Topic 11 - Logical Efforts
Topic 11 - Logical Efforts
[email protected]
o Multi-stage Logic Networks
o Choosing the Best Number of Stages
o Example
Harvey Mudd College
o Asymmetric & Skewed Logic Gates
Claremont, CA
o Circuit Families
o Summary
Introduction Example
Chip designers face a bewildering array of choices. ? ? ? Ben Bitdiddle is the memory designer for the Motoroil 68W86, an embedded
processor for automotive applications. Help Ben design the decoder for a
o What is the best circuit topology for a function?
register file: a<3:0> a<3:0> 32 bits
o How large should the transistors be?
o How many stages of logic give least delay?
4:16 Decoder
Decoder specification:
16 words
Register File
Logical Effort is a method of answering these questions: o 16 word register file 16
o Each word is 32 bits wide
o Uses a very simple model of delay
o Back of the envelope calculations and tractable optimization
o Each bit presents a load of 3 unit-sized transistors
o True and complementary inputs of address bits a<3:0> are available
o Gives new names to old ideas to emphasize remarkable symmetries
o Each input may drive 10 unit-sized transistors
Who cares about logical effort?
Ben needs to decide:
o Circuit designers waste too much time simulating and tweaking circuits
o High speed logic designers need to know where time is going in their logic
o How many stages to use?
o How large should each gate be?
o CAD engineers need to understand circuits to build better tools
o How fast can the decoder operate?
Logical Effort David Harris Page 3 of 56 Logical Effort David Harris Page 4 of 56
Outline Delay in a Logic Gate
Logical Effort David Harris Page 5 of 56 Logical Effort David Harris Page 6 of 56
6 g= 6 g = 4/3
p=
D
Normalized delay: d
5
AN
d= p=2
D
Normalized delay: d
g= 5
AN
N
g=1d = (4/3)h + 2
4 p=
ut
N
r
rte
inp
d= 4 p=1
ut
r
ve
rte
inp
d=h+1
2-
3
in
ve
2-
effort How about a 3
in
2 delay 2-input NOR? effort
2 delay
1
parasitic delay 1
parasitic delay
1 2 3 4 5
Electrical effort: h = Cout / Cin 1 2 3 4 5
Electrical effort: h = Cout / Cin
o d = f + p = gh + p o d = f + p = gh + p
o Delay increases with electrical effort o Delay increases with electrical effort
o More complex gates have greater logical effort and parasitic delay o More complex gates have greater logical effort and parasitic delay
Logical Effort David Harris Page 7 of 56 Logical Effort David Harris Page 45 of 56
Computing Logical Effort A Catalog of Gates
DEF: Logical effort is the ratio of the input capacitance of a gate to the input Table 1: Logical effort of static CMOS gates
capacitance of an inverter delivering the same output current. Number of inputs
Gate type
o Measured from delay vs. fanout plots of simulated or measured gates 1 2 3 4 5 n
Logical Effort David Harris Page 8 of 56 Logical Effort David Harris Page 9 of 56
Example Example
Estimate the frequency of an N-stage ring oscillator: Estimate the frequency of an N-stage ring oscillator:
Logical Effort David Harris Page 10 of 56 Logical Effort David Harris Page 46 of 56
Example Example
Estimate the delay of a fanout-of-4 (FO4) inverter: Estimate the delay of a fanout-of-4 (FO4) inverter:
d d
This is about 60 ps
in a 0.18 μm process.
Logical Effort David Harris Page 11 of 56 Logical Effort David Harris Page 47 of 56
Can we write F = GH ?
