Two-Level Logic Synthesis

-- Quine-McCluskey Method
Two-Level Logic
> What is Two-Level Logic?
> Why Two Levels?
= Universal
= Speed
= Simplicity
> Typical Two Level Circuits:
AND-OR, OR-AND, NAND-NAND, NOR-NOR
> Cost Functions in Two-Level Circuits
= Number of gates
= Number of fanin (gate inputs)

ENEE 644 2
Quine’s PI Theorem
> Review: Implicants and PIs:
Let f: Bn→B
= A minterm is an implicant if the corresponding discriminant is 1.
= An implicant with k literals has 2n-k such minterms, who share these
k literals.
= An implicant l1•••lk is prime iff l1•••lj-1lj+1•••lk is not an implicant for all
1≤j≤k.
= In the cubical representation, for an implicant with k literals,
• k = n: a vertex (minterm)
• k = n-1: an edge
z x’yz yz xyz
• k = k: an (n-k)-dimension face (plane)
xy’z
> (Quine’s Theorem [1952]) A minimal SOP must
y always
x
xyz’
consist of a sum of prime implicants.
> Minimal in terms of number of literals xy’z’ x
ENEE 644 3
Computing PIs: Tabular Method
f(w,x,y,z) = x’y’ + wxy + x’yz’ + wy’z
1. Rewrite in minterm canonical form;
f(w,x,y,z) = (wx’y’z+wx’y’z’+w’x’y’z+w’x’y’z’)+
(wxyz+wxyz’)+(wx’yz’+w’x’yz’)+(wxy’z+wx’y’z)
2. Group by number of complement literals;
3. Merge terms in adjacent groups;
Xy + Xy’ = X -- distance-1 merging
4. Repeat step 3. until no new term is created.

ENEE 644 4
f(w,x,y,z) = (wx’y’z+wx’y’z’+w’x’y’z+w’x’y’z’)+
(wxyz+wxyz’)+(wx’yz’+w’x’yz’)+(wxy’z)
wxyz x wxy
Check for:
wxz Xy + Xy’ = X
wxyz’ x wyz’
wxy’z x wy’z
x x x’y’
All prime implicants:
wx’y’z wx’y’
wxy, wxz, wyz’,
wx’yz’ x x’y’z x x’z’
wy’z, x’y’, x’z’
wx’z’ x
x’yz’ x
wx’y’z’ x x’y’z’ x More on Quine’s Method:
w’x’y’z x w’x’y’ x ↑ Easy to implement
x x
↑ Can handle don’t cares
w’x’yz’ w’x’z’
↓ inefficient
w’x’y’z’ x

ENEE 644 5
Computing PIs: Iterated Consensus Method
> Start with: SOP standard form (as compared to
the canonical form in Tabular Method)
> Goal: sum of all PIs -- complete sum
> Basic idea: xY + x’Z = xY + x’Z + YZ
-- consensus law
> Theorem: A SOP formula is a complete sum iff
= No term includes any other term;
= The consensus of any two terms either does not exist
or is contained in some term.

ENEE 644 6
Example: Iterated Consensus Method
f(w,x,y,z) = wx + x’y + xyz xY + x’Z =
= wx + x’y + xyz + wy xY + x’Z + YZ
= wx + x’y + xyz + wy
= wx + x’y + xyz + wy + yz
= wx + x’y + wy + yz

↑ Start with any SOP form

ENEE 644 7
Computing PIs: Recursive Method
> Basic idea: if F1 and F2 are complete sums, the
complete sum of F1•F2 can be obtained by:
= Expanding F1 and F2 to POS (Boole’s Expansion)
= Multiplying out F1 and F2 by distributive law
= Applying x •x=x and x •x’=0
= Eliminating all terms that are contained in others
> Example: f(w,x,y,z) = (w+x)(x’+y)(y+z)
= (wx’+wy+xx’+xy)(y+z)
= wx’y+wx’z+wyy+wyz+xyy+xyz
= wy+xy+wx’y+wx’z+wyz+xyz
= wy+xy+wx’z
ENEE 644 8
Computing the PIs: Recursive Method
> Given f(x,y,z….)= (x’ + f(1,y,z..))(x + f(0,y,z…))
(Boolean Expansion)
> C-S(f) = ABSORB((x’ +C-S(f(1,y,z..))(x + C-
S(f(0,y,z…)))

