SiGe HBT BiCMOS Field Programmable Gate Arrays for Fast Reconfigurable Computing Bryan S. Goda Rensselaer Polytechnic Institute Troy, New York.

SiGe HBT BiCMOS Field Programmable Gate Arrays for

Fast Reconfigurable Computing

Bryan S. Goda

Rensselaer Polytechnic Institute

Troy, New York

Agenda

• Introduction

• BiCMOS FPGA History

• SiGe HBT BiCMOS Process

• Current Mode Logic

• Xilinx 6200 FPGA Design

• Configuration Memory

• Performance Results

• Conclusions and Future Work

Current Role of SiGe

• “More Zip per Chip”

• Wireless Phones -> Watch Sized Phone

• Direct Broadcast Satellite

• Fiber-Optic Lines, Switches, and Routers

Programmable Bipolar Logic

1983: Fairchild ECL Field Programmable Logic Array• Fuse Based• 4ns Cycle Rate• High Power• Scaling Problems

1990: Algotronix 1.2uM 256 Cell Configurable Logic Array• fT 6 GHz, 200ps Gate Delay• 4 Transistor Static RAM Memory Cells• ASIC Emulation and Signal Processing• Forerunner of XC6200

V+

a a

Vref

a a

EN1 EN2

V-

Y1 Y1 Y2 Y2

US Patent CMOS Switchable 2 Input Multiplexer

SiGe Heterojunction Bipolar Transistor

• Selectively introduce Ge into the base of a Si BJT

• Smaller Base Bandgap increases e- injection, higher Beta (100)

• Higher Beta allows more heavily doped base RB (125 Ohm)

• Graded Bandgap decrease base transit time fT

SiGe HBT

• 50Ghz Process, 100Ghz process within a year (30uA at 50 Ghz)

• 5 layers of metal• Used in RPI VLSI Class• co-integrated with CMOS process

– can have HBT logic with CMOS memory– low power and high speed

fT Curves for Various Emitter Lengths

SiGe HBT Layout

Base

Emitter

Collector Sub-Collector

EC

EV

e-

h+n+ Siemitter

p-Si

Ge

p-SiGebase

Eg,Ge(x=0)

Eg,Ge(grade)= Eg,Ge(x=Wb)- Eg,Ge(x=0)

Drift Field

n- Sicollector

Band Diagram

=0.031 ev

Dielectric ConstantSi = 11.7Ge =16.2SiGe (7.5% Ge)=12.03

CML Branch Current vs. Differential DC Voltage

IBM SiGe and CMOS Load Gate Delays on M1, M2, LM

Current Steering Logic

Level 1

Vcc 0 V

-250 mV

Level 2

-950 mV

-1.2 V

Level 3

-1.90 V

-2.15 V

Vee 4.5 V

Fastest Logic LevelLimited Drive Capability

Inter-block Signal LevelGood Fan-Out (10)

Clock SignalSlowest Level

Level 4 Possible

Current Steering Logic In SiGe• 13ps Transistor Switching Time (75 Ghz)

– 6ps Process Next Year

• Small Voltage Swings (250mv) vs 3.3 or 5 V– Less Power– Smaller Swing = Faster

• “Steer” Currents, Use Differential Logic– Less Switch Noise

• Less Transistors needed, Complement Signal Present

• Flip-Flops and Multiplexers Easy to Implement

Vcc

AA A

BB

Vee

Vref

A XOR B

A XOR B

O V

Level 10 -0.25 V

Level 2-0.95 -1.2V

-4.5V

A B A XOR B0 0 00 1 11 0 11 1 0

CMLXOR Logic Schematic

1 0 1 1 0 1 1 1 0

0 0 0 1 1 0 0 1 0

1 0 1 0 1 1 1 0

Alevel1

Blevel 2

A XOR B

Logic Cell

General FPGA Structure

I/O Cell

Routing Network

Configuration Memory

High Speed FPGA Applications

• Real Time Image Processing- Radar- Pattern Recognition

• Digital Networks- Mobile Subscriber Equipment- Command Information Systems- High Speed Switching Nodes

• Control Systems- Guidance Systems- Reprogrammable Survivability

• Satellite Systems

Desired Image

Search Image

1. Desired Image is programmed into chip (1 pixel = 1CLB)2. Load a section of search image3. If enough pixels match, then turn found bit on4. Load another section, or reprogram with new desired image

Image Correlation

Samples FromXC6200 CAD Tools

CLBsIO Blocks

Pins

FPGA Drawbacks

• Slowdown – 200 Mhz Internal Speed down to 30-60 MHz External – Pass Transistor = Low Pass Filter