Logical Effort David Harris Page 12 of 56 Logical Effort David Harris Page 13 of 56
Branching Effort Branching Effort
No! Consider circuits that branch: No! Consider circuits that branch:
15 G = 15 G =1
H = H = 90 / 5 = 18
90 GH = 90 GH = 18
5 5
h1 = h1 = (15+15) / 5 = 6
15 h2 = 15 h2 = 90 / 15 = 6
90 F = = GH? 90 F = 36, not 18!
Logical Effort David Harris Page 14 of 56 Logical Effort David Harris Page 48 of 56
We can now compute the delay of a multi-stage network: Gate sizes can be found by starting at the end of the path and working backward.
o At each gate, apply the capacitance transformation:
o Path Effort Delay: DF = ∑ fi C out • g i
PF = ∑ pi
i
o Path Parasitic Delay: C in = ----------------------
-
i ˆf
o Path Delay: DF = ∑ di = D F + P o Check your work by verifying that the input capacitance specification is satis-
fied at the beginning of the path.
We can prove that delay is minimized when each stage bears the same effort:
ˆf = g h = F 1 ⁄ N
i i
Therefore, the minimum delay of an N-stage path is:
1⁄N
NF +P
o This is a key result of logical effort. Lowest possible path delay can be found
without even calculating the sizes of each gate in the path.
Logical Effort David Harris Page 15 of 56 Logical Effort David Harris Page 16 of 56
Example Example
Select gate sizes y and z to minimize delay z Select gate sizes y and z to minimize delay z
9C 9C
from A to B y from A to B y
z z
A 9C A 9C
C C
Logical Effort: G = y z 3 y z
B Logical Effort: G = (4 ⁄ 3) B
9C 9C
Electrical Effort: H = C out
Branching Effort: B = Electrical Effort: H = --------- = 9
C in
Path Effort: F = Branching Effort: B = 2•3 = 6
Best Stage Effort:
ˆf = Path Effort: F = GHB = 128
Delay: D = ˆf = F 1 ⁄ 3 ≈ 5 Work backward for sizes:
Best Stage Effort:
9C • (4 ⁄ 3)
z = ----------------------------- = 2.4 C
Work backward for sizes: 5
z = Delay: D = 3 • 5 + 3 • 2 = 21 • ( 4 ⁄ 3)
y = 3z
---------------------------- = 1.92 C
y = 5
Logical Effort David Harris Page 17 of 56 Logical Effort David Harris Page 49 of 56
Datapath load
64 64 64 64
Fastest
N: 1 2 3 4
f: 64 8 4 2.8
D: 65 18 15 15.3
1⁄N 1⁄N
D = NF + P = N ( 64 ) + N assuming polarity doesn’t matter
Logical Effort David Harris Page 18 of 56 Logical Effort David Harris Page 19 of 56
Derivation of the Best Number of Stages Best Number of Stages (continued)
Suppose we can add inverters to the end of a path without changing its function. p inv + ρ ( 1 – ln ρ ) = 0 has no closed form solution.
o ˆ
How many stages should we use? Let N be the value of N for least delay. o Neglecting parasitics (i.e. pinv = 0), we get the familiar result that ρ = 2.718 (e)
Logic Block: N-n1 extra inverters o For pinv = 1, we can solve numerically to obtain ρ = 3.59
n1 stages How sensitive is the delay to using exactly the best number of stages?
