<wpwrak>
ah, new feature at fedex: you can now pick the time zone relative to which the scans times are shown. e.g., Jan 21, 2011 8:08 PM Shipment information sent to FedEx
<wpwrak>
now let's try something easy, "origin" time zone. that would be the US, fedex home turf. undefined NaN, NaN NaN:NaN AM Shipment information sent to FedEx
<wpwrak>
very interesting. at least we know it's morning there ;-)
<kristianpaul>
nice tftpclient works with a simple string, lets try something bigger
<qi-bot>
[commit] Werner Almesberger: stdpass.fpd: set outline clearance (courtyard) to 0.15/0.25 mm of IPC-7351B http://qi-hw.com/p/kicad-libs/7c62277
<roh>
wpwrak: oh.. i see you had fun with footprints
<roh>
nice mail
<wpwrak>
still having more fun. the qfns still lurk ...
<roh>
:)
<roh>
see it that way: anything you do there, lots of embedded hackers will profit from
<wpwrak>
the rules adam suggested are a bit more sophisticated than the virtual shrug you get from most data sheets, but if you look at NXP's material, then you see that there's quite a lot more to them
<roh>
the qi-hw 'development' of getting a toolchain and data to build hw from ideas is a really important step in my opinion. we need an alternative to eagle for the community.
<wpwrak>
yeah. the next step is then a mass-produced cheap mill for making pcbs :)
<roh>
most importantly: getting all these rules being developed real-world tested
<wpwrak>
and then, pick and place :)
<wpwrak>
yup. the real-world tests are the scary part :)
<roh>
wpwrak: hehe... true. i'd like to own a picknplace machine... just need.. more space first
<roh>
but i would be glad to have something which generates proper layouts from completely foss tools
<wpwrak>
for p&p, i wonder if one couldn't make a machine that just picks components from a high-density loose container, such as a pill box or my vial tray. so you'd de-tape first, and then the p&p would figure out how to get the components out and oriented
<roh>
i dont need to manufacture all parts myself. i would be happy with having the possibility to order proper pcbs without using the vendor tools of the service company
<wpwrak>
(generates proper layouts) you mean design or produce (i.e., make boards) ?
<roh>
eh. yes
<wpwrak>
oh, that should be fine. we have nice gerbers :) we're not the first ones to use kicad :)
<roh>
true. but there is a lot of difference between a tool spitting out 'something you can work with' or something 'you can mass produce' with a serious yield
<roh>
thats where your tolerances make a big difference
<wpwrak>
ah yes, that's true
<kristianpaul>
(yeah. the next step is then a mass-produced cheap mill for making pcbs) wow, really? :-)
<roh>
these yamaha picknplace machiens are funny to watch
<roh>
kristianpaul: no. milling is nice for prototypes. completely crap for anything automated or mass-produced
<kristianpaul>
agree
<roh>
you by far cannot reach the details needed to live without rework. also running-costs are quite high for all rapid prototyping technologies
<roh>
milling is fine for 'simple' stuff.. werner is already kicking the limits of home-brewn pcb witj the qfn... but these were etched and toner-transfered, not milled, right?
<wpwrak>
roh: (milling) i mean for cutting/drilling the boards. i'd still use etching for the layout.
<roh>
atben/atusb boards
<roh>
i see.
<wpwrak>
roh: with HCl+H2O2, you basically have no waste acid to dispose, because the volume is so small. just put it outside, let the water evaporate, then throw away the solids.
<wpwrak>
well, or collect them, if you want
<roh>
wpwrak: mmh..  still not good for me. i dont have any space left for a chemics lab stm
<roh>
eh atm
<wpwrak>
you need about 50x50 cm, maybe less. some large plastic basic to work inside, so that things don't spill outside
<roh>
wpwrak: we currently already host a small biolab... could use a second floor at the lab
<wpwrak>
then i have one of the large low glass forms used for the oven inside, again to contain spills. this one is expected to get spills from time to time.
<roh>
and the rooms with all the dirt(saw, milling, turning dust) arent a good fit for chemicals
<wpwrak>
ah .. biolab sounds like a good excuse for not cleaning. i have to remember this ;-))
<wpwrak>
the chemistry is quite simple
<roh>
naah. real biolab. a friend of ours experiments with gen-foo
<wpwrak>
(chem) the main problem are the corrosive vapours. already if you keep a bottle of HCl inside, some will escape and look for a victim
<roh>
wishes for _much_ more space including funding and doing some proper seperation of the dirt problem
<wpwrak>
(chem) then prepared acid also produces corrosive fumes
<roh>
wpwrak: true. i know corrosion in all forms now.
<wpwrak>
if you have a balcony, that would be the place to put all these things
<roh>
when playing with kilograms of batteries.... one sees nasty things
<roh>
ever held about 4.5kg of lithium?
<wpwrak>
or i have a little green box on the terrace ... with contents that aren't all that "green" ;-)
<wpwrak>
urgh
<wpwrak>
real men don't just hold it. they LICK it ;-)
<roh>
no balcony... also that wouldnt work well.. its berlin so outside temperatures are about 4°C atm. and can be between -20°C and +35°C in the shadow easily over the year
<wpwrak>
sounds okay
<roh>
real lab would mean having a second fridge with labels and the chemicals in there
<wpwrak>
don't know at what point HCL freezes, though
<roh>
not having the vodka bottle for drinking right next to a box with stuff i know it is 'mutagen'
<wpwrak>
sure. real lab. but you can get things done with a lot less
<wpwrak>
;-)
<roh>
true. but i want that warm fuzzy feelting back
<wpwrak>
all my nasty chemicals vanish quite easily when necessary
<roh>
ever had a battery acid spill to clean up?
<wpwrak>
the workspace is a box with drawer. normally sits in the kitchen, for easy access to water, but when the kitchen has to be more habitable, i carry it over to the mech lab
<wpwrak>
yes, in a car
<wpwrak>
eventually, the car suffered structural failure, but at a completely different place
<roh>
i was cleaning the table and noticed some clean liquid
<wpwrak>
naw, what i'm saying is that the HCl+H2O2 process is ideal for home use because it's so efficient and produces little waste
<roh>
when it started doing things with the piece of cloth i was using i washed my hands and started using gloves and one-way towels
<roh>
ok. i will come back to that when i get annoyed doing outline milling ;)
<roh>
s/outline/isolation
<wpwrak>
you just have to make sure the vapours don't have too much time to mess up things. so keep the HCl outdoors and unused acid as well. (well, you could collect it, but that seems somewhat pointless)
<wpwrak>
(acid surprise) nice ;-)
<roh>
yeah. i am quite lucky that all MY batteries are lithium based or encapsuled lead-gel. no single one with liquid stuff in it.. thats all the motorcycle crews stuff
<roh>
they are also removing the acid now. replacing their batteries with lifepo4.. helps reducing weight of the cacle
<roh>
s/cacle/cylce
<roh>
some even remove the generator (lichtmaschine) to reduce the weight. lifepo4 has enough energy to run the electronics for regular trips (recharing at home)
<roh>
crazy stuff
<wpwrak>
:) maybe search the bins of project orion for some more efficient power sources
<roh>
i think our issue is storing and transporting energy. converting them i see as widely solved issue
<wpwrak>
hence project orion :)
<rjeffries>
werner are you here?
<wpwrak>
eating .. can't type much, but i can read what you write
<rjeffries>
your priorites (eating over hacking) sure are screwed up. <g>
<rjeffries>
anyhow
<rjeffries>
what would be the technical challenge to a card that plug into 8:10 (gag) slot on Ben, that has send and recieve LEDs to control a TV etc?
<rjeffries>
there are chips that decode RC5/RC6 I am guessing that this would be sorta ugly but that's OK
<rjeffries>
with such a chip, Ben read  and writes a serial data stream to the remoote control dongle
<rjeffries>
not sure if BEN can also supply enough power however
<wpwrak>
mmh, you could use a chip that just detects the carrier
<wpwrak>
steve|m: nice :) didn't know you could do all this without even etching
<wpwrak>
kristianpaul: yeah, i know ... sins of the youth ...
