Download Lecture 25 MOSFET Basics - Doolittle

Lecture 25
MOSFET Basics (Understanding with Math)
Reading: Pierret 17.1-17.2 and Jaeger 4.1-4.10 and
Notes
Georgia Tech
ECE 3040 - Dr. Alan Doolittle
MOS Transistor I-V Derivation
With our expression relating the Gate voltage to the surface potential and
the fact that S=2F we can determine the value of the threshold voltage
VT  2 F 
VT  2 F 
S
2qN A
C ox
S
2 F 
S
2qN D
 2 F 
C ox
S
(for n - channel devices)
(for p - channel devices)
where,
C ox 
 ox
xox
is the oxide capacitance per unit area
Where we have made use of the use of the expression,
 S  KS o
Georgia Tech
ECE 3040 - Dr. Alan Doolittle
MOS Transistor I-V Derivation
Coordinate Definitions for our “NMOS” Transistor
x=depth into the semiconductor
from the oxide interface.
y=length along the channel from
the source contact
z=width of the channel
xc(y) = channel depth (varies
along the length of the channel).
n(x,y)= electron concentration
at point (x,y)
n(x,y)=the mobility of the
carriers at point (x,y)
Georgia Tech
Device width is Z
Channel Length is L
Assume a “Long Channel” device (for
now do not worry about the channel
length modulation effect) ECE 3040 - Dr. Alan Doolittle
MOS Transistor I-V Derivation
Concept of Effective mobility
The mobility of carriers near the interface is
significantly lower than carriers in the
semiconductor bulk due to interface
scattering.
Since the electron concentration also varies
with position, the average mobility of
electrons in the channel, known as the
effective mobility, can be calculated by a
weighted average,
n


x  xc ( y )
x 0

 n ( x, y )n( x, y )dx
x  xc ( y )
x 0
Empirically
n 
n( x, y )dx
or defining ,
QN ( y )   q 
x  xc ( y )
x 0
n( x, y )dx
ch arg e / cm 
 q x  xc ( y )
n 
 n ( x, y )n( x, y )dx

x

0
QN ( y )
Georgia Tech
o
1   VGS  VT 
where,  o and  are constants
2
ECE 3040 - Dr. Alan Doolittle
MOS Transistor I-V Derivation
Drain Current-Voltage Relationship
In the Linear Region, VGS>VT and 0<VDS<Vdsat
J N  q n nE  qDN n
Neglecting the diffusion current, and recognizing the current is
only in the y-direction,
J N  J Ny
Georgia Tech
d
 q n nE y  q n n
dy
ECE 3040 - Dr. Alan Doolittle
MOS Transistor I-V Derivation
Drain Current-Voltage Relationship
In the Linear Region, VGS>VT and 0<VDS<Vdsat
I D     J Ny dxdz  Z 
x  xc ( y )
x 0
J Ny dx
x  xc ( y )

d 
  q 
 n ( x, y )n( x, y )dx 
   Z
x 0

dy 

d
  Z  n QN
dy

yL
y 0
 VDS
I D dy   Z  n 
 0
 VDS
I D L  Z  n 
 0
 Z n
ID 
L
Georgia Tech
 VDS

0
Q N d
QN d
Q N d
To find ID, we need an expression relating the
electrostatic potential,  and QN ECE 3040 - Dr. Alan Doolittle
MOS Transistor I-V Derivation
“Capacitor-Like” Model for QN
Assumptions:
•Neglect all but the mobile inversion charge (valid for deep inversion)
•For the MOSFET, the charge in the semiconductor is a linear function of
position along the semiconductor side of the plate. Thus,  varies from 0 to VDS
Since Cox 
dQ
,
dV
Source
MOS Capacitor
Only voltages above threshold create inversion charge
QN  Cox VGS  VT  for VGS  VT
Neglect the depletion region charge
Georgia Tech
Drain
MOS Transistor
QN  Cox VGS  VT   
for VGS  VT
Note: Assuming a linear variation of potential along the channel
leads to an underestimation of current but is a good estimate for
hand calculations.
ECE 3040 - Dr. Alan Doolittle
MOS Transistor I-V Derivation
Using “Capacitor-Like” Model for QN we can estimate ID as:
 Z  n  V
ID 
  0 QN d
DS
L
 Z n
ID 
L
  VDS

Z  n C ox
ID 
L
0
 C ox VG  VT   d
2


VDS
VGS  VT V DS 

2


0  V DS  V Dsat
and VGS  VT
This is known as the “square law” describing the
Current-Voltage characteristics in the “Linear” or
“Triode” region.
Note the linear behavior for small VDS (can neglect VDS2 term). Note
the negative parabolic dependence for larger VDS but still VDS<VDsat
(can NOT neglect VDS2 term).
Georgia Tech
ECE 3040 - Dr. Alan Doolittle
MOS Transistor I-V Derivation
“Capacitor-Like” Model for QN
But what about the saturation region?
For VDS>Vdsat the voltage drop across our channel is VDsat with the remaining voltage
(VDS-VDsat) dropped across the pinch-off region
I D  I Dsat
Z  n C ox

