Interconnect Delay Compensation in Timing Analysis for. Designs Containing Multiple Voltage Domains

Similar documents
Amber Path FX SPICE Accurate Statistical Timing for 40nm and Below Traditional Sign-Off Wastes 20% of the Timing Margin at 40nm

Principles of Current Source Modeling

Andrew Clinton, Matt Liberty, Ian Kuon

PROCESS-VOLTAGE-TEMPERATURE (PVT) VARIATIONS AND STATIC TIMING ANALYSIS

Current Based Delay Models: A Must For Nanometer Timing

Appendix. RF Transient Simulator. Page 1

Chapter 4. Problems. 1 Chapter 4 Problem Set

Experiment 8 Frequency Response

Modeling the Effect of Wire Resistance in Deep Submicron Coupled Interconnects for Accurate Crosstalk Based Net Sorting

Chapter 3 Novel Digital-to-Analog Converter with Gamma Correction for On-Panel Data Driver

UNIT-III POWER ESTIMATION AND ANALYSIS

Timing analysis can be done right after synthesis. But it can only be accurately done when layout is available

Characterization and Variation Modeling for 22FDX. Ning Jin Digital Design Methodology Team

Managing Cross-talk Noise

Accurate Timing and Power Characterization of Static Single-Track Full-Buffers

Accurate and Efficient Macromodel of Submicron Digital Standard Cells

DIGITAL INTEGRATED CIRCUITS A DESIGN PERSPECTIVE 2 N D E D I T I O N

EE 434 ASIC and Digital Systems. Prof. Dae Hyun Kim School of Electrical Engineering and Computer Science Washington State University.

Microcircuit Electrical Issues

Low Jitter, Low Emission Timing Solutions For High Speed Digital Systems. A Design Methodology

ASICs Concept to Product

TFA: A Threshold-Based Filtering Algorithm for Propagation Delay and Output Slew Calculation of High-Speed VLSI Interconnects

Digital Integrated Circuits Designing Combinational Logic Circuits. Fuyuzhuo

Accurate In Situ Measurement of Peak Noise and Delay Change Induced by Interconnect Coupling

CROSS-COUPLING capacitance and inductance have. Performance Optimization of Critical Nets Through Active Shielding

ICCAD 2014 Contest Incremental Timing-driven Placement: Timing Modeling and File Formats v1.1 April 14 th, 2014

PROGRAMMABLE ASIC INTERCONNECT

CS250 VLSI Systems Design. Lecture 3: Physical Realities: Beneath the Digital Abstraction, Part 1: Timing

Low Power Design Methods: Design Flows and Kits

EDA Challenges for Low Power Design. Anand Iyer, Cadence Design Systems

CharFlo-Cell! Cell! Next-Generation Solution for Characterizing and Modeling Standard Cell and I/O Library

Digital Integrated Circuits Designing Combinational Logic Circuits. Fuyuzhuo

Lecture #2 Solving the Interconnect Problems in VLSI

Introduction to Digital VLSI Design מבוא לתכנון VLSI ספרתי

Signal Integrity Management in an SoC Physical Design Flow

DATASHEET CADENCE QRC EXTRACTION

An Initial Case Study for BIRD95: Enhancing IBIS for SSO Power Integrity Simulation

Operational Amplifiers

Analysis and Reduction of On-Chip Inductance Effects in Power Supply Grids

ECEN 720 High-Speed Links: Circuits and Systems

Power Consumption and Management for LatticeECP3 Devices

A Current-based Method for Short Circuit Power Calculation under Noisy Input Waveforms *

Low Power System-On-Chip-Design Chapter 12: Physical Libraries

Capturing Crosstalk-Induced Waveform for Accurate Static Timing Analysis

DesignCon On-Chip Power Supply Noise and Reliability Analysis for Multi-Gigabit I/O Interfaces

Design of Adders with Less number of Transistor

An Energy Scalable Computational Array for Energy Harvesting Sensor Signal Processing. Rajeevan Amirtharajah University of California, Davis

Advanced Techniques for Using ARM's Power Management Kit

Semiconductor Technology Academic Research Center An RTL-to-GDS2 Design Methodology for Advanced System LSI

The Physical Design of Long Time Delay-chip

EE584 (Fall 2006) Introduction to VLSI CAD Project. Design of Ring Oscillator using NOR gates

The Digital Abstraction

DESIGNING powerful and versatile computing systems is

ES250: Electrical Science. HW6: The Operational Amplifier

Power Estimation. Naehyuck Chang Dept. of EECS/CSE Seoul National University

EE434 ASIC & Digital Systems. Partha Pande School of EECS Washington State University

IBIS-AMI Terminology Overview

An EM-aware methodology for a high-speed multi-protocol 28Gbps SerDes design with TSMC 16FFC

Gate Delay Estimation in STA under Dynamic Power Supply Noise

A Bottom-Up Approach to on-chip Signal Integrity

Logic Synthesis. Logic synthesis transforms RTL code into a gate-level netlist. RTL Verilog converted into Structural Verilog

Chapter 13: Comparators

3 Circuit Theory. 3.2 Balanced Gain Stage (BGS) Input to the amplifier is balanced. The shield is isolated

Signal integrity means clean

Vishay Siliconix AN724 Designing A High-Frequency, Self-Resonant Reset Forward DC/DC For Telecom Using Si9118/9 PWM/PSM Controller.

Using IBIS Models for Timing Analysis

The Need for Gate-Level CDC

TODAY, ubiquitous electronic systems are an inseparable

Mixed-Signal Simulation of Digitally Controlled Switching Converters

UNIT-II LOW POWER VLSI DESIGN APPROACHES

Lecture 11 Circuits numériques (I) L'inverseur

EC 1354-Principles of VLSI Design

CHAPTER 5 DESIGN AND ANALYSIS OF COMPLEMENTARY PASS- TRANSISTOR WITH ASYNCHRONOUS ADIABATIC LOGIC CIRCUITS

Reading. Lecture 17: MOS transistors digital. Context. Digital techniques:

DIGITAL SIGNAL PROCESSING WITH VHDL

Single-Ended to Differential Converter for Multiple-Stage Single-Ended Ring Oscillators

20V, 2A Buck Switching Regulator

EE241 - Spring 2013 Advanced Digital Integrated Circuits. Projects. Groups of 3 Proposals in two weeks (2/20) Topics: Lecture 5: Transistor Models

Power Supply Noise Induced Jitter Estimation in High Speed Clock Tree for Full Chip Timing Analysis

4. Subtracting an even number from another even number gives an odd number. 5. Subtracting an odd number from another odd number gives an even number

Clock-Powered CMOS: A Hybrid Adiabatic Logic Style for Energy-Efficient Computing

IT HAS become well accepted that interconnect delay

Guaranteeing Silicon Performance with FPGA Timing Models

LSI Design Flow Development for Advanced Technology

Design of High-Speed Op-Amps for Signal Processing

Digital Control Technologies for Switching Power Converters

Analog-aware Schematic Synthesis

The Digital Abstraction

Lecture 11: Clocking

Lecture 11 Digital Circuits (I) THE INVERTER

LM139/LM239/LM339 A Quad of Independently Functioning Comparators

A New network multiplier using modified high order encoder and optimized hybrid adder in CMOS technology

Relationship Between Signal Integrity and EMC

Impedance Matching: Terminations

Static Timing Analysis Taking Crosstalk into Account 1

Fast Estimation and Mitigation of Substrate Noise in Early Design Stage for Large Mixed Signal SOCs Shi-Hao Chen, Hsiung-Kai Chen, Albert Li

QUICKSWITCH BASICS AND APPLICATIONS

Single Event Transient Effects on Microsemi ProASIC Flash-based FPGAs: analysis and possible solutions

Boost Logic : A High Speed Energy Recovery Circuit Family

ETE 112. Structured Programming Laboratory

Transcription:

Interconnect Delay Compensation in Timing Analysis for Designs Containing Multiple oltage Domains Incentia Design Systems, Inc. 1. Introduction A timing signal may flow from one voltage domain to another different voltage domain in a circuit that contains multiple voltage domains. When a cell drives its receiver cell with a different supply voltage, the measuring trip-points at the inputs of the receiver cell (usually 50% of for the interconnect delay and 20%-80% (or 10%-90%) of for the transition delay) could be different from those of the driving cell. A frequently used solution is to insert level shifters between two different voltage domains so that the circuit timing can work as desired. However, a robust timing analysis tool still needs to handle the situation and compensate the difference when level shifters are missing between different voltage domains. This paper presents a compensation technique that has been adopted in Incentia s TimeCraft static timing analyzer to properly model interconnect delay and transition under such a situation. 2. Compensation Method for oltage Difference Assume that a cell is driving an interconnect network, as shown in Figure 1, and its supply voltage is. Figure 1: Interconnect network example 1

We can analyze the output waveform of each receiver point at 1 & 2, by taking the convolution integral with the impulse response. Since the waveform equation at the receiver point can be obtained, we can measure the upper / lower thresholds (20%-80% or 10%-90%) of the transition delays exactly by Newton-Raphson iteration. Figure 2: Waveforms at driving point and receiver points The supply voltage from driving cell is. Thus, the waveform at the receiver point has the same voltage swing from GND to. When the receiver cell has a different supply voltage, such as high or low, as shown in Figure 2, the interconnect delay is compensated by estimating Δt in Figure 2. First, we obtain the measured output transition delay, OUT as shown in Figure 2. Then, we extend the transition delay by dividing it with the trip-point reference (UpperTh & LowerTh from the Liberty). EXT OUT _ (1) FULLSW ( UpperTh LowerTh) _ is the time consumed for the output waveform to take a full swing from EXT FULLSW GND to. To estimate Δt, it is necessary to calculate the voltage difference, Δ, by considering the input threshold trip-point (usually 50%). 2

