Introducing the Xilinx Spartan Series High Performance, Low Cost FPGAs with on- chip Select-RAM.

Introducing the Xilinx Spartan Series

High Performance, Low Cost FPGAs with on-chip Select-RAM

Xilinx Spartan Families Complete High Volume FPGA Solution

Spartan: 5 Volt with up to 40K System Gates SpartanXL: 3.3 Volt with up to 40K System Gates Xilinx Alliance and Foundation Software

Process technology leap UltraDenseTM 0.5m for 5 Volt Spartan Family UltraDenseTM 0.35m for 3.3 Volt SpartanXL Family

XC4000 heritage Powerful software tools Most successful FPGA architecture

No compromises Performance, RAM, Cores, and Low Price

Basic Architecture

Array of CLBs surrounded by perimeter IOBs

Special functions in corners

Spartan Meets ASIC RequirementsLow Price

Unit Cost

Development Cost

Unit Cost

Development Cost

NRE

LostOpportunity

Cost of Ownership

Per Unit ($)

FPGAASIC

FPGA Cost of Ownership Advantage No test vectors required Limited or no simulation Automatic Place and Route Re-spins in hours not months Faster Time-to-Market No NRE

1x

2x

10x

Total Cost Management

Leading edge process technology Smallest die size

Streamlined feature set Synchronous single/dual-port Select-RAM Popular serial configuration modes JTAG support

Optimized manufacturing test flow for low cost

Focused package offering Highest volume plastic PLCC, VQ, TQ, PQ, BGA packages

0

1

2

3

4

0.6u FPGA2LM

0.6u FPGA3LM

0.5u FPGA3LM

Ultradense0.5u FPGA

AssemblyTestSilicon

Total Cost Management

Relative Cost

1993

1995

1996

Spartan Series addresses all aspects of cost

1997

Majority of cost is back end (assembly, test, overhead)

Spartan die are pad limited

Packages Focus on popular, low-cost packages

Highest-volume ASIC plastic packages

Only six packages: PC84 (05/10) VQ100 (05/10/20/30) TQ144 (10/20/30) PQ208 (20/30/40) PQ240 (30/40) BG256 (30/40)

Optimized PQ208 pinout Adds more I/O

UltraDenseTM 0.5 Process

ChipTransistor gates 0.5- allows 3.3 V supply

All other features 0.35- small size- low capacitance- performance- low power

Combines 5 V operation with 0.35 benefits

UltraDenseTM Process

Core

core-limited

I/O padsGate count determines

die size

Spartan Die Size for High I/O packageNearly Equivalent to Gate Arrays

pad-limited

Core

I/O pads

I/O countdetermines

die size

Advanced Process Technology eliminates the “RAM compromise”

Die Sizewithout RAM

1.0

Logicand RAM

I/O pads Die Sizewith RAM

1.0

Logic

Empty, wasted die

area

Technology advances have reduced die size faster than pad technology

Xilinx Spartan Series

5 Volt -> XCS05 XCS10 XCS20 XCS30 XCS40

3.3 Volt -> XCS05XL XCS10XL XCS20XL XCS30XL XCS40XL

System Gates 2K-5K 3K-10K 7K-20K 10K-30K 13K-40K

Logic Cells 238 466 950 1368 1862

Max Logic Gates 3,000 5,000 10,000 13,000 20,000

Flip-Flops 360 616 1120 1536 2016

Max RAM bits 3,200 6,272 12,800 18,432 25,088

Max I/O 80 112 160 192 224

Performance 80MHz 80MHz 80MHz 80MHz 80MHz

No Compromises: Performance, RAM, Cores, and Low Price

Spartan Series Footprint Compatibility

5 Volt XCS05 XCS10 XCS20 XCS30 XCS40

3.3 Volt XCS05XL XCS10XL XCS20XL XCS30XL XCS40XL

PC84 PC84

VQ100 VQ100 VQ100 VQ100

TQ144 TQ144 TQ144

PQ208 PQ208 PQ208

PQ240 PQ240

BG256 BG256

Highest volume ASIC plastic packages

Footprint compatible in common packages

Xilinx Introduces First Fully Programmable

System Solution

First FPGA Architecture Designed for Intellectual Property

System Level Design Trend

DSP

CustomLogic

BusI/F

RAMI/F

High-DensityHigh-PerformanceCustom Device

PCI

Scratch PadSRAM

PC Board

High SpeedSynchronous

DRAM(Mbytes)

