Joerg-Neo900 has quit [Killed (orwell.freenode.net (Nickname regained by services))]
DocScrutinizer05 has quit [Disconnected by services]
DocScrutinizer05 has joined #neo900
pagurus has quit [Ping timeout: 240 seconds]
pagurus has joined #neo900
Airwave has joined #neo900
l_bratch has quit [Ping timeout: 240 seconds]
freemangordon has joined #neo900
freemangordon has quit [Client Quit]
freemangordon has joined #neo900
l_bratch has joined #neo900
xmn has quit [Quit: Leaving.]
chainsawbike has quit [Ping timeout: 246 seconds]
chainsawbike has joined #neo900
infobot has quit [Ping timeout: 260 seconds]
paulk-collins_ has joined #neo900
weebear has joined #neo900
weebear has quit [Quit: Page closed]
weebear has joined #neo900
tsuggs has quit [Ping timeout: 264 seconds]
jonsger has joined #neo900
weebear has quit [Ping timeout: 260 seconds]
Pali has joined #neo900
chomwitt has joined #neo900
jonsger has quit [Remote host closed the connection]
jonsger has joined #neo900
ecloud has quit [Ping timeout: 256 seconds]
ecloud has joined #neo900
jonsger has quit [Remote host closed the connection]
jonsger has joined #neo900
jonsger has quit [Ping timeout: 264 seconds]
xmn has joined #neo900
atk has quit [Quit: Well this is unexpected.]
cc___ has joined #neo900
atk has joined #neo900
rhombus has joined #neo900
<rhombus>
I recently acquired an N900 in pristine condition, as I am really stoked about this project.
<rhombus>
The radio silence and the difficulties with the GTA04 run have been sobering, though.
<rhombus>
I'm also a little bummed about the Gemalto chips -- it seems there's no satisfying combination of network support. I'd want European LTE and UMTS with GSM quad-band support, as I'm often in NAm.
<rhombus>
The PHS8 might be a poor compromise.
<rhombus>
But I guess you can't always get what you want, and if the project dies on the vine, it won't matter anyway.
<rhombus>
How can I help the project succeed?
<rhombus>
Donating is the obvious way, but it's feeling like more of a gamble than before, especially now that it's 2017 and we're hearing of challenges
<rhombus>
which I suppose are normal in any project of this nature
<rhombus>
Freedom isn't free :)
rhombus has quit [Quit: KVIrc KVIrc Aria 4.3.1, revision: 6250, sources date: 20120701, built on: 2016-09-17 11:52:23 UTC http://www.kvirc.net/]
<MonkeyofDoom>
it is tragic that cycle times on open hardware are so long
mzki has quit [Quit: leaving]
chomwitt has quit [Ping timeout: 240 seconds]
mzki has joined #neo900
chomwitt1 has joined #neo900
<DocScrutinizer05>
a pity rhombus vanished so quickly
<DocScrutinizer05>
I have the perfect answer to >>How can I help the project succeed?<< : Provide manpower! I can't keep up with the workload, particularly on the web update and communications front. And more so since the project had massive impact on my health and thus performance, manifesting in an extreme form since last new year
<DocScrutinizer05>
on the SoC warp soldering issue: I had a chat with Nikolaus and the problem boils down to: I have a possible solution worth trying, but that's a method not usually used in standard industry production processes (so complicated and in consequence more expensive (maybe a 10 bucks or more per device, ballpark guesstimate), and Nikolaus has no resources left to do further experiments to try that method
<atk>
What is this SoC warp soldering issue?
<DocScrutinizer05>
the putative solution in telegram: solder and glue (with ->underfill) the SoC flimsy "PCB ring" around the CPU die to the rigid RAM PoP chip. AKA pre-stacking
<Joerg-Neo900>
the "problems" with pre-stacking: the fab lines are not made to deal with prestacked components
<Joerg-Neo900>
neither are there tools to do the prestacking
<Joerg-Neo900>
and applying underfill to the stack is probably a whole new variant in this special case, not done before in industry
<atk>
So for my non-EE mind, the thing happening here is that the ram for the cpu is stacked on the CPU
<Joerg-Neo900>
so, while by mere logic this should be a sure cure for the issue, *doing* it is another issue
<Joerg-Neo900>
yes
<Joerg-Neo900>
that's PoP aka Package on Package
<atk>
And presumably normally what happens is the CPU is pick and placed on the board, that gets reflowed, then you place the ram on that and reflow again?
