681Topic1

Open Discussion of Design Flow: 

Open Discussion of Design Flow Today’s task: Design an ASIC that will drive a TV cell phone Exercise objective: Importance of codesign

Layout Styles: 

Layout Styles Full-custom Gate-array Standard-cell Macro-cell FPGA (Field Programmable Gate-Array)

Full-custom: 

Full-custom Complete control over transistor and interconnect dimensions (within design rule constraints) Produces Optimized Design (density, power, performance) Circuit Designers create application-specific building blocks Technology Provider (foundry) provide SPICE/HSPICE transistor models, parasitic extraction tools Models are used to drive transistor sizing/layout constraints Layout technician creates graphics from design schematic Top level down/Bottom level up Continual verification of design as it becomes more defined

Design Rule Constraints: 

Design Rule Constraints Minimum Spacing Between metal lines (varies per layer) Line width Transistor channel lengths Active area Via stacks (Check from work)

Full-custom Disadvantages: 

Full-custom Disadvantages Most time-consuming Most error-prone Highest NRE (non-recurring expense) Design time Layout time Mask costs Longest time to manufacture

Gate-array Layout: 

Gate-array Layout Transistors pre-placed, fixed in size Personalized by metal routing Fastest to manufacture Lowest mask cost Lends itself to automated placement and wiring

Gate Array: 

Gate Array Vdd Gnd Horizontal Routing Channel Vertical Routing Channel Sea of Gates: Routing Channels removed, route at higher metal layers

Gate Array Example: 

Gate Array Example Vdd Gnd A A B B Vdd Vdd Schematic A A B B Out Out

Gate-array Disadvantages: 

Gate-array Disadvantages Non-optimized spacing Limited transistor sizing options Density Performance Power Wiring blockages/inefficiencies Excess circuitry

Standard Cell Layout: 

Standard Cell Layout Design partitioned into cells of standard height Power and Ground (Power grid) wiring preset Technology provider supplies libraries of pre-designed cell elements for usage (utilize varying numbers of cells) Primitives (NAND, NOR, etc.) Storage Elements (DFF) Libraries can be tailored to specific applications (e.g., low power vs. high performance) Requires full manufacturing sequence Typically automated place and wiring

Standard Cell Layout: 

Standard Cell Layout Routing Channel Routing Channel Feed-through cell Note uniform height

Our Cell Library: 

Our Cell Library Need specifics on library students will be using

Standard Cell Disadvantages: 

Standard Cell Disadvantages Cell height restrictions limits cell library contents Full set of masks Longer manufacturing times

Macro-cell Layout: 

Macro-cell Layout Library elements provided by technology supplier (e.g., foundry) Elements can be of varying heights and widths Richer variety of library elements (IP friendly)

Macro-cell Disadvantages: 

Macro-cell Disadvantages Similar to Standard-cell in length of manufacturing times, mask costs Placement and wiring more complex Pre-layout of power grid more difficult, may not be possible

FPGA: 

FPGA Field Programmable Gate Array Array of logic blocks (Configurable Logic Blocks CLB) Switchable interconnect resources Wire segments of varying lengths Programmable switches that connect logic resources to wire segments Final user sets switches (CLB and interconnect) Immediate Use (“zero” fab time) Minimal expense Great for hardware prototyping

Slide17: 

FPGA: Virtex-II Architecture Virtex™-II architecture’s core voltage operates at 1.5V I/O Blocks (IOBs) Configurable Logic Blocks (CLBs) Clock Management (DCMs, BUFGMUXes) Block SelectRAM™ resource Dedicated multipliers Programmable interconnect © 2005 Xilinx, Inc. All Rights Reserved For Academic Use Only

Slices and CLBs: 

Slices and CLBs Each Virtex-II CLB contains four slices Local routing provides feedback between slices in the same CLB, and it provides routing to neighboring CLBs A switch matrix provides access to general routing resources CIN COUT COUT Local Routing CIN SHIFT © 2005 Xilinx, Inc. All Rights Reserved For Academic Use Only

Simplified Slice Structure: 

Slice 0 D Q CE PRE CLR D Q CE PRE CLR Simplified Slice Structure Each slice has four outputs Two registered outputs, two non-registered outputs Two BUFTs associated with each CLB, accessible by all 16 CLB outputs Carry logic runs vertically, up only Two independent carry chains per CLB © 2005 Xilinx, Inc. All Rights Reserved For Academic Use Only

FPGA Disadvantages: 

FPGA Disadvantages Least efficient use of silicon/wiring resources Limited size options Limited performance Not good for high volume applications If used for prototyping, still may have significant changes when migrate to higher performance design and package solution