Path effort F
1 .6
D(N) / D(N)
n1 1 .5 1
1 .4
1 .2 6
1⁄N
D = NF +
∑ p i + ( N – n 1 ) p inv 1 .2
1 .0
1 .1 5
I like to use
1 ρ=4
(ρ = 2 .4 ) (ρ = 6 )
∂D 1⁄N 1⁄N 1⁄N
------- = – F ln ( F )+F + p inv = 0
∂N
ˆ 0 .0
1⁄N 0 .5 0 .7 1 .0 1 .4 2 .0
o Define ρ ≡ F to be the best stage effort. Substitute and simplify:
N /N
p inv + ρ ( 1 – ln ρ ) = 0 o 2.4 < ρ < 6 gives delays within 15% of optimal -> we can be sloppy
Logical Effort David Harris Page 20 of 56 Logical Effort David Harris Page 21 of 56
Outline Example
o Introduction Let’s revisit Ben Bitdiddle’s decoder problem using logical effort:
a<3:0> a<3:0> 32 bits
o Delay in a Logic Gate
o Multi-stage Logic Networks
4:16 Decoder
16 words
o Choosing the Best Number of Stages Decoder specification:
Register File
o Example o 16 word register file 16
o Asymmetric & Skewed Logic Gates o Each word is 32 bits wide
o Circuit Families o Each bit presents a load of 3 unit-sized transistors
o Summary o True and complementary inputs of address bits a<3:0> are available
o Each input may drive 10 unit-sized transistors
Logical Effort David Harris Page 22 of 56 Logical Effort David Harris Page 23 of 56
Example: Number of Stages Example: Number of Stages
How many stages should Ben use? How many stages should Ben use?
o Effort of decoders is dominated by electrical and branching portions o Effort of decoders is dominated by electrical and branching portions
o Electrical Effort: H = o Electrical Effort:
32 • 3- = 9.6
H = --------------
o Branching Effort: B = 10
o Branching Effort: B = 8 because each address input
If we neglect logical effort (assume G = 1), controls half the outputs
o Path Effort: F =
If we neglect logical effort,
Remember that the best stage effort is about ρ = 4 o Path Effort: F = GBH = 8 • 9.6 = 76.8
o Hence, the best number of stages is: N =
Remember that the best stage effort is about ρ = 4
o Hence, the best number of stages is: N = log 476.8 = 3.1
o Let’s try a 3-stage design
Logical Effort David Harris Page 24 of 56 Logical Effort David Harris Page 50 of 56
Example: Gate Sizes & Delay Example: Gate Sizes & Delay
Lets try a 3-stage design using 16 4-input NAND gates with G = Lets try a 3-stage design using 16 4-input NAND gates with G = 1 • 2 • 1 = 2
a0a0 a1a1 a2a2 a3a3 a0a0 a1a1 a2a2 a3a3
10 unit input capacitance 10 unit input capacitance
y out0 y out0
z z
96 unit wordline 96 unit wordline
capacitance capacitance
y out15 y out15
z z
o Actual path effort is: F = o Actual path effort is: F = 2 • 8 • 9.6 = 154 Close to
o Therefore, stage effort should be: f = o Therefore, stage effort should be: f = ( 154 ) = 5.36
1⁄3 4, so f is
reasonable
o Gate sizes: z = y = o z = 96 • 1 ⁄ 5.36 = 18 y = 18 • 2 ⁄ 5.36 = 6.7
o Path delay: D = o D = 3f + P = 3 • 5.36 + 1 + 4 + 1 = 22.1
Logical Effort David Harris Page 25 of 56 Logical Effort David Harris Page 51 of 56
Example: Alternative Decoders Outline
Logical Effort David Harris Page 26 of 56 Logical Effort David Harris Page 27 of 56
Asymmetric logic gates favor one input over another. Asymmetric logic gates favor one input over another.
Example: suppose input A of a NAND gate is most critical. Example: Suppose input A of a NAND gate is most critical:
o Select sizes so pullup and pulldown still match unit inverter o Select sizes so pullup and pulldown still match unit inverter
o Place critical input closest to output o Place critical input closest to output
2 2
2 2
x
x a 4/3
a 4/3
4 b
4 b
Effort on A
o Logical Effort on input A: g A = 10 ⁄ 9 goes down at
o Logical Effort on input A: gA = expense of
o Logical Effort on input B: gB = 2 effort on B and
o Logical Effort on input B: gB = total gate effort
o Total Logical Effort: g tot = g A + g B = 28 ⁄ 9
o Total Logical Effort: g tot = g A + g B
Logical Effort David Harris Page 28 of 56 Logical Effort David Harris Page 52 of 56
Symmetry Factor Skewed Gates
In general, consider gates with arbitrary symmetry factor s: Skewed gates favor one edge over the other.
o s = 1/2 in symmetric gate with equal sizes Example: suppose rising output of inverter is most critical.