L
2


V Dsat




V
V
V
 GS

T
Dsat
2


V Dsat  V DS
But the charge at the end of the channel is zero due to the pinched off channel,
Q N ( y  L)  C ox VGS  VT  V Dsat   0
VGS
or
 VT  V Dsat
Thus,
I D  I Dsat 
Georgia Tech

Z  n C ox
VGS  VT 2
2L

V Dsat  V DS
ECE 3040 - Dr. Alan Doolittle
MOS Transistor I-V Derivation
Summary of MOSFET IV Relationship
Z  n C ox
ID 
L
0  VDS
2


VDS




V
V
V
 GS

T
DS
2 

 V Dsat and VGS  VT
I D  I Dsat 

Z  n C ox
VGS  VT 2
2L

V Dsat  V DS
VDsat  VGS  VT
Georgia Tech
ECE 3040 - Dr. Alan Doolittle
MOS Transistor Applications
Voltage variable Resistor
An n-channel MOSFET has a gate width to length ratio of Z/L=100, un=200
cm2/Vsec, Cox=0.166 uF/cm2 and VT=1V. We want to develop a resistor that has
a resistance that is controlled by an external voltage. Such a device would be used
in “variable gain amplifiers”, “automatic gain control devices”, “compressors” and
many other electronic devices. Define what range of VDS must be maintained to
achieve proper “voltage variable resistance” operation. Find the “On-resistance”
(VDS/ID) of the transistor from 1.5V<VGS<4Vfor small VDS .
First, to achieve voltage variable resistance operation, we must operate in the
linear region. Otherwise, the current is either a constant regardless of drain
voltage (saturation region) or is approximately zero (cutoff due to the capacitor
being in either accumulation and depletion).
Thus, VGS -VT>VDS. Given the values above, 0<VDS<0.5V
Continued...
Georgia Tech
ECE 3040 - Dr. Alan Doolittle
MOS Transistor Applications
Voltage variable Resistor
Using the linear region ID equation:
Z  n C ox
ID 
L
RDS
RDS
2

 Z  n C ox
VDS


VGS  VT VDS  for small VDS
V

V
V

 GS

T
DS
2 
L

L
V DS
V DS
Z



ID
Z  n C ox
 n C ox VGS  VT 
VGS  VT VDS 
L
0.01

200 0.166e  6 F / cm 2 VGS  1


Thus,
100   RDS  600 
Georgia Tech
ECE 3040 - Dr. Alan Doolittle
MOS Transistor Applications
Current Source
The same transistor is to be used for a “Current Source”. Define the range of
drain-source voltage that can be used to achieve a fixed current of 50 uA.
For a constant current regardless of Drain-Source voltage, we must use the
saturation region:


Z  n C ox
VGS  VT 2
VDsat  VDS
2L
100 200cm 2 / VSec 0.166uF / cm 2
VGS  12
50uA 
2
VGS  1.173V
I D  I Dsat 


This source will operate over a VDS>VGS-VT or VDS>0.173 V
Georgia Tech
ECE 3040 - Dr. Alan Doolittle
MOS Transistor: Deviations From Ideal
Channel Length Modulation Effect
Above “pinch-off” (when VDS>VDsat=VGS-VT) the channel length reduces
by a value L.
Thus, the expression for drain current,
I D  I Dsat 

Z  n C ox
VGS  VT 2
2L
Becomes,
I D  I Dsat

Z  nCox
VGS  VT 2

2L  L 
or since * L L,
I D  I Dsat

V Dsat  VDS

VDsat  VDS
1
1  L 
 1 

L  L L 
L 


L 
Z  nCox
2 
VGS  VT  1  

2L
L 

VDsat  VDS
*In many modern devices, this assumption does not hold. Thus, the channel length modulation parameter we are deriving does not describe the IV
expressions well.
Georgia Tech
ECE 3040 - Dr. Alan Doolittle
MOS Transistor: Deviations From Ideal
Channel Length Modulation Effect
But the fraction of the channel that is pinched off
depends linearly on VDS because the voltage
across the pinch-off region is (VDS-VDsat) so,
L
  V DS
L
Channel Length Modulation
causes the dependence of drain
current on the drain voltage in
saturation.
where  is known as the Channel-Length
Modulation parameter and is typically:
0.001 V-1 < V
I D  I Dsat
Georgia Tech


