CMOS Logic Families

Many families of logic exist beyond Static CMOS Comparison of logic families for a 2-input multiplexer Briefly overview
pseudo-nMOS differential (CVSL) dynamic/domino complementary pass-gate
ECE 410, Prof. A. Mason Advanced Digital.1

nMOS Inverter
Logic Inverter nMOS Inverter
x y x y=x

assume a resistive load to VDD nMOS switches pull output low based on inputs
nMOS Inverter (a) nMOS is off, (b) nMOS is on

0 1 1 0

Active loads
use pMOS transistor in place of resistor resistance varies with Gate connection
Ground always on Drain=Output turns off when Vout > VDD-Vtp Vbias
VSG = VSD so always in saturation
Vbias

can turn Vbias for needed switching characteristics

ECE 410, Prof. A. Mason Advanced Digital.2

Pseudo-nMOS
full nMOS logic array replace pMOS array with single pull up transistor Ratioed Logic generic pseudo-nMOS logic gate Advantages
requires proper tx size ratios less load capacitance on input signals fewer transistors
faster switching higher circuit density

pseudo-nMOS inverter

Disadvantage
VOL > 0

pull up is always on

significant static power dissipation

ECE 410, Prof. A. Mason

pseudo-nMOS Advanced Digital.3 NAND and NOR

Output High Voltage, VOH (Maximum output)

Pseudo nMOS DC Operation

occurs when input is low (Vin = 0V), nMOS is OFF triode operation pMOS has very small VSD pMOS pulls Vout to VDD VOH = VDD

Output Low Voltage, VOL (Minimum output)

occurs when input is high (Vin = VDD) both nMOS and pMOS are ON
nMOS is on stronger; pulls Vout low

VOH = VDD

as Vout goes low, nMOS enters triode

continues to sink current from pMOS load

Logic Swing (max output swing)

VL = VOH - VOL < VDD

VOL > 0 V (active load always pulling)

pseudo nMOS inverter VTC VOL > Ground

Advanced Digital.4

ECE 410, Prof. A. Mason

Pseudo nMOS Transient Analysis

Rise and Fall Times
harder to analyze for pseudo nMOS due to always on active load
slow rise time faster fall time
but does not fall to 0 volts

ECE 410, Prof. A. Mason

Advanced Digital.5

Differential Logic
Cascode Voltage Switch Logic
aka, Differential Logic
(CVSL)

Performance advantage of ratioed circuits without the extra power Requires complementary inputs
produces complementary outputs

generic differential logic gate

Operation
two nMOS arrays one for f, one for f differential AND/NAND gate cross-coupled load pMOS one path is always active since either f or f is always true other path is turned off (logic arrays turns off one load)
no static power
ECE 410, Prof. A. Mason Advanced Digital.6

Differential Logic
Advantages of CVSL
low load capacitance on inputs no static power consumption automatic complementary functions

Disadvantages
requires complementary inputs more transistors
for single function

Very useful in some circuit blocks where complementary signals are generally needed
interesting implementation in adders

differential 4-input XOR/XNOR

ECE 410, Prof. A. Mason

Advanced Digital.7

Dynamic Logic
Advantages of ratioed logic without power consumption of pseudo-nMOS or excess tx of differential Dynamic operation: output not always valid Precharge stage
clock-gated pull-up precharges output high logic array disabled

Evaluation stage
precharge pull-up disabled logic array enabled & if true, discharges output
generic dynamic logic gate

ECE 410, Prof. A. Mason

Advanced Digital.8

Dynamic Logic
Example: Footed dynamic NAND3 Footed vs. Unfooted
foot tx ensures nMOS array disabled during precharge

unfooted

footed

ECE 410, Prof. A. Mason

Advanced Digital.9

Charge Redistribution in Dynamic Logic

Major potential problem
during evaluation, precharge charge is distributed over parasitic capacitances within the nMOS array
causes output to decrease (same charge over larger C less V)

if the function is not true, output should be HIGH but could be much less than VDD

One possible solution

keeper transistor
injects charge during evaluation if output should be HIGH keeps output at VDD

keeper controlled by output_bar

on when output is high
charge distribution over nMOS parasitics during evaluation ECE 410, Prof. A. Mason Advanced Digital.10

Domino Logic
Dynamic logic can only drive an output LOW
output HIGH is precharged only with limited drive

Domino logic adds and inverter buffer at output Cascading domino logic
must alter precharge/eval cycles clock each stage on opposite clock phase

generic domino logic gate

NP dynamic logic ECE 410, Prof. A. Mason

NO RAce (NORA) domino logic Advanced Digital.11

Pass Transistor (PT) Logic

B A 0 B F=AB A B 0 F=AB B

Gate is static a low-impedance path exists to both supply rails under all circumstances N transistors instead of 2N No static power consumption Ratioless Bidirectional (versus undirectional)
ECE 410, Prof. A. Mason Advanced Digital.12

