Faculty of Engineering

Electrical Engineering Department

Communications and Electronics Program

Logic Circuit Design

Class Notes
EE242

SPRING 2015
 LOGIC CIRCUIT DESIGN EE242

Term: SPRING 2015

Instructor: Dr. M. El-Banna, Room: EE-Building 4th floor
Classes : Group1 SUN 10:10 – 11:40
Group2 SUN 2:00 – 03:30

Topics to be covered:
Chapter 1: Implementation of Logic Functions
Chapter 2: Design of Combinational Circuits
Chapter 3: Sequential Circuits
Chapter 4: Design of Synchronous Sequential Circuits
Chapter 5: Introduction to Counters, Registers, and HDL language

Text Book: Fundamental Logic Design, Thomson, Charles Roth

Further Reading: Contemporary Digital Design, Johnson and Karim

Grading Policy:
Midterm 30
LAB 30
Final 90
 ILOs of the Logic Circuit Design EE242

1. Comprehend and use the main blocks of combinational circuits, MUXs, ROMs, PLAs, PALs,
Decoders and Encoders
2. Design combinational circuits using different blocks.
3. Carry out a project using the NI- LabVIEW software package to design and test combinational
circuits, if time allows.
4. Differentiate between combinational and sequential circuits
5. Review all types of Flip Flops used in sequential circuits and represent their functions by state
diagrams.
6. Convert verbally stated design problems into state diagrams and hence state tables.
7. Differentiate between Mealy and Moore finite state machines.
8. Follow up the design procedure of sequential circuits starting from implication tables through
partition tables, state transition tables, excitations maps and eventually the hardware
implementation.
9. Use the Verilog/VHDL language to design come known logic circuits, combinational and/or
sequential
10. Comprehend and design synchronous and asynchronous counters
..

CHAPTER
FOUR
Implementation of
Logic Functions _ _ _ _ ----I

4.1 Introduction The design mechanism of combinational logic circuits is usually a

multi-step process. The realization, and the subsequent minimiza­
tion, of the logic function is not the end of the design. We are
already familiar with the various schemes for coming· up with the
reduced logic function either in the SOP or in the POS format.
These forms can be translated easily into either a familiar AND­
OR or OR-AND pattern of logic circuits. However, we have also
seen in the last chapter that digital ICs have several practicallimi­
tations that may affect the implementation of circuits. These
.
include the fan-in and fan-out limitationS and the fact.
that ICs are
more frequently available in the NAND and NOR form than in the
AND and OR form. NAND and NOR gates are easier to realize
with electronic components and are, therefore, the basic ingredients
used in ali of the logic families. Consequently, it is important for the
designer to be familiar with the techniqueS for translating the
reduced function so that either NAND gates or NOR gates may be
used.
Combinational circuits may be realized using a standardized
combinational unit called a multiplexer (MUX). In the MUX
some of the input variables are used as input selectors for the unit
and the remaining variables are entered as data inputs. Two other
devices, read-only memories .(ROMs) and programmable logic
arrays (PLAs), are also frequently used to implement combina­
tional networks. This chapter will explore the possibilities of using
only one type of gate or one of the modules-MUX, ROM, PAL, or
PLA-for the realiz~tion of combinational circuits. Such exploration
is extremely useful ~use most· designers usually choose to use
only one type of basic gate unless there is a particular reason for
doing otherwise. After studying this chapter, you should be able to:
o Design combinational circuits using only NAND gates;
97
98 CHAPTER FOUR Implementation of Logic Functions

o Design combinational circuits using only NOR gates;

o Design combinational circuits using single- or multi­
level multiplexers (MUXs);
o Design combinational circuits using read-only memories
(ROMs);
o Design combinational circuits using programmable
logic arrays (PLAs);
o Construct a complex circuit using the outputs of a
known functional unit.

4.2 Universal Logic In practice many logic circuits are built using only NAND and
NOR gates because the basic gates in some of the logic families such
Elements as TTL and CMOS are NAND and NOR, respectively. NAND
and NOR gates are considered universal logic elements since they
both can be easily manipulated to obtain all possible logic func­
tions. This simplification follows directly from, Boolean theorems
that we have discussed in the earlier chapters.
A close inspection of the truth table of these two functions, as
described in Section 1. 7, reveals that the NAND and NOR opera­
tors are duals of each other. Recall also from Chapter 1 that the
dual of a Boolean expression is obtained by replacing every OR
with AND, AND with OR, 0 with 1, and 1 with O. Six of these dual
properties are listed as follows:

NAND NOR

l.a·O=l a+l=O
2. a:l = a a + 0 =a
3. a' a = a a+a= a
4. a' h = a+ b a + b = a' b
5. a. h = a + b a+ b = a • b
6. a' = a • b a+ b= a+ b

All of the logic functions may be generated using these properties of

NAND and NOR logic. The corresponding NAND and NOR logic
circuits for various functions are shown in Figure 4.1.
The circuits of Figure 4.1 show how NAND-and NOR gates may
be cascaded to form each of the logic functions NOT, AND, OR, .
and X-OR Since either NANDs or NORs may be used to imple­
ment all of the logic operations, designers may prefer to use only
. NANDs or only NORs in order to decrease the inventory oflSpare
parts. One of the methods by which to realiZe this is the brute force
 4.2 Universal Logic Elements 99

FIGURE 4.1 Logic Functions

Using NANDs and NORs.

Gate NAND Circuit NOR Circuit

NOT A-QJ-A A-£D-X

A
AND

OR A+B ~=L>O-A+B
B

X-DR A A
B B

scheme, where each of the logic operations of the Boolean function

is replaced by the corresponding NAND/NOR circuit. Note, how­
ever, that restricting the number of inputs to the gates will not cause
any major problem if proper use of the involution and DeMorgan's
laws are made.

EXAMPLE4.i SOLUTION
Implement A El) B using only There are two different ways to implement this function: (a) using NAND
NAND gates. equivalents ofan X-OR gate and of a NOT gate, and (b) using the NAND
equivalents of either the SOP or the POS tenns.
a. The first possibility results in the circuit of Figure 4.2.
h. Otherwise, A El) B can be ~ressed in the SOP fonn as AB
+ All Consequently: the circuit appears as in Figun: 4.3. The
circuit of Figure 4.3 can be reduced further since X = X. There­
fore, the circuit reduces to that of Figure 4.4. Note also that the
function could be expressed in the POS form. Consequently, A El)
B = (A + B)(A + B), which leads to another variation of an X­
NOR circuit as shown in Figure 4.5. The circuit of Figure 4.5
100 CHAPTER FOUR Implementation of Logic Functions

FIGURE 4.2

A--'--i

B--+--;

X-OR NOT

FIGURE 4.3

A~~--~-------;~

B -!---.-------'-------i--i

D--A(±)B

NOT AND OR

FIGURE 4.4

A--~----------------------~

B--r~------~------------~

may be reduced further by making use of the law of involution.

The resulting circuit is shown in Figure 4.6.