2 2
o s = 1/4 in previous example
x
o Downsize noncritical NMOS transistor to reduce total input capacitance
a 1/(1-s)
2 2 1
1/s b x x x
Logical effort of inputs: a 1/2 a 1 a 1/2
1
------------ + 2
1
--- + 2
1
-------------------- + 4
1 –s s s (1 – s) HI-Skewed inverter
Unskewed w/ Unskewed w/
gA = ---------------------- gB = ------------- g tot = ------------------------------ equal rise equal fall
3 3 3
Compare with unskewed inverter of the same rise/fall time to compute effort.
o Critical input approaches logical effort of inverter = 1 for small s
o Logical Effort for rising (up) output: gu =
o But total logical effort is higher for asymmetric gates
o Logical Effort for falling (down) output: gd =
o Average Logical Effort: g avg = ( g u + g d ) ⁄ 2
Logical Effort David Harris Page 29 of 56 Logical Effort David Harris Page 30 of 56
Skewed gates favor one edge over the other. DEF: Logical effort of a skewed gate for a particular transition is the ratio of the
Example: suppose rising output of inverter is most important. input capacitance of that gate to the input capacitance of an unskewed
inverter delivering the same output current for the same transition.
o Downsize noncritical NMOS transistor to reduce total input capacitance
a a a
o LO-Skewed gates favor falling outputs by downsizing PMOS transistors
1/2 1 1/2
o Logical effort is smaller for the favored input due to lower input capacitance
Skewed inverter
Unskewed w/ Unskewed w/ o Logical effort is larger for the other input
equal rise equal fall
Compare with unskewed inverter of the same rise/fall time Critical rising
effort goes down
o Logical Effort for rising (up) output: gu = 5 ⁄ 6 at expense of
noncritical and
o Logical Effort for falling (down) output: g d = 5 ⁄ 3 average effort
o Average Logical Effort: g avg = ( g u + g d ) ⁄ 2 = 5 ⁄ 4
Logical Effort David Harris Page 53 of 56 Logical Effort David Harris Page 31 of 56
Catalog of Skewed Gates Outline
1 2
gu = 2
1
gd = 1
1
2 gavg = 3/2
LO-Skew
1 gu = 4/3 2
gu = 2
gd = 2/3 1
2
gd = 1 1
gavg = 1
gavg = 3/2
Logical Effort David Harris Page 32 of 56 Logical Effort David Harris Page 33 of 56
Pseudo-NMOS Pseudo-NMOS
Pseudo-NMOS gates replace fat PMOS pullups on inputs with a resistive pullup. Pseudo-NMOS gates replace fat PMOS pullups on inputs with a resistive pullup.
o Resistive pullup must be much weaker than pulldown stack (e.g. 4x) o Resistive pullup must be much weaker than pulldown stack (e.g. 4x)
o Reduces logical effort because inputs must only drive the NMOS transistors o Reduces logical effort because inputs must only drive the NMOS transistors
o However, NMOS current reduced by contention with pullup o However, NMOS current reduced by contention with pullup
o Unequal rising and falling efforts o Unequal rising and falling efforts
o Quiescent power dissipation when output is low o Logical effort can be applied to domino, pseudo-NMOS, and other logic families
2/3 2/3
Example: Pseudo-NMOS inverter Example: Pseudo-NMOS inverter x
x
o Logical Effort for falling (down) output: gd = a
o Logical Effort for falling (down) output: gd = 4 ⁄ 9 a 4/3
4/3
o Logical Effort for rising (up) output: gu = o Logical Effort for rising (up) output: gu = 4 ⁄ 3
o Average Logical Effort: g avg = ( g u + g d ) ⁄ 2 o Average Logical Effort: g avg = ( g u + g d ) ⁄ 2 = 8 ⁄ 9
Logical Effort David Harris Page 34 of 56 Logical Effort David Harris Page 54 of 56
Pseudo-NMOS Gates Dynamic Logic
Dynamic logic replace fat PMOS pullups on inputs with a clocked precharge.
