Passive device fabrication in Integrated circuits

Passive Device Fabrication in IC
By
Prof. Abhishek Kadam
(M.Tech Electronics)
Recourse's Credit : RF Microelectronics By B. Razavi

Contents
• Fabrication of inductors
• fabrication of transformers
• fabrication of varactors
• fabrication of fixed value capacitors

Need of Inductors
• Modern RF design needs many inductors
• This topology suffers from two serious drawbacks
• the bandwidth at node X is limited to 1/[(𝑅 𝐷| 𝑟𝑂1
𝐶 𝐷 ]
• the voltage headroom trades with the voltage gain, g 𝑚1
(𝑅 𝐷| 𝑟𝑂1
• CMOS technology scaling tends to improve the former but at the cost of the latter
• For example, in 65-nm technology with a 1-V supply, the circuit provides a
bandwidth of several gigahertz but a voltage gain in the range of 3 to 4.

Need of Inductors
• 𝐿 𝐷 resonates with 𝐶 𝐷, allowing operation at much higher
frequencies (albeit in a narrow band)
• 𝐿 𝐷 sustains little dc voltage drop, the circuit can comfortably
operate with low supply voltages while providing a reasonable
voltage gain (e.g., 10).

Drawback of using off chip Inductors
• External inductors present cost penalties
• The bond wires and package pins connecting the chip to the outside world
may experience significant coupling creating crosstalk between different
parts of the transceiver.
• external connections introduce parasitics that become significant at higher
frequencies.
• it is difficult to realize differential operation with external loads because of
the poor control of the length of bond wires

Disadvantages of on chip inductors
• On-chip inductors exhibit a lower quality
factor than their off-chip counterparts, leading
to higher “phase noise” in oscillators
• Unlike resistors and Capacitors, inductors are
much more difficult to model

INDUCTORS
• Basic Structure
– Integrated inductors are typically realized as metal
spirals
– Owing to the mutual coupling between every two
turns, spirals exhibit a higher inductance than a
straight line having the same length
– To minimize the series resistance and the parasitic
capacitance, the spiral is implemented in the top
metal layer

INDUCTORS
• Fig shows typical spiral structure of Integrated inductor
• There are three turns AB, BC and CD
• Considering mutual inductance Total inductance is given by
• 𝐿 𝑡𝑜𝑡𝑎𝑙 = 𝐿1 + 𝐿2 + 𝐿3 + 𝑀12 + M13 + M23
• Due to geometry inner turns exhibit lower inductance
• Mutual inductance M13 is smaller than 𝑀12

Two-dimensional square spiral
• A two-dimensional square spiral is fully
specified by four quantities
– The outer dimension, 𝐷 𝑜𝑢𝑡
– The line width, W
– The line spacing, S
– Number of turns, N

• Comment on following two spiral structures

Inductor Geometries
a) Circular
b) Octagonal
c) Symmetric

Inductor Geometries
d) Stacked
e) With ground shield
f) Parallel spirals

Inductance Equation
• An empirical formula that has less than 10% error
for inductors in the range of 5 to 50 nH is given by
Where 𝐴 𝑚 is metal area and 𝐴 𝑡𝑜𝑡𝑎𝑙 is total inductor
area

For N turns 𝑙 𝑡𝑜𝑡 is given by
Calculate 𝑙 𝑡𝑜𝑡 for fig

Thus Approximated value of L is given by

• a given length of wire yields roughly a
constant inductance regardless of how it is
“wound.”

Parasitic Capacitances
• As a planar structure built upon a substrate,
spiral inductors suffer from parasitic
capacitances of two types
1. The metal line forming the inductor exhibits
parallel plate and fringe capacitances to the
substrate

2.The adjacent turns also bear a fringe
capacitance, which equivalently appears in
parallel with each segment

Loss mechanism
Suppose the metal line forming an inductor exhibits a series resistance
The Q may be defined as the ratio of the desirable impedance, 𝐿1 𝜔0, and the
undesirable impedance, 𝑅 𝑆
𝑄 = 𝐿1 𝜔0/𝑅 𝑠