The first and the third fonns, as shown in Figures 4.2 and 4.4, require five
NAND gates each, and the fifth, as shown in Figure 4.6, requires a total of
six. NAND gates.
 4.3 Function Implementation Using NANDs 101

FIGURE 4.5
\

A---e>--...

NOT OR

AND
B -......

FIGURE 4.6

B - -.....

4.3 Function It is quite easy to realize any SOP function using two levels of
NAND gates. This method makes use of the fact that comple­
Implementation Using menting a function twice returns the function to its original form.
NANDs This result is achieved in two steps:
1. The function is complemented by complementing the
ANDed terms and replacing the OR signs with AND signs.
2. The original function is then recovered by complementing
the complement function .. "

It is not necessary to perform this operation each time a NAND

realization is required. SOP forms always assume the same two­
level NAND form.
The output of a NAND gate is also equivalent to the ORed out­
put of the complements of the input. This statement follows directly
from DeMorgan's theorem. Consequently, we are led to the follow­
102 CHAPTER FOUR Implementation of Logic Functions

ing set of rules for obtaining the output function of a multi-level

NAND circuit:
Rule 1. Consider the gate from which the output signal is derived
as the first level, the preceding gate as the second level, and so on.
Rule 2. In odd-numbered levels the NAND gates perform OR
operations. All ungated input variables entering the odd-level
NAND gates will appear complemented in the final expression.
Rule 3. In even-numbered levels the NAND gates perform AND
operations. All input variables entering the even-level NAND gates
will appear uncomplementedin the final expression.

EXAMPLE 4.2 SOLUTION

Using only NAND gates, implement f = ABC + ABc + CD
the function given by
f(A,B,C,D) = ABC + ABc + CD which yidds
=~=~=

f = ABC' ABc· CD

The final circuit, therefore, is obtained as shown in Figure 4.7. The circuit
requires three three-input NAND gates and one two-input NAND gate
provided Band C inputs are also available in the complemented form.

FIGURE 4.7 A \r,

B
C

A
h
I"'" f(A,B,C,D)
C

C
D
\.

t,

4.4 Function The implementation of an SOP function using only NOR gates is
Implementation Using possible only if the function is first converted to the equivalent POS
form. The process includes the following steps:
NORs
1. Plot the function on a K-map and obtain the comple­
mented function by grouping all zeros.
 4.4 Function Implementation Using NORs 103

2. Expand each of the ANDed terms by using DeMorgan)s

theorem.
3. Complement the whole Boolean expression.

NOR realizations of SOP functions always have the same two-level

structure. Steps 1 through 3 should be followed until the designer is
confident of the result.
DeMorgan's theorem may be used to interpret the NOR opera­
tion as well. The output of the NOR gate is equivalent to the
ANDed output of the complements of the inputs. Rules for the
interpretation of multi-level NOR circuits are listed as follows:
Rule 1. Consider the gate from which the output signal is derived
_L·C as the first level, the preceding gate as the second level, and so on.
Rule 2. In odd-numbered levels the NOR gates perform AND
operations. All ungated input variables entering the odd-level NOR
gates will appear complemented in the final-expression.
Rule 3. In even-numbered levels the NOR gates perform OR
operations. Input variables entering the even-level NOR gates will
appear uncomplemented in the final Boolean expression.

EXAMPLE 4.3 SOLUTION

Using only NOR gates) implement The minterms are plotted in a four-variable K-map and the corresponding
the function given by zeros are grouped as shown in Figure 4.8. This gives
j(A)B)G,D) = ~m(O)2)4)5,8,1O,13) J= CD + BC +BD + ABD
=c+D+B+c+B+ +A+B+D
Therefore,
j=C+D+B+C+B+D+A+B+D

FIGURE4.B ~
A

-
1 1 W 1
r0­

O 1 1 0

0 0 0 01
,.......,
I­
1 0 0 1
'----y----'

The equivalent NOR circuit, therefore, may be obtained as shown in Fig­

ure 4,9. It requires three two-input NOR gates, one three-input NOR gate,
104 CHAPTER FOUR Implementation of Logic Functions

and one four-input NOR gate provided the inputs are als9 available in the
complemented form.

FIGURE 4.9 c----\

5----1
"---­

8----\
C----I
x..r--- f

B----\

5----1"---­

A----;
0----1
"---­

EXAMPLE 4.4 SOLUTION

Write the output expression for the FIGURE 4.10

multi-level circuit shown in

Figure 4.1 O. A-----------\
2
B
4
C
f(A,B,C,O,E,F)

o
E
3
F----------~

This is a three-level circuit. Therefore, the involved operations may be

summarized as follows: .,
a. Gate 1 performs tge OR operation;
h. Gate 2 performs the OR operation, and A must appear uncom­
plemented;
c. Gate 3 performs the AND operation, and F must appear uncom­
plemented;
 4.5 Function Implementation Using MUXs 105

d. Gate 4 perfonns the OR operation, and both Band C must

appear complemented;
e. Gate 5 performs the AND operation, and both D and E must
appear complemented.

Therefore,

f(A,B,C,D) = [(D' E)' F] + [A + (8 + C)] = DEF + A + B + C

Function output expressions may also be determined by starting at the

inputs and making repeated use of DeMorgan's theorem. This latter tech­
nique is best when there is a mix of NOR and NAND gates in the circuit.

4.5 Function A multiplexer (MUX), known also as a data selector, is a combinational

network that has up to 2n data inputs, n control inputs, and an out­
Implementation Using put line. Commercial MUXs are limited to values of n of 1 through
MUXs 4. An additional input is available that allows cascading of mul­
tiplexers to obtain higher-order devices. The MUX allows the selec­
tion of one of the 2n data inputs as the device output. This selection
is made by the control lines. A block diagram of a MUX with eight
data input lines, Do, DI , D2, D3, D4, DSl D6, and D7, is shown in Fig­
ure 4.11[a]. Most MUXs are provided with at least two additional
lines: Jfor the complemented output and E for enabling the device.
The internal circuit configuration of the corresponding MUX is
shown in Figure 4.l1[b]. For every 2n inputs the MUX has exactly n
control lines. By applying appropriate signals to the control lines,
anyone of the data lines may be selected. For example, when 121110
= 011, the D3 input is routed to the output provided E = O. Note
that whenever E = 1, the MUX is completely disabled; that is,
regardless of the control variables or the data inputs, the output is O.
The enable input allows several of these devices to be cascaded
together.
A MUX with 2n input lines and n selection lines (such a device is
also referred to as a 1-of-2n MUX) may be wired to realize any
Boolean function of n + 1 vaI'iables. This fact will be illustrated by
implementing the function f(A,B,C,D) = 2:m(0,2,4,5,6,8,1O,13)
using a 1-of-8 MUX. A 1-of-8 ~X has three control inputs where
all but anyone of the 'input variables may be entered. For example,
we may consider A, B, and C inputs as the three control inputs, /2,
/1' and /0, respectively. The technique, therefore, consists of deter­
mining the function output in terms of the fourth input, D, for every
possible combination of the control inputs. The values so obtained
are then entered at the respective data inputs of the MUX. The
truth table for the said function may then be reorganized as shown
106 CHAPTER FOUR Implementation of Logic Functions

FIGURE 4.11 Eight-Input Enable

MUX: [aJ Block Diagram and [b]
Circuit.
----100 E
----10 1
-----\0 2
-----\ 0 3 f
Data inputs 1-of-8 } MUX output
---104 MUX f
---105
---10 6

'---y------'

Data select
[a]

00­
)­
r--­
0 1­
}­
f--­

O2­
)­
~
0 3­ ~ )~ 7

0 4­
)i

0 5­
)­
06

0 ( )
~

0 7­
)­ '.

