EE141 s09 Hw3 Solution
EE141 s09 Hw3 Solution
College of Engineering
Department of Electrical Engineering and Computer Sciences
1. We sometimes design the inverter to favor one transition. Fig 1 shows two examples with
unequal rising and falling delays.
(a) (b)
Figure 1 Skewed inverters.
(1) Calculate the logical effort of the skewed inverters shown in Fig. 1 for both rising and
falling output transitions. Which inverter does favor rising output transition? Why? (10
pts)
(2) Suppose the skewed inverters shown in Fig. 1 are included in your digital cell library.
Choose and design the inverter m1, m2, and m3 to minimize the delay Td for the low-to-
high input transition, as shown in Fig. 2. Calculate the delay Td and the corresponding
delay for the high-to-low input transition. (15 pts)
Solution:
(1) For skewed inverter (a):
Logical effort for the rising output transition
Logical effort for the falling output transition
For skewed inverter (b):
Logical effort for the rising output transition
Logical effort for the falling output transition
Skewed inverter (b) has the smaller logical effort for the rising output transition and
therefore favors rising output transition.
(2) For the low-to-high input transition, the output of m1 is pulled high, the output of m2 is
pulled low, and the output of m3 is pulled high. As a result, we can choose inverter (b) as
the second stage, inverter (a) as the third stage, and inverter (b) as the final stage to
minimize Td.
, , ,
, , ,
2. Following schematic implements the logic equation of (A+B+C+D). The four inverters
connected to inputs are minimum sized.
Figure 3 4-Input OR logic.
(1) If Cload is 25xCgmin, where Cgmin is the gate capacitance of a minimum-size inverter. Size
the NAND, NOR and INV gates to minimize the propagation delay from input to output.
(10 pts)
(2) Repeat part (1) if Cload is 1xCgmin. Be reminded that the input inverters are already
minimum sized. Subsequent gates therefore cannot be smaller than unit size. (10 pts)
(3) Can you design another circuit that implements the same logic but has shorter
propagation delay for the case of Cload =1xCgmin? The four input inverters come from the
I/O interface design. They can neither be replaced nor be resized. (10 pts)
Solution:
(1)
, , ,
(Above sizing is based on unit-size NAND and NOR gates that have equal resistance as a
unit-size inverter. If unit-size NAND and NOR gates are sized with equal input
capacitance as a unit-size inverter, sizing should then be: , s2 = 2.7 , s3 = 5.5 and
s4 = 8.7 .)
€ €
€
unit size (You get full credits too if you argue that the NAND gate can be sized
down to 0.5x, because the minimum transistor width in a unit-size NAND gate is 2. This
apparently does not apply to an NOR gate.)
Now apply logical effort analysis to the 3-stage circuit starting with NAND since its size
has been fixed.
, ,
Therefore,
, . Again this is impractical
Finally apply logical effort analysis to the 2-stage circuit starting with NOR.
, , ,
By rearranging the circuit to the following, propagation delay can be reduced (there are
other options such as using 4-input NOR):
(1) Size the inverter m1 and m2 for the minimum delay. (5 pts)
(2) Assume all inverters share the same supply Vdd. What is the total energy drawn from the
supply when the input switches from 0 to Vdd? What is the total energy dissipated as
heat? (10 pts)
Solution:
(2) The total energy drawn from the supply when the input switches from 0 to Vdd is: