VLSI Design

Unit- 2
MOS and BiCMOS Circuit Design Processes
• Methods of realizing circuit design in silicon
• The design process can be understood by means of stick diagrams and symbolic
diagrams along with set of design rules.
• Design rules: is a communication link between designers specifying the requirements
and the fabricator.
MOS Layers:
• MOS circuits are basically formed by 4 layers
§ Metal
§ Polysilicon
§ N diffusion
§ P diffusion
• Here all the 4 layers are isolated from each other through thick or thin oxide layer
(i.e., silicon dioxide layer)
• The thin oxide (thinox) layer includes n-diffusion, p-diffusion and transistor
channel.
Stick diagram:
• Stick diagrams are a means of capturing topography and layer information using
simple diagrams.
• They convey layer information through color codes (or monochrome encoding).
• Acts as an interface between symbolic circuit and the actual layout.
• Stick diagrams do show all components/vias(contacts), relative placement of
components and helps in planning and routing. It goes one step closer to layout.
• However they do not show exact placement of components, transistor sizes, length and
width of wires also the boundaries. Thus we can say that it does not give any low level
details.
• The color encodings chosen for different technologies is shown below.
• Encodings for NMOS process:

Dept. of ECE, Dr. M. Durga Prakash

VLSI Design

Procedure to draw Stick Diagram:

Nmos Design Process.
1) Draw two metal lines/ power rails providing sufficient space to accommodate all
transistors. i.e: Vdd & Vss.
2) Draw common n+ diffusion layer for all the transistors.
3) Provide Vdd and Vss contacts.
4) Draw polysilicon to cross n+ diffusion layer to form transistors.
5) Create buried contact for depletion transistor.
6) Provide input and output connection.

CMOS Design Process:

• Two type of transistors are used i.e: Nmos and Pmos, thus in stick diagram
demarcation line is used to separate them.
• All Pmos transistors are placed above Demarcation line and Nmos transistors below
demarcation line.
• While drawing stick Diagram
1. Diffusion paths must not cross the demarcation line
2. N-diffusion and P-diffusion wires must not join.
3. Nmos and Pmos transistors are joined by Metal layer when it is required.
4. Cross must be placed on Vdd and Vss which represent substrate and P-well
connection respectively.
Encodings for CMOS process:

Dept. of ECE, Dr. M. Durga Prakash

VLSI Design

Procedure to draw Stick Diagram:

1) Draw two metal lines/ power rails providing sufficient space to accommodate all
transistors. i.e: Vdd & Vss.
2) Draw demarcation line in the middle of the two power lines.
3) Draw P+ diffusion above demarcation and N+ diffusion below demarcation
4) Draw polysilicon to represent Pmos and Nmos which represents gates of the transistor.
5) Connect source terminal of transistors to supply.
6) Drain terminals of transistor are connected using metal 1.
7) Place contact cuts wherever necessary.
8) Draw X which represents substrate and P-well contact on power lines.

Layout: describes actual layers and geometry on silicon substrate to implement a

function(Expressions).
[Diffusion region where transistor can be formed is called active region, polysilicon serves as
the gate of MOS transistor. L defines channel length and W represents width of channel/active
region]
Design rules: are set of guidelines which specify minimum dimension and spacing allowed in
layout drawing. Design rules also acts a communication link between circuit designers and
process engineers during manufacturing phase.
Goal of design rule: is to achieve optimum yield. Yield = (No. of good chips on
wafer)/(Total no. of chips on wafer).

Design rules are also called layout rules. If the circuit performance has to be increased then
rules must be more aggressive. Else this leads to non-function of the circuit or yield reduction.
There are two rules.
1. Micron Rule - Absolute Dimension rule, here all sizes and spacing are specified in micron.
Here the circuit density is the important goal.
2. Lambda (λ) Based Rules - The Lambda based design rules are Proposed by Mead and
Conway. Scalable design rules, here this design rule normalizes all geometric design rule by
parameter lambda (λ) also called as scaling factor/feature size. In this all mask patterns are
expressed as multiples of lambda.
Advantages of lambda based design rules:
1. The mask layout can be scaled down proportionally if the feature size of the fabrication
process is reduced.
2. Design rules are conservative.
3. This rule enable technology changes and design reuse and reduced design cost.