{'. ­

[b]

\
 4.5 Function Implementation Using MUXs 107

in Figure 4.12. The entries under thef(D) column are determined

by comparing the entries of the column D with that of the column!
For any combination of A, B, and C, the following are true:
1. Iff = 1 irrespective of D, thenf(D) = 1.
2. Iff = 0 irrespective of D, thenf(D) = O.
3. Iff = xwhenD = x, thenf(D) = D.
4. Iff = xwhen D = x, thenf(D) = D.

FIGURE 4.12 Truth Table for A B C D f f(D)

f(A,B,CJJ)
13).
= Im(0,2,4,5,6,8,10, 0 0 0 0
1
1
0
l5 ,
0
v
0 0 1 0 1 0 (/)
l5 i)
1 0
0 1 0 0 1
\)
1 D
1 1
0 1 1 0 1
l5
I --­'P
1 0 ~
0
1 0 0 0 1 (f)
l5 0
1 0
1 0 1 0 1 0 Q
1 0
i5 0
1 1 0 0 0
D
1 1
1 1 1 0 0
0
1 0

The function, therefore, is implemented as shown in Figure 4.13[a].

The eight values of f(D) are fed respectively into data inputs Do
through D7• Consequently, for any combination of the three control
inputs, f(D) would actually appear at the MUX output. The
implementations of the same function for the other three combina­
tions of the control inputs-B, C, and D; A, C, and D; and A, B, and
D-can also be obtained in like-manner. The corresponding MUX
configurations are shown in Figures 4.13[b-d]. The designer might
even change the order of the control inputs, resulting in a total of 24
different circuit configurations. In addition, if one of the control
variables is available only in coIJlplemented form, the variable may
be used without invertiiig and the inputs rearranged accordingly
(see Problem 3b).
It is quite obvious that the use of MUXs provides the designer
with numerous choices. Consequently, it is necessary to consider
each of the solutions to determine which is the optimum. In Figure
108 CHAPTFJl FOUR Implementation of Logic Functions

FIGURE 4.13 MUX

(j
Implementation off(A,B,C,D)
= 2:m(O,2,4-6,8,1O,13): [a] 121tlo
= ABC, [b] 121lo = BCD, [c] 121110 Do E Do E
= ACD, and [d] 121110 = ABD. 01 01
O2 O2
03 03 f
1-of-8 f 1-of-8
04 MUX 04 MUX
05 05
0 06 A 06
0 07
07
12 I, 10
0
A B C B C 0

[a] [b]

B E Do
E
Do
01

01
O2

O2 03

f f
03 1-of-8
1-of-8 04 f
f
04 MUX 05 MUX

05 06
06 07

0 7 12 0
B c
0 A C 0 A B 0
[c] [d]

4.13[a], lJ is seen to be tied to five of the data inputs. The gate that
provides lJ must then have a fan~out of at least five. If we limit our­
selves to the four choices of Figq.re 4.13, it is apparent that Figure
4.13[d] provides the most preferable circuit, because the Cvariable
needs to be fed directly to only two of the data inputs. Example 4.5
illustrates the meChanism of obtaining a multi-level multiplexer
circuit.
 4.5 Function Implementation Using MUXs 109

EXAMPLE 4.5 SOLUTION

Implement the function a. The function may be expressed in the SOP form and then
regrouped as
f(A,B,C,D,) = ~m(3,4,8---10,13-15)
f(A,B,C,D) = ABCD + ABCD + ABCD + ABCD + ABCD
a. by using a l-of~4 MUX and a
few assorted gates,
+ ABCD + ABCD + ABCD
= AB(CD) + AB(CD) + AB(CD + CD + CD)
h. by using l-of-2 and l-of-4
MUXs in two levels.
+ AB(CD + CD + CD)
= AB(CD) + AB(CD) + AB(C + D) + AB(C + D)
Accordingly, the resultant circuit is obtained as shown in Figure
4.14.

FIGURE 4.14:

C E
Do
o
01 7
1-of-4
O2 MUX
03 f ((A,B,C,D)
C
i5
11 10

A B

h. Also,
f(A,B,C,D) = AB[C(O) + C(D)] + AB[C(D) + C(O)]
+ AB[C(l) + C(D)] + AB[C(D) + C(l)]
This implies that a two-level MUX circuit would be able to gen­
erate this function. The firSt level of a 1-of-2 MUX essentially
eliminates the need for discrete gates. The resultant circuit is
obtained as shown in Figure 4.15. However it can be shown that
C(O) + C(D) == C(l) + C(D)
and
C(D) + C(O) = C(D} + C(l)
Note also that the MUXs usually are provided with an additional
output for providing the complemented result. Consequently, two
of the first-level l-of-2 MUXs may be removed. The resulting
reduced multi-level MUX circuit is obtained as shown in Figure
4.16.
110 CHAPTER FOUR Implementation of Logic Functions

FIGURE 4.15

D [j 0 t5 D 0

01 Do D1 Do D1 Do D1 Do
1-of-2 1-of-2 1-of-2 E 1-of-2
C /0 E C /0 E C /0 C /0 E
MUX MUX MUX MUX
f 7 f 7 f f f

A
1-of-4 MUX E
a
f 7

f(A,a,C,D)

FIGURE 4.16

0 Do
r
1-of-2
MUX
f
D D1

/0 Do
D1 f
C
1-of-4
I D2 MUX
!(A,a,C,D)

E 03
/1 10
t5 Do
1
1-of- 2
MUX A a
f
0 Dl

/0

 4.5 Function Implementation Using MUXs 111

MUXs can also be applied in a more brute force manner. This

technique involves 1-of-2n MUXs for functions of n variables, while
Example 4.5 used l-of-2 n - 1 MUXs to implement a function. The
function values from a function's truth table are transformed
directly to the inputs of the MUX. The n variables are treated as
control inputs to the MUX. Figure 4.17 shows the truth table for a
three-variable function and the corresponding 1-of-8 MUX imple­
mentation. For two functions ofthe same variables, two MUXs may
be used as shown in Figure 4.18.

FIGURE 4.17 Realization of o

f(A,B~C) =
i":m(1-3 J5J6) Usihg
l-of-8 MUX.
E
Do
01
A B C f O2
1-of-8
0 0 0 0 03 f
MUX
0 0 1 1 04 f f
0 1 0 1 05
0 1 1 1 06
1 0 0 0 07
1 /2 11 /0
0 1 1
1 1 0 1
1 1 1 0
ABC

FIGURE 4.18 Realization of

Ji(A~~C) = i":m(I,3,6,7) and
fo(A~~C) = i":m(O,3-5) Using Two
l-of-8 MUXs.