Z  n C ox
VGS  VT 2 1   VDS 

2L
V Dsat  VDS
ECE 3040 - Dr. Alan Doolittle
MOS Transistor: Deviations From Ideal
Body Effect (Substrate Biasing)
Until now, we have only considered
the case where the substrate (Body)
has been grounded….
…but the substrate (Body) is often
intentionally biased such that the
Source-Body and Drain-Body
junctions are reversed biased.
The body bias, VBS, is known as the backgate bias and can be used to modify the
threshold voltage.
Note that now our channel potential has an offset equal to VBS, ….
Georgia Tech
ECE 3040 - Dr. Alan Doolittle
MOS Transistor: Deviations From Ideal
Body Effect (Substrate Biasing)
Thus, our threshold potential with the body grounded,
VT  2 F 
VT  2 F 
S
2qN A
C ox
S
2 F 
S
2qN D
 2 F 
C ox
S
(for n - channel devices)
(for p - channel devices)
Surface Potential S
The Gate- Body Threshold becomes,
2qN D

2 F  VBS  (for p - channel devices)
VGB
 2 F  VBS  S
Cox
S
Threshold
VGB Threshold  2 F  VBS 
S
Cox
2qN A
S
 2 F  VBS 
(for n - channel devices)
But we would like to have this in terms of VGS instead of VGB.
Since, VGS =VGB+VBS
VGS Threshold  2 F 
VT=
Georgia Tech
VGS Threshold  2 F 
S
2qN A
C ox
S
2 F  VBS 
S
2qN D
 2 F  VBS 
C ox
S
(for n - channel devices)
(for p - channel devices)
ECE 3040 - Dr. Alan Doolittle
MOS Transistor: Deviations From Ideal
Body Effect (Substrate Biasing)
This can be rewritten in the following form (more convenient to reference the threshold
voltage to the VBS=0 case).
 2
   2
VT Pierret   VTN  Jaeger   VTO  
F
 VBS   2 F
VT Pierret   VTP  Jaeger   VTO
F
 VBS   2 F


(for n - channel devices)
(for p - channel devices)
where,
 
Georgia Tech
2qN A  S
C ox
is known as the body effect parameter
ECE 3040 - Dr. Alan Doolittle
MOS Transistor:
Enhancement Mode verses Depletion Mode MOSFET
We have been studying the “enhancement mode” MOSFET
(Metal-Oxide-Semiconductor Field Effect Transistor). It is called
“enhancement” because conduction occurs only after the channel
conductance is “improved” or “enhanced”. In this case,
VTN>0 and VTP<0
Transistors can be fabricated such that: VTN  0 and VTP  0
These transistors have conduction for VGS=0 due to a channel
already existing without the need to “invert the near surface
region”. To modulate currents, a field must applied to the gate that
depletes the channel. Thus, transistors of this nature are called
“Depletion mode MOSFETs”.
Georgia Tech
ECE 3040 - Dr. Alan Doolittle
MOS Transistor:
Enhancement Mode verses Depletion Mode MOSFET
Georgia Tech
ECE 3040 - Dr. Alan Doolittle
MOS Transistor:
Summary
4-Terminal
Enhancement
Depletion
3-Terminal
Enhancement
Depletion
NMOS
(n-channel)
PMOS
(p-channel)
Jaeger uses the notation:
NMOS
K n  K n'
W
W
  n Cox
where W is the Gate Width (Z in Pierret)
L
L
PMOS
K p  K p'
Georgia Tech
W
W
  p Cox
where W is the Gate Width (Z in Pierret)
L
L
ECE 3040 - Dr. Alan Doolittle
MOS Transistor:
Summary
NMOS
PMOS
Regardless of Mode
K n  K n'
Cutoff
W
W
W
W
  p Cox
(Note : W  Z in Pierret)
  n C ox
(Note : W  Z in Pierret) K p  K 'p
L
L
L
L
I DS  0
Linear
I DS
Z  n C ox

L
VGS  VTN
Saturation
for VGS  VTN
2


VDS


V
V
V



 GS
TN
DS
2 

and
VGS  VTN  VDS  0

VGS
Threshold Voltage
VTN  VTO  
VT for Enhancement
Mode
VT for Depletion
Mode
Georgia Tech