Disadvantages:
1. Linear scaling cannot be extended and is limited over range of dimension (1-3 μm)
2. As rules are conservative, results in over dimension and density of design is less.

The Design rules can be conveniently set out in diagrammatic form as shown in fig. 1
for width and separation of conducting path. In fig. 2 shows the design rules associated with
extensions and separations with transistor. Fig. 3 and 4 demonstrates the design rules for

Dept. of ECE, Dr. M. Durga Prakash

VLSI Design

contacts between layers. Table below also gives the layer and distance of separation
dimensions.

Layer Dimension Layer -Layer Dimension

n-diffusion 2λ n-diffusion – n-diffusion 3λ
p-diffusion 2λ p-diffusion – p-diffusion 3λ
Polysilicon 2λ n/p diffusion - polysilicon 1λ
Metal 1 3λ Poly-poly 2λ
Metal 2 4λ Metal 1 3λ
Metal 2 4λ
Layer dimension Distance of Separation

Fig. 1 Design rules for wires and separations (nMOS and CMOS)

Fig. 2 Design rules for Transistors (nMOS, pMOS and CMOS)

Transistor design rules

§ Minimum dimension of transistor is 2λ × 2λ – overlapping of diffusion and ploy
§ Poly and diffusion both must extend beyond the boundary of transistor at least by 2λ

Dept. of ECE, Dr. M. Durga Prakash

VLSI Design

§ Implant for depletion mode transistor is 6λ × 6λ i.e., implant must extend boundary of
transistor by at least 2λ in all direction.
§ From the boundary/ implant of one transistor, the next transistor should maintain min
distance of 2λ
§ The distance from contact cut to transistor should be at least 2λ

Metal contact – contact between metal 1 to polysilicon OR metal 1 to diffusion (active

region) is called metal contact. This is shown in fig. 3
§ A 2λ × 2λ cut centered on 4λ × 4λ superimposed area is used to connect layers
§ In case of multiple contacts the distance between adjacent contacts should be 2λ

Fig. 3. Contacts (nMOS and CMOS)

Via contact – the contact between metal 1 and metal 2 is called via contact as shown in fig. 4.
§ A 2λ × 2λ cut centered on 4λ × 4λ superimposed area is used to connect layers
§ To connect metal 2 with diffusion via and cut both are used

Fig. 4 contacts

Contact Cuts:
• Electrical connection between layers can be done using special structures 'contact
cuts'.
• There are 3 approaches for contacts between polysilicon and diffusion in nMOS
circuits. They are
1. Polysilicon to metal and then to diffusion
2. Buried contact - polysilicon to diffusion
3. Butting contact - polysilicon to diffusion using metal
ü Among the three buried contact is most used as it gives economy in space and reliable
contact.
Ø Buried contact is distinguished feature in nMOS for connection between poly and
diffusion and this is most widely used than butting contact.
Dept. of ECE, Dr. M. Durga Prakash
VLSI Design

Buried Contact (nMOS):

§ Layer is joined over the area of 2λ × 2λ with buried contact cut extending by 1λ in all
directions except in the diffusion path. It extends by 2λ in order to avoid formation of
unwanted transistors.
§ The contact cut shown in broken line indicates the region where thinox is removed on
the silicon wafer and polysilicon gets deposited on wafer.
§ When impurities are added, it diffuse into poly and also to diffusion region within the
contact area. This provides satisfactory contact between ploy and diffusion as shown
in fig 5.

Ø In CMOS poly to diffusion connections are made through metal. The process of
making connection between metal and either of 2 layers (poly or diffusion) is by
buried contact.

