cherrycyanide
Member
Thanks for the update. Im expecting pandora next year.
Oh. heh. :unsure: *sheepish*WizardStan said:It was a general query, not necessarily aimed at you.
MWeston said:Unfortunately, right now I am seeing a failure rate of at least 25%.
We need to get the failure rate down to about 3% to get production going full speed.
------------------
For reference the xbox 360 came out with a :angry: shipped :angry: failure rate of 30% in 2007 and is reportedly at 16-17% failure in 2008. These numbers are based on retailers reports.
I think I can cheerfully wait a few weeks to get my toy done right the first time instead of wishing the pandora folks were as eager to push out alpha quality hardware as microsoft
Belthos.
Can you say.....
Red Ring of Death?
"Very generous of you to let even Chip write it for them, especially after it has been announced that Chip is going to do the PR for them... "7zark7 said:I cannot see what harm a small update would do (heck chip can even write it for them).
7zark7 said:If it was taken in any other way it was not intended
No worries, it wasn't taken wrong, hence the smiley. It just read a little funny.
Wow! How many layers in the board did it take to get connections to all of those balls? Are you doing buried vias and vias in the pads?chiark said:Just to give some insight into this...
10mmx10mm 0.4mm pitch BGA with 529 connectors against a dime. A dime is 17mm in diameter... Slightly smaller than a pound coin, or a euro coin.
Five hundred and twenty nine connectors in less space than a pound coin.
Yeah, that sort of puts the difficulty of assembling it into context!
belthos said:------------------MWeston said:Unfortunately, right now I am seeing a failure rate of at least 25%.
We need to get the failure rate down to about 3% to get production going full speed.
For reference the xbox 360 came out with a :angry: shipped :angry: failure rate of 30% in 2007 and is reportedly at 16-17% failure in 2008. These numbers are based on retailers reports.
I think I can cheerfully wait a few weeks to get my toy done right the first time instead of wishing the pandora folks were as eager to push out alpha quality hardware as microsoft
Belthos.
Can you say.....
Red Ring of Death?
Remember that Microsofts problems were failures in the field caused by design problems, not manufacturing problems. I am quite sure that every single one of those XBox 360s worked before it was put in the box and shipped.
This is not at all the same type of failure that MWeston is talking about.
Beagleboard is 6 layer, but their target is 4. Looking at the gerber files from beagle I don't see any buried vias, but yes vias in pads are used. For fun you can check the gerbers for yourself...DAP said:Wow! How many layers in the board did it take to get connections to all of those balls? Are you doing buried vias and vias in the pads?
Trevor Bradley said:The frightening thing for most of the "concerned" camp has been the lack of information, rather than the content of that information. (I've worked in too many places where a lack of information meant that something very very bad was going on. After a while you begin to assume that lack of information means the worst.)
I was never in the "concerned" camp (I have a lot of confidence in the people behind the Pandora), but I agree that this is usually the case.
thanks so much for this update! it boggles my mind as a non-expert to think how soldering a 0.4mm BGA might even work. i just hope that the contracts you have with the producer put the quality issues firmly into their share of problems, at least financially.MWeston said:Okay, here's a little update just from the assembly side of things. Some people are going to bitch and moan and raise hell for anything I say other than "shipping today!", but I will share some insight into the development process of a new project. This is basically as Squidge explained it before.
The factory will not do all 4000 at once, even if they had absolutely every part in front of them. This is a brand new design using technology that still isn't adopted by all assemblers. That is the POP design. Also, the 0.4mm pitched parts tends to cause troubles and requires very specific and repeated procedures to be followed to guarantee success. Basically, we have to do this in small runs. I tested a few boards of the new design, then batches get larger like 20 boards then 100, then 400 and then maybe the balance but the procedure has to be nailed down before they will pump out thousands.
Unfortunately, right now I am seeing a failure rate of at least 25%. That doesn't really scare me as it doesn't mean the boards are junked. It just means the stencil thickness, reflow oven temperature profile and dwell times and maybe even the visual inspection stages have not been tweaked yet. Reflowing the boards and doing some touch ups brings them back to life and then they work fine. They just have to keep trying things slowly and figure out all the tough areas so that those get extra attention and the areas of the boards that never fail can get a quick inspection. A final burn in test of each board will confirm that everything works.
We need to get the failure rate down to about 3% to get production going full speed. The manual labor on patching or re-running 3000 boards is financial suicide so you can see why things are not going as we would have liked. The biggest mistake any manufacturer can do is ship boards without feeling 100% confident in the hardware. It's common practice to use customers to test your beta software *cough* *Archos* *cough* but never to test hardware.
The other good news is that once this process is nailed down, we shouldn't see such headaches and delays for future batches. Everything is new right now and even though I have the most experience with this stuff in the group, I have never worked with such tiny, tiny little parts before. Going from 0.5mm to 0.4mm pitch BGA parts is a bigger leap than the numbers would suggest. Any fellow Engineers on the board might know what I'm talking about.
Very, very small pixies with very, very, very small soldering irons.hch said:it boggles my mind as a non-expert to think how soldering a 0.4mm BGA might even work.
chiark said:Very, very small pixies with very, very, very small soldering irons.hch said:it boggles my mind as a non-expert to think how soldering a 0.4mm BGA might even work.
Hope this helps
http://www.instructables.com/id/Toaster-Ov...-Soldering-BGA/
Poor man's version of bga soldering.
Belthos
I'm sure the Pandora Dev Team and the Factory know how tiny the OMAP Chip is since it was choosen, so I think they will finaly manage all this Stuff.chiark said:Just to give some insight into this...
10mmx10mm 0.4mm pitch BGA with 529 connectors against a dime. A dime is 17mm in diameter... Slightly smaller than a pound coin, or a euro coin.
Five hundred and twenty nine connectors in less space than a pound coin.
Yeah, that sort of puts the difficulty of assembling it into context!
The picture wasn't to update the dev team on the challengefusion_power said:I'm sure the Pandora Dev Team and the Factory know how tiny the OMAP Chip is since it was choosen,