<Joerg-Neo900>
no, usually you stack all and reflow once
<atk>
Right
<atk>
so what is pre-stacking?
<Joerg-Neo900>
reflow CPU and RAM, and then place the stack on PCB
<atk>
Ah, I had imagined that might be the case
<Joerg-Neo900>
placer engines can't handle stacks
<atk>
and the underfill trick you say is to reflow the CPU and RAM, then put the underfill in-between the two layers and then place the stack on the PCB?
<Joerg-Neo900>
yes
<atk>
Interesting
<atk>
How resistant is this underfil to temperatures?
<Joerg-Neo900>
underfill is basically cyanoacrylate glue
<atk>
as in, does it get soft?
<atk>
Ok, so pretty much not affected.
<Joerg-Neo900>
no, there are variants that are sufficiently heat resiliant
<atk>
But couldn't it still crack because of the difference in expansion of the two chips?
<Joerg-Neo900>
the trick is to apply the underfill to the presoldered stack
<Joerg-Neo900>
((crack)) to be tested. I'd hope it can't
<Joerg-Neo900>
I'd guess if the forces that occur during heating are so large that they can crack underfill, nobody would use pre-stacking since same forces easily crack the solder points or pads
<atk>
Who invented this bizarre "POP" design? It seems very ... prone to issues..
<atk>
But yes, 30% yield was it? that seems pretty unacceptable.
<Joerg-Neo900>
yep
<atk>
This theoretical method of yours - how much would it cost to test?
<Joerg-Neo900>
of course for Neo900 we'd try PoP soldering on special 'low-cost' test PCB, that 'only' cost us a CPU, a RAM chip and a few bucks for PCB, to verify yield before we start building hundreds and see 2/3 paperweights fall off the end of assembly line
<Joerg-Neo900>
test is "cheap", in tests you place components manually anyway
<Joerg-Neo900>
the 'killer' is you probably have to do same in production since the method is not really compatible with fab line placer tools
<atk>
What's the worst case for cost on using this method in production?
<Joerg-Neo900>
as stated above: a maybe 10 or 20 bucks per device
<atk>
I think that's not too bad
<Joerg-Neo900>
plus one-time expense for tooling and setup
<atk>
And it seems like this might produce more reliable and rugged devices
<Joerg-Neo900>
no idea how much that would be, I don't think it's astronomically high
<Joerg-Neo900>
yes, that's the idea :-)
<Joerg-Neo900>
((tooling, setup, overhead)) you need pre-stacking trays like shown in http://www.ti.com/lit/an/swpa182c/swpa182c.pdf#19 (or similar), you need somebody manually place a few hundred stacks into those and apply paste/flux and reflow, then have the pre-satcked components placed back into plastic trays or whatever, and finally find out how to place those on PCB, if placer engine can handle them or not
<Joerg-Neo900>
ooh, I forgot the underfill. This needs to get applied carefully so it soacks into the capillary gap beween CPU and RAM but doesn't spill on the CPU bottom BGA balls
<Joerg-Neo900>
soaks*
<Joerg-Neo900>
having a 'bath' with 0.2mm high of underfill fluid in it and then dipping the *sides* of the pre-satck into that bath may do the trick
<Joerg-Neo900>
other possible approaches involve a thin 'wick' or a razorblade to apply from top
<atk>
The sides? Wouldn't that leave some underfill on the bottom of the stack?
<Joerg-Neo900>
well, the fluid must not be high enough to reach the bottom BGA balls
<Joerg-Neo900>
well, sometimes you simply can't think out of the box :-/ Why not stick the complete top RAM chip into underfill fluid, deep enough so the fluid just reaches the gap between chips but not the higher bottom balls?
<Joerg-Neo900>
I think there's a number of ways to test how to apply underfill, just nobody tested them yet
<atk>
That's what I was thinking, but wouldn't air just get trapped in the middle?
<Joerg-Neo900>
one of those should turn out to be feasible
<Joerg-Neo900>
to avoid air trapping, drown one side first, then lower the opposite side too
<atk>
I was thinking why not just use a syringe with a fine needle, then mark on the needle how deep it is to get half way under the chip, then get the syringe between the balls (I'm pretty sure they make needles this thin) and apply it there.
<atk>
That would be a manual process though
<Joerg-Neo900>
there's no syringe that small
<atk>
How small are the gaps?