Fig. 5 Buried and butting contacts only for nMOS

Dept. of ECE, Dr. M. Durga Prakash

VLSI Design

Fig 5. Cross section through contact structures

Butting contact
§ Butting contact process is complicated and done when two layers do not overlap. Contact
cut of 2λ × 2λ is made until each of layers is joined. The layers are held in such a way that
these two contacts become continuous.
§ The poly and diffusion outlines overlap and thinox under ploy acts as mask during
diffusion process. Finally contact between two butting layers is done by a metal. This can
be seen in fig. 5 cross-sectional view.

Double Metal MOS Process Rules:

• If to process considered till now introduction of second metal layer will boosts the
design capabilities. It gives more freedom. Ex. this will be helpful for power rail (Vdd
and Vss/Gnd) distribution and also for clock.
• This process is called Double Metal MOS Process
• This technique involves connecting metal 1 and metal 2 contacts called 'via'. This is
shown in fig. 4 and fig. 6

Fig. 6 cross section of via contact structure

• The 2nd metal layer is coarser than 1st metal layer (conventional) and the isolation layer
between the 2 is usually thicker than normal.
• To distinguish contacts between 1st and 2nd metal layer they are called as ‘vias’ rather
than contact cut
• In stick diagram representation its color code is dark blue or purple.

Dept. of ECE, Dr. M. Durga Prakash

VLSI Design

The steps of fabrication process is as follows:

1. Using chemical vapor deposition oxide layer under 1st metal layer is deposited.
2. using same method oxide between 2 metal layers are formed.
3. Selected areas of oxide are removed by using plasma etching. The etching process is
done under high vertical ion bombardment to get high and uniform etching.

The layout strategy used with double metal process is summarized as below
1. Second metal layer is usually used for global power railings and clock lines
2. First metal layer is used for local power distribution and signals.
3. The layout of the two metals are such that are mutually orthogonal wherever
possible

§ Similar to double metal process, other process allows second poly layer. The process
steps are similar to previously described process.
§ The first polysilicon (poly 1) layer is deposited and patterned on this a second thinox
(thin oxide) layer is grown. On this the second polysilicon (poly 2) layer is deposited
and patterned. Thus 2nd thinox isolated the poly layers.
§ Presence of poly 2 provides greater flexibility in interconnections and allows
transistors to be formed by intersection of poly 2 and diffusion.
CMOS Lambda-based Design Rules:
• Comparing to Nmos fabrication process, CMOS fabrication is more complex.
• Extending the Nmos design rules, Noting exclusion of butting contact and buried
contact rules.
• Additional rules associated with CMOS process concerned with unique feature p-well
CMOS, i.e: p-well and P+ Mask and Substrate contact.

Dept. of ECE, Dr. M. Durga Prakash

VLSI Design

Problems on stick diagram and layouts.

Nmos Inverter

Dept. of ECE, Dr. M. Durga Prakash

VLSI Design

CMOS Inverter

Dept. of ECE, Dr. M. Durga Prakash

VLSI Design

Diffusion lines running horizontal and polysilicon lines in vertical direction.

Dept. of ECE, Dr. M. Durga Prakash

VLSI Design

• When more number of inputs available, Euler’s path is determined to know gate
ordering.
• Advantage of using Euler’s path is to that a common diffusion line can be used which
reduces number of contact cuts.
• Uninterrupted path in both pull-up and pull down network represents optimized gate
ordering which helps in drawing layout without breaking the diffusion layer.

Dept. of ECE, Dr. M. Durga Prakash

VLSI Design

Dept. of ECE, Dr. M. Durga Prakash

VLSI Design

Dept. of ECE, Dr. M. Durga Prakash

VLSI Design

Stick diagram of 2:1 MUX using transmission gates.

Two input XOR gate realization using transmission gates.

