Layout Solns 3

Submitted by L.
Bollschweiler
[email protected]
3.1 Redraw the layout and cross-scctional views of a pad, similar to Fig. 3.2, if
the final pad size is 50µm
x 75µm with a scale factor 100nm.
# Units = Size / scale. W = 70µm/100nm = 700units. L = 50µm/100nm = 500 units.
700
500 x-sect.
M2 Insulator
Insulator
Insulator
FOX
p-substrate
M2
3.2 Estimate the capacitance to ground of the pad in Fig.3.20 made with both metal1
and metal2.
Soln:
M1 and M2 are connected through vias, thus capacitance from M2 to M1 or M2 to

substrate is not
applicable, leaving only the capacitance from M1 to P-Sub.
2
1sub plate fringe
C area C Scale permeter C Scale = × × + × ×
Getting the values of C

plate
and C
fringe
from table 3.1 on page 61
C
1sub
= 2000* 2000* 23aF /um2 *(50nm)
2
+ 2000* 4*79aF /um*50nm
C
1sub
= .26 pF
Mahesh Balasubramanian
Harish Reddy Singidi
3.3) Suppose a parallel plate capacitor was made by placing a 100µm square piece of
metal1 directly over metal1 below metal2 in Fig: 3.2. Estimate the capacitance
between
the two plates of the capacitor. Estimate the capacitance from metal1 to substrate.
Layout
Cross section view
Metal2 Insulator
C
M2-M1
Metal1 Insulator
Insulator
C
M1-Sub
Insulator
FOX
P-Substrate
C
M1-M2
= Area . Plate capacitance + perimeter . Fringe capacitance
C
M1-M2
= (100µm)
2
. 35 aF/µm
2
+ 400 µm . 100aF/µm
C
M1-M2
= 390fF
C
M1-Sub
= (100µm)
2
. 23 aF/µm
2
+ 400 µm . 79aF/µm
C
M1-Sub
= 261.6fF
HW3 Shambhu Roy
3.4 Sketch the cross section for the layout seen in Fig 3.26
Fig. 3.26 Layout used in Problem 3.4
Sol:
Insulator
Insulator
FOX
p-substrate
Insulator
Overglass layer
Metal 1
KRISHNAMRAJU KURRA
Q #3.5: Sketch the cross-sectional view, at the dashed line, for the layout seen in
Fig.3.27. What is the
contact resistance between metal3 and metal2.
Problem 3.6 Edward Kunz
Estimate the intrinsic propagation delay through a metal line encapsulated in an

ILD
with a relative dielectric constant of 1.5. What value metal sheet resistance,
using the
values from ex 3.3 would be required if the RC delay through metal the metal line
is
equal to the intrinsic delay?
t
d intrinsic =
= =
8 10 * 3
5 . 1
E c
ε
4ns/meter = 4ps/mm
t
d metal =
C R* * 35 .
4ps = .35*R*162fF
R =
Ω = 70
162 * 35 .
4
fF
ps
R sheet =
=
Ω
5000
70
.014Ω
3.7
Using CV=Q rederive the results in Ex. 3.5.
We have 10 by 10 square piece of metal1 and an equal-size piece of metal2 placed

exactly above the metal1 piece.
Scale factor = 50nm.
C. V = Q
For metal 2 to metal 1 C

12
. V
12
= Q
12
For metal 1 to substrate C

1sub
V
1sub
= Q
1sub
From the circuit Q

12
= Q
1sub
Implies C
12.
V
12
= C
1sub
V
1sub
Implies C
12
(∆Vm2 – ∆ V
m1
) = C
1sub
. (∆V
m1
– 0) ---- (1)
From ex 3.5:
C
12
= 10. 10. (0.05)
2
. 35 aF + 40 (0.05). 100 aF
Implies C
12
= 209 aF
C
1sub
= 10. 10. (0.05)
2
. 23 aF + 4. 10. (0.05) 79 aF
Implies C
1sub
= 164 aF
Equation ---(1) implies 209 aF (1- ∆V

m1
) = 164 aF (∆V
m1
)
Implies ∆V
m1
= 209/ (209+ 164) = 560 mV
Therefore ∆V
m1
= 560 mV
PROBLEM 3.8) For the layout seen in figure below sketch the cross-sectional
view(along the dotted line) and estimate the resistance between points A and B.
Remember a via is sized 1.5 by 1.5.
Solution:-
Given metal sheet resistance =0.1 ohm/square