1 0 fo

! ~
E E
..­ Do Do
01 ~ 01
.....­ O2 ~.D2

A B C t1 to 0 3 1-of-8 t.b­ 03 1-of-8 to­

~ 0 4 MUX f­ 04 MUX f­
0 0 0 0 1 - 05 05
0 0 1. 1 0 06 ~ 06
0 1 0 0 0 07 '--­ 07
0 1 1 1 1 /2 11 /0 /2 11 /0
1
1
0
0
0
1
0
0
1
1 A 1 I
1 1 0 1 0 B
1 1 1 1 0 C
112 CHAPTER FOUR Implementation of Logic Functions

The technique just discussed is not as efficient as the one before

in terms of MUX size, but it is certainly easy to implement and is a
good way to introduce another device that may be used to imple­
ment combinational logic functions. If we put the circuit implemen­
tation of Figure 4.18 in a box, we could show the circuit symboli­
cally as illustrated in Figure 4.19. Such a representation is called a
black-box representation since the user is not required to know about
the details of the internal logic. Not knowing what is in the box in
Figure 4.19, we might explain the circuit action by simply saying
that ABC forms an address, and iI and Jo are what is stored there.
This is precisely the explanation of the operation of a device with
which we can implement multiple functions of a set of variables.
This device is called a read-only memory (ROM). It can be considered
as a set of storage cells with every cell having a value for each of the
multiple functions. These .ROMs can be programmed (Is and Os
applied to the inputs of each of the function's MUX inputs) by the
manufacturer or by the designer if he or she has available blank
ROM chips and an appropriate programmer. The information
stored in the ROM remains there permanently. The ROM equiva­
lent of the logic of Figure 4.18 is shown by the block diagram of Fig­
ure 4.19. When power is applied,iI andJo have the same values as
before power was removed.

FIGURE 4.19 Black-Box

Representation of the Circuit of
Figure 4.18.
Inputs 1:
12

11 ROM
III Outputs

10 fo

ROMs come in many sizes. The sizes are determined by the

number of storage cells (corresponding to the number of MUX
inputs) and the number of bits stored in each cell (the number of
functions that can be implemented). In Figures 4.17 and 4.18 there
are three variables that allow addressing 23 storage cells in our
MUX-implemented ROM. In general, n variables would require 2n
storage cells and correspond to· n address lines. Each storage cell
would have one bit (lor 0) for each function of the n variables.
Commercially available ROMs come in many sizes. They may
be listed as 2K X 1 or 2K X 8, meaning 2048 storage locations one
bit wide in the first case and 2048 storage locations eight bits wide
in the second case. The first would allow implementing a function of
up to 11 variables, the second up to eight functions of 11 variables.
The designer selects a ROM of adequate size to implement the
desired function. .
 4.5 Function Implententation Using MUXs 113

The following section discusses ROMs in more detaiL For most

function generation applications, ROMs can be vi&ualized as a set
of MUXs-one for each function of the input variables (addresses).

EXAMPLE 4.6 SOLUTION

Obtain a multi-bit shifter (see The characteristics of the shifter could be summarized as follows:
Example 2.4 for definition) thathas
an (n + 2)-bit input, x, an n-bit' if d = 1, s = 1, and E = 1 (left-shift)
output,), and three control inpu~: if d = 0, s = 1, and E = 1 (right-shift)
)i=

s> d, and E. You may use only Xi if s = 0 and E = 1 (no-shift)

MUXs for the deSign.'

o ifE=O

where 0 < i < n - 1.

A I-of 4 MUX could be used corresponding to each bit of). The vari­
ables sand d can be fed as its selectors, and E as the enable input. The
inputs Xi) Xi- hand xi+ 1 could be introduced at the MUX data inputs, and
they could be suitably selected as the shifter output. The multi-bit shifter is
obtained accordingly as shown in Figure 4.20. Whenever E = 0, the
MUXs are disabled and the output becomes zero.

FIGURE 4.20

Xn Xn-l Xn -2 x, Xo X-l

D2 D1 Do
S 11 ~ "-~~-'
11
1-of-4 MUX E 1-of-4 MUX E E

d 10 10
.f f f f

Yn-l Yo
114 CHAPTER FOUR Implementation of Logic Functions

4.6 Function A read-onry memory (ROM), as the name implies, is intended to hold
fIxed information that can only be read, not altered. The primary
Implementation Using use of the ROM is to provide a means for storing binary informa­
ROMs tion. The storing is done during the fabrication of the ROM and
may not be altered without undergoing a signifIcantly involved pro­
cess. The same is true for a combinational network that has been
designed, fabricated, tested, and encapsulated with only the inputs
and the outputs available. The ROM has become an important
part of many digital systems because of the ease with which com­
plex functions such as code conversion, program storage, and char­
acter generation can be implemented. The chip count of circuits, for
which the access time of the ROM is not a restriction, may be
greatly reduced by using ROMs.
A 2m X n ROM is an array of memory cells organized into 2m
wordsofn bits each, as shown by the block diagram of Figure 4.21.
Such a ROM is accessed by means of m address lines and the stored
information is retrieved via a total of n data-out lines, one for each
bit of the word. The ROM corresponds to a combinational network
with n outputs, where each of the outputs is associated with up to 2m
different minterms. A ROM may be provided with one or more
chip-select lines to permit cascading smaller ROMs to form a ROM
with more words (allowing implementation of functions of more
variables).

FIGURE 4.21 Block Diagram of Address lines -+/--:..~

Em
a ROM. 2,"l xn 1---7'/'--+ Output lines
ROM n
Chip-select --1'---+1

As indicated in the previous section, a ROM is a combinational

circuit. A ROM can be implemented by using only diodes, bipolar
transistors, or MOS transistors. Although the diode matrix ROM
no longer represents the current ROM technology, it serves as a
simple model to show the basic concept. Figure 4.22 shows a simple
diode ROM, where the row and column lines are interconnected
via diodes placed at the respective intersections. The absence or
presence of a diode indicates that the corresponding row and col­
umn intersection is programmed with a 1 or a O. If the output is
buffered with inverterS; the converse is true. The output for the zth
address depends on the ORing diodes connected to that line. For
example, if A~lAo = 100, the output will be DaD7 ... DI =
01000001. For A~iAo = 100 input, a low is produced on the
decoder output numbered 4. This low. and the pull-up resistor for­
ward biases each of the diodes connected to this row and pulls down
 4.6 Function Implementation Using ROMs 115

FIGURE 4.22 8 X 8 Diode

ROM.

Pull-up
resistors

01
O2
A2 12
3~8jine 03
A1 . 11 decoder 0
4 OR matrix
Ao 10

Outputs

the corresponding colW1Ul outputs to a low. In the absence of a

diode, a high is maintained at the output due to the pull-up resis­
tors. The capacity of a ROM usually is quoted as the number of
possible intersections in the matrix, for example, 8 X 8 = 64 bits =
(1/16)K bits in this case. A total of 1024 intersections is usually
referred to as lK bits. Examples 4.7 and 4.8 illustrate the use of
diode ROMs in combinational problems.

