CMOS Inverter: Propagation Delay

advertisement
I. CMOS Inverter: Propagation Delay
A. Introduction
• Propagation delays tPHL and tPLH define ultimate speed of logic
• Define Average Propagation Delay
t

+t
PHL PLH 

t = ----------------------------------------p
2
• Typical complex system has 20-50 propagation delays per clock
cycle.
• Typical propagation delays < 1nsec
B. Hand Calculation
• Use an input signal that has tr=0 and tf=0 for hand calculation
• Calculate current drive
• Calculate capacitance being driven
VIN
VOH
tCYCLE
VOL
t
VOUT
tPHL
tPLH
VOH
VOH
50%
tCYCLE
VOL
t
EECS 6.012 Spring 1998
Lecture 13
C. Load Capacitance: Input to Next Stage
• Typical configuration: load capacitance CL consists of the input
capacitances of the next stage of inverters plus parasitic drain/bulk
capacitance and wiring capacitance
VDD
W
L p2
VDD
VDD
2
W
L n2
W
L p1
VIN
+
VOUT
CL
VIN
1
VDD
W
L n1
W
L p3
−
3
W
L n3
(a)
(b)
• Estimation of the input capacitance: n- and p-channel transistors in
the next stage switch from triode through saturation to cutoff during
a high-low or low-high transition
• Requires nonlinear charge storage elements to accurately model
• Hand Calculation use a rough estimate for an inverter
C
in
=
C
ox
( W ⋅ L ) p + C ox ( W ⋅ L ) n
• CG above is
C
G
=
C
ox
( WL ) p2 + C ox ( WL ) n2 + C ox ( WL ) p3 + C ox ( WL ) n3
EECS 6.012 Spring 1998
Lecture 13
D. Parasitic Capacitance-Drain/Bulk Depletion
gate oxide
n+ polysilicon gate
source
interconnect
drain
interconnect
bulk
interconnect
,
,,,,
,,,,
,,
,,,,
,,,,
,,,
,,,,
,,,,
deposited
oxide
,
,,
,,
,,
L
field
oxide
n+ drain diffusion
Ldiff
n+ source diffusion
p+
[ p-type ]
(a )
,,,,,
,
,,,,
,,,
, , ,,,,,,,,,,,,,,
,,,,, ,,,,,,,,,,,,
,,,,, , , , , , ,
,,,,, , , , , ,
,
,
,,,,,,,,,,,,
,,,,,,,,,,,,
,
,
,
,
,
,
active area (thin
oxide area)
gate contact
gate
interconnect
polysilicon gate
contact
n+ polysilicon gate
A
metal
interconnect
source contacts
W
bulk
contact
source
interconnect
drain
interconnect
L
drain
contacts
edge of
active area
(b)
L
W
Ldiff
(c)
EECS 6.012 Spring 1998
Lecture 13
Calculation of Parasitic Depletion Capacitance
• Depletion qJ(vD) is non-linear --> take the worst case and use the
zero-bias capacitance Cjo as a linear charge-storage element during
the transient.
• “Bottom” of depletion regions of the inverter’s drain diffusions
contribute a depletion capacitance
CBOTT = CJn(WnLdiffn) + CJp(WpLdiffp)
• CJn and CJp being the zero-bias junction capacitances (fF/µm2) for
the n-channel MOSFET drain-bulk junction and the p-channel
MOSFET drain-bulk junction, respectively.
Typical numbers: CJn and CJp are about 0.2 fF/µm2
• “Sidewall” of depletion regions of the inverter’s drain diffusions
make an additional contribution:
CSW = (Wn + 2Ldiffn)CJSWn + (Wp + 2Ldiffp)CJSWp
• CJSWn and CJSWp being the zero-bias sidewall capacitances
(F/µm) for the n-channel MOSFET drain-bulk junction and the
p-channel MOSFET drain-bulk junction, respectively.
Typical numbers: CJSWn and CJSWp are about 0.5 fF/µm
• The sum of CBOTT and CSW is the total depletion capacitance, CDB
EECS 6.012 Spring 1998
Lecture 13
,,
E. Parasitic Capacitance-Wires
• “Wires” consist of metal lines connecting the output of the inverter
to the input of the next stage
,,
,,,,,
,,,,,
,,,,,
,,,,,
,,,,,
,,,,,
,,,,,
,,,,,
,
,
,
metal interconnect
(width Wm, length Lm)
polysilicon
gate
0.6 µm deposited oxide
0.5 µm thermal oxide
p+
p
(grounded)
gate oxide
• The p+ layer (i.e., heavily doped with acceptors) under the thick
thermal oxide (500 nm = 0.5 µm) and deposited oxide (600 nm = 0.6
µm) depletes only slightly when positive voltages appear on the
metal line, so the capacitance is approximately the oxide
capacitance:
C
wire
=
C
thickox
( W m ⋅ Lm)
• where the oxide thickness = 500 nm + 600 nm = 1.1 µm.
• For large digital systems, the parasitic capacitance can dominate the
load capacitance
CL = CG + CP = CG + (CDB + Cwire)
EECS 6.012 Spring 1998
Lecture 13
F. High-to-Low Transition (at Output)
• AssumeVIN switches instantly from low to high.
• Driver transistor (n-channel) switches from cutoff to saturation
• p-channel load switches from triode to cutoff
• Circuit during high-to-low transition:
IDn
vOUT (t)
+
CG+ CP
VIN (t > 0) = VOH
_
• The voltage on the load capacitor at t = 0- was V+
• Since n-channel MOSFET is initially saturated and the input voltage
is a constant, the drain current is initially IDn(sat) for VGS = V+.
2
I Dn ( sat ) = ( 1 ⁄ 2 ) µ n C ox ( W ⁄ L ) n ( V OH – V Tn )
VOH = V+
• Propogation delay is (by definition) the time required for VOUT to
reach V+ / 2:
dv
– I Dn ( sat )
d  QL 
OUT
--------------- = -----  -------------------  = ----------------------dt  C G + C P 
dt
C +C
G
P
EECS 6.012 Spring 1998
Lecture 13
G. Hand Calculation of tPHL
• The output waveform is:
vOUT (t)
VOH
slope = dvOUT / dt
VOH/2
0
tPHL
t
• The high-to-low propagation delay is given by
( V OH ⁄ 2 ) – V OH
– I Dn ( sat )
--------------- = ---------------------------------------- = ----------------------t
dt
C +C
PHL
G
P
dv
OUT
• Solving for the delay:
t
PHL
( C G + C P ) ( V OH ⁄ 2 )
= -----------------------------------------------------------------------------------2
( 1 ⁄ 2 ) µ n C ox ( W ⁄ L ) n ( V OH – V Tn )
EECS 6.012 Spring 1998
Lecture 13
H. Graphical Interpretation
• The n-channel driver remains saturated throughout the first half of
the transition from high-to-low... all that matters, according to the
definition of propagation delay for hand analysis.
ID
VOUT
t = 0+
VIN = VOH
t = tPHL
VOH
t = 0−
VIN = 0V
VOH
2
0
0
VOH
2
VOH
VOUT
(a)
0
0
t
tPHL
(b)
I. Hand Calculation of tPLH
• low-to-high transition, the p-channel load is supplying a constant
current -IDp(sat) to charge up the load and parasitic capacitance.
t
PLH
( C G + C P ) ( V OH ⁄ 2 )
= -------------------------------------------------------------------------------------2
( 1 ⁄ 2 ) µ p C ox ( W ⁄ L ) p ( V OH + V Tp )
• For identical propagation delays, the (W/L) of the p-channel load is a
factor of two higher than for the n-channel driver, to compensate for
the lower hole mobility in the channel.
EECS 6.012 Spring 1998
Lecture 13
II. Power Dissipation
• Energy from power supply needed to charge up the capacitor:
E
charge
=
∫V
+
i(t) dt
=
+
V Q
+ 2
= ( V ) ( CG + C P)
• Energy stored in the capacitor:
+ 2
1