VideoData In

Frame DataBlock

SelectRAMMemory(Kbytes)

BlockSelectRAM

Memory(kbytes)

DistributedSelectRAM

Memory(bytes)

Line Data

DistributedSelectRAM

Memory(bytes)

Pixel Data

Video PixelProcessing

Function(logic)

ProcessedVideo Out

Real Time Video Processor

Virtex FPGA

Hierarchy of RAM provides efficient and very high bandwidth data processing

Virtex Family Overview

0.25um, 5 layer metal processDensity: 50 thousand to 1 million system gates

Performance 100+ MHz performance (3-4 levels) 160 MHz system performance

First device in 2Q98 250,000 system gates One million system gate device by end of 1998

Functional Block Diagram

CLB Segmented routing

SelectI/OPins

DistributedSelectRAMMemory

BlockSelectRAMMemory

PLL

66 MHz PCI SSTL3

Vector BasedInterconnectdelay=f(vector)

2 LCs 2 LCs

CLB

4 InputLUT

RegisterCarryand

Control

I3I2I1I0

O

WI DI

DCE

CLK

Q

CO

CI

4 InputLUT

RegisterCarryand

Control

I3I2I1I0

O

WI DI

DCE

CLK

Q

CO

CI

PR

RS

PR

RS

Polarity of all control signals selectable

Fast arithmetic and multiplier circuitry

Optimized for synthesis

Virtex Configurable Logic Block

Segmented Routing Interconnect

3-STATE BUSSES

SWITCH

MATRIX

2 LCs 2 LCs

CA

RR

Y

CA

RR

Y

CLB

CA

RR

Y

CA

RR

Y

Fast local routing within CLBs

General purpose routing between CLBs

Fast Interconnect 8ns across 250,000

system gates

Predictable for early design analysis

Optimized for five layer metal process

Block RAM

Configure as: 4096 bits with variable aspect ratio

8-32 blocks per device

True dual-port, fully synchronous operation Cycle time <10 ns

Flexible block RAM configuration 5 blocks: 2K x 10 video line buffer 1 block: 512 x 8 ATM buffer (9 frames) 4 blocks: 2K x 8 FIFO 9 blocks: 4K x 9 FIFO with parity

WEAENACLKAADDRADINA

DOA

DOB

RAMB4

WEBENBCLKBADDRBDINB

Virtex FPGA Performance 100+ MHz internal speeds

155 MHz SONET data stream processing 100+ MHz Pipelined Multipliers 66 MHz PCI

100+ MHz system interface speeds

without PLL with PLL

Tco (output register) 6 ns 3.5 nsTsu (input register) 3 ns 3 nsTh (input register) 0 ns 0 nsMax I/O performance 110 MHz 160 MHz

Virtex FPGAs Deliver World-Class Performance!

Today’s Gate Array

Gate Array

XC5200 Series

XC5200 Family Overview

Flexible FPGA Product Family: - 5 Devices Ranging from 3,000 - 23,000 usable gates- 5-Volt and 3-Volt Devices

Low cost, process-optimized architecture Xilinx FPGA Innovations:

- VersaBlockTM

Logic Module- VersaRing

TM I/O Interface

6 Levels of Interconnect Hierarchy Performance Up to 50MHz Robust Feature Set:

- Carry Logic - 3-state buffers- Cascade Chain - 4 Global Nets- JTAG Logic - Slew rate control

