<wpwrak>
wolfspraul: what happened: /home of my workstation died. a bit too quickly to recover any data :-( now i'm restoring things ... from repositories, memory, ancient backups, ...
<wolfspraul>
oh
<wolfspraul>
not much I can do to help I guess
<wpwrak>
wolfspraul: i'll have a few more days of fun... ;)
<wolfspraul>
yeah, sorry to hear that.
<wolfspraul>
I have two questions for you still :-)
<wolfspraul>
first is about the resistor on the jtag-serial board, Yanjun Luo did change it to a 250V rating now
<wolfspraul>
1M/250V
<wpwrak>
wolfspraul: what threw me off was that - for some mysterious reason - the pc didn't mount root when a first looked at the problem. that's why i thought it was the other disk. otherwise, i may have had a bit more time to recover things while the failing disk still had some life in it.
<wolfspraul>
my problem is in digikey I cannot find 'voltage rating' as a parameter for resistors
<wpwrak>
wolfspraul: (resistor) heh ;-)
<wolfspraul>
it has Power (Watts)
<wpwrak>
wolfspraul: BOOM to the rescue ;-)
<wolfspraul>
is that the right value and do I need to do some math?
<wolfspraul>
well yes true, but then I would need to encode the fields correctly for boom
<wolfspraul>
will 1M/250V work? (as the value field)
<wolfspraul>
1206 in the footprint
<wolfspraul>
I could just try, but I am worried it will pick the wrong part.
<wpwrak>
power is an indirect indicator. voltage and power both depend on the package size (and a bit on the type of resistor)
<wpwrak>
1M/250V will work if you teach the .sub file how to handle it :)
<wolfspraul>
can I copy it from somewhere?
<wpwrak>
there's a new tool that can help you with this ... lemme fetch eda-tools ...
<wpwrak>
ah no, it's in the openmoko svn ... testsub
<wpwrak>
and how much are those critters nowadays ? i once bought one for USD 1k. i think openmoko's was 2k.
<wolfspraul>
1650 USD
<wolfspraul>
this is for Qi Hardware now, so I plan a similar system with a wiki page and all Qi projects can use those numbers
<wolfspraul>
in fact that's what I wanted to discuss
<wpwrak>
very moderate inflation then :)
<wolfspraul>
the IEEE really is keen on making sure that all numbers are used
<wolfspraul>
not very wasteful 'blocks'
<wolfspraul>
so how do we do that?
<wolfspraul>
should we just start with 00-00-01 to 00-00-06 for the 6 Milkymist One RC1 boards?
<wpwrak>
yeah, at least they say that they're very rigid on this
<wolfspraul>
well I want to play by the book
<wpwrak>
not sure how bad they really are
<wolfspraul>
sure but why waste...
<wolfspraul>
so we could have a wiki page
<wolfspraul>
and whenever people need some, even for small runs, they just increment
<wolfspraul>
that's one way
<wolfspraul>
or we do have 'blocks'
<wolfspraul>
but then first of all some blocks may stay very empty, and we need to 'fill' those blocks later, or a block may get full and another one is needed
<wpwrak>
you could make small blocks
<wolfspraul>
I almost feel a sequential model is best, with the wiki it should be possible to administer it
<wolfspraul>
I don't expect thousands of runs, and even if there are hundreds after some years, fine there can be a table with some x hundred rows
<wolfspraul>
well, before I start I thought I discuss with you
<wolfspraul>
take your time...
<wolfspraul>
my vote is probably for a simple sequential model
<wpwrak>
if you make 1000 devices in one go, you can just allocate 1000 sequential numbers :)
<wolfspraul>
all products mixed, just increment as many as you need for the run right in front of you
<wolfspraul>
ah yes, sure
<wolfspraul>
I would allocate the first 6 (in retrospect) for the first 6 Milkymist One devices
<wolfspraul>
the RC1 run
<wolfspraul>
and another 35 for the RC2 run
<wolfspraul>
if you need some for ben-wpan you just add to the table, even down to 1 if you like
<wolfspraul>
at least it's an official table and we can really make sure those numbers are 'officially' and globally unique
<wpwrak>
sounds good to me, at least for now
<wolfspraul>
ok great
<wpwrak>
there's a tiny privacy/security concern with having MACs that are traceable to eactly production dates, but that's probably nothing really worth worrying about
<wolfspraul>
hmm
<wolfspraul>
we discussed this subject (unique/serial number) in general before you joined
<wpwrak>
oh, and WPAN uses some new IEEE MAC addresses with 64 bits. i haven't studied them yet, but they probably just added some 16 more bits. so no risk to run out of these anytime soon :)
<wolfspraul>
back then the consensus was that we are not against unique/serial numbers, because they can have many positive benefits
<wolfspraul>
but we have two requirements with each such number:
<wolfspraul>
1) easily/conveniently removable
<wolfspraul>
2) documented
<wpwrak>
unique is a requirement anyway. what i meant was that you could cross-link it to some other data, e.g., build date, hardware revision, etc.
<wolfspraul>
yes
<wpwrak>
removable is good, yes
<wolfspraul>
like I said
<wolfspraul>
that is possible, and there are many positive things that you can do with the cross-linking
<wolfspraul>
so we are not saying 'copyleft hardware = anonymous hardware'
<wolfspraul>
as in - never any unique/serial numbers
<wpwrak>
no tin foil hat :)
<wolfspraul>
the consensus was: a) removable b) documented
<wolfspraul>
but with those 2 requirements, bring them on, no problem
<wpwrak>
sounds good to me
<wolfspraul>
maybe I mention this policy/idea at the top of the page
<wolfspraul>
the IEEE doesn't like 'MAC-48', they want to say 'EUI-48'
<wolfspraul>
I will just make multiple tables then, if ben-wpan only needs EUI-64
<wolfspraul>
if it's indeed EUI-64 only, you can go into a range 'outside' of the EUI-48 range, thereby not using up any EUI-48 numbers
<wolfspraul>
up to you
<wolfspraul>
let's just start the wiki page, to track/coordinate who is using what
<wpwrak>
(eui-48/64) hmm, i wonder if they form completely separate namespaces. good question ...
<wolfspraul>
well we can see later
<wolfspraul>
if someone needs a EUI-64 number only, it should be chosen outside of the EUI-48 range, to conserve those numbers
<wolfspraul>
if you need a EUI-64 number that can 'fallback' to EUI-48, then of course you just take real EUI-48 numbers and encode them as EUI-64
<wolfspraul>
that's a requirement of the product, the wiki page is just to track who took what
<wpwrak>
seems that they defined it the other way around: there's a 0xffff codepoint in eui-64 that allows you to encapsulate eui-48 :)
<wolfspraul>
yes
<wolfspraul>
so like I said, whoever needs/wants a EUI-64 needs to know whether there is any reason to take a 0xFFFF encodede EUI-48, or one outside of that
<wolfspraul>
I will start 2 tables then, to track EUI-48/MAC-48 and (real) EUI-64 assignments
<wolfspraul>
with 'real' I mean non-0xFFFF EUI-64
<wolfspraul>
those can show up in the EUI-48 table
<wpwrak>
alright. maybe reserve some small range for "local use experiments". e.g., 00-00-xx[-xx-xx] or such
<wpwrak>
wikipedia (http://en.wikipedia.org/wiki/MAC_address) is slightly too dismissive about that difference. after all, it was IEEE worth one extra codepoint (0xfffe) to make this distinction
<wpwrak>
ah, but they mention the encapsulation as well
<wolfspraul>
the encapsulation in EUI-64 is different?
<wolfspraul>
I was just thinking about making just one table for both MAC-48 and EUI-48
<wpwrak>
good. now i know that thingy too. i had seen it before and couldn't quite make sense of it back then (lacking context)
<wolfspraul>
if the EUI-48 is in fact another namespace then I make the table only for MAC-48 :-)
<wolfspraul>
for Milkymist One, we need MAC-48 numbers, no?
<wpwrak>
it seems that the EUI-48 and MAC-48 aren't quite different but not quite the same either :)
<wpwrak>
for MM1, tht would be ethernet ? that should be MAC-48
<wpwrak>
the wikipedia article says that MAC-48 and EUI-48 are from the same address pool
<wpwrak>
the distinction looks like a case of a bit of overengineering
<wolfspraul>
ok I start with a MAC-48 only table, assign 6 + 35 for Milkymist One RC1 and RC2
<wolfspraul>
after that, if ben-wpan comes along and needs something, I'm sure you/we have figured out what the best number is
<wolfspraul>
maybe it's a case of corporate/authority envy, there are so many articles about what a MAC-48 is, but they were not written by the IEEE
<wolfspraul>
so they had to do something about that, and said "bah", it's EUI-48 now :-)
<wolfspraul>
and _WE_ will say what that is :-)
<wpwrak>
trademark it too, for good measure :)
<wolfspraul>
I am wondering whether we should also assign proper MAC addresses to NanoNotes, for USB-ethernet networking?
<wolfspraul>
it wasn't needed for now at least, well we can do it when someone understands the need and implements it too
<wpwrak>
dunno what the implications are. as far as i recall, usb-ether by default just borrows from some company's MAC address space.
<wolfspraul>
why does Openmoko have one then?
<wpwrak>
you can distinguish devices that way. if you have more than one, things seem to get funny with certain host operating systems. mac and such.
<wolfspraul>
were the Openmoko numbers ever used? I never heard about it...
<wpwrak>
i think they were. at least qi (the boot loader) passed them to the kernel. not sure about u-boot.
<wpwrak>
but about the only context where i remember having heard about it mattering was with macs.
<wolfspraul>
wpwrak: hmm. when I do this calculation my result is P = 250*250/1000000 = 0.0625
<wolfspraul>
but when I look up 1M 1206 resistors, there is not one below 0.125, most are 0.25W, some higher
<wolfspraul>
so maybe that's why Yanjun Luo first said the "/250V" rating would not be needed, in his experience (he said in 10+ years and lots of products he never needed it)
<wolfspraul>
because a 1M 1206 resistor will implicitly be able to hold up to 250V and higher anyway?
<wolfspraul>
well, something may be wrong with my math of course. Let me know.
<nitin_gupta>
NN kernel is not stable
<nitin_gupta>
getting lot many problems while using it
<rafa>
nitin_gupta: which kind of problems?
<cj101>
hi, is this a channel where i can get help with making a pcie card for my computer?
<wpwrak>
wolfspraul: resistors should have a V=... field in the characteristics
<wpwrak>
wolfspraul: you can't calculate P and V from each other. P depends on how quickly the heat can get out of the package. the smaller, the slower. V depends on how far the terminals are from each other, arcing and such.
<wpwrak>
wolfspraul: 1206 is probably big enough ... lemme check ...
<wpwrak>
hmm, tricky. panasonic and stackpole only go up to 200 V.
<wpwrak>
but that's continuous voltage, not peak. i didn't record peak. maybe i should ...
<wolfspraul>
hmm
<wpwrak>
the higher-voltage resistors have their voltage encoded separately, so just the package doesn't tell the value. you could grep the .chr files for it, though
<wolfspraul>
Yanjun Luo said this is totally unnecessary anyway :-)
<wpwrak>
;-))
<wpwrak>
so what changed his mind to put a voltage there in the first place ?
<wolfspraul>
your comment in the TODO file
<wpwrak>
the whole circuit is admittedly a little mysterious to me. what exactly is supposed to happen when there's a discharge into the shield ?
<wolfspraul>
after I repeatedly pointed your comment out to him he basically said "if you want to change it go ahead, I don't care. It's not necessary anyway" :-)
<wolfspraul>
the .chr files indeed show a V= value for the resistors
<wolfspraul>
why is my math so wrong?
<wpwrak>
ah :) well, i was just commenting on the inconsistency. if it matters for the cap, it ought to matter for the R. unless you actually expect to arc over the R, which may or may not be the case.
<wolfspraul>
maybe I just change it back, and done
<wpwrak>
because the math is about a different physical phenomenon
<wolfspraul>
we have already made fully working boards, so this whole thing seems like a time-sink operation
<wolfspraul>
I just leave the 1206 footprint, and remove the /250V indicator
<wpwrak>
i would like to understand how this circuit is supposed to behave. as opposed to, say, just connecting shield to the ground plane.
<wolfspraul>
let it match to the cheapest 1206 1M R resistor
<wpwrak>
;-)
<wpwrak>
it would have been such a nice way to exercise your .sub skills ;-)
<wolfspraul>
I will get to it, but the motivation must come from more valuable targets... (xue)
<wpwrak>
it's not *that* difficult :)
<wolfspraul>
not from a jtag-serial board that we already know works, even with a 0402 resistor that is most likely even weaker
<wpwrak>
well, it's ESD protection. did you ESD-test the board ?
<wolfspraul>
it's not difficult to parse and match fields, but I don't trust the meaning of the fields I am matching
<wolfspraul>
and as long as that's the case I hesitate to put some automatic matching in place that I will never question again in the future
<wolfspraul>
you are not even sure about 'continuous' or 'peak' etc.
<wolfspraul>
in digikey there is no 'voltage' at all, only power (watts)
<wpwrak>
all the values in the .chr at the moment are continuous.