• Limited Bandwidth

• Relatively Long Configuration Times (Seconds)

• Vender Guarded Information

• More Expensive than Comparable ASIC

Equivalent Circuit from Node 3 to Node 2

M

M

M M

MM

1

On

23

4

3

2

1 4

Interconnect

4

1

23

Pass Transistor (Memory)

Pass Transistor Interconnect Modeling

Field Programmable Gate Arrays (FPGA)

• Hierarchy Level Organization (Sea of Gates)– Simple Cells (Configurable Logic Blocks)– 4x4, 16x16, 64x64 groupings– Hierarchy of routing resources at each level– I/O Blocks (external interface)

Design Parameters

• Logic Swings Levels- Based on Differential Pair Switching- Current Levels

• Redesign of the Configurable Logic Block- Take Advantage of Differential Wiring- What Parts Can be Turned off if not Used?

• Supply Levels- How Many Levels of Logic?

• Routing Resources

• CMOS Voltage Levels- Integrate CMOS into Bipolar Current Tree

a

VCC 0 V

OUT

OUT

a b b c c d d

S1 S1 S1 S1

S2 S2

Vee -3.4 V

Level 3-1.9 -2.15V

Level 2-0.95 -1.2V

Level 10 -0.25V

Replace with

Vref

Current Tree with CMOS Routing

Pulse Width 50ps 60ps 70ps 100ps

CMOSBipolar

Bipolar vs Bipolar/CMOS Current Trees

4:1 MultiplexerLevel 1 Inputs

Level 1Output

Level 1Output

Level 2Input

Level 2Input

Level 3Input Level 3

Input

CMOSVersion

W/L 5:1

If a=1 then select Y2output = b

If a=0 then select Y3output = 0

A OR B

A and B

If a=1 then select Y2output = 1

If a=0 then select Y3output = b

X2:= a

1

0

X1:= a

X3:= b

Y2

Y3

X2:= b

1

0

X1:= a

X3:= a

Y2

Y3

Sample Logic Using Multiplexers

Non-Inverted Output

Inverted Output

X2:=b

1

0

X1:= a

X3:= a

Y2

Y3

X2:= b

1

0

X1:= a

X3:= a

Y2

Y3

Redesign of XC6200 Logic

Original XC6200 Design• Have to Track Inversions

Revised Design• Use Differential Pair Logic• Eliminate XC6200 Fast Logic• No Inversion Tracking

X1Y2X2

X3 Y3

1

0

D Q

Clk Q

FC

S

RP Multiplexer

CS Multiplexer

Clr

Original XC6200Architecture

RedesignedArchitecture

X1Y2X2

X3 Y3

1

0

D Q

Clk Q

FC

S

RP Multiplexer

CS Multiplexer

Clr Switchable

Bipolar with CMOS Routing

10 Ghz Three CLB Simulation

4:1 MuxHigh Speed Logic

2:1 MuxCMOS Control Buffer

4:1 Mux (off switchable)CMOS Control Master/Slave Latch (off switchable)

(off switchable)

CLB Layout

Sample CLB Test Circuit

CLB8:1 Mux

Pad Drivers

8/1 DivideBufferVrefVref

Actual Fabricated Test Circuit

Pads (110u x 110u)

NSEWN4S4E4W4

NSEWN4S4E4W4

F

N S E W N4 S4 E4 W4

X1 X2CLB

X3

Outgoing CLB Routing Incoming CLB Routing

4x4 Block Boundary Routing

S Switches

E S

witch

es

N Switches

W S

wit

ches

S Switches

E S

witch

es

N Switches

W S

wit

ches

Local RoutingMagic Routing

Length 4 FastLane (4x4)Length 16 Fastlane (16x16)Chip Length Fastlane (64x64)

NSEWN4S4E4W4

NSEWN4S4E4W4

F

N S E W N4 S4 E4 W4

X1 X2CLB

X3

Local CLB Routing

N S

W FW

out

S E W F

Sout

N S E

F

Eou

t

N E W F

Nout

• Nearest Neighbor Routing• Output (F) or Local Through

Example: Route East Signal Through to Next CLBNote: Can’t Route Signal Back to Origin at this Level

New

Co

nfi

gu

rati

on

Dat

a

VEE

VEE

VSS VSS

SRAM Bits In Memory Planes CMOS to CML Buffer

decode

CLBMultiplexer

InputsVREF

Normal CMOS Memory-CML Interface

D Latch M/S40 Transistors

D Latch M/S18 Transistors

DQ

Clock

QCLK

D

CLK

QQ

Data Data

WordOut Out

RAM Cell6 TransistorsParallel Load

Memory Design

Memory Planes

CLBs

3-D Chip Stacking

• Shorter Wires• More CLBs/Area• Optimize Memory

CLB with Routing and RAM (2)

