| Packaging's Value Added:
Greater Functionality for Integrated Circuits |
|
By Dr.
Tom Di Stefano Contributing Editor
'Designers
are finding ways to redesign DRAM chips by employing wafer
stacking to integrate functions that are distributed across
a chip stack in new ways.'
|
 |
Wafer-level packaging began, as the name implies,
as the packaging of integrated circuits on the wafer before dicing
it into chips.
The rationale is transparent: At the wafer level,
you reduce backend costs by packaging chips in a batch, thereby
simplifying the process. Rather than performing one-at-a-time mechanical
processes for handling and wire bonding chips, we can now process
tens of thousands of bonds on a wafer at one time.
Wafer-level packaging promises to do for packaging
what the integrated circuit did for the transistor half a century
ago!
New
Functionality
This story does not end with cost reduction.
As was the case with integrated circuits, wafer-level packaging
allows us to add completely new functionality to the chip. The strategic
allure of wafer-level packaging is the capability of integrating
additional interconnects, passives, optical links and as yet unforeseen
elements into the final packaged device.
It is difficult to project the details, but
this capability will find applications that will add new chapters
to the semiconductor saga.
Wafer stacking is emerging as a technology with
the potential to impact the industry greatly. This new dimension
to wafer-level packaging has not gone unnoticed. Several semiconductor
industry giants are developing stacked chips as a way to produce
DRAM or higher density and performance at lower cost.
These programs go well beyond 3-D packaging
of memory chips - that has been relegated to the sidelines.
Chip
Redesign
Designers are finding ways to redesign DRAM
chips by employing wafer stacking to integrate functions that are
distributed across a chip stack in new ways.
Links from one chip to another have minimum
capacitance and inductance compared to conventional packaging. Chip-to-chip
communication in the stack is greatly simplified. Off-chip drivers
for interchip communications now look more like on-chip circuitry.
Large and power hungry CMOS drivers are not
needed for these links. Lower cost nMOS may be adequate for many
of the wafers in the stack, where CMOS can be used on only those
layers that require I/O from the stack.
We could go on about ways to rearrange memory
arrays to make use of 3-D connectivity. Lower power dissipation
of the nMOS layers allows stacking more nMOS layers without exceeding
thermal limits.
The possibilities are enticing, and hard to
predict beyond the first few steps. But that's what makes the field
exhilarating.
Wafer stacking is not without challenges. Because
the rewards are so great, you can be sure that teams of engineers
are working to break though the roadblocks. Assembling chips in
a wafer stack presents testing and yield problems.
Is a stack of chips yieldable? How do we connect
front-to-back? Must we thin the wafer to a membrane? How do we do
test and repair? What about wafer-level burn-in? Is it necessary?
What about logistics? These and other questions are being addressed,
and they must be answered before wafer stacking becomes a reality.
DRAM products using wafer stacking are a long
way from a market reality, but the development in this activity
bears careful watching.
Without question, wafer stacking is one more
in an expanding list of wafer-level packaging technologies that
go beyond simply "packaging."
|
Dr. Di Stefano is a chip-scale packaging
pioneer, a prolific author and inventor, and the founder of
Tessera. He is currently president of Decision Track, Mountain
View, Calif., and may be reached at tdistefano@decisiontrack.com.
|
|