Вы находитесь на странице: 1из 121

Introduction to Microeletronics

Diomadson Belfort, Ph. D.


Explore the history of electronics.

Quantify the impact of integrated circuit
Describe classification of electronic signals.
Review circuit notation and theory.
Introduce tolerance impacts and analysis.
Introduce CMOS Technology and Logics Circuits
The Start of the Modern Electronics Era

Bardeen, Shockley, and Brattain at The first germanium bipolar

Bell Labs - Brattain and Bardeen transistor. Roughly 50 years later,
invented the bipolar transistor in electronics account for 10% (4
1947. trillion dollars) of the world GDP.
Electronics Milestones
1874 Braun invents the solid-state rectifier. 1958 Integrated circuits developed by Kilby
1906 DeForest invents triode vacuum tube. and Noyce
1907-1927 1961 First commercial IC from Fairchild
First radio circuits developed from
diodes and triodes. 1963 IEEE formed from merger of IRE and
1925 Lilienfeld field-effect device patent
filed. 1968 First commercial IC opamp
1947 Bardeen and Brattain at Bell 1970 One transistor DRAM cell invented by
Laboratories invent bipolar transistors. Dennard at IBM.
1952 Commercial bipolar transistor 1971 4004 Intel microprocessor introduced.
production at Texas Instruments. 1978 First commercial 1-kilobit memory.
1956 Bardeen, Brattain, and Shockley 1974 8080 microprocessor introduced.
receive Nobel prize. 1984 Megabit memory chip introduced.
2000 Alferov, Kilby, and Kromer share
Nobel prize
2009 Boyle and Smith share Nobel prize
Evolution of Electronic Devices

Vacuum Discrete
Tubes Transistors


Integrated Surface-Mount
Circuits Circuits
Microelectronics Proliferation

The integrated circuit was invented in 1958.

World transistor production has more than doubled every
year for the past twenty years.
Every year, more transistors are produced than in all
previous years combined.
Approximately 1018 transistors were produced in a recent
Roughly 50 transistors for every ant in the world.

*Source: Gordon Moores Plenary address at the 2003 International

Solid-State Circuits Conference.
Device Feature Size

Feature size reductions

enabled by process
Smaller features lead to
more transistors per unit
area and therefore higher
Rapid Increase in Density of Microelectronics

Memory chip density Microprocessor complexity

versus time. versus time.
Signal Types

Analog electrical signals

take on continuous values -
typically current or
Digital signals appear at
discrete levels. Usually we
use binary signals which
utilize only two levels.
One level is referred to as
logical 1 and logical 0 is
assigned to the other level.
Analog and Digital Signals

Analog electrical signals After digitization, the

are continuous in time - continuous analog signal
most often voltage or becomes a set of discrete
current. (Charge can also values, typically separated
be utilized as a signal
conveyor.) by fixed time intervals.
Notational Conventions
Total signal = DC bias + time varying signal

vT = VDC + vsig
iT = I DC -+Risig
Resistance and conductance and G with same
subscripts will denote reciprocal quantities. The most
convenient form will be used within expressions.

1 1
Gx = and g =
Rx r
Problem-Solving Approach
Make a clear problem statement.
List known information and given data.
Define the unknowns required to solve the problem.
List assumptions.
Develop an approach to the solution.
Perform the analysis based on the approach.
Check the results and the assumptions.
Has the problem been solved? Have all the unknowns been found?
Is the math correct? Have the assumptions been satisfied?
Evaluate the solution.
Do the results satisfy reasonableness constraints?
Are the values realizable?
Use computer-aided analysis to verify hand analysis
What are Reasonable Numbers?
If the power suppy is 10 V, a calculated DC bias value of 15 V (not within
the range of the power supply voltages) is unreasonable.
Generally, our bias current levels will be between 1 microamp and a few
hundred milliamps.
A calculated bias current of 3.2 amps is probably unreasonable and should
be reexamined.
Peak-to-peak ac voltages should be within the power supply voltage range.
A calculated component value that is unrealistic should be rechecked. For
example, a resistance equal to 0.013 ohms or 1012 ohms
Given the inherent variations in most electronic components, three
significant digits are adequate for representation of results. Three
significant digits are used throughout the text.
Circuit Theory Review: Voltage Division

v1 = ii R1 and v 2 = ii R2
Applying KVL to the loop,

v i = v1 + v 2 = ii (R1 + R2 )
and ii =
R1 + R2
Combining these yields the basic voltage division formula:

R1 R2
v1 = v i v2 = vi
R1 + R2 R1 + R2
Circuit Theory Review: Voltage Division (cont.)