Metal contact resistance = 10 ohms
I assumed that there are 7 squares in each metal (between A and B)
The total resistance between A and B = (6+5+5+5+6) x (0.1ohm/square) + 4 x (10
ohms)
= 2.7 + 40
= 42.7 ohms(approximately)
--Steve Bard
Problem 3.9
Question: Laying out two metal wires directly next to each other, and with minimum
spacing, for a long distance increases the capacitance between the two conductors,
C
m
. If
the two conductors are V
DD
and ground, is this a good idea? Why or why not?
Answer: Yes, this is a good idea. The mutual capacitance between the wires, C
m
, serves
as a decoupling capacitor that supplies charge during transient periods. That
capacitance
helps keep the wires at their respective voltages, V
DD
and ground, when the circuit in the
chip begins drawing power.
EE 510 SPRING 2004 HOMEWORK 4

MAEZLIN J AVILA TAYLOR
P3.10 Consider the following schematic:
This circuit can be used to model ground bounce and VDD droop. Show, using SPICE,
that a decoupling capacitor can be used to reduce these effects for various
amplitude and
duration current pulses.
The schematic drawn in LASI creates the following .CIR file:
*** SPICE Circuit File of 3_10_SCH
.control
destroy all
run
plot vx-vy
.endc
.tran 1n 50n
* End of E:\Lasi7\Mosis\3_10_sch.hdr
* MAIN 3_10_SCH
Cdecoupling Vx Vy 1p
Ipulse Vx Vy DC 0 PULSE 0 10u 10n 1n 1n 2n 50n
R1 vn1 Vx 5k
R2 0 Vy 5k
VDD vn1 0 DC 1
.END
This file plots the difference between the model VDD and model ground nodes:
The 1pF capacitor controls the droop/bounce caused by a 10uA current spike
resulting in
a minimal 2.5% change.
A longer pulse, 10ns, of the same amplitude results in a slightly more significant
change
of 4.75%:
A bigger pulse, such as 50uA results in a significant 15% change:
3.11) Lay out the pad frame specified by the information in Table 3.2. Assume a 3
metal CMOS
process is used. Comment on how the scale factor affects the (drawn) layout size.
Sol: - As stated in the book Table 3.2 is given as below:-
Layout of single pad & pad frame is as shown below:-
Scale factor determines the actual size of the feature being fabricated on the
wafer. Actual size =scale
factor ×drawn size; smaller the scale factor smaller is the actual size and thus
more number of die per
wafer can be processed which makes the product cheaper because, processing cost per
wafer is fixed.
-Shantanu Gupta
100µ
100µ
Metal-3 to Metal-2 to Metal-1 Pad
88µ×88µ
Over glass layer
130µ
center tocenter distance
24 pin pad frame
Geng Zheng
3.12 Propose, and lay out, a test structure to measure the sheet resistance oI
metal3. Comment on
the trade-oIIs between accuracy and layout size. Using your test structure provide
a numerical
example oI calculating sheet resistance Ior metal3.
Solution:
Let's get started by writing down the equation Ior calculating the resistance
R=R
E
N+
_
i=1
N
AR
i
(1)
where R
E
is the sheet resistance,
AR
i
is the variation oI the i -th sheet resistance, and N is
the total number oI R
E
. II the layout is relatively large and symmetrical, the error term is
averaged to zero
_
i=1
N
AR
i
-0 (2)
Also notice that the resistance oI the conductor (metal3) is probably very small.
Intuitively we
know that a number oI R
E
is necessary in order to minimize the error and to avoid
electromigration. Since we also don't want the layout size to be too large, the
test structure
may be laid out in a serpentine pattern, Fig. 1.
The overall resistance between A and B is
R
AB
=
J
AB
I
AB
=R
E
( N
R
N
L
+N
J
+0.6N
C
+2) (3)
where
N
R
is the number oI rows,
N
L