MUX MUX MUX

CLB

MUX Selects

CLBSelect

RAM1

RAM2

Layout of Configurable Logic Block with 2 sets of RAMRAM 2:1 Mux

8:1Mux (routing)CMOS Selects CLB (logic)

Master/Slave Latch(memory)

Circuit Elements:240 nfets122 pfets36 resistors98 npn1 HBTs16 npnhb1 HBTs

Circuit Type Buffer CMLXOR,AND,OR

MUXXOR,AND,OR

CLB

Propagation Delay 17ps 22-25ps 23-26ps 100ps

SiGe Performance

* Projected Power Levels for 7HP Process:At 50Ghz, 30 uA, 20x+ reduction in power

Power Decreasing Ideas

Date Idea Power Consumption/CLBDec 98 Original CLB 73 mWJune 99 CLB Redesign I 34 mWAug 99 CLB Redesign II 24 mWDec 99 Widlar Current Mirror

with CMOS Control, CMOS Routing 10.8 mW

Mar 00 Supply Voltage 4.5 -> 3.3V 7 mWDec 00* 7HP Process 0.3 mW

Multiplexer Performance vs Temperature

Normal 250 mV Swing

200 mV Min Swing

Vcc

Vref

Vee

Input

Widlar Current Mirror with CMOS Control

XC6200 Design Improvements

• Developed at the University of Scotland

• Inversion of Signal at Every CLB- Taken care of due to differential pair wiring

• No Pass Transistors, Use Multiplexers for Routing

• Able to turn off unused parts with CMOS controlled current mirror

• No CMOS-CML Conversion circuits needed, CMOS in current trees

• Handcrafted, dense layouts

• Context Switching

Power Delay Product

0.001

0.01

0.1

1

1998 1999 2000 2001 2002

Year

uW

/gat

e/M

hz

(log

sca

le)

PDP BiCMOS

PDP CMOS High

PDP CMOS Low

5HP

7HP8HP

A A B B C C

A A B B C C

Slow Transition

Fast Transition

Data Dependent Switching

Could Vary Signals Up to 30%

Setup Time Violations

Differential Logic hasComplement Switching In Opposite Direction

Bit Line Twisting

Future Work

• Testing

• Overall FPGA Architecture

• Scaling

• Integrate with Other Systems

• Projected Graduation May 2001, work to continue at USMA

• Power Reduction- 7HP Process

Pattern1000110010070ps ~ 7.1 GHz

Pattern2101101110070ps

Select

00011001001011011100

0001000100 AND1011111100 OR

AND OR AND OR

CLB Context Switch Example

Redesigned CLB Cell with Routing and Memory (2x)

2x24 BitRAM

Three 8-1 InputMux

CLBFour 4-1 Output

Mux

M1M2M3M4

CLB Row 4x1

Switch

Circuit Elements

1520 Nfets792 Pfets260 Resistors140 NPN1 HB576 NPN1

Memory Bus Lines N/S Input Output

Device XC6209 XC6216 XC6236 XC6264

Gate Count 9-13K 16-24K 36-55K 64-100K

Number Cells 2304 4096 9216 16384

I/O Blocks 192 256 384 512

Row x Col 48x48 64x64 96x96 128x128

XC6200 Device Family

Typical Routing Delays

Symbol Parameter XC6200 SiGe Redesign TNN Route Nearest Neighbor 1 ns 23 ps

Tmagic Route X2/X3 to Magic Out 1.5 ns 47 ps

TL4 Length 4 FastLane 1.5 ns 47 ps

TL16 Length 16 FastLane 2 ns 70 ps

TCL64 Chip-Length (64) Delay 3 ns 94 ps

~31x improvement

4x4 CLB Layout Cell

• Largest Basic Block

• Over 13,000 Transistors

• CommercialProduct Size is a 4x4 Arrayof this Cell

5 Stage Ring Oscillator

Schematic 6.36 Ghz -- 8.4mA

Parasitics 5.71 Ghz 89% 8.6mA

50oC 5.26 Ghz 82% 8.85 mA

75oC 4.87 Ghz 76% 9.1 mA

100oC 4.16 Ghz 65% 9.34 mA

125oC 3.12 Ghz 49% 9.5 mA

Speed Relative to Schematic Current