08 3D Integration 2012
08 3D Integration 2012
Section France
3D-IC Integration
Agenda
Chapitre Franais Section France
Introduction
Process overview
Akasaka, Y., and Nishimura, T., "Concept and Basic Technologies for 3-D IC Structure IEEE Proceedings of International Electron Devices Meetings, Vo. 32, 1986, pp. 488-491.
Source IBM
http://www.research.ibm.com/journal/rd/526/knickerbocker.html
Moores law by scaling conventional CMOS involves huge investments. 3D IC processes : An opportunity for another path towards continuing the scaling, involving less investments. Like for conventional CMOS, infrastructures are needed to promote 3D-IC integration, making it available for prototyping at reasonable costs.
CMP annual users meeting, 25 January 2012, PARIS
Monolithic
Distributing a whole system across several tiers
Heterogeneous Multi layer 3D-IC TSV integrated 3D-IC TSV integrated 3D-IC face to face
n o i t a r g e Int
Die to Die Integrated package Multi-Chip Module
Discrete
Assembly of Known Good Dies
CMP annual users meeting, 25 January 2012, PARIS
Discrete : 3D packaging, stacked dies, 1- Design a whole system. 2- Split it in subsystems. 3- Place the subsystems as predefined Known Good Dies (IPs). 4- Determine and place the interfaces in between. 5- The system is done Monolithic : 3D-IC Integration 1- Design a whole system. 2- Split it in subsystems. 3- Determine and place the interfaces in between. 4- Generate and Place the subsystems in between the interfaces. 5- The system is done
Here comes the difference : The key for a true 3D-IC Integration
CMP annual users meeting, 25 January 2012, PARIS
Chapitre Franais
Section France
3D-IC Applications
Chapitre Franais
Section France
"Implementing a 2-Gbs 1024-bit -rate Low-Density Parity-Check Code Decoder in ThreeDimensional Integrated Circuits" Lili Zhou, Cherry Wakayama, Robin Panda, Nuttorn Jangkrajarng, Bo Hu, and C.-J. Richard Shi University of Washington International Conference on Computer Design, ICCD, Oct. 2007
R8051 CPU
80MHz operation; 140MHz Lab test (VDD High) 220MHz Memory interface
IEEE 754 Floating point coprocessor 32 bit Integer coprocessor 2 UARTs, Int. Cont., 3 Timers, Crypto functions 128KBytes/layer main memory
CMOS Image Sensor (Sensor + Processor + Memory) 3D stacked Memories (Flash, DRAM, etc) Multi-cores Processor + Cache Memory NoC (Network on Chip) Processor + DRAM + RF + MEMS + Optical communication +
CMP annual users meeting, 25 January 2012, PARIS
Design Methodology
Chapitre Franais Section France
The more Design Automation is performed on the 3rd dimension, the more is the 3D-IC Integration.
Processor + DRAM + RF + MEMS + Optical communication
NoC Multi-Processors + Memory Pixel Sensor (HEP) Simple Imaging Sensor Memory Stack
System Complexity
CMP annual users meeting, 25 January 2012, PARIS
Chapitre Franais
Section France
CMC-CMP-MOSIS Collaboration
Join forces for the technical support, and dedicate the roles for each partner.
Make easier the tech support for local users respectively by each local center.
Because there is no standard for the 3D-IC integration, it is urgent to setup an infrastructure making possible a broad adoption of 3D-ICs. That will have a beneficial effect on prices, more frequent MPW runs, and more skilled engineers.
CMP annual users meeting, 25 January 2012, PARIS
CMC supporting Canadian Customers CMP supporting European Customers MOSIS supporting US Customers Each may support other locations
Chapitre Franais
Section France
Tezzaron Process Flow for TSV and DBI (using Via Middle process)
Chapitre Franais Section France
Tezzaron Process Flow for TSV and DBI (using Via Middle process)
Chapitre Franais Section France
Cu
Cu
Cu
Cu
Cu
Cu
Cu Cu
Cu
Cu
Cu
Resulting 2-tier 3D-IC integration TSV and DBI (Via Middle Process)
DBIs continuing the stacking
Chapitre Franais Section France
Source Tezzaron
Resulting 2-tier 3D-IC integration TSV and DBI (Via Middle Process)
Chapitre Franais Section France
Source Tezzaron
CMP annual users meeting, 25 January 2012, PARIS
ce a l np
3D-IC Users
In discussion
In p rep ara tio n
Critical mass will allow frequent MPW runs and low pricing
CMP annual users meeting, 25 January 2012, PARIS
Chapitre Franais
Section France
The Design Platform is modular. It has all features for full-custom design or semi-custom automatic generation design. PDK : Original PDK from GF + (TSV / DBI) definition from Tezzaron Libraries : CORE and IO standard libraries from ARM Memory compilers : SPRAM, DPRAM and ROM from ARM 3D-IC Utilities : Contributions developments embedded in the platform Tutorials, Users setup.
