5. Blown up
3. Trench isolation
3.1 Definition of shallow trench
3.2 Definition of deep trench isolation
3.3 Comparison of STI & LOCOS
3.4 Shallow trench isolation
3.5 CMP for STI
3.6 Deep trench isolation
3.7 Fabrication process of deep trench isolation
4. Silicon on insulator techniques(SOI)
4.1 Definition of SOI
4.2 Industrial need of SOI
4.3 SOI techniques
4.4 Methods of SOI isolation
4.5 Dielectric isolation
4.6 Wafer bonding
6. Blown up
5. Metallization
5.1 Definition
5.2 Multilevel metallization
5.3 Interconnection material
5.4 Metal requirement
5.5 Junction spiking
5.6 Stress migration
5.7 Electromigration
6. Planarization
6.1 Definition
6.2 Process of planarization
6.3 Working principle
7. NMOS IC technology
7.1 Definition & basic concepts
7.2 Fabrication process for NMOS
7.3 Advantages/Disadvantages of NMOS
7. Blown up
8. CMOS IC technology
8.1 Basic concepts
8.2 Fabrication process for CMOS
8.3 Advantages/Disadvantages
9. Bipolar IC technology
9.1 Basic concepts
9.2 Steps of fabrication
9.3 Advantages and limitations
10. Fault diagnosis & characterization technique
10.1 Basic concepts
10.2 Combinational fault diagnosis methods
10.3 Sequential fault diagnosis methods
10.4 Characterization techniques
8. BASIC DEFINITIONS
1. VLSI Very-large-scale integration (VLSI) is the process of
creating an integrated circuit (IC) by combining
thousands of transistors into a single chip.
2. Device isolation It is the ability of the technology to allow each device to
operate independently of the state of the other. Unless the
technology is limited to discrete devices.
3. LOCOS LOCOS, short for LOCal Oxidation of Silicon, is
a microfabrication process where silicon dioxide is formed in
selected areas on a silicon wafer having the Si-SiO2 interface
at a lower point than the rest of the silicon surface.
4. Trench isolation A trench is a type of excavation or depression in the ground
that is generally deeper than it is wide (as opposed to a
wider gully, or ditch), and narrow compared to its length (as
opposed to a simple hole).
5. Shallow trench
isolation
Shallow trench isolation (STI), also known as box isolation
technique, is an integrated circuit feature which prevents
electric current leakage between adjacent semiconductor
device components. STI is generally used on CMOS process
9. 6. Deep trench
isolation
It is an integrated circuit feature which prevents electric
current leakage between adjacent semiconductor
device components using trenches of fixed width.
7. Silicon on
insulator
Silicon on insulator (SOI) technology refers to the use of a layered
silicon–insulator–silicon substrate in place of conventional silicon
substrates in semiconductor manufacturing, especially
microelectronics, to reduce parasitic device capacitance, thereby
improving performance.
8. Dielectric
isolation
Dielectric isolation, is the process of electrically isolating various
components in the IC chip from the substrate and from each other by
an insulating layer.
9. Metallization Metallization is the process that makes accessible the IC to the outside
world through conducting pads
10. Junction
spiking
The penetration of a junction by aluminum, which occurs when
silicon near the junction dissolves in aluminum and migrates
along the interconnect lines. Aluminum then replaces silicon at
the junction.
10. 11. Stress migration Stress migration is a failure mechanism that often occurs
in integrated circuit metallization (aluminum, copper).
Voids form as result of vacancy migration driven by
the hydrostatic stress gradient. Large voids may lead to
open circuit or unacceptable resistance increase that
impedes the IC performance.
12. Electromigration Electromigration is the transport of material caused by
the gradual movement of the ions in a conductor due to
the momentum transfer between conducting electrons
and diffusing metal atoms.
13. Planarization Chemical mechanical polishing/planarization is a process
of smoothing surfaces with the combination of chemical and
mechanical forces. It can be thought of as a hybrid of chemical
etching and free abrasive polishing.
14. NMOS IC
technology
N-type metal-oxide-semiconductor logic uses n-type field
effect transistors to implement logic gates and other digital
circuits.
15. Diagnostic resolution A unit under test (UUT) fails when its observed behavior is
different from its expected behavior. Diagnosis consists of
locating the physical fault(s) in a structural model of the UUT.
The degree of accuracy to which faults can be located is called
diagnostic resolution.
11. 16. Functionally
equivalent faults
Functionally equivalent faults (FEF) cannot be distinguished. The
partition of all faults into distinct subsets of FEF defines
the maximal fault resolution. A test that achieves the maximal fault
resolution is said to be a complete fault-location test.
17. Fault table A fault table is a matrix where columns Fj represent faults,
rows Ti represent test patterns, and aij = 1 if the test pattern
Ti detects the fault Fj, otherwise if the test pattern Ti does not
detect the fault Fj, aij = 0.
18. Fault dictionaries Fault dictionaries (FD) contain the same data as the fault tables
with the difference that the data is reorganized. In FD a mapping
between the potential results of test experiments and the faults is
represented in a more compressed and ordered form.
19. Adaptive testing In sequential fault diagnosis the process of fault location is carried
out step by step, where each step depends on the result of the
diagnostic experiment at the previous step. Such a test experiment
is called adaptive testing.
20. Diagnostic tree Sequential experiments can be carried out either by observing only
output responses of the UUT or by pinpointing by a special probe
also internal control points of the UUT (guided probing).
Sequential diagnosis procedure can be graphically represented
as diagnostic tree.
12. VLSI
DEFINITION: Very-large-scale integration (VLSI) is the
process of creating an integrated circuit (IC) by combining
thousands of transistors into a single chip.
https://www.youtube.com/watch?v=Q5paWn7bFg4&list=PLJqog
rYlIWSsCOGWb5MmwohySBWrM2ofm
14. DEVICE ISOLATION
Definition: It is the ability of the technology to allow each device to
operate independently of the state of the other. Unless the
technology is limited to discrete devices.
Isolation Techniques:
Diffusion isolation with reverse biased diodes:
• Historically used for bipolar
• Currently used to isolate NMOS from PMOS through a well.
Oxide isolation:
• Used in early days of MOS
• Field can’t be implanted for parasitic transistor Vt control
• Step height is too much
15. CONTD..
Local oxidation of silicon (LOCOS)
• Main method used today in a variety of forms e. g., semi-recessed
LOCOS, fully-recessed LOCOS, SWAMI, poly-buffered LOCOS.
• 0.6 µm pitch: LOCOS limit if thick (>300 nm) field oxides are
required. 0.4 µm pitch with recessed LOCOS (200 nm field oxide)
Trench isolation
• Cutting edge technology today
16. JUNCTION AND OXIDE ISOLATION
There are several options for device isolation in Si and GaAs
technologies and evaluate each according to the density, yield,
planarity and parasitic effects.
Steps for construction of Bipolar ICs:
1. Diffuse a deep p layer and a shallower N+ layer into an n type
substrate.
2. The substrate acts as the common collector.
3. An insulating layer is then deposited on substrate, contacts are
patterned and etched and the interconnect is applied and patterned.
https://www.youtube.com/watch?v=slxAEd
mgODg
19. CONTD..
4. Emitter isolation: Emitter of two adjacent transistors are
automatically isolated from each other by the fact that each emitter
is totally enclosed in the base diffusion.
5. Base layer isolation: for base layer to be isolated from each other,
there must be a large energy barrier between the holes in the two
base regions.
As the height of the barrier decreases, the leakage between the
bases increases exponentially.
Measure of isolation:
Depletion layer associated with the two base collector junction do
no touch.
20. CONTD..
For a simple one sided step junction, the depletion layer thickness is:
Where,
= the relative permittivity of Si
= built in voltage of junction
= intrinsic carrier concentration
D
CBbis
D
qN
VVk
W
)(2 0
sk
biV
2
ln
i
DA
bi
n
NN
q
kT
V
in
21. Simple calculation of average isolation
distance required between transistors as a
function of device density
22. CONTD..
Isolation length: to get an estimate for the isolation length required
for a technology, consider the densest portion of the circuit.
According to the graph displayed in previous slide; it reads off the
maximum permissible isolation distance.
If the densest portion of the circuit is transistors/sqcm and 50% of
the area is active in that region, the device separation must be 10µm
or less.
For a random logic circuit, the isolation distance should be less than
twice the first metal pitch.
