CMOS Transistor

Layout KungFu

Lee Eng Han

Valerio B. Perez
Mark Lambert Cayanes
Mary Grace Salaber
Copyright © 2005 by Lee Eng Han.

All rights reserved. No part of this publication may be reproduced,

stored in a retrieval system or transmitted in any form or by any means,
mechanical, photo-copying, recording, or otherwise, without the prior
written permission from Lee Eng Han, who is one of the authors of the
publication.
Table of Contents

Preface

1. Introduction .................................................................. 1

2. MOS Transistors ........................................................... 2

3. Fabrication of MOS Transistor ..................................... 5

4. Layout a Single Transistor .......................................... 10

First Stroke The basic transistor layout ....................... 11

Second Stroke Compact the transistor layout .................. 12

Third Stroke Speed up the transistor ............................. 16

Forth Stroke Clean up the substrate Disturbances ........ 18

Fifth Stroke Balancing area, speed and noise .............. 24

Sixth Stroke Relief the stress ........................................ 26

Seventh Stroke Protect the gate ........................................ 27

Eighth Stroke Improve yield ............................................28

Preliminary Release i www.eda-utilities.com

5. Layout Several Transistors ......................................... 30

Ninth Stroke Close proximity .........................................31

Tenth Stroke Interdigitated layout ................................. 31

Eleventh Stroke Dummy transistor ..................................... 36

Twelfth Stroke Two-dimension interdigitated layout ....... 38

Thirteenth Stroke Guard ring for the matched transistors ..... 40

Fourteenth Stroke Keep NMOS away from N-well .............. 40

Fifteenth Stroke Orientate the transistor ............................. 41

Sixteenth Stroke Match the interconnects ........................... 42

Seventeenth Stroke The unmatchable ...................................... 44

6. Verifying the Transistor Layout ............................................. 46

Eighteenth Stroke Physical verification beyond DRC and

LVS .......................................................... 55

Bibliography ..................................................................... 57

Preliminary Release ii www.eda-utilities.com

Preface

Many IC design books emphasize on circuit design theories and there is

little coverage on custom circuit layout techniques. Hence this book is
specially written to focus on the custom circuit layout techniques. It is
an easy book for the custom layout engineer as prior knowledge on
circuit design is not necessarily required to understand the content of the
book.

This book is written for two groups of audiences – New & Experienced
custom layout designers. The new comer to custom layout would benefit
from the wider perspectives in implementing the design to a layout. The
experienced custom layout designers, on the other hand, will have a
better appreciation of the rationale behind the layout practices.

Feedback to the authors

Please feel free to let us have your valuable inputs for future
improvement. We would also appreciate if you could let us know if the
book is of value to you. We can be contacted at
layout@eda-utilities.com. Your feedback is most welcome.

Acknowledgements
The authors wish to thank Hwee Ling Goh from STMicroelectronics for
reviewing the technical details. The authors also like to thank Karen
Phang for her effort in proof-reading the book and giving invaluable
suggestions in making the book easier to read.

Preliminary Release iii www.eda-utilities.com

 Chapter 1

Introduction

Welcome to Layout KungFu!

There are many levels of Layout KungFu, but we will focus on the
fundamental and CMOS transistor layout is what you will find in this
KungFu book. If you enjoy the KungFu and want to find out more, we
could start a Layout KungFu series!

Please place your feet firmly on the floor, bend your knee at 90 degree
and take a deep breath. Now, are you sitting comfortably on the chair?
We shall begin the KungFu journey.

Preliminary Release 1 www.eda-utilities.com

 Chapter 2

MOS Transistor

There are two types of MOS transistors. They are called n-channel MOS
transistor (NMOS) and p-channel MOS transistor (PMOS). Each
transistor has 4 terminals, namely drain (D), gate (G), source (S) and
bulk (B) as illustrated in the transistor symbols.

The bulks of the PMOS and the NMOS are usually connected to power
and ground respectively. If the bulk terminal is omitted from the
schematic symbol, the connections can be assumed to be what is shown
in the following diagram.

