whitequark changed the topic of #tinyqma to: design of an open hardware DDS-based QMA with a low-voltage dc/rf stage :: http://irclog.whitequark.org/tinyqma
mne has joined #tinyqma
<whitequark>
bofh__: any idea which dielectric to use?
<whitequark>
needs to be easily machinable and at same time with low thermal expansion coefficient
<whitequark>
and rigid
<whitequark>
maaaaybe macor?
<bofh__>
macor would definetly work, honestly any machinable ceramic insulator should work here
<bofh__>
(yes I realise "machinable ceramic" is a bit of an oxymoron)
<whitequark>
wellll I am told alumina is "machinable"
<whitequark>
in the sense that, say, glass is "machinable"
<whitequark>
I've yet to try this bc I don't have spare alumina
<whitequark>
one of the compositions sinters in a commercial ceramic oven, and I have all the necessary oxides for it and a ball mill
<bofh__>
also isn't alumina's thermal expansion coeff a bit on the high end?
<whitequark>
hmm
<whitequark>
well it is less than the other thing i am considering
<whitequark>
which is teflon
<bofh__>
I don't quite have a feel for how low we need it
<whitequark>
which expands so hard you can't machine it to better than .5mm
<whitequark>
solely because of the heat liberated during cutting
<whitequark>
also squishy
<whitequark>
alumina... hm
<whitequark>
5.4
<bofh__>
I thought perhaps aluminum nitride, but that doesn't work well as a dielectric when you get to high-frequency RF
<whitequark>
errr it is machinable?
<whitequark>
sounds like it would be hard as hell
<whitequark>
heck there are tools with AlN coatings
<whitequark>
lol pine wood is 5
<whitequark>
that... actually
<whitequark>
...
<whitequark>
why not
<bofh__>
yeah no, I wouldn't really want to machine AlN
<bofh__>
rofl
<bofh__>
as long as it's completely dry, it would be a suitable dielectric here
<whitequark>
the trouble with it is that it's porous
<whitequark>
but I want to try it solely for the sheer shock value it would have on anyone in the field
<bofh__>
yeah that's basically the only reason I'm considering it
<bofh__>
oh god
<bofh__>
thing is we're under a kV and aren't running particularly high currents, so the dielectric properties aren't really hard to come by
<whitequark>
sure
<whitequark>
practically anything dielectric works here
<whitequark>
I'm more concerned of mechanical properties
<bofh__>
yeah that's basically the deciding factor. I wasn't even aware you *could* machine teflon, iirc it melts at like 300C or thereabouts.
<whitequark>
um, duh
<whitequark>
it softens way earlier
<whitequark>
you don't want it to get over fifty.
<whitequark>
but the thing is, you can machine acrylic
<bofh__>
yeah that's what I thought.
<whitequark>
it's a pain in the ass
<whitequark>
but I did it
<whitequark>
basically just soak it in water
<whitequark>
then you can actually cut chips
<whitequark>
the moment the endmill is not in water, it melts, solidifies in a cylindrical thingy on the endmill, and breaks it
<bofh__>
when you say acrylic, do you mean PMMA, or?
<whitequark>
PMMA.
<bofh__>
wow
<whitequark>
the feeds/speeds for PMMA are basically the same as for aluminium
<whitequark>
people call it "transparent aluminium"
<whitequark>
Al also tends to stick to endmill, though the mechanism is of course different
<whitequark>
also a pain to machine
<whitequark>
fun fact: PMMA and pureish Al have the same Mohs hardness
<bofh__>
I assume just lasercutting it wasn't sufficient for the detail needed?
<whitequark>
I don't have a laser cutter
<whitequark>
but I have a mini mill
<whitequark>
CNC
<whitequark>
20x30x5cm area
<whitequark>
plastics all seem to have expansion coefficients of over 25, and most over 50 even
<whitequark>
that's micron per meter per kelvin
<bofh__>
fun thing with a CNC is if you retrofit it with a 445nM 1W laser diode (cheap as hell on ebay), you can use this to do photoresist etching of PCBs.
<whitequark>
yeah I have considered this
<bofh__>
yeah I'm not really considering any plastic here for exactly that reason
<whitequark>
I currently do photolithography
<whitequark>
using transparencies printed on a laser printer
<whitequark>
the diode thing is slooooow and also doesn't give you as nice resolution unless you focus it somehow
<bofh__>
I've done this, it works well, and you can route traces much more precisely than by doing toner-transfer (which I have also done). More of a pain in the ass to setup, though, and slow.
