Photo soon (I've yet to make one that isn't super ugly), but the copper disc to flared end seal is repeatably vacuum tight.
In a pinch, the metal needs to expand in all directions, which is very hard. Because of this, the stress builds up until the glass breaks. In a disc seal, it only stretches in one direction which it can easly do by getting thinner.
Failed glass-to-metal seal attempt #10: Copper foil. Since thermal expansion is a fraction of the total size, a thinner wire will expand less and has more tolerance for mismatched CTEs.
10 um wire is very difficult to handle and can't carry much current, but a 10 um foil is has none of those problems.
I tried sealing some copper foil in in glass (copper's oxide forms a very strong bond), and while it looks really nice, it's very leaky. I guess letting the copper expand in one direction isn't enough.
The proper way to do glass-copper is a houskeeper/knife edge seal: A very thin walled copper tube can be sealed inside a glass tube because the walls are flexible enough to take up any stress.
... but I don't have any way to manufacture a 20-50 μm thick tube in the required diameter.
First successful (non-leaking) glow tube: The lead ins are 10 um tungsten wire through borosilicate glass. Tube is full of low pressure air.
Neon sign style with the electrodes on opposite sides... which pushed the operating voltage above what my DC power supply can do. The photo shows it running on AC (hence the glow in the evacuation stem)
Homemade glow lamp: it's borosilicate glass melted with a oxy-propane torch. The outer diameter is 11mm.
The vacuum from my rotary vane pump isn't perfect, and there's a lot of residual air left over (that's what's glowing), but it should still be good enough for a triode or small CRT.
... although I haven't figured out a good way to do glass-to-metal seals yet. This one has copper plated steel wire (CTE=~10 um/[m*K]) passing through borosilicate glass (CTE=~3 um/[m*K]), which aren't close enough to stop thermal expansion from breaking the bond: The tube only worked for a few hours.
Might have to buy some tungsten wire (CTE=~4 um/[m*K]) or figure out a graded seal: soda lime glass bonds to the wire quite well, but it's is very annoying to work with.
More questionable quality space photos: Arp 114/Arp 25/NGC 2276 (bottom) and NGC 2300 (top).
Two hours of integration with a C9.25 under the full moon.
The moon I guess. Taken with a C9.25 (230mm, fl=2300m, f/10)
Made a box by gluing together some scrap wood and CNC milling a clock-sized pocket. Looks ok as long as you don't get too close. (mostly because the workholding failed half way through)
Built a clock using some VFDs pulled from an old calculator. Much cheaper then nixie tubes and they run on 12 volts so there's no shock hazard.
... and here's a 3x3 font. Surprisingly legible:
