Home » Intel Manuals » Motherboards » Intel VC820 » Manual Viewer

Intel VC820 Design Guide - Page 109

Monte Carlo Analysis, 3.2.3.4 Simulation Criteria, both fast and slow corner conditions

Add to My Manuals
Save this manual to your list of manuals

Page 109 highlights

Advanced System Bus Design 3.2.3.3 3.2.3.4 Monte Carlo Analysis Perform a Monte Carlo analysis to refine the passing solution space region. A Monte Carlo analysis involves randomly varying parameters (independent of one another) over their tolerance range. This analysis intends to ensure that no regions of failing flight time and signal quality exists between the extreme corner cases run in pre-layout simulations. For the example topology, vary the following parameters during Monte Carlo simulations: • Lengths L1 through L3 • Termination resistance RTT on the processor cartridge #1 • Termination resistance RTT on the processor cartridge #2 • Z0 of traces on processor cartridge #1 • Z0 of traces on processor cartridge #2 • S0 of traces on processor cartridge #1 • S0 of traces on processor cartridge #2 • Z0 of traces on baseboard • S0 of traces on baseboard • Fast and slow corner processor I/O buffer models for cartridge #1 • Fast and slow corner processor I/O buffer models for cartridge #2 • Fast and slow package models for processor cartridge #1 • Fast and slow package models for processor cartridge #2 • Fast and slow corner 82820 MCH I/O buffer models • Fast and slow 82820 MCH package models Simulation Criteria Accurate simulations require that the actual range of parameters be used in the simulations. Intel has consistently measured the cross-sectional resistivity of the PCB copper to be approximately 1 Ω*mil2/inch, not the 0.662 Ω*mil2/inch value for annealed copper that is published in reference material. Using the 1 Ω*mil2/inch value may increase the accuracy of lossy simulations. Positioning drivers with faster edges closer to the middle of the network typically results in more noise than positioning them towards the ends. However, Intel has shown that drivers located in all positions (given appropriate variations in the other network parameters) can generate the worstcase noise margin. Therefore, Intel recommends simulating the networks from all driver locations, and analyzing each receiver for each possible driver. Analysis has shown that both fast and slow corner conditions must be run for both rising and falling edge transitions. The fast corner is needed because the fast edge rate creates the most noise. The slow corner is needed because the buffer's drive capability will be a minimum, causing the VOL to shift up, which may cause the noise from the slower edge to exceed the available budget. Slow corner models may produce minimum flight time violations on rising edges if the transition starts from a higher VOL. So, Intel highly recommends checking for minimum and maximum flight time violations with both the fast and slow corner models. The fast and slow corner I/O buffer models are contained in the processor and Intel 820 chipset electronic models provided by Intel. The transmission line package models must be inserted between the output of the buffer and the net it is driving. Likewise, the package model must also be placed between a net and the input of a receiver model. Editing the simulator's net description or topology file generally does this. Intel®820 Chipset Design Guide 3-9

