Page 42 - ChipScale_Jan-Feb_2020-digital
P. 42
Figure 2: HDAP connectivity checking without a
design intent.
Figure 3: Running HDAP LVS without die pins.
no design intent with which to compare), it’s
still a valuable open/short checking process Assembly-level LVS/alignment “interface physical geometry,” not the
that can be used without a source netlist (yet). verification. In an HDAP, alignment “interface logical pins.” As for solving these
Figure 2 shows the block diagram of this verification is equally important as LVS, pin names/net names mismatches between
simple open/short checker. especially when rotation or scaling is the SoCs and HDAPs, EDA companies are
Package/interposer LVS without die required (e.g., a die is shrunk by 10%, supplying novel approaches to accommodate
pins. In the next case, the verification data rotated 90°, then placed on the package). such issues without creating false LVS
includes both a package design database and HDAP designers must ensure that, violations. For example, HDAP designers
a source netlist, so HDAP designers should even after such processing, there is still can denote such inconsistencies as waivers
be able to run a true LVS verification. sufficient overlap between the package on input. Figure 4 shows the assembly-level
While their first thought is probably to use bumps and die bumps. There are two LVS/alignment with die pins flow.
the same LVS tools and methodologies levels of alignment verification possible,
used for SoC LVS, thinking a little further depending on what data the HDAP
reveals that that approach is not suitable for designer can obtain.
this specific process. The concept of LVS Geometrical die interface only. If
in an SoC is dependent on the existence HDAP designers only have access to the
of “devices” in both the SoC layout design die interface (bumps) GDSII layer, they can
database and the source netlist. These still perform both alignment and package
devices can be active (transistors) or passive LVS verification. Designers can use EDA
(resistors, capacitors, etc.). Because most tools to automatically place the die(s) on
2.5D-IC and FOWLP designs don’t include top of the package/interposer and verify Figure 4: Full assembly-level sign-off LVS/
alignment, including die pins.
any devices (just metal routing), SoC LVS the alignment (as required). However, if an
tools simply aren’t practical. extra pin exists in the SoC design database
The more viable solution is to use or in the package design database, it is not Summary
assembly-level LVS tools that, by default, captured when using this flow. As HDAP technologies become
account for the non-existence of devices Geometrical die interface + logical die mainstream, designers require qualified
in packages. However, these tools require pins. In this flow, all interface information design rules from foundries/OSATS,
more than just the package layout design (geometry and pins) from the die(s) is a nd rel iable sig n- of f automat ed
database for assembly-level LVS—they available. HDAP designers have all the verification flows from EDA companies.
also require the layout design databases for required data for comprehensive HDAP The development of package ADKs is
the dies. But there’s a solution! As long as LVS/alignment verification, and can run addressing the first requirement. While
the input spreadsheet netlist includes die this flow as a sign-off flow. Designers can current HDAP verification flows are far
bump (x,y) coordinates and pin names, the validate all possible violations: 1) assembly- from mature when compared to well-
functionality in assembly-level LVS tools level LVS; 2) assembly-level alignment; and established SoC verification flows, EDA
can be enhanced to automatically generate 3) extra/missing pins. companies are providing automated tools
“placeholder” die bumps (representing From a connectivity perspective, and flows that can take into account
the die). This “die placeholders” design however, there is an issue of which the various levels of data availability,
database can then be used in the assembly designers should be aware when including while still enabling HDAP designers to
design database to successfully run HDAP both the die(s) and the package in assembly- perform useful and valuable HDAP LVS/
LVS using only the package layout design level LVS. If the SoC team uses a different alignment verification flows.
database and input spreadsheet netlist. A naming methodology for the die pins/nets
block diagram for the HDAP LVS without than the package team, the SoC layout References
die pins flow is shown in Figure 3. design database may contain a die pin called 1. J. Ferguson, T. Ramadan, “Assembly
Although the flows described above “A,” while in both the system source netlist design kits are the future of package
provide obvious value to the customer and the package design database, the pin is design,” 3DInCites, Aug. 3, 2015,
(capturing any physical shorts/opens in the referred to as “A_B.” Such disparities result https://www.3dincites.com/2015/08/
package design database), the full HDAP in many false LVS violations. This issue was assembly-design-kits-are-the-future-
assembly still can’t be guaranteed to function irrelevant in the geometrical die interface of-package-design-verification/
successfully. What if there is a misalignment only flow, because the only data the HDAP 2. J. Ferguson, “Why do we need
between a die and the package in the HDAP? designer imported from the die was the assembly design kits for packages?,”
40 Chip Scale Review January • February • 2020 [ChipScaleReview.com]
40