<DocScrutinizer>
aah gta04, yeah. That'S about the whole story
<wpwrak>
gta04 .. "your baby" :)
<DocScrutinizer>
yeah
<wpwrak>
or shall i say "your love child" ? ;-)
<DocScrutinizer>
hmm
<DocScrutinizer>
iirc you at least didn't feel much love for it :-)
<kristianpaul>
Who loves arm around this channel?.. :p
<wpwrak>
DocScrutinizer: let's say that i smelled the rat :)
<DocScrutinizer>
yup
<DocScrutinizer>
anyway, I guess kristianpaul referred to dash
<kristianpaul>
i have news ! seems repeated data is normal after all, i'm just reading short intervals from buffer address and i'm getting repeated data in samples of 128, 256 and  512 bits
<DocScrutinizer>
it's neither non-deployed nor super alpha, but it clearly was a navi
<kristianpaul>
s/arm/arm processors or SoC
<kristianpaul>
nv about arm, i'm kind of sleep now
<kristianpaul>
i need my tea but i dint go shopping yet
<qi-bot>
[commit] Werner Almesberger: usrp/plscan: comment out yrange (the experiments now cover all sorts of ranges) http://qi-hw.com/p/ben-wpan/9c2f37b
<qi-bot>
[commit] Werner Almesberger: atusb/cam/mkmk: update for on-going work; showcase use of "rotate" http://qi-hw.com/p/ben-wpan/155dfba
<kristianpaul>
hey, gnu coreutils is good source of code sample for implementing a command line parameters :-)
<kristianpaul>
zzZ
<bartbes>
I wonder
<bartbes>
does Jane Andreas ever drop by here?
<zrafa>
bartbes: maybe he is the gforth guy? ;-))
<bartbes>
hahaha
<bartbes>
no that was David Kuehling iirc
<zrafa>
two faces man :)
<bartbes>
it's just annoying that I know their (supposedly) real names
<bartbes>
but not who they are
<wolfspraul>
bartbes: no I've never seen Jane here
<wpwrak>
bartbes: welcome to the age of networking, where you communicate with everyone, yet have never seen anyone's face :)
<wpwrak>
yizhang: wheee, toys are coming ! good news, thanks !! :)
<yizhang>
wpwrak: you are welcome.
<yizhang>
wpwrak: thanks a lot
<yizhang>
i'm looking forward to see your great results
<wpwrak>
wolfspraul: the wpan experimental boards will need some kind of "case". i guess i real case is unlikely at the time, so what shall be the alternative ? "naked" and hope for the best ? a bit of acrylic / silicone / PUR spray ? (used for conformal coating)
<wpwrak>
wolfspraul: dip in a jar of any of the above ? other ? what does the average ACME do when they need a "case" but don't have time to make one ?
<wpwrak>
yizhang: hehe, me too :) and let's hope they are as you say :)
<wpwrak>
wolfspraul: btw, sharism.cc really ought to be de-museified. maybe something for your return flight :) how about just replacing all the "content that should be updated from time to time" (but isn't) with links to where the real stuff is ?
<kristianpaul>
wpwrak: toys from digikey?, wow so fast !
<wpwrak>
kristianpaul: no, these are toys from sharism (need more bens to experimenting - i have a tendency to leave them in various states of dismantlement and immobility that makes it hard to use them for more than one kind of experiment)
<wpwrak>
kristianpaul: the digi-key toys should already be somewhere over the southern US :)
<kristianpaul>
ah i see, i was about to think that think digikey is getting _fast_ to you :-)
<wpwrak>
kristianpaul: they're much faster than you thought :) they shipped something like one hour after i ordered. i think that's a new record.
<wpwrak>
alas, at the end, i'll lose a day between customs and fedex. customs, by delaying things beyond the fedex cutoff time for that day, and fedex by having a relatively early cutoff time (something around 10 am for clearing customs)
<wolfspraul>
wpwrak: [sharism.cc] yes of course I know.
<wolfspraul>
it will need a few more weeks, but I'm getting there
<wpwrak>
wolfspraul: sorry for the nagging ... it's just that everytime i look at it, something at the back of my head starts screaming :)
<wolfspraul>
oh sure
<wolfspraul>
thanks for the reminder
<wolfspraul>
about the case, that's interesting
<wolfspraul>
acrylic/silicone/pur
<wpwrak>
what's "the chinese way" if it has to be quick and dirty ?