VTC of PT AND Gate

B
2.4/0.6 1.2/0.6 Vout, V
2

B=VDD, A=0VDD
1

A 1.2/0.6 B 0
1.2/0.6

F= AB

0 0 1

A=VDD, B=0VDD A=B=0VDD

2

Vin, V

Pure PT logic is not regenerative - the signal gradually degrades after passing through a number of PTs (can fix with static CMOS inverter insertion)
ECE 410, Prof. A. Mason Advanced Digital.13

nMOS Only PT Driving an Inverter

In = VDD A = VDD VGS
D B S
M1

Vx = VDD-VTn

M2

Vx does not pull up to VDD, but VDD VTn Threshold voltage drop causes static power consumption (M2 may be weakly conducting forming a path from VDD to GND) Notice VTn increases of pass transistor due to body effect (VSB)

ECE 410, Prof. A. Mason

Advanced Digital.14

Voltage Swing of PT Driving an Inverter

3

In = 0 VDD
1.5/0.25

In x = 1.8V

VDD

D 0.5/0.25

x
0.5/0.25

Out

Voltage, V

Body effect large VSB at x - when pulling high (B is tied to GND and S charged up close to VDD) So the voltage drop is even worse Vx = VDD - (VTn0 + ((|2f| + Vx) - |2f|))

0 0

Out
0.5 1 1.5 2

Time, ns

ECE 410, Prof. A. Mason

Advanced Digital.15

TG Full Adder
Cin B

Sum

16 Transistors; full swing transmission gates

ECE 410, Prof. A. Mason

Cout

Advanced Digital.16

Basic CMOS Isolation Structures

LOCOS Local Oxidation of Silicon STI Shallow Trench Isolation LDD Lightly-Doped Drain
Used to reduce the lateral electric field in the channel

SOI Silicon on Insulator BiCMOS -Bipolar and CMOS on same chip

ECE 410, Prof. A. Mason

Advanced Digital.17

LOCOS
Isolation between transistor
Field Oxide (FOX)

FOX formed by
masking active regions thermal oxidation of non-masked areas

n+ n

p+

p+

n+

n+

p+

FOX p-type substrate

Self-aligned gate
S/D formed after poly gate S/D automatically aligned to gate

Cross section view

Layout view
ECE 410, Prof. A. Mason Advanced Digital.18

Problems with LOCOS

Device packing density limited by birds beak effect of FOX isolation layer.
effects the width (W) of the transistor

Can improve density with Shallow Trench Isolation (STI)

birds beak

ECE 410, Prof. A. Mason

Advanced Digital.19

Shallow Trench Isolation (STI)

Form Gate and S/D first Then isolate devices by
etching trench (~0.4um) in substrate between devices filling trench with deposited oxide

Eliminates the area lost to birds beak effect of LOCOS Well doping and channel implants done later in process via high energy ion implantation

ECE 410, Prof. A. Mason

Advanced Digital.20

Lightly-Doped Drain (LDD)

Create lightly doped regions in S/D near the channel Necessary for submicron fabrication
reduces the electric field across the channel reduces the velocity of electrons in the channel
(hot carrier effect)

reduces performance, but allows higher density

Can be used with any isolation process

ECE 410, Prof. A. Mason

Advanced Digital.21

Silicon On Insulator (SOI)

Buried SiO2 layer beneath surface of active singlecrystal Si substrate More expensive, but excellent isolation
no leakage current to substrate no latchup high transconductance good subthreshold performance reduced short channel effects radiation immunity
ECE 410, Prof. A. Mason Advanced Digital.22

BiCMOS

Advantage

Disadvantage

both Bipolar and CMOS transistor Increased process complexity Reduced density (just no way to make small BJTs)
ECE 410, Prof. A. Mason Advanced Digital.23

Scaling Options
Constant Voltage (CV)
voltage remains constant as feature size is reduced causes electric field in channel to increase
decreases performance

but, device will fail if electric field gets too large

Constant Electric Field (CE)

scale down voltage with feature size keeps electric field constant
maintain good performance

Electric Field in channel vs. Channel Length at various Supply Voltages constant voltage

but, limit to how low voltage can go

constant electric field

ECE 410, Prof. A. Mason

Advanced Digital.24

Low Voltage Issues

Reasons modern/future circuits have lower voltage
lower voltage = lower dynamic power: Pdyn = CL VDD2 fCLK lower voltage required for smaller feature size
feature sizes reduced to improve performance/speed smaller features = shorter distances across the channel
stronger electric fields in channel poorer performance or device failure unless voltage is reduced also

Side effects of reducing voltage

reduces current (speed) and increases noise problems as supply voltage decreases, must also reduce threshold voltage
otherwise, circuits will not switch correctly

reduced threshold voltage = increased subthreshold current

increased leakage currents
increased static power consumption
ECE 410, Prof. A. Mason Advanced Digital.25

Short Channels
Effective channel length
must account for depletion region spreading into the channel more important as channel length gets shorter S
L (drawn)
G xd D LD ~xd

Leff = L(drawn) 2 LD X d

For short channels, roughly measured by L < 1m

Source-Substrate and Drain-Substrate junction depletion layer extend noticeably into the channel will reduce the about of bulk charge, QB, in the channel thus reduce the threshold voltage as channel length decreases called the short channel effect need a new way to calculate QB

