IRC log for #qi-hardware on 20150316

00:14.50qi-bot[commit] Werner Almesberger: cameo/: new command write_gerber, for Gerber output (master) http://qi-hw.com/p/cae-tools/c7ff028
00:14.51qi-bot[commit] Werner Almesberger: cameo/lang.l: require all commands followed by file name to end with whitespace (master) http://qi-hw.com/p/cae-tools/096b91e
00:19.58whitequarkhow do these people come to the channel?..
00:56.49wpwrakirc --list-channels | grep -i hardware
01:56.54DocScrutinizer05did I miss something?
01:57.18DocScrutinizer05ooh, "whice psu is better"
01:57.45DocScrutinizer05exactly what wpwrak said
01:58.32DocScrutinizer05wpwrak: btw kudos for the PCB shaping :-)
01:59.17DocScrutinizer05you already mentioned the root question "can you use it?"
02:00.38DocScrutinizer05rather " like to know if this is useful for you"
02:01.29*** join/#qi-hardware atommann (~atommann@58.251.2.94)
02:12.57DocScrutinizer05seems DXF is a format eagle could import
02:15.18wpwraki hope he can handle gerber :) i don't know dxf. i could offer gnuplot, though. that's the easiest of them all :)
02:16.37whitequarkdxf, not really
02:16.47whitequarkimporting into eagle is a pain overall
02:16.50DocScrutinizer05anyway I exported http://wstaw.org/m/2015/03/16/plasma-desktophp1895.png to dxf and got http://paste.opensuse.org/22000300
02:17.21DocScrutinizer05whitequark: exactly my point
02:17.52DocScrutinizer05whitequark: how would you import a PCB shape to eagle?
02:19.18DocScrutinizer05tbh I don't even understand how the shape is defined inside eagle
02:20.00*** join/#qi-hardware Haswell (~broaderba@unaffiliated/haswell)
02:20.15DocScrutinizer05I mean, I can draw wires and arcs, but how's that becoming the outer edges of the allowed PCB area for eagle?
02:21.10DocScrutinizer05simply by traces not being allowed to cross wires, arcs etc?
02:23.22DocScrutinizer05or even simpler by traces only being allowed on background color? then we could import bmp as well prolly
02:24.33whitequarkum, eagle does not have an allowed PCB area.
02:24.57whitequarkwhen you export it or process it into gerber, it just uses the widest extent of anything on the board
02:25.38whitequarkthere's no "background color" and the traces can intersect if you so desire
02:26.16whitequarkif you mean the autorouter, i think you need to define corresponding keepout areas to make it stay clear.
02:26.16DocScrutinizer05from eagle "help": >> ==Platinenumrisse und Plazierung festlegen==     Gegebenenfalls die Leerplatine in Größe und Form verändern (MOVE, SPLIT).<<
02:26.25wpwrakthat sounds quite normal. e.g., in kicad, you have a specific layer for the board outline. eagle must have something very similar.
02:26.27whitequarki don't speak german
02:26.44wpwrakwhitequark: weren't you learning that weird language ? :)
02:26.58whitequarki kinda skipped on it
02:27.04whitequarkanyway, used GT
02:27.10DocScrutinizer05wpwrak: I guess "layer" is the key word
02:27.16whitequarkeagle does have an outline layer and it has a rectangle by default there
02:27.22whitequarkthe layer doesn't affect anything at all
02:27.36whitequarksome board houses like oshpark use it internally
02:27.53whitequarki don't think eagle itself cares
02:27.54wpwrakwhitequark: you'll be sorry for it when those alternate histories come true :) well, or perhaps for skipping french. can't be so sure about which one will win.
02:28.05DocScrutinizer05well, the autorouter must have an idea of where the PCB ends, no?
02:28.19wpwrakthere must be a cut, too .. :)
02:28.27whitequarkhm, not sure
02:28.54whitequarkah, yes, the help says autorouter also respects
02:28.55whitequarkthat
02:29.07wpwrakthe pcb outline is the only thing that defines the pcb outline. no traces, no pads, etc. so it must be there, somehow :)
02:29.12whitequarki don't use autorouter, as it's evident
02:29.22DocScrutinizer05so it's based on "pixels" in PCB outline layer?