Flexible XC5200 Feature SetXC5200 Feature BenefitArchitecture-optimized 0.6u process Smallest die size - lowest cost

VersaBlockTM

logic module Versatile logic PLUS abundant local routing

VersaRingTM

I/O interface High logic utilization AND increased pin

assignment flexibility

Simple logic cell resources Efficient synthesis results

Carry logic Fast arithmetic functions

Cascade chain Efficient wide-input functions

Logic cell feed through Logic cell = logic and routing at the same time

Independent logic cell I/Os Easy access to logic resources

Boundary Scan logic Enhanced board / system testability

3-state buffers On-chip bussing

4 global buffers Fast, low skew signal distribution

Slew rate control High performance and reduced noise

XC5200 Architecture Overview

XC5200 first FPGA to combine: 0.6um CMOS SRAM process Triple-layer metal (TLM) New, process-optimized

architecture

Result: FPGA Efficiency Breakthrough

Most gates / mm2 More routing for less cost

Architecture Highlights: VersaBlockTM logic module VersaRingTM I/O interface General Routing Matrix (GRM)

Abundant Routing with VersaBlockTM

VersaBlock equals:— Configurable Logic Block (CLB)

– 4 identical Logic Cells (LC0-LC3)– 4 3-state buffers

— Local Interconnect Matrix (LIM)– 100% local connectivity– 24 bi-directional nodes connecting to GRM– Up to eight outputs (4 tri-statable)– Up to 23 inputs

— High Density Direct Connects– Fast paths to neighboring CLBs– 4 direct connects to and from neighbors

Result: Logic + abundant local routing— Minimizes local routing congestion— Granular and symmetrical structure

XC5200 Configurable Logic Block

Configurable Logic Block (CLB) 4 Identical Logic Cells 20 independent inputs 12 independent outputs 2 independent 5-input functions

Logic Cell (LC0 - LC3) Function generator, register, & control logic Independently usable F & FD Programmable flip flop or latch option Fast carry logic Cascade chain Independent feed-through capability

Abundant Routing Resources

6 Levels of Hierarchy

General Routing Matrix:

1. 10 single length lines

2. 4 double length lines

3. 8 longlines per row and channel

VersaBlock:

4. Local Interconnect Matrix (LIM)

5. Direct connects to all neighbors:

- 4 uni-directional lines in

- 4 uni-directional lines out

6. Logic cell feedthrough

SingleLengthLines

DoubleLengthLines

DirectConnects

Longlines

LocalInterconnect

Matrix

VersaRingTM: High Utilization & Pin Assignment Flexibility

Versatile interface between internal logic and I/O:

— I/O decoupled from core logic— Incremental edge routing resources

VersaRing resources include:— 8 horizontal longlines— 8 vertical longlines— 4 direct connects in per side— 4 direct connects out per side— 4 double length lines to GRM— 10 single length lines to GRM— 8 single length lines to adjacent VersaRing tile

XC5200 Family

XC5202 XC5204 XC5206 XC5210 XC5215Max Logic Gates 3,000 6,000 10,000 16,000 23,000Typical Gate Range 2-3K 4-6K 6-10K 10-16K 15-23KLogic Cells 256 480 784 1,296 1,936Flip-Flops 256 480 784 1,296 1,936Max I/O 84 124 148 196 244Performance -6/-5/-4/-3 -6/-5/-4/-3 -6/-5/-4/-3 -6/-5/-4/-3 -6/-5/-4/-3Packages: PC84 PC84 PC84 PC84

PQ100 PQ100 PQ100 VQ100 VQ100 VQ100 TQ144 TQ144 TQ144 TQ144

PQ160 PQ160 PQ160 PQ160 TQ176 TQ176PQ208 PQ208 HQ208

BG225 BG225PQ240 HQ240

HQ304 BG352

PG156 PG156 PG191 PG223 PG299

100% 100% Footprint Footprint Compatibility in Compatibility in

Common PackagesCommon Packages

What is the CPLD

• Similar to having multiple PAL devices interconnected in one chip

• Best applications Wide functions Fast arithmetic Complex counters Complex state machines PAL/GAL or TTL integration Non-volatile

