<jow_laptop>
kyak: yes it would, but give me some time
<jow_laptop>
I already have far too much stuff in the queue
<jow_laptop>
kyak: oh btw, I'll commit an mtd-utils update in a minute that will replace underscores in the packagenames with dashes, you'll need to update the Qi default config I assume
<jow_laptop>
e.g.  mtd-utils-flash_eraseall -> mtd-utils-flash-eraseall
<jow_laptop>
opkg does not like underscores there
<kyak>
jow_laptop: thanks for heads up, will do so
<kyak>
jow_laptop: btw, there is a problem with dejavu-font-ttf package
<kyak>
it had been splitted in several packages for each font, and each of these packages (like dejavu-fonts-ttf-DejaVuSans) is made dependant on dejavu-fonts-ttf package; but there is no such package as dejavu-fonts-ttf
<kyak>
so when someone tries to install these fonts later via opkg, it has dependency issues
<kyak>
probably the same problem exists for mtd-utils...
<kyak>
ah, perhaps not anymore after your latest commit..
<jow_laptop>
kyak: ah yes, will fix that one too
<jow_laptop>
kyak: should be fixed as well now
<jow_laptop>
the problem is that mirko always forgets to define an install section, therfore no physical .ipk is built
<kyak>
heh :)
<kyak>
ok, thanks
<kyak>
btw i assumed that physical ipk would build (with nothing but CONTROL files inside) even if there is no install section, as long as there is $(eval $(call BuildPackage...
<wolfspraul>
oops, back again
<wolfspraul>
maybe someone is working on it...
<wolfspraul>
I didn't do anything
<wpwrak>
that's what they all say ;-)
<wolfspraul>
but strange, it left with 'operation timed out'
<wolfspraul>
something strange, but it seems recovered by itself
<wolfspraul>
wpwrak: in tuxbrain's atben/atusb run, did the rf testing include rf performance testing too?
<wolfspraul>
I mean in terms of speed or antenna reach
<wolfspraul>
and if it did, did you see much differences between boards?
<wpwrak>
it included signal strength measurements, relative to a reference device (although i think the latter was actually from the production run - but we verified against one of the prototypes after)
<wpwrak>
some differences, but not a lot. i also don't know how stable the test environment itself was, with respect to obstacles, position/rotation, interference, etc.
<wolfspraul>
ok but it sounds like there weren't any clues that setting up a more rigid/restricted test environment for RF performance would be worth doing in the next run
<wpwrak>
plus there were two (one for each direction) unidirectional transfer tests, to check the frame loss
<wpwrak>
so far no clues that anything escaped. but the numbers aren't very large, so maybe we'll still find something
<wpwrak>
the atusb with a bad solder joint on the balun was properly identified by the spectrum test
<wolfspraul>
usually you can assume that if you don't see something in a run of 250, you can safely scale up
<wpwrak>
i mean the boards that went to customers and are being used there
<wolfspraul>
that's because you will always find whatever new things as you scale, and random/rare/unexplained things, but they may not be worth the time to investigate further anyway
<wpwrak>
so we don't know yet if the tests found all the problems
<wolfspraul>
I think the fact that those 130 showed so few problems is a good sign
<wolfspraul>
that's a solid number
<wpwrak>
but they found at least that one thing that wasn't just a catastrophic system-wide failure
<wolfspraul>
with 10 or 20 you may overlook a serious manufacturing problem, but with 130 it's much less likely, and like I said, with 250 you are assumed to be (economically) safe
<wpwrak>
yes, i'd say the smt went quite well. the rotated chip and the missing resistor are clear smt mistakes. perhaps need to see if we can improve QC there.
<wolfspraul>
that's normal, can happen
<wpwrak>
there could also be additional SMT mistakes the tests wouldn't find. e.g., absence of a varistor. we could test for it, but that would need a fairly complex test rig
<wolfspraul>
aoi should catch that
<wolfspraul>
'should' as it is most likely not used on such a small run and small smt shop
<wpwrak>
yeah, it's more a case for optical inspection than anything else
<wpwrak>
the balun with the one bad solder joint is actually one of the bits i felt the least confident about. the other was the usb connector, where i had even more doubts. but that one went perfectly, it seems.
<wpwrak>
(balun) the problem there is that the footprints recommended by the manufacturers barely touch the contacts. so i already extended them a bit beyond the most generous interpretation.
<wpwrak>
but i guess i can't complain with only 1/250 showing problems. there could be more with invisible problems, because there are several ground pins, and you can lose at least one of them without significant loss of function.
<wpwrak>
i have a bug in the silk screen of some components, including all the 0402. the outlines need to be pushed out a little further. or omitted. didn't cause any problems this time, but worth improving for the future.
<wolfspraul>
well sure, that all sounds reasonable. those things and improvements will never stop though.
<wpwrak>
and my smallest silkscreen text is at the limits of the printing process of the PCB house, making some things hard to read. again, nothing horrible, and probably not worth fixing on atben/atusb, but good to remember for the future.
<wpwrak>
yeah, there's always one more thing ;-))
<wolfspraul>
so once a problem is rare enough, it may not be worth economically to even track it down, because such rare problems will pop up forever
<wolfspraul>
it's manufacturing, so every piece is unique
<wolfspraul>
there is no point to hope for the perfect binary world one day, it will not come
<wolfspraul>
so you have to make economic decisions, not technical ones
<wpwrak>
we're already amazingly close to that perfect world, if you think of it :)
<wpwrak>
"publish to create prior art" isn't exactly a new concept ;-)
<wpwrak>
and i think it's a good one, yet one also has to understand its limitations
<wpwrak>
e.g., a patent may still be granted no matter how widely you disseminate your prior art. and it can still be upheld for a number of reasons, including your publication not being "visible" enough
<kristianpaul>
(isn't exactly a new concept) thats truth. maye be re-enforce it is what is worth
<kristianpaul>
(not being "visible" enough) yeah, i was afraid you think that as well.. i but i wanted a second opinion :-)
<wpwrak>
so, for example, if we disseminate some novel idea on the qi-hw list, would that count ? or in the Buga Journal of Applied Sciences, written in the local indigenous dialect ?
<wpwrak>
or maybe even in spanish - would that be enough to invalidate a US patent ?
<kristianpaul>
yeah...
<wpwrak>
so the problem with prior art is that it's no the existence of the idea that counts but the realistic possibility of knowing of it
<kristianpaul>
visibillity usually will mean mainline meadia, wich very good filtering what is not intented to be so public when is convient
<kristianpaul>
or may be university media is enought?..
<wpwrak>
technical news of the university of buga ?
<wpwrak>
or do you mean conference papers ?
<kristianpaul>
yeah, thats the point :-)
<wpwrak>
there's also the tricky question of whether online publications count or if it has to be printed
<wpwrak>
and if online publications count, how obscure can they be ? qi-hw planet ? kristianpaul's blog ? qi-bot's irc channel log ?
<wpwrak>
werner's blog on neo-existentialism and herbal medicine where by sheer accident something technical is dropped ?
<wpwrak>
but i think bringing those ideas out into the open is a good first step. better yet if they spread a little bit to other places than just the qi-hw list.
<wpwrak>
and also judges know about google. and google reads and indexes everything, including our hypothetical blog on herb lore.
<wpwrak>
whoa. darkness falls ... at 3 pm. that'll be a heck of a rain