Title: The Importance of Adopting a PackageAware Chip Design Flow
1R I O D E S I G N A U T O M A T I O N
The Importance of Adopting aPackage-Aware Chip
Design Flow
2Package Aware Chip Design Value Proposition
IO, Bump, PIN placement optimization
- IC/Package Prototype and co-optimization
exploration - SI analysis, planning (Power planes analysis)
- Optimize IO/Bump/Pin placement
- Bump pattern/Tile Synthesis with RDL route
- Escape route for bumps with minimum package
layers
Technology
Lower Cost
Quicker to Market
TTM
-
- Save weeks to months in design time
- Early exposure to entire Package/Pad/IO
- Ability to handle changes late in design cycle
- First Pass success for Chip/Package/PCB
Cost Savings
- Chip Die area reduction (10)
- Package/PCB area/layers reduction
- NRE cost of third-party (300K-500K per
project) - Fewer resources per chip (1 per project from 6)
3Chip/Package Convergence Toward Single Pass
Verification
Design Convergence
- Bottom-up planning w/DEF floorplan
- Major Steps
- Create power groups and voltage domains
- Map power groups/regions to new core plan
- Place final I/O cells and route RDL
- Route final package/s
- Extract/simulate I/O to ball interconnect
- Value Results
- Export chip to PR tools for extraction and
verification - Export package file for final routing
- Top Down Planning w/netlist
- Major Steps
- Generate I/O netlist, power groups and area
blocks - Split power groups and assign regions
- Synthesize I/O rings and place I/O cells
- Value Results
- I/O placement for core planning
- Detailed package routing
- Initial package design start
- Early Prototype phase
- Major Steps
- Compile I/O ring
- Explore package options
- Value Results
- Initial die size and I/O sequencing
- Package options and total packaged/chip cost
estimate
4Unified Data Model Enables Cross Domain Synthesis
- All elements of the interconnectare captured and
modeled - Cross domain co-optimization
- Repository for golden interconnect between
chipand package - Connectivity management from start of design
through final verification
5 Technology Advances
Core Limited Die
Die/Package with planar routing
I/O Cells
Pad limited Die
Core area
I/O Cells
Open Space
Core area
Open Space
- Package optimization
- Exploration of die/package I/O plan minimizes
layer count and enables accurate package cost
estimates - Package level power distribution planning
facilitates balanced power deliver to die and
minimizes surprises - Concurrent design results in timely 1st package
delivery - Intelligent bump/ball assignment enables planer
routes that frees space for plane stapling and
decaps for better power distribution
- Die optimization
- In core limited die, greater optimization of
I/O pad ring can increase core size - In pad limited designs, variable pitch bump
patterns can reduce overall die size - Bi-directional flow supports I/O placement based
on core or package driven constraints - Signal/Power/Ground ratio automation
- 65/45 nm equals more pad limited designs
6 Chip/Package Interconnect
- Interconnect synthesis must take entire
interconnect into consideration to enable
chip/package co-optimization - Package route feasibility is key to up front
final cost estimates - Package tradeoffs can lead to significant cost
reduction
7Complete Interconnect Analysis
Early driver to PCB extraction and analysis
avoids surprises at verification
8Design Initialization
Initial chip/package
CO-DESIGN FLOW
Design Initialization
Tiling
Global I/O Placement
- Full integration with
- OpenAccess enables interface to third party PR
tools - Die input
- LEF for standard cells and hard macros and I/Os
- DEF for netlist
- IBIS I/O driver models
- Support for existing tile libraries if
presentbut not required - Package input
- TechFile from package design team or
sub- contractor - Net constraints
- Voltage domains, differential pair and
bus constraints
I/O Ring Synthesis
Detailed I/O Placement
Package Escape Route Ball/Pin Assignment
9 Prototype Global I/O Placement
CO-DESIGN FLOW
Global I/O Placement
Design Initialization
Global I/O Placement
I/O Ring Synthesis
- Must support fixed or partially placed floorplans
- Global placement is flyline based and establishes
first pass at I/O sequencing - Automatic grouping and regioningof I/O
- Ever greater need for prototype flow to support
accurate Request for Quote estimates for both die
and package - Manual placement capability to fine tune initial
placement
Detailed I/O Placement
Package Escape Route Ball/Pin Assignment
10I/O Ring Synthesis
I/O Cell Floor Plan
CO-DESIGN FLOW
Design Initialization
Global I/O Placement
- I/O Planning synthesizes a set of templates and
bump macros for each power group or area I/O
block - Cell instances define the I/O ring where I/O
cells are to be placed - Need to ensure that there are an adequate number
of I/O sites to support the number of signals on