PAL

PAL

PAL

PAL

Swi-tch

Mat-rix

Prog.ANDarray

FixedOR

array

FF/Macro-

cell

FF/Macro-

cell

Functions of 5 Product Term are fastest

Xilinx CPLDs use a Sum of Products Architecture

From 5 P-terms to 15 P-terms cost under 1ns of tpd

AND functions are “free”, Ors are not. An OR gate equals a Macrocell

Inverters are free

36 Input 5 Product Term

Pin Selection For CPLD

Remember the JTAG programming pins

Join the CPLDs in the same JTAG Daisy Chain for board debug

The Xilinx CPLD JTAG cable has 4mA drive, be sure to buffer the JTAG Signals if more than 5 devices in chain

Why not EEPROM Technology?

3XRouting

Switches

Product benefits due to 67% smaller FLASH cell More routing switches in the same area supports pinlocking Lower parasitic capacitance improves performance Long term cost improvements due to scalability

Greatly improved endurance with FLASH Lower program/erase voltages

Source

Data

Word Line

Control

Ground

Data

Word Line

Control

Write

Floating Gate

Typical E2 CPLD Cell FastFLASH Cell

Control

Write

Data

Ground

Word Line

Source

Data

Word LineControl

S1

S2

S3

Source

Data

Word Line

Control

Source

Data

Word Line

Control

S1

ISP Market Growth

1997

Non-ISP70%

ISP30%

2000

Non-ISP40%

ISP60%

XC9500 CPLDs

“The New Standard in ISP”

XC9500 Key Features

5 Volt In-System Programmable (ISP) CPLDs

Most complete IEEE 1149.1 JTAG for ISP and Test

High performance: 5ns pin-to-pin

High density: up to 6,400 usable gates

Highest reprogramming reliability 10,000 program/erase cycles 20-year data retention

Industry’s best pin-locking architecture

XC9500 Architecture

FunctionBlock 1

JTAGController

FunctionBlock 2

I/O

FunctionBlock n

3

I/O - Global Tri-States

2 or 4

FunctionBlock 3

I/O

In-SystemProgramming Controller

FastCONNECTSwitch Matrix

JTAG Port

3

I/O

I/O

I/O - Global Set/Reset

I/O - Global Clocks

I/OBlocks

1

FunctionBlock

FastCONNECT

Complete Interconnectivitywith FastCONNECT™

JTAG

GlobalS/R

GlobalClocks

FunctionBlock

FunctionBlock

FunctionBlock

FunctionBlock

FunctionBlock

FunctionBlock

FunctionBlock

Global3-State

XC9500 FastCONNECT

Pin Inputs(~ 3 entries / FB)

Macrocells(36 entries / FB)

1

2

3

4

36

FastFLASH Function Block

ToFastCONNECT

FromFastCONNECT

23 Global3-State

GlobalClocks

I/O

I/O

36

Product-Term

Allocator

Macrocell 1

ANDArray

Macrocell 18

XC9500 Macrocell

FromFastCONNECT

36

SUM-TermLogic

D/T Q

R S

to/from other macrocells

RegisterXOR 18

GlobalClocks

GlobalOEs

P-term ClkP-term R&SP-term OE

3 2 or 4

to/from other macrocells

GlobalR/S

P-TermAllocator

Feedback Paths

FastCONNECT

Pin

LocalMacrocell

FastCONNECT

FB X

Macrocell

Macrocell

Pin feedback

FastCONNECT feedback

Local feedback

What is Pin-Locking?

Ability to retain device pin assignments for small to medium design changes

introducing a new variable to existing terms adding input signals inverting signals introducing 1or 2 buried flip flops adding p-terms

Requires a symmetric, uniform architecture

Requires software focus on pin-locking

3 Keys to Pin-Locking

FunctionBlock

FastCONNECT

JTAG

Global3-State

GlobalS/R

GlobalClocks

FunctionBlock

FunctionBlock

FunctionBlock

FunctionBlock

FunctionBlock

FunctionBlock

2nd generation switch matrix:

Complete interconnect of all pins & blocks

All pin-to-pin paths at full speed

FunctionBlock

Fully PopulatedSwitch Matrix

Keys to Pin-Locking

ToFastCONNECT

23 Global3-State

GlobalClocks

I/O

I/O

FromFastCONNECT

36

Macrocell 1

ANDArray

Macrocell 18

ProductTermAllocator

•

Wide FunctionBlock Fan-in

Flexible MacrocellLogic Allocation

XC9500 Supports Design Changeswith Fixed Pinouts

Design Change XC9500 Feature

Add another input FastCONNECT switch matrixpin or FB output with 100% connectivity

Add more logic in XC9500 allows expansionthe macrocell up to 90 P-terms

Add additional input 36 total inputs are available connections to the FB plus FastCONNECT

AND gate capability

CPLD Architecture Designed for Superior Pin-Locking

Routability Excellent Good* Poor Good

Notes: * Decreases with density

Xilinx

XC9500

Altera

Max7KS

Lattice

1K/2K/3K

AMD

Mach5

36

Yes

90

Yes

32

No No

No

36

No

18/24

No No

32 32

32

Fully populated switch

Maximum pterms/Mcell

Bi-directional individual

product term allocation

Function block fan-in

Requirements to Supportthe Product Life Cycle

Prototyping SystemIntegration

FieldUpgrades

ManufacturingPre-production

ManufacturingHigh Volume

In-System Program Pin Locking JTAG High Endurance

High program/erase reliability Thousands of program erase cycles required

Pin Locking Capability Pinout remains the same after design changes

Industry standard JTAG Device programming, logic debugging, manufacturing

board test

XC9500 Other System Features

Enhanced Data Security Features Read security bits prevent unauthorized reading Write security bits prevent accidental program/erase

Reduced power option per macrocell

3.3v/5v outputs

24 mA, 100% PCI compliant

Output Noise Reduction Slew rate control User programmable ground pin capability

User Programmable Ground Pin

User I/O Pin

Ground Pin

User I/O Pin

Internal Logic

Additional Ground Pin• Lower ground inductance• Reduce ground noise

XC9500 Product Family

9536

Macrocells

UsableGates

tPD (ns)

Registers

Max. UserI/Os

36 72 108 144 216

800 1600 2400 3200 4800

5 7.5 7.5 7.5 10

36 72 108 144 216

34 72 108 133 166

Packages 44VQ44PC 44PC

84PC100TQ100PQ

84PC100TQ100PQ160PQ

100TQ100PQ160PQ

288

6400

15

288

192

208HQ352BG

160PQ208HQ352BG

9572

95108

95144 95216 95288

FLASH Process Technology Roadmap

1994 1996 1998 2000

.6

.4

.2

5v

2002

.8

5v

3.3v2.5v

1.8v

5v

Year

Featu

re S

ize (

µ)

1995 1997 1999 2001

Address Decoder Benchmark

Tests effects of routing and function block fan-in on pin locking

a[N:0] out1

b[N:0] out2

FB1

FB2

Address Decoder Design Iteration

a[N:0] out1

b[N:0] out2

FB1

FB2

MODULE swapTITLE 'Bus Swap'

equations

out1 = b_bus == 24;out2 = a_bus == 24;

END

Wrong address bits decoded, address bus rerouted

Address Decoder Pinlock Performance

Altera

Xilinx

Initial Compile

Pinlocked with Changes

All XC9500 -10

16-Bit Bus

10

XC9572 -10and larger

32-Bit Bus

10

XC95108-10and larger

36-Bit Bus

10tPD

(nS)