02:29.26whitequarklines
02:29.28whitequarkit's all vector
02:29.43wpwrakin kicad, that's just a drawing. lines and arcs
02:29.46DocScrutinizer05ok
02:30.03DocScrutinizer05yes, like in eagle
02:30.37DocScrutinizer05I picked the lines + arcs idea for my example of above from another board that works
02:31.55DocScrutinizer05eagle help is for the bin
02:32.07DocScrutinizer05:-S
02:32.23wpwrakif all else fails, i just look things up in the source ;-)
02:32.51wpwrakkinda the ultimate "help"
02:36.28*** join/#qi-hardware archang (~archang@111.198.29.54)
02:37.06DocScrutinizer05layer 20 "Dimension"
02:40.20DocScrutinizer05gotta love eagle: http://wstaw.org/m/2015/03/16/plasma-desktopjj1895.png  window switching, it fails on refresh/redraw
02:40.26DocScrutinizer05sometimes
02:41.21whitequarkwait
02:41.28whitequarkis "Golden Delicious" a kind of apples?
02:41.34DocScrutinizer05yes
02:41.40whitequarklol
02:41.52DocScrutinizer05actually green apples
02:42.00DocScrutinizer05THE apple
02:42.13whitequarkthey look yellow
02:42.24DocScrutinizer05yellow-greenish
02:42.25whitequark"The Golden Delicious is a cultivar of apple with a yellow color"
02:42.28whitequarknot just me
02:42.52DocScrutinizer05they're not red as usual apples
02:43.18DocScrutinizer05guess that's why they#re called/named "golden"
02:44.08DocScrutinizer05http://en.wikipedia.org/wiki/Golden_Delicious has good pictures
02:47.04DocScrutinizer05anyway
02:47.08DocScrutinizer05wpwrak: http://wstaw.org/m/2015/03/16/plasma-desktopPh1895.png
02:48.35DocScrutinizer05wpwrak: we need to provide a way to actually import the stuff into eagle, otherwise it will again be "please send pencil sketch on napkin"
02:54.42*** join/#qi-hardware kristianpaul (~kristianp@181.137.226.143)
02:54.42*** join/#qi-hardware kristianpaul (~kristianp@unaffiliated/kristianpaul)
02:55.22DocScrutinizer05/whois stallmanu
02:55.41DocScrutinizer05err
02:57.52*** join/#qi-hardware kristianpaul (~kristianp@181.137.226.143)
02:57.52*** join/#qi-hardware kristianpaul (~kristianp@unaffiliated/kristianpaul)
03:06.51*** join/#qi-hardware atommann (~atommann@58.251.2.94)
03:08.19*** join/#qi-hardware kristianpaul (~kristianp@unaffiliated/kristianpaul)
03:10.16*** join/#qi-hardware kristianpaul (~kristianp@unaffiliated/kristianpaul)
03:14.02DocScrutinizer05wpwrak: maybe actually just this: http://wstaw.org/m/2015/03/16/plasma-desktopOQ1895.png ;  detail: http://paste.opensuse.org/96310635 ; full file: http://paste.opensuse.org/50264494
03:21.03DocScrutinizer05wpwrak: added a arc:  <wire x1="82.55" y1="43.18" x2="62.23" y2="22.86" width="0.4064" layer="20" curve="-266.304779"/>   http://wstaw.org/m/2015/03/16/plasma-desktopnt1895.png
03:21.47DocScrutinizer05I guess that should have been -270
03:22.04DocScrutinizer05which is actually how I drawn it
03:22.10whitequarknonono don't modify xml directly
03:22.19whitequarkwhat you should do is to make a "script"
03:22.34whitequarkwhich is what's accepted in the bar above the drawing area and also in .scr files
03:22.36DocScrutinizer05whywhywhy when we have no better API ?  :-)
03:22.51whitequarkwe do, that's my point
03:23.30DocScrutinizer05yeah, a script is prolly fine, as long as I can convince wpwrak to learn about the available commands and their syntax
03:23.49whitequarkthe embedded help describes the scripting commansd pretty well
03:23.59whitequarki worked extensively with both ulp and scr languages of eagle
03:24.02whitequarkthey're surprisingly nonhorrible
03:24.04DocScrutinizer05though non-comprehensive
03:25.22DocScrutinizer05I completely agree with you, but I'd be willing to "convert" files even manually, for this one-shot
03:25.46whitequarkit's probably faster to generate the scr
03:25.51whitequarkfor one, you don't have to parse and serialize xml
03:26.22DocScrutinizer05lemme try a lil gimmick...