the die - Calculate the number of power/ground sites
required to meet the signal/power/ground ratios
(SSO) - Need to ensure that there are an adequate amount
of bumps to support the number of signals and
associated power and ground - Calculation of bump spacing must be based on
package routing constraints and enable DRC clean
escaping - Full support for area I/O and SerDes blocks
I/O Ring Synthesis
Detailed I/O Placement
Package Escape Route Ball/Pin Assignment
11Detailed I/O Placement
Detailed Placed I/O
CO-DESIGN FLOW
Design Initialization
I/O Sequence w/RDL
Global I/O Placement
- Final step in I/O planning process
- Automatic placement of I/O cells establishes
connectivity between cells and bumps - Must honor user defined placement constraints
- Need to insert the proper number of power and
ground cells based on established SPG ratio
requirements - Must assign bumps in cells as well as bump macros
to I/O nets - Finally, this step must perform RDL routing from
I/O pads to bumps to establish an accurate
interconnect
I/O Ring Synthesis
Detailed I/O Placement
Package Escape Route Ball/Pin Assignment
12Package Escape Route and Ball/Pin Assignment
Routed Package
CO-DESIGN FLOW
Design Initialization
Escape Routing
Global I/O Placement
- Automatic escape layer assignment
- Must honor techfile design rules when escaping
die - Must automatically escape power and ground
netsto assigned power/ground layers on the
package - Escape signal nets to assigned package layers
- Route escape routes using min DRC rules
toperimeter of the die - Assign nets to balls to establish line of
sight,single layer planar routing - Must support any previously assigned balls
- Must honor net constraints such as differential
pair when assigning balls to nets - Route from die perimeter to balls using relaxed
line/spacing rules
I/O Ring Synthesis
Detailed I/O Placement
Package Escape Route Ball/Pin Assignment
13Prototype Timing and SI Analysis
CO-DESIGN FLOW
Design Initialization
Embedded SI Analysis
Global I/O Placement
- RLC estimation is performed on final routed nets
- Parasitics values need to be generated for both
primary and coupled nets - On-chip parasitics need to be generated basedon
the per-unit value for each layer in the
techfile - Package estimates can be based on empirical
models and package vias are modeled as lumped RLC
values - Integrated parasitic (R,L,C, Mutual and coupling)
extraction - Load models and terminations can be used forthe
PCB - Timing analysis for delay, over/undershootand
ringback
I/O Ring Synthesis
Detailed I/O Placement
Package Escape Route Ball/Pin Assignment
14Prototype PI Analysis
CO-DESIGN FLOW
Embedded PI Analysis
Design Initialization
- Power analysis support for time domain and
frequency domain analysis - For packages characterization, a field solveris
needed to compile a set of templates based on the
package stack-up and mesh structure - Equation based models can be used to quickly
estimate the capacitance effect in the package - Voltage domains should be automatically cut and
modeled based on electrical and physical
constraints - Adaptive meshes need to be generated for each
voltage domain as described by the plane
geometries and associated traces and shapes for
rapid analysis - Quick analysis capability enables exploration
- Reports and plots can be generated
Global I/O Placement
I/O Ring Synthesis
Detailed I/O Placement
Package Escape Route Ball/Pin Assignment
15Design Export
CO-DESIGN FLOW
Example of design design exported to package and
IC tools
Design Initialization
Global I/O Placement
- Final chip netlist along with associated macro
cell libraries need to be exported with cover
macrosand RDL to any chip tool supporting
LEF/DEF - DRC clean final package escape routing would
thenbe exported directly back into package
design tool - Must also include bump to ball netlist
- Constraints defined in such as differential
pairsmust also be automatically captured and
passedalong down stream - To support existing flows, netlist, including all
I/O-bump-ball connectivity should be exported in
spreadsheet format - Final verification must be done using 3D
analysisand final verification tools
I/O Ring Synthesis
Detailed I/O Placement
Package Escape Route Ball/Pin Assignment
16Key Value Proposition in Concurrent Design
- Lighting fast prototype capability
- More exploration/what if analysis on chip
floorplan, package substrate and design rules
(results in optimal die size and package) - Reduce cost
- Saves engineering resources and results in lower
development cost - Smaller die size and less expensive package
- Correct by construction and co-optimization
- IO ring synthesis capability guarantees package
routability - One pass package design flow eliminates
iterations and reducestime to market - Early (prototype) signal and power integrity
analysis (no more surprises at verification)