16-Bit Bus

10

EPM7128S-10and larger

32-Bit Bus

10

EPM7128S-10and larger

36-Bit Bus

10

EPM7128S-10and larger

tPD

(nS)

16-Bit Bus

10

32-Bit Bus

10

ispLSI1048-80

36-Bit Bus

10

20 20

30

20

30

16.419.2 19.2

15

23.319.2

23.3

31.6

23.3

4039.9

12 12

a b c d

ispLSI1016-110ispLSI1024-90ispLSI1032-90ispLSI1048-80

abcd

a b

ispLSI1032-90ispLSI1048-80

ab

tPD

(nS)

Lattice

Address Counter Benchmark

Tests effects of routing and function block fan-in on pinlocking when macrocell feedback is required

adata[N:0] qa[N:0]

bdata[N:0] qb[N:0]

bload

aload

clockhold

ACNT

BCNT

Address Counter Design Iteration

adata[N:0]

bdata[N:0]

qa[N:0]

qb[N:0]

clockhold

bload

aload

ACNT

BCNT

MODULE cntswapequationsacount := bdata & aload # acount & !aload & hold # (acount + 1) & !aload & !hold;acount.clk = clock;

bcount := adata & bload # bcount & !bload & hold # (bcount + 1) & !bload & !hold;bcount.clk = clock;

Counter load data incorrect, must be swapped

Address Counter Pinlock Performance

Altera

Xilinx

16-Bit Address

50

XC9572 -10and larger

66.7

100

24-Bit Address

50

XC95108 -10and larger

66.7

100

32-Bit Address

50

XC95144 -10and larger

66.7

100

16-Bit Address

50

EPM7096-10EPM7128S-10and larger

77100

24-Bit Address

50

77100

32-Bit Address

50

EPM7256S-10100

EPM7096 Won't Pinlock

83 83 83

EPM7160E Won't PinlockEPM7128S Won't Pinlock

83

EPM7256S Won't Pinlock

a b ca b

ab

EPM7160E-10EPM7128S-10EPM7192S-10and larger

abc

fMAX

(MHz)

fMAX

(MHz)

ispLSI2064-100and larger

ispLSI2128 -100

16-Bit Address

50 32

100

24-Bit Address

5025

100

32-Bit Address

50

20

100

25 20 25 24 17

a b

ispLSI2096-100ispLSI2128-100

ab

fMAX

(MHz)Lattice

XC9500 Fits In Industry Standard JTAG Chains

XC9500 XC4000EX

uP

DSP

ASICXC9500

TDI

TDO

TMSTCK

Advanced JTAG Download Software

Multi-chip JTAG strings supported automatically

XChecker and Parallel download cables available

Easy-To-Use XACTstepM1.4 for CPLDs

Windows-based Design Manager & Flow Engine

Automatic device fitting

Static Timing Analyzer

XACT Performance for timing driven fitting

HardWireTM FpgASICThe Superior ASIC Solution

FPGA FpgASIC

CORES

FPGA + HardWire Advantage

Cost Reduction Alternatives

Opportunity Cost

Re-engineering Cost(Risk, Engr, Time to Market)

XilinxFPGA

XilinxHardWire

Custom ASIC

Un

it P

rice

1X

0.8X

0.6X

0.4X

0.2X

0.1X

Zero High

Cost Reduction =

(unit price x unit volume) - (Opportunity Cost)

- (Re-engineering Cost)

Cost Reduction =

(unit price x unit volume) - (Opportunity Cost)

- (Re-engineering Cost)

Gate Array Conversion HardWire

Functional • Simulation vectors must • No simulation vectorsCompatibility be developed

• Redesign required • 100% Pin & Functional• Some gates and nets are lost • CLBs and Nets preserved

Timing • Post route timing simulation • Simple timing analysisCompatibility always required without simulation

• Non-exhaustive • Asynchronous path focus

Testability • Fault Coverage varies 70 - 95% • >95% Fault Coverage• Test Vectors development and • Test vectors created

