ECE380 Digital Logic

Optimized Implementation of
Logic Functions:
Multilevel Synthesis and Analysis

Electrical & Computer Engineering Dr. D. J. Jackson Lecture 14-1

Multilevel synthesis
• For the previous minimization problems, the goal
was to always find a minimum SOP or POS
realization of a given logic function
• Circuits of this type have 2 levels (stages) of logic
– For SOP form, the first level consists of only of AND gates
that connect to a second level OR gate
– For POS form, the first level consists of only of OR gates that
connect to a second level AND gate
• We assume that both true and complement forms of
the input variables are available.
• A two-level realization is usually efficient for
functions of a few variables

Electrical & Computer Engineering Dr. D. J. Jackson Lecture 14-2

1
Multilevel synthesis
• As the number of inputs increases, a two-
level circuit may result in fan-in problems
(depending on the technology used to
implement the circuit)
• Fan-in: The number of inputs to a particular
gate or circuit component
• Consider the following minimum cost SOP
expression
– f(x1,…,x7)=X1X3X6’+X1X4X5X6’+X2X3X7+X2X4X5X7

Electrical & Computer Engineering Dr. D. J. Jackson Lecture 14-3

Multilevel synthesis
• Consider implementing f in two types of PLDs:a CPLD and an
FPGA
• This CPLD implementation works because we have enough
inputs (at least 7), enough AND gates (one per product term),
and enough OR gate inputs (one for each AND gate output)
x x x x x x x
1 2 3 4 5 6 7 unused

PAL-like block
0 1
0
f
D Q

Electrical & Computer Engineering Dr. D. J. Jackson Lecture 14-4

2
Multilevel synthesis
• If we have an FPGA that has only 2-input
LUTS, we cannot implement this function
directly as written
– Since the minimum SOP form had terms with
three and four literals (requiring three- and four-
input AND gates), and
– There are four product terms needing to be ORed
together (requiring a four-input OR gate)
• The fan-in required to implement this
function is too high for an FPGA with only 2-
input LUTs

Electrical & Computer Engineering Dr. D. J. Jackson Lecture 14-5

Multilevel synthesis
• To solve this problem, the function must be
expressed in a form that has more than two
levels of logic operations
– Such a form is called a multilevel logic
expression
• Two common techniques for synthesis of
multilevel logic functions are:
– Factoring
– Functional decomposition

Electrical & Computer Engineering Dr. D. J. Jackson Lecture 14-6

3
Factoring
• Factoring utilizes the distributive property to
rewrite the expression in a form that
generally has fewer literals per term
– f(x1,…,x7)=X1X3X6’+X1X4X5X6’+X2X3X7+X2X4X5X7
– f(x1,…,x7)=(X1X6’+X2X7)(X3+X4X5)
• In this form, the function has no more than
two literals comprising each ‘term’
• It can be implemented using only 2-input
LUTs

Electrical & Computer Engineering Dr. D. J. Jackson Lecture 14-7

Factoring
0
x1 x1x6’
0
x6 1 0
0 1 x1x6’+x2x7
1
0
x2 1 0
0
0 x2x7 0
x7 f
1 0
0 1
x3
0 1
x4
0 1
x5 0 x4x5 1 x3+x4x5
1
Electrical & Computer Engineering Dr. D. J. Jackson Lecture 14-8

4
Fan-in problems
• Fan-in restrictions are not just a problem in PLDs as
in the previous case
• Fan-in is also a problem for individual logic gates
• In general, as the number of inputs to a gate
increases the propagation delay increases
• Propagation delay is the total amount of time needed
for a change at a gate input to cause a change at the
gate output
• Therefore, we may wish to limit the number of inputs
to a given gate (5 is a typical maximum)

Electrical & Computer Engineering Dr. D. J. Jackson Lecture 14-9

Fan-in problems
• Given the function
– f=X1X2’X3X4’X5X6+ X1X2X3’X4’X5’X6
• The direct solution for this would require 2
six-input AND gates and 1 two-input OR gate
(plus appropriate NOT gates)
• Factoring the function to the following form
– f= X1X4’X6(X2’X3X5+ X2X3’X5’)
• Gives a solution requiring 2 three-input AND
gates, 1 two-input OR gate, and 1 four-input
AND gate

Electrical & Computer Engineering Dr. D. J. Jackson Lecture 14-10

5
Fan-in problems
• Factor the following expression so that the
solution requires only 2-input AND and OR
gates
• Hint: The solution will require 4 AND and 2
OR gates (plus NOT gates)

– f(x1,…,x7)=x1x2’x4’x5+x1x2’x6x7’+x3’x4’x5+x3’x6x7’

Electrical & Computer Engineering Dr. D. J. Jackson Lecture 14-11

Impact on wiring complexity

Electrical & Computer Engineering Dr. D. J. Jackson Lecture 14-12

6
Functional decomposition
• Complexity of a logic circuit, in terms of
wiring and logic gates, can often be reduced
by decomposing a two-level circuit into
subcircuits
– One or more subcircuits implement functions that
may be used in several places to construct the
final circuit
• A single two-level logic expression is replaced
by two or more new expressions
– The new expressions are combined to define a
multilevel circuit

Electrical & Computer Engineering Dr. D. J. Jackson Lecture 14-13

Decomposition example
• Consider the following expression
– f(w,x,y,z)=xyw’+x’z+y’z
• In this form, the function requires 1 three-
input AND gate, 2 two-input AND gates, and
1 three-input OR gate
• COST=4 gates + 10 inputs = 13
• COST=19 if NOT gates (and their inputs) are
included
• Rewrite f into the following form
– f(w,x,y,z)=(xy)w’+(x’+y’)z
• Let g(x,y)=xy and note that g’=x’+y’
Electrical & Computer Engineering Dr. D. J. Jackson Lecture 14-14

7
Decomposition example
• The function f becomes
– f(g,w,z)=gw’+g’z
• The circuit would be the following with a cost of 16
(including NOT gates and their inputs)

Electrical & Computer Engineering Dr. D. J. Jackson Lecture 14-15

Practical issues
• Functional decomposition is a powerful tool
for reducing the complexity of logic functions
• It can also be used to implement general
logic functions that have built-in constraints
– For example, in PLDs, it is necessary to ‘fit’ a
desired logic circuit into logic blocks that are
available on these devices
– Available logic blocks are a target for the
decomposed subfunctions that are then used to
form larger functions
• CAD tools make extensive use of the concept
of functional decomposition

Electrical & Computer Engineering Dr. D. J. Jackson Lecture 14-16