Skin Effect
• At high frequencies, the current through a conductor prefers
to flow at the surface.
• The actual distribution of the current follows an exponential
decay from the surface of the conductor inward, 𝐽 𝑠 =
𝐽0exp(2𝑥/𝛿), where 𝐽0 denotes the current density (in 𝐴/𝑚2
)
at the surface, and δ is the “skin depth.” The value of δ is
given by
𝛿 = 1/ 𝜋𝑓𝜇σ
• The extra resistance of a conductor due to the skin effect is
equal to 𝑅 𝑠𝑘𝑖𝑛 = 1/𝜎𝛿

• In spiral inductors, the proximity of adjacent
turns results in a complex current distribution.
• the current may concentrate near the edge of
the wire. To understand this “current
crowding” effect, consider the more detailed
diagram shown below

• The current in one turn creates a time-varying
magnetic field, B, that penetrates the other
turns, generating loops of current Called
“eddy currents,”

• Thus expression for the resistance of a spiral
inductor is
Where

• Current crowding also alters the inductance and
capacitance of spiral geometries. Since the current is
pushed to the edge of the wire, the equivalent
diameter of each turn changes slightly, yielding an
inductance different from the low-frequency value

Alternative Inductor Structures
• Symmetric Inductors

• The principal drawback of symmetric inductors is
their large interwinding capacitance

• How do we reduce the interwinding
capacitance?
• We can increase the line-to-line spacing, S,
but, for a given outer dimension, this results in
smaller inner turns and hence a lower
inductance.
• As a rule of thumb, we choose a spacing of
approximately three times the minimum
allowable value.

Inductors with Ground Shield
• The magnetic coupling from an inductor to the substrate can be
understood with the aid of basic electromagnetic laws:
1. Ampere’s law states that a current flowing through a conductor generates a magnetic
field around the conductor;
2. Faraday’s law states that a time-varying magnetic field induces a voltage, and hence a
current if the voltage appears across a conducting material;
3. Lenz’s law states that the current induced by a magnetic field generates another
magnetic field opposing the first field.
• Ampere’s and Faraday’s laws readily reveal that, as the current through an
inductor varies with time, it creates an eddy current in the substrate
• Lenz’s law implies that the current flows in the opposite direction.

• both the electric coupling and the magnetic coupling to the
substrate are eliminated if a grounded conductive plate is
placed under the spiral
• This method indeed reduces the path resistance seen by both
displacement and eddy currents but the equivalent
inductance also falls with 𝑅 𝑠𝑢𝑏,
• we observed that eddy currents in a continuous shield
drastically reduce the inductance and the Q

• The shield can provide a low-resistance termination for electric field
lines even if it is not continuous
• a “patterned” shield, i.e., a plane broken periodically in the
direction perpendicular to the flow of eddy currents, receives most
of the electric field lines without reducing the inductance.
• A small fraction of the field lines sneak through the gaps in the
shield and terminate on the lossy substrate. Thus, the width of the
gaps must be minimized.
• The use of a patterned shield may increase the Q by 10 to 15% but
the improvement comes at the cost of higher capacitance

• Stacked Inductors
– At frequencies up to about 5 GHz, inductor values
encountered in practice fall in the range of five to several
tens of nanohenries
– If realized as a single spiral, such inductors occupy a large
area and lead to long interconnects between the circuit
blocks.
– In stacked Inductors place two or more spirals in series,
obtaining a higher inductance not only due to the series
connection but also as a result of strong mutual coupling.

Transformers
• Integrated transformers can perform a number of useful
functions in RF design:
1. impedance matching
2. feedback or feed forward with positive or negative polarity
3. ac coupling between stages.
4. single-ended to differential conversion or vice versa
• A well-designed transformer must exhibit the following:
1. low series resistance in the primary and secondary windings
2. high magnetic coupling between the primary and the secondary
3. low capacitive coupling between the primary and the secondary,
4. Low Parasitic capacitances to the substrate.

Transformer Structures
• An integrated transformer generally comprises
two spiral inductors with strong magnetic
coupling.

