Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
1
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
Lecture 19
High Frequency Analysis of FET Circuits
In this lecture you will learn:
• High Frequency Analysis of FET Circuits• Miller Effect and the Miller Capacitance
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
Phasor Analysis: Complex Impedances
R
C
tjevtv Re
tjeiti Re
+
-
tjrr evtv Re
Capacitive Impedance/Admittance:
CjYZ
11
Inductance Impedance/Admittance:
LjY
Z
1
L~
2
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
Phasor Analysis: Calculations with Impedances
R
C
v
i
+
-
rv cv
+ -
vRCj
v
CjR
Cjvc
1
11
1
vRCj
RCjv
CjR
Rvr
11
One can compute the voltage phasors using the impedances:
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
Phasor Analysis: Bode Plots
R
C
v
i
+
-
rv cv
+ -
HRCjv
vc
1
1
210log10 H
RC1
~3 dB
Slope = -20 dB/decade
A transfer function with a pole at frequency 1/RC
3dB break point
3
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
Poles and Zeroes of Transfer Functions
R
C
v
i
+
-
rv cv
+ -
1
1cv
v j RC
1
rv j RC
v j RC
Pole at the complex frequency j
RC
Pole at the complex frequency j
RC
Zero at the complex frequency 0
With a little abuse of language we will say a pole at the frequency 1
RC
With a little abuse of language we will say pole at the frequency and a zero at 0 frequency
1
RC
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
Laplace Transform Calculations: Language Differences to Note
R
C
v s
i s
+
-
rv s cv s
+ -
1
11 1c
sCv s v s v ssRCR
sC
1 1r
R sRCv s v s v s
sRCRsC
One can compute the voltages using Laplace transforms as well:
s j
Pole at 1
RC
Pole at 1
RC
Zero at 0
4
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
Phasor Analysis: Bode Plots
R
C
v
i
+
-
rv cv
+ -
HRCj
RCjvvr
1
210log10 H
RC1
~3 dB
Slope = +20 dB/decade
A transfer function with a zero at zero frequency and a pole at frequency 1/RC
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
The Miller Effect and the Miller CapacitanceJohn A. Miller (1920)
Consider a voltage amplifier:
sv
sR
out in sv Av Av inv
Consider now a capacitor sitting at the input of an amplifier:
1
out in
ss
v Av
Av
j R C
sv
sR inv
C
Input voltage to the amplifier decreases, and so does the output voltage of the amplifier, at high frequencies (but not too bad…..)
|A| >> 1A < 0
1
in sin s
s in s
Z vv v
R Z j R C
1inZ
j C
Open circuit voltage gain
5
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
Consider now a capacitor straddling the input and the output of an amplifier:
out inv Av tv
inv
The Miller Effect and the Miller Capacitance
C ti
1
1
1
t t out t t
t
tin
t
i j C v v j C v Av
j C A v
vZ
i j C A
|A| >> 1A < 0
The capacitance, as seen from the input end, is effectively very large!
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
Capacitor straddling the input and the output of an amplifier:
ss
inout
vACRj
A
Avv
11
sv
sR
inv
The amplifier input voltage, and the output voltage, will now begin to drop-off at a much lower frequency!! This is the Miller effect and the capacitance positioned this way is called the Miller capacitance
The Miller Effect and the Miller Capacitance
C
1 1
in sin s
s in s
Z vv v
R Z j R C A
1
1inZj C A
The pole has shifted to a much lower frequency!
|A| >> 1A < 0
6
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
The Miller Effect and the Miller Capacitance: Summary
ss
inout
vACRj
A
Avv
11
In both cases:
A A
1
1inZj C A
Lesson:
A capacitor C straddling the input-output terminals of an amplifier of gain A = -|A| behaves like a very large capacitor C(1-A) sitting between the input and the ground terminals.