SOLUTION
.. Use ROM to realize the implemen~ The function truth table is obtained as shown in Figure 4.23. A 2-4 line
tation of the integer function decoder would
. be .sufficient to decode the numbers 0, 1, 2, and 3, and it is
apparent that a maximum of five output lines are needed to represent the
forO < x <3
cube of the largest number. The two select lines, Ao and Ah can be used to
. select anyone of these four outputs. .
116 CHAPTER FOUR Implementation of Logic Functions

FIGURE 4.23 x f(x} f(x} in Binary

0 0 00000
1 1 00001
2 8 01000
3 27 11011

The resultant diode matrix implementation of il ROM, therefore, is

obtained as shown in Figure 4.24.

r---~--~---'--~~-------+V
FIGURE 4.24

Pull-up
resistors
E

11 0 1 [)-4I!----t-4t---Hl---+___~f__--+----
2-4 line
decoder
~ 02O'--+---~.-~~~~--+----

output

EXAMPLE 4.8 SOLUTION

Using a minimal ROM FIGURE 4.25

implementation, design a seven­
segment-to-BCD code converter.
The seven-segment display device
cons~ts of seven LEDs, as shown in
Figure 4.25, arranged in such a way
that they could be used for
displaying data.

The seven-segment-to-BCD truth table is obtained as shown in Figure

4.26. The input entries are selected in such a way that the output would be
displayed only if the corresponding LED segments are lighted. For exam­
ple, when all of the seven segments are turned on, the display will be an 8.
 4.6 Function Implementation Using ROMs 117

FIGURE 4.26
Inputs Outputs
P Q R S T U V D C B A
1 ·1 1 1 1 1 0 0 0 0 0
0 1 1 0 0 0 0 0 0 0 1
1 1 0 1 1 0 1 0 0 1 0
1 1 1 1 0 0 1 0 0 1 1
0 1 1 0 0 1 1 0 1 0 0
1 0 1 1 0 1 1 0 1 0 1
0 0 1 1 1 1 1 0 1 1 0
1 1 1 0 0 0 0 0 1 1 1
1 1 1 1 1 1 1 1 0 0 0
1 1 1 0 0 1 1 1 0 0 1

Note, however, that there could be up to two choices for displaying a 1:

either U and Tor Q and R.
FIGURE 4.27
A direct implementation of this table requires a 27 X 4 = 512-bit
ROM. Note that the partitioned table (with inputs P, (b R, S, and T), as
shown in Figure 4.27, would require only seven of the possible 25 = 32
p Q R S T Z Y X combinations. 'This situation suggests that a reasonable improvement is
1 1 1 1 1 0 0 0 possible if the designer is willing to cascade at least two smaller ROMs.
0 1 1 0 0 0 0 1 One of these ROMs should have at the least three outputs-z, Y, and X­
1 1 0 1 1 0 1 0 since 23 > 7. The output of the first ROM-Z, Y, and X-can be fed along
1 1 1 1 0 0 1 1 with the other two inputs-U and V-into the second ROM. However, care
1 0 1 1 0 1 0 0 must be taken in organizing the second ROM so that its output becomes
0 0 1 1 1 1 0 1 equivalent to that of Figure 4.26. Accordingly, the compressed truth table
1 1 1 0 0 1 1 0 of Figure 4.28 is obtained such that it incorporates the same logic as that of
Figure 4.26. The seven inputs have now been replaced by only five inputs,
where the first three are functions of P, (b R, S, and T and the other two are
U and V themselves.
Outputs Z, Y, and X are realizable using a 96-bit (3 X 25) ROM. A
second ROM can be used where Z, Y, x, U, and V are the inputs. The
second ROM size is 4 X 25 = 128 bits. Therefore, the total ROM size
requirement is reduced to only 96 + 128 = 224 bits. 'This solution reduces

FIGURE 4.28
z y X U V D C B A
0 0 0 1 0 0 0 0 0
0 0 1 0 0 0 0 0 1
0 1 0 0 1 0 0 1 0
0 1 1 0 1 0 0 1 1
0 0 1 1 1 0 1 d 0
1 0 0 1 1 0 1 0 1
1 0 1 1 1 0 1 1 0
1 1 0 0 0 0 1 1 1
0 0 0 1 1 1 0 0 0
1 1 0 1 1 1 0 0 1
118 CHAPTER FOUR Implementation of Logic Functions

the ROM size to about half of the originaL The ROM implementation of
the circuit, therefore, is given by the multi-level circuit of Figure 4.29.

FIGURE 4.29 p
Q

jf
R

Inputs
Outputs
B
A

-
Although the diode matrix serves to demonstrate the ROM con­
cept, ROMs are presently manufactured using bipolar and MOS
transistors, as shown in Figure 4.30. The presence of a connection
from a row line to either a transistor base or a MOSFET gate repre­
sents a logic 0, and the absence of such a connection represents a
logic 1. For economical reasons MOS ROM is preferred to bipolar
ROM for large numbers of bits. Access time for bipolar ROM,
however,is much less than that ofMOS ROM.

FIGURE 4.30 [a] Bipolar m X n

ROM and [b] MOS m X n ROM.

--~t---+ Vee -----41-- +Vee

~--~--~---+--. ··-4-----jr--Row1 -+----t--- Row 1

_4,------jI-- Row 2 -4-~r--'---~-"'-'---4---Row2

-4--~-Rowm -+----f--- Row m

Column 1 Column 2 Column n Column 1 Column 2 Column n

[a] [b1
 4.7 Function Implementation Using PLAs and PAls 119

4.7 Function Examples 4.7 and 4.8 illustrated how a ROM may be used to
implement SOP logic expressions. There is another aspect of the
Implementation Using ROM that deserves attention, however. A ROM consists of a level
PLAs and PALs of AND gates, which constitute the decoder part, followed by a sec­
ond level of OR gates (made up of diodes or transistors), which con­
stitute the encoder section. A ROM may be thought of as a
programmable array of logic gates. With this array of AND and
OR gates, every combination of minterms of the input variables
(addresses) can be formed. This flexibility is costly in that, when
implementing complex functions, not all minterms are necessary to
realize a given expression. For example, a ROM that processes 12
variables requires a total of 4K byte (eight bits are called a byte)
memory. For example, such an arrangement is needed for the Hol­
lerith code conversion circuit that has up to 12 input variables, but
has only 96 eight-bit output combinations of these variables. This
situation implies that 4000 out of 4096 bytes will remain unused.
Such waste can be eliminated by the use of a programmable logic array
(PLA).
A PLA consists of an array of AND-OR logic along with inverters
that may be programmed to realize the desired output. In essence a
PLA may be regarded as being made up of two separate ROMs: an
AND ROM and an OR ROM. A typical PLA configuration is
shown in Figure 4.31 in a 12 X 32 X 8 format. The circuit consists

FIGURE 4.31 12 x 32 x 8 PLA

Using Diodes.

<f-4- ­ AND array ---+. +-- OR array --t>

+v .---e-----+----i II

. . .... . . ..

. .....