Preliminary Release 2 www.eda-utilities.com

A circuit design usually uses more than one type of PMOS or NMOS to
cater for different power supply voltages. For example, a design may use
thicker gate-oxide transistors to operate in higher power supply at the
I/O interfaces. In order to differentiate the different voltage range
transistors, sometimes circuit designers conveniently make use of the
“depletion-mode” transistor symbol for this purpose. Typically, the
schematic symbols for the depletion-mode transistors are represented
with a thicker gate drawing.

The MOS transistor’s performance varies with its channel length (L) and
channel width (W). The drain current (ID) that flows through the
transistor operating in the saturation mode is shown in the following
equation.

W 
(
I D = K ∗   ∗ VGS − Vt ) ∗ (1 +λV
2
DS )
L

Preliminary Release 3 www.eda-utilities.com

where K and λ can be taken as process technology constants.

Note that ID is proportional to the ratio of W over L. Typically L is kept

to the minimum dimension allowed in the design rule, and is to be
layout exactly as indicated in the schematic. However, this is not always
the case for W. We will elaborate more in chapter 4.

Transistor Spice Model

The Fab supports different types of transistor. For example,

transistors can have different types of VT to compromise
between leakage power and speed. Transistors can also have
different types of gate oxide thickness to allow the transistors
to operate at different voltage range.

Every type of transistor is associated with its own transistor

spice model. PMOS and NMOS also have different transistor
model. A transistor model includes a set of parameters that
define the electrical performance of the transistor. The design
engineer uses the transistor spice model and the circuit netlist
to simulate their design. The layout engineer has to craft the
transistor layout that match the transistor spice model and the
designed W and L.

Preliminary Release 4 www.eda-utilities.com

 Chapter 3

Fabrication of MOS Transistor

This chapter briefly describes a very simplified version of the

fabrication of a transistor on the silicon wafer. The ability to visualize
the cross-section of a layout is a basic skill that all layout designers
should master.

Step 1 : Well formation stage. Implants n-type impurities into the wafer
followed by diffusing the impurities deep into the substrate to form the
N-Wells. For CMOS process, the silicon substrate is usually p-type.

Step 2 : Active & isolation stage. Thick oxide is grown outside the active
areas. Active areas are defined as areas where the CMOS transistors are
fabricated. Thick oxide is also known as field oxide. Field oxides isolate
the transistors from one another.

Preliminary Release 5 www.eda-utilities.com

Step 3 : Gate oxide formation stage. A thin gate oxide is grown across
the wafer. Gate oxide of only tens of silicon oxide atoms thick is created
during the fabrication process with the current technology. Gate oxide is
the insulator between the transistor’s gate and its channel. Gate oxide
refers to the “O” in “MOS” which stands for Metal-Oxide-
Semiconductor.

Step 4 : Gate formation stage. Poly (i.e. polysilicon) are deposited on the
wafer. The poly that are deposited on the gate oxides are the gates of the
transistors which are usually known as gate poly. The gate poly will
incline upward when it extends over the field oxide. The gate oxide in
the active area that are not covered by the gate poly will be etched away
to form the source and the drain of the transistor.

Preliminary Release 6 www.eda-utilities.com

Step 5 : Source and drain formation stage. P-type and n-type impurities
are implanted into the active areas. The impurities are diffused into the
silicon to form the source terminals and the drain terminals. As the
impurities diffuse both vertically and laterally, the gate poly will slightly
overlap the sources and the drains which will result in gate overlap
capacitances. The diffusions for the sources and the drains of NMOS
and PMOS are N-diffusion (N-diff) and P-diffusion (P-diff) respectively.

P-diff in p-substrate is known as p-tap, while n-diff in N-well is known

as n-tap. Connections from the metal routings to the substrate and the N-
wells are made through the p-tap and the n-tap. This is necessary to
ensure that the wells are properly tied down and the transistors are
isolated. The p-substrate should be biased to the lowest voltage potential
while the N-well should be biased to the highest voltage potential. In
this way, all the P-N junctions are reverse biased and hence the
transistors are electrically isolated from one another as shown in the
diagram below.

Preliminary Release 7 www.eda-utilities.com

Isolating the transistors with thick field oxide is commonly found in
0.35um1 and larger technologies. For 0.25um and smaller technologies,
shallow trench isolation (STI) shown in the diagram below is more
commonly used to isolate the transistors. In STI fabrication, trenches are
etched into the wafer and filled with silicon oxide to isolate the islands
of transistor active area.