is the number oI R
E
within a row excluding the
corners,
N
J
is the total number oI R
E
between corners in the vertical lines, and
N
C
is the
number oI corners (note that the corner sheet resistance is estimated to be 0.6 R
E
). II we lay
out the test structure so that in Fig. 1
W
Gap
=W
L
, we have
N
J
=N
R
-1
N
C
=2 N
R
-2
(4)
Figure 1: A test structure in serpentine pattern.
Thus rewriting Eq. (3) we get
R
E
=
J
AB
I
AB
1
|( N
L
+2.2) N
R
-0.2¦
(5)
Knowing
N
L
and
N
R
we can use Eq. (5) to estimate the sheet resistance.
Suppose we are tasked to measure the metal3 sheet resistance oI the ON
Semiconductor C5
process (the scaling Iactor \=300 nm ) |1|. BeIore we starting layout, let's take
a look at the
sheet resistance Ior diIIerent processes |2|:
Process Metal3 Sheet Resistance ( D)
IBM 130 nm 0.05
IBM 90 nm 0.08
TSMC 180 nm 0.04
TSMC 350 nm 0.07
We can make a guess that metal3 sheet resistance oI the C5 process may be around
0.05D.
Lets lay out a test structure consists oI approximately 5000 R
E
in order to reduce the error
term in Eq. (1). Lets also try to avoid elecromigration as well as large layout
size by setting
the width oI the metal3 to be twice oI the minimum in DRC rules, or a drawn width
oI 10. II
there are 100 R
E
in a single row oI the structure, we then need about 50 rows. This layout is
shown in Fig. 2.
Figure 2: A test structure layout in a serpentine pattern.
Since there are corners in the serpentine pattern, the layout in Fig. 2 only has 49
rows.
ThereIore
N
R
=49
and
N
L
=100
. Using Eq. (5), the relation between R
E
and test voltage
J
AB
and measured current
I
AB
is
R
E
=
J
AB
I
AB
199.7×10
-6
(6)
A chip Iabricated through MOSIS may be 1.5mm×1.5 mm . The width oI this structure
is
roughly 10010\=0.3mm , or 1/5 oI the total chip width. II more accurate result is
needed,
we can increase the total number oI R
E
(a larger layout size).
A Iinal comment is about the electromigration. The metal width in this layout is
10\ , or
3um . A typical current threshold Ior avoiding electromigration eIIect is 1mA/
um . Thus the
maximum allowed current Ilowing through this structure is 3mA . Assume R
E
=0.05D,
using Eq. (6), the voltage applied to the test structure should not exceed 0.75V.
Reference:
|1| ON Semiconductor C5 Process, http://www.mosis.com/on¸semi/c5/
|2| MOSIS Iabrication processes, http://www.mosis.com/products/Iab/vendors/

Layout Solns 3

Uploaded by

Copyright:

Available Formats

Layout Solns 3

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Layout Solns 3

Uploaded by

Copyright:

Available Formats

Submitted by L.

# Units = Size / scale. W = 70µm/100nm = 700units. L = 50µm/100nm = 500 units.

M1 and M2 are connected through vias, thus capacitance from M2 to M1 or M2 to

Getting the values of C

Cross section view

Fig. 3.26 Layout used in Problem 3.4

Problem 3.6 Edward Kunz

Estimate the intrinsic propagation delay through a metal line encapsulated in an

We have 10 by 10 square piece of metal1 and an equal-size piece of metal2 placed

For metal 2 to metal 1 C

For metal 1 to substrate C

From the circuit Q

Equation ---(1) implies 209 aF (1- ∆V

Given metal sheet resistance =0.1 ohm/square

EE 510 SPRING 2004 HOMEWORK 4

P3.10 Consider the following schematic:

The schematic drawn in LASI creates the following .CIR file:

*** SPICE Circuit File of 3_10_SCH

A bigger pulse, such as 50uA results in a significant 15% change:

Sol: - As stated in the book Table 3.2 is given as below:-

Layout of single pad & pad frame is as shown below:-

You might also like