In reply to a comment I had, Mr. Mitchell (if it's really Mitchel, sorry)
of that famous Si-Valley board shop; Zycon, asked:
> Jerry,
>
> Compaq has also realized an order of magnitude improvement in their
> assembly DPPM by going to BGA. Would that cover the cost difference?
>
> [log in to unmask]
I had said:
>Seems like everything I pick up these days has an article about BGA's and
>how they will revolutionize the world. From the talk at the shows, the
>magazine articles, and they equipment brochures I've read in the last year
>or so, you'd think they were flooding the market.
>
>I'm still skeptical.
(snipped off some stuff here)
To answer your question:
No way, not here. In round numbers our direct cost to assemble _and_ test a
board assembly is perhaps $20. If I have an initial failure rate due to
shorts of say 5% (which is exaggerated) and a rework cost of $10 (including
re-test) and the use of a BGA totally eliminates that, then our cost is:
$20 plus ($10 x 0.05) = $20.50.
I believe the above is realistic. IMO, the device needs to cost us the SAME
or nearly so, else we would say "please give us good old QFP's".
Many people use inflated overhead rates to pump up cost savings. This might
be explained as follows:
cost of initial assembly and test (assume all pass)
$20 direct
70 overhead (at 350%)
----
90
cost of troubleshoot/repair
$10 direct
35 overhead (350%)
----
$45
So you could say that the cost of the "bad" process is $90 plus 5% times
$45 or $92.50. Therefore, they save $2.50 per unit to move to the BGA and
they only pay $2 more. If you make 10 thousand units per day, you likely
cannot deal with rework effectively, because it is erratic and stops
product flow in continous assembly. They have some justification.
The numbers could be even higher, but if any company building PC's is
spending more than $20 (direct) on a motherboard for labor, then they will
soon be bankrupt or bought by the Taiwanese, renaming themselves after an
old radio company.
So, even if you give me cosmic magnitudes of improvment, we'd rather keep
our $15 per part, thank you.
I am impressed by what the Compaq people say about their good process
results with BGA's, make no mistake. I believe them. They are breaking the
ground in this area, and will help to popularize the package. The presence
of a major buyer will help make this a more common (and economical)
package. Currently, though, QFP's are very established and automated within
the IC packaging industry, it will be a while before all those leadframes,
wire bonders, and transfer molding machines are used up and worn out. But
even then, how many >200 pin devices go on the average board?
The reliablility of a BT resin/glass package as large a 35 mm or so is
prolly better than a plastic QFP 304 on standard glass/epoxy boards. There
are good reasons to adopt BGA's. Motorola, IBM and several other
semiconductor houses are waving the flag bigtime.
It's an interesting discussion topic. We will likely stand on the sideline
and cheer until the price moves into sight. Us Texans are like that, except
maybe in Houston. ;-)
And meanwhile, I'm gettin' kinda tired of hearing about the plaguey things.
Seems to me, everyone keeps telling me over and over all about them 'cause
they want to sell me something.
Jerry Cupples
Interphase Corporation
Dallas, TX
http://www.iphase.com
p.s. look for the "Zc" logo on selected 14 layer 155 megabit/sec FDDI
interface cards (those cost us way more than $20 to make)...
|
