Cadence Kits
Incisive functional verification
Encounter digital IC design
Virtuoso custom design
Allegro IC-PKG-PCB co-design
OrCAD PCB design
System-in-package design
Design for manufacturing
PRODUCTS
Assura DRC
Assura LVS
Cadence CMP Predictor
Cadence Litho Electrical Analyzer
Cadence Litho Physical Analyzer
Cadence MaskCompose Suite
Cadence QuickView Viewer
Cadence QRC Extraction
Diva Physical Verification
Dracula
Encounter Test
Physical Verification System
DESIGN TASKS
Silicon analysis
Analysis and signoff
IP catalog
Print-friendly version
COSMIC

COSMIC TECHNICAL DESCRIPTION

TABLE OF CONTENTS:

  1. Objectives
  2. COSMIC Methodology Overview
  3. COSMIC Test Structures
  4. COSMIC Schematics
  5. Register to Download Test Structures

1.0 Objectives

This document describes and illustrates the test structures that can be used for total interconnect capacitance characterization with semiconductor wafers. Measurement results from these structures fabricated in silicon provide a standard for comparing the accuracy of interconnect extraction software.The original structure was published by U.C. Berkeley.



2.0 COSMIC Methodology Overview

COSMIC is an active test chip approach, also known as an on-chip method, for measuring interconnect (wiring) capacitance in an IC. The active approach has many advantages over the commonly used passive approach of measuring wire capacitance, such as:

  • Realistic interconnect patterns
  • Measurements down to femtofarads
  • Smaller test structures

The principle behind the COSMIC approach of measuring total wire capacitance for any interconnect pattern is similar to the one proposed by Chen, et al. (1996), and is shown in Figure 1. It consists of nMOS and pMOS transistors configured to function like an inverter. The two "pseudo-inverter" configurations are identical in every respect except that only one includes the capacitance of the interconnect to be measured. The Vp and Vn signals of Figure 1 are two non-overlapping signals that ensure that, at a given time, only one of the two transistors in the pseudo-inverter is conducting. These two signals are generated from an external dual-pulse generator that must be able to control pulse width independently for the two signals, Vp and Vn. These signals are applied through on-chip inverters to the gates of the pseudo-inverters.



Figure 1. The COSMIC Approach


The capacitance, C, in Figure 1 is determined by measuring the difference in the drain current of the two inverters and deriving the value using the following equations:

Substituting capacitive current:

Results in this equation:

In these equations,

  • T = time period of the pulses Vp and Vn
  • f = frequency of the pulses
  • Vdd = DC power supply
  • i1 = total drain current on left hand side (LHS) peudo inverter
  • i2 = total drain current on right-hand side (RHS) pseudo inverter

The current is normally in the pA to nA range and thus can easily be measured by a current meter. The equipment required for measuring the capacitance using COSMIC structures is:

  • A dual pulse generator
  • A current meter
  • A DC power supply with low noise (microvolt range)

An oscilloscope is helpful for looking at the pulses being applied to the gates of the pseudo inverter but not required.

Since the reference structures were first published by U.C. Berkeley in 1996, changes have been made to the methodology enabling whereby not only total capacitance but also coupling capacitance can be measured between any two lines in the presence of any other lines.

The COSMIC methodology has been used and tested to measure total capacitance of a line in the presence of any other lines for any process down to 90nm technology node. In addition, the methodology has been extended to measure coupling capacitance of any two lines in the presence of any other lines. As an example, Table I shows measured coupling capacitances using COSMIC approach for 180nm Al process. The measured data is compared with field solver and full-chip parasitic extractor Fire & Ice®. The results validated Fire & Ice to be within 10% of silicon. Note that structures A, B and C are inter-digited lines, two parallel lines, and two crossing lines, with different line width and spacing.

Test structure Measured coupling capacitance (fF) Field solver capacitance (fF) Error (FS vs. model) (%) ICE (fF) Error (ICE vs. model) (%)
A36.516.834.926.672.46
A419.1919.551.8818.51-3.54
B16.687.136.746.943.89
B27.858.325.997.971.53
B33.0862.81-8.942.857-7.42
B43.0862.95-4.412.891-6.32
C13.7583.60-4.203.7760.48
C23.4143.23-5.393.4410.79
C32.9782.85-4.302.714-8.87

To decrease interconnect delay, copper and low k dielectrics have been introduced at 130nm mode to reduce interconnect resistance and coupling capacitance. Manufacturing effects such as wide edge enlargement (WEE) and pattern density dependent metal width and thickness variations etc. have made modeling and characterization of capacitance become difficult. Therefore COSMIC has been enhanced to provide coupling capacitance extraction that takes above manufacturing effects into consideration, and at atto-farad level accuracy, enabling accurate interconnect extraction verification at advanced process nodes. Since COSMIC enables capacitance measurement between two lines in the presence of any other lines, by using comb type test structure between the two metal plates (see Figure 2), the non-ideal trapezoidal Cu metal line widths at the top (Wt) and bottom (Wb) surfaces can be measured electrically.


Figure 2. Inter-digital metal lines between metal plates

Inter-digited metal lines (shown here as 2-D cross-section) sandwiched between metal plates on the top and bottom. COSMIC measures Cc, Ct and Cb, from where we can find Wt, Wb, St, Sb and T.