<wolfspraul>
hard to say. probably reuse a USB stick case.
<wpwrak>
heh :)
<wolfspraul>
but that will be butt ugly
<wpwrak>
pity that the board is a bit too large for that
<wpwrak>
too wide and too long actually. well, not too long for some of the older ones.
<wolfspraul>
why don't we present it to roh ? :-)
<wolfspraul>
I'm willing to pay a (little) money if he starts experimenting.
<wolfspraul>
you want the case for the 8:10 card or the usb card?
<wpwrak>
both will need a "case"
<bartbes>
hey, does anyone know how to release the fb in case the program which grabbed it crashed?
<wpwrak>
bartbes: are you familiar with the wrapper-cleaner approach ? start process that ignores all signals it can ignore then forks the real thing. when the real thing exits, clean up.
<bartbes>
wpwrak: that..
<bartbes>
doesn't sound too bad
<bartbes>
I can do that
<bartbes>
a signal handler
<bartbes>
thanks for the idea
<wpwrak>
(case) if something silicone-based works, that could be pretty easy to make in small quantities. silicone being soft may also tolerate a fair amount of inaccuracy
<wpwrak>
dabs some silicone on wood treated with improvised release agents ... oil and alcohol. we'll know how that went in ~24 hours.
<wpwrak>
bartbes: not just a signal handler. the first process just waits for the other one to terminate by whatever means, be it crash, kill, or regular exit. then it cleans up the mess that was left behind.
<wolfspraul>
silicone is used for low-volume plastic injection
<wolfspraul>
from one silicone mould, you can make 10-20 pieces, of course it looses accuracy fast
<wpwrak>
bartbes: for an even tougher variant, open a pipe to a daemon, let the daemon do the dirty work, the daemon waits until the pipe is closed, then cleans up
<wpwrak>
bartbes: you're guaranteed that, no matter how your process died, its file descriptors will be closed (well, excepting catastrophic system failure :)
<bartbes>
yeah the thing is
<bartbes>
how would I release the fb from that 2nd (host) process?
<wpwrak>
bartbes: the advantage over the wrapper approach is that they can live in different domains of the system. e.g., the daemon can be unkillable by the (non-root) user.
<bartbes>
you're seriously overthinking this..
<wpwrak>
bartbes: the 2nd process would "own" the fb. e.g., open it, set it up, then either tell the 1st process to use it, or - if necessary - pass a file descriptor. when you close the pipe, clean up.
<bartbes>
yeah, that isn't going to work
<wpwrak>
bartbes: naw, that's how such things are done if failure is not an option :)
<bartbes>
I just need to clean up the fb afterwards
<bartbes>
or use a signal handler to catch sigsegv :P
<wpwrak>
bartbes: why is it not going to work ?
<bartbes>
your solution is hugely impractical
<wpwrak>
sigsegv has the disadvantage that you won't catch user kill, abort(), assert(), any function doing a fprintf(stderr, "moan\n"); exit(1);, etc.
<wpwrak>
it's not completely tested because the underlying resource didn't work as intended at that time (kernel problem), but it should be close
<wpwrak>
what you do is that you fork your daemon and open a pipe (unix domain socket) to it. the daemon opens the resource, sets it up, and passes a file descriptor back to you
<wpwrak>
when you're done, you exit. daemon notices that you closed the pipe, cleans up (it still has the file descriptor), and exits as well
<wpwrak>
a lot of programs rely on this mechanism in one way or another. not all of them use file descriptor passing, though, but that's just a shortcut so that i can keep my daemon simple (i.e., i don't worry much about the application doing bad things with the file descriptor, which it could in this case)
<wpwrak>
wolfspraul: (silicone mould) yes. i wonder where to get this sort of materials from (here in buenos aires)
<wpwrak>
wolfspraul: the alternative would be to have a mold made of something more solid, e.g., wood, and just pour the silicone inside. so the "case" itself would be silicone.
<wolfspraul>
I don't know, too little experience. My guess would be that a silicone case is too weak.
