<wpwrak>
yes ! finally killed a nasty bug in fped. figuring out that one took something like a year :)
<qi-bot>
[commit] Werner Almesberger: fped/obj.c (generate_items): handle failure to expand package name gracefully (master) http://qi-hw.com/p/fped/b8f1add
<freemor>
@wpwrak congrats.. know how annoying those hard to track down bugs can be
<Freemor>
@Ayla isn't Enlightenment very pointer dependant (haven't looked at it in years)
<mth>
AwAyla: when cleaning up existing code, you have something usable all the time that gradually becomes better, while if you rewrite from scratch you have nothing usable for a long time, then something usable but lacking features and in the end something good
<mth>
it's very easy to underestimate the amount of time before you reach usability parity with the existing program
<mth>
so by all means write your own menu as a long-term replacement for gmenu2x, but we should maintain gmenu2x until then
<wolfspraul>
100% agree
<larsc>
yep
<larsc>
for the nanonote kernel the existing ingenic drivers allowed us to replace one at a time while still having a more or less running system all the time
<wpwrak>
larsc: in other words, do not cut branch before having taken residence on different branch. given our ancestry, the concept should be intuitive :)
<larsc>
hmpf. stupid. before wondering why the devices does not respond apply power
<wpwrak>
(M1 is done with altium ... unfortunately)
<whitequark>
wpwrak: my one is done in Eagle & TopoR, so the output is just the strict RS48whatever gerber file set
<whitequark>
well, it looks easy to tweak
<whitequark>
*looks like it is
<wpwrak>
prettygerbv should be easy. the drill file conversion may be harder. there's a lot of weird stuff in excellon and i only implement what's actually used
<whitequark>
wpwrak: I've adjusted the file names, and it kind of works
<whitequark>
every view except "front" has its layers' positions hopelessly screwed up
<wpwrak>
heh, that's where the fun begins :)
<wpwrak>
you may also want to check the holes. chances are they're not at the right place (if you drew them at all)
<whitequark>
wpwrak: the holes were misplaced even just after CAM script my PCB manufacturer has provided
<whitequark>
I've adjusted gerber files manually in the gerbv. I hope they don't account for the offset in their own software.
<wpwrak>
(holes bad with script from manufacturer) nice :)
<whitequark>
or, actually I think that everything except holes was not at its place
<wpwrak>
with the kicad process we have now, it's pretty much "make fab" and mail them the files :)
<whitequark>
because when I ran the autorouter (another perversion: it works only with pcad, and a some weird format I never heard of, and an Eagle ULP exports into the latter), it gave me two copper layers in gerber, and they were on par with holes (but moved relative to everything else)
<whitequark>
kicad... I find it rather impenetrable somehow.
<whitequark>
Eagle is quirky and open-source, but I managed to understand how to actually do something useful in it
<whitequark>
*not open-source
<wpwrak>
(kicad) it's not so bad. some things are still weak, but yuo can work around most of them.
<wpwrak>
e.g., don't draw footprints in kicad - use my fped :)
<wpwrak>
yeah, the non-openness of eagle is its big problem. i know that it's extremely popular, but ...
<whitequark>
I was impressed by gerbv, through (it's in geda suite right?), first because it actually worked (not in the way 100s of windows crappy gerber viewers work), and second was now, when I've spotted Scheme code in your script
<wpwrak>
heh ;-)
<whitequark>
(I've got severely schemed recently, to the extent of porting it to Cortex-M3's)
<wpwrak>
gerbv is not as flexible as it may seem. but i like what i was able to do with it. for more advanced things, i may have to write my own gerber renderer, though
<whitequark>
what's I'm trying to do is to show my first PCB which will be actually manufactured at the fab
<wpwrak>
how many layers ?
<whitequark>
two
<whitequark>
I was able to make one-layered boards myself, but it looks like vias are too hard
<wpwrak>
then the prettygerbv process should suit you well. you just need to work out the quirks eagle puts.
