<wpwrak>
a build .. that means that xiangfu is hiding somewhere :)
<solrize>
wpwrak, i dunno anything about the mechanical stuff, i'm just a coder ;).  I think it's done by an outside contractor so i don't know what they use
<solrize>
but the first few units were 3d printed, then we got the molds made
<wpwrak>
would be interesting what Free tools people who do lots of such things use. particularly if these tools allow parametric designs
<solrize>
yeah i dunno what's out there... the reprap community might have some good stuff though
<wpwrak>
i saw some mention of AoI and lately OpenSCAD. AoI isn't parametric. OpenSCAD is, but has its shortcomings. e.g., no automatic filleting and chamfering. such tasks get a little boring if you have to do them with CSG ...
<wpwrak>
ah, they use FreeCAD. interesting. a while ago, it seemed a bit feature-poor, but maybe it has caught up since.
<solrize>
i haven't followed this stuff at all; i just found that on wp ;)
<solrize>
i'd like to know what's around in free vlsi cad stuff
<solrize>
mechanical stuff is less worrisome since the results are out where you can see them ;)
<solrize>
it would sure be nice to have a series of standardized enclosures for different types of products, that were available in small qty and a bit nicer than altoids tins ;)
<wpwrak>
naw, i think you need to be able to fully control the case-making process. for anything resembling CE devices, you need customized cases
<solrize>
hmm, yeah if you want to sell a lot
<wpwrak>
don't we all ? :)
<solrize>
i'd be happy with a standardized box and small quantity... worry about custom cases once there was revenue from the non-custom one
<wpwrak>
have you made any contact with google.org yet ?
<wolfspraul>
wpwrak: rejon is in Taipei!
<wpwrak>
wolfspraul: ah ! well, he could use e-mail :)
<wolfspraul>
I'm slightly depressed about the m1 rc3 run. early results indicate some smt problem, phew.
<solrize>
hey rejon
<wolfspraul>
rc2 was great on that end, and same smt shop, but rc3 looks suspicious. we know more in a few days (xray etc).
<wpwrak>
so no clear suspect yet ?
<wolfspraul>
so after I'm over digesting the reality, I probably need to expedite rc4
<wolfspraul>
no not yet, nothing much I can do now in support anyway
<wolfspraul>
I just want to get the product back into the shop asap
<wpwrak>
yeah, it's all in adam's able hands
<wolfspraul>
but my feeling is that the yield for this run will be bad
<wpwrak>
will you send a board to sebastien, for parallel debugging ?
<wolfspraul>
nothing needs to be debugged
<wpwrak>
all those high-current boards are a little suspicious indeed
<wpwrak>
(debugging) well, something's wrong :)
<wolfspraul>
yes but not design related, that would be a bad idea to defocus Sebastien with manufacturing issues
<wpwrak>
;-)
<wolfspraul>
Adam and suppliers will sort this out, it's just slightly depressing as I said, because rc2 was so good so the expectations are high.
<wpwrak>
so he didn't express interest in lending a hand there. okay. just thought two pairs of eye seeing more than one, etc.
<wolfspraul>
no that's not a good idea
<wpwrak>
but yes, if it's purely SMT, then there's not much he can do
<wolfspraul>
the boards need to stay in one playce, so they can be xrayed in bulk etc.
<wolfspraul>
it needs to be efficient
<wpwrak>
so rc4 may be very close. let's see about those gates then ...
<wolfspraul>
I need to sort out the economics too, this is all getting tough to finance (but will find a solution)
<wolfspraul>
bad yield is not good :-)
<wolfspraul>
but bad yield or not, I think my reaction has to be to expedite the next run and focus on finding the rc3 smt issues fast and then immediately making another run
<wpwrak>
yeah, sucks. that one will cost you :-(
<wolfspraul>
(no, not pushing about the gates, I'm just reporting rc3 thinking)
<wpwrak>
so you'd defer the gates until rc5 ?
<wolfspraul>
no, rc4 of course :-)
<wpwrak>
just checking ;-)
<wolfspraul>
first we need more rc3 data, which Adam will collect all of next week
<wolfspraul>
then hopefully we get a number of 100% pass boards on one side, and can turn them into full products with case etc. in parallel need to understand the root cause of the smt issues.