Fault Grading required automatically

Result • Resource intensive • Easy, virtually no work• Medium to high risk • Lowest risk

The HardWire Advantage

Xilinx Fills the ASIC Gap

Low

Low High

50Kg

200Kg

High System-on-a-Chip

(SLI)

Glue Logic

TheASICGap

LSI Logic, Lucent, T.I., VLSI,NEC, SGS, Motorola

Matra-Temic, ESI, AMI, Orbit, et.al

TERRITORY

(FPGA) (HARDWIRE ASIC)

Volume

Den

sity

HardWire Methodology“Design Once”

Xilinx

ATPGXilinx

ATPG

Prototypes Prototypes

Test

Development

Test

Development

VerificationVerification

Place and RoutePlace and Route

VerificationVerification

CaptureCapture

Typical ASIC Design Phases

FPGA

Design

Xilinx HardWire Methodology

Production ReadyPrototypesProduction Ready

Prototypes

Physical Data BasePhysical Data Base

Iterations

Gate ArrayRedesign Path

Physical Data Base Design File Conversion

HardWire Advantage“Design Once” Compatibility

HardWire Industry GenericFeature

Array Gate Array

Identical Pinout as FPGA Yes ?

Identical Features as FPGA Yes No

Convert From Design file (verified database) Schematic

Simulation Vectors Not Needed Required

Test Vectors Not Needed Required

Fault Coverage >95% ?

Easy qual(same vendor/process)

Yes No

Risk Almost Zero Medium-High

Engineering Cost Almost Zero Medium-High

HardWire DESIGNLOCK ?Reduces Re-design Risk

FPGA HardWire

All CLBs and nets are preserved - No netlist transformation

Unused CLB/logic is not mapped for optimal die size CLBs & nets are maintained in the same relative location on the die.

HardWire CLB Based Mapping Preserves Functionality

CLB Mapping Preserves Placement

ASIC Place and Route is Random

By DesignFPGA Placement

FPGA DEVICE ROUTING

Generic ASIC Routing Can Change Critical Paths

HardWire ASIC DesignLocktm Methodology:Placement, Routing and Relative Timing Preserved

DesignLocktm MethodologyMinimizes Change

XH3 Architecture is Xilinx Specific

Efficient Hi-PerformanceASIC Core

FPGA I/O(Driver, ESD,5V Tolerant,JTAG)

PAD

OSC

POR,CE

VREF,CLKS

FPGA Features

Core

Cor

e

FPGA Cores

The Competitive Battleground:Reducing Time to Volume

Iterations (ASIC)

Iterations

Iterations

Iterations

FPGA

FPGA +HardWire ASIC

FPGA +Generic ASIC

Generic ASIC

Iterations (ASIC)

Design/Verification

SystemVerification

ASICRe-engrg.

Prototypes

Production

With Iterations

With Iterations

What is a Prototype?Standard Gate Array vs. the HardWire Array

StandardGate Array

HardWireArray

Wafer FabSpecial quick-turn

fab line (notproduction fab)

Production Fab

Assembly

Opens/Shorts

Productionassembly line

Package Type

Quick-turnassembly line

Production Package

Test

Ceramic

Production Test

Typical gate array prototype is a functional prototype only HardWire Array prototype truly represents the production part

HardWire Success I

Application: Arcade Video Game

FPGA: XC4020E-2PQ208

Features: Dual port RAM, PCI Logicore (Initiator and Target)

Comment: Attempted to do standard ASIC with several manufacturers with no success.

Solution: Only the HardWire ASIC solution was able to incorporate the logic and features providing a cost effective solution.HardWire Device: XC4420-PQ208CProjected Volume: 120-200KU per year.

HardWire Success II

Application: Controller chip in three chip telecom chipset.

FPGA: XC4020E-3PQ208I

Features: Dual port RAM, JTAG and Configuration Emulation.

Comment: Needed fast time to market cost effective solution. Internal solution could not meet time to market requirements.