Two way selector with enable

Dept. of ECE, Dr. M. Durga Prakash

 VLSI Design

Basic circuit concepts

• In MOS technology, Active devices are dealt with some measurement.
• Wiring up of circuits is done through various conductive layers which is produced by
MOS Processing.
• Therefore it is necessary to be aware of resistive and capacitive characteristics of each
layers.
• For evaluating the effects of wiring, input and output capacitances, sheet resistance and
standard unit capacitances are used.
• Further delay associated with wiring, inverters are evaluated by the term delay unit τ.

Sheet Resistance 𝑹𝒔

Consider a transistor with a channel having resistivity ρ, width W, thickness t and length
between source and drain is L.

Resistance of the channel between drain and source is expressed as.

𝜌. 𝐿𝑒𝑛𝑔𝑡ℎ 𝜌. 𝐿
𝑅𝐷𝑆 = =
𝐴𝑟𝑒𝑎 𝑜𝑓 𝑐𝑟𝑜𝑠𝑠 𝑠𝑒𝑐𝑡𝑖𝑜𝑛 𝑡. 𝑊

𝐿
𝑅𝐷𝑆 = 𝑅𝑠
𝜌
𝑊

Where 𝑅𝑠 = is a constant and it is called sheet resistance.

𝑡

From the above equation, sheet resistance can be defined as resistance of the channel whose
length and width are equal.

𝑅𝑠 is completely independent of area square. Ex: 1μm per side square slab of material has
exactly same resistance as 1cm per side square slab of same material if thickness is same.

Dept. of ECE, M. Durga Prakash

 VLSI Design

Area capacitance

In between gate and channel exists a capacitance and it is called gate capacitance and
denoted by 𝐶𝑔.

From the above diagram.

∈0∈𝑟 𝐴
𝐶𝑔 =

𝐷
A is area of the channel or surface area of the gate
𝐶𝑔 ∈0 ∈𝑟
= 𝑝𝐹/(𝜇𝑚)2
𝐴 𝐷
∈0 ∈𝑟
𝐶𝐴 =

𝐷
∈0 = permittivity of free space = 8.854x 10−12 F/m.

∈𝑟= relative permittivity of a given material

𝐷 = thickness of sio2 constant for a given technology.

Area capacitance is defined as capacitance per unit area at the gate of transistor and denoted
by 𝐶𝐴.

Standard unit of capacitance (□𝑪𝒈)

The standard unit of capacitance is defined as the capacitance at the gate of 1:1 transistor.

Ex: consider a 1:1 transistor where L = 2λ and W = 2λ.

Gate area of transistor = L x W

Dept. of ECE, M. Durga Prakash

 VLSI Design

A = 2λ x 2λ = (2λ)2

Actual capacitance at the gate of transistor C = 𝐶𝐴. 𝐴 = 𝐶𝐴. (2λ)2

C = 4x10−4 𝑃𝐹/(𝜇𝑚)2. (2λ)2

Consider 5 𝜇𝑚 technology, i.e: 2λ = 5 𝜇𝑚

C = 4x10−4 𝑝𝐹/(𝜇𝑚)2. (5 𝜇𝑚)2 = 4x10−4 x 25 = 0.01 pF

C = 1□𝐶𝑔

Standard Delay Unit (τ)

Time delay is measured in terms of standard unit τ.

It is defined as product of 𝑅𝑠 and 𝐶𝑔. i.e: τ = 𝑅𝑠. □𝐶𝑔

Measurement of τ.

Consider nmos driven by pass transistor shown in below figure and the dimensions are
indicated. Pass transistor is ON for given gate voltage 𝑉𝐺𝐺 . Pass transistor is represented by
𝑅𝑠 due to its equal length and width. Pull down transistor of inverter is represented by
capacitance □𝐶𝑔. Since pull down has minimum dimensions.

τ is defined as time taken by capacitor to charge from 0 to 63.2% of maximum value as

shown in below figure.