<wolfspraul>
and in parallel prepare rc4
<wolfspraul>
my own goal/expectation for 100% pass rc3 boards is now between 30 and 60
<wolfspraul>
but who knows, maybe there's a miracle and it will suddenly improve, but somehow I doubt that
<wpwrak>
100% pass ? hmm, 0% ;-)
<wpwrak>
>90% pass, maybe
<wolfspraul>
no I mean boards that are in sellable condition
<wolfspraul>
that means they 100% pass all tests, are considered stable etc.
<wpwrak>
i wonder how that through-hole dip thing works. sounds mildly scary to me
<wpwrak>
aah, i see
<wpwrak>
yes, that number sounds reasonable
<wolfspraul>
well. that would make rc3 my worst post-om run ever :-)
<wolfspraul>
we could learn something very interesting about potential smt failure next week, at least that. and expensive for me too :-)
<wpwrak>
so .. despite it all still not a new record in your life ? :)
<wolfspraul>
no, but I want to learn what happened
<wolfspraul>
4 (5) boards so far are 100% ok, so it's not a positioning problem I think
<wpwrak>
yeah, it happened just at a "good" time. just when you took a larger risk but still had no return from previous boards.
<wolfspraul>
must be reflow related?
<wolfspraul>
oh sure, and based on the assumption that rc2 was so good
<wolfspraul>
:-)
<wolfspraul>
something wrong with the solder material? something wrong with pcb surface?
<wolfspraul>
what else?
<wpwrak>
how's the financial side with the fab handled in such cases ? do they wave part of their fees ?
<wolfspraul>
exactly, that will also be interesting
<wpwrak>
did you change anything in the pcb making ?
<wolfspraul>
also I am wondering about the life expectancy of the boards that pass all tests
<wolfspraul>
I'm not saying I'm worried about that but I want to be able to have the knowledge and answer all such challenges/questions.
<wolfspraul>
need to ask adam [pcb making]
<wolfspraul>
I think it was all the same, same pcb shop, same smt shop
<wpwrak>
(life expectancy) good point
<wolfspraul>
maybe we are discussing ghosts and the sample size is just too small now
<wolfspraul>
but I looked at the rc2 results earlier today and the difference to the first rc3 results is rather striking :-)
<wolfspraul>
so yeah, for those who are curious, interesting times ahead :-)
<wolfspraul>
for my wallet, not so interesting times :-)
<wolfspraul>
it helps when the next armchair general is coming and talks about bom costs and things being 'expensive'
<wolfspraul>
and they will surely come
<rejon>
wpwrak: i contacted google ppl
<rejon>
playing out various scenarios
<rejon>
no results
<rejon>
or rather in-progress
<wpwrak>
rejon: great. it ought to take a while, i guess.
<wpwrak>
rejon: what are the scenarios you described ?
<wolfspraul>
I just want to learn as quickly as possible and move forward.
<wpwrak>
wolfspraul: maybe you need a nice metal case. then people will automatically assume a higher price ;-)
<wpwrak>
of course, then roh needs a beefier laser :)
<wolfspraul>
the diversity of the early rc3 results is puzzling to me
<wolfspraul>
something new in all those years
<wpwrak>
or he finally has to unwrap the cnc mill :)
<wolfspraul>
if we are lucky, we can trace it down to 1 (fixable) failure point. that'd be cool.
<wpwrak>
(rc3 diversity) indeed. sounds like too much solder paste or marginal positioning.
<wolfspraul>
I don't think positioning
<wpwrak>
adam could slaughter one board, cut power traces until he can identify one chip/area
<wpwrak>
a positioning error at the margin of tolerances would explain it
<wolfspraul>
I think the tool of choice is xray.
<wpwrak>
all you need is one of those many VDD balls to end up on one of those many GND pads. or vice versa.
<wolfspraul>
so the current plan is to fix2 and test another 20 boards, and then take all failed ones to an xray session
<rejon>
wpwrak: emails sent, but no decisions made on these things unless you face to fae with people...engineering that right now...taking some time
<wolfspraul>
wpwrak: have you done xray work?