• The transformer structure of previous diagram
suffers from low magnetic coupling, an
asymmetric primary, and an asymmetric
secondary.
• To remedy the former, the number of turns
can be increased as shown in fig below but at
the cost of higher capacitive coupling

• It is possible to construct planar transformers
having a turns ratio greater than unity

• Figure below shows two other examples of planar
transformers. Here, two asymmetric spirals are inter
wound to achieve a high coupling factor.

• Transformers can also be implemented as three-
dimensional structures. Similar to the stacked inductors
• It is important to recognize the following attributes
– the alignment of the primary and secondary turns results in a
slightly higher magnetic coupling factor here than in the planar
transformers
– unlike the planar structures, the primary and the secondary can
be symmetric and identical
– the overall area occupied by 3D transformers is less than that of
their planar counterparts

• Another advantage of stacked transformers is
that they can readily provide a turns ratio
higher than unity
• the idea is to incorporate multiple spirals in
series to form the primary or the secondary.

• Stacked transformers must, however, deal
with two issues
– the lower spirals suffer from a higher resistance
due to the thinner metal layers.
– the capacitance between the primary and
secondary is larger here than in planar
transformers
• To reduce this capacitance, the primary and secondary
turns can be “staggered,” thus minimizing their overlap

VARACTORS
• A varactor is a voltage-dependent capacitor.
• Two attributes of varactors become critical in
oscillator design:
– the capacitance range, i.e., the ratio of the
maximum and minimum capacitances that the
varactor can provide 𝐶 𝑚𝑎𝑥 / 𝐶 𝑚𝑖𝑛
– the quality factor of the varactor, which is limited
by the parasitic series resistances within the
structure

• In older generations of RF ICs, varactors were
realized as reverse-biased pn junctions

The capacitance range and Q of pn junctions
• At a reverse bias of 𝑉𝐷, the junction capacitance, 𝐶𝑗, is
given by
where 𝐶𝑗0
is the capacitance at zero bias, 𝑉0 the built-in
potential, and m an exponent around 0.3 in integrated
structures
• 𝐶𝑗 has very weak dependence on 𝑉𝐷
• In today’s technology with 𝑉𝐷𝐷 ≈ 1𝑉,𝐶𝑗max
/𝐶𝑗m𝑖𝑛
=1.23

• The Q of a pn-junction varactor is given by the total
series resistance of the structure.
• this resistance is primarily due to the n-well and can
be minimized by selecting minimum spacing
between the 𝑛+
and 𝑝+
contacts
• to lower the resistance in two dimensions each 𝑝+
region can be surrounded by an 𝑛+ ring.

• Due to the two-dimensional nature of the flow, it is difficult to
determine or compute the equivalent series resistance of the
structure
• This issue arises partly because the sheet resistance of the n-
well is typically measured by the foundry for contacts having a
spacing greater than the depth of the n-well
• Since the current path in this case is different from that in Fig.
below the n-well sheet resistance cannot be directly applied
to the calculation of the varactor series resistance

• In modern RF IC design, MOS varactors have
supplanted their pn-junction counterparts.
• A regular MOSFET exhibits a voltage-dependent
gate capacitance but the non monotonic behavior
limits the design flexibility.

• A simple modification of the
MOS device avoids the before
mentioned issues
• Accumulation-mode MOS
varactor are shown in fig.
near
• this structure is obtained by
placing an NMOS transistor
inside an n-well
– If 𝑉𝐺 < 𝑉𝑆, then the electrons in
the n-well are repelled from the
silicon/oxide interface and a
depletion region is formed and
the equivalent capacitance is
given by the series combination
of the oxide and depletion
capacitances

– As 𝑉𝐺 exceeds 𝑉𝑆, the interface
attracts electrons from the 𝑛+
source/drain terminals,
creating a channel
– The overall capacitance
therefore rises to that of the
oxide, behaving as shown
• These varactors operate with low supply voltages better
than their pn-junction counterparts.
• Another advantage of accumulation-mode MOS
varactors is that, unlike pn junctions, they can tolerate
both positive and negative voltages.