One more aspect about the graph is that the circuit density is a
sensitive function of the isolation distance
Reducing the isolation distance by a factor of 3 allows an order of
magnitude increase in the circuit density.
5
10
23. CONTD..
Steps for construction of MOSFET:
1. Add an insulating layer and a metal line that happens to pass over
on the bipolar transistor, but does not contact both transistors.
2. The collector of the two transistors acts as the source and drain of
the MOSFET, and the metal line acts as gate.
3. If a large positive bias exists on the line, the surface under the line
may invert, turning on the parasitic MOSFET.
4. This effect will short out the two collectors, even if they are
sufficiently far apart so that the depletion regions do not touch.
24. Cross section of simple bipolar technology with a
metal line crossing the junction isolation region,
forming a parasitic MOSFET
25. CONTD..
The MOSFET threshold for NMOS transistors on p type substrate is
given by:
Where:
= the metal semiconductor work function
0
4
2
s
fA
ox
oxs
fmsT
k
qN
k
tk
V
i
A
f
n
N
q
kT
ln
ms
26. CONTD..
Since the intrinsic carrier concentration increases with temperature,
the parasitic threshold voltage may shift by several volts when
operating at high temperatures.
The source to drain current of an MOS device increases
exponentially with gate voltage in the sub-threshold regime.
We need to make the threshold of these parasitic transistors at least 2
times the supply voltage.
It ensures that they will not turn on, even in the presence of an
excessive supply voltage and voltage spikes on the supply line, nor
will the leakage currents be excessive.
27. as a function of the assuming no . Solid
lines are a perfect interface, dotted lines are for
TV AN 0ms
211
10
cmNit
28. CONTD..
The graph shows a plot of threshold voltage versus the substrate
concentration with oxide thickness as a parameter.
He oxide thickness was varied from 0.2 to 1.0 µm in 0.2µm
increments.
Also included is the threshold voltage if a total fixed charge of
is present.
To achieve a suitably large parasitic threshold, one must select a
thick field oxide and a large substrate concentration.
The large substrate concentration degrades the performance due to
junction capacitance.
The thicker oxide improves both performance and parasitic turn on,
but obviously due to effect of oxide charge, some substrate doping is
required even for very thick field oxide.
211
10
cm
30. CONTD..
The device separation can be reduced and the parasitic threshold
increased by adding a p+ diffused barrier ring around each device.
Guard rings require an additional masks that must be aligned to the
transistor.
The guard ring must be deep(at least 2µm) or the depletion region
will simply extend beneath the guard ring, shorting out the devices.
The large thermal cycle needed to produce such a deep junction
must be done early in process.
31. ADVANTAGES
Junction isolation is simple and produce a planar isolation.
It have a high yield .
Density is not large.
Increasing the substrate concentration increases the density but also
increases the capacitance.
Guard rings improves the prevention of turning on parasitic MOS
devices.
It is cheap.
32. LOCOS
Definition: LOCOS, short for LOCal Oxidation of Silicon, is
a microfabrication process where silicon dioxide is formed in
selected areas on a silicon wafer having the Si-SiO2 interface at a
lower point than the rest of the silicon surface.
https://www.youtube.com/watch?v=9IbID_a
jhak
34. LOCOS DEFINED
LOCOS = LOCal Oxidation of Silicon
Defines a set of fabrication technologies where
the wafer is masked to cover all active regions
thick field oxide (FOX) is grown in all non-active regions
Used for electrical isolation of CMOS devices
Relatively simple to understand so often used to introduce/describe
CMOS fabrication flows
Not commonly used in modern fabrication
other techniques, such as Shallow Trench Isolation (STI) are
currently more common than LOCOS
36. LOCOS –STEP 1
Form N-Well regions
Grow oxide
Deposit photoresist
Pattern photoresist
NWELL Mask
expose only n-well
areas
Layout view
Cross section view
p-type substrate
NWELL mask
NWELL mask
oxide photoresist
37. LOCOS –STEP 1
Form N-Well regions
Grow oxide
Deposit photoresist
Pattern photoresist
NWELL Mask
expose only n-well areas
Etch oxide
Remove photoresist
Layout view
Cross section view
p-type substrate
oxide
38. LOCOS –STEP 1
Form N-Well regions
Grow oxide
Deposit photoresist
Pattern photoresist
NWELL Mask
expose only n-well areas
Etch oxide
Remove photoresist
Diffuse n-type dopants
through oxide mask layer
Layout view
Cross section view
p-type substrate
n-well
39. LOCOS –STEP 2
Form Active Regions
Deposit SiN over wafer
Deposit photoresist over
SiN layer
ACTIVE mask
ACTIVE mask
SiN photoresist
p-type substrate
n-well
40. LOCOS –STEP 2
Form Active Regions
Deposit SiN over wafer
Deposit photoresist over
SiN layer
Pattern photoresist
*ACTIVE MASK
ACTIVE mask
ACTIVE mask
SiN photoresist
p-type substrate
n-well
41. LOCOS –STEP 2
Form Active Regions
Deposit SiN over wafer
Deposit photoresist over
SiN layer
Pattern photoresist
*ACTIVE MASK
Etch SiN in exposed areas
leaves SiN mask which
blocks oxide growth
ACTIVE mask
SiN photoresist
p-type substrate
n-well
42. LOCOS –STEP 2
Form Active Regions
Deposit SiN over wafer
Deposit photoresist over
SiN layer
Pattern photoresist
*ACTIVE MASK
Etch SiN in exposed areas
leaves SiN mask which
blocks oxide growth
Remove photoresist
Grow Field Oxide (FOX)
thermal oxidation
ACTIVE mask
p-type substrate
n-well
FOX
43. LOCOS –STEP 2
Form Active Regions
Deposit SiN over wafer
Deposit photoresist over SiN layer
Pattern photoresist
*ACTIVE MASK
Etch SiN in exposed
areas
leaves SiN mask which
blocks oxide growth
Remove photoresist
Grow Field Oxide (FOX)
thermal oxidation
Remove SiN
ACTIVE mask
p-type substrate
n-well
FOX
44. LOCOS –STEP 3
Form Gate (Poly layer)
Grow thin Gate Oxide
over entire wafer
negligible effect
on FOX regions
gate oxide
45. LOCOS –STEP 3
Form Gate (Poly layer)
Grow thin Gate Oxide
over entire wafer
negligible effect on
FOX regions
Deposit Polysilicon
Deposit Photoresist
gate oxide
POLY mask
POLY mask
polysilicon
46. LOCOS –STEP 3
Form Gate (Poly layer)
Grow thin Gate Oxide
over entire wafer
negligible effect on
FOX regions
Deposit Polysilicon
Deposit Photoresist
Pattern Photoresist
*POLY MASK
Etch Poly in exposed
areas
Etch/remove Oxide
gate protected by
poly
gate oxide
POLY mask
POLY mask
47. LOCOS –STEP 3
Form Gate (Poly layer)
Grow thin Gate Oxide
over entire wafer
negligible effect on
FOX regions
Deposit Polysilicon
Deposit Photoresist
Pattern Photoresist
*POLY MASK
Etch Poly in exposed
areas
Etch/remove Oxide
gate protected by
poly
gate oxide
48. LOCOS –STEP 4
Form pmos S/D
Cover with photoresist
PSELECT mask
PSELECT mask
55. LOCOS –STEP 6
Form Contacts
Deposit oxide
Deposit photoresist
Pattern photoresist
*CONTACT Mask
One mask for both
active and poly
contact shown
CONTACT mask
p+p+ p+
n
n+ n+ n+
CONTACT mask
56. LOCOS –STEP 6
Form Contacts
Deposit oxide
Deposit photoresist
Pattern photoresist
*CONTACT Mask
One mask for both
active and poly
contact shown
Etch oxide
p+p+ p+
n
n+ n+ n+
57. LOCOS –STEP 6
Form Contacts
Deposit oxide
Deposit photoresist
Pattern photoresist
*CONTACT Mask
One mask for both
active and poly
contact shown
Etch oxide
Remove photoresist
Deposit metal
immediately after
opening contacts so
no native oxide
grows in contacts
Planerize
make top level
p+p+ p+
n
n+ n+ n+
58. LOCOS –STEP 7
Form Metal 1 Traces
Deposit photoresist
p+p+ p+
n
n+ n+ n+
METAL1 mask
METAL1 mask
59. LOCOS –STEP 7
Form Metal 1 Traces
Deposit photoresist
Pattern photoresist
*METAL1 Mask p+p+ p+
n
n+ n+ n+
METAL1 mask
METAL1 mask
60. LOCOS –STEP 7
Form Metal 1 Traces
Deposit photoresist
Pattern photoresist
*METAL1 Mask
Etch metal
p+p+ p+
n
n+ n+ n+
metal over poly outside of cross section
61. LOCOS –STEP 7
Form Metal 1 Traces
Deposit photoresist
Pattern photoresist
*METAL1 Mask
Etch metal
Remove photoresist
p+p+ p+
n
n+ n+ n+
62. LOCOS –STEP 8
Form Vias to Metal 1
Deposit oxide
Planerize oxide
Deposit photoresist
p+p+ p+
n
n+ n+ n+
VIA mask
VIA mask
63. LOCOS –STEP 8
Form Vias to Metal1
Deposit oxide
Planerize
Deposit photoresist
Pattern photoresist
*VIA Mask
p+p+ p+
n
n+ n+ n+
VIA mask
VIA mask
68. LOCOS –STEP 9
Form Metal2 Traces
Deposit photoresist
Pattern photoresist
*METAL2 Mask
Etch metal
p+p+ p+
n
n+ n+ n+
69. LOCOS –STEP 9
Form Metal2 Traces
Deposit photoresist
Pattern photoresist
*METAL2 Mask
Etch metal
Remove photoresist
p+p+ p+
n
n+ n+ n+
70. LOCOS –STEP 10+
Form Additional Traces
Deposit oxide
Deposit photoresist
Pattern photoresist
Etch oxide
Deposit metal
Deposit photresist
Pattern photoresist
Etch metal
Repeat for each
additional metal
p+p+ p+
n
n+ n+ n+
p-type substrate
71. SIMPLIFICATIONS FROM COMPLETE
PROCESS
Skipped several substrate doping steps:
channel implant to adjust threshold voltages
surface implant to increase breakdown voltage
no LDD, lightly-doped drain
no deposition of contact interface materials
metal patterning simplified
more complex “lift-off” process often used
no overglass (thick top dielectric) layer
no bonding pad layer
simplified use of dark/clear field masks and positive/negative
photoresist
72. TRENCH ISOLATION
Definition of Trench: A trench is a type of excavation or
depression in the ground that is generally deeper than it is wide (as
opposed to a wider gully, or ditch), and narrow compared to its
length (as opposed to a simple hole).