<whitequark>
toner transfer is shit
<whitequark>
not even considering
<whitequark>
I got photolithography figured out fairly well
<whitequark>
ironically I have trouble etching lately. all my etchants either immediately work shitty or start working shitty after one etch or so
<whitequark>
I suspect my lack of deionized water is the culprit
<whitequark>
I have no clue where to get deionized water. I knew one place which sold it bottled but it no longer operates
<bofh__>
so I just use Na4S2O8 for both prep and trace etchant, saves me from having to get fecl3
<bofh__>
you want it more dilute for the latter
<whitequark>
yeah I figured that out
<whitequark>
but yu want to heat it
<whitequark>
I have nothing to heat it in
<bofh__>
oh so fecl3 etches slow as shit in my experience unless heated
<whitequark>
faster than persulphate ime
<whitequark>
marginally, perhaps
<whitequark>
anyway lately I switched to just making pcbs at board houses, my needs are a tad too complex
<whitequark>
the QMA board will be 4 layer obviously
<bofh__>
yeah this method doesn't scale to multilayer nicely obv
<whitequark>
well i have considered making multilayer
<whitequark>
you can't buy prepregs
<whitequark>
but you can buy .2mm copperclad fiberglass here
<bofh__>
so I've done photolithography but my issue is the boards themselves (the ones with the preapplied photoresist) tend to be more expensive than I'd like
<bofh__>
rofl I keep forgetting Inkscape technically will try to import postscript
<whitequark>
there is a reason i painstakingly compose these lists
<whitequark>
INKSCAPE IS FUCKING AWFUL GARBAGE I HATE INKSCAPE
<bofh__>
it's generated broken svgs every single time I've tried it
<whitequark>
EXACTLY
<whitequark>
IT HAS ONE JBO
<whitequark>
AND IT CAN'T DO IT
<whitequark>
anyway basically any photoresist which has color like that is the same as mine
<whitequark>
approximately
<whitequark>
there are different colored resists
<bofh__>
it's not like it's a hard task, they're both text-based vector image formats and often both support the same scaling/dimensions/colourspaces/etc
<bofh__>
fuck I've regexed a ps to svg before (it was a simple ps)
<whitequark>
lol
<whitequark>
hardcore
<bofh__>
no, hardcore is applying regex on an instruction sequence in a binary and hoping you don't hit anything else with it
<bofh__>
successfully worked 3/5 times so far
<whitequark>
i had a crack for sublime text that consisted of dd piped into sed
<bofh__>
rofl
<bofh__>
also I saw your post on fixing sublime/gtk's xcompose support
<bofh__>
that was pretty cool
<whitequark>
yeah
<whitequark>
i was at the point where pressing compose and not seeing the result just sent me into nine hours of blind rage
<whitequark>
and ... apparently this is what results when that happens
<bofh__>
I'm impressed you managed to do that in 10 hours not having any idea of the internals of libgtk
<whitequark>
I was really angry.
<bofh__>
it feels like you'd need 10 hours to just track down the sequence of calls it uses to do anything
<bofh__>
gtk is kind of a mess. and huge.
<whitequark>
rofl that is mostly what i was doing
<whitequark>
you can't successfully use gtk without going knee deep in its source
<whitequark>
when i ported solvespace to gtk, i spent like half of the time figuring out how exactly something works
<bofh__>
I feel like this is a disturbingly common thing with a lot of opensource software.
<whitequark>
like... filechooser? I don't remember
<bofh__>
the docs are shit and often don't cover a lot of things at all
<whitequark>
in case of gtk, everything in it is shit
<whitequark>
right so alumina can be machined with diamond tools and water coolant
<whitequark>
as expected really
<whitequark>
bofh__: do you know anything about ionizer design?
<whitequark>
usually it's an electron gun and a few accelerating electrodes
<whitequark>
but I don't really know anything about fabricating these
<bofh__>
so the only hard part of that is the electron gun if we need something stronger than what is used in, say, an old CRT
<whitequark>
I'm not quite sure of the geometry mainly
<bofh__>
most QMAs use I've seen use direct-current or inductively-coupled argon plasma ion sources, which are more complicated than just using RF arc sources
<bofh__>
oh you mean for focusing the beam?
<bofh__>
I have a document that describes the geometry relatively well, now if only I can find it.
<whitequark>
indeed that is not ... very complex machining
<bofh__>
a pair of plates might even be sufficient actually
<bofh__>
the beamforming is done mostly via the actual emitted EM fields, so the shape of the emitting elements isn't necessarily going to be complex, and certainly doesn't have to be.