Using the derived equations with the

indicated values,

8 k
v1 = 10 V = 8.00 V
8 k + 2 k
2 k
v 2 = 10 V = 2.00 V
8 k + 2 k

Design Note: Voltage division only applies when both resistors are
carrying the same current.
Circuit Theory Review: Current Division

vi vi
ii = i1 + i2 where i1 = andi2 =
R1 R2
Combining and solving for vi,

1 R1 R2
v i = ii = ii = ii (R1 || R2 )
1 1 R1 + R2
R1 R2
Combining these yields the basic current division formula:

R2 R1
i1 = ii i2 = ii
R1 + R2 R1 + R2
Circuit Theory Review: Current
Division (cont.)

Using the derived equations with the

indicated values,

3 k
i1 = 5 ma = 3.00 mA
2 k + 3 k
2 k
i 2 = 5 ma = 2.00 mA
2 k + 3 k

Design Note: Current division only applies when the same voltage
appears across both resistors.
Circuit Theory Review: Thvenin and Norton
Equivalent Circuits


Circuit Theory Review: Find the Thvenin
Equivalent Voltage

Problem: Find the Thvenin

equivalent voltage at the output.
Known Information and Given
Data: Circuit topology and values
in figure.
Unknowns: Thvenin equivalent
voltage vth.
Approach: Voltage source vth is
defined as the output voltage with
no load (open-circuit voltage).
Assumptions: None.
Analysis: Next slide
Circuit Theory Review: Find the Thvenin
Equivalent Voltage

Applying KCL at the output node,

vo - vi vo
i1 = + = G1 (v o - vi )+ GS v o
Current i1 can be written i1 = G1 v o - vi )
Combining the previous equations

G1 ( +1)v i = G1 ( +1) + GS v o

G1 ( +1) R1RS ( +1)RS

vo = vi = vi
G1 ( +1) + GS R1RS ( +1)RS + R1
Circuit Theory Review: Find the Thvenin
Equivalent Voltage (cont.)

Using the given component values:

( + 1)RS (50 + 1)1 k

vo = vi = v i = 0.718v i
( +1)RS + R1 (50 + 1)1 k +1 k

v th = 0.718v i
Circuit Theory Review: Find the Thvenin
Equivalent Resistance

Problem: Find the Thvenin

equivalent resistance.
Known Information and
Given Data: Circuit topology
and values in figure.
Unknowns: Thvenin
equivalent Resistance Rth.
Test voltage vx has been added to
Approach: Find Rth as the
the previous circuit. Applying vx
output equivalent resistance
and solving for ix allows us to find
with independent sources set
to zero. the
Thvenin resistance as vx/ix.
Assumptions: None.

Analysis: Next slide

Circuit Theory Review: Find the Thvenin
Equivalent Resistance (cont.)

Applying KCL,

i x = -i1 - i1 + G S v x
= G1v x + G1v x + G S v x
= G1 ( + 1) + G S v x
vx 1 R1
Rth = = = RS
i x G1 ( + 1) + G S +1

R1 20 k
Rth = RS = 1 k = 1 k 392 = 282
+1 50 + 1
Circuit Theory Review: Find the Norton Equivalent

Problem: Find the Norton

equivalent circuit.
Known Information and
Given Data: Circuit topology
and values in figure.
Unknowns: Norton
equivalent short circuit current A short circuit has been
in. applied across the output.
Approach: Evaluate current The Norton current is the
through output short circuit. current flowing through the
Assumptions: None. short circuit at the output.
Analysis: Next slide
Circuit Theory Review: Find the Norton Equivalent
Circuit (cont.)

Applying KCL,

in = i1 + i1
= G1v i + G1vi
= G1 ( +1)v i
v i ( +1) Short circuit at the output
= causes zero current to flow
R1 through RS.
Rth is equal to Rth found
50 +1 vi
in = vi = = (2.55earlier.
mS)v i
20 k 392
Final Thvenin and Norton Circuits

Check of Results: Note that vth = inRth and this can be used to
check the calculations: inRth=(2.55 mS)vi(282 ) = 0.719vi,
accurate within round-off error.

While the two circuits are identical in terms of voltages and

currents at the output terminals, there is one difference between
the two circuits. With no load connected, the Norton circuit still
dissipates power!
Amplifier Basics
Analog signals are typically manipulated with linear
Although signals may be comprised of several different
components, linearity permits us to use the superposition
Superposition allows us to calculate the effect of each of
the different components of a signal individually and then
add the individual contributions to create the total
resulting signal.
Amplifier Linearity

Given an input sinusoid: v i = Vi sin( i t + )

For a linear amplifier, the output is
at the same frequency, but v o = Vo sin( i t + + )
different amplitude and phase.

In phasor notation:
v i = Vi v o = Vo( + )

v o Vo( + ) Vo
Amplifier gain is: A= = =
vi Vi Vi
Amplifier Input/Output Response

vi = sin 2000t V

Av = -5

Note: negative
gain is
equivalent to 180
degrees of
phase shift.
Ideal Operational Amplifier (Op Amp)

Ideal op amps are assumed to have

infinite voltage gain, and
infinite input resistance.

These conditions lead to two assumptions useful in analyzing ideal op-amp


1. The voltage difference across the input terminals is zero.

2. The input currents are zero.
Ideal Op Amp Example
Find the voltage gain of an op amp with resistive feedback
v i - ii R1 - i2 R2 - v o = 0
Writing a loop equation:
vi - v-
ii = i2 =
From assumption 2, we know that i- = R1
0. vi
ii =
Assumption 1 requires v- = v+ = 0.
vo R2
Av = =-
Combining these equations yields: vi R1

Assumption 1 requiring v- = v+ = 0
creates what is known as a virtual
ground at the inverting input of the
Ideal Op Amp Example
(Alternative Approach)

vi v- - v o -vo
From Assumption 2, i2 = ii : ii = and i2 = =
R1 R2 R2
vi -vo
i2 = ii gives =
R1 R2
Yielding: vo R2
Av = = -
vi R1
Design Note: The virtual ground is
not an actual ground. Do not short
the inverting input to ground to
simplify analysis.
Amplifier Frequency Response

Amplifiers can be designed to selectively amplify specific ranges of

frequencies. Such an amplifier is known as a filter. Several filter types are
shown below:

Low Pass High Pass Band Pass Band Reject All Pass
Circuit Element Variations
All electronic components have manufacturing tolerances.
Resistors can be purchased with 10%, 5%, and
1% tolerance. (IC resistors are often 10%.)
Capacitors can have asymmetrical tolerances such as +20%/-50%.
Power supply voltages typically vary from 1% to 10%.
Device parameters will also vary with temperature and age.
Circuits must be designed to accommodate these variations.
We will use worst-case and Monte Carlo (statistical) analysis to
examine the effects of component parameter variations.
Tolerance Modeling
For symmetrical parameter variations
Pnom(1 - ) P Pnom(1 + )
For example, a 10 k resistor with 5% percent tolerance
could take on the following range of values:
10k(1 - 0.05) R 10k(1 + 0.05)
9500 R 10500
Circuit Analysis with Tolerances

Worst-case analysis
Parameters are manipulated to produce the worst-case min and max
values of desired quantities.
This can lead to over design since the worst-case combination of
parameters is rare.
It may be less expensive to discard a rare failure than to design for
100% yield.
Monte-Carlo analysis
Parameters are randomly varied to generate a set of statistics for
desired outputs.
The design can be optimized so that failures due to parameter
variation are less frequent than failures due to other mechanisms.
In this way, the design difficulty is better managed than a worst-case
Worst Case Analysis Example
Problem: Find the nominal and
worst-case values for output
voltage and source current.
Known Information and Given
Data: Circuit topology and
values in figure.

VOnom , VOmin , VOmax , I Inom , I Imin , I Imax

Nominal voltage solution:
Approach: Find nominal values
and then select R1, R2, and VI R1nom
values to generate extreme cases VOnom = VInom
of the unknowns. R1nom + R2nom
Assumptions: None.
Analysis: Next slides
= 15V = 5V
18k + 36k
Worst-Case Analysis Example (cont.)

Nominal Source current:

VI nom 15V
I I = nom nom
= = 278 mA
R1 + R2 18k + 36k

Rewrite VO to help us determine how to find the worst-case values.

R1 VI VO is maximized for max VI, R1 and min

VO = VI = R 2.
R1 + R2 R2
VO is minimized for min VI, R1, and max
15V (1.1) R 2. min 15V (0.95)
VOmax = = 5.87V VO = = 4.20V
36K(0.95) 36K(1.05)
1+ 1+
18K(1.05) 18K(0.95)
Worst-Case Analysis Example (cont.)