1
A technology of 0.35um means that the shortest channel length (L) of a transistor is
0.35um.

Preliminary Release 8 www.eda-utilities.com

 Source, Drain Gate and Bulk

The drains, sources, gates and bulks of the NMOS and the
PMOS are illustrated in the figure below. Observe the cross
section of the layout and you will find the followings.

• The drain and the source are fabricated in the same

way.
• The bulk of all the NMOS are connected together.
• The bulk of all the PMOS in the same N-well are
connected together.

Preliminary Release 9 www.eda-utilities.com

 Chapter 4

Layout a Single Transistor

Before we can start to layout the transistor, we have to settle a few

logistics.

Firstly, the following legends are adopted.

Secondly, there are more than one ways to draw an identical layout. For
example, the two layouts in the following diagram are the same. Layout
in “Method A” uses P-diffusion layer and N-diffusion layer. Layout in
“Method B” uses diffusion layer and implant layer. The illustrations in
this book use “Method A”.

Preliminary Release 10 www.eda-utilities.com

First Stroke. The basic transistor layout

The basic transistor layout as illustrated has a channel length (L) of

0.2um and a channel width (W) of 20um. The source diffusion and the
drain diffusion should be filled with the maximum number of contacts to
reduce the resistance of the connection from the metal to the diffusion,
and to maximize the amount of current that can flow through the
contacts.

Preliminary Release 11 www.eda-utilities.com

 Bulk Connection

The substrate and the N-well are doped lightly. A direct

connection from the metal routing layer to the bulk is not
allowed. The connection should be made through a higher
doped diffusion such as the p-diffusion and the n-diffusion in
order to establish a good contact. In some fabrication
process, two additional layers are used for these connections.
These layers have much higher doping than the diffusions for
the source and the drain.

Second Stroke. Compact the transistor layout

The basic transistor layout from the first stroke has a rather awkward
aspect ratio. Putting transistors with fixed aspect ratio together will not
give a compact layout. Fortunately, the aspect ratio of the transistor can
be modified by using the transistor current equation shown in page 3.
For example, the transistor with a width of 20um and a length of 0.2um
is similar to having four transistors connected in parallel, each with a
width of 5um and a length of 0.2um.

The next layout shows the transistor with four fingers. The layout has a
better aspect ratio than the one in the first stroke. Note that the
connection to the bulk is omitted in the layout. Bulk connection is
discussed later.

Preliminary Release 12 www.eda-utilities.com

So far everything looks fine. But it is not. The transistor with a width of
20um and a length of 0.2um is not exactly the same as four transistors
that are connected in parallel, each with a width of 5um and a length of
0.2um.

Layout cannot be fabricated exactly as drawn in the layout due to the

limitations in the manufacturing process, such as process tolerances and
mask misalignment. Some of the manufacturing limitations are captured
in the Spice transistor model. Two of the main parameters in the Spice
transistor model are DW and DL. DW shows the delta difference of
drawn W from effective W. DL shows the delta difference of drawn L
from effective L.

The next diagram shows the transistor effective channel length can be
affected by under-etching or over-etching of the poly, as well as the
amount of lateral diffusion under the gate. The effective channel width
of the transistor is affected by the “bird peak” of the isolation scheme. In
addition, the inclination of the gate poly up to the field oxide makes it
difficult to determine the exact width of the transistor.

Preliminary Release 13 www.eda-utilities.com

Considering an example where DL is 0.015um and DW is 0.045um. To
simulate a typical “fast” corner transistor, the transistor is modeled to
have a narrower L and a wider W. On the contrary, a “slow” corner
transistor is modeled with a wider L and a smaller W. Hence, at the fast
corner DL is negative (i.e. -0.015um) and DW is positive while at the
slow corner DL is positive and DW is negative (i.e. -0.045um). The
following table shows L and W of the transistors from the first stroke
and the second stroke, and at slow, typical and fast corners to emulate
manufacturing process tolerances.

L (um) W (um)
Transistor from first stroke Fast corner 0.185 20.045
Typical corner 0.200 20.000
Slow corner 0.215 19.955
Transistor from second stroke Fast corner 0.185 20.180
Typical corner 0.200 20.000
Slow corner 0.215 19.820

The width of the transistor from the first stroke differs by 0.045um
between the slow or the fast corner from the typical corner. However,
the width of the transistor from the second stroke differs by 0.18um
between the slow or the fast corner from the typical corner. This could
pose circuit performance deviation for the circuit designer if left
unaccounted for.