Section	Page
Introduction 1	13
1.1 About This Design Guide	13
1.2 References	14
1.3 System Overview	14
1.3.1 Chipset Components	15
1.3.2 Bandwidth Summary	16
1.3.3 System Configuration	17
1.4 Platform Initiatives	20
1.4.1 Direct Rambus*	20
1.4.2 Streaming SIMD Extensions	20
1.4.3 AGP 2.0	20
1.4.4 Hub Interface	20
1.4.5 Manageability	21
1.4.6 AC’97	22
1.4.7 Low Pin Count (LPC) Interface	23
Layout/Routing Guidelines 2	27
2.1 General Recommendations	27
2.2 Component Quadrant Layout	27
2.3 Intel® 820 Chipset Component Placement	29
2.4 Core Chipset Routing Recommendations	30
2.5 Source Synchronous Strobing	31
2.6 Direct Rambus* Interface	33
2.6.1 Stackup	34
2.6.2 Direct Rambus* Layout Guidelines	34
2.6.2.1 RSL Routing	34
2.6.2.2 RSL Termination	37
2.6.2.3 Direct Rambus* Ground Plane Reference	39
2.6.2.4 Direct Rambus* Connector Compensation	40
2.6.2.5 RSL Signal Layer Alternation	46
2.6.2.6 Length Matching Methods	47
2.6.2.7 VIA Compensation	49
2.6.2.8 Length Matching & Via Compensation Example	49
2.6.3 Direct Rambus* Reference Voltage	51
2.6.4 High-speed CMOS Routing	51
2.6.4.1 SIO Routing	52
2.6.4.2 Suspend-to-RAM Shunt Transistor	52
2.6.5 Direct Rambus* Clock Routing	54
2.6.6 Direct Rambus* Design Checklist	54
2.7 AGP 2.0	57
2.7.1 AGP Interface Signal Groups	58
2.7.2 1X Timing Domain Routing Guidelines	59
2.7.3 2X/4X Timing Domain Routing Guidelines	59
2.7.4 AGP 2.0 Routing Summary	61
2.7.5 AGP Clock Routing	62
2.7.6 General AGP Routing Guidelines	62
2.7.7 VDDQ Generation and TYPEDET#	63
2.7.8 VREF Generation for AGP 2.0 (2X and 4X)	65
2.7.9 Compensation	67
2.7.10 AGP Pull-ups	67
2.7.10.1 AGP Signal Voltage Tolerance List	68
2.7.11 Motherboard / Add-in Card Interoperability	68
2.8 Hub Interface	69
2.8.1 Data Signals	70
2.8.2 Strobe Signals	70
2.8.3 HREF Generation/Distribution	70
2.8.4 Compensation	71
2.9 System Bus Design	72
2.9.1 100/133 MHz System Bus	72
2.9.2 System Bus Ground Plane Reference	73
2.10 S.E.C.C. 2 Grounding Retention Mechanism (GRM)	73
2.11 Processor CMOS Pullup Values	75
2.12 Additional Host Bus Guidelines	78
2.13 Ultra ATA/66	82
2.13.1 Ultra ATA/66 Detection	82
2.13.2 Ultra ATA/66 Cable Detection	83
2.13.3 Ultra ATA/66 Pullup/Pulldown Requirements	86
2.14 AC’97	87
2.14.1 AC’97 Signal Quality Requirements	89
2.14.2 AC’97 Motherboard Implementation	89
2.15 USB	91
2.16 ISA (82380AB)	92
2.16.1 ICH GPIO connected to 82380AB	92
2.16.2 Sub Class Code	92
2.17 IOAPIC Design Recommendation	92
2.18 SMBus/Alert Bus	93
2.19 PCI	93
2.20 RTC	93
2.20.1 RTC Crystal	94
2.20.2 External Capacitors	94
2.20.3 RTC Layout Considerations	95
2.20.4 RTC External Battery Connection	95
2.20.5 RTC External RTCRST Circuit	96
2.20.6 RTC Routing Guidelines	96
2.20.7 VBIAS DC Voltage and Noise Measurements	97
Advanced System Bus Design 3	101
3.1 Terminology and Definitions	101
3.2 AGTL+ Design Guidelines	104
3.2.1 Initial Timing Analysis	105
3.2.2 Determine General Topology, Layout, and Routing Desired	108
3.2.3 Pre-Layout Simulation	108
3.2.3.1 Methodology	108
3.2.3.2 Sensitivity Analysis	108
3.2.3.3 Monte Carlo Analysis	109
3.2.3.4 Simulation Criteria	109
3.2.4 Place and Route Board	110
3.2.4.1 Estimate Component To Component Spacing for AGTL+ Signals	110
3.2.4.2 Layout and Route Board	110
3.2.4.3 Host Clock Routing	112
3.2.4.4 APIC Data Bus Routing	112
3.2.5 Post-Layout Simulation	113
3.2.5.1 Intersymbol Interference	113
3.2.5.2 Cross-Talk Analysis	113
3.2.5.3 Monte Carlo Analysis	113
3.2.6 Validation	114
3.2.6.1 Measurements	114
3.2.6.2 Flight Time Simulation	114
3.2.6.3 Flight Time Hardware Validation	115
3.3 Theory	115
3.3.1 AGTL+	115
3.3.2 Timing Requirements	116
3.3.3 Cross-talk Theory	116
3.3.3.1 Potential Termination Cross-Talk Problems	118
3.4 More Details and Insight	119
3.4.1 Textbook Timing Equations	119
3.4.2 Effective Impedance and Tolerance/Variation	120
3.4.3 Power/Reference Planes, PCB Stackup, and High Frequency Decoupling	120
3.4.3.1 Power Distribution	120
3.4.3.2 Reference Planes and PCB Stackup	121
3.4.3.3 High Frequency Decoupling	122
3.4.3.4 SC242 Connector	123
3.4.4 Clock Routing	123
3.5 Definitions of Flight Time Measurements/ Corrections and Signal Quality	124
3.5.1 VREF Guardband	124
3.5.2 Ringback Levels	124
3.5.3 Overdrive Region	124
3.5.4 Flight Time Definition and Measurement	125
3.6 Conclusion	126
Clocking 4	129
4.1 Clock Generation	129
4.2 Component Placement and Interconnection Layout Requirements	134
4.2.1 14.318MHz Crystal to CK133	134
4.2.2 CK133 to DRCG	134
4.2.3 MCH to DRCG	135
4.2.4 DRCG to RDRAM Channel	136
4.2.5 Trace Length	136
4.3 DRCG Impedance Matching Circuit	138
4.3.1 DRCG Layout Example	139
4.4 AGP Clock Routing Guidelines	139
4.5 Series Termination Resistors for CK133 Clock Outputs	139
4.6 Unused Outputs	140
4.7 Decoupling Recommendation for CK133 and DRCG	140
4.8 DRCG Frequency Selection and the DRCG+	140
4.8.1 DRCG Frequency Selection Table and Jitter Specification	140
4.8.2 DRCG+ Frequency Selection Schematic	141
System Manufacturing 5	145
5.1 In Circuit FWH Flash BIOS Programming	145
5.2 FWH Flash BIOS Vpp Design Guidelines	145
5.3 Stackup Requirement	145
5.3.1 Overview	145
5.3.2 PCB Materials	146
5.3.3 Design Process	146
5.3.4 Test Coupon Design Guidelines	147
5.3.5 Recommended Stackup	147
5.3.6 Inner Layer Routing	147
5.3.7 Impedance Calculation Tools	148
5.3.8 Testing Board Impedance	148
5.3.9 Board Impedance/Stackup Summary	149
System Design Considerations 6	153
6.1 Power Delivery	153
6.1.1 Terminology and Definitions	153
6.1.2 Intel® 820 Chipset Customer Reference Board Power Delivery	154
6.1.3 64/72Mbit RDRAM Excessive Power Consumption	157
6.1.3.1 Option 1: Reduce the Clock Frequency During Initialization	158
6.1.3.2 Option 2: Increase the Current Capability of the 2.5V Voltage Regulator	158
6.2 Power Plane Splits	159
6.3 Thermal Design Power	159
6.4 Glue Chip 3 (Intel® 820 Chipset Glue Chip)	160
Appendix A: Reference Board Schematics: Uni-Processor	161
Reference Design Schematics: Uni-Processor A	163
Appendix B: Reference Design Schematics: Dual-Processor B	201