Worst-case source currents:

VImax 15V (1.1)

IImax = min min = = 322 mA
R1 + R2 18k(0.95) + 36k(0.95)

VImin 15V (0.9)

I min
I = max max = = 238 mA
R1 + R2 18k(1.05) + 36k(1.05)

Check of Results: The worst-case values range from 14-17

percent above and below the nominal values. The sum of the
three element tolerances is 20 percent, so our calculated values
appear to be reasonable.
Monte Carlo Analysis

Parameters are varied randomly and output statistics are gathered.

We use programs like MATLAB, Mathcad, SPICE, or a spreadsheet to
complete a statistically significant set of calculations.
For example, with Excel, a resistor with a nominal value Rnom and
tolerance can be expressed as:
R = Rnom (1+ 2 ( RAND() - 0.5))

The RAND() function

returns random numbers
uniformly distributed
between 0 and 1.
Monte Carlo Analysis Results

VO (V)
Average 4.96
Nominal 5.00
Standard Deviation 0.30
Maximum 5.70
W/C Maximum 5.87
Minimum 4.37
W/C Minimum 4.20

Histogram of output voltage from 1000 case Monte Carlo

Monte Carlo Analysis Example

Problem: Perform a Monte Carlo

analysis and find the mean, standard
deviation, min, and max for VO, II,
and power delivered from the source.
Known Information and Given
Data: Circuit topology and values in
Unknowns: The mean, standard
deviation, min, and max for VO, II,
and PI. Monte Carlo parameter
Approach: Use a spreadsheet to definitions:
evaluate the circuit equations with VI = 15(1+ 0.2(RAND() - 0.5))
random parameters.
Assumptions: None.
R1 = 18,000(1+ 0.1(RAND() - 0.5))
Analysis: Next slides R2 = 36,000(1+ 0.1(RAND() - 0.5))
Monte Carlo Analysis Example (cont.)
Monte Carlo parameter
definitions: VI = 15(1+ 0.2( RAND() - 0.5))
R1 = 18,000(1+ 0.1( RAND() - 0.5))
R2 = 36,000(1+ 0.1( RAND() - 0.5))
Circuit equations based on Monte Carlo parameters:
VO = VI II =
R1 + R2 R1 + R2
Avg Nom. Stdev Max WC-max Min WC-Min
Vo (V) 4.96 5.00 0.30 5.70 5.87 4.37 4.20
II (mA) 0.276 0.278 0.0173 0.310 0.322 0.242 0.238
P (mW) 4.12 4.17 0.490 5.04 -- 3.29 --
Temperature Coefficients

Most circuit parameters are temperature sensitive.

P = Pnom(1+1T+ 2T2) where T = T-Tnom
Pnom is defined at Tnom
Most versions of SPICE allow for the specification of
TNOM, T, TC1(1), TC2(2).
SPICE temperature model for resistor:
R(T) = R(TNOM)*[1+TC1*(T-TNOM)+TC2*(T-TNOM)2]
Many other components have similar models.
Numeric Precision

Most circuit parameters vary from less than 1 % to

greater than 50%.
As a consequence, more than three significant digits is
Results in the text will be represented with three
significant digits: 2.03 mA, 5.72 V, 0.0436 A, and so
However, extra guard digits are normally retained during
Integrated Circuit Fabrication Overview

Top view of an integrated pn diode.

Integrated Circuit Fabrication (cont.)

(a) First mask exposure, (b) post-exposure and development of photoresist,

(c) after SiO2 etch, and (d) after implantation/diffusion of acceptor dopant.
Integrated Circuit Fabrication (cont.)

(e) Exposure of contact opening mask, (f) after resist development and etching of
contact openings, (g) exposure of metal mask, and (h) After etching of aluminum and
resist removal.
MOS Capacitor Structure

First electrode - Gate:

Consists of low-resistivity
material such as metal or
doped polycrystalline silicon
Second electrode - Substrate
or Body: n-- or p-type
Dielectric - Silicon dioxide:
stable high-quality electrical
insulator between gate and
Substrate Conditions for Different Biases

Accumulation Depletion
Inversion VG > VTN
Low-frequency C-V Characteristics for an MOS
Capacitor on p-type Substrate

MOS capacitance is a non-

linear function of voltage.
Total capacitance in any
region is dictated by the
separation between
capacitor plates.
Total capacitance modeled
as series combination of
fixed oxide capacitance and
depletion-layer capacitance.
NMOS Transistor:

4 device terminals:
Gate(G), Drain(D),
Source(S) and Body(B).
Source and drain regions
form pn junctions with
vSB, vDS and vGS always
positive during normal
vSB always < vDS and vGS
to reverse bias pn
NMOS Transistor:
Qualitative I-V Behavior

VGS << VTN : Only small leakage

current flows.
VGS < VTN: Depletion region formed
under gate merges with source and
drain depletion regions. No current
flows between source and drain.
VGS > VTN: Channel formed between
source and drain. If vDS > 0, finite iD
flows from drain to source.
iB = 0 and iG = 0.
NMOS Transistor:
Triode Region Characteristics

iD = Kn vGS - VTN - v DS
for vGS - VTN vDS 0

Kn = Kn' W L
Kn' = mnCox" (A/V2 )
Cox" = ox Tox
ox = oxide permittivity (F/cm)
Tox = oxide thickness (cm)
NMOS Transistor:
Triode Region Characteristics (cont.)

Output characteristics
appear to be linear.
FET behaves like a
gate-source voltage-
controlled resistor
between source and
drain with

i 1 1
Ron = D = =
v DS W W
v DS 0
Q -Pt Kn' (VGS - VTN - VDS ) Kn' (VGS - VTN )
Voltage-Controlled Resistor

Example 1: Voltage-Controlled Attenuator

vO Ron 1
= =
vS Ron + R 1+ Kn R(VGG - VTN )

If Kn = 500mA/V2, VTN = 1V, R = 2k

and VGG = 1.5V, then,
vO 1
= = 0.667
vS mA
1+ 500
V 2 (2000)(1.5 -1)V
To maintain triode region operation,
vDS vGS - VTN or vO VGG - VTN
0.667vS (1.5 -1)V or vS 0.750 V
MOSFET as a Voltage-Controlled Resistor (cont.)

Example 2: Voltage-Controlled High-Pass Filter

O( )
V s s
Voltage Transfer T (s) = =
VS (s) s + o
where, cut-off 1 Kn (VGS - VTN )
o = =
frequency RonC C

If Kn = 500mA/V2, VTN = 1V, C =

0.02mF and VGG = 1.5V, then,
1.5 -1)V
fo = = 1.99 kHz
2 (0.02 mF )
To maintain triode region
operation, v V - V = 0.5 V
NMOS Transistor:
Saturation Region

If vDS increases above triode region limit, the

channel region disappears and is said to be
Current saturates at constant value, independent
of vDS.
Saturation region operation mostly used for
analog amplification.
NMOS Transistor:
Saturation Region (cont.)

Kn' W
(vGS - VTN )
iD = for vDS vGS - VTN
2 L

vDSAT = vGS - VTN is termed the saturation or pinch - off voltage

Transconductance of an MOS Device
Transconductance relates the change in drain current to a
change in gate-source voltage

gm =
vGS Q - pt
Taking the derivative of the expression for the drain
current in saturation region,

W 2I D
gm = K '
(VGS - VTN ) =
Channel-Length Modulation

As vDS increases above vDSAT, the length

of the depleted channel beyond the
pinch-off point, L, increases and the
actual L decreases.
iD increases slightly with vDS instead of
being constant.

Kn' W
( TN ) (1+ v DS )
iD = vGS - V
2 L

= channel length modulation

Depletion-Mode MOSFETS

NMOS transistors with VTN < 0

Ion implantation process is used to form a built-in n-
type channel in the device to connect source and drain
by a resistive channel
Non-zero drain current for vGS = 0
Negative vGS required to turn device off.
Transfer Characteristics of MOSFETS
Enhancement- & Depletion-Mode Devices

Transfer Characteristic: Plots drain current versus

gate-source voltage for a fixed drain-source voltage
Body Effect or Substrate Sensitivity

Non-zero vSB changes threshold voltage,

causing substrate sensitivity modeled by

VTN = VTO + g (v SB + 2 F - 2 F )
VTO = threshold voltage for vSB = 0
g = body - effect parameter ( V)
2 F = surface potential (V)
PMOS Transistors:
Enhancement-Mode Structure
p-type source and drain regions
in an n-type substrate.
vGS < 0 required to create p-type
inversion layer in channel
For current flow, vGS < VTP
To maintain reverse bias on
source-substrate and drain-
substrate junctions, vSB < 0 and
vDB < 0
Positive bulk-source potential
causes VTP to become more
PMOS Transistors:
Output Characteristics