Z  n C ox
VGS  VTN 2 1   VDS 
2L
 VTN
and
V DS  VGS  VTN  0
I DS 

2 F  VSB   2 F 
I DS  0
for VGS  VTP
2

Z  n C ox 
VDS
ID 
VGS  VTP V DS 

L
2 

VGS  VTP
and
VGS  VTP  VDS  0
VGS


Z  n C ox
VGS  VTP 2 1   VDS 
2L
 VTP
and
VDS  VGS  VTP  0
I DS 
VTP  VTO  

2 F  VBS   2 F 
VTN  0
VTP  0
VTN  0
VTP  0
ECE 3040 - Dr. Alan Doolittle
MOS Transistor:
Bias Circuitry-Enhancement Mode NMOS
Due to zero DC current flow in the gate, the bias analysis of a MOSFET
is significantly easier than a BJT.
A
B
C
Georgia Tech
•Form Thevenin
circuits looking out
the gate, drain, and
source
ECE 3040 - Dr. Alan Doolittle
MOS Transistor:
Bias Circuitry-Enhancement Mode NMOS
3V  I G Rth  VGS
10V  I DS R3  VDS
IG
•But IG=0 so VGS=3V
•Assume Saturation operation (selected for
easy math because IDS does not depend on
VDS since no  was given – =0):

IDS

Kn
vGS  VTN 2 for v DS  vGS  VTN  0
2
25 x10 6
3  12  50 A
i DS 
2
Check V DS
i DS 


10V  50uA(100k )  VDS
VDS  5V  VGS  VTN  2V
•Assumption of Saturation operation was correct! If it were not correct simply
make another assumption (I.e. linear region) and resolve.
Georgia Tech
ECE 3040 - Dr. Alan Doolittle
MOS Transistor:
Bias Circuitry-Depletion Mode NMOS
•Bias circuit of a depletion mode device is much simpler due to the fact
that the device conducts drain current for VGS=0V
VTO  3V
 0 V
K n  200 uA
•What value of R1 results in 100 uA drain current?
•Again Assuming saturation:
V2

Kn
vGS  VTN 2 for v DS  vGS  VTN  0
2
2 I DS
2 100uA
 VTN 
  3V 
  2V
Kn
200uA / V 2
i DS 
VGS

IDS
R1  
VGS
2

 20 K
I DS
100uA
Check V DS
10V  I DS R1  V DS  100uA20k   V DS
V DS  8V  VGS  VTN  2V  (3V )  1V
•Assumption of Saturation operation was correct! If it were not correct simply
make another assumption (I.e. linear region) and resolve.
Georgia Tech
ECE 3040 - Dr. Alan Doolittle
PMOS Transistor:
Bias Circuitry-Enhancement Mode PMOS
VTO  1V
 0 V
K p  25 uA
ID  IS 
VGS  VTP
V2



K
Z  n C ox
VGS  VTP 2 1   VDS   p VGS  VTP 2
2L
2
and
V DS  VGS  VTP  0

1. 5 M
 6V R EQ  1M || 1.5M  600 K
1.5M  1M
 I S RS  VGS  I G RG  V EQ
V EQ  10V
V DD
V DD  V EQ  RS
Kp
2
V
GS
2


25 E  6
 2.66  12  34.4A
2
 I D RS  V DS  I D R D  V DS  6.08V
ID 
V DD


25 E  6
VGS  12  VGS
2
 7.21  0  VGS  2.71 or  2.66V
10  6  4  39,000
VGS2  0.051VGS

 VTP   VGS
Check V DS
V DS  VGS  VTP  0   6.08   2.66  ( 1)  0
Georgia Tech
ECE 3040 - Dr. Alan Doolittle
MOS Transistor:
Bias Circuitry-Possible Combinations
1) Vth _ Base  VGS  I DS Rth _ Source  Vth _ Source
2) I DS
and


K
2
 n VGS  VTN  1  VDS 
2
optionally , Assume
or
I DS
2


VDS
 K n VGS  VTN VDS 

2 

either saturated or linear/triode.
3) Vth _ Drain  VDS  I DS Rth _ Source  Rth _ Drain   Vth _ Source
•Always: Solve 1) for VGS and plug into 2).
•In certain cases, VDS will need to be eliminated by using
3) solved for VDS and plugged into 2).
•Case A: Saturated, and =0 and no source resistor –
only 1 and 2 required. Results in 1st order polynomial.
•Case B: Saturated, and >0 and no source resistor – all
3 equations needed. Results in 1st order polynomial.
•Case C: Saturated, and =0 and a source resistor – all 3
equations needed. Results in 2nd order polynomial.
•Case D: Saturated, and >0 and a source resistor – all 3
equations needed. Results in 3rd order polynomial.
•Case E: Linear/Triode, with or without a source resistor
– all 3 equations needed. Results in 2nd order
polynomial.
Georgia Tech
ECE 3040 - Dr. Alan Doolittle
Useful Formulas for DC Bias Solutions
If a 3rd order polynomial results, try factoring it into a linear and quadratic term 1st. If
this is not easy for your case, a longer but sure fire way is listed below.
Georgia Tech
ECE 3040 - Dr. Alan Doolittle