Now That We Have An Error Rate

[chuckr]

Some questions that could only be managed AFTER testing, here:

First, now that we have a history, how much elapsed time will testing each unit (each one that's tested) take? 2nd, what's the pct. of the 3895 units will have to be fully tested? (I'm looking to see if I can take those two items and dope out for myself how long the testing will have to take). Will that test automation that's been talked about work for final testing, and could that help any?

Thanks. I do REALLY hope you folks will decide to write up your experiences, and what were the surprises you hit.

 

[droneb]

im not with the OP Team,
but been working sometimes with some QC teams


usually a system doesnt get all the full testing because it is time consuming and high cost.

so in order to keep a threshold of faulty boards (not so notorious fault like "it doesnt turn on")
they pull up a sample on each production and if they get more than an x ammount of faulty boards (calculated by using normal,Z component and range).
they will have to make a further check in order to keep the faults under the thresholds.

heres an example using the data we already have

ok using binomial distribution we get the number of probable faulty boards in a x sample.

for samples of 20 (arbitrary selected)
an expected value of np=20 * (2 / 105)= 0.380952381 ~boards
and variance of npq=20 * (2 / 105) * (103 / 105) = 0.373696145

if they use a safety range of 95% this means +1.65 times the variance (using normal dist transformation)
(0.373696145 * 1.65) + 0.380952381 = 0.99755102 boards
so at most they should expect that on samples of 20 no more than 1 should be faulty

 

[kloplop321]

gotta love statistics. I understand it (since I am taking a university level class of it)

 

[borgqueenx]

I dont want a failed pandora here. Its already so much evil in such a small box
are pandoras tested before they are shipped? And batteries precharged?

 

[dflemstr]

And batteries precharged?

I hope not... Or, well, they have to be charged somewhat to make the Lithium ions chargeable in the first place (A Lithium ion battery "dies" when it's completely uncharged as you all know), but a full charge... Why?

 

[lulzfish]

And batteries precharged?

"Nah, he's been waiting like a year for this thing, he can wait until tomorrow when the battery's charged."
"Yeah. Also, it would be poor form to ship an uncharged device that users can't play with immediately (unless they plug it in)"

No, OP isn't Apple or whoever does that kind of stuff. It will probably be around 40% charged or whatever the battery's dropped to.

Also, inb4 let's start 2 polls, 3 discussions, and a flamewar speculating whether OP will charge the batteries or not.
Or let's leave it to them.

 

[fiori_musicali]

Also, inb4 let's start 2 polls, 3 discussions, and a flamewar speculating whether OP will charge the batteries or not.

Lmao

 

[gp32rich]

103 out of 105 passed. The other two may come up after another reflow but sometimes they are just dead and get tossed aside. This whole time has been spent working on the ones that have been functional and building the test jig with pins that probe the PCB. We've also been focusing efforts on adding more features to the software so that an automated application can be run on top of the OS to check all of the hardware functionality with as little human interaction as possible.

I hope this means that all boards will have some (at least rudimentary) testing through the JTAG port if not full burn in

Expect few failed delivery units!

 

[droneb]

103 out of 105 passed. The other two may come up after another reflow but sometimes they are just dead and get tossed aside. This whole time has been spent working on the ones that have been functional and building the test jig with pins that probe the PCB. We've also been focusing efforts on adding more features to the software so that an automated application can be run on top of the OS to check all of the hardware functionality with as little human interaction as possible.

I hope this means that all boards will have some (at least rudimentary) testing through the JTAG port if not full burn in

Expect few failed delivery units!

CMON!, what kind of product has a 100% no fail rate? theres none

and thats because most of the products get damaged either while on distribution (shipping) or while being retailed.
after that, well they can still have a working unit that has a hidden failure (i got my laptop and 1 month later i had to call and change the optical drive)

and plus, nowadays RoHS will shorten the life on every new electronic device so i would expect more on warranty products

 

[chuckr]

The actual fact of the way testing was to be accomplished was defined (at least) months ago, so the statistics course wasn't needed (while I took it too, I really never liked stats). However, since we DO have the actual failure rate now established on a large enough test sampling to be relied upon, I wanted to have the ACTUAL testing ratio given (not just the general fact, all the factors which contribute now having been established. So, what's the number? Given the failure modes and the established testing procedures and automation, what's the per unit test time required? What were the failure modes?

Some of us (who maybe aren't so experienced in hardware manufacturing) are curious, aren' t you?

 

[gp32rich]

103 out of 105 passed. The other two may come up after another reflow but sometimes they are just dead and get tossed aside. This whole time has been spent working on the ones that have been functional and building the test jig with pins that probe the PCB. We've also been focusing efforts on adding more features to the software so that an automated application can be run on top of the OS to check all of the hardware functionality with as little human interaction as possible.

I hope this means that all boards will have some (at least rudimentary) testing through the JTAG port if not full burn in

Expect few failed delivery units!

CMON!, what kind of product has a 100% no fail rate? theres none

and thats because most of the products get damaged either while on distribution (shipping) or while being retailed.
after that, well they can still have a working unit that has a hidden failure (i got my laptop and 1 month later i had to call and change the optical drive)

and plus, nowadays RoHS will shorten the life on every new electronic device so i would expect more on warranty products

What? Where'd you get 100% no fail from?

What I am expecting is that all units that leave the factory will be fully working,
so apart from blundering posties I'm expecting that my unit will fire up first time

 

[droneb]

103 out of 105 passed. The other two may come up after another reflow but sometimes they are just dead and get tossed aside. This whole time has been spent working on the ones that have been functional and building the test jig with pins that probe the PCB. We've also been focusing efforts on adding more features to the software so that an automated application can be run on top of the OS to check all of the hardware functionality with as little human interaction as possible.

I hope this means that all boards will have some (at least rudimentary) testing through the JTAG port if not full burn in

Expect few failed delivery units!

CMON!, what kind of product has a 100% no fail rate? theres none

and thats because most of the products get damaged either while on distribution (shipping) or while being retailed.
after that, well they can still have a working unit that has a hidden failure (i got my laptop and 1 month later i had to call and change the optical drive)

and plus, nowadays RoHS will shorten the life on every new electronic device so i would expect more on warranty products

What? Where'd you get 100% no fail from?

What I am expecting is that all units that leave the factory will be fully working,
so apart from blundering posties I'm expecting that my unit will fire up first time

i cant recall why wuss dat, (maybe because pointing something so obvious sounded sarcastic.)

anyway, ill wait until the first panda is born before making any calculations