home..

Week 06 - machine prototype

group page // repo source files // objectives

meta
tape macro placement
loaded macro (h-substrate) placement
deltaxy marshmallow manipulator
lumenpnp fixturing
tiles
h geometry improvements
media
dice hierarchy

meta >

geometry
- ordered some new macro tiles (4x4, 4x8, 8x8) and new resistor tile
machine
- lumenpnp placed tiles, on both double-sided tape and loaded macro-tile
marshmallow deltaxy innovation

tape macro placement >

compared to earlier placement tests, tuned z to be more into the substrate (resulting in a morer reliable grip) and made sure to also apply force to substrate itself, creating aligned tiles. necessary for connecting tiles to loaded macro substrate as well.

loaded macro (h-substrate) placement >

loaded macro substrate performed nearly on par with tape macro placement, observed that some tiles didn’t quite seat properly and may result in interference w/ neighboring tiles. despite seating failure, forces involved do not appear to damage connectors, which makes it easier to run continuous tests w/o fear of damaging substrate.

some improvements can likely be made to vision pipeline for better rotational correction, Shravika suggested trays that deliberately rotate to stress test the system. good candidate for cadquery.

deltaxy marshmallow manipulator >

some advanced manipulator prototyping at my cohort retreat, credit to Ilan Moyer for the original delta xy design, and Nathan Perry for suggesting the marshmallow implementation:

lumenpnp fixturing >

spent a chunk of time iterating on getting standoffs to fit monitors; they are all stable now (yay finally)
had to re-cad the pcb fixture and iterate because couldn’t find lumenpnp cad files online, annoying
- finally have the soldered h macro board installed on the lumenpnp, identifying that will be difficult to fixture new 0-offset macrotiles to lumenpnp bed w/ existing fixture strategies, will need to identify a better way to install those
- (use an offset board as the base and install new macros ontop? macro on macro?)
ring oscillator
- shravika is working on characterizing the fet board

tiles >

ordered some new macro tiles (4x4, 4x8, 8x8) and new resistor tile (1206_vert). new macro tiles feature no border, which enable them to tile just like the 1x1 tiles. new resistor tile features integrated power on one of the terminals, and can be rotate 180 deg to connect to vcc or gnd.

new tiles - batch 2 >

old tiles - batch 1 >

h geometry improvements >

h geometry needs improvement, noticing a couple of issues:

eon base
- the pitch is slightly off, but also the geometry doesn’t allow for compliance of pitch
- the geometry makes seating h difficult; the seated portion of the geometry is the widest part of eon and leads to plastic deformation, so solution is to neck down and create an energy well for latching action
- need to include some in-cut features such that base of eon can do compliance
tie-bar
- tie-bar gate leads to a protruding feature after manual singulation; in-set this feature so that it doesn’t interfere w/ seating
- make it easier to remove from manual singulation

media >

dice hierarchy >

thinking about the atomic levels of DICE currently:

1x1 tiles are atoms
1x2 tiles might be molecules
4x4, 4x8, 8x8 might be proteins(?) (macro-tiles)
additional levels of hierarchy from assembling macro-tile-sandwiches (DICE voxels)
functional organelles

in a sense, for purposes of risc-v serv computer, molecules might be the smallest unit required, and maybe not atoms