Thus COSMIC enables one to characterize the back end process accurately as in an example, measurements were made on inter-digited test structure similar to Figure1, for 130nm Cu process, the results are shown in Table II. The top and bottom widths were compared with SEM cross-section. The COSMIC results are in excellent agreement with SEM data.

Table II. SEM measured cross-section and COSMIC measured physical dimensions of structures similar to that shown in Figure 2 for 130nm Cu process.

Orthogonal Drawn SEM COSMIC Error %
Thickness0.350.400.3892.8
Top ILD thickness0.400.400.401-0.3
Bottom ILD thickness0.400.410.4012.2
Wire top width0.220.260.2455.8
Wire bottom width0.220.210.1938.1
Top spacing0.220.190.195-2.6
Bottom spacing0.220.230.247-7.4


3.0 COSMIC Test Structures

The basic structure layout has six pads and assumes n-well CMOS process, where the well pad is connected to Vdd.

The structure layout samples are drawn for Metal 1 and are provided in MOSIS process based on units of Lambda. For complete characterization, these structures should be drawn for all other metal lines for varying widths and spacing, including poly lines. You can download the following cells for Metal 1, in GDSII format, for detailed review from the COSMIC homepage.

In all modules, LHS stub is included to represent parasitic capacitance of the RHS that is not intended to be measured.

Note: "2X" and "4X" in instance names indicate two and four times, respectively, the minimum width and spacing of the line.


Cell Name Module Label Comments Figure
MOSIS_A_M1 A1 Single line to substrate. Node to be measured is 600 Lambda units long. LHS has stub 60 units long. Figure 3
MOSIS_B_M1M2PLATE B1 Single line 600 Lambda units long under Metal 2 plate (grounded). LHS has stub 60 units long. Figure 4
MOSIS_C1_2X_M1 C1 Interlacing comb structure over substrate. 10 fingers. LHS stub is 120 units long. Figure 5
MOSIS_C2_2X_M1 C2 Non-interlacing comb structure over substrate. 10 fingers. LHS stub is 120 units long. Figure 6
MOSIS_D1_2X_M1M2 D1 90-degree single crossing line. Node to be measured is 600 Lambda units. Crossing line is 300 units. LHS stub is 60 units long. Figure 7
MOSIS_F1_4X_M1M2 F1 Three crossing lines. Node to be measured is 600 units long. Crossing line is 300 units. Two side crossing lines are grounded. LHS stub is 60 units long. Figure 8
MOSIS_G_M1 G1 Lateral three lines. RHS top and bottom lines grounded. LHS stub is 60 units long. Figure 9
MOSIS_G2_M1 G2 Lateral two lines. LHS stub is 60 units long. Figure 10
MOSIS_I_2X_M1M2 I1 Stacked lines 600 Lambda long with Metal 2 line grounded. LHS stub is 60 units long. Figure 11
Note: "2X" and "4X" in instance names indicate two and four times, respectively, the minimum width and spacing of the line.


4.0 COSMIC Schematics

The following schematics represent the test structures with the cell and module name. In all the figures. Metal 1 is not filled, while Metal 2 is shaded. For clarity, the well pad connected to Vdd is not shown in the figure.


Figure 3. Module A1

Cell Name: MOSIS_A_M1
Single line to substrate. Node to be measured is 600 Lambda units long. LHS has stub 60 units long.



Figure 4. Module B1

Cell Name: MOSIS_B_M1M2PLATE
Single line 600 Lambda units long under Metal 2 plate (grounded). LHS has stub 60 units long.



Figure 5. Module C1

Cell Name: MOSIS_C1_2X_M1
Interlacing comb structure over substrate. 10 fingers. LHS stub is 120 units long.



Figure 6. Module C2

Cell Name: MOSIS_C2_2X_M1
Non-interlacing comb structure over substrate. 10 fingers. LHS stub is 120 units long.



Figure 7. Module D1

Cell Name: MOSIS_D1_2X_M1M2
90-degree single crossing line. Node to be measured is 600 Lambda units. Crossing line is 300 units. LHS stub is 60 units long.



Figure 8. Module F1

Cell Name: MOSIS_F1_4X_M1M2
Three crossing lines. Node to be measured is 600 units long. Crossing line is 300 units. Two side crossing lines are grounded. LHS stub is 60 units long.



Figure 9. Module G1

Cell Name: MOSIS_G_M1
Lateral three lines. RHS top and bottom lines grounded. LHS stub is 60 units long.



Figure 10. Module G2

Cell Name: MOSIS_G2_M1
Lateral two lines. LHS stub is 60 units long.



Figure 11. Module I2

Cell Name: MOSIS_I_2X_M1M2
Stacked lines 600 Lambda long with Metal 2 line grounded. LHS stub is 60 units long.

5.0 Register to Download Test Structures

You may now register to download COSMIC test structures.





COSMIC footnotes

The COSMIC test structures developed by Simplex are used for total interconnect capacitance characterization with semiconductor wafers.

Measurement results from these structures fabricated in silicon provide a standard for comparing the accuracy of interconnect extraction software. COSMIC was based on the original structure which was published by U.C. Berkeley:

James C. Chen, Bruce W. McGaughy, Dennis Sylvester, and Chenming Hu,"An On-Chip, Attofarad Interconnect Charge-Based Capacitance Measurement (CBCM) Technique," Proc. of IEDM 1996, pp. 69-72.