For vGS > VTP, transistor is

For more negative vGS, drain
current increases in
PMOS device is in the triode
region for small values of VDS
and in saturation for larger
Remember VTP < 0 for an
enhancement mode transistor
MOSFET Circuit Symbols
(g) and(i) are the
most commonly
used symbols in
VLSI logic design.
MOS devices are
In NMOS, n+
region at higher
voltage is the drain.
In PMOS p+ region
at lower voltage is
the drain
Internal Capacitances in Electronic Devices
Limit high-frequency performance of the electronic device
they are associated with.
Limit switching speed of circuits in logic applications
Limit frequency at which useful amplification can be
obtained in amplifiers.
MOSFET capacitances depend on region of operation and
are non-linear functions of voltages at device terminals.
NMOS Transistor Capacitances:
Triode Region

Cox" = Gate-channel
capacitance per unit
CGC = Total gate channel
CGS = Gate-source
CGC WL CGD = Gate-drain
CGS = + CGSOW = Cox" + CGSOW
2 2 capacitance.
C WL CGSO and CGDO = overlap
2 2 capacitances (F/m).
NMOS Transistor Capacitances:
Triode Region (cont.)

CSB = Source-bulk capacitance.

CDB = Drain-bulk capacitance.
CDB = CJ AD + CJSW PD AS and AD = Junction bottom
area capacitance of the
source and drain regions.
PS and PD = Perimeter of the
source and drain junction
NMOS Transistor Capacitances: Saturation Region

Drain no longer connected to channel

2 2
3 3
NMOS Transistor Capacitances:
Cutoff Region

Conducting channel region

completely gone.

CGB = gate-bulk capacitance

CGBO = gate-bulk
capacitance per unit
CGD = CGDOW width.
SPICE Model for NMOS Transistor

Typical default values used by

KP = 50 or 20 mA/V2
VTO = 1 V
mn or mp = 500 or 200 cm2/V-s
2FF = 0.6 V
Tox= 100 nm
MOS Transistor Scaling
Scale Factor
Drain current: ox W W
Kn* = mn = mn ox = K n
Tox L Tox L
ox W vGS VTN vDS vDS iD
i = mn
D - - =
Tox L 2 2
Gate Capacitance:
ox W CGC
( )
* " * *
CGC = C ox W L = =
Tox L
V *
* *
GC =
where is the circuit delay in a logic circuit.
MOS Transistor Scaling
Scale Factor (cont.)

Circuit and Power Densities:

* *
P * = VDD iD = =
P* P* P 2 P
= = = extremely important result!
A* W * L* (W )(L ) A

Power-Delay Product:
PDP * = P * * = =
2 3
Cutoff Frequency:
gm mn
T = = 2 (VGS - VTN )

fT improves with square of channel length reduction

MOS Transistor Scaling (cont.)

High Field Limitations:

High electric fields arise if technology is scaled down with
supply voltage constant.
Causes reduction in mobility of MOS transistor,
breakdown of linear relationship between mobility and
electric field and carrier velocity saturation.
Ultimately results in reduced long-term reliability and
breakdown of gate oxide or pn junction.
Drain current in saturation region is linearized to

Cox" W
iD =
(vGS - VTN )v SAT where vSAT is carrier saturation
MOS Transistor Scaling (cont.)

Sub-threshold Conduction:
iD decreases exponentially for
vGS < VTN.
Reciprocal of the slope in
mV/decade gives the turn-off
rate for the MOSFET.
VTN should be reduced if
dimensions are scaled down.
However, curve in sub-
threshold region shifts
horizontally instead of scaling
with VTN
Process-defining Factors

Minimum Feature Size F : Width of smallest line or space that can

be reliably transferred to the wafer surface using a given generation
of lithographic manufacturing tools
Alignment Tolerance T: Maximum misalignment that can occur
between two mask levels during fabrication
Mask Sequence
Polysilicon-Gate Transistor

Mask 1: Defines active area or

thin oxide region of transistor
Mask 2: Defines polysilicon gate
of transistor, aligns to mask 1
Mask 3: Delineates the contact
window, aligns to mask 2.
Mask 4: Delineates the metal
pattern, aligns to mask 3.
Channel region of transistor
formed by intersection of first
two mask layers. Source and
Drain regions formed wherever
mask 1 is not covered by mask 2
Basic Ground Rules for Layout

T = F/2 =
could be 1, 0.5, 0.25
mm, etc.