The layout in the second stroke can be statistically significant. The

circuit designer would perform Monte Carlo simulation to “center the
design” so as to improve manufacturing yield. In Monte Carlo
simulation, a small statistical variation is added to W and L of every

Preliminary Release 14 www.eda-utilities.com

transistor in the circuit. Folding a transistor to four fingers means that a
larger variation in the circuit performance since the variation made to
the four fingers is four times larger than the same variation made to a
single transistor. The layout can be challenging for the circuit designer
to optimize the design for better yield.

Operating Corners

Variations in fabrication process, ambient temperature and

supply voltage affect the electrical performance of the
transistors. For example, a higher temperature and a lower
supply voltage make the transistor operate slower. Many
companies require the circuit designers to verify the
operation of the circuit design by simulating the design in
slow (SS) corner, typical corner (TT) and fast corner (FF).
Some companies also require the circuit designers to simulate
the design with fast NMOS and slow PMOS corner (FS), and
slow NMOS and fast PMOS corner (SF). Some companies
require the circuit to work correctly within a 3 sigma spread
from the typical corner (FF3, SS3).

Is folding a transistor into multiple fingers a bad idea? Actually, it is an

excellent idea. Everyone uses it. Second stroke makes the layout
compact. Third stroke is a continuation of the second stroke and it will
enable the design to run at a higher speed!

Before we leave the second stroke, remember that whatever trick you
used in the layout, it must be communicated to the circuit designer. The
circuit designer needs to update the schematic to reflect the layout
implementation. Using the same layouts from the first stroke and the
second stroke as an example, the original spice netlist (i.e. the first
stroke) of the transistor is

M1 D G S B N_model L=0.2u W=20u

The ‘N_model’ in the netlist is the name of the NMOS transistor model.
The spice netlist for the layout in the second stroke is

Preliminary Release 15 www.eda-utilities.com

M1 D G S B N_model L=0.2u W=5u M=4

The circuit designer simulates the design using the spice netlist. Thus it
is important for the spice netlist to represent the layout implementation
as closely as possible.

Third Stroke. Speed up the transistor

What can you do in the layout to make the transistor operates faster?

Reducing the parasitic capacitance and resistance would increases the

speed of the transistor. We need to understand where the parasitic
capacitances and resistances are. The following diagram shows the
simplified capacitances associated with the drain, gate and source of a
transistor to the bulk.

Overall Csb is dependent on the area of source (AS) and perimeter of

source (PS). Similarly, overall Cdb is dependent on the area of drain
(AD) and perimeter of drain (PD). Both Csb and Cdb have components
that are dependent on the diffusions in the proximity. The values of AS,
AD, PS and PD of a transistor can be extracted from the layout. Post-
layout spice netlist should include these parameters.

Preliminary Release 16 www.eda-utilities.com

The frequency response of the transistor can be improved if the source
capacitance and drain capacitance are reduced. Study the transistors
from the first stroke and the second stroke. Can you see that the
transistor from the second stroke has the drain area reduced by half and
the source area reduced by a quarter when compared to the transistor
from the first stroke?

The folded transistors have smaller gate resistance as shown in the

diagram below. This will make the transistors turn on and off faster than
the transistor in the first stroke.

In general, try to fold a transistor to an even number of fingers. For

example, the transistor from the second stroke is folded to four
transistors, and is drawn to optimize the frequency response of the drain.

Resistivity of the poly is a few orders higher than the resistivity of the
metal. The parasitic capacitances between the poly and the substrate,
and between the metal and the poly, are also very much larger than the
parasitic capacitance between the metal to the substrate. Hence, using
poly for interconnect could degrade the frequency response of the
transistor if the poly routing is not optimized carefully. Refer to the
following layout. The layout at the top uses both poly and metal to

Preliminary Release 17 www.eda-utilities.com

connect the gates. The first layout at the bottom uses only metal to
connect the gates. The second layout at the bottom is a popular method
to have multiple contacts per gate. To improve yield, the contacts for the
gates in this layout are placed slightly further from the transistors so as
to increase the distance between the diffusion and the poly that are
running in parallel. The layout at the top will have poorer frequency
response due to additional parasitic capacitances from the metal to the
poly, and from the poly to the substrate.

Forth Stroke. Clean up the Substrate Disturbances

Performances of analog designs are sensitive to electrical disturbance.

Disturbance in the substrate should be minimized as much as possible.
Two common types of substrate disturbance are

• Disturbance from minority carrier

• Substrate coupling noise

Preliminary Release 18 www.eda-utilities.com

Disturbance from minority carrier

Minority carriers are injected into the substrate from the source
diffusions and the drain diffusions when

• The source potential or the drain potential of NMOS is below the

substrate potential

• The source potential or the drain potential of PMOS is above the

N-well potential

There are several possibilities for the above conditions to happen.

Examples are

• Inductive ground path causes the ground in the substrate to

bounce

• Resistive power and ground path from the power pins to the
substrate and the N-well

• Fast switching signal with significant overshoot

The drifting of the minority carriers in the substrate and the N-well
create a potential different that can affect the performance of the circuit,
or trigger a latch-up.

Substrate coupling noise

A reverse biased diode has the electrical properties of a capacitor.

Circuit signals can be coupled through the substrate as illustrated in the
diagram below.

Preliminary Release 19 www.eda-utilities.com

To reduce disturbances from minority carrier, you may use guard ring in
the following configuration around noisy transistors.

• Surround NMOS in the p-substrate with N-well guard ring. Tie

the N-well guard ring to VDD. The N-diffusions from the
NMOS could inject stray electrons into the substrate. These stray
electrons could be collected efficiently by the N-well guard ring
that is biased to VDD to attract the electrons.

• Surround the PMOS in the N-well with P-diffusion guard ring.

Tie the P-diffusion guard ring to ground. P-diffusions from the
PMOS inject stray holes into the N-well. These stray holes could
be collected efficiently by the P-diffusion guard ring that is
biased to ground to attract the holes.

For the guard rings to be effective, the resistance in the path from the
straying minority carrier to the guard ring and then to the voltage source
must be kept as low as possible. Hence, the minority carrier noise guard
rings are made wider so as to decrease it resistance. Ideally, the guard
rings should be placed as closely to the likely noise sources as possible.
The guard rings are also placed around the critical transistors to
minimize stray electrons and stray holes from affecting the critical
transistors.

To reduce substrate coupling noise, you may use guard ring in the
following configuration around critical transistors.

• Surround NMOS in the p-substrate with p-tap guard ring that is

connected to ground.

• Surround PMOS in the N-well with n-tap guard ring that is

connected to VDD.

The following layouts show both PMOS and NMOS surrounded with
double guard rings.

Preliminary Release 20 www.eda-utilities.com

It is generally believed that N-well guard ring in p-substrate and P-
diffusion guard ring in N-well are not of much use. The stray electrons
and holes travel deep into the substrate and are not collected by the

Preliminary Release 21 www.eda-utilities.com

guard rings. However, putting a guard ring is better than leaving the
space empty since we still have to keep the noisy transistors at a distance
away from the other transistors.

Double Guard Ring for PMOS

The most common
way to add a double
guard ring to PMOS
is shown here. The
outer P-guard ring
acts as both a guard
ring for PMOS, and
as p-tap for the
substrate. However,
the P-guard ring
might be too far
from the PMOS to be effective in collecting stray hole
from the PMOS.

An important layout practice is to ensure that there is no (or very little)

current flowing through any part of the guard ring. Consider the layout
in the diagram below. A current flowing in the p-type guard ring raises
its potential above VSS. If the potential of the n-diffusion next to the p-
type guard ring (shaded in the diagram) is at VSS, the PN junction
potentially becomes forward biased, results in hole injected from the
guard ring into the substrate!

Preliminary Release 22 www.eda-utilities.com

A more subtle case is shown below. In the circuit design, the source of
the transistor is connected to VSS. It is very convenient to tap the VSS
from the guard ring as shown in the layout below. If the peak current
going through the transistor is very small, the layout may be acceptable.
Otherwise, tapping power from the guard ring is not allowed.

Preliminary Release 23 www.eda-utilities.com

 Floating Well

Wells that are tied to and varies with

the source terminal of the transistor
are called floating well. Such an
example is shown on the right.
Floating well is quite commonly used
in analog design to reduce the “body
effect” of the transistor. The PMOS
with this type of bulk connection needs an N-well on it
own, and hence the layout has a huge area overhead. It is
important to identify all the PMOS that sit on a floating
well before floor plan the layout.

Fifth Stroke. Balancing Area, Speed and Noise

The guard rings in the forth stroke take up a lot of area. The guard rings
also add capacitive load to the transistor as illustrated in the diagram
below.

Instead of using a full guard ring, you may consider using a U-shape
guard ring. Some designers do not favor the use of U-shape guard ring
while some designers use U-shape guard ring only for p-tap and n-tap.
The following layout shows a NMOS with U-shape guard rings.

Preliminary Release 24 www.eda-utilities.com

If reducing the area of the layout is of utmost priority and you do not
expect much minority carrier noise in the silicon, then you may consider
reducing the numbers of local N-well guard rings and P-diffusion guard
rings.

Guard Ring

The guard rings do not have to be rectangular. Two ways to

add a guard ring around a group of transistors is shown here.

Preliminary Release 25 www.eda-utilities.com

Sixth Stroke. Relief the Stress

The stress from the STI onto the drain and source has effect on the
performance of NMOS and PMOS. The impact of the STI stress
depends on the source and drain overhang (which are indicated as SA
and SB in the following diagram) of the transistor active island.

To reduce the effect of STI stress, the source and drain diffusions need
to be extended when they are next to the STI. However, a large diffusion
also increases parasitic capacitance and layout area. A better approach is
to insert one or more dummy transistors at each end of the transistor as
illustrated in the diagram below. Note that the dummy transistor must
share the diffusion with the non-dummy transistor.

Preliminary Release 26 www.eda-utilities.com

Seventh Stroke. Protect the Gate

The gate oxide underneath the poly is incredibly thin. If the charges
accumulated on the poly is sufficiently large, the charges accumulated
can damage the gate oxide. This is known as process antenna effect.

The maximum amount of charges that can be accumulated on the poly is

proportional to the area of the poly1. Thus, an effective layout practice to
prevent process antenna violation is to stay within the antenna ratio
design rule of the respective technology. Some general guidelines are

• Minimize the use of poly for routing

• Minimize the use of poly to connect the gates together

Some knowledgeable readers might suggest using diode to protect the

poly from antenna ratio violation. Antenna diode is only effective in
preventing antenna violation from metal routing, and does not help in
antenna violation due to poly. The reason is simple. The diodes are
made from diffusions, but the poly is deposited onto the wafer before the
diffusions are implanted into the wafer. Hence, the diode does not exist
at the time the poly is fabricated on the wafer!

Besides protecting the gates from process antenna effect, other measures
to protect the gate are

• Do not place contact and via directly on top of the transistor’s

gate.

• Avoid routing over the gates of critical transistors. Refer to the

following diagram for an example.

• Avoid routing over active areas of critical transistors.

An exception to the guidelines is to allow routing over decoupling MOS

capacitors. This is done to compromise for a shorter routing.

1
It is more correct to say that charges are accumulated on the perimeter side-wall area
of the poly, which can be calculated as the perimeter of the poly multiple by the
thickness of the poly.

Preliminary Release 27 www.eda-utilities.com

Eighth Stroke. Improve Yield

The most compact layout does not give the best manufacturing yield.
Use the layout practices discussed here to enhance yield. The practices
are also illustrated in the following layout.

Avoid use of single contact or via

Avoid using single contact or via. Use at least double contact or double
via whenever possible. A high percentage of IC manufacturing defects is
related to faulty contact and via issues.

Metal coverage of contacts and vias

Always give additional metal coverage on the contacts and the vias if
they are located at the end of the metal line.

Preliminary Release 28 www.eda-utilities.com

Poly extension from diffusion

Exceed the DRC requirement for poly overhang rule and minimum
distance from poly to diffusion rule whenever possible. In particular, do
not run poly near to the diffusion edge.

Note

Chapter 5 and chapter 6 will not be

included in the pdf release.

Preliminary Release 29 www.eda-utilities.com