<whitequark>
at least, for me
<wpwrak>
(vias) just make a hole and solder wire
<wpwrak>
if you have a cnc mill, you can make very precise holes and use a matching wire that doesn't fall out. easy to solder that way.
<wpwrak>
things get harder if you have big ground planes in addition to this.
<whitequark>
(mill) no, I don't. I only have a manual one, through with a support, but I always miss one or two holes destroying something in the process
<whitequark>
that is, while milling holes for 0.1" pinheads.
<whitequark>
and really thin drill bits which are suitable for vias just break because of miscentered something
<whitequark>
(vias) I've tried that once. it does not work well under TSSOP, at least for me.
<wpwrak>
a cnc mill shouldn't be all too expensive. maybe starting at usd 1500 (incuding taxes and such), you should be able find something decent
<whitequark>
usd 1500 is currently a lot for me
<rjeffries>
wpwrak I have a dumb question about that Panasonic EVQ77 siwith/joystick  I am not clear hwo many i/os it needs, and don't know if they are digital or analog.
<wpwrak>
(vias under tssop) ah, i wouldn't try that either. keep them out in the open
<whitequark>
wpwrak: if I manage to get images for that board somehow, you'll understand why I can't
<wpwrak>
rjeffries: all digital. 5 i/os
<wpwrak>
if you have enough space for tssop, you have enough space for vias not under components ;-)
<rjeffries>
wpwrak thanks. so it is like a D-Pad in a sense. yes?
<wpwrak>
rjeffries: maybe .. dunno what a d-pad is :)
<wpwrak>
it's kinda like the joysticks you got for the vic-20
<rjeffries>
wpwrak yes you do. one of those funking littel pinting devices on many mobile phones used
<whitequark>
a) I can throw out the boards they'll do; b) I don't know how to make my autorouter write the drills file.
<wpwrak>
*smt* is for populating the boards
<wpwrak>
(just look at the latest version, of course. we went through a few iterations, also because they had some change requests)
<wpwrak>
(drill file) don't you re-import the layout into eagle ? the drill file may be called "excellon"
<whitequark>
wpwrak: hm. I've found a tutorial for my autorouter and Eagle, and it mentions nothing about re-importing. so I've just replaced top/bottom copper gerber files with the one autorouter has written
<whitequark>
well, I doubt that import is possible. After all, you need to know which wire belongs to which net, and gerber files does not contain this kind of information at all
<wpwrak>
,aybe that's the way to do it. then you just need to find a way to get your drill file.
<whitequark>
ah yes, I remember something about Excellon. not that I knew what it is at that moment, of course
<wpwrak>
(import) well, depends on what your autorouter provides. there are relatively "fat" interfaces that should have all the information.
<wpwrak>
but ... it may be faster to just route the board manually :)
<whitequark>
it took me around of five hours to study that autorouterit has a really good interface, and the end result is pretty good too (I think)
<wpwrak>
it seems that someone studied their steiner trees well :)
<wpwrak>
some things look a little odd, though. e.g., the route to the pin below "JP1". or the little arc near "C1".
<wpwrak>
als the route between R6 to R7 is unnecessarily close to the pads.
<wpwrak>
either the router or eagle also has problems with filled areas. e.g., C3 and C2 has pads that are parts of the ground area, without thermal relief. they'll be fun to hand-solder :)
<wpwrak>
in general, 100% autorouting is a bad idea. not only do most autorouters get things wrong, but there's also a lot of implicit information that's not properly codified. e.g., you'd keep areas sensitive to interference apart, but unless you've told your autorouter explicitly about that, it may not
<wpwrak>
or maybe it does on all the runs before the last one, for the factory :)
<whitequark>
wpwrak: (quirks) yes, it has some, and I've fixed a few. looks like not enough of them
<whitequark>
wpwrak: (100% autorouting) I've set up several (three) keepout areas and so on, without them the result was much worse. It wasn't hard, through
<whitequark>
(filled areas) that's the gerber export. the area should have been hatched, but somehow it exports to gerber as solid
<wpwrak>
hmm, hatched became solid ? that's nasty ...