<wpwrak>
wood may not be the greatest of materials for this sort of experiment, but it's easy to source and machine. also, if i find a release agent that works with wood, it'll work with anything ;-)
<wolfspraul>
but I don't know what happens if you make a case out of relatively thin silicone... no idea
<wolfspraul>
the moulds I know are big blocks
<wpwrak>
i was thinking of a maybe 5 mm thick blob of silicone into which the pcb (or a suitable victim pcb/object) is stuck
<wpwrak>
so it would hug the board very closely
<wpwrak>
of course, it wouldn't really be a case, but perhaps good enough for pioneering experimenters
<wolfspraul>
so if we want proper high quality copyleft work, my minimum cost will be ca. 3000 USD
<wolfspraul>
you can say 1000 USD for PCB, 1000 USD for SMT, 1000 USD for components
<wolfspraul>
(I haven't even looked at the bom yet, just giving you ballpark numbers)
<wolfspraul>
the PCB can be cheaper, especially if it's 2-layer or so
<wpwrak>
hmm, ~300-500 units then. quite a lot
<wolfspraul>
but the SMT will easily be higher
<wolfspraul>
lowest maybe 800 USD
<wolfspraul>
so let's look at the bom now :-)
<wpwrak>
err, s/units/bundles/ (one atusb plus one atusd)
<wolfspraul>
there's always the option of doing things 'on the street'
<wolfspraul>
then the pcb will be shared with others, some kids will solder it together by hand, etc.
<wolfspraul>
parts will be taken out of whatever spare leftover box flying around
<wolfspraul>
and so on
<wolfspraul>
but it works
<wolfspraul>
:-)
<wolfspraul>
for 10 units, that system is unbeatable
<wpwrak>
soldering is a bit tricky in this case, with all the RF and nasty ground planes
<wolfspraul>
the gnd planes are on the pcb, right?
<wolfspraul>
why does it affect soldering?
<wpwrak>
yup
<wolfspraul>
you mean the amount of solder has to be controlled well? must not spill onto the gnd planes?
<wpwrak>
the ground acts as a heat sink. so when you try to solder something nearby, the ground will suck away the heat
<wolfspraul>
well I am describing process. I can do anything 'on the street'. whether it will work, or how well, that's another question.
<wolfspraul>
the lowest cost there is maybe 150 USD :-)
<wpwrak>
and often enough, the solder just solidifies on the (cool) ground and not on the pins
<wolfspraul>
for pcb, smt, components, testing (ahem) - everything
<wpwrak>
still need to figure out testing. the good news is that it'll get a little easier with the new chip.
<wolfspraul>
that's the bom for 100 units?
<wolfspraul>
you cut off the total...
<wpwrak>
oh, sorry ... updating
<wolfspraul>
well it looks like the 1000 usd I was guessing :-)
<wolfspraul>
it's always the same...
<wolfspraul>
so yeah, if we make 100, that means each will cost 30 USD
<wolfspraul>
to make
<wolfspraul>
very very roughly
<wolfspraul>
of course my time for free, and Adam also (maybe there is room for a little money for Adam, but not much)
<wolfspraul>
if you want to bring the costs down, we have to make more, like you said, maybe only 500
<wolfspraul>
but that may be too risky
<wolfspraul>
maybe we discover serious improvements
<wolfspraul>
or we make a case, and need to make a modification to the boards
<wpwrak>
we probably will
<wpwrak>
that too
<wolfspraul>
at 500, PCB cost is still 1000 USD, SMT still 1500 USD
<wolfspraul>
for 100, maybe PCB is a bit cheaper
<wolfspraul>
maybe even 400 USD, don't know
<wolfspraul>
but SMT will still be 1000 USD
<wolfspraul>
it's exactly the same kind of trap like with jtag-serial
<wolfspraul>
I bit the bullet there and made 100 in the end.
<wolfspraul>
very hard
<wolfspraul>
each one costs me 30 USD, just cash costs
<wolfspraul>
500 would have been much better, but that is just way too risky
<wpwrak>
updating doesn't seem work. i'll just paste the totals here: atusb  13 items, USD 932.4
<wolfspraul>
maybe we can make 8:10 and usb together
<wpwrak>
atusd 7 items, USD 400.31
<wpwrak>
(that's for 100 units each)
<wolfspraul>
why is atusb 500 USD more?