Δ ( DelayTh ) ( DelayTh ) or receiver high driver ( DelayThreceiver low ) ( DelayThdriver ) (2) Since _ is the time for full swing from GND to, we can estimate the EXT FULLSW time difference for voltage difference, Δ, by Δt EXT _ FULLSW Δt Δ EXT _ FULLSW Δ (3) Substituting Eqn (1) & (2) into Eqn (3) gives the compensation equation, as follows. OUT Δt ( UpperTh LowerTh) or OUT ( UpperTh LowerTh) ( DelayTh DelayTh ) receiver ( DelayTh DelayTh ) receiver high low driver driver (4) In case of rising transition, this time difference is added to the measured interconnect delay under. On the other hand, in case of falling transition, this time difference is subtracted from the measured interconnect delay. For the output transition delay compensation, we multiply the ratio of the receiver supply voltage over the driver supply voltage to the measured output transition delay. COMPENSATED or OUT high (6) OUT low 3

3. Consideration of Level Shifter When a driver or receiver cell is a level shifter, we do not need to compensate the delay difference. A level shifter is a cell that has a level-shifter attribute defined in the cell section of a Liberty file. The level shifter is designed to bridge signals between two different supply voltage domains, and is already characterized with respect to its different input and output supply voltages. Thus, no compensation is needed. 4. Other Compensations Two additional compensations are also considered. Slew derate difference If the driver and receiver cells have different slew derates for upper / lower thresholds, we should also consider compensation. This is always considered, regardless of supply voltage difference, for transition delay compensation. Upper / lower threshold difference This upper / lower threshold difference is only applied when wire-load model is used for this net. When RC interconnect network is annotated to the net, we measure the upper / lower thresholds exactly from the output waveform equation. Thus, it is automatically considered. Refer to Section 5 to see when these compensations are applied. 5. Parameter to Control Compensation Compensation due to the supply voltage difference is based on the assumption that the interconnect delay should be measured at the threshold point of the receiver s supply voltage reference. For example, assume that the driver cell is in 1.2 voltage domain, and the receiver cell is in 0.9 voltage domain. When the driver is taking the rising transition, the interconnect delay should be measured between the time points of 0.6 (50% of 1.2) and 0.45 (50% of 0.9 ). Since 0.45 can come first, it requires compensation. In timing analysis, we can address this issue by using automatic compensation. However, this may be a design problem since a level shifter is needed from design point of view. Furthermore, consider that a cell is driving the interconnect network again in Figure 1. Since the driver cell has as the supply voltage, the output response at each receiver point has voltage swing too. If the driver cell is in 0.9 voltage domain and the 4

receiver is in 1.2 voltage domain, we have to measure 0.24 (20% as lower threshold) and 0.96 (80% as upper threshold) for the transition delay. But the supply voltage of the driver cell is 0.9, and thus, the output waveform never reaches 0.96 at the receiver point. Although the compensation can be done in timing analysis, it is not done correctly. In the early design stage, you may want to quickly get the preliminary timing analysis result of the design without fixing the problem caused by signals crossing different voltage domains. In this case, automatic compensation can be used. On the other hand, in the later design stage when place and route are done, you should make sure level shifters are inserted correctly whenever signals are crossing different voltage domains. Automatic compensation should not be applied. And signals crossing different voltage domains without proper level shifters in place should be detected and fixed at this stage. Therefore, a parameter is introduced to turn on / off the automatic compensation. The following table summarizes what compensations, including voltage, slew derate, and upper/lower threshold, are considered together with the parameter setting in each case. When proper level shifters are in place, no voltage compensation will be done when crossing different voltage domains no matter you turn on this parameter or not. In the early design stage, usually wire load model is used for timing analysis. We suggest you turn on the automatic compensation because level shifters between different voltage domains may not be inserted at this time. In the later design stage, detailed RC information is available and back annotated to timing analysis. At this stage, you should turn off the automatic voltage compensation so that any remaining design problems caused by voltage differences can be observed. When RC is annotated When WLM is used Level Shifter without Level Shifter with Level Shifter parameter true false true false delay Δ - - - transition Δ ΔSlewDerate ΔSlewDerate ΔSlewDerate ΔSlewDerate Level Shifter without Level Shifter with Level Shifter parameter true false true false delay Δ - - - transition Δ ΔSlewDerate ΔSlewDerate ΔSlewDerate ΔSlewDerate Δupper/lowerTh Δupper/lowerTh Δupper/lowerTh Δupper/lowerTh 5

6. Incentia Timing Analysis Products Incentia offers static timing analysis (STA), signal integrity analysis, and statistical STA through its TimeCraft product line. TimeCraft is a full-chip, gate-level static timing analyzer for timing sign-off. It is the fastest STA available in the market and has been proven through numerous customer tape-outs. TimeCraft has demonstrated unparalleled advantages in runtime and capacity that dramatically reduce total timing verification turnaround time. TimeCraft s built-in delay calculator takes in parasitic RC information through SPEF back annotation file, and libraries in NLDM, CCS, or ECSM format. It generates accurate timing analysis results, within 2% of SPICE results in general. It also has special algorithms to handle extreme cases, such as long wires, long wires with fat metals, high fan-out nets, mesh RC networks, parallel buffers, and cross-linked non-tree clock networks. TimeCraft performs automatic interconnect delay compensation when crossing different voltage domains as described in this paper. 6