Inverter NAND2 NOR2 o Reduces logical effort because inputs must only drive the NMOS transistors
o Eliminates pseudo-NMOS contention current and power dissipation
2/3 2/3 2/3 o Only the falling (“evaluation”) delay is critical
x x x
4/3
o Downsize noncritical precharge transistors to reduce clock load and power
a 4/3 a 8/3 a 4/3
b
b 8/3 φ
Example: Footless dynamic inverter 1
x
gd = 4/9 gd = 8/9 gd = 4/9 o Logical Effort for falling (down) output: gd = a
gu = 4/3 gu = 8/3 gu = 4/3 1
gavg = 8/9 gavg = 16/9 gavg = 8/9
Robust gates may require keepers and clocked pulldown transistors (“feet”).
Tradeoffs exist between power and effort by varying P/N ratio. o Feet prevent contention during precharge but increase logical effort
o Weak keepers prevent floating output at cost of slight contention during eval
Logical Effort David Harris Page 35 of 56 Logical Effort David Harris Page 36 of 56
Dynamic logic replace fat PMOS pullups on inputs with a clocked precharge. Inverter NAND2 NOR2
o Reduces logical effort because inputs must only drive the NMOS transistors
φ 1 φ 1 φ 1
o Eliminates pseudo-NMOS contention current and power dissipation
x
x x
o Critical pulldown (“evaluation”) delay independent of precharge size Footless
a
a 1 2 a 1 1 b
gd = 1/3 gd = 2/3 gd = 1/3
b 2
φ 1
Example: Footless dynamic inverter
x
o Logical Effort for falling (down) output: gd = 1 ⁄ 3 a 1
φ 1 φ 1 φ 1
x x x
Footed
Robust gates may require keepers and clocked pulldown transistors (“feet”). a 2 a 3 a 2 2 b
o Feet prevent contention during precharge but increase logical effort gd = 2/3 b 3 gd = 1 gd = 2/3
φ 2 φ 2
o Weak keepers prevent floating output at cost of slight contention during eval φ 3
Logical Effort David Harris Page 55 of 56 Logical Effort David Harris Page 37 of 56
Domino Gates Domino Gates
Dynamic gates require monotonically rising inputs. Dynamic gates require monotonically rising inputs.
o However, they generate monotonically falling outputs o However, they generate monotonically falling outputs
o Alternate dynamic gates with HI-skew inverting static gates o Alternate dynamic gates with HI-skew inverting static gates
o Dynamic / static pair is called a domino gate o Dynamic / static pair is called a domino gate
Logical Effort David Harris Page 38 of 56 Logical Effort David Harris Page 56 of 56
Assumptions: o Introduction
o PMOS transistors have half the drive of NMOS transistors o Delay in a Logic Gate
o Skewed gates downsize noncritical transistors by factor of two o Multi-stage Logic Networks
o Pseudo-NMOS gates have 1/4 strength pullups o Choosing the Best Number of Stages
Table 4: Summary of Logical Efforts o Example
Inverter g n-input NAND g n-input NOR g o Asymmetric & Skewed Logic Gates
Circuit Style
gu gd gu gd gu gd o Circuit Families
Static CMOS 1 (n+2)/3 (2n+1)/3 o Summary
HI-Skew 5/6 5/3 (n/2+2)/3 (n+4)/3 (2n+.5)/3 (4n+1)/3
LO-Skew 4/3 2/3 2(n+1)/3 (n+1)/3 2(n+1)/3 (n+1)/3
Pseudo-NMOS 4/3 4/9 4n/3 4n/9 4/3 4/9
Footed Dynamic 2/3 (n+1)/3 2/3
Footless Dynamic 1/3 n/3 1/3
Logical Effort David Harris Page 39 of 56 Logical Effort David Harris Page 40 of 56
Summary Method of Logical Effort
Table 5: Key Definitions of Logical Effort Logical effort helps you find the best number of stages, the best size of each gate,
Term Stage expression Path expression and the minimum delay of a circuit with the following procedure:
∏ gi
Logical effort g (seeTable 1) G =
Electrical effort C C
o Compute the path effort: F = GBH
out out (path)
h = --------- H = ---------------------
- ˆ ≈ log F
C in C in (path) o Estimate the best number of stages: N 4
Branching effort n/a ˆ
B = ∏ bi o Estimate the minimum delay: D = N̂F
1⁄N
+P
Effort f = gh F = GBH o Sketch your path using the number of stages computed above
Effort delay f ˆf = F 1 ⁄ N
DF = ∑ fi o Compute the stage effort:
Number of stages 1 o Starting at the end, work backward to find transistor sizes:
N
C out • g i
∑
Parasitic delay p (seeTable 2) P = pi i
C in = ----------------------
-
i ˆf
Delay d = f+p D = DF + P
Logical Effort David Harris Page 41 of 56 Logical Effort David Harris Page 42 of 56
Logical effort is not a panacea. Some limitations include: Logical effort is a useful concept for thinking about delay in circuits:
o Facilitates comparison of different circuit topologies
o Chicken & egg problem o Easily select gate sizes for minimum delay
how to estimate G and best number of stages before the path is designed o Circuits are fastest when effort delays of each stage are equal and about 4
o Simplistic delay model o Path delay is insensitive to modest deviations from optimal sizes
neglects effects of input slopes o Logic gates can be skewed to favor one input or edge at the cost of another
o Interconnect o Logical effort can be applied to domino, pseudo-NMOS, and other logic families
iteration required in designs with branching and non-negligible wire C or RC
same convergence difficulties as in synthesis / placement problem Logical effort provides a language for engineers to discuss why circuits are fast.
o Maximum speed only o Like any language, requires practice to master
optimizes circuits for speed, not area or power under a fixed speed constraint
A book on Logical Effort is available from Morgan Kaufmann Publishers
o http://www.mkp.com/Logical_Effort
o Discusses P/N ratios, gate characterization, pass gate logic, forks, wires, etc.
Logical Effort David Harris Page 43 of 56 Logical Effort David Harris Page 44 of 56
Delay Plots Example
D
Normalized delay: d
5
AN
g=1d = (4/3)h + 2
N
4 p=1
ut
r
rte
inp
d=h+1
ve
2-
3 g≡1
in
Logical Effort:
effort
2 delay C out
Electrical Effort: h = --------- = 1
1 C in
parasitic delay
Parasitic Delay: p = p inv ≈ 1
1 2 3 4 5
Electrical effort: h = Cout / Cin
Stage Delay: d = gh + p = 2 A 31 stage ring
oscillator in a
1 1 0.18 μm process
Oscillator Frequency: F = ------------------- = ----------- oscillates at about
o d = f + p = gh + p 2Nd abs 4N τ 670 MHz.
o Delay increases with electrical effort
o More complex gates have greater logical effort and parasitic delay
Logical Effort David Harris Page 45 of 56 Logical Effort David Harris Page 46 of 56
Estimate the delay of a fanout-of-4 (FO4) inverter: No! Consider circuits that branch:
d 15 G =1
H = 90 / 5 = 18
90 GH = 18
5
h1 = (15+15) / 5 = 6
15 h2 = 90 / 15 = 6
90 F = 36, not 18!
This is about 60 ps
in a 0.18 μm process. Now we can compute the path effort:
∏ hi = BH ≠ H
in circuits that branch
o Path Effort: F = GBH
Logical Effort David Harris Page 47 of 56 Logical Effort David Harris Page 48 of 56
Example Example: Number of Stages
Select gate sizes y and z to minimize delay z How many stages should Ben use?