<wpwrak>
xray is best, yes. whether it makes sense to try something else depends a bit on the queue before xray.
<wolfspraul>
it's very cool. I have heard that you could actually use the xray machine you can find at many dentists as well.
<wpwrak>
wolfspraul: no, didn't do xray yet. i saw some images but had a hard time figuring out what they show
<wolfspraul>
maybe you should ask your dentist next time whether you can get free or paid access to his xray in the evening :-)
<wolfspraul>
seriously
<wpwrak>
i think the expensive part of xray is getting a digital film
<wolfspraul>
you mean take pictures?
<wpwrak>
well, the dentist may just take the chemical route
<wpwrak>
that's inexpensive. but slow.
<wolfspraul>
no so what we do now in Taipei is we know an smt shop that has an xray device, and they are flexible to rent out access to it, by half hour
<wolfspraul>
no
<wolfspraul>
just a monitor
<wpwrak>
how much do they charge ?
<wolfspraul>
the xrays stay on, continuously
<wolfspraul>
not much, maybe 100 USD/hr or so
<wpwrak>
not too bad
<wolfspraul>
and they will give you an operator as well
<wpwrak>
rejon: alright. at some point in time we should discuss the scenarios
<wolfspraul>
so the total is: pack boards (1hr), drive there (1hr), operate xray & take pics (1hr), drive back (1hr)
<wolfspraul>
something like that
<wpwrak>
(give you an operator) sweet. so they don't want customers to radiation-poison themself by accident ;-)
<wolfspraul>
xray machines differ quite a bit in their performance, but I don't know much about that. just that there are cheap and expensive xray machines.
<wolfspraul>
no it's not that
<wpwrak>
and then brood over the results for the rest of the night :)
<wolfspraul>
it's super easy to operate
<wolfspraul>
but the problem is to zoom in quickly and get a sharp picture
<wolfspraul>
then you can 'roam' around the pcb, it's very nice
<wolfspraul>
you can take pictures of an entire pcb the size of m1 in 5 minutes, no problem
<wolfspraul>
just clik clik clik, 10 pics, then you got it all
<wolfspraul>
you can also go through vertically of course, but like I said the operator may be helpful in getting the positioning done fast
<wolfspraul>
but if you get access to your dentist's xray, you have time to play so that's not so important :-)
<wolfspraul>
I can definitely say that if you have an xray nearby, I would always prefer that over any electrical way of tracking down those impedance/short or cold soldering problems
<wolfspraul>
the only problem is how quickly accessible the xray is to you, if it comes with a big overhead then it's not good. but if not, it's far better than electrical searches.
<wolfspraul>
you could even effectively use it without taking pictures, by just writing down which pins you saw with a failure
<wolfspraul>
thinking this further, you can even attempt to do a hot-air fix right there, and put it back into the xray to see whether the bug is fixed :-)
<wolfspraul>
which I do not understand. but the prices of xray systems can go into the millions of USD, and I'm sure that's not just better recording software.
<wolfspraul>
the one at that SMT place costs 200,000 USD I think, or something like that (can ask again)
<wolfspraul>
actually I may have a picture and model number, one sec
<wpwrak>
yeah. may be that they use different energy levels
<wpwrak>
and they probably have some motor-controlled table such
<wpwrak>
(and now the link will go to the IRC log and hopefully boost the page rank a little more ;-)
<wpwrak>
hmm, should get ready to go shopping
<wpwrak>
will continue with the component research later ...
<kristianpaul>
about smt issues, you changed this time smt right?
<kristianpaul>
but also the run was not just straight like rc2? i meean the boards are been delivered by packages of ie 20
<kristianpaul>
i just wondering how smt internal  procedures could change because of it...
<kristianpaul>
i pointed the smt changes, becasuei dunno SMT/DIP Process Flow is same on all of then...
<wolfspraul>
kristianpaul: no all the same, same smt shop.
<kristianpaul>
hum..
<wolfspraul>
but in hardware, every piece is unique, every everything is unique.
<kristianpaul>
ok, i dont have more to say then :)
<wolfspraul>
that's why it's hardware
<wolfspraul>
so we see
<wolfspraul>
it will be exciting :-)
<kristianpaul>
sure :)
<wolfspraul>
by Wednesday we know more I think
<kristianpaul>
Wed, is full boards deliver?