This capacitor reduces the gain bandwidth of the circuit
inin s
s in
Zv v
R Z
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
NFET: Capacitances
P-Si Substrate (or Bulk)
Gate
+_
_
N-Si
GSV
+_
DSV
SBV
+
SourceDrain
y0y Ly
N-Si
Csb Cdb
Cgs Cgd
7
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
NFET: High Frequency Small Signal Model
+gsv-
Gate
+bsv-Source
Base
Draindi
gsmvg bsmbvg+dsv-
oggsC
dbCsbC
gigdC
gbC
FET high frequency model:
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
NFET: High Frequency Small Signal Model
+gsv-
Gate
+bsv-
Source Base
Draindi
gsmvg bsmbvg+dsv
-
oggsC
gigdC
Our simplified high frequency model:
8
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
NFET Gate ChargeThe total gate charge QTG (units: Coulombs) consists of the image charge due to:
1) The total inversion layer charge QTN
2) The total depletion layer charge QTB
TG TN TBQ Q Q
The inversion layer charge QN (units: Coulombs/unit area) is a function of position “y” in the device:
N ox GS TN CSQ y C V V V y
The total inversion layer charge QTN (units: Coulombs) can be found by integration over the entire area of the FET:
0
0
L
TN N
L
ox GS TN CS
Q W Q y dy
WC V V V y dy
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
NFET Gate Charge
0
0
DS
L
TN ox GS TN CS
V
ox GS TN CS CSCS
Q WC V V V y dy
dyWC V V V dV
dV
Recall that (Lecture 10):
D N n y
CSn ox GS TN CS
n ox GS TN CS
CS D
I WQ y E y
dV yW C V V V y
dy
W C V V V ydy
dV I
2 2
0
2 3 3
3
DSVn
TN ox GS TN CS CSD
nox GS TN DS GS TN
D
Q WC V V V dVI
WC V V V V VI
9
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
Drain
Gate
SiO2
Ly
L
The channel has been “pinched off”
Channel potential: TNGSCS VVyV
Channel potential: DSCS VyV
For :DS GS TNV V V
Drain
Gate
SiO2
Ly
Channel potential: DSCS VyV
Linear Region:
For :DS GS TNV V V Saturation Region:
NFET: Inversion Layer Charge in Linear and Saturation
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
NFET Gate Charge
Inversion layer charge QN :
2 3 3
3n
TN ox GS TN DS GS TND
Q WC V V V V VI
In saturation (VDS > VGS - VTN) but the inversion charge remains fixed at the value when VDS = VGS – VTN because of pinch-off:
2
3TN ox GS TNQ WLC V V
Expression only works in the linear region
, ,
2
3GD GB GD GB
TG TN TB
TG TNgs ox
GS GSV V V V
Q Q Q
Q QC WLC
V V
Therefore, in saturation (since the depletion region thickness and charge do not change with the gate voltage), we get:
10
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
NFET Capacitances in Saturation: A Simple Model
, , ,
0
GS GB GS GB GS GB
TG TG TNgd
GD DS DSV V V V V V
Q Q QC
V V V
In saturation, because of pinch off, the inversion layer charge QTN is not affected by the drain voltage:
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
Capacitances
In Saturation:
,
2 2
3 3GS SB
Ggs ox ov p ox
GS V V
QC WLC WC WC WLC
V
,
0 0
GS SB
Ggd ov p
GD V V
QC WC WC
V
Overlap capacitances
Parasitic capacitances
11
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
NFET: High Frequency Small Signal Model for Saturation Region
+gsv-
Gate
+bsv-
Source Base
Draindi
gsmvg bsmbvg+dsv
-
oggsC
gigdC
Our simplified high frequency model:
2
3gs ox ov pC WLC WC WC
gd ov pC WC WC
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
+
INV
DDV
R
tvV outOUT
tvs
+
The Common Source Amplifier
+ inv
-
di
gsmvg+ outv
-oggsC
gigdC
R sv
A high frequency small signal model can be built as follows:
+ gsv-
sR
sR
tjss evtv Re
12
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
The Common Source Amplifier: Open Circuit Voltage Gain
KCL at (1) gives:
gdinoutinmoutod Cjvvvgvgi
Also:
Riv dout
The above two give:
gdo
gdm
in
outv RCjRg
RCjRg
v
vA
1
+ inv
-
di
gsmvg+ outv
-oggsC
gigdC
R+ gsv-
(1)
Need to find:
in
outv v
vA and
s
out
vv
H
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
The Common Source Amplifier: Open Circuit Voltage Gain
ogd
m
gd
omgdo
gdm
in
outv rRCj
g
Cj
rRgRCjRg
RCjRg
vv
A||1
1||
1
210log10 vA
gdo CrR ||
1
Slope = -20 dB/decade
210 ||log10 om rRg
gd
m
Cg
A relatively largefrequencyThese poles might not limit the frequency performance…….PTO
21010log 1 0
13
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
The Common Source Amplifier: Open Circuit Voltage Gain
ogd
m
gd
omgdo
gdm
in
outv rRCj
g
Cj
rRgRCjRg
RCjRg
v
vA
||1
1||
1
vA
gdo CrR ||
1
gd
m
Cg
0
2
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
+ inv
-
di
gsmvg+ outv
-oggsC
gigdC
R sv+ gsv-
sR
The Common Source Amplifier: Input-Output Response
So far we have found the open circuit voltage gain:
ogd
m
gd
omgdo
gdm
in
outv rRCj
g
Cj
rRgRCjRg
RCjRg
v
vA
||1
1
||1
What about:
??out
s
v
v
14
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
The Common Source Amplifier: Input Impedance
+ inv
-
di
gsmvg+ outv
-oggsC
ti gdC
R tv + gsv-
(2)
KCL at (2) gives:
1
1
1
t gs t gd t out
gs gd v t
tin
t gs gd v
i j C v j C v v
j C j C A v
vZ
i j C j C A
vin
out Av
v
vgdgsin ACCj
Z
1
1The input impedance is:
gi
+
-
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
The Common Source Amplifier: Input Impedance
+ inv
-
di
gsmvg+ outv
-oggsC
gigdC
R sv+ gsv-
sR
svgdgsins
in
s
in
RACCjZRZ
vv
111
Looking in from the input terminal the capacitance Cgd seems to be magnified by a factor proportional to the open circuit voltage gain of the amplifier!!
Input voltage divider Miller Effect!!
15
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
The Common Source Amplifier: Total Gain
+ inv
-
di
gsmvg+ outv
-oggsC
gigdC
R sv+ gsv-
sR
Finally:
Where:
svgdgs
v
ins
inv
s
in
in
out
s
out
RACCjA
ZRZ
Avv
vv
vv
H
11
ogd
m
gd
omin
outv rRCj
g
Cj
rRgv
vA
||1
1||
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
The Common Source Amplifier: Poles and Zeros of the Total Gain
21
3
11
10
11
jj
jH
RACCjA
Hsvgdgs
v
A little tedious algebra can show that the transfer function above has two poles and a zero
1) Assuming gmRS >> 1, these poles are:
om rRgH ||0
gd
m
Cg
3
1
ogdgs
m
rRCCg
||11
1
somgdgs RrRgCC ||
11
2
This pole will likely determine the smallest frequency at which the total gain rolls over
2) Assuming gmRS << 1, these poles are:
gssCR11
1
This pole will likely determine the smallest frequency at which the total gain rolls over
ogd rRC ||11
2
16
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
The Common Source Amplifier: The Miller Approximation
0 ||
1m gdout
v v m oin o gd
g R j RCvA A g R r
v g R j RC
Miller approximation
If the poles in Av() are at high enough frequencies such that the lowest poles of the full transfer function are determined by other poles (which would be the case if gmRS >> 1 ) then one may approximate the open circuit gain Av() by its low frequency value:
And then:
1 1
||
1 1 ||
out v
s gs gd v s
m o
gs gd m o s
v AH
v j C C A R
g R r
j C C g R r R
Looking in from the input terminal the capacitance Cgd seems to be magnified by a factor proportional to the low frequency open circuit voltage gain of the amplifier!!