<whitequark>
wpwrak: yeah. and google does not reveal anything on that topic, either.
<wpwrak>
i wonder if kicad could do hatched. never tried that. there's a gazillion options in that dialog, though ...
<wpwrak>
of course, in practical terms, i don't think hatched makes all that much sense. what you really want are "moats" without copper around pads, and the traces connecting them to the copper zone. that way, the copper zone doesn't suck away all the heat, but you still get all the shielding benefits.
<wpwrak>
DocScrutinizer: maybe you know some uses for hatched copper fill ?
<whitequark>
wpwrak: hm, there's such an option, I'll try it out
<whitequark>
by the way, the drill file exported by the autorouter is displayed by gerbv, but it shows several errors about junk in place of tool definition
<whitequark>
and, worse, it, by some reason, does not want to set offset for that layer
<wpwrak>
is the drill file in excellon or in gerber ?
<whitequark>
ah, sorry, the latter is I'm being stupid and using mms where it expects mils
<whitequark>
the global switch at the bottom of workspace is quite confusing in terms of UX
<wpwrak>
the invisible hand of the us-american royalists ;-)
<whitequark>
well, often I prefer mils, too, because a lot of stuff is inch-based, but the autorouter only understands offsets in mm
<wpwrak>
pretty much all the modern stuff is metric. once the chinese buy digi-key, the last few imperial parts will vanish, too ;-) but yes, mils can stick at times. e.g., i wouldn't know my trace widths and clearances in mm without using "units".
<whitequark>
wpwrak: what about DIP? SO-1.27? pinheads? SMD passive component sizes?
<wpwrak>
some evil folks even metricise traditionally imperial parts. such as 100 mil becoming 2.50 mm. and then the really smart ones adjust the drill holes so both variants still somehow fit. sometimes ;-)
<wpwrak>
pretty everything smt below SOIC is metric or has metric equivalents. alas, not all of them happy choices. e.g., 0603 and 0402 exists in both systems.
<whitequark>
yes, that's quite confusing
<wpwrak>
DIP ... well yes, still convenient for opto-couplers. and then you get all those 0.1" headers. or, as i said, their metric counterparts
<whitequark>
the pseudo-metric headers will bite you at more than 10 rows, I suppose
<whitequark>
(or lines?)
<wpwrak>
i think they went a bit too far with metricising the passive smt components. they're more like type codes anyway. the precise size of the chip scarcely matters.
<wpwrak>
i'd start to worry around the 4th pin :)
<wpwrak>
of course, you can bend them a little, ...
<whitequark>
was just going to say the same.
<wpwrak>
... and then, when you plug in the connector, bend them back a little, ... ;-)
<whitequark>
(digikey) darn. I've ordered three buck converters, three LVDS serializers and a bunch of passive components from them. and now they're asking about the country in which I'll "assemble that parts"
<whitequark>
are all US suppliers equally idiotic/paranoid?
<wpwrak>
i think that's just gov.us nonsense
<wpwrak>
i doubt anyone actually reads what you write there ;-)
<whitequark>
no, they've actually sent me a email from their Intl Sales Dept. you know, by a real human
<whitequark>
he has signed it as "Bobbie" (quoted)
<whitequark>
1.) Who is the buyer of these parts? 2.) You list the parts for individual use, but we are shipping to a company, please clarify. 3.) Are these parts being assembled in the states or in Russia?
<wpwrak>
wow ;-) maybe answer that you'll assemble in the principality of sealand, but you didn't find it in their menu. that'll give them something to ponder ;-)
<whitequark>
hm, company.
<whitequark>
I'm using a mail forwarding service (their fee for delivery to russia is $120. a hundred and twenty, that is), so I've unchecked the "ship to residental" checkbox