• we approximate the characteristic of
Accumulation-mode MOS varactor by
Here, a and 𝑉0 allow fitting for the intercept and the slope,
respectively, and 𝐶 𝑚𝑖𝑛 and 𝐶 𝑚𝑎𝑥 include the gate-drain and gate-
source overlap capacitance.

• The Q of MOS varactors is determined by the
resistance between the source and drain
terminals
• This resistance and the capacitance are
distributed from the source to the drain and
can be approximated by the lumped model
depicted below

• Finding 𝐶𝑡𝑜𝑡 𝑎𝑛𝑑 𝑅𝑡𝑜𝑡
• Let us first consider only one-half of the structure as
shown and We turn the circuit upside down, arriving
at the more familiar topology illustrated

For general T-line shown below, input impedance, 𝑍𝑖𝑛,
is given by
Thus for equivalent MOS structure 𝑍𝑖𝑛 is given by

• At frequencies well below 1/(𝑅𝑡𝑜𝑡 𝐶𝑡𝑜𝑡/4), the
argument of 𝐭𝐚𝐧𝐡 is much less than unity,
allowing the approximation,
It follows
Hence

• It is desirable to maximize the Q of varactors for
oscillator design.
• From our foregoing study of MOS varactors, we
conclude that the device length (the distance
between the source and drain) must be minimized
• but for a minimum channel length, the overlap
capacitance between the gate and source/drain
terminals becomes a substantial fraction of the
overall capacitance, limiting the capacitance range
• yielding a Low ratio of

CONSTANT CAPACITORS
• RF circuits employ constant capacitors for various
purposes
1. To adjust the resonance frequency of LC tanks
2. to provide ac coupling between stages
3. to bypass the supply rail to ground
• MOS Capacitors
– MOSFETs configured as capacitors offer the highest
density in integrated circuits because 𝐶 𝑜𝑥 is larger
than other capacitances in CMOS processes.
– But the use of MOS capacitors entails two issues
• To provide the maximum capacitance, the device requires
a 𝑉𝐺𝑆 higher than the threshold voltage
• The channel resistance limits the Q of MOS capacitors at
high frequencies

• Consider the circuit given
• 𝑉𝐺𝑆3
= 𝑉𝐷𝐷 − 𝑉𝐺𝑆22
Thus for Low 𝑉𝐷𝐷, 𝑉𝐺𝑆3
is
Low and Hence Due to Low Overdrive Voltage
𝑅 𝑜𝑛 is high
• For these reasons, MOS capacitors rarely serve
as coupling devices

• Other application of MOS capacitors is in
supply bypass
• the supply line may include significant bond wire inductance,
allowing feedback from the second stage to the first at high
frequencies
• The bypass capacitor, M3, creates a low impedance between
the supply and the ground, suppressing the feedback.
• if the equivalent series resistance of the device becomes
comparable with the reactance of its capacitance, then the
bypass impedance may not be low enough to suppress the
feedback.

• Comment of 𝑅 𝑜𝑛 of the following citcuit
modification

Metal-Plate Capacitors
• If the Q or linearity of MOS capacitors is
inadequate, metal-plate capacitors can be
used instead.
• The parallel plate structure is shown below
However, even with all metal layers and a
poly layer, parallel-plate structures achieve
less capacitance density than MOSFETs do.
For example,
with nine metal layers in 65-nm
technology, the former provides a density
of about 1.4 fF/𝜇𝑚2 and the latter, 17 fF/
𝜇𝑚2

• Parallel-plate geometries also suffer from a parasitic capacitance to
the substrate
• In a typical process, this value reaches 10%, leading to serious
difficulties in circuit design.
• To alleviate the above issue, only a few top metal layers can be
utilized

• An alternative geometry utilizes the lateral
electric field between adjacent metal lines to
achieve a high capacitance density
This “fringe” capacitor consists of narrow metal lines with the
minimum allowable spacing

Passive device fabrication in Integrated circuits

More Related Content

What's hot (20)

Similar to Passive device fabrication in Integrated circuits (20)

Recently uploaded (20)

Passive device fabrication in Integrated circuits