~---.,v~-~

Inputs Outputs
120 CHA.PTER FOUR Implementation of Logic Functions

of an initial AND array, which can implement anyone or more of

the 32 product teIlllS of up to 12 variables. The inclusion of an
inverter with each input variable allows any minterm to be formed.
A PLA may be used as a Boolean function generator in much the
same way as a ROM. As a simple example, Figure 4.32 shows a
small PLA layout with six inputs, twelve product teIlllS, and four
outputs. The dots in the matrix of the top section can be thought of
as AND inputs and those on 'the bottom part can be interpreted as
OR inputs to generate the outputs. The output functions are easily
determined as follows:

01 = liD + D + AD + lieF = A + D + lieF

O2 = BDF + DE + eE + "CDEF

0 3 = "CD + AE + DE

04 = "CD + liDF

FIGURE 4.32 PLA

Implementation.

B
Ltxt
c
Ltxt
Inputs
D
4: AND matrix

'"'"
E
Lr>~
F
Ltxt

OR matrix Outputs
03
04 ,
 4.7 Function Implementation Using PLAs and PALs 121

Larger numbers of prod~ct terms and/or outputs may be

obtained when more than one PLA is cascaded. Some of these
expansion schemes are shown in Figure 4.33. The product terms
essentially are increased by tying the outputs in parallel. This con­
figuration resembles the wiring of open-collector outputs. Corre­
spondingly, an increase in the word size could be accomplished by
unhooking the outputs.

FIGURE 4.33 PIA Expansion

Scheme: [a] Product Term
Expansion and [6] Output
-
--;
...-­
.­
Expansion.
·· PLA ·· PLA
· ·
-4' ,.. 0 8 08

L...­ '-­

PLA
·· PLA
·
- -
[a] [b]

The designer must use care in choosing the mintel1llS to be

formed in the AND section of the PLA. In order to use PLAs opti­
. mally, it is necessary to have as many output functions as possible
that have common minterms. It is not necessary for each function to
be minimized; the goal is to minimize the total number of minterms
required to implement the set of functions.
It would be appropriate now to discuss an additional program­
mable device known as programmable array logic (PAL). It also allows
the systems engineer to design his or her "own chip" by fusible links
to configure AND and OR gates to perfonn the desired logic func­
tions. The PAL is b~ically a programmable AND array driving a
fixed OR array. In comparison, both of the arrays of PLA are
programmable, while the programmable version of ROM, known
commonly as PROM (to be discussed in detail in Chapter 12), has
a fixed AND matrix and programmable OR array.
122 CHAPTER FOUR Implementation of Logic Functions

In the PAL circuit, as shown in Figure 4.34, an AND array

allows the designer to specify the product terms required and con­
nect them to perform the required SOP logic functions. The PALs
are available in a number of different part types, however, that vary
the OR gate format. Specifying the OR gate connection, therefore,
becomes a task of device selection rather than of programming.

FIGURE 4.34 Logic Diagram of A B

a Simple PAL.

Fusible links

Output

A B

Consequently, PALs totally eliminate the need for a second matrix

without any significant loss of flexibility.
In general PALs offer cost-effective capabilities for improving the
effectiveness of existing logic designs by expediting and simplifying
prototypes and board layouts. Figures 4.35[a-b] respectively show
the PLA and PAL configurations of a four-input-four-output
AND-OR circuit. The PLA provides the most flexibility for imple­
menting logic functions since the designer is equipped with com­ .i
J
plete control over all inputs and outputs. However, this flexibility .j

makes PLAs expensive and somewhat formidable to comprehend.

In comparison, the PAL combines much of the flexibility of the

PLA with the low cost and easy programmability of the PROM.

 4.8 Bridging Technique 123

FIGURE 4.35 A Four-Input,

Four-Output AND-OR Circuit
Using [a] PLA and [6] PAL

A B C o

A B c o

~ ~ ~ ~
OR array t-- ..- ...- ...- OR array
(programmable)
A fixed

~
"------~

~, ( () () ( () ()

,
=<
--I

y v
AND a!ray AND array

(programmable)
(programmable)

[a] [b]

" indicatesa.fusible link. Links are removed where no connection is desired.

4.8 Bridging The bridging technique is not so much a self-contained design algo­
rithm as it is a way to bend the c4aracteristics of a Boolean function
Technique that cannot be reduced further. If after using K-maps or the Q-M
technique, the function is still large and unwieldy due to the
minterms being logically separated, the function might be bridged by
using known functions that exhibit similar patternS of logically sep­
124 CHAPTER FOUR Implementation of Logic Functions

arated minterms. The X-OR function frequently is used in the

bridging process.
Consider the K-maps of several X-OR functions, shown in Fig­
ure 4.36. It is obvious that such K-maps are not reducible without
FIGURE 4.36 Examples of the X-OR function. A close examination of these maps reveals that
Several X-OR Functions. there are equal numbers of Is and Os on each half. In addition, the

A A A A
-.~-

1 0 1 0 0 1 0 1 0 0 1 1

L--­
0 1 0 1
_
1
..
0 1 0 1 1 0 0

B
AEBC

A A

0 1 0 1 0 1 1 0 1 0 1 0

C{ 0 1 0 1 1 0
'---y------'
0 1 1 0 1 0
'---v----'
B B

BEBC ACB8

A A

0 1 0 1 0 0 1 1 0 1 0 1 I 1 1 0 0

1 0 1 0 1 1 0 0 1 0 1 0 0 0 1 1

0 1 0 1 0 0 1 1 1 0 1 0 1 1 0 0

1 0 1 0 1 1 0 0 0 1 0 1 0 0 1 1

AEBBEBCEBD AEBCEBD

0 0 0 0 0 1 1 0 0 0 1 1 1 0 1 0

1 1 1 1 1 0 0 1 0 0 1 1 0 1 0 1

0 0 0 0 1 .0 0 1 1 1 0 0 1 0 1 0 ID

1 1 1 1 0 1 1 0 1 1 0 o i 0 1 0 1
'---y------' '---v-----'
B B
BEBD AEBC
 4.8 Bridging Technique 125

complement of the X-OR function may be obtained either by com­

plementing the whole function or by complementing odd numbers
of variables. An X-OR function can be changed to another equiva­
lent X-OR function as long as either (a) an even number of vari­
ables have been complemented, or (b) the entire function and an
odd number of variables have been complemented. For example,
AffiBffiCffiD= AffiBffiVffiD=A ffiBffiVffiD
= A ffi B ffi C ffi D = A ffi B ffi C ffiD
The X-OR functions are often very useful in implementing func­
tions that have logically isolated minterrns.
As long as the function to be implemented bears the characteris­
tic of an X-OR function, it can be realized using one or more X­
OR gates. However, the problem becomes more difficult when the
K-map closely resembles that of an X-OR function but is not one.
The bridging technique then is used to connect the desired function
and a closely resembling X-OR function. The technique consists of
the following steps:
1. Match the function K-map, F, as closely as possible to a
known X-OR K-map,j.
2. Realize the function F using bridging such that F = f' X
+ Y, where X and Yare two separate functions of the same
input variables. X and Y are determined by closely com­
paring K-maps of F andj.