03:26.42whitequark... don't tell me you're going to inject that as a string using sed
03:29.24DocScrutinizer05http://paste.opensuse.org/43156895
03:29.42DocScrutinizer05hehe I already thought about that, but not yet
03:30.02whitequarkugh you're the worst
03:30.24DocScrutinizer05but what motivated me for http://paste.opensuse.org/43156895 is pretty close to that and also somewhat related to sed
03:31.43DocScrutinizer05tbh the xml looks astonishingly similar to what I think the wire command syntax will be
03:33.04DocScrutinizer05hmm
03:33.05DocScrutinizer05WIRE ['signal_name'] [width] [ROUND | FLAT] • [curve | @radius] •..
03:33.50DocScrutinizer05• means coords iirc
03:33.55whitequarkyes
03:33.58whitequark(1 1)
03:34.01whitequarkof this form
03:34.29whitequarksometimes it accepts input in really strange form to make it uniformly fit • format
03:34.51whitequarkoh, i just remembered how exactly
03:35.01whitequarkusually • means (x y)
03:35.20whitequarkbut sometimes • means (major minor) where major is the major grid spacing and minor is the minor (alt) one
03:35.30DocScrutinizer05ugh
03:35.42whitequarkit is fucking beyond me who thought it is a good idea to reuse that syntax
03:35.47whitequarkbut at least i'll save you some time
03:35.58DocScrutinizer05:-)
03:38.42DocScrutinizer05btw pre eagle6.x .brd files looked way more nasty ;-)
03:38.52DocScrutinizer05binary crap
03:41.42whitequarkyeah
03:42.17DocScrutinizer05anyway I guess it's rather trivial to convert whatever vector format drawing into eagle WIRE
03:42.51whitequarknot always
03:42.56whitequarkif it uses arcs and lines, yes
03:43.00whitequarkif it uses B-splines, then no
03:43.00DocScrutinizer05yes
03:43.10DocScrutinizer05that's what I assumed
03:43.48DocScrutinizer05I hope wpwrak didn't use B-splines to describe teh shape of a 2.5D-scan
03:44.15DocScrutinizer05;-)
03:44.16whitequarkB-splines are nice
03:44.29whitequarkrepresenting circles and ellipses exactly
03:45.03whitequarkespecially NURBS surfaces. otherwise you're left with triangle crap like OpenSCAD
03:45.18DocScrutinizer05yes, maybe. But I doubt I'll find splines in any vectoriced bmp
03:46.15DocScrutinizer05I even doubt there will be arcs in there
03:46.20whitequarkhrm
03:46.42whitequarkwell, I would probably use splines to avoid the "pixelation" effect
03:47.01whitequarkit's not like we're living inside minecraft
03:48.49DocScrutinizer05for the problem at hand: convert http://maemo.cloud-7.de/Gallery-N900-exploded/platine_01.jpeg shape into http://wstaw.org/m/2015/03/16/plasma-desktopJb1895.png
03:49.36DocScrutinizer05wpwrak alrady did all the heavy lifting
03:49.40whitequarkdo you even need bitmap tracing?
03:49.43whitequarkI would use calipers
03:50.05DocScrutinizer05nah, the shape is pretty irregular
03:50.26whitequarkhm, maybe
03:50.57DocScrutinizer05http://neo900.org/stuff/werner/tmp/pcb-outline-visual-comp.jpg
03:52.16DocScrutinizer05kudos to werner for that
03:52.26DocScrutinizer05awesome work
03:53.47whitequarkwhat i would do is to use the edge detector in GIMP and then trace the resulting contour in inkscape
03:54.03DocScrutinizer05also see latest commits at http://neo900.org/git/?p=misc
03:54.09whitequarkthe former works surprisingly well, just used it today
03:56.26DocScrutinizer05yes, that's what I meant by "vectorized bmp"
03:56.49whitequarkinkscape would produce a B-spline though
03:56.51DocScrutinizer05anyway we're already there, courtesy wpwrak
03:56.52whitequarksince that is what svg uses
03:56.59DocScrutinizer05OUCH!