All modules inside the platform refer to a unique variable, making it portable to any site. The installation procedure is straightforward. Support of CDBA and OpenAccess databases.
CMP annual users meeting, 25 January 2012, PARIS
chrt13lprf_DK009_Rev_1D (Version issued in Q2 2011) assura calibre cds_cdb cds_oa doc eldo hercules hspice prep3DLVS skill spectre strmMaptables_ARM strmMaptables_Encounter
assura: FILLDRC LVS QRC calibre: 3DDRC 3DLVS DRC FILLDRC calibreSwitchDef
Cadence CDBA 5.1.41 Cadence OA 6.1.4 Design-rules manuals Simulation Models 3D-LVS Preprocessor Design-kit Customization
HEP labs contributing with Programs, Libraries, and Utilities. All included in the Design Platform
DBI (direct bonding interface) cells library. (FermiLab) 3D Pad template compatible with the ARM IO lib. (IPHC) Preprocessor for 3D LVS / Calibre (NCSU) Skill program to generate an array of labels (IPHC) Calibre 3D DRC (Univ. of Bonn) Dummies filling generator under Assura (CMP) Basic logic cells and IO pads (FermiLab) Floor-planning / automatic Place & Route using DBIs, and TSVs (CMP) Skill program generating automatically sealrings and scribes (FermiLab) MicroMagic PDK (Tezzaron/NCSU)
CMP annual users meeting, 25 January 2012, PARIS
DBI
CMP annual users meeting, 25 January 2012, PARIS
Choosing 2D or 3D LVS
3D-LVS fully functional both for the CDB and OA. Graphical interface for the NCSUs preprocessor utility for merging 2 tiers GDSII. Corrected LEF file for the M6 / two thick metal option. Assura 2D LVS is functional. Walk-Through Encounter tutorial with both DBI and TSV scripting for automatic P&R. Master file browsing the documentation.
3 package options : Complete design-platform TDP = (TDK + libraries + compilers) TDK only TDK + libraries with reduced layout views (memory blocks generated on request with reduced layout views)
CMP annual users meeting, 25 January 2012, PARIS
MicroMagic 3D viewer
Chapitre Franais Section France
MicroMagic 3D crossection
Chapitre Franais Section France
System Level Partitioning 3D Floor-Planning DBI, TSV, IO placement Automatic Place & Route
Design exploration at system level Design exploration at the physical level DBI, TSV, and IO placement & optimization Cells and blocks place & route can be done tier by tier To be done for each tier, then combined for back-annotation to the 3D top level system
Extraction, Timing Analysis Physical verification 3D DRC, 3D LVS Dummies Filling Final 3D DRC
- DBIs Placement - TSVs Placement - Obstructions on TSVs - Clock routing - Final routing
- Encounter natively refuses to make the routing for pins on DBIs. - Custom scripts solved the problem. Its a workaround. - The resulting layout is compliant to the Tezzaron DRC, LVS etc
Saving the floor plan for the bottom tier, and apply it for top tier so the automatic Place & Route run the placement and routing taking into account the DBI locations. The place & route for both tiers is optimal for timing, buffer sizing and power performance.
Chapitre Franais
Section France
HAMBURG
GERMANY
19.7 mm2
ORSAY
FRANCE
25 mm2
ISEA
Toulouse
FRANCE
6.25 mm2
Conclusion
A very collaborative work has been achieved and still ongoing between the parties CMC, CMP, MOSIS, FermiLab, Tezzaron, HEP Labs, NCSU.
Chapitre Franais
Section France
Design Platform 2011q2v3 in use since June 2011. (Release 2012q1v1 will be available in February, 2012)