Example 4.9 illustrates the idea behind the bridging scheme.

EXAMPLE 4.9 SOLUTION

Using X-OR and other·· assorted The function K-map is obtained as shown in Figure 4.37. By comparing
gates, implement the function the K-map of Figure 4.37 with those of Figure 4.36, it would become obvi­
ous that the closest match occurs with! = A El1 B El1 C El1 D. However,
F(A,B,C,D) = ~m(O,1,3,6,9,lO,15)

FIGURE 4.37
)
1 0 0 o i

1 0 0 1

1 0 1 O·

0 1 0 1
'-----v----'
B
126 CHAPTER FOUR Implementation of Logic Functions

they are not exactly alike. They differ at three minterm locations: 1,5, and
12. The two functions may be bridged, therefore, as shown in Figure 4.38.
The bridging between the two functions F and f required that certain con­
straints, as listed in Figure 4.39, be met in determining X and Y functions.
These constraints follow directly from the equation F = f . X + Y.

FIGURE 4.38

1 0 0 0 1 0 1 0 d - 0 - d 0 I 0 0

1 0 0 1 0 1 0 1 0 d 1 0 0 d
+

1 0 1 0 1 0 1 0 d d d 0 d 0

0 1 0 1 0 1 0 1 d d 0 d 0 d
'--v----' ~ ~ -----.,,-­
B B B B

FIGURE 4.39 The d's in the table of Figure 4.39 indicate that either X or Y or both
must equal 1. In other words, X and Y cannot simultaneously be 0 when F
= f = 1. This use of d, however, permits many choices for the selection of
F f X Y
X and Y. It can be seen that if all d's in X are set equal to 0, then F = Y,
0 0 - 0 which is contrary to what is expected in bridging. When F = Y no bridg­
0 1 0 0 ing is needed. On the other hand, if all d's in X are set equal to 1, then
1 0 1
Y =ABCD
1 1 d d
X=B+C
Therefore,

F = (A EB B EB C EB D) • (B + C) + ABC

The resultant circuit is obtained as shown in Figure 4.40. This bridged cir­
cuit is certainly better than the circuit that could be obtained by using only

FIGURE 4.40
------j.....-:~\­
oA -------I'--....J

B-+......-I
 Problems 127

a K-map. If a designer was limited to using only a K-map, the function

would have reduced instead to

F(A,B,C,D) = ABC + ABD + BCD + ABCD + ABCD + ABCD

The bridge scheme is very general and it does not have to involve
only X-OR functions. This technique ~an be used for generating a
complex function when a like function of the same variables already
exists (see Problem 16).

4.9 Summary In this chapter various practical techniques were introduced for
realizing combinational circuits. In particular, circuits using only
NAND gates, only NOR gates, only MUXs, only ROMs, only
PLAs, and only PAls were discussed. In addition, the bridging
technique was introduced to handle functions that are otherwise not
reducible.

Problems 1. Obtain the circuit for the following functions using only
NAND gates:
a. f(A,B,C,D) = ~m(I,4,10,11,13,15)
b. f(A,B,G,D) = ~m(I,3,4;9,1O,13)
c. f(A,B,C,D) = ~m(I,8-1O,15)
d. f(A,B,G,D,E) = ~m(I,3-7,11,14-17,22,24-27,30)
e. f(A,B,C,D,E) = ~m(I,8-1O,13-17,21,25-27,30,31)
2. Obtain NOR circuits for the functions of Problem 1.
3. a. Use a single level of l-of-8 MUXs and a few assorted
gates (if needed) to obtain a combinational circuit for
each of the functions of Problem 1.
b. For each of your solutions, complement one variable and
rearrange the inputs so that the function is still correct.
4. Using l-of-4 MUXs, obtain a two-level MUX circuit for each
of the functions of Problem 1.
5. Obtain the circuit for the functionf(X,Y,2:,U, V) = ~m(0,1,6,
7,9,12,13,15,18,20,22,24-26,28) using two levels of l-of-4
MUXs and a few assorted gates.
6. Use bridging to implement the following functions using X­
OR gates:
a. f(A,B,C,D) = ~m(I,2,5,7,8,10,13,14)
b. f(A,B,G,D) = ~m(2,3,6,7,9,11,12,13)
c. fCW,X,Y,Z) = ~m(0,2,3,6-8,1O,13)
d. f(W,X,Y,Z) = ~m(0,6,9,1O,15)
7. Implement the functions of Problem 6 using ROMs.
8. Implement the functions of Problem 6 using PLAs.
128 CHAPTER FOUR Implementation of Logic Functions

9. Given the function j{A,B,C,D) = ~m(0,4,9,1O,11 ,12) and Jl

= B EB D, determineJ2 andJ3 for the circuit of Figure 4.PI.

FIGURE4.Pl
;:---lr--)----i-~)r---f(A,B,C,D)
~------------------~--~-

10. Use a ROM to design a binary-to-Gray code converter.

11. Draw the logic diagram of an 8 X 2 ROM that produces the
full adder function as described in Chapter 1 (Table 1.4).
12. Use a ROM to achieve four-bit by four-bit binary multiplica­
tion.
13. Design a four-input network that squares each of the binary
inputs using (a) a ROM, (b) a PLA, and (c) a PAL. Assume
that both input and output are unsigned.
14. Design a five-input logic network that fmds the 2's comple­
) ment of a positive number using (a) a ROM, (b) a PLA, and
(c) a PAL.
15. Design a network that accepts trigonometric angles in degrees
(between 0° and 10° in steps of 1°) and gives out the corre­
sponding tangent value correct up to five significant places.
Use (a) a ROM and (b) a PLA.
16. Consider the truth table for the full adder of Table 1.4. Bridge
the carry-out function with Ai EB Bj , where Ai and Bj are the
augend and addend, respectively.

Bartee, T. C. "Computer design of multiple output logical networks." IRE

Suggested Readings Trans. Elect. Compo vol. EC-lO (1961): 21.
Cerny, E., and Marin, M. A. "A computer algorithm for the synthesis of
memoryless logic circuits." IEEE Trans. Compo vol. C-23 (1974): 455.
Cerny, E., and Marin, M. A "An approach to unified methodology of
combinational switching circuits," IEEE Trans. Compo vol. C-26 (1977):
745.
Culliney,j. N.; Young, M. H.; Nakagawa, T.; and Muroga, S. "Results of
the synthesis of optimal networks of AND and OR gates for four-vari­
able switching functions." IEEE Trans. Compo vol. C-27 (1979): 76.
Curtis, H. A "Short-cut method of deriving nearly optimal arrays of
NAND trees." IEEE Trans. Compo vol. C-28 (1979): 521.
Davio, M. "Read-only memory implementation of discrete functions."
IEEE Trans. Compo vol. C-29 (1980): 931.
Dietmeyer, D. L., and Su, Y. H. "Logic design automation of fan-in lim­
ited NAND networks." IEEE Trans. Compo vol. C-18 (1969): 11.
Ektare, A. B., and Mital, D. P. "Multiplexer logic circuit design using
cubical complexes." Elect. Lett. vol. 16 (1980): 495.
Ektare, A B., and Mital, D. P., "Probabilistic approach to multiplexer
logic circuit design," Elect. Lett. vol. 16 (1980): 686.
Suggesled Readings 129