03:58.33DocScrutinizer05well, cool for inkscape and SVG, not cool for eagle
04:02.49DocScrutinizer05I'm pretty sure wpwrak will only shrug about converting whatever thin red line in whatever format he got now, into a loooong sequence of adjacent (staight) WIRE commands aproximating the red line's shape for arbitrary precision
04:03.47DocScrutinizer05I bet it's not even a question whether or not the length of each vector will be maximized
04:06.57DocScrutinizer05actually the latter is extremely simple: when (x,y) == (x[-1],y[-1]) * n, simply replace old (x[-1],y[-1]) by new (x,y)
04:07.49DocScrutinizer05err sorry, of course polar coords apply here
04:09.28DocScrutinizer05so when (x,y) - (x[-1],y[-1]) == ( (x[-1],y[-1]) - (x[-2],y[-2]) ) * n, simply replace old (x[-1],y[-1]) by new (x,y)
04:13.43DocScrutinizer05whitequark: ((scripts)) actually that's *exactly* what I suggested as "interchangeable eagle format" for Neo900 project
04:14.22DocScrutinizer05whitequark: you know there are ULPs that can export whole eagle projects as a single huge script
04:14.31whitequarkyes
04:16.28DocScrutinizer05alas these ULP are a tad... bitrotten, would need some revisiting to add a few lacking (new?) command syntax details of eagle6
04:21.31DocScrutinizer05and *really* great would be to directly access and read out the undo/redo buffer - I could bet on it containing complete plaintext commands. What a nice session log as well as obviously also a script
04:21.57whitequarkunlikely plaintext
04:22.04DocScrutinizer05hmm, dunno
04:22.04whitequarksince you can't easily revert them
04:22.13whitequarknot much point in reparsing and such
04:22.20DocScrutinizer05aah right
04:22.22DocScrutinizer05nm
04:22.22whitequarknote that solvespace does something very similar
04:22.36whitequarkits savefiles are "forth" programs that recreate the environment from scratch
04:23.19whitequarkdoesn't quite map to an undo/redo buffer--it just copies the entire savefile (in-memory) when you do anything
04:23.25whitequarksince it's small
04:23.29whitequarkbut it's still neat
04:23.49DocScrutinizer05well, I'm just musing about a theoretical approach: save to .brd/.sh after each single command, do diffs
04:24.03whitequarkthat'd work. but what for?
04:24.05DocScrutinizer05.sch
04:24.17DocScrutinizer05for "patches"
04:25.08DocScrutinizer05solving the "send a pencil sketch on a napkin please" problem
04:28.57DocScrutinizer05wpwrak: OOOPS we need BOB too :-o
04:29.37DocScrutinizer05not to same precision, but anyway we need some 3rd PCB shape
04:32.03*** join/#qi-hardware archang (~archang@111.198.29.53)
04:41.54*** join/#qi-hardware archang (~archang@111.198.29.54)
04:56.21*** join/#qi-hardware xiangfu (~xiangfu@111.198.29.54)
05:22.11*** join/#qi-hardware atommann (~atommann@58.251.2.94)
06:15.44*** join/#qi-hardware porchao (~quassel@218.231.187.90.eo.eaccess.ne.jp)
06:32.03*** join/#qi-hardware xiangfu (~xiangfu@111.198.29.53)
08:57.45*** join/#qi-hardware pcercuei (~paul@137.71.226.54)
09:31.35*** join/#qi-hardware xiangfu (~xiangfu@111.198.29.54)
09:41.33*** join/#qi-hardware xiangfu (~xiangfu@111.198.29.54)
09:47.09*** join/#qi-hardware xiangfu (~xiangfu@111.198.29.53)
09:53.53*** join/#qi-hardware xiangfu (~xiangfu@111.198.29.53)
10:24.44*** join/#qi-hardware wej (~j@2001:1af8:4010:a08c:1::6e)
10:33.16*** join/#qi-hardware fengling (~fengling@2002:6fc6:1d36:0:6479:8604:932:e140)
11:34.19*** join/#qi-hardware wej (~j@2001:1af8:4010:a08c:1::6e)
12:09.52*** join/#qi-hardware wej (~j@2001:1af8:4010:a08c:1::6e)
12:42.26*** join/#qi-hardware wej (~j@2001:1af8:4010:a08c:1::6e)
12:55.11wpwrakDocScrutinizer05: (b-splines) no, it's all list line segments. some 10+ k of them ;-) (either connected or unconnected, depending on where in the process you are)
13:18.34*** join/#qi-hardware Nik05 (~Nik05@unaffiliated/nik05)
13:49.00*** join/#qi-hardware wej (~j@2001:1af8:4010:a08c:1::6e)
13:59.51DocScrutinizer0510+k ? sounds like it could use some "smoothing"
14:01.56DocScrutinizer05what do you think about providing a "script" that eagle could readily run?