<wolfspraul>
don't know what that means. maybe digikey is missing a field, maybe I don't understand what I'm dealing with :-)
<wolfspraul>
like I said, I will just leave 1206, because I think he changed the layout as well, and remove the /250V
<wolfspraul>
done :-)
<wpwrak>
digi-key don't put everything into their tables. i think you also don't have temperature coefficients.
<wolfspraul>
they are there
<wolfspraul>
the list is quite long for resistors, but no 'voltage'
<wpwrak>
ah, okay
<wolfspraul>
suspicious to me, given my large degree of cluelessness around this
<wpwrak>
anyway, if you look at the vendor data sheets, voltages are always there.
<wpwrak>
maybe digi-key are afraid people will get confused and think they can just pick a sufficient voltage without looking at power.
<wpwrak>
most of the time, power is what limits you, not voltage.
<wolfspraul>
if digikey doesn't even have it in their tables, for something as common as a 'resistor', then there must be more to it
<wolfspraul>
yeah, something like that :-)
<wolfspraul>
so I am not the right person to do the matching, let alone automatic matching
<wpwrak>
well, you don't have to understand the deeper meaning to do the matching. neither does BOOM itself :)
<wolfspraul>
in all 1392 different 1206 resistors I have in my boom .chr, there is not one with V != 200V
<wolfspraul>
I will keep my fingers off the matching until I know what I'm doing, or until I can copy/paste from a known/proven/verified design.
<wpwrak>
but i;m still curious why you need that circuit at all. in most designs i've seen, shield is just connected to ground. maybe incorrectly so. however, in some cases, you don't even have a choice. e.g., all common coax cables don't have a separate shield, so signal ground is protective ground. and it's certainly easy enough to touch the shield.
<wpwrak>
footprint size is a very good indicator for voltage (also for power)
<wolfspraul>
sure, that's why I will leave the change from 0402 to 1206 that Yanjun made
<wpwrak>
so it's less about "making it work" but more about documenting it properly. if you specify the voltage, then it's clear what you're after.
<wolfspraul>
but I remove the /250V indicator which creates this whole matching uncertainty, for unknown return value
<wpwrak>
your logic doesn't quite add up, but i agree that the result will work :)
<wpwrak>
hmm, i shall refer to my holy book on circuit design for an answer to that riddle ...
<wpwrak>
hmmm, the holy book agrees with not connecting directly in the case of non-RF, but suggests a choke or resistor instead of a capacitor. kinda the opposite. let's see if there's more in the in-depth chapters ...
<wpwrak>
wolfspraul: my holy book is more concerned with ground loops than with ESD effects on the shield. if making a connection at all, it suggests a bead. USB standard just mumbles something about consistency with accepted practices and regulatory standards, without offering any specific suggestions.
<wolfspraul>
also I think there is no issue, I think we are just going in circles
<wolfspraul>
all fine
<wolfspraul>
board works, designer is happy
<wolfspraul>
forget the 250V thing
<wpwrak>
wolfspraul:Â Â so i do wonder where this RC circuit comes from. i've seen that the Ben has the same sort of setup.
<wolfspraul>
Yanjun is super busy, I will hold my limited fire until I have something really pressing I need from him
<wolfspraul>
he thinks this is good, we have made working boards - done
<wpwrak>
did he also design the USB connection of the Ben ?
<wolfspraul>
no
<wpwrak>
so there's a meme in the EE community, or at least in our corner of it. wonder where it comes from ... at least intel's docs have nothing that suggests this sort of setup
<wpwrak>
the ben's is even a bit worse, because it has no bleed resistor. so i think that you could build up some high voltage differential between shield and the rest of the circuit. e.g., when the other end of the cable isn't connected to anything or if the shield is broken
<sujan>
hi, i've just copied nupdf to nanonote and when i tried to open the pdf file, it says cannot find config file, quitting... any suggestion? thanks
<qbject>
sujan: Let me think. I know I had that problem too at first....
<sujan>
ok thanks
<sujan>
following the readme.. i just copied over but there is that issue
<qbject>
Try using cd to get into the folder with nupdf in it, then select the PDF you want to read with an absolute path.
<qbject>
Hola, heberth. Hablo solo ingles pero estoy feliz esta aqui.
<heberth>
qbject:Â Â i don't spoke ingles
<heberth>
jejeje
<qbject>
heberth: hehe. Seremos amigos silencios
<heberth>
kristianpaul:Â Â sera nuestro traductor
<heberth>
qbject: ^
<qbject>
Anybody like William Gibson?