<kristianpaul>
steve|m: cool, 8:10 card
<wpwrak>
kristianpaul: looks like a 9:24 card to me :)
<kristianpaul>
:p
<kristianpaul>
yes
<rjeffries>
steve|m so you do some MSP430 hacking? their Launchpad for $4.30 was irresistable for me. n=3
<rjeffries>
wpwrak waht frequency is the clock that come out the 8:10 connector
<steve|m>
rjeffries: well, I got 2 launchpads some weeks back and definitely will do something with them.. I want to build a ultra low-power epaper clock with a MSP430 and the display of a motorola F3
<steve|m>
I was quite amazed, since everything worked out of the box.. and the "mspdebug" tool is awesome
<steve|m>
2KB flash isn't that much, though
<rjeffries>
steve|m yes, that MSP430 Launchpad kit is very nice. TI must be subsiding it $10 or $15 a pop, just to capture new designs. Great for the hobby crowd
<rjeffries>
steve where are you located
<steve|m>
near FRA/germany
<roh>
rjeffries: msp430 community is strong but small i think
<roh>
also not all variants are supported by gcc yet i think
<steve|m>
but the "support" mostly is register definitions which can be easily added
<roh>
sure. still a lot of people are quite resistant to hacking up a working compiler before doing a project ;)
<wpwrak>
rjeffries: you can program the clock dividers. you can probably set it within a wide range, particularly if you're not afraid of changing the system's main clock :)
<rjeffries>
wprak thx for clock info. you are NOT using theclick from BEN on atben correct? (you had to add  xtal?)
<wpwrak>
yeah. i tried to use it, but it turned out to have slight imperfections that made it not work properly for atben.
<wpwrak>
rjeffries: and if all else fails, you could also generate the ~38 kHz carrier also with software. less efficient, but it'll work.
<roh>
nice hack. unfortunately the devices aren't avail in .de
<steve|m>
roh: yeah, only amazon.co.uk for 9.95 pound + 9,95 for shipping, which is something like 23¬ currently
<steve|m>
basically you get 2 radios - it comes with an usb stick
<roh>
hm... i never ordered at amazon
<roh>
does that work with a german amazon account?
<steve|m>
sure.. but needs a credit card for payment
<roh>
nih
<steve|m>
roh: you want one? there's a pack of 2 for £17.84, so (price + 3,90 dhl shipping to you) / 2 would be something like 17,95¬
<roh>
for something like 10Euro, sure
<steve|m>
yeah, true.. it's much cheaper in the US
<roh>
i am just hesitant to waste too much time getting weird shit sent from 'out of europe'. eats a lot of time and money usually
<steve|m>
17,95 would be the price for one actually
<roh>
uh. i see.
<roh>
well.. 20E is also ok for some new toy
<steve|m>
roh: okay, then I'll order a double-pack
<roh>
:)
<kristianpaul>
damit ttcp gives and nicelly RX Fifo overflow...
<kristianpaul>
lekernel: when no rtems support, how do you tested minimac and basic tcp/ip comunication on mm1?
<kristianpaul>
s/and/a
<lekernel>
what's ttcp?
<kristianpaul>
nice net tool for measuring troughtput
<kristianpaul>
So far my tests shows that there is a problem with tcp replies from rtems...
<kristianpaul>
in wich tftpd is not affected as it just pull? files from the indicated server
<lekernel>
there are tons of problems with ethernet on rtems
<kristianpaul>
but when i try the httpd or telnet examples, oh dear..
<lekernel>
dunno. works for me ~8% of the time
<lekernel>
80%
<kristianpaul>
yeah i'm just realizing it now ;-)
<kristianpaul>
bah
<lekernel>
I have no time, really
<kristianpaul>
jeje
<kristianpaul>
sure
<lekernel>
Fallenou was supposed to fix it, but he didn't
<kristianpaul>
ok, i'll check what seems to be the problem, i really need ethernet for my tests
<kristianpaul>
suscribes ro rtems mail list :-)
<kristianpaul>
oh this is qi :p
<larsc>
this is SPARTA(N)!
<wpwrak>
and packet seem to be jumping off the FIFO well enough ... :)
<lekernel>
kristianpaul: please ask fallenou about network problems
<kristianpaul>
lekernel: i did on irc yday, but no reply so far..
<lekernel>
he was paid for that
<kristianpaul>
I think i'll better ask on ml
<kristianpaul>
ah yes,, gsoc stundents !!
<kristianpaul>
:-)
<lekernel>
yeah, and I didn't test his stuff well enough. if i ever participate in gsoc again, i'll have a totalitarian policy about bugs and stability
<wpwrak>
all those executions at dawn ...
<lekernel>
i.e. you won't pass until your code is rock solid under the worst possible conditions of stress
<kristianpaul>
stress ethernet !! this is_ real_time_ so  i want the 10MBps troughput ;-)
<wpwrak>
roh: do Atmel publish footprints anywhere ? all i see are package definitions
<wpwrak>
lekernel: reminds me of the time when i worked at IBM research. back then, IBM still liked token ring a lot more than anything else. we had just gotten new workstations. they had as a brand-new feature ... ethernet. i did a ping -f on a colleague's station. the box crashed to hard even the reset button didn't work ;-)) so you see, even major products from reputable companies aren't necessarily tested all that well.
<lekernel>
unfortunately, I did test the thing with ping -f
<lekernel>
it works
<lekernel>
but large packets do a lot more damage
<lekernel>
it's actually so crappy that transferring a few megabytes over FTP often crashes it
<lekernel>
as long as you have big packets, the bugs manifest themselves
<wpwrak>
that's nasty
<kristianpaul>
no icmp works!!, not fully tcp :/
<kristianpaul>
ah is packet size a problem..
<kristianpaul>
i'll try hping later
<wpwrak>
where does rtem's tcp/ip stack actually come from ? diy ?
<kristianpaul>
bsd
<kristianpaul>
or bsd-like..
<roh>
wl
<wpwrak>
kristianpaul: ping -s  for the packet size
<kristianpaul>
wl?
<roh>
wpwrak: dunno. they got a lot of pdfs online
<wpwrak>
roh: ;-))
<wpwrak>
roh: i've been sifting through that haystack for a while ... but haven't spotted any needles yet :)
<kristianpaul>
wpwrak: yeah sure, but i dont transmit valuable data on icmp,  i need some tcp ack,rst,syn stress tests :-) hping to rescue !
<wpwrak>
kristianpaul: the losses probably occur even if the data isn't valuable ... :)
<kristianpaul>
i also had bad luck trying make a session with a socket,  used netcat, but timeout popup first, i'll re check but is almost same behavior than in telnet, httpd, ...
<wpwrak>
kristianpaul: guess you have to fix that ethernet first
<kristianpaul>
wpwrak: may be UDP it is fast i used it for VPN at work
<kristianpaul>
it seems
<kristianpaul>
I just make sure what is the problem..
<wpwrak>
kristianpaul: if it's a packet size problem, you could try to lower the MTU
<kristianpaul>
apart from the TX FIFO thign :\
<kristianpaul>
hmm i dint ought on MTU, thanks for poiting !
<kristianpaul>
s/ought/tought
<wpwrak>
there ought to be a switch somewhere for changing the MSS, too. that would be less intrusive. but PMTU discovery may do a good enough job if you tweak the MTU. just don't make it too small
<roh>
well.. in the end we cant have support for all possibilities. and to be fair, i dont think any vendor does. my bet is that most just have one footprint for every 'string' like 'qfn84' and use that for all chips with the same 'string'
<wpwrak>
roh: oh, you ain't seen nothing yet ;-)
<wpwrak>
roh: NXP alone have about three variants
<wpwrak>
roh: intersil have 4 (for QFN32) at last count
<wpwrak>
roh: (3 NXP variants) for QFN28 and QFN32 each
<kristianpaul>
Is all this discuss about NXP because a foot print-fix recomendation from Adam btw?
<wpwrak>
roh: some of them have very detailed land patterns. alas, the ones that are similar to what i need don't
<roh>
wpwrak: the question is: do they fail when soldering in reality?
<wpwrak>
kristianpaul: yeah. they don't resonate well with the material i found
<roh>
i mean... are you saying we cant have one 'qfn32'-layout which works for all components?
<wpwrak>
roh: would kinda suck to find out :)
<wpwrak>
roh: the problem is the center pad. huge variations where.
<kristianpaul>
wpwrak: dont trust manuals? :-)
<wpwrak>
kristianpaul: i would trust them if they did actually document that stuff ...
<roh>
wpwrak: sure.. 'special' stuff needs special footprints (amps, ldo, pmu, stuff that dissipates heat that way)
<wpwrak>
(intersil) sorry, not four. it's 5.
<wpwrak>
roh: things also seem to depend on how the case is cut and such. sure, you can have a "generic" footprint that works most of the time in the lab, but ...
<roh>
sure. i still think we should only add footprint deviations to 'the lib' when needed, not for every new chip
<roh>
that way we would go nuts and not solve any real problems but more the theoretical ones... and it can still go bad. you know yourself how wrong documentation can be. rather stay on the design, prototype, test, rethink loop evaluating the real world
<roh>
the latter one is known to work well, and the foss community gives us good chance for crowdsourcing some of the rather annoying gruntwork which could be needed for some kind of testing, or inputting data. i was quite impressed about that in gta02 (how many people startet looking at the schems and commented on stuff)
<wpwrak>
roh: i'm just doing the chips i need for atben/atusb ...