2 (V (VG Vt )) Xd = s D qN A

Leff

some bulk charge lost to depletion layers

(from Kuo and Lou, p. 43)

ECE 410, Prof. A. Mason

Advanced Digital.26

Short Channel Effects

Short channel lengths introduce effects which must be considered
in selecting process technologies for a given application in design of circuits

Short Channel Effects

mobility degradation threshold voltage degradation velocity saturation hot carrier effects

Other effects made worse by short channels

leakage currents latch-up subthreshold operation

ECE 410, Prof. A. Mason

Advanced Digital.27

Velocity Saturation
Charge Velocity Velocity Saturation
vn = E, is mobility and E is electric field valid for small E, as assumed in previous I-V equations if E > Ec (critical field level), velocity will reach a maximum vsat = saturation velocity, velocity at E > Ec lateral field near drain is very high charge can experience velocity saturation even at VDS voltages before pinchoff occurs here nCOX W 2
ID = 2 L

Short Channel Devices

[2(V

GS

Vt )V 'DSAT V 'DSAT

Velocity Overshoot

where VDSAT is the drain-source voltage which generates the critical electric field, Ec valid when VDSAT < VGS-Vt (when velocity saturation occurs before channel pinchoff)

with very short channels, carriers can travel faster than saturation velocity occurs in deep submicron devices with channel lengths less than 0.1m the velocity saturation equation above becomes inaccurate for very small channel lengths and more detailed models are required.
ECE 410, Prof. A. Mason Advanced Digital.28

Hot Carrier Effects

High E-field in channel will accelerate charge carriers Accelerated carriers can start colliding with the substrate atoms
generates electron-hole pairs during the collision these will be accelerated, collide with substrate atoms and form even more electron-hole pairs: called impact ionization

Impact ionization can lead to

avalanche breakdown within the device large substrate currents degradation of the oxide
high energy electrons collide with gate oxide and become imbedded causes a shift in threshold voltage considered catastrophic effect
leads to unstable performance

ECE 410, Prof. A. Mason

Advanced Digital.29

Hot Carrier Effects II

Supply voltages dropping slower than channel length
as a result electric fields in channel continue to increase.

Hot carrier effect must be considered for submicron devices

may potentially be a limiting factor in how far devices can be scaled down
unless we can reduce electric fields in channel

To reduce hot carrier effects, increase channel length pMOS devices may be better overall for deep submicron circuits
hole mobility is closer to electron mobility under high electric fields which occur in submicron devices hot carrier effects are worse in nMOS devices than pMOS

ECE 410, Prof. A. Mason

Advanced Digital.30

Leakage Currents
All p-n junctions in the MOSFET structure will have a reverse bias leakage current qA j ni Leakage Current, Ilk
where

I lk =

2 0

xd

Undesired effects of leakage currents

add to unwanted static power consumption limit the charge storage time of dynamic circuits

Aj is the junction area ni is the intrinsic carrier concentration 0 is the effective minority carrier lifetime xd is the depletion layer thickness, xd=f(VR)

Factors in leakage
ni is a strong function of temperature (doubles every 11C)
significant in high power density circuit that generate heat
ECE 410, Prof. A. Mason

Advanced Digital.31

Latch-Up
Latch-up is a very real, very important factor in circuit design that must be accounted for Due to (relatively) large current in substrate or n-well
create voltage drops across the resistive substrate/well
most common during large power/ground current spikes

turns on parasitic BJT devices, effectively shorting power & ground

often results in device failure with fused-open wire bonds or interconnects

hot carrier effects can also result in latch-up

latch-up very important for short channel devices

Avoid latch-up by
including as many substrate/well contacts as possible limiting the maximum supply current on the chip

ECE 410, Prof. A. Mason

Advanced Digital.32

Subthreshold Operation
Weak inversion, when VG > 0 but < Vt Subthreshold operation
some channel change and the drain current is small but not zero. referred to as the subthreshold region the drain current is an exponential function of the gate voltage the current increase sharply with VGS until the device turns on

W ( qVGS I D = I D0 e L
where

nkT )

ID

Channel Length

ID0 is a process dependant constant, typically ~20nA n is a process dependant constant, typically n=1.5 kT/q = 26mV at room temperature

Vt

VGS

subthreshold currents are much larger in short channel devices

due to high e-fields at the drain reducing effective channel length effect can be reduced my using lightly-doped drain regions under the gate
ECE 410, Prof. A. Mason Advanced Digital.33

Subthreshold Operation
Analog Circuits
subthreshold operation is exploited for low power operation in low frequency applications W I D = I D 0 e ( qVGS nkT ) Digital Circuits L subthreshold current serves as undesired leakage current want to quickly transition in/out of subthreshold thinner gate oxide = faster transition
same as current technology trend

How can we decrease subthreshold currents and speed transition?

lighter substrate doping = faster transition
opposite of current technology trend

faster transition = less subthreshold leakage current = lower power process must be chosen to match the speed, power, and performance specs for each circuit
what is best for DRAM is not best for microprocessors, etc.

higher doping needed for better short channel performance

ECE 410, Prof. A. Mason

Advanced Digital.34