Solution: HardWire ASIC provided production quality samples 7 weeks after the design submittal kit was received. Including package change.

HardWire Device: XC4420-TQ100C

Projected Volume: 100-300KU per year. Sold to several phone companies

HardWire Success III

Application: six chip encryption chipset for next generation phone/radio base station.

FPGA: XC3030-5PC84I, XC5204-4PQ100, XC4005E-3PQ100.

Features: SelectRAM, JTAG and Configuration Emulation.

Comment: Limited engineering resources available to complete conversion process. Time to completion critical for end customer.

Solution: Used different HardWire technologies to match features and performance with most cost effective solution. Production quality samples delivered 6 weeks after complete design submittal kit received.

HardWire Device: XC3330-PC84I, XC5404-PQ100I, XC4405-PQ100C Projected Volume: 10-20KU per year.

XACT Step M1XACT Step M1

Design Flow for Implementation

1. Invoke Design Manager

2. Start a Project or Open a Project

3. Specify Back Annotation File for Simulation

4. Implement the Design

5. Check Timing Results

6. Download the FPGA/CPLD

Step 1: Invoke Design Manager

Configurable Flow EngineControls start/stop points and custom options

Timing AnalyzerReport on net and path delays

PROM File FormatterCreate file to program configuration file into PROM

Hardware DebuggerDownload configuration file

Design EditorDevice-level view of routing

• Invoke the Design Manager – Command line usage: dsgnmgr <cr>. PC users: select icon

• The Design Manager window appears:

Status Window

Step 2: Start a Project

Set up the design for implementation

Select Part Type

B. Select Part

C. Select Options (...see next foil)

A. Select Design -> Implement

Step 3: Specify Back Annotation

Select Design -> Implement -> Options

Enable “Produce Timing Simulation Data”

Design -> Implement -> Options

Step 3: Specify Back Annotation

Format options:VHDLVerilogXNFEDIF

EDIF formats:Generic - Standard 2000ViewlogicMentor EDIF

• Specify the Back Annotation Netlist Format• To use this form, select Design -> Implement ->

Options -> Edit Implementation Template -> Interface

Design -> Implement -> Options -> Edit Template

Step 4: Implement the Design (1) Select Design -> Implement -> Run

Or press right arrow icon “Run-only” Flow Engine appears

– The design is implemented and the configuration file is created

Flow Engine Status WindowMain Window

Step 4: Implement the Design (2)

The Flow Engine appears after implementation is started

— Allows further customization Tracks status of the design

Flow Engine Status Window

Step 5. Check Results (1) Select Utilities -> Report Browser from the Design Manager

Reports are shown once available Double-click to open Yellow sparkles indicates new (not yet read)

Main Window

Report Browser

Step 5: Check Results (2)

Key Report Files Map Report

Displays result of DRC (Design Rule Check) Indicates if the design will fit into the specified part Identifies ways to improve the design Reports nets with no source or load

Logic Level Timing Report provides delay estimates Created after Map Reports longest paths in the design Created before placement Based on block delays and minimum net delays

Step 5: Check Results (3) Placement and Routing Report includes resource summary

Indicates the percentage of utilization The number of I/O and flip-flops is specified Reports if the design routed Gives an overall timing score

– Score of zero indicates all timing specifications were met

Post Layout Timing Report Based on block delays and net delays after routing Used for detailed delay analysis after implementation

DLY Report Lists delay for each net

Pad Report Cross reference of Input/Output components and package pins

Step 5. Check Results (4)

Select the Timing Analyzer Icon Analyze delays before and after implementation

Timing Analyzer

Step 6: Configure the FPGA/CPLDTo configure, or program FPGAs, a bit file is

downloaded Configuration can be driven by a microprocessor Data can be stored in a PROM

To configure CPLDs, a JED file is downloaded CPLDs can be configured In-System, or CPLDs can be configured using third party programmers

More information on Configuration is given later today