Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
ECE6421 Homework 3
R, L, C, Inverter and Gate Delay Due 10/07 in class
1. For the inverter in the figure below, with an output load of 3pF and 0.25 um technology.
a. Calculate tpHL, tpLH, and tp.
b. Are the rising and falling delays equal? Why and why not?
Answer: During the pull-down (discharging) process, the NMOS M1 is beginning from
saturation mode. Vout begins from VDD and decreases. At this point,
1 W 1 1.5
I dn = kn n (VDD − VTn ) 2 = ×115 × × (2.5 − 0.43)2 = 739.1452 µ A
2 Ln 2 0.5
V
Req 0 = DD = 3.3823K Ω
I dn
VDD
When Vout = , the discharging process is ended. At this point, M1 is in linear region. Thus:
2
VDD − Vout = I RL × RL = Idn 1 × RL
and
Wn VDS2 Wn Vout2
I dn1 = kn (V − V
DD Tn DS )V − =
n k (V − V )V
DD Tn out −
Ln 2 Ln 2
We can get I dn1 = 353.6250 µ A .
Req1 = Vout / I dn1 = 3.5348K Ω
Req 0 + Req1 3.3823K + 3.5348 K
So Req = = = 3.4586 K Ω . Thus
2 2
t pHL = ln(2) × ( Req || RL ) × CL = 0.69 × 3.3061K × 3 ×10−12 = 6.8436ns
Similarly:
t pLH = ln(2) × RL × CL = 0.69 × 75K × 3 ×10−12 = 155.25ns
Then, we get:
t +t
t p = pLH pHL = 81.0468ns
2
3.2. Assume that the propagation delay of a minimum size inverter is 70 ps. Also assume
that the input capacitance of a gate is proportional to its size. Determine the sizing of the
two additional buffer stages that will minimize the propagation delay. (Assume γ =1 )
CL 20 pF
Answer: F = = = 2000 , let CL / Ci be evenly distributed over the 3 inverters, we
Ci 10 fF
Figure 3.2 Buffer insertion for driving large loads
get: f = 3 F = 3 2000 = 12.5992 .
So the sizing for the two additional buffers is 12.6 and 158.76 times of the minimum
size, respectively.
b. If you could add any number of stages to achieve the minimum delay, how many stages
would you insert? What is the propagation delay in this case?
N
F ln F
Answer: Assume N stages are inserted, we should have: γ + N F − = 0 . When γ =1,
N
f opt is around 3.6. Thus N = log 3.6 2000 ≈ 6(integer) . Thus 5 stages should be inserted.
So t p = Nt p 0 (1 + N F / γ ) = 6 × 70 p × (1 + 6 2000 /1) = 1910.8 ps .
c. Describe the advantages and disadvantages of the methods shown in (a) and (b).
Answer: For method in (a), the overall delay is larger than the delay by method (b). But
compared with method (a), method (a) will insert more inverter, it needs more area (inverters
and interconnects), which introduce more resistance and capacitance. This will lead to more
power consumption and potential degradation to the circuit, for example, crosstalk.
G 12
24 24 24
C D E
12
F
24 24
A B
F 4
8 12
A C
12 2
D G
8 12
B E
For pull-up, when all PMOSs are on, the pull-up capability is highest, thus
ABCDEFG = 0000000 . When only one pull-up path is on, and the current from VDD goes
through most PMOSs, it is a worst case. This kind of paths include:
ABCDEFG = {0101110,1001110, 0110110,1010110, 0111010,1011010}
For pull-down, when all NMOSs are on, the pull-down capability is highest, thus
ABCDEFG = 1111111 . When only one pull-down path is on, and the current from Vout goes
through most NMOSs, it is a worst case. This kind of paths include:
ABCDEFG = {0000001,1100010, 0011110}
5. Given a chain of gates as shown below in the figure. Select gate sizes x and y for least
delay from A to B. Also for each stage, in order to get equivalent tpHL and tpLH, what size
should their PMOS and NMOS be? Note that the load capacitance CL = 45 Cunit, the input
gate capacitance of any gate is proportional to its size.
x
y
x
45
A 8
x
y B
45