CHAPTER 14 CMOS DIGITAL LOGIC CIRCUITS

advertisement
CHAPTER 14 CMOS DIGITAL LOGIC CIRCUITS
Chapter Outline
14.1 Digital Logic Inverters
14.2 The CMOS Inverter
14.3 Dynamic Operation of the CMOS Inverter
14.4 CMOS Logic-Gate Circuits
14.5 Implications of Technology Scaling: Issues in Deep-Submicron Design
NTUEE Electronics III
14-1
14.1 DIGITAL LOGIC INVERTERS
The Voltage-Transfer Characteristic (VTC)
 The function of the inverter is to invert the logic value of its input signal
 The voltage-transfer characteristic is used to evaluate the quality of inverter operation
 VTC parameters
 VOH: output high level
 VOL: output low level
 VIH: the minimum value of input interpreted by the inverter as a logic 1
 VIL: the maximum value of input interpreted by the inverter as a logic 0
 Transition region: input level between VIL and VIH
NTUEE Electronics III
14-2
Noise Margins





The VTC is generally non-linear
VIH and VIL are defined as the points at which the slope of the VTC is 1
Robustness (noise margin at a high level): NMH = VOH  VIH
Robustness (noise margin at a low level): NML = VIL  VOL
Static inverter characteristics for ideal VTC:
 VOH = VDD
 VOL = 0
 VIH = VIL = VDD/2
 NMH = NML = VDD/2
Ideal VTC
NTUEE Electronics III
14-3
Power Dissipation
 Static power dissipation: power dissipated when the inverter stays in logic 0 or logic 1
 Dynamic power dissipation: power dissipated as the output is switching
2
PD  f CVDD
Propagation Delay





tPHL : high-to-low propagation delay
tPLH : low-to-high propagation delay
tP (propagation delay) = ( tPLH + tPHL )/2
Maximum switching frequency fmax = 1/2tP
The output transient of the inverter can be characterized by a RC charge/discharge model
vO (t )  V  (V  V0 )e t / RC
NTUEE Electronics III
14-4
Power-Delay Product and Energy-Delay Product




Power and delay are often in conflict for inverter operation
Power-delay product is a figure-of-merit for comparing logic-circuit technologies or families
2
Power-delay product is defined as PDP  PD t P  CVDD
/2
2
Energy-delay product is defined as EDP  CVDD
tP / 2
Silicon Area
 Area reduction through advances in processing technology
 Area reduction through advances in circuit design techniques
 Area reduction through careful chip layout
Fan-In and Fan-Out
 Fan-in of a gate is the number of its inputs
 Fan-out is the maximum number of similar gates that a gate can drive
Logic-Circuit Families
NTUEE Electronics III
14-5
Inverter Implementation
 Simplest implementation of the inverter with a MOSFET and a load
 Inverter implementation with complementary switches
 Inverter implementation with a double-throw switch
NTUEE Electronics III
14-6
14.2 THE CMOS INVERTER
Circuit Operation





A CMOS inverter consists of an n-channel and a p-channel MOSFET
The n-channel device turns on and the p-channel device turns off as the input level goes high
The p-channel device turns on and the n-channel device turns off as the input level goes low
1
1
The turn-on device is modeled by a resistance: rDSN  kn' W / L n (VDD  Vtn ) and rDSP  k p' W / L  p (VDD  | Vtp |)
VOH = VDD and VOL = 0 for any CMOS inverter
NTUEE Electronics III
14-7
The Voltage-Transfer Characteristic
 The transistors go through five different operation regions as the input goes from 0 to VDD
 The operating point is obtained by making iDN = iDP
 Region I: (QN off; QP tri.)
iDN ( 0)  iDP
 Region II: (QN sat.; QP tri.)
iDN 
 Region III: (QN sat; QP sat)
iDN
 Region IV: (QN tri.; QP sat.)
iDN
 Region V: (QN tri.;
tri ; QP off)
iDN
1
1


k n (vI  Vtn ) 2  iDP  k p (VDD  vI  | Vtp |)(VDD  vO )  (VDD  vO ) 2 
2
2


1
1
 k n (vI  Vtn ) 2  iDP  k p (VDD  vI  | Vtp |) 2
2
2
1 2
1

 k n (vI  Vtn )vO  vO   iDP  k p (VDD  vI  | Vtp |) 2
2 
2

 iDP ( 0)
iD
iD
Region I
iD
vO
Region II
iD
Region III
vO
vO
iD
Region IV
vO
NTUEE Electronics III
Region V
vO
14-8
Static Characteristics of the CMOS Inverter
 Ratioless logic: VOH and VOL are independent of ratio of the transistors
 VOH = VDD
 VOL = 0
 Static power dissipation is zero for both states
 Noise margins can be determined by the VTC
 The switching voltage (when vI = vO) is defined by
VM 
r (VDD  | Vtp )  Vtn
r 1
where r 
kp
kn

 p (W / L) p
 n (W / L) n
 VM increases (VTC shifts) with r
 NML increases and NMH decreases as r increases
 NML decreases and NMH increases as r decreases
NTUEE Electronics III
14-9
The Matched Inverter
 A matched inverter has equivalent pull-up and pull-down device with kn = kp and Vtn = |Vtp| = Vt
 The VTC is symmetric
 Determine VIL from the VTC in Region II:
1
1


(vI  Vt ) 2  (VDD  vI  Vt )(VDD  vO )  (VDD  vO ) 2 
2
2


vI  Vt  (VDD  vO )  (VDD  vI  Vt )
dvO
dv
 (VDD  vO ) O
dvI
dvI
1
VIL  (3VDD  2Vt )
8
 Determine VIH from the VTC in Region IV:
1
1
(vI  Vt )vo  vO2  (VDD  vI  Vt ) 2
2
2
dv
dv
vO  (vI  Vt ) O  vO O  (VDD  vI  Vt )
dvI
dvI
1
VIH  (5VDD  2Vt )
8
 Noise margins: NMH = NML = (3VDD + 2Vt)/8
 Switching voltage: VM = VDD/2
NTUEE Electronics III
14-10
14.3 DYNAMIC OPERATION OF THE CMOS INVERTER
Determining the Propagation Delay
 Evaluated by charging/discharge the output capacitor C through QP and QN
 Average current method:
 tPHL:
1
iDN ( E )  iDN (M )
2
1
iDN ( E )  k n (VDD  Vtn ) 2
2
2

 VDD  1  VDD  
iDN ( M )  k n (VDD  Vtn )
 
 
2
 2  2  


CVDD
C
t PHL 
 n
k nVDD
2 I av
I av 
 7 3Vtn Vtn2 
 2 
4
V
VDD 
DD

where  n  2 /  
 tPLH:
t PLH 
 pC
CVDD

k pVDD
2 I av
 7 3 | Vtp | Vtp2 
 2 
where n  2 /  
VDD 
 4 VDD
 Propagation delay: tP = (tPHL+tPLH)/2
NTUEE Electronics III
14-11
 An alternative approach:
 Modeling the turn-on device as a resistance
 Use RC charge/discharge behavior to evaluate the propagation delay
 The empirical values of the resistors are given by
RN 
12.5
(k)
(W / L) n
and
RP 
30
(k)
(W / L) p
 tPHL= 0.69RNC
 tPLH= 0.69RPC
NTUEE Electronics III
14-12
Determining the Equivalent Load Capacitance
 Components accountable for the equivalent load capacitance
 Transistor parasitic capacitances
 Wiring capacitance or interconnect capacitance
 Input capacitance of the following stages
C  2C gd 1  2C gd 2  Cdb1  Cdb 2  C g 3  C g 4  Cw
NTUEE Electronics III
14-13
Inverter Sizing
 Minimum length permitted by the technology is usually used as the length for all channels
 Device aspect ratio (W/L)n is usually selected in the range 1 to 1.5
 The selection of (W/L)n is relative to (W/L)n
 Matched inverter by (W/L)p : (W/L)n = n: p
 (W/L)p = (W/L)n: minimum area, small propagation delay
 (W/L)p = 2(W/L)n: a frequently used compromise
 Transistor sizing (aspect ratios are increased by a factor of S) versus propagation delay
 Load capacitance: C  Cint  Cext  SCint 0  Cext
 Equivalent resistance: Req  1 ( RN  RP ) 
2 S
S
Req 0
S
 Req 0 
1


( SCint 0  Cext )  0.69 Req 0Cint 0  Req 0Cext 
S


 S 
 Propagation delay: t P  0.69
Dynamic Power Dissipation
2
 Dynamic power dissipation: PD  fCVDD
 Peak current: I peak  1 kn  VDD  Vtn 
2
 2
2