<qbject>
SiENcE: I'm confused. Is nLove for BNN up to the same status as the Caanoo version?
<SiENcE>
qbject, please ask bartbes, i think he is doing the nlove version for bnn
<SiENcE>
i also compiled a dingux one...so you might try this?
<qbject>
Gotcha. THanks!
<SiENcE>
but i dont know if dingux apps work on bnn
<qbject>
SiENcE: Some do, and I've stolen some Dingux libs to make even more run.
<bartbes>
qbject: no, I haven't updated the makefile
<qbject>
bartbes: Well I appreciate the time you and SiENcE have spent. Where would you announce if you did do that?
<SiENcE>
i have an strange issue with my toolchain. sometimes make doesn't find a code file. next i start make it finds the file. mh
<kanzure>
where did wolfgang get the money for the ben nanonote production runs?
<qbject>
kanzure: I'm curious about that, too.
<qbject>
He's usually in here on the opposite side of the clock. Might just ask him directly.
<qbject>
Maybe wpwrak knows.
<rafa>
anyone here have tested the wikireader plugin?
<kanzure>
qbject: will do.
<bartbes>
qbject: added it to my todo
<qbject>
bartbes: eeeexcellent. thanks very much.
<wpwrak>
kanzure: (money) as far as i know, personal savings
<kanzure>
hm
<qbject>
wpwrak: I assumed that he had prior electronics-manufacturing experience, given the smoothness that the Ben release has had, hardware-wise. Is that correct?
<kanzure>
i was thinking ben nanonote might have been leftover stuff from a previous gig?
<wpwrak>
qbject: oh yes. remember openmoko ? that's where he was before qihw.
<wpwrak>
kanzure: nope, it's based on a dictionary design that was commercialized by some chinese (?) company. wolfgang licensed it, made some changes, and had the company/companies that had made the dictionary produce it.
<wpwrak>
kanzure: that way, the usually quite high cost of just getting started (case design, prototypes, etc.) was saved.
<kanzure>
how did he convince them to use a copyleft license?
<kanzure>
also, please keep prepending kanzure: plz (i might miss messages otherwise)
<qbject>
wpwrak: you were on openmoko too, right?
<wpwrak>
qbject: yup
<wpwrak>
kanzure: not everything of the ben is under an open license. e.g., we don't have the cad files. for the electronics, dunno. i also don't know how much of that was redone.
<kanzure>
ah i see
<qbject>
wpwrak: but it seems like he does have permission to reverse-engineer and modify.
<wpwrak>
qbject: yup. besides, i don't think people in china care about such things a lot :)
<qbject>
wpwrak: Oh, that's a good point. I forget about Loongson. Nice that there is a society so confident of their own future success who still believe, essentially, in sharing.
<wpwrak>
qbject: maybe they just haven't had sufficient excess wealth for long enough to grow a large number of parasitic lawyers and lobbyists :)
<qbject>
wpwrak: makes me wish I knew more about their pre-revolutionary society.
<wpwrak>
;-)
<wpwrak>
(catching up with the irc backlog from the last days ...)
<wpwrak>
kristianpaul: for presentations, why not use latex ?
<wpwrak>
(sharism in spanish) why not some spanglish ? "sharismo" ;)
<qbject>
wpwrak: Well _I_ like it.
<qbject>
wpwrak: reminds me of "charisma"
<qbject>
wpwrak: and "fortissimo"
<wpwrak>
charisma is a good association
<qbject>
I agree.
<qbject>
(To be honest, "sharism" always seemed a bit...truncated to me.)
<kristianpaul>
wpwrak: latex right is nice i need learn it or get a friendly editor...
<rafa>
kristianpaul: for standar texts like articles, papers, letters you just need to edit templates ;)
<rafa>
kristianpaul: if you did some html doc before it is something similar : if you knoe <h1>tit</h1> <h2>tit2</h2> then in latex you can have something like \section{tit} \subsection{tit2}
<rafa>
if you knoe=if you know
<kristianpaul>
rafa: html yes i know
<kristianpaul>
interesting
<kristianpaul>
(need get some templates)
<rafa>
kristianpaul: check some examples.. I used pdflatex mydoc.tex to convert from latex to pdf directly
<rafa>
kristianpaul: what do you need to write?
<kristianpaul>
rafa: as wpwrak pointed, presentation slides
<rafa>
convert=pdf output from tex
<rafa>
kristianpaul: ah.. there are a lot of examples or templates for that :)
<kristianpaul>
looking now
<rafa>
kristianpaul: if you want I can give you an example :)