<kristianpaul>
also, as i pointed before i dont see this like a whole run, i mean first 20 boards, then.. xx
<kristianpaul>
but i guess it was you desition, and this number should not affect internal SMT workflow, no matter the nummer, same grade of preparation i guess :)
<wolfspraul>
kristianpaul: no, it's always one run.
<kristianpaul>
ah, i was confused then...
<wolfspraul>
the pick & place machine is loaded with m1 parts, then one board is produced (pick & place, reflow)
<wolfspraul>
Adam checks that board, which all was perfect (of course)
<wolfspraul>
then the rest of the boards run through pick & place, reflow, manual or automatic optical inspection
<wolfspraul>
at that point it stops
<wolfspraul>
pick & place machine is unloaded and loaded for next board
<wolfspraul>
now... some more manual steps are done later, like DIP soldering, cleaning the pcb with a brush
<wolfspraul>
those can be done in batches sometimes, depends on staff load and shifts etc.
<wolfspraul>
and Adam insisted on early delivery, so they delivered 48 pieces first. in the end every piece is unique and at some point that unique one is finished, right? :-)
<wolfspraul>
the only question is whether you see the uniqueness, or you wait until you get them all "at once" :-)
<roh>
rjeffries: you are aware that besides people in the us nobody uses F?
<roh>
well.. and belize.
<roh>
;)
<kristianpaul>
those imperial measures ;)
<rjeffries>
roh I thought of that when I posted. As a global citizen I really should have converted to C for my qi-friends
<rjeffries>
I've visited Belize. Nice place. Go to Cay Caulker if you ever get the chance.
<wpwrak>
roh: you can make fun of them now, but they'll be vindicated in the zombie apocalypse, when all the dead from past millennia crawl out of their graves and all measure distances in paces, feet, and worse
<wpwrak>
"to use iOS developers" nice typo. but there's an "ab" missing :)
<roh>
well.. patent-traders need to be burnt to the ground same as investment banks if society wants to survive...*shrug*
<wpwrak>
yeah, nothing new there
<roh>
wpwrak: *nod*
<roh>
.oO(trousers... sunglasses... back to the lab!)
<wpwrak>
i still have that idea to create a very long island that spans the atlantic, declare an independent nation, then force all airplanes entering my airspace to briefly land and the passengers to go through immigration. pretty mcuh what the US do with international-to-international connections. anyone involved in patent trafficking will be arrested and judged. my nation will have the death penalty for selected heinous crimes :)
<wpwrak>
alas, after sealand, they tightened the requirements for what can become an independent nation, so that plan will be a little more difficult to implement. i think new volcanic islands would still be eligible. so i shall have vulcanologists among my minions when i'm ready to make my move.
<kristianpaul>
i think you'll need to build a vessel first, and considering sea is not even well mapped in osm, you may have luck and find interesting islands :-)
<qi-bot>
[commit] Werner Almesberger: components/: changed output pin tyoe of 74x1g07_*.lib to open collector (master) http://qi-hw.com/p/kicad-libs/21c356e
<wpwrak>
kristianpaul: i suspect that all currently existing islands have been found by now. satellites are kinda handy for that ... hmm, i'll need a few myself. equipped with death ray guns, of course. BWAHAHA ! :)
<kristianpaul>
lol
<gnutoo>
gnoon
<gnutoo>
oops
<gnutoo>
I'll relog in kde
<kristianpaul>
I'll do as sebastien, i'll ignore not verified functionallity and make code works upon that..
<kristianpaul>
heck, wishbone cross domain clock, after adding a slight delay in software it dint crash anymore
<kristianpaul>
cross domain clock bridge*
<kristianpaul>
~3.26 Msps (4 bytes/sample), not bad, and milkymist dint crashed, take that metastabillity !
<GitHub155>
[milkymist/gps-sdr-testing] namuru debug signals enabled, fixed extra bit in irq reg, and removed uncesary commented code - Cristian Paul Peñaranda Rojas
<GitHub155>
[milkymist/gps-sdr-testing] fixed port with wrong direction - Cristian Paul Peñaranda Rojas