14:03.05DocScrutinizer05(I guess converting DXF to a script is what needs to get done anyway)
14:04.39wpwrakdxf is pure horror. my "native" format is gnuplot. plot "foo.gp" with lines
14:06.09wpwrakbut i suspect that nik would not want to use a generated outline directly anyway, but rather adjust his existing design with it. also because some features depend on component shapes, which in some cases will be different anyway
14:06.42wpwrakalso, he may have item at a higher layer of abstracion there. arcs, etc.
14:06.50wpwrakand anyway, what is this doing on #qi-hw ? :)
14:07.37DocScrutinizer05there are no component dependant shapes. And I'd prefer the thing getting defined without manual tweaking
14:08.07DocScrutinizer05(qi) I discussed stuff with whitequark
14:09.26DocScrutinizer05when the DXF-whatever gets used to do manual shape edits then we have exactly same unfortunate situation we already got
14:10.35DocScrutinizer05neglecting details since they are "too complicated" and supposedly "not needed" and adjusting dimensions to own expectations
14:11.27DocScrutinizer05I don't care when instead of arcs a 3000 straight vectors approximate a curve
14:12.35*** join/#qi-hardware sb0 (~lekernel@217.9.101.140)
14:13.02DocScrutinizer05for the approximation I _know_ about how exact it is. Fitting an arc manually to match whatever mask/template will rsult in arbitrary errors again
14:13.52DocScrutinizer05prolly I gotta do it myself
14:14.10DocScrutinizer05convert to a script and import to eagle, that is
14:26.52whitequarkyou can fit an arc to the same error as those 3000 straight segments
14:26.58whitequarkand it's not going to make eagle grind to a halt
14:37.55wpwrakmanual fitting is often remarkably accurate, especially if you have zoom :) also, you can deliberately introduce small deviations, e.g., to fit other patterns you need to consider (grid or such)
14:46.42whitequarkwpwrak: isn't dxf a fairly simple text based format?
14:47.02whitequarkor was that dwg?
14:47.19whitequarki looked at export code for some of those and it was pretty trivial
14:47.22wpwraktext-based but rather chaotic. not in the xml sense but rather in the "too many punch cards" sense
14:48.13wpwrakand autocad were busy augmenting it. so the first step would be to find out which of those hundreds of possible commands you really need
14:48.19whitequarkah, right
17:09.02*** join/#qi-hardware wolfspraul (~wolfsprau@brln-d9bada21.pool.mediaWays.net)
17:22.47DocScrutinizer51hi wolfspraul :)
17:37.29*** join/#qi-hardware paul_ (~paul@bcdd3b18.skybroadband.com)
19:25.05*** join/#qi-hardware pcercuei (~pcercuei@2a02:810d:1b40:1118::a)
19:30.19*** join/#qi-hardware wolfspraul (~wolfsprau@brln-d9bada21.pool.mediaWays.net)
20:29.23*** join/#qi-hardware pcercuei (~pcercuei@2a02:810d:1b40:1118:1af4:6aff:fec5:1c8f)
22:34.09*** join/#qi-hardware wej (~j@2001:1af8:4010:a08c:1::6e)

Generated by irclog2html.pl Modified by Tim Riker to work with infobot.