Ellipse :
I wish you lots of success with this - but I want to pass on a few risks I see, as the little bit I can do to ameliorate them.
I have a large stash of parts accumulated over the years, so I'm set, personally. For everyone else, however, I'd love to see progress against many of the challenges you are facing this summer.
The flippies, keystems and springs are some parts of the F keyboards that people have had trouble replicating.
I find Unicomp's single-piece keys do not so much replicate the feel of the original single piece IBM keys as found on the XT/AT/bigfoot/4077s, though the difference between them and the two piece model M's is not troubling to me, since I prefer one-piece. This is very irritating for me, since I would love to be able to source new keys for fun F builds.
The barrels are likely to be successful fast, as they are relatively flexible tolerance simple molds and are not picky about plastic, so long as it slides well with old PBT and whatever Unicomp is currently using. In a pinch, unicomp barrel-plates can be chopped up, but I seriously doubt this will be needed.
I'm glad you are reducing risk on the case-side with CNC - reduces options for bulk cost reduction, but makes it way more likely a nice case will be available soon within the cost window.
No-one I know has successfully replicated the flippies, whether in resin, home IM, or whatever - nevermind with conducting material to get a working one that both flips well and also scans well with minimal bounce. I don't even know if they require a minimum resistance to reduce ringing - conductive plastics can be much lower resistance now than 30-40 years ago in the F/beamspring era. Bit of fiddling with a scope testing some some fiddled keys could likely answer this part. I can't tell from your image of the flippie (IBM called them capacitive flip-plates) to what degree you have replicated IBM's work, as only the topside is shown in the rendering above.
So - I feel you need to make sure your model *flips* well, and you need to also confirm it scans and bounces nicely. A nicely behaved F switch has a single, short 1.5ms bounce at the front for the make, imo. The release is smooth and clean, generates a simple near linear signal on break.
I guess I'm trying to warn that it might take a few more tries than one might consider reasonable to get all the variables playing nicely with each other. If your time and personal energy hold out - many of the problems that surprise you can be managed without huge amounts of additional cash - but it will take time. If you stick it out, ultimately, they will ship.
Luckily some risk amelioration can be had for those who can source an old XT or 122 as a stopgap - lots of handy parts in there. The 4704 case is currently unobtanium, regardless of price - pcb is a little fiddly, but copying the original exactly is an option.
Effective springs can be had by swapping out from unicomp flippies - they aren't the same strength, but they are made to sufficiently close tolerances, so work fine. Slight feel difference, but not one I mind, personally.
I see the biggest risk - and the biggest win, with the case. The other dependencies are less risky, or less important, imo.
I'd love to see working flippies though - I see them as a bit risky, but totally hope you get them working well so that the community has an arbitrary supply going forward. : )
Bestest Lucks!
dfj