<wpwrak>
yup, that would be the idea. you need them both anyway
<wpwrak>
because of it has an MCU, a USB connector, etc.
<wpwrak>
ah, and a crystal
<wolfspraul>
both are 2-layer?
<wpwrak>
yes. 0.8 mm
<wolfspraul>
let's say we make a panel, maybe both atusb and atusd on the same panel?
<wolfspraul>
hey, btw
<wolfspraul>
atusd = secure digital?
<wolfspraul>
:-)
<wpwrak>
dunno. is there any other name ? :)
<wolfspraul>
you don't want to take the glory of these famous and valuable trademarks, and steal intellectual property, do you?
<wolfspraul>
I was laughing when I got a letter from these intel lawyers once.
<wolfspraul>
and their 'famous' client
<wolfspraul>
they really write like that
<wolfspraul>
so many things one could reply, but what a waste of time...
<wpwrak>
(panel) you'll probably have something like four of each per panel :) they're not exactly huge
<wolfspraul>
I'm thinking about ATUS-b and ATUS-d
<wolfspraul>
make ATUS the name
<wolfspraul>
that cuts right through the problems :-)
<wpwrak>
naw, the basename is atrf :)
<wolfspraul>
the panel is about cost
<wpwrak>
i already did one of the huge renamings
<wolfspraul>
if we can have both atusd and atusb on the same panel, that means only one time smt programming, for example
<wolfspraul>
I mean pick & place programming
<wolfspraul>
ah ok
<wpwrak>
yup. they also share many components
<wolfspraul>
so I will just call both boards atrf then
<wolfspraul>
atrf USB version, and atrf 8:10 version?
<wolfspraul>
at = atmel?
<wpwrak>
what's 8:10 ? :) i religiously follow the mailing list, but i never saw such a thing mentioned there :)
<wpwrak>
at = atmel, yes
<wolfspraul>
yeah
<wolfspraul>
I started to write, but then didn't want to wake up sleeping dogs.
<wolfspraul>
atrf is the name?
<wolfspraul>
'at' rf sounds cool, I can live with that
<wolfspraul>
it's not at = atmel
<wolfspraul>
it's at = at something
<wpwrak>
atrf is what i use when referring to the whole thing. e.g., for the tools. atrf-id and such.
<wolfspraul>
up to you, I follow your names
<wolfspraul>
so we put both atusb and atusd on the same panel
<wpwrak>
originally, i thought i'd also try ti's ccXXXX series. so the naming scheme was intended for at... and cc...
<wpwrak>
then i found out that TI's chips all have export restrictions
<wolfspraul>
as for how many in the panel, typically you ask the pick&place people first
<wpwrak>
(panel) yes, there ought to be a minimum panel size and it's probably larger than atusb plus atusd
<wolfspraul>
for example, when it gets too long, the arm that puts the components onto the board may push it down slighly, and lift up on the other side, shifting (unsoldered) components
<wolfspraul>
oh sure, conveyor belt
<wolfspraul>
you need to have a few mm margin on the sides, and if possible optical mark as 0,0 point
<wpwrak>
also, atusd is L-shaped, so you probably want two together. maybe 2*atusb + 2*atusd per panel.
<wolfspraul>
yes so anyway, both minimum and maximum size are typically foudn out by talking to the particular smt shop
<wpwrak>
yup. adam should be able to provide some estimates
<wolfspraul>
my guess would be maybe 4x4 boards or so
<wolfspraul>
if it gets too big it's not good, adds new problems
<wolfspraul>
the atusb has a connector that extends beyond the edge of the board?
<wolfspraul>
but atusd has not, right?
<wpwrak>
(names) i guess i'll change them at some point. there are more "SD" references to kill anyway, including a beautiful "uSD-Card" over in kicad-libs :)
<wolfspraul>
that needs to be considered for panels, so that everything can be soldered before the boards are cut
<wolfspraul>
the cutting normally will just go all the way through, you cannot cut a pcb when you have a connector on the other side
<wolfspraul>
or at least that adds extra complexities at that step
<wpwrak>
(minus the green cable, minus the 100 mil header at the bottom, and it will be a few mm longer)
<wolfspraul>
nice work
<wpwrak>
(cutting) do as we did at openmoko ? get the board perforated by not completely cut
<wolfspraul>
and then how do you cut them?
<wpwrak>
thanks :)
<wolfspraul>
I'm just describing the most common/best understood/cheapest way
<wpwrak>
how I cut them ? i cut them before soldering.
<wolfspraul>
so on a panel, you put all connectors that are reaching over the edge of the board pointing outwards
<wolfspraul>
not somewhere inside the panel
<wpwrak>
how we "cut" them at openmoko ? dunno for MP. probably manually - you can just break them off at the perforation
<wolfspraul>
ok
<wolfspraul>
like I said, I just describe what I think is the normal/cheapest way
<wolfspraul>
don't break
<wolfspraul>
yes you can, but people try to avoid it
<wolfspraul>
you cut after reflow & dip
<wolfspraul>
after everything is on
<wpwrak>
yeah, mechanical stress and such
<wolfspraul>
(if possible)
<wolfspraul>
but then you cannot cut where you have connectors over the edge of the board
<wpwrak>
i can also try to optimize away the usb connector if you want ;-)
<wpwrak>
i've done that too ... lemme find it ...
<wolfspraul>
because the normal cutters they have (and operators), will just cut all the way through, if you know what i Mean... :-)
<wolfspraul>
but it's no big deal here, with only atusb having a connector
<wolfspraul>
we can position it pointing outwards
<wolfspraul>
so maybe let's make 100 of each?
<wolfspraul>
pcb should be... let's say 600 USD (guess)
<wolfspraul>
smt... 1200 USD
<wolfspraul>
components 1300 USD
<wolfspraul>
3100 USD
<wolfspraul>
yield?
<wolfspraul>
how should we sell them?
<wolfspraul>
in pairs?
<wolfspraul>
shipping is also a pain, so cheap for fedex
<wolfspraul>
maybe just throw in a regular mail letter :-) that would be quite hardcore though...
<wolfspraul>
no tracking, no nothing
<wolfspraul>
what do you think about making 100 each?
<wolfspraul>
how many do you need? (of course you get them for free)
<wpwrak>
100 sounds like a reasonable quantity to me
<wolfspraul>
100 each?
<wolfspraul>
or 50+50
<wpwrak>
yup
<wolfspraul>
yes I think so too
<wolfspraul>
I hope you feel good about the design :-)
<wolfspraul>
and I'll just make it, he he
<wolfspraul>
but no worries, no risk no fun if it doesn't work, I think I made the decision with a clear state of mind
<wolfspraul>
how many do you need?
<wpwrak>
(sell in pairs) sounds reasonable as well. dunno if people will object if they're "forced" to get one they don't think they want. but we can probably worry about that later as well :)
<wolfspraul>
I'm not so worried about 'forced'
<wolfspraul>
I'm worried to reach people that don't have much money, in the cheapest way possible.
<wolfspraul>
if we have 50 USD fedex cost on top, shipping an item for 20 USD is stupid
<wolfspraul>
it's not about forcing anyone
<wpwrak>
i would need maybe 4-5 pairs for myself. have some units for rework and experiments.
<wpwrak>
yes, fedex is the killer here
<wolfspraul>
for example, if we say 'you can only buy this in 10 pieces', then someone has to step up and organize a little group purchase
<wolfspraul>
if that happens, everybody can save a lot
<wpwrak>
that's messy
<wolfspraul>
if it does not happen, we shoot ourselves in the foot
<wpwrak>
you could try the distributors, though
<wpwrak>
(foot) yup
<wolfspraul>
I am just explaining my thinking.
<wolfspraul>
it's not about forcing someone to buy something they don't want. not at all.
<wpwrak>
lemme show you the non-connector solution. a minute ...
<wolfspraul>
I can ship them in a regular untrackable letter.
<wolfspraul>
but that will also be messy.
<wolfspraul>
people have to pay first, cc or paypal, and then just wait and pray.
<wolfspraul>
we can always offer fedex, but that only makes sense if you order 2, or even 4 right away
<wolfspraul>
let's say we sell them for 20 USD a piece?
<wolfspraul>
or 30?
<wolfspraul>
at 30 we could actually recoup all costs, more or less
<roh>
ha. found a packetmarke. no standing in line.
<kristianpaul>
wolfspraul: (Shenzhen) nice pics, so much to see...