Match case Limit results 1 per page

Intel

820 Chipset

Design Guide

3-9

Advanced System Bus Design

3.2.3.3

Monte Carlo Analysis

Perform a Monte Carlo analysis to refine the passing solution space region. A Monte Carlo analysis

involves randomly varying parameters (independent of one another) over their tolerance range.

This analysis intends to ensure that no regions of failing flight time and signal quality exists

between the extreme corner cases run in pre-layout simulations. For the example topology, vary the

following parameters during Monte Carlo simulations:

•

Lengths L1 through L3

•

Termination resistance R

on the

processor cartridge #1

•

Termination resistance R

on the

processor cartridge #2

•

of traces on

processor cartridge #1

•

of traces on

processor cartridge #2

•

of traces on

processor cartridge #1

•

of traces on

processor cartridge #2

•

of traces on baseboard

•

of traces on baseboard

•

Fast and slow corner

processor I/O buffer models for cartridge #1

•

Fast and slow corner

processor I/O buffer models for cartridge #2

•

Fast and slow package models for

processor cartridge #1

•

Fast and slow package models for

processor cartridge #2

•

Fast and slow corner 82820 MCH I/O buffer models

•

Fast and slow 82820 MCH package models

3.2.3.4

Simulation Criteria

Accurate simulations require that the actual range of parameters be used in the simulations. Intel

has consistently measured the cross-sectional resistivity of the PCB copper to be approximately

Ω

*mil

/inch, not the 0.662

Ω

*mil

/inch value for annealed copper that is published in reference

material. Using the 1

Ω

*mil

/inch value may increase the accuracy of lossy simulations.

Positioning drivers with faster edges closer to the middle of the network typically results in more

noise than positioning them towards the ends. However, Intel has shown that drivers located in all

positions (given appropriate variations in the other network parameters) can generate the worst-

case noise margin. Therefore, Intel recommends simulating the networks from all driver locations,

and analyzing each receiver for each possible driver.

Analysis has shown that

both fast and slow corner conditions

must be run for both rising and

falling edge transitions. The fast corner is needed because the fast edge rate creates the most noise.

The slow corner is needed because the buffer’s drive capability will be a minimum, causing the V

to shift up, which may cause the noise from the slower edge to exceed the available budget. Slow

corner models may produce minimum flight time violations on rising edges if the transition starts

from a higher V

. So, Intel

highly recommends

checking for minimum and maximum flight time

violations with both the fast and slow corner models. The fast and slow corner I/O buffer models

are contained in the processor and Intel 820 chipset electronic models provided by Intel.

The transmission line package models must be inserted between the output of the buffer and the net

it is driving. Likewise, the package model must also be placed between a net and the input of a

receiver model. Editing the simulator’s net description or topology file generally does this.