Feedstock Hopper: Design notes
Why the geometry is the way it is, and where the numbers came from. Open work is in the issues.
Funnel angle is measured on the diagonal corner
Section titled “Funnel angle is measured on the diagonal corner”On a rectangular funnel the corner runs further out than either flat face over the same drop, so it is always the shallowest surface. Pellets bridge on the shallowest surface, not the average one, so constraining a face leaves the corner well under target – the imported design sat at 32-50 degrees on its faces but 27-36 degrees at the corner.
A rectangular bin also cannot hold equal X and Y angles for one drop, so a single “wall angle” has to name a surface. The corner is the honest choice: set it, and both faces come out steeper.
Hopper literature usually quotes the angle from vertical; ours is from horizontal, which is 90 minus theirs. 60 degrees here is 30 degrees from vertical.
Feedstock minimum angles: 60 degrees virgin, 70 degrees regrind
Section titled “Feedstock minimum angles: 60 degrees virgin, 70 degrees regrind”Design targets from general bulk-solids practice, not measurements. The critical mass-flow angle depends on the material and on the wall friction of the surface it slides on [J], so a printed wall is not interchangeable with the steel these figures assume. Regrind is set steeper because irregular flake nests and bridges where free-flowing pellets do not [W].
Treat these as provisional. A hopper that bridges in service wants a measured wall friction angle against an actual printed surface, not a nudge to these.
Bulk densities: 0.62 kg/L virgin (assumed), 0.489 kg/L regrind (measured)
Section titled “Bulk densities: 0.62 kg/L virgin (assumed), 0.489 kg/L regrind (measured)”Pellet bulk density spans roughly 450-850 kg/m3 across resin types [C]; virgin
PLA sits near the middle of that. Regrind flake is markedly lighter and its
density is variable rather than a property [W] – which is exactly why the
regrind figure is now measured rather than assumed: 0.489 kg/L on our own
shred. See feedstock.md. Virgin is still the general-practice
0.62 and worth measuring the same way if it ever becomes the main feedstock.
These only affect the reported capacity, never the geometry. Every capacity figure in this repo inherits their uncertainty.
min_wall is perpendicular material, compensated on the corner
Section titled “min_wall is perpendicular material, compensated on the corner”Wall thickness is applied by insetting the inner surface horizontally, but on a
face sloped at theta the material perpendicular to it is only inset * sin(theta). The imported design applied 3 mm horizontally, which is 1.61 mm of
real material on the 10 kg funnel.
The inset is sized on the corner angle – the shallowest surface – so nothing finishes thinner than asked; flat faces and the vertical bin come out thicker. One inset for the whole profile also keeps every inner surface parallel to its outer, so the funnel meets the bin with no internal ledge to catch pellets.
Verified on the rendered mesh: at the funnel corner a 1.44 mm sphere sits entirely inside the wall and a 1.56 mm one breaks out, bracketing the 3.00 mm target.
Inner corner radii must follow the wall inward
Section titled “Inner corner radii must follow the wall inward”An inner radius clamped at a floor is no longer a parallel offset, and it thins the wall exactly at the corner that is already shallowest – about 8% in the imported geometry. So the radii subtract the inset and the geometry asserts the result stays at or above 0.8 mm, roughly two 0.4 mm extrusions, rather than clamping quietly.
Physical rules live in functions, not in comments
Section titled “Physical rules live in functions, not in comments”Anything the geometry depends on physically – how much room a particle needs,
how steep a wall has to be, how much a litre of feedstock weighs – is a
function or a registry row in hopper_specs.scad, never a bare number with a
note beside it.
That lets the same rule be read both ways. Give it a requirement and it sizes
the geometry: design_particle_size = 5 with regrind selected demands a 50 mm
outlet, and the build refuses to render if the coupling cannot provide one. Give
it the geometry and it reports the capability: every render echoes the largest
particle the whole path will actually pass, which is the narrowest section, not
the one you were thinking about.
That is how the coupling came to be 58 mm rather than 44. The requirement drove it; the assert refused the old size; the resize followed.