NTUEE Electronics III
14-14
14.4 CMOS LOGIC-GATE CIRCUITS
CMOS Logic-Gate Structure
Implementation of PDN
Implementation of PUN




The PDN can be most directly synthesized by expressing Y .
The PUN can be most directly synthesized by expressing Y .
The PDN can be obtained from the PUN (and vice versa) using duality property.
However, duality of the PDN and PUN is not a necessary condition.
NTUEE Electronics III
14-15
Transistor Sizing
 The (W/L) ratios are chosen for a worst-case gate delay equal to
that of the basic inverter
 The derivation of equivalent (W/L) ratio is based on the equivalent
resistance of the transistors
rDS  (W / L) 1
 1

1

 ...
Series Connection  (W / L) eq  
 (W / L)1 (W / L) 2

1
Parallel Connection  (W / L) eq  (W / L)1  (W / L) 2  ...
Effects of Fan-In and Fan-Out






Each additional input to a CMOS gate requires two additional transistors
Increases the chip area and the propagation delay due to excess capacitive loading
The number of NAND gate is typically limited to 4
Redesign the logic design may be required for a higher number of inputs
Advantages of using CMOS logic: static power dissipation, ratioless design, noise margin
Disadvantage of using CMOS logic: area, complexity, capacitive loading, propagation delay
NTUEE Electronics III
14-16
Examples for CMOS Logic Gates
NTUEE Electronics III
14-17
14.5 IMPLICATIONS OF TECHNOLOGY SCALING
Moore’s Law
 A new technology is developed for every 2~3 years due to cost and speed requirement
 The trend was predicted more than 40 years ago by Gordon Moore
 For every new technology generation:
 The minimum length is reduced by a factor of 1.414 and the area is reduced by a factor of 2
 The cost is reduced by half or the circuit complexity is doubled
 Device scaling generally decreases the parasitics and enhances the operating speed
 The operating power is reduced
 The current technology node advances into deep-submicron
 Issues in deep
deep-submicron
submicron technologies have to be taken into account for circuit designs
NTUEE Electronics III
14-18
Scaling Implications
NTUEE Electronics III
14-19
Velocity Saturation
 Long-channel devices:
 Drift velocity: vn = nE
 Electric field in the channel: E = vDS/L
 Short-channel devices:
 Velocity saturates at a critical field Ecr with vsat  107 cm/s
 The vDS at which velocity saturates is denoted by VDSsat
 VDSsat = EcrL = vsatL/n
 VDSsat is a device parameter
The II-V
V Characteristics
 Long-channel devices
1
W
 Saturation current: iD   nCox (VGS  Vt ) 2
2
L
 Short-channel devices
 For VGS-Vt < VDSsat: same as long-channel devices
 For VGS-Vt > VDSsat:
1 2 
W
(VGS  Vt )VDSsat  VDSsat


2
L
1


 WCox vsat VGS  Vt  VDSsat 
2


I Dsat   nCox
NTUEE Electronics III
14-20
Current Equation for Velocity Saturation
 For vGS-Vt  VDSsat and vDS  VDSsat, the drain current is given by
iD   n Cox
W
1


VDSsat  vGS  Vt  VDSsat (1  VDS )
L
2


 The current is reduced from the predication of a long-channel device
 The dependence on vGS is more linear rather than quadratic
 Four regions of operation: cutoff, triode, saturation and velocity saturation
 Short-channel PMOS transistors undergo velocity saturation at the same value of vsat
 The effects on PMOS are less pronounced due to lower mobility and higher VDSsat
NTUEE Electronics III
14-21
Subthreshold Conduction
 The device is not complete off in deep-submicron devices as vGS < Vt
 The subthreshold current is exponentially proportional to vGS: iD = Isexp(vGS/nVT)
 It is a problem in digital IC design for two reasons:
 Such current leads to nonzero static power dissipation for CMOS logics
 May cause undesirable discharge of capacitors in dynamic CMOS logics
The Interconnect
 The width of the interconnect scales down with the CMOS technology
 The metal wire is no longer an ideal short
 Series parasitic resistance may cause undesirable voltage drop and excess delay
 Parasitic capacitance to ground may lead to speed degradation and additional dynamic power
NTUEE Electronics III
14-22
Download