E
= - (C + C ) (V )
store  2 
G
P
• Energy lost in p-channel MOSFET during charging:
E
diss
=
E
 1-  ( C + C ) ( V + ) 2
E
–
=
 2 G
charge
store
P
• During discharge, the n-channel MOSFET driver dissipates an
identical amount of energy.
• If the charge/discharge cycle is repeated f times/second, where f is
the clock frequency, the dynamic power dissipation is:
P
= ( 2E diss ) ⋅
f
+ 2
= ( CG + C P) ( V )
f
• In practice, many gates don’t change state for every clock cycle,
which lowers the power dissipation
• Additional source of dissipation: power flow from V+ to ground
when both transistors are saturated. Can be signifcant, but hard to
estimate by hand. Eliminate with Circuit Techniques.
EECS 6.012 Spring 1998
Lecture 13
III. CMOS Static Logic Gates
A. CMOS Static NAND Gate
VDD
M3
A
M4
A
B
+
M2
B
VOUT
_
M1
(b)
B. I-V Characteristics of n-channel devices
VDD
VM
M3
M4
ID
VM
VM
VGS1 = VM
M2
+
V
M1 DS1
VGS2 = VM − VDS1
ID1 = ID2
−
VDS
(a)
(b)
EECS 6.012 Spring 1998
Lecture 13
C. Define kneff and kpeff for the NAND Gate
• Effective width of two p-channel devices is 2Wp
• BUT worst case only 1 device is on...kpeff = kp3 = kp4
,,
,,
• Effective length of two n-channel devices is 2Ln (Derivation in Text)
• kneff = kn1/2 = kn2/2
VM
,,
,,
source
VM
gate
gate
M1
M2
n+
L1
L2
(a)
,,
,,
drain
VM
,,
source
VM
gate
gate
M1
M2
L1
L2
,,
drain
(b)
D. Calculation of static and transient performance
• kpeff = kneff is desirable (equal propagation delay; symmetrical
transfer curve),
W
W
 =  ----• Recall µn = 2µp
 Ln
 L-  p 2-input NAND Gate
• For an M-input NAND Gate
M W
W
 = --- ----- Ln
2 Lp
EECS 6.012 Spring 1998
Lecture 13
E. CMOS Static NOR Gate
VDD
M4
A
M3
B
+
M1 B
A
M2 VOUT
_
(a)
• Effective width of two n-channel devices is 2Wn ...kneff = 2kn1 = 2kn2
• BUT worst case only 1 device is on...kneff = kn1 = kn2
• Effective length of two p-channel devices is 2Lp (Same reasoning)
• kpeff = kp3/2 = kp4/2
• An M-input NOR gate-requires very wide p-channel devices since
µp = µn/2
W
 = 2M W

---- Lp
 Ln
EECS 6.012 Spring 1998
Lecture 13
Download