9C
from A to B y z o Effort of decoders is dominated by electrical and branching portions
A 9C 32 • 3- = 9.6
C
y o Electrical Effort: H = --------------
3 z B 10
Logical Effort: G = (4 ⁄ 3) 9C o Branching Effort: B = 8 because each address input
C out controls half the outputs
Electrical Effort: H = --------- = 9
C in
Branching Effort: B = 2•3 = 6 If we neglect logical effort,
Path Effort: F = GHB = 128 o Path Effort: F = GBH = 8 • 9.6 = 76.8
ˆf = F 1 ⁄ 3 ≈ 5 Work backward for sizes:
Best Stage Effort: Remember that the best stage effort is about ρ = 4
9C • (4 ⁄ 3)
z = ----------------------------- = 2.4 C
5 o Hence, the best number of stages is: N = log 476.8 = 3.1
Delay: D = 3 • 5 + 3 • 2 = 21 • ( 4 ⁄ 3) o
y = 3z
---------------------------- = 1.92 C Let’s try a 3-stage design
5
Logical Effort David Harris Page 49 of 56 Logical Effort David Harris Page 50 of 56
Lets try a 3-stage design using 16 4-input NAND gates with G = 1 • 2 • 1 = 2 Asymmetric logic gates favor one input over another.
a0a0 a1a1 a2a2 a3a3 Example: Suppose input A of a NAND gate is most critical:
10 unit input capacitance o Select sizes so pullup and pulldown still match unit inverter
o Place critical input closest to output
y out0 2 2
z
96 unit wordline x
capacitance
a 4/3
y out15 4 b
z
Effort on A
o Actual path effort is: F = 2 • 8 • 9.6 = 154 Close to o Logical Effort on input A: g A = 10 ⁄ 9 goes down at
expense of
4, so f is o
o Therefore, stage effort should be: f = ( 154 )1 ⁄ 3 = 5.36 reasonable
Logical Effort on input B: gB = 2 effort on B and
total gate effort
o z = 96 • 1 ⁄ 5.36 = 18 y = 18 • 2 ⁄ 5.36 = 6.7 o Total Logical Effort: g tot = g A + g B = 28 ⁄ 9
o D = 3f + P = 3 • 5.36 + 1 + 4 + 1 = 22.1
Logical Effort David Harris Page 51 of 56 Logical Effort David Harris Page 52 of 56
Skewed Gates Pseudo-NMOS
Skewed gates favor one edge over the other. Pseudo-NMOS gates replace fat PMOS pullups on inputs with a resistive pullup.
Example: suppose rising output of inverter is most important. o Resistive pullup must be much weaker than pulldown stack (e.g. 4x)
o Downsize noncritical NMOS transistor to reduce total input capacitance o Reduces logical effort because inputs must only drive the NMOS transistors
o However, NMOS current reduced by contention with pullup
2 2 1 o Unequal rising and falling efforts
x x x o Logical effort can be applied to domino, pseudo-NMOS, and other logic families
a 1/2 a 1 a 1/2
Logical Effort David Harris Page 53 of 56 Logical Effort David Harris Page 54 of 56
Dynamic logic replace fat PMOS pullups on inputs with a clocked precharge. Dynamic gates require monotonically rising inputs.
o Reduces logical effort because inputs must only drive the NMOS transistors o However, they generate monotonically falling outputs
o Eliminates pseudo-NMOS contention current and power dissipation o Alternate dynamic gates with HI-skew inverting static gates
o Critical pulldown (“evaluation”) delay independent of precharge size o Dynamic / static pair is called a domino gate
Logical Effort David Harris Page 55 of 56 Logical Effort David Harris Page 56 of 56