W/L = 10/2 = 5/1

Transistor Area = 120 2
MOSFET Biasing
Bias sets the dc operating point around which the device operates.
The signal is actually comprised of relatively small changes in the
voltages and/or currents.
Remember (Total = dc + signal): vGS = VGS + vgs and iD = ID + id
Bias Analysis Approach
Assume an operation region (generally the saturation
Use circuit analysis to find VGS
Use VGS to calculate ID, and ID to find VDS
Check validity of operation region assumptions
Change assumptions and analyze again if required.

NOTE: An enhancement-mode device with VDS = VGS is

always in saturation
CMOS Technology
Inverter Circuit

(a) Circuit schematic for a CMOS inverter

(b) Simplified operation model with a high input applied
(c) Simplified operation model with a low input applied
CMOS Technology
Inverter Operation

When vI is pulled high (to VDD), the PMOS

transistor is turned off, while the NMOS device is
turned on pulling the output down to VSS

When vI is pulled low (to VSS), the NMOS

transistor is turned off, while the PMOS device is
turned on pulling the output up to VDD
CMOS Technology
Inverter Fabrication

The CMOS inverter consists of a PMOS device stacked on top

on an NMOS device, but they need to be fabricated on the
same wafer

To accomplish this, the technique of n-well implantation was

developed as shown in this cross-section of a CMOS inverter
CMOS Technology
Inverter Layout

Two methods of
laying out a CMOS
inverter are shown
transistors lie within
the n-well, whereas
the NMOS transistors
lie in the p-substrate
Polysilicon is used to
form common gate
connections, and
metal is used to tie
the two drains
CMOS Inverter
Static Characteristics: vI = VL

For vI = VL VTN, MN is off, and MP is on. Therefore VH

= VDD, ID = 0, and there is no static power dissipation.
CMOS Inverter
Static Characteristics: vI = VH

For vI = VH = VDD, VL = 0 V, and ID = 0 A which means

that there is no static power dissipation
CMOS Inverter
Voltage Transfer Characteristics

The VTC shown is for a

CMOS inverter that is
symmetrical (Kp = Kn).
CMOS Inverter
Voltage Transfer Characteristics (cont.)

The simulation
results show the
varying VTC of the
inverter as VDD is
The theoretical
minimum voltage
supply for CMOS
technology is
VDD = 2VT ln(2) V
or only 18 mV!
CMOS Inverter
Voltage Transfer Characteristics (cont.)

Simulation results
show the varying VTC
of the inverter as KR =
KN/KP is changed

For KR > 1 the NMOS

current drive is
greater, and the logic
transition occurs for vI
< VDD/2

For KR < 1 the PMOS

current drive is
greater, the logic
transition occurs for vI
> VDD/2
CMOS Inverter
Noise Margins

Noise margins
are defined by
the points shown
in the given figure
CMOS Inverter
Noise Margins (cont.)

2K R (VDD - VTN + VTP ) (VDD - K RVTN + VTP )
VIH = -
(K R -1) 1+ 3K R K R -1

( K R + 1)VIH - VDD - K RVTN - VTP
2K R
2 K R (VDD - VTN + VTP ) (VDD - K RVTN + VTP )
VIL = -
(K R -1) K R + 3 K R -1

( K R + 1)VIL + VDD - K RVTN - VTP
CMOS Inverter
Propagation Delay Estimate

The two modes of capacitive charging/discharging

that contribute to propagation delay
CMOS Inverter
Propagation Delay Estimate (cont.)
V - V 2V
PHL = RonN Cln4 H TN
-1 +

V H + VL V H - VTN
RonN =
K n (V H - VTN )
p = = PHL = 1.2RonN C
If it is assumed the inverter in symmetrical with
(W/L)P = 2.5(W/L)N, then PLH = PHL
CMOS Inverter
Rise and Fall Times

The rise and fall times are given by the following

approximate expressions:

t f = 3 PHL
t r = 3 PLH
CMOS Inverter
Design Example
Design a reference inverter to achieve a delay of 250
ps with a 0.2 pF load given the following information:

VDD = 3.3 V
C = 0.2 pF
p = 250 ps
VTN = -VTP = 0.75 V
CMOS Inverter
Design Example (cont.)
Assuming the inverter is symmetrical and using the
values given in Table 7.1:

' mA
K = 100
' mA
K p = 40 2
p = PHL = PLH = 250 ps
CMOS Inverter
Design Example (cont.)

Solving for RonN:

Then solve for the transistor ratios:

W 1 3.77
= ' =
L n K n RonN (VDD - VTN ) 1
W K n' W W 9.43
= ' = 2.5 =
L p K p L n L n 1
CMOS Inverter
Performance Scaling
State-of-the-art short gate length
technologies are hard to analyze
Scaling can be used to properly set W/L for a
given load capacitance relative to reference
gate simulation with a reference load.