<Joerg-Neo900>
we're talking 0.4mm BGA pith here, the clear distance between balls and the gap between chip surfaces in in the <0.1mm range
<Joerg-Neo900>
pitch*
<atk>
hmm, the thinnest syringes made are apparently 0.0826 mm
<atk>
wait, that's id
<Joerg-Neo900>
you hardly caninsert any wire in there, no matter how fine
<atk>
0.1842 od, nevermind
<Joerg-Neo900>
http://susepaste.org/80972434 might be to scale, with the upper ball pitch being 0.5mm (not 0.4, that's lower iirc)
<atk>
If the gap is so thin, if the underfill has a suitably low viscosity it should just soak in even if you dip it at an angle with the back facing the liquid.
<atk>
Ah, so there's divets in the bottom chip to hold the balls
<Joerg-Neo900>
yep
<Joerg-Neo900>
and yes, the underfill is supposed to be very low viscosity so it soaks into capillary gaps
<Joerg-Neo900>
usually you apply it to the corner between chip and pcb after assembly/reflow
jonsger has joined #neo900
<Joerg-Neo900>
atk: if you want to help, you could sum up this conversation on tmo. Would you?
<atk>
I'll need to make an account but I can give a summary and link to the full log
<Joerg-Neo900>
sure thing. Every help appreciated
* Joerg-Neo900
afk. BBL. sorry
<Joerg-Neo900>
though, you might face some limitation with a fresh account
<atk>
If anything I'll just summarise it and send you the text
<Joerg-Neo900>
that's fine with me, I can post it in your name
infobot has joined #neo900
<Joerg-Neo900>
~wb
<infobot>
thx
arnaudj has joined #neo900
<cc___>
DocScrutinizer05: I know how unhealthy it can be to work too much, I hope your health issues are not too bad
<cc___>
Joerg-Neo900: forgive my newbie question but there's something I don't understand in all this : how are you responsible of the soldering process ? I thought you were only designing the device and the soldering was done by a company
<cc___>
somebody else should answer actually I don't want to waste joerg's time on my noobish question
<ravelo>
Joerg-Neo900 is running the company
<ravelo>
if he only gets 30% yield he loses 70% of his own money
<cc___>
ok
<ravelo>
so this problem should be resolved
<DocScrutinizer05>
Thanks! and yes, originally the concept was: GolDeliCo hass the 'shop' and manages the hw manufacturing, together with their partner fab (http://www.global-components.de). This changed when Nikolaus bailed out of project management and I founded Neo900 UG, so everything dropped on my feet entirely
<cc___>
and what is the solution nokia used ? the thread says they didn't waste 70% of the production. Did they use the method joerg was talking about ?
<DocScrutinizer05>
Now Nikolaus reports he can't reproduce the building of GTA04 which Neo900 is basically based on, regarding processes etc
<DocScrutinizer05>
yes
<DocScrutinizer05>
sorry, afk again
<DocScrutinizer05>
my soup gets cold
<ravelo>
Gta04 was already built a few times, every time Nikolaus resolved the problems
<ravelo>
now he has also quite low volume, only 30 pieces
<arnaudj>
enjoy your soup Joerg!
<ravelo>
so no money left to make more tests
<ravelo>
in total already a few 100 where built
<ravelo>
(gta04's)
<ravelo>
Mahlzeit :)
<arnaudj>
hi everyone
<cc___>
so does that mean that the open hardware community can't afford to make phones as thin as the ones produced by the industry ?
<cc___>
I mean with the same thickness/power ratio
<MonkeyofDoom>
define "can't afford"
<MonkeyofDoom>
they certainly can't have the same profit margins (or even the same prices for the same kind of products, nonprofit)
<MonkeyofDoom>
as I understand it scale is almost everything
<MonkeyofDoom>
(developer-power, incl. hardware and mechanical developers, is the rest)
<atk>
DocScrutinizer05: that's what I've got, I think it's short yet informative, and makes all the necessary references for anyone who wants to read more into it
<atk>
It would be appreciated if someone else here quickly proof read it just in case I made some silly spelling mistake somewhere.
<MonkeyofDoom>
%30 -> 30%
<atk>
ah, yes
<atk>
I thought that looked wrong
<MonkeyofDoom>
"production," -> "production;", caps on "Samsung"
<atk>
Great, anything else?
<MonkeyofDoom>
, -> ; on "reflowed," as well
<MonkeyofDoom>
"has been filling" -> "is filling"
<MonkeyofDoom>
otherwise mostly fine, no hard errors at all