> Essentially a recursive procedure. ABSORB is a

ENEE 644 9
 Compute Complete Sum by Recursion Tree

f(w,x,y,z)=w’y’z+xyz+wyz’+wx’y’
f(0,x,y,z)=y’z+xyz = (y+z)(y’+xz) = y’z+xz
0 w 1 f(0,x,0,z) = z
y y f(0,x,1,z) = xz = (x+0)(x’+z) = xz
0 1 0 1 f(0,0,1,z)=0
f(0,1,1,z)=z
z x x’ z
0 1
f(1,x,y,z)= xyz+yz’+x’y’ =(y+x’)(y’+x+z’)
0 1
=xy+yz’+x’y’+x’z’
0 z 1 x f(1,x,0,z) = x’
f(1,x,1,z) = xz+z’ = (z+1)(z’+x) = x+z’
f(w,x,y,z) f(1,x,1,0) =1
f(x)= (x + f(0))
= (w+y’z+xz)(w’+xy+yz’+x’y’+x’z’) f(1,x,1,1) = x
= wxy+wyz’+wx’y’+wx’z’+w’y’z+x’y’z+w’xz+xyz • (x’+f(1))
ENEE 644 10
Quine-McCluskey Method
Problem: Given a Boolean function f (may be
incomplete), find a minimum cost SOP formula.
# of literals
Q-M Procedure:
1. Generate all the PIs of f, {Pj}
2. Generate all the minterms of f, {mi}
3. Build the Boolean constraint matrix B, where Bij is 1 if
mi∈ Pj and is 0 otherwise
4. Solve the minimum column covering problem for B

ENEE 644 11
Example: Quine-McCluskey Method
f(w,x,y,z) = x’y’ + wxy + x’yz’ + wy’z
wxy wxz wyz’ wy’z x’y’ x’z’
wx’y’z’ 1 1
w’x’y’z 1 minimum cover(s):
w’x’y’z’ 1 1 {x’y’, x’z’,wxy, wxz},
wxyz 1 1 {x’y’, x’z’,wxy, wy’z},
wxyz’ 1 1 {x’y’, x’z’,wxz, wyz’}.
wx’yz’ 1 1
w’x’yz’ 1
wxy’z 1 1
wx’y’z 1 1

ENEE 644 12
More on Quine-McCluskey Method
> Goal: find a minimum SOP form
> Why We Need to Find all PIs?
f(w,x,y,z) = x’y’ +wxy+x’yz’+wy’z
= x’y’+x’z’+wxy+wy’z 1. Are all terms PIs?
2. Is the form optimal?
= x’y’+x’z’+wxy+wxz 3. Is the form unique?
= x’y’+x’z’+wxz+wyz’
> How We Find Them?
= Quine’s tabular: start with minterm, the smallest I
= Iterated consensus: complete sum theorem
= Recursive: complete sum theorem

ENEE 644 13
 x’yz yz xyz
Quine’s PI Theorem z
xy’z
y x
xyz’
> Review: Implicants and PIs:
Let f: Bn→B xy’z’ x
= A minterm is an implicant if the corresponding discriminant is 1.
= An implicant with k literals has 2n-k such minterms, who share these
k literals.
= An implicant l1•••lk is prime iff l1•••lj-1lj+1•••lk is not an implicant for all
1≤j≤k.
= In the cubical representation, for an implicant with k literals,
• k = n: a vertex (minterm)
• k = n-1: an edge
• k = k: an (n-k)-dimension face (plane)
> (Quine’s Theorem [1952]) A minimal SOP must
always consist of a sum of prime implicants.

ENEE 644 14
Testability
Stuck
Stuck--at
at--0?
0?
0 0 ? 1
> Testability x y y z x’ z

> Stuck-at-0/1 Fault Model A1 A3 A2

> Redundant Gate

An area-optimal circuit must be fully testable for

ENEE 644 15