the bay (|Y| 14-15 mm)
and the JP1/JP4 or JP3/JP6 pair touches the end stop; servo plugs and
the battery need to share that space.
4. Leg geometry: in the original the femur servo sits further out than
`COXA_LEN_MM` = 55 suggests. Check when porting the legs.