Trench isolation is a method used to prevent latch-up and
isolate transistors from each other.
Many new isolation process approaches have been developed around
the idea of etching away part of the substrate and refilling it with an
insulator.
These can be divided into two classes:
Shallow trench isolation
Deep trench isolation
73. TYPES OF TRENCH ISOLATION
Definition of Shallow trench isolation: Shallow
trench isolation (STI), also known as box isolation
technique, is an integrated circuit feature which prevents
electric current leakage between adjacent semiconductor
device components. STI is generally used
on CMOS process technology nodes of 250
nanometers and smaller.
Definition of Deep trench isolation: it is an integrated
circuit feature which prevents electric
current leakage between adjacent semiconductor
device components using trenches of fixed width.
75. ISOLATION OF TRANSISTORS
Isolation of Transistors :The use of reverse bias pn junctions to
isolate transistors becomes impractical as the transistor sizes
decrease.
76. USE OF SHALLOW TRENCH ISOLATION
TECHNOLOGY
Use of Shallow Trench Isolation Technology: Shallow trench
isolation (STI) allows closer spacing of transistors by eliminating the
depletion region at the surface.
77. COMPARISON OF STI AND LOCOS
If the n+ to p+ spacing is large, the Bird’s beak can be compensated
using techniques such as poly buffered LOCOS
At some point as the n+ to p+ spacing gets smaller, the restricted
bird’s beak leads to undesirable stress effects in the transistor.
An important advantage of STI is that it minimizes the heat cycle
needed for n+ or p+ isolation compared to LOCOS. This is a
significant advantage for any process where there are implants
before STI.
https://www.youtube.com/watch?v=MGNM
AKiXV4Y
79. THE SHALLOW TRENCH ISOLATION
(STI)
It is the preferred isolation technique for the sub-0.5m technology,
because it completely avoids the bird's beak shape characteristic.
With its zero oxide field encroachment STI is more suitable for the
increased density requirements, because it allows to form smaller
isolation regions.
https://www.youtube.com/watch?v=CllgoL
mICWo
81. CONTD..
The STI process starts in the same way as the LOCOS process. The
first difference compared to LOCOS is that a shallow trench is
etched into the silicon substrate, as shown in Fig. 1.2a.
After underetching of the oxide pad, also a thermal oxide in the
trench is grown, the so-called liner oxide (see Fig. 1.2c).
But unlike with LOCOS, the thermal oxidation process is stopped
after the formation of a thin oxide layer, and the rest of the trench is
filled with a deposited oxide (see Fig. 1.2d).
Next, the excessive (deposited) oxide is removed with chemical
mechanical planarization. At last the nitride mask is also removed.
The price for saving space with STI is the larger number of different
process steps.
82.
83. SHALLOW TRENCH ISOLATION (STI)
1. Cover the wafer with pad oxide and silicon
nitride.
2. First etch nitride and pad oxide. Next, an
anisotropic etch is made in the silicon to a depth
of 0.4 to 0.5 microns.
3. Grow a thin thermal oxide layer on the trench
walls.
4. A CVD dielectric film is used to fill the trench.
5. A chemical mechanical polishing (CMP) step is
used to polish back the dielectric layer until the
nitride is reached. The nitride acts like a CMP
stop layer.
6. Densify the dielectric material at 900°C and strip
the nitride and pad oxide.
84. CMP FOR STI
STI is the mainstream CMOS isolation technology
In STI, substrate trenches filled with oxide surround devices or group of
devices that need to be isolated
Relevant process steps:
Diffusion (OD) regions covered with nitride (acts as CMP-stop)
Trenches created where nitride absent and filled with oxide
CMP to remove excess oxide over nitride (overburden oxide)
Si
Oxide Nitride
Before CMP After Perfect CMP
CMP goal: Complete removal of oxide over nitride, perfectly planar nitride and
trench oxide surface
85. IMPERFECT CMP
Planarization window: Time window to stop CMP
Stopping sooner leaves oxide over nitride
Stopping later polishes silicon under nitride
Larger planarization window desirable
Step height: Oxide thickness variation after CMP
Quantifies oxide dishing
Smaller step height desirable
CMP quality depends on nitride and oxide density
Control nitride and oxide density to enlarge planarization window and to
decrease step height
Failure to clear oxide Nitride erosion Oxide dishing
Key Failures Caused by Imperfect CMP
86. CONCLUSIONS
Imperfect STI CMP causes functional and parametric yield loss
Our fill insertion approach focuses on:
(1) oxide density variation minimization,
(2) nitride density maximization
Large nitride fill features contribute to nitride and oxide densities, small ones to
nitride only shape fill to control both densities
Proposed max. nitride fill insertion with holes to control oxide density and
achieve high nitride density
Results indicate significant decrease in oxide density variation and increase in
nitride density over tile-based fill
CMP simulation shows superior CMP characteristics, planarization window
increases by 17%, and step height decreases by 9%
87. DEEP TRENCH ISOLATION
A high quality thermal oxide liner is grown
along the side-walls of the deep trench and the
remaining oxide is deposited at low
temperature.
A polysilicon stress-relief layer is deposited so
that it fills the deep trench and is recessed
below the silicon surface.
The deep trench process integration is
designed to minimize the impact on the
shallow trench isolation module which is used
for logic isolation.
Hence, maximum re-use of shallow trench
process steps becomes critical.
88. CONTD..
The deep trench oxide thickness is sufficient to sustain a breakdown
voltage from nwell to substrate of 74V.
This breakdown voltage has been found to be stable even with
repeated stress, which is made possible by interface between the
deep trench silicon sidewall and the high quality thermal oxide liner.
The deep trench in this technology is used to significantly increase
the analog packing density by bringing devices adjacent to each
other across the deep trench, thereby enabling shrinks ranging from
50% for medium-voltage analog to >80% for high-voltage analog
components compared to the 0.35µm SMOS7 technology.
In SMOS7, high-energy implant chains replaced the deep diffused
implants, which were present in older diffusion based technologies
89. FABRICATION PROCESS
1. Take a standard LOCOS structure.
2. After nitride patterning trenches are etched.
3. It must have smooth walls at no more than 85º with respect the
plane of a wafer.
4. Trench etch is made by depositing while etching silicon
anisotropically. This will create a small cusp of at the top of
the trench.
5. The thickness of this cusp increase with time, producing the
desired taper.
6. A thin local oxidation is done by using thinner oxides to increase
capacitance.
7. Finally a layer of polysilicon is deposited etched back.
2SiO
2SiO
90.
91. CONTD..
we have developed and qualified a 0.25µm CMOS based high-side
capable 70V smart power process on a P++ substrate with a deep
trench high-voltage isolation and logic shallow trench isolation.
By using a deep trench combined with a P++ substrate, we have
realized significant analog shrink, reduction of substrate parasitics
and 74V high side capability without affecting analog matching and
process complexity
92. SILICON ON INSULATOR ISOLATION
TECHNIQUES(SOI)
Definition: Silicon on insulator (SOI) technology refers to the use
of a layered silicon–insulator–silicon substrate in place of
conventional silicon substrates in semiconductor manufacturing,
especially microelectronics, to reduce parasitic device capacitance,
thereby improving performance.
SOI-based devices differ from conventional silicon-built devices in
that the silicon junction is above an electrical insulator,
typically silicon dioxide or sapphire (these types of devices are
called silicon on sapphire, or SOS).
93. CONTD..
To completely encase each device in an insulating material
The choice of insulator depends largely on intended application,
with sapphire being used for high-performance radio frequency (RF)
and radiation-sensitive applications, and silicon dioxide for
diminished short channel effects in microelectronics devices.
94. INDUSTRY NEED OF SOI
Lower parasitic capacitance due to isolation from the bulk silicon,
which improves power consumption at matched performance.
Resistance to latch up due to complete isolation of the n- and p-well
structures.
Higher performance at equivalent VDD. Can work at low VDD's.
Reduced temperature dependency due to no doping.
Better yield due to high density, better wafer utilization.
Reduced antenna issues
No body or well taps are needed.
Lower leakage currents due to isolation thus higher power
efficiency.
Inherently radiation hardened ( resistant to soft errors ), thus
reducing the need for redundancy
95. SOI TECHNIQUE
1. In this process, a thin layer of single-crystal silicon can be
produced on top of a thermal SiO2 layer on a silicon wafer.
2. Strips of oxide are produced by patterning the oxide layer using
photolithography.
3. a thin layer of silicon is then deposited on the wafer.
4. It will be polycrystalline in the regions where the deposited silicon
layer overlays the oxide and it will be single crystal in the regions
where there is direct contact with silicon substrate.
5. In the next step we will directionally recrystalise the silicon layer,
which in turn recrystallises the substrate to act as the nucleation
centre.
6. As the heated zone is scanned across the wafer the crystal growth,
propagates from these nucleation regions to the regions of the
silicon film on top of the oxide islands or strips.
7. Thus we form a complete single crystal layer of silicon.
96. METHODS OF SOI ISOLATION
1. Dielectric isolation
2. Wafer bonding
97. DIELECTRIC ISOLATION
Definition: Dielectric isolation, is the process of electrically
isolating various components in the IC chip from the substrate and
from each other by an insulating layer.
It's main use is to eliminate undesirable parasitic junction
capacitance or leakage currents associated with certain applications.
It is used to build high voltage telecommunication ICs that required
electrically isolated bidirectional switches.
98.
99. V-GROOVE ISOLATION
1. V-groove isolation is formed with an n-type substrate, on which an
n+ diffusion is performed.
2. An SiO2 layer is formed, which is then patterned to form a grid of
intersecting lines opening in the oxide
3. The wafer formed is then exposed to an orientation dependent
etching (ODE) process, where the patterned layer is used as the
etching mask; which results in the formation of V-shaped grooves
as shown in the picture (b).
4. In this the <111> plane sidewalls are at an angle of 54.74 degree
with respect to the <100> top surface of the silicon wafer.
101. CONTD..
5. As a result the starting material is <111> oriented crystal, which is
normally used for p-n junction isolation. But for dielectric isolation
the starting material is <100> oriental silicon.
6. The etchant used in the above step etches away the exposed silicon
anisotropically, this means that the etch rate is much faster along
the <111> planes than along the <100> crystal planes.
7. This kind of preferential etching is the key reason behind the
formation of V-groove.
102. CONTD..
8. The depth D of the isolation groove can be determined in the
initial oxide cut width W as
9. Cover the sidewalls of the V-groove with an oxide layer, therefore
the wafer is subjected to a thermal oxidation process.
10. After completing the oxide layer, a very thick layer of
polycrystalline silicon is deposited as shown in picture (c).
2
W
D
103. CONTD..
11. The most critical step in the V-groove isolation process is
explained in figure (d).
12. Keeping polycrystalline surface side of the wafer down, silicon
wafers are mounted on the lapping plate.
13. In the next step, n-type silicon substrate is then carefully lapped
down to the level at which the vertices of the V-grooves become
exposed.
14. So now we get an array of n-type single crystal silicon regions that
are isolated from the polycrystalline silicon substrate.
15. Polycrystalline silicon now serves to provide the mechanical
support for the IC.
16. This material is ideal for the function because of its good thermal
expansion coefficient, it can withstand high processing
temperatures, and is a good match to single crystal silicon.
104. CONTD..
17. The n-type silicon has now moved down to vertices of the V-
grooves because of the lapping operation.
18. If the lapping is recessive, then proper isolation will not be
achieved.
19. But if excessive lapping is done, it may lead to thinner n-type
regions.
20. Wafer diameter is approx 100mm and the V-groove depth is about
10 micro meters, thus precise lapping is necessary.
21. The n+ diffused layer serves as a buried layer to reduce the
collector series resistance of the n-p-n transistors.
22. The rest of the processing sequence for the dialectically isolated
ICs follows along the same line as for the conventional junction
isolated IC.
105. ADVANTAGES OF DIELECTRIC
ISOLATION
Permittivity of SiO2 is one reason, which is 1/3rd of Silicon and
hence capacitance is reduced.
Oxide is thicker than the depletion region of the substrate junction
and capacitance is inversely proportional to the thickness of oxide.
No need of applying negative potential to the substrate.
106. DRAWBACKS OF DIELECTRIC ISOLATION
The wafer are not as planar as normal starting material.
Wafers made by this process are expensive.
107. WAFER BONDING
In this process two wafers
are pressed together at high
temperature until they fuse.
The wafers are fused at low
temperature by anodic
bonding.
If the wafers are oxidized
before bonding, a layer of
oxide remains at the centre
of the fused wafer.
108. CONTD..
The wafer can be ground back to thickness of 2 to 3µmusing
standard grinding and polishing techniques.
If thinner layers are required, additional processing can be done to
produce submicrometer semiconductor films on top of the oxide.
Device can be isolated with a simple etch process that produces
single crystal islands on top of the insulating oxide.
109. METALLIZATION
Definition: Metallization is the process that makes accessible the IC
to the outside world through conducting pads.
Doped silicon conduct electricity but have large resistance and lack
interconnecting facility.
Thin conductive metal films (Al, Cu, Au, Ag etc) are used as
interconnects between Si and external leads.
110. WHY INTERCONNECT STRUCTURES
ARE IMPORTANT?
Rough Estimation of Interconnect RC Time Delay
As technology progresses, Ls decreases RC delay increases.
To decrease RC delay - ρ, ε, L should take low values
111. NEEDS OF NEW TECHNOLOGY
Lower resistivity metal for interconnect wiring.
Lower dielectric constant material for the interlayer dielectric.
Smaller wire lengths-Multilevel Metallization
112. MULTI LEVEL METALLIZATION
Metal interconnections.
Span several planes.
Isolated by the insulating dielectric layers
Interconnected by the wiring in the third dimension through the
holes in the dielectric planes.
Three dimensional network of interconnections is given the name
multilevel interconnections.
116. METALS REQUIREMENTS
Low resistivity
Easy to deposit
Easy to etch and planarize
High melting point
High electromigration resistance
Mechanical stability, adherence to interlayer dielectrics and
other materials on chip
Substrate matched coefficient of thermal expansion
Low stress, high stress migration resistance
https://www.youtube.com/watch?v=8wYI7E
FAeMw
118. CONTD..
Controlled microstructure
Preferably uniform large grains and smooth surfaces
Oxidation/corrosion resistance
Low chemical reactivity
Ideally passivates itself
Compatible with surrounding materials and their processing
Bondable to wirings in package
Environmentally safe material during processing and actual use, and
recyclable
Reliable
Low cost
119. Property/ metal Cu Ag Au Al w
Resistivity 1.67 1.59 2.35 2.66 5.65
Youngs modulus 12.98 8.27 7.85 7.06 41.1
Thermal conductivity 3.98 4.25 3.15 2.358 1.74
Coeff. Of thermal
expansion CTE
17 19.1 14.2 23.5 4.5
M.P(ºC) 1085 962 1064 660 3387
Specific heat capacity 38 234 132 917 138
Corrosion in air Poor Poor Excellent Good Good
Adhesion to Sio2 Poor Poor Poor Good Good
Delay 2.3 2.2 3.2 3.7 7.8
Thermal stress per
degree for films on Si
2.5 1.9 1.2 2.1 0.8
120. ALUMINUM
Early ICs used pure Al as the interconnect material
Low resistivity
Strong adhesion with Si
Corrosion resistant
Problems with pure Al
Junction spiking
Electromigration
Stress migration
Later ICs used Al alloyed with Cu
121. JUNCTION SPIKING
Definition of junction spiking: the penetration of
a junction by aluminum, which occurs when silicon near
the junction dissolves in aluminum and migrates along the
interconnect lines. Aluminum then replaces silicon at
the junction.
Consider Al-Si contact
Solubility of Si in Al is 0. 5 wt% at 4500C
•Si will dissolve into the Al during annealing (at 4500C)
Solution
Add Si to the Al
Introduce a barrier metal layer between the
Al and the Si substrate.
(TiN)
122. STRESS MIGRATION
Definition: Stress migration is a failure mechanism that often
occurs in integrated circuit metallization (aluminum, copper).
Voids form as result of vacancy migration driven by
the hydrostatic stress gradient. Large voids may lead to open
circuit or unacceptable resistance increase that impedes the
IC performance.
Due to difference between coefficient of thermal expansion for Al
and Si.
Al – 23 x 10-6 0C-1 and Si – 2.6 x 10-6 0C-1
High compressive stresses in Al at high temperatures
123. CONTD..
Movement of Al occurs along grain boundaries
Whole grains of Al pushed upward forming hillocks
Under tensile stress voids are formed
Consequences
• Electrical shorts between interconnect levels
• Rough surface topography making lithography and etch difficult
Solution
Addition of elements that have limited solubility
Ex:- Cu atoms segregate and precipitate preferentially along the
grain
boundaries suppressing hillock formation
124. ELECTROMIGRATION
Definition: Electromigration is the transport of
material caused by the gradual movement of the
ions in a conductor due to the momentum transfer
between conducting electrons and diffusing metal
atoms.
125. ELECTROMIGRATION
Transport of mass in metals under the influence of high current
Occurs by transfer of momentum from electrons to the positive
metal ions
High current densities in the smaller devices are responsible for
electron migration
Grain boundary diffusion is the primary vehicle of mass transport
Metal in some regions pile up and voids form in other regions
126. SOLUTIONS - ELECTROMIGRATION
Alloying with copper (Al with 0.5%Cu)
• Multilayer structure
– Shunt layer provides alternative path for current flow
– If shunt layer has high melting point and strong mechanical
properties,
they can be more rigid and act as barrier to hillock and void
formation
127. PROPERTIES OF METALS
Tungsten Gold copper
Good corrosion resistance Low resistivity Higher conductivity
Electromigration and
stress migration stability
Very inert More electromigration
resistance
Excellent deposition
methods
Adheres poorly Higher ultimate tensile
strength
Sometimes used for filling
of vias called plugs
Very costly Higher melting poit, low
CTE
High resistivity High thermal conductivity
Poor adhesion High specific heat
128. PLANARIZATION
DEFINITION: Chemical mechanical polishing/planarization is a
process of smoothing surfaces with the combination of chemical and
mechanical forces. It can be thought of as a hybrid of chemical
etching and free abrasive polishing.
129. PROCESS OF PLANARIZATION
The process uses an abrasive and corrosive chemical slurry in
conjunction with a polishing pad and retaining ring, typically of a
greater diameter than the wafer.
The pad and wafer are pressed together by a dynamic polishing
head and held in place by a plastic retaining ring.
The dynamic polishing head is rotated with different axes of
rotation .
This removes material and tends to even out any irregular
topography, making the wafer flat or planar.
https://www.youtube.com/watch?v=2z4lq-Ms_OU
131. CONTD..
This may be necessary to set up the wafer
for the formation of additional circuit
elements.
For example, CMP can bring the entire
surface within the depth of field of
a photolithography system, or selectively
remove material based on its position.
Typical depth-of-field requirements are
down to Angstrom levels for the latest
22 nm technology.
132. WORKING PRINCIPLE
Physical action
Typical CMP tools, such as the ones seen on the right, consist of a
rotating and extremely flat platen which is covered by a pad.
The wafer that is being polished is mounted upside-down in a
carrier/spindle on a backing film.
The retaining ring (Figure 1) keeps the wafer in the correct
horizontal position.
During the process of loading and unloading the wafer onto the tool,
the wafer is held by vacuum by the carrier to prevent unwanted
particles from building up on the wafer surface.
133. CONTD..
A slurry introduction mechanism deposits the slurry on the pad,
represented by the slurry supply in Figure 1.
Both the platen and the carrier are then rotated and the carrier is kept
oscillating; this can be better seen in the top view of Figure 2.
A downward pressure/down force is applied to the carrier, pushing it
against the pad; typically the down force is an average force, but
local pressure is needed for the removal mechanisms.
Down force depends on the contact area which, in turn, is dependent
on the structures of both the wafer and the pad.
134. CONTD..
Typically the pads have a roughness of 50 µm; contact is made by
asperities (which typically are the high points on the wafer) and, as a
result, the contact area is only a fraction of the wafer area.
In CMP, the mechanical properties of the wafer itself must be
considered too.
If the wafer has a slightly bowed structure, the pressure will be
greater on the edges than it would on the center, which causes non-
uniform polishing.
In order to compensate for the wafer bow, pressure can be applied to
the wafer's backside which, in turn, will equalize the centre-edge
differences.
135. CONTD..
The pads used in the CMP tool should be rigid in order to uniformly
polish the wafer surface.
However, these rigid pads must be kept in alignment with the wafer
at all times.
Therefore, real pads are often just stacks of soft and hard materials
that conform to wafer topography to some extent.
Generally, these pads are made from porous polymeric materials
with a pore size between 30-50 µm, and because they are consumed
in the process, they must be regularly reconditioned.
In most cases the pads are very much proprietary, and are usually
referred to by their trademark names rather than their chemical or
other properties.
136. CHEMICAL ACTION
Before about 1990 CMP was
viewed as too "dirty" to be included
in high-precision fabrication
processes, since abrasion tends to
create particles and the abrasives
themselves are not without
impurities.
Since that time, the integrated
circuit industry has moved
from aluminium to copper conduct
ors.
137. CONTD..
This required the development of an additive patterning process,
which relies on the unique abilities of CMP to remove material in a
planar and uniform fashion and to stop repeatably at the interface
between copper and oxide insulating layers.
Adoption of this process has made CMP processing much more
widespread.
In addition to aluminum and copper, CMP processes have been
developed for polishing tungsten, silicon dioxide, and (recently)
carbon nanotubes.
138. LIMITATIONS OF CMP
There are currently several limitations of CMP that appear during
the polishing process requiring optimization of a new technology.
In particular, an improvement in wafer metrology is required.
In addition, it was discovered that the CMP process has several
potential defects including stress cracking, delaminating at weak
interfaces, and corrosive attacks from slurry chemicals.
The oxide polishing process, which is the oldest and most used in
today's industry, has one problem: a lack of end points requires blind
polishing, making it hard to determine when the desired amount of
material has been removed or the desired degree of planarization has
been obtained.
139. CONTD..
If the oxide layer has not been sufficiently thinned and/or the desired
degree of planarity has not been achieved during this process, then
(theoretically) the wafer can be repolished, but in a practical sense
this is unattractive in production and is to be avoided if at all
possible.
If the oxide thickness is too thin or too non-uniform, then the wafer
must be reworked, an even less attractive process and one that is
likely to fail. Obviously, this method is time-consuming and costly
since technicians have to be more attentive while performing this
process.
140. NMOS IC TECHNOLOGY
Definition: N-type metal-oxide-semiconductor logic uses n-
type field effect transistors to implement logic gates and other digital
circuits.
These nMOS transistors operate by creating an inversion layer in
a p-type transistor body.
This inversion layer, called the n-channel, can
conduct electrons between n-type "source" and "drain" terminals.
The n-channel is created by applying voltage to the third terminal,
called the gate.
Like other MOSFETs, nMOS transistors have four modes of
operation: cut-off (or subthreshold), triode, saturation (sometimes
called active), and velocity saturation.
141. P-N JUNCTIONS
A junction between p-type and n-type semiconductor forms a diode.
Current flows only in one direction.
p-type n-type
anode cathode
142. NMOS TRANSISTOR
Four terminals: gate, source, drain, body
Gate – oxide – body stack looks like a capacitor
Gate and body are conductors
SiO2 (oxide) is a very good insulator
Called metal – oxide – semiconductor (MOS) capacitor
Even though gate is
no longer made of metal
n+
p
GateSource Drain
bulk Si
SiO2
Polysilicon
n+
143. NMOS OPERATION
Body is commonly tied to ground (0 V)
When the gate is at a low voltage:
P-type body is at low voltage
Source-body and drain-body diodes are OFF
No current flows, transistor is OFF
n+
p
GateSource Drain
bulk Si
SiO2
Polysilicon
n+
D
0
S
https://www.youtube.com/watch?v=0FwfSMxBU3s
145. CONTD..
When the gate is at a high voltage:
Positive charge on gate of MOS capacitor
Negative charge attracted to body
Inverts a channel under gate to n-type
Now current can flow through n-type silicon from source
through channel to drain, transistor is ON
n+
p
GateSource Drain
bulk Si
SiO2
Polysilicon
n+
D
1
S
146. NMOS IC TECHNOLOGY
The basic n channel circuit consists of NMOS transistors.
MOS transistors consists of a source, a drain a and a gate region.
Each transistor is isolated form its neighbor transistors and other
devices by a thick field oxide.
Usually phosphorus doped SiO2 , called P – glass is used as
insulating layer.
Under this SiO2 layer, a thin layer of dopant is used, called Chan-
stop region.
The Chan-stop region serves to improve isolation between
transistors.
147. CONTD..
When +ve charge is applied at the gate, then –ve charge starts to
move from the source to drain.
Following points will be discussed.
Fabrication process sequence for NMOS
Special Considerations for NMOS ICs
148. FABRICATION PROCESS SEQUENCE
FOR NMOS
1. The starting Si wafer is a lightly doped p – type substrate.
2. First step is to oxidize the Si to form a layer of SiO2 .
153. CONTD..
Now deposit a layer of thin oxide in order to form gate oxide of the
NMOS transistor.
154. CONTD..
Now on the top of the thin oxide layer, a layer of polysilicon is
deposited. It is used as gate electrode material for MOS to
interconnect it.
155. CONTD..
After deposition of the polysilicon layer, it is patterned and etched to
form the interconnects and the MOS transistor gate.
156. CONTD..
The thin oxide not covered by polysilicon is also etched away so that
source and drain junctions may be formed.
157. CONTD..
The entire silicon surface is then doped with a high concentration of
impurities either by diffusion or ion implantation.
158. CONTD..
Once the source and drain regions are completed, the entire surface
is again covered with and insulating layer of SiO2 .
159. CONTD..
The insulating oxide is then patterned in order to provide
contact window for drain and source junctions.
160. CONTD..
The surface is now covered with evaporated aluminium which will
form the interconnections.
161. CONTD..
Finally the metal layer is patterned and etched, completing the
interconnections of the MOS transistor on the surface.
162. CONTD..
MOS transistors must be electrically isolated from each other during
fabrication.
Isolation is required to prevent unwanted conduction path between
devices.
163. ADVANTAGES OF NMOS TECHNOLOGY
Since electron mobility is twice (say) that of hole mobility, an n-
channel device will have one-half the on-resistance or impedance of
an equivalent p-channel device with the same geometry and under
the same operating conditions.
Thus n-channel transistors need only halt the size of p-channel
devices to achieve the same impedance. Therefore, n-channel ICs
can be smaller for the same complexity or, even more important,
they can be more complex with no increase in silicon area.
NMOS circuits offer a speed advantage over PMOS due to smaller
junction areas. Since the operating speed of an MOS IC is largely
limited by internal RC time constants and capacitance of diode is
directly proportional to its size, an n-channel junction can have
smaller capacitance. This, in turn, improves its speed.
164. DISADVANTAGES
The n-channel device has following problems in the device
processing.
Most of the mobile contaminants are positively charged.
Since NMOS operates with the gate positively based with
respect to the substrate, these ions collect along the oxide-
silicon interface. This charge causes a shift in VTh.
Also, there is fixed positive charge at the Si-SiO2 interface
resulting from various steps of the manufacturing process.
This also shifts the threshold voltage. Both these charges have
tendency to make the device normally on.
These two charges exist in PMOS device too, but the positive
ions are pulled to the AI-S1O2 interlace by the negative bias
applied to gate. There, they cannot affect the device threshold
severely.
165. CONTD..
Another problem with NMOS device occurs during the oxidation of
silicon which takes place at the Si-SiO2 interface.
No real abrupt change occurs between silicon and Si02; rather there
is a transition zone.
This transition zone contains positively charged Silicon atoms
which increase the absolute magnitude of the threshold voltage for a
p-channel device and decrease the absolute magnitude of the
threshold voltage for an n-channel device.
This means it is difficult to make an n-channel device that is off at
zero gate voltage. This is why it is more difficult to make an n-
channel device than a p-channel device.
166. WHAT IS THE ADVANTAGE OF CMOS OVER NMOS
?
CMOS is preffered over NMOS:
As CMOS propogates both logic '1', and '0', without a voltage drop
when using NMOS only, logic '1' (i.e Vdd) suffers a thresold drop
and the output after passing through one NMOS gate would be Vdd-
Vt(thresold voltage of the NMOS gate).
Hence signal margin is very important in NMOS causing possible
SI(signal integrity) issues.
Hence CMOS is preferred. By the way CMOS and NMOS and also
PMOS are all low powered. Static power consumption is the same,
dynamic power consumption depends on signal swing (i.e number of
times data line varies)
167. CMOS IC TECHNOLOGY
A CMOS inverter is realized by the series combination of a
PMOS and NMOS transistors.
Transfer characteristic of the CMOS inverter is output voltage as
a function of input voltage.
The circuit diagram of a CMOS inverter is shown on next slide.
The cross section of the inverter structure shows the n-channel
transistor formed in a p-region called tub or well.
The gates of the transistors are connected to from the input.
In order to understand the operation of the CMOS inverter,
define the threshold voltages of NMOS and PMOS transistors.
Let VTn = 1 V & VTp = -1 V and VDD = 5V.
https://www.youtube.com/watch?v=OBiu2agne_U
170. CONTD..
The operation of the CMOS inverter can be divided into 5 regions.
NMOS (OFF) will be in cutoff region.
PMOS (ON) will be in linear region.
So, V0 = VDD.
NMOS (ON) Saturation region.
PMOS (ON) Linear region
100:ARegions intnin VVV
5.21
2
V:BRegion tn in
DD
in V
V
V
171. CONTD..
NMOS (ON) Saturation region.
PMOS (ON) Saturation region.
PMOS (ON) saturation region.
NMOS (ON) Linear region.
5.2
2
V:CRegion in in
DD
V
V
5.35.2
22
V
:DRegion DD
intp
DD
in VV
V
V
183. ADVANTAGES OF CMOS TECHNOLOGY
1. High input impedance. The input signal is driving electrodes with
a layer of insulation (the metal oxide) between them and what they
are controlling. This gives them a small amount of capacitance, but
virtually infinite resistance. The current into or out of CMOS input
held at one level is just leakage, usually 1 nanoAmpere or less
2. The outputs actively drive both ways
3. The outputs are pretty much rail-to-rail
4. CMOS logic takes very little power when held in a fixed state. The
current consumption comes from switching as those capacitors are
charged and discharged. Even then, it has good speed to power
ratio compared to other logic types.
5. CMOS gates are very simple. The basic gate is an inverter, which
is only two transistors. This together with the low power
consumption means it lends itself well to dense integration. Or
conversely, you get a lot of logic for the size, cost and power.
185. BIPOLAR IC FABRICATION
It is high speed technology.
Speed of operation of Bipolar IC is determined by base width of the
devices.
Base width is determined by difference between two impurity
diffusion profiles.
Devices with very thin base width has high speed of operation.
Bipolar ICs requires buried layer of dopant by growing an epitaxial
layer on silicon.
https://www.youtube.com/watch?v=2fOtOp4KXbM
187. BIPOLAR INTEGRATED CIRCUITS.
FABRICATION
If EDP is used, a bipolar npn transistor can be formed by a sequence
of such steps:
Step 1. n-type epilayer growth.
Step 2. Oxidation.
Step 3. Photolithography forming windows for isolation diffusion.
Step 4. Isolation (separation) diffusion.
188. CONTD..
Step 5: Oxidation.
Step 6: Photolithography forming windows for base diffusion.
Step 7: Base diffusion
190. CONTD..
Step 11. Oxidation.
Step 12. Photolithography forming windows for contact areas.
Step 13. Metallization (deposition of a thin aluminium layer by
vacuum evaporation).
Step 14. Photolithography (selective etching of the metal layer).
Interconnections are formed at this step.
191. CONTD..
Step 15. Anneal in hydrogen to form ohmic contacts where the
aluminium meets n silicon region.
Step 16. Passivation, i.e. deposition of silicon dioxide or silicon
nitride by lowtemperature CVD (chemical vapour deposition).
Step 17. Photolithography exposing bonding pads.
192.
193. ADVANTAGES/DISADVANTAGES OF
BIPOLAR IC TECHNOLOGY
Bipolar devices can switch signals at high speeds
Can be manufactured to handle large currents so
that they can serve as high-power amplifiers in
audio equipment and in wireless transmitters.
Bipolar devices are not especially effective for
weak-signal amplification, or for applications
requiring high circuit impedance
194. FAULT DIAGNOSIS
Definition: A unit under test (UUT) fails when its observed behavior
is different from its expected behavior. Diagnosis consists of
locating the physical fault(s) in a structural model of the UUT. The
degree of accuracy to which faults can be located is called
diagnostic resolution.
Definition: Functionally equivalent faults (FEF) cannot be
distinguished. The partition of all faults into distinct subsets of FEF
defines the maximal fault resolution. A test that achieves the
maximal fault resolution is said to be a complete fault-location test.
195. BASIC CONCEPTS
Repairing the UUT often consists of substituting one of
its replaceable units (RU) referred as a faulty RU, rather than in an
accurate identification of the real fault inside an RU.
We characterize this process by RU resolution. Suppose that the
results of the test do not allow to distinguish between two suspected
RUs U1 and U2.
We could replace now one of these RUs, say U1 with a good RU,
and return to the test experiment.
If the new results are correct, the faulty RU was the replaced one;
otherwise, it is the remaining one U2.
This type of procedure we call sequential diagnosis procedure.
196. TYPE OF DIAGNOSIS PROCESS
The diagnosis process is often hierarchical:
1. Top-down approach (system boards ICs) first-level diagnosis
may deal with "large" RUs like boards called also field-replaceable
units. The faulty board is then tested in a maintenance center to
locate the faulty component (IC) on the board. Accurate location of
faults inside a faulty IC may be also useful for improving its
manufacturing process.
2. Bottom-up approach (ICs boards system) a higher level is
assembled only from components already tested at a lower level.
This is done to minimize the cost of diagnosis and repair, which
increases significally with the level at which the faults are
detected.
197. COMBINATIONAL FAULT DIAGNOSIS
METHODS
This approach does most of the work before the testing experiment.
It uses fault simulation to determine the possible responses to a
given test in the presence of faults.
The database constructed in this step is called a fault table or a fault
dictionary.
To locate faults, one tries to match the actual results of test
experiments with one of the precomputed expected results stored in
the database. The result of the test experiment represents
a combination of effects of the fault to each test pattern.
That's why we call this approach combinational fault diagnosis
method. If this look-up process is successful, the fault table indicates
the corresponding fault(s).
198. FAULT TABLE
Definition: A fault table is a matrix where
columns Fj represent faults, rows Ti represent test patterns,
and aij = 1 if the test pattern Ti detects the fault Fj, otherwise if
the test pattern Ti does not detect the fault Fj, aij = 0.
Denote the actual result of a given test pattern by 1 if it differs
from the precomputed expected one, otherwise denote it by 0.
The result of a test experiment is represented by a
vector where ei = 1 if the actual result of the test patterns
does not match with the expected result, otherwise ei = 0.
Each column vector fj corresponding to a fault Fj represents a
possible result of the test experiment in the case of the
fault Fj.
199. CONTD..
Three cases are now possible depending on the quality of the test
patterns used for carrying out the test experiment:
1. The test result E matches with a single column vector fj in FT. This
result corresponds to the case where a single fault Fj has been
located. In other words, the maximum diagnostic resolution has
been obtained.
2. The test result E matches with a subset of column vectors {fi, fj …
fk} in FT. This result corresponds to the case where a subset of
indistinguishable faults {Fi, Fj … Fk} has been located.
3. No match for E with column vectors in FT is obtained. This result
corresponds to the case where the given set of vectors does not
allow to carry out fault diagnosis. The set of faults described in the
fault table must be incomplete (in other words, the real existing
fault is missing in the fault list considered in FT).
200. CONTD..
In the example the results of three test experiments E1, E2, E3 are
demonstrated. E1 corresponds to the first case where a single fault is
located, E2 corresponds to the second case where a subset of two
indistinguishable faults is located, and E3 corresponds to the third
case where no fault can be located because of the mismatch of
E3 with the column vectors in the fault table.
201. FAULT DICTIONARY
Definition: Fault dictionaries (FD) contain the same data as the
fault tables with the difference that the data is reorganized. In FD a
mapping between the potential results of test experiments and the
faults is represented in a more compressed and ordered form.
For example, the column bit vectors can be represented by ordered
decimal codes (see the example) or by some kind of compressed
signature.
202. MINIMIZATION OF DIAGNOSTIC DATA
To reduce large computational effort involved in building a fault
dictionary, in fault simulation the detected faults are dropped from
the set of simulated faults.
Hence, all the faults detected for the first time by the same vector
will produce the same column vector (signature) in the fault table,
and will be included in the same equivalence class of faults.
In this case the testing experiment can stop after the first failing test,
because the information provided by the following tests is not used.
Such a testing experiment achieves a lower diagnostic resolution.
A tradeoff between computing time and diagnostic resolution can be
achieved by dropping faults after k>1 detections.
203. CONTD..
Example:
In the fault table produced by fault simulation with fault dropping,
only 19 faults need to be simulated compared to the case of 42 faults
when simulation without fault dropping is carried out (the simulated
faults in the fault table are shown in shadowed boxes). As the result
of the fault dropping, however, the following faults remain not
distinguishable: {F2, F3},{F1, F4},{F2, F6}.
204. FAULT LOCATION BY STRUCTURAL
ANALYSIS
Assume a single fault in the circuit. Then there should exist a path
from the site of the fault to each of the outputs where errors have
been detected. Hence the fault site should belong to the intersection
of cones of all failing outputs. A simple structural analysis can help
to find faults that can explain all the observed errors.
205. SEQUENTIAL FAULT DIAGNOSIS
METHODS
Definition: In sequential fault diagnosis the process of fault location
is carried out step by step, where each step depends on the result of
the diagnostic experiment at the previous step. Such a test
experiment is called adaptive testing.
Definition: Sequential experiments can be carried out either by
observing only output responses of the UUT or by pinpointing by a
special probe also internal control points of the UUT (guided
probing). Sequential diagnosis procedure can be graphically
represented as diagnostic tree.
1. Fault location by edge-pin testing
2. Generating tests to distinguish faults
3. Guided-probe testing
4. Fault location by UUT reduction
206. 1. FAULT LOCATION BY EDGE-PIN
TESTING
In fault diagnosis test patterns are applied to the UUT step by step.
In each step, only output signals at edge-pins of the UUT are
observed and their values are compared to the expected ones.
The next test pattern to be applied in adaptive testing depends on the
result of the previous step. The diagnostic tree of this process
consists of the fault nodes FN (rectangles) and test nodes TN
(circles).
A FN is labeled by a set of not yet distinguished faults. The starting
fault node is labeled by the set of all faults. To each FN k a TN is
linked labeled by a test pattern Tk to be applied as the next.
207. CONTD..
Every test pattern distinguishes between the faults it detects and the
ones it does not. The task of the test pattern Tk is to divide the faults
in FN k into two groups - detected and not detected by Tk faults.
Each test node has two outgoing edges corresponding to the results
of the experiment of this test pattern.
The results are indicated as passed (P) or failed (F). The set of faults
shown in a current fault node (rectangle) are equivalent (not
distinguished) under the currently applied test set.
208. CONTD..
Example:
We can see that most of the faults are uniquely identified, two faults
F1,F4 remain indistinguishable. Not all test patterns used in the fault
table are needed. Different faults need for identifying test sequences
with different lengths. The shortest test contains two patterns the
longest four patterns.
209. CONTD..
Rather than applying the entire test sequence in a fixed order as in
combinational fault diagnosis, adaptive testing determines the next
vector to be applied based on the results obtained by the preceding
vectors.
In our example, if T1 fails, the possible faults are {F2,F3}. At this
point applying T2 would be wasteful, because T2 does not distinguish
among these faults. The use of adaptive testing may substantially
decrease the average number of tests required to locate a fault.
210. GENERATING TESTS TO DISTINGUISH
FAULTS
To improve the fault resolution of a given test set T, it is necessary to
generate tests to distinguish among faults equivalent under T.
Consider the problem of generating a test to distinguish between
faults F1 and F2. Such a test must detect one of these faults but not
the other, or vice versa. The following cases are possible.
1. F1 and F2 do not influence the same set of outputs.
Let OUT(Fk) be the set of outputs influenced by the fault Fk. A
test should be generated for F1 using only the circuit feeding the
outputs OUT(F1), or for F2 using only the circuit feeding the
outputs OUT(F2).
2. F1 and F2 influence the same set of outputs. A test should be
generated for F1 without activating F2, or vice versa, for F2
without activating F1.
211. CONTD..
Three possibilities can be mentioned to keep a fault F2: xk=e not
activated, where xk denotes a line in the circuit, and e{0,1}:
1. The value e should be assigned to the line xk.
2. If this is not possible then the activated path from F2 should be
blocked, so that the fault F2 could not propagate and influence the
activated path from F1.
3. If the 2nd case is also not possible then the values propagated from
the sites F1 and F2 and reaching the same gate G should be
opposite on the inputs of G.
212. CONTD..
Example:
1. There are two faults in the circuit: F1: x3,10, and F2: x41. The
fault F1 may influence both outputs, the fault F2 may influence only
the output x8. A test pattern 0010 activates F1 up to the both outputs,
and F2 only to x8. If both outputs will be wrong, F1 is present, and if
only the output x8 will be wrong, F2 is present.
2. There are two faults in the circuit: F1: x3,20, and F2: x5,21. Both of
them influence the same output of the circuit. A test pattern 0100
activates the fault F2. The fault F1 is not activated, because the line
x3,2 has the same value as it would have had if F1 were present.
213. CONTD..
3. There are the same two faults in the circuit: F1: x3,20, and F2:
x5,21. Both of them influence the same output of the circuit. A test
pattern 0110 activates the fault F2. The fault F1 is activated at its
site but not propagated through the AND gate, because of the value
x4 = 0 at its input.
4. There are two faults in the circuit: F1: x3,11, and F2: x3,21. A test
pattern 1001 consists the value x11 which creates the condition
where both of the faults may influence only the same output x8. On
the other hand, the test pattern 1001 activates both of the faults to
the same OR gate (i.e. none of them is blocked).
5. However, the faults produce different values at the inputs of the
gate, hence they are distinguished. If the output value on x8 will be
0, F1 is present. Otherwise, if the output value on x8 will be 1,
either F2 is present or none of the faults F1 and F2 are present.
214. GUIDED-PROBE TESTING
Guided-probe testing extends edge-pin testing process by monitoring
internal signals in the UUT via a probe which is moved (usually by
an operator) following the guidance provided by the test equipment.
The principle of guided-probe testing is to backtrace an error from
the primary output where it has been observed during edge-pin
testing to its physical location in the UUT.
Probing is carried out step-by-step. In each step an internal signal is
probed and compared to the expected value. The next probing
depends on the result of the previous step.
A diagnostic tree can be created for the given test pattern to control
the process of probing. The tree consists of internal nodes (circles) to
mark the internal lines to be probed, and of terminal nodes
(rectangles) to show the possible result of diagnosis.
The results of probing are indicated as passed (P) or failed (F).
215. CONTD..
Typical faults located are opens and defective components. An open
between two points A and B in a connection line is identified by a
mismatch between the error observed at B and the correct value
measured at A.
A faulty device is identified by detecting an error at one of its
outputs, while only correct values are measured at its inputs.
The most time-consuming part of guided-probe testing is moving the
probe. To speed-up the fault location process, we need to reduce the
number of probed lines. A lot of methods to minimize the number of
probings are available.
216. CONTD..
Example:
Let have a test pattern 1010 applied to the inputs of the circuit. The
diagnostic tree created for this particular test pattern is shown. On
the output x8 , instead of the expected value 0, an erroneous signal 1
is detected. By back tracing (indicated by bold arrows in the
diagnostic tree) the faulty component NOR- x5 is located.
217. CONTD..
Diagnostic tree allows to carry out optimization of the fault location
procedure, for example to generate a procedure with minimum
average number of probes.
218. FAULT LOCATION BY UUT
REDUCTION
Initially the UUT is the entire circuit and the process starts when its
test fails. While the failing UUT can be partitioned, half of the UUT
is disabled and the remaining half is tested. If the test passes, the
fault must be in the disabled part, which then becomes the UUT. If
the test fails, the tested part becomes the UUT.
219. CHARACTERIZATION TECHNIQUE
CHARACTERIZATION PROCEDURE :An outline of factors
that should be considered when establishing a characterization
procedure is provided here for every supplier to establish a
characterization procedure.
Device Characterization Plan: The characterization plan should
include the following major activities for the device to be
characterized:
1. Review the Characterization Checklist.
2. Determination of if a matrix lot is necessary for the device
characterization.
3. Determination of the characterization method to be used.
4. Establishment of the parameters and conditions to be
characterized.
5. Define format of the characterization report.
220. CONTD..
Matrix Lot Characterization: When characterizing a matrix lot,
the number of split cells, samples per cell and the data analysis
methods should also be defined in the plan.
Sample Sizes :When deciding on sample sizes for characterization,
two important factors are to be considered: confidence interval and
confidence level.
221. CONTD..
Characterization Report :The characterization report should include
the following:
1. A copy of the characterization plan.
2. A detailed discussion of the characterization methods used
3. A listing of parameters and conditions used in characterization.
4. Characterization data analysis and conclusions.
5. Document simulation results including brief explanations on
methods applied – for parameters that are not measurable and/or
tested in production and covered by design simulation only.
6. Identify part weaknesses and reliability concerns and define
corrective actions.
222. CONCLUSIONS
FDI: a mature field
Huge literature
SAFEPROCESS
European projects like MONET
Further research focuses on:
New class of systems (e.g. Hybrid systems)
Applications
Fault tolerance issues
223. REFERENCES
www.web.stanford.edu/class/ee311/NOTES/Isolation.pdf
https://en.wikipedia.org/wiki/LOCOS
www.textofvideo.nptel.iitm.ac.in/117106093/lec33.pdf
www.iue.tuwien.ac.at/phd/filipovic/node77.html
www.iue.tuwien.ac.at/phd/hollauer/node7.html
www.https://en.wikipedia.org/wiki/Shallow_trench_isolation
https://en.wikipedia.org/wiki/Silicon_on_insulator
www.sctest.cse.ucsc.edu/lavo/Fault_Diagnosis_Overview.ppt
https://www.edgefx.in/understanding-cmos-fabrication-technology/
www.iaa.ncku.edu.tw/~aeromems/MEMSDesign/Ch2.pdf
www.circuitstoday.com/bipolar-ic-manufacturing-process
Book: The Science and Engineering of Microelectronic Fabrication,
Stephen A. Campbell, Oxford University Press, 2001