And now the single pole is at: somgdgs RrRgCC ||1
11
1 1
out v
s gs gd v s
v AH
v j C C A R
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
+ inv
-
di
gsmvg+ outv
-oggsC
gigdC
R sv+ gsv-
sR
The Common Source Amplifier: Total Gain Under the Miller Approx
+ inv
-
di
gsmvg+ outv
-og
gsC
gi omgd rRgC ||1
R sv+
gsv-
sR
Miller approximation is equivalent to saying that the following circuit will describe the input-output response at not too high frequencies………
17
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
The Common Source Amplifier: Input Impedance
+ inv
-
di
gsmvg+ outv
-og
gsC
gi omgd rRgC ||1
R sv+
gsv-
sR
Looking in from the input terminal the capacitance Cgd seems to be magnified by a factor proportional to the low frequency open circuit voltage gain of the amplifier!!
omgdgsin rRgCCj
Z||1
1
Miller approximation is equivalent to saying that the following circuit will describe the input-output response at not too high frequencies………
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
210log10 H
-20 dB/dec
210 ||log10 om rRg
The Common Source Amplifier: Total Gain Under the Miller Approx
Hv
v
s
out
||1
1
omgdgss rRgCCR Can be large
+ inv
-
di
gsmvg+ outv
-og
gsC
gi omgd rRgC ||1
R sv+
gsv-
sR
18
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
gsmvg oggsC
gdC
R+ gsv-
sR
The Common Source Amplifier: Output Impedance
This gives:
ti(1)
tv
KCL at (2) gives:
s
gsgsgsgdgst R
vvCjCjvv
(2)
t
sgdgs
sgdgs v
RCCj
RCjv
1
Need to find:
outZ
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
gsmvg oggsC
gdC
R+ gsv-
sR
The Common Source Amplifier: Output Impedance
KCL at (1) gives:
1||11
||
sgdgs
sgdgsm
oout
gdgstgsmo
tt
RCCj
RCjCjg
rRZ
CjvvvgrR
vi
ti(1)
tv
(2)
t
sgdgs
sgdgs v
RCCj
RCjv
1
Not a bad approximation even at moderately high frequencies
19
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
+ inv
-
outi
gsmvg oggsC
ing ii gdC
R si+ gsv-
sR
The Common Source Amplifier: Short Circuit Current Gain
Need to find:
in
out
ii
gdgs
m
inmingdmin
in
in
out
in
out
CCjg
ZgZCjgiv
v
i
i
i
gdm
in
out
outinmgdin
Cjgv
i
ivgCjv
00
Start from KCL at (1):
(1)
gdgsin CCj
Z
1
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
Short Circuit Current Gain and the Transition Frequency (fT or T)
For most transistors, the short circuit current gain falls of with frequency with a -20 dB/dec slope (at high enough frequencies)
The frequency at which the short circuit current gain equals unity is called the transition frequency (fT or T)
The transition frequency expresses the maximum useful frequency of operation of the transistor
2
10log10
in
out
i
i
0
Slope = -20 dB/decade
T
???
20
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
The Common Source Amplifier: The Transition Frequency
gdgs
m
in
out
CCjg
ii
2
10log10
in
out
i
i
0
Slope = -20 dB/decade
T
gs
m
gdgs
mT C
gCC
g
The transition frequency is:
ECE 315 – Spring 2005 – Farhan Rana – Cornell University
The Common Source Amplifier: The Transition Frequency
gs
mT C
g
The FET transition frequency is:
This is the highest frequency at which the transistor is still usefulQ: What is its physical significance?
0y Ly
Gate
yESource Drain
VCS=VDSVCS=VGS-VTN
VCS=0
oxgs
TNGSoxnm
WLCC
VVCLW
g
32
22
3
L
VVCg TNGS
ngs
mT
Therefore the transition frequency is:
t
nynTNGS
nTNGS
nT Lv
L
E
LL
VV
L
VV
1
2
Electron transit time through the FET
Therefore, the electron transit time sets the maximum frequency of operation of the FET!!
21
ECE 315 – Spring 2005 – Farhan Rana – Cornell University