Dept. of ECE, M. Durga Prakash

VLSI Design

Inverter Delays

Consider basic 4:1 nmos inverter. To achieve 4:1 𝑍𝑝𝑢 𝑡𝑜 𝑍𝑝𝑑 ratio, 𝑅𝑝𝑢 will be 4𝑅𝑝𝑑. Clearly
resistance 𝑅𝑝𝑢 value is 𝑅𝑝𝑢 = 4 𝑅𝑠 = 40KΩ. Meanwhile 𝑅𝑝𝑑 value is 10KΩ.

Delay associated with inverter depends on ON and OFF condition of transistors.

Consider a pair of cascaded inverter, delay in this pair will be constant irrespective of sense
of logic level transition. The overall delay of nmos inverter is τ + 4τ = 5τ. Shown in below
figure.

In general term delay through nmos inverter pair is given by 𝑇𝑑 = (1 + 𝑍𝑝𝑢⁄𝑍𝑝𝑑) τ

So single 4:1 inverter exhibits asymmetric delays, delay in turning on τ (capacitor

discharging condition) and delay in turning off is 4 τ (capacitor charging condition).
Asymmetry becomes worse for inverter with 8:1 ratio.

For CMOS inverter, nmos rules no longer applies, but we need to consider natural
asymmetry of equal size pull up and pull down transistors.

Gate capacitance is double compare to nmos inverter since input is connected to both
transistors and delay associated with pair of minimum size inverters is shown in below
figure.

Dept. of ECE, M. Durga Prakash

VLSI Design

Asymmetry of resistance is eliminated by increasing the width of p- device channel by factor

of two or three, but gate capacitance increases by the same factor.

Driving large capacitive loads

• A large capacitive loads problem arises when a signal to be transmitted from On chip
to Off chip destinations.
• Off chip capacitance is is generally higher than On chip □𝐶𝑔.And it is denoted by 𝐶𝐿.
𝐶𝐿 ≥ 104□𝐶𝑔
• A capacitance of this order to be driven through low resistance otherwise long delays
will occur.

Cascaded Inverters as drivers

• Inverters to drive large capacitive loads resistance associated with pull up and pull
down transistors to be low.
• Low resistance values of 𝑍𝑝𝑢 and 𝑍𝑝𝑑 implies low L:W ratio or channel width must
be made wider to reduce channel resistance but consequently inverter occupies large
area.
• Gate area LxW is more significant and large capacitance present at input which slows
down rate of change of voltage at input.
• Remedy to use N cascade inverter is by maintaining L to a minimum feature size and
width of each successive stage is increased by factor f as shown in below figure.

• With increase in width factor increases capacitive load at input side and area occupied
by the inverter also increases.
• The rate of width increase influence on number of stages to be cascaded to drive
particular 𝐶𝐿 value.
• Total delay associated with nmos pair is 5 τ and cmos pair is 7 τ.
Let y = 𝐶𝐿⁄□𝐶 = 𝑓𝑁 , 𝑓 and N are interdependent.
𝑔
To determine value of 𝑓 to minimize overall delay for given y
ln(𝑦) = 𝑁 ln(𝑓)

Dept. of ECE, M. Durga Prakash

 VLSI Design

ln(𝑦)
N=
ln(𝑓)
For N even, total delay = 𝑁 5 𝑓 τ = 2.5 𝑓 τ (nmos) or
2
= 𝑁 7 𝑓 τ = 3.5 𝑓 τ (cmos)
2