<kristianpaul>
he, too much questions, i cant answer now ;-)
<kristianpaul>
give me some days
<roh>
;)
<kristianpaul>
btw you have a M1 too isnt??
<kristianpaul>
if so, grab rtems examples and see it your self :-)
<roh>
i only have a mechanical protoype right now (no chips soldered, only sockets)
<roh>
will get a rc2 next month
<kristianpaul>
good :-D
<roh>
the .v files are verilog?
<kristianpaul>
yes sr
<roh>
somehow it seems much less code than i expected
<kristianpaul>
what are you looking at?
<roh>
the .v files in boards/milkymist-one/rtl
<wpwrak>
mmu.v :)
<kristianpaul>
wpwrak: ;_)
<kristianpaul>
roh: top files are not so big (besides wishbone blablahblah..), check the cores :-)
<roh>
sure. but isnt that the idea? splitting complexity into modules till small enough for our tiny puny human brainz?
<kristianpaul>
ah, you mean sebastien brain? :-)
<roh>
i dont yet get even half of what this source does i guess ;)
<roh>
but i dont know verilog.. so i have atleast something to learn
<kristianpaul>
wpwrak: you think mmu is trivial to do? (actually so far i dint check literature about it)
<roh>
1'b0 means 1 'line' with the Binary value of 0 ?
<kristianpaul>
s/do/model
<kristianpaul>
1 bit
<roh>
4'hd is funny.
<wpwrak>
kristianpaul: trivial if you don't care about efficiency. non-trivial if you do :)
<roh>
what does 1'bz mean? is z something like 'tristate' (not 1, not 0, please float)?
<roh>
i guess i should find the right book for me.
<kristianpaul>
:-)
<wpwrak>
kristianpaul: for a dead-simple MMU, you just need one register set: current page virtual address, current page physical address. for each mmu'ed access, you compare with the virtual address. if there's a match, you substitute the upper address bits with the phy addr register.
<wpwrak>
kristianpaul: if there's a mismatch, you do a trap/supervisor call/whatever. the trap also disables the mmu until you return from it.
<wpwrak>
kristianpaul: the trap handler then walks the page tables in software and provides the new entry
<wpwrak>
kristianpaul: (for the trap handler, you also need to provide the address that caused the trap)
<wpwrak>
kristianpaul: all this assumes that your instructions that touch memory are restartable after a trap. if not, things get more complicated.
<kristianpaul>
What think linux about that?
<wpwrak>
kristianpaul: well, you have to implemenent the usual paging code of course
<wpwrak>
kristianpaul: that is page table handling. define a structure, access functions, intialization, and so on.
<larsc>
roh: z is floating
<roh>
and what does x mean?
<larsc>
undefined
<larsc>
matches everything if used in a case statement
<roh>
so 64'bx just means '64 lines, dont care which state yet' ?
<larsc>
i guess so
<roh>
hmm. that needs some getting used to
<larsc>
if run an simulation and see an x it could aslo be that there are multiple drivers setting different values for the bus
<larsc>
if you run ...
<kristianpaul>
roh: what also need time to used: how to rightly wire modules; the concurrent "programing"Â Â thinking
<kristianpaul>
anyway verilog is interesting as it looks like C :-)
<kristianpaul>
and be friendly with your synthesis tool :-)
<roh>
ack. i somehow like verilog more than vhdl already
<roh>
the vhdl code i got flying around is much more verbose and obscure
<kristianpaul>
:-|
<roh>
verbose as in 'i dont get why they write down the same twice'
<wpwrak>
roh: vhdl:verilog = COBOL:C :)
<roh>
hrhr
<larsc>
roh: you mean architecture and component?
<wpwrak>
similarly, SDL:Promela = COBOL:C :)
<kristianpaul>
wait wait, was not vhdl Ada like?
<wpwrak>
well, ADA is not necessarily a more flattering comparison ;-)
<larsc>
kristianpaul: yes it was/is
<roh>
larsc: more like syntax and concept
<larsc>
ok
<qi-bot>
[commit] Werner Almesberger: qfn.fpd: added measurements and improved design of the central pad http://qi-hw.com/p/ben-wpan/8703fc1
<qi-bot>
[commit] Werner Almesberger: atben.sch: rearranged spacing in schematics to make more room at the feed line http://qi-hw.com/p/ben-wpan/77a3029