( W / L ) CL ' W W Pr ef CL '

P = Pr ef or =
(W / L ) ' CLref L L P CLref

Scaling allows us to calculate a new geometry (W/L)' in

terms of a target load and delay.
CMOS Inverter
Performance Scaling

Consider a reference inverter with a delay of

3.16 ns.
What is the delay if an inverter has a W/L 4x
larger than the transistors of the reference
inverter and twice the load capacitance.
(2 /1) 2 pF'
P = 3.16 ns = 1.58 ns
(8 /1)' 1pF
Scaling allows us to calculate a new geometry (W/L)' or
delay relative to a reference design.
CMOS Logic
Delay of Cascaded Inverters
An ideal step was used to derive the previous delay
equations, but this is not possible to implement
By using putting the following circuit in SPICE, it is
possible to produce more accurate equations
CMOS Logic
Delay of Cascaded Inverters (cont.)

The simulated output of the previous circuit appears below,

and it can be seen that the delay for the nonideal step
input is approximately twice than the ideal case:

PHL @ 2.4RonN C
PLH @ 2.4RonP C
t f = 2 PHL
t r = 2 PLH
CMOS Logic
Static Power Dissipation

CMOS logic is considered to have no static power


This is not completely accurate since MOS transistors

have leakage currents associated with the reverse-
biased drain-to-substrate connections as well as sub-
threshold leakage current between the drain and
CMOS Logic
Dynamic Power Dissipation

There are two

components that
add to dynamic
power dissipation:
1) Capacitive load
charging at a
frequency f given
by: PD = CV2DDf
2) The current that
occurs during
switching which
can be seen in the
CMOS Logic
Power-Delay Product

The power-delay product is given as:

The figure shows a symmetrical

inverter switching waveform
CMOS Logic
NOR Gate

Y = A+B
Basic CMOS logic gate CMOS NOR gate Reference
structure implementation Inverter
Transistor Sizing
When sizing the transistors, we attempt to keep the
delay times the same as the reference inverter

To accomplish this, the on-resistance in the PMOS

and NMOS branches of the NOR gate must be the
same as the reference inverter

For a two-input NOR gate, the (W/L)p must be made

twice as large
Body Effect
Since the bottom PMOS body contact is not
connected to its source, its threshold voltage
changes as VSB changes during switching

Once vO = VH is reached, the bottom PMOS is not

affected by body effect, thus the total on-resistance of the
PMOS branch is the same

However, the rise time is slowed down slightly due to

|VTP| being a function of time
Two-Input Layout
Three-Input Circuit

It is possible to extend this same design technique to

create multiple input NOR gates

Y = A+ B+C
CMOS Logic
NMOS and PMOS Transistor Symbols

Shorthand Notation
CMOS Logic
NAND Gates

Y = AB

CMOS NAND gate Reference Inverter

Transistor Sizing

The same rules apply for sizing the NAND gate

devices as for the NOR gate, except now the
NMOS transistors are in series

For the two-input NAND, (W/L)N will be twice the

size of that of the reference inverter
Five Input NAND

Complex CMOS Logic Gate
Design Example

Design a CMOS logic gate for (W/L)p,ref = 5/1 and for (W/L)n,ref = 2/1 that
yields the function:
Y = A + BC + BD
By inspection (knowing Y), the NMOS branch of the gate can drawn as
the following with the corresponding graph, while considering the
longest path for sizing purposes:
Complex CMOS Logic Gate
Design Example (cont.)
By placing nodes in the interior of each arc, plus two more outside
the graph for VDD (3) and the complementary output (2), the PMOS
branch can be realized as shown on the left figure
Connect all of the nodes in the manner shown in the right figure, and
the NMOS arcs that the PMOS arcs intersect have the same inputs
Complex CMOS Logic Gate
Design Example (cont.)

From the PMOS

graph, the PMOS
network can now
be drawn for the
final CMOS logic
gate while once
considering the
longest PMOS
path for sizing
Two equivalent forms of the final circuit
Minimum Size Gate
Design and Performance

With CMOS technology, there is an area/delay

tradeoff that needs to be considered

If minimum feature sized are used for both

devices, then the PLH will be increased compared
to the symmetrical reference inverter
Minimum Size Complex Gate
Design and Layout

The following shows the layout of a complex minimum

size logic gate