In all cases, delay ἀ N 𝑓 τ = ln(𝑦) 𝑓 τ
ln(𝑓)
• Total delay is minimized If 𝑓 assumes the value e. i.e: each stage is approximately 2.7
times wider than its predecessor and it is applicable for both cmos and nmos inverters.
Thus assuming 𝑓 = 𝑒, we have
Number of stages N= ln(𝑦)
And overall delay 𝑡𝑑
N even: 𝑡𝑑= 2.5 N τ (NMOS) or 𝑡𝑑= 3.5 N τ (CMOS)
N odd: 𝑡𝑑= [2.5 (N-1)+1]e τ (NMOS) or 𝑡𝑑= [3.5 (N-1)+1]e τ (CMOS)
For ΔVin which indicates logic 0 to 1 transistion of Vin.
𝑡𝑑= [2.5 (N-1)+4]e τ (NMOS) or 𝑡𝑑= [3.5 (N-1)+5]e τ (CMOS)
For ΔVin which indicates logic 1 to 0 transistion of Vin.

Super buffers

• Asymmetry of conventional inverter gives rise to significant delay problems when used
to drive large capacitive loads.
Common approach used in nmos inverter is to use super buffers an inverting type nmos super
buffer is shown in figure.

• Consider input Vin = 1, the inverter formed by T1 and T2 is turned On and thus gate of
T3 is pulled down to zero volts with small delay. So T3 is in cut off and T4 is turned On
and output is pulled down.

Dept. of ECE, M. Durga Prakash

VLSI Design

• When Vin = 0, gate of T3 is allowed to rise to Vdd. Thus T4 turned Off, T3 is made to
conduct with Vdd on its gate. The voltage applied to gate is twice the average voltage of
conventional nmos inverter.
• Doubling effective Vgs will increase current, thus reduces the delay in charging capacitor
at output, so symmetry is achieved.
• The Non-inverting type nmos inverter is shown in below figure.

BICMOS Inverter

• Bipolar transistor availability in Bicmos technology presents possibility of using bipolar

transistors as drivers at the output stage of the inverter.
• Transconductance and current/area characteristics are superior than MOS devices. so it
has high current driving capability.
• Bipolar transistor has exponential dependence of output current on base emitter voltage
which means transistor can operate with small input voltage swing compared to MOS
transistors and switches large current.
• So the bipolar transistors have better switching performance results in small input voltage
swing and switch large current.
• Switching performance of transistor driving capacitive load can be seen from simple
model.
• The time required to change output voltage Vout by an amount equal to input voltage is
given by
𝐶𝐿
Δ𝑡 =
𝑔 𝑚
Where 𝑔𝑚 is trans conductance of bipolar transistor. As 𝑔𝑚 increases Δ𝑡 decreases.
• Bipolar transistor delay has 2 main components Tin and 𝑇𝐿.
• Tin is the initial time required to charge the B-E junction of the transistor. It is time taken
to charge the input gate capacitance.

Dept. of ECE, M. Durga Prakash

VLSI Design

• 𝑇𝐿 is time taken to charge the output load capacitance 𝐶𝐿. This value is less for bipolar
by factor of ℎ𝑓𝑒.
• As BJT has higher Tin, 𝑇𝐿 is small and because of this faster charging takes place
and helps in reducing the delay.
• Combined effect of Tin & 𝑇𝐿 is in in graph. There is 𝐶𝐿 critical load capacitance below
which BICMOS driver is shown than CMOS driver.

• Delay of BICMOS is described by T = Tin+(𝑉⁄𝐼𝑑)(1/ ℎ𝑓𝑒)𝐶𝐿.

• Delay for BICMOS inverter is reduced by a factor of ℎ𝑓𝑒 when compared with CMOS
inverter.

Dept. of ECE, M. Durga Prakash

VLSI Design

Basic circuit concepts

Problems related to sheet resistance

Dept. of ECE, M. Durga Prakash

VLSI Design

Total power dissipation Pt = 0.

Problems related to standard unit of capacitance.

Dept. of ECE, M. Durga Prakash

VLSI Design

Dept. of ECE, M. Durga Prakash

VLSI Design

Dept. of ECE, M. Durga Prakash

VLSI Design

Dept. of ECE, M. Durga Prakash

VLSI Design

Dept. of ECE, M. Durga Prakash