Bayonet joint uses bayonet-lock-scad [B]
Section titled “Bayonet joint uses bayonet-lock-scad [B]”The joint was hand-rolled with rectangular tabs. It now comes from our own
bayonet-lock-scad, which brings a hard end stop, a keyed pattern, and a
thicker socket wall outside the channel. scadman.lock pins it to the v0.12.0 tag (it
was pinned by commit 85c43ae, v0.11.0, before the library published tags);
at least 0.9.1 is required, which fixed the z alignment of the two halves when
entry_depth is not half of part_height, exactly the case used here.
The decisions behind how it is used, and how they were verified, are in the design log.
Anti-rotation is not a screw. A radial self-tapping M4 into a pocket in the
neck was tried and removed. Two sockets are not the same case, and treating them
as a symmetric pair was the mistake: disengaging a bayonet means rotating back
through sweep_angle and then travelling entry_depth axially, and gravity
sits on opposite sides of that travel at each end. At the upper socket the
body must be lifted 12 mm against its own loaded 4.24 kg, so rotation alone
releases nothing and the screw guarded a failure gravity already prevents. At
the lower socket the outlet simply drops out, and the hub is an open bore,
so the hopper would drain through it.
The lower screw was also unbuildable as drawn: it could only exit on the hub skirt’s 45 degree cone, where no head seats.
So the outlet’s retention rests on the bayonet, and on the hose thread that screws into it. It has held in service; strain-relieving the hose near the outlet is still the cheap insurance, because it removes the torque instead of resisting it.
The fed machine’s opening caps the system
Section titled “The fed machine’s opening caps the system”The machine’s own feed opening is the last section of the path, and it is metal and not ours to size, so it is a target registry row rather than a number in the check. On the GreenBoy3D toolhead it was taken off the vendor STEP files: dia 18.30 mm, centred on the extruder axis, identical in both sliding-pit parts at z 9.5 and 10.1. That is where material drops onto the screw.
Against the parallel rule that gives a system ceiling of 6.1 mm for virgin pellets and 4.6 mm for regrind flake – below the 5 mm the build is otherwise specified for. Our own sections are all sized for 5 mm, which is harmless headroom, but the assembled machine tops out lower and the spec echo now says so. Leaving the machine’s opening out of the path would report a system limit we do not have.
It also explains the 3 mm sieve the feedstock is currently prepared to: the bore is 6.1x the sieve, comfortably inside, rather than an arbitrary choice.
Build volume 250 x 210 x 210 mm
Section titled “Build volume 250 x 210 x 210 mm”The Original Prusa i3 MK3S envelope [P], and the largest printer available here
for the foreseeable future. Reported on every render; require_printable turns
it into a hard error. It is on, and every preset passes with it.
Known limits
Section titled “Known limits”Built, verified and working; nothing to do unless something changes.
- The coupling is past the bayonet library’s exercised range. Its examples
use
interface_radiusof 15 or less and its angular corrections are tuned forpin_radiusof 3.0 or less; we run 29.15 at 3.0. Rendering shows it seats, captures and keys, but the detent’s snap-past grew fourfold with the radius, because it is placed by angle. Re-verify if the coupling grows again. - Bin wall creep. The safety factors in
loads.mdare against short-term yield and say little about months of sustained load on a thermoplastic in a warm enclosure. The failure would be slow bowing, not a break. In order of cost: raisemin_wall, add perimeters, add a rib at mid-height.
References
Section titled “References”- [C] Conair – What is Bulk Density? https://www.conairgroup.com/resources/resource/what-is-bulk-density/
- [J] Jenike & Johanson – Designing bulk material storage and feeding systems https://jenike.com/designing-end-to-end-bulk-material-storage-and-feeding-systems-hoppers-silos-mass-flow-feeders/
- [W] Wijay Systems – Plastic pellet conveying systems https://wijaysystems.com/plastic-pellet-conveying-systems/
- [P] Prusa Research – Original Prusa i3 MK3S+ https://www.prusa3d.com/product/original-prusa-i3-mk3s-3d-printer-3/
- [B]
bayonet-lock-scadhttps://github.com/CameronBrooks11/bayonet-lock-scad - GreenBoy3D toolhead:
GB3DPE.mdin the firmware repo.