<roh>
wolfspraul: you want with tracking and insurance, right?
<roh>
hmmm.. what about transparent shrink-wrap-tube?
<roh>
as 'case' ?
<wolfspraul>
roh: you mean your shipment to me? if you can send with DHL, sure that's cool
<wolfspraul>
otherwise regular mail should not get lost either, I would think
<wpwrak>
(atusb case) the board is a bit larger (wider and longer) than the usual usb sticks
<wolfspraul>
roh: it's just a heads up now about these boards, but I think Werner and I are quite determined to pull it off. So unless you hide and resists, you will have some of these beauties at raumfahrtagentur in a month or two.
<wpwrak>
(heat-shrink tubing) that could be an option for atusb. would leave one side wide open, though
<roh>
wolfspraul: :) sure.. send them over
<wpwrak>
i don't see it for atusd though. shape is too weird.
<wolfspraul>
ah nice idea, you mean those normal heat shrink tubes?
<roh>
i think we also got a atmel raven kit around (same transcievers)
<roh>
wolfspraul: quite normal.. let me find a picture
<wpwrak>
wolfspraul: heat-shrink tubing was used in the little jtag dongle that in turn inspired the openmoko debug board
<wpwrak>
wolfspraul: of course, this one had connectors on both sides
<wpwrak>
roh: my thoughts on silicone would basically be to make a tight wrap of silicone, mainly held by friction and elasticity. also easy to remove for rework ;-)
<wpwrak>
roh: wood would enter the picture at the material for the mold. not because it's ideal for a mold, being porous and everything, but it's easy to source and machine
<wpwrak>
roh: now i just need to find a good release agent for wood + silicone
<roh>
hm. mould...  to do what? completely submerse the board in a 'plastic block' ?
<wolfspraul>
kristianpaul: what these pics don't know is the size of it. I estimate 10-20k shops (like the ones pictured there one by one), and maybe 100-200k people working and trading there.
<wolfspraul>
yes, that's a 'k' for the number of shops
<wpwrak>
roh: the ideal would be a silicone solution where one could just "dunk" the board and it would form something like a thick glove. repeat a few times for thickness. not sure if this is realistic, though.
<roh>
i mean.. from the balun? to the antenna feeding point via some R or C .. why is the trace so wide? and the gnd so thin?
<wpwrak>
roh: otherwise, make a u-shape mold, fill it with "thick" silicone, then suspend the board in it
<wpwrak>
roh: (trace) the calculators say a 50 R feed line has to be around 60 mil on this kind of board
<wpwrak>
roh: you mean the connection ? the ground has a bigger ground plane underneath. or do you mean the size of the ground area per se ?
<wpwrak>
roh: the board will also get more vias. more like the via density of the atusb board. this is just an older version where i had the wrong number for the minimum via spacing
<roh>
wpwrak: ah i see. (ground plane)
<roh>
how were these boards made?
<roh>
etched?
<wpwrak>
cnc-cut, then toner transfer, yes
<roh>
so 2 sided?
<wpwrak>
aye
<roh>
very nice result for home-made stuff
<wpwrak>
thanks :) it took a bit of experimenting :)
<roh>
i can imagine :)
<wpwrak>
oh, and i cheated a little. that old atusd board still has leaded solder. i've switched to lead-free since. so soldering quality has gone down a bit.
<wpwrak>
but via quality has gone up, since i switched to cnc-drilling them as well and using a wire that would fill the hole completely, to it wouldn't move around when soldering. that used to be a pain.
<roh>
:)
<wpwrak>
bartbes: your posting doesn't mention if you had to use the .INF file that was posted before. may be worth clarifying.
<wpwrak>
hmm, oil seems to work well as a release agent for silicone. talcum powder is even better, but leaves residues on the silicone. let's see what happens with an oiled circuit submerged in silicone. good that i have so many broken boards to throw away ;-)
<wpwrak>
also, silicone responds reasonably well to paint thinner, so that may be a way to get the stuff to act a bit more like paint. it doesn't like acetone - builds small chunks in it.
<wpwrak>
now .. what to do with the pc controlling my mill ... suffered what looks like a disk failure. of course, the disk is a maxtor, so it's either the very first failed maxtor in my life, or something else blew ...