Ellis, D. T. "A synthesis of combinational logic with NAND or NOR ele­

ments." IEEE Trans. Elect. Compo vol. EC-14 (1965): 701.
Fleisher, H., and Maissel, L. "An introduction to array logic." IBM] Res.
& Dev. vol. 19 (1975): 98.
jones,]. W. "Array logic macros." IBM] Res. & Dev. vol. 19 (1975): 120.
jullien, G. A. "Residue number scaling and other operations using ROM
arrays." IEEE Trans. Compo vol. C-27 (1978): 325.
Kambayashi, Y. "Logic design of programmable logic arrays." IEEE
Trans. Compo vol. C-28 (1979): 609.
Lai, H. C., and Muroga, S. "Minimum parallel binary adders with
NOR(NAND) gates." IEEE Trans. Compo vol. C-28 (1979): 648.
Li, H. F. "Variable selection in logic synthesis using multiplexers." Int.]
Electron. vol. 49 (1980): 185.
Liu, T. K; Hohulin, K R.; Shiau, L. E.; and Muroga, S. "Optimal one­
bit full adders with different types of gates." IEEE Trans. Compo vol. C­
23 (1974): 63.
Logue,]. c.; Brickman, N. F.; Howley, F.; jones,]. W.; and WU, W. W.
"Hardware implementation of a small system in programmable logic
arrays." IBM] RES. & Dev. vol. 19 (1975): 110
Lotfi, Z. M., and Tosser, A.]. "Systematic search for minimum synthesis of
logical functions with multiplexers." Int.] Electron. vol. 47 (1980): 569.
Nagle, H. T.,jr.; Carroll, B. D.; and Irwin,]. D. An Introduction to Computer
Logic. EnglewOO<;YCliffs, N.].: Prentice-Hall, 1975.
Nakamura, K "Synthesis of gate-minimum multi-output two-level nega­
tive gate networks." IEEE Trans. Camp. vol. C-28 (1979): 768.
Papachristou, C. A. "An algorithm fot optimal NAND cascade logic syn­
thesis." IEEE Trans. Compo vol. C-27 (1978): 1099.
Peatman,]. B. Digital Hardware Design. New York: McGraw-Hill, 1980.
Preparta, F. P. "On the design of universal Boolean functions." IEEE
Trans. Compo vol. C-20 (1971): 418.
Sasao, T. "Input variable assignment and output phase optimization of
PLA's." IEEE Trans. Camp. voL C-33 (1984): 879.
Sasao, T. "An algorithm to derive the complement of a binary function
with multiple-valued inputs." IEEE Trans. Compo vol. C-34 (1985): 131.
Shannon, C. E. "A symbolic analysis of relay and switching circuits."
Trans. AlEE. vol. 57 (1938): 713.
Weinberger, A. "Device sharing in array logic." IBM Tech. Disc. Bult. vol.
19 (1976): 1357.
Weinberger, A "High-speed programmable logic array adders." IBM]
Res. & Dev. vol. 23 (1979): 163.
Whitehead, D. G. "Algorithm for logic circuit synthesis by using mul­
tiplexers." Elect. Lett. vol. 13 (1977):,355.
Wood, R. "A high density programmable logic array chip." IEEE Trans.
Compo vol. C-28 (1979): 602.
Design of . CHAPTER
FIVE
Combinational
Circuits-______ ----I

5.1 Introduction A combinational logic circuit, as shown by the block diagram of Figure
5.1, is defined as a combination of logic devices whose output is a
function of the present values of the input variables and indepen­
dent of the past values. After propagation time through the circuit,
input variable changes cause output changes that are dependent
orily on the present input values.
\

FIGURE 5.1 Block Diagram of a

Combinational Network. Inputs ·· •... Combinational
··· Outputs
· logic

In Chapters 1, 2, and 4 we introduced the necessary tools to

design combinational logic circuits. The design algorithm leading to
the realization of a complex combinational circuit consists of the fol·
lowing essential steps:
1. The complex logic problem is intuitively analyzed and
decomposed into a set of smaller but nontrivial functional
units.
2. The number and characteristics of both input and output
variables for each of the functional units are identified.
3. A truth table for each of the functional units is determined.
4. The output functions for each of the functional units are
simplified using one of the ininimization schemes covered
in previous chapters.
5. The circuits corresponding to each of the functional units
are assembled and tested individually and then connected
to form the desired complex function.
130
 5.2 Binary Adders 131

In practice, the designer would have to consider various practical

limitations such as the number of logic gates, interconnections, gate
inputs, fan-out, and the length of propagation delay. In the not too
distant past the only option available to the designer was to assem­
ble the entire logic circuit with a sack full of SSI chips. However, at
this time we have better alternatives because many more complex
logic circuits are available in IC form. The only limitation to the
use ofeither the MSI or the LSI is that we may not be able to locate
a device that exactly meets our requirements. In that event the
designer must modify the standard device by externally combining
it with other SSI or MSI chips.
In this chapter applications of these combinational logic design
tools will be considered. The design and application of MSI devices,
including adders, subtracters, decoders, encoders, and error-control
logic, is presented. These devices play an important part in the
development of more advanced digital systems. The application of
the MSI devices considered is not governed by a set of design proce­
dures as well defined as those procedures for individual logic gates.
Experience and intuition (horse sense) become important. There is
no substitute for understanding exactly what the MSI devices can
do. After studying this chapter, you should be able to:
o Break a complex design into manageable subunits;
o Design individual subunits and be able to cascade them
together;
o Understand the working principles and design process
of various combinational binary adders and/or sub­
tracters;
o Understand the working principles and design process
of various code converters;
o Understand the working principles and design of BCD
arithmetic circuits;
o Understand the working principles and design of vari­
ous decoders and encoders;
o Understand the working principles and design of vari­
ous error-correcting circUits.

5.2 Binary Adders

Most arithmetic operations are reducible to simple addition or sub­
traction processes that can be performed repetitively for more com­
plex operations such as multiplication and division. If the addition
circuit has no carry-in, the addition of the least significant bits
involves only two operancls, the addend and augend. The addition
of the remaining bits, however, requires the carry-in from the addi­
tion of the previous column.
132 CHAPTER FIVE Design of Combinational Circuits

A multi-bit adder can be realized in various ways, each having

different speed and cost characteristics. A two-level network would
obviously prove to be the fastest. However, this network would
require a large number of gates and gate inputs. It would be neces­
sary to have 22n NAND gates of 2n + 1 inputs and one NAND gate
of 22n inputs to add two n-bit numbers. This number of gates and
inputs is quite significant for even small values of n. The alternative
to this expensive design is quite straightforward. It follows directly
from our observation of the algorithm of an n-bit add operation.
Irrespective of the number of bits, the process of adding augend and
addend is identical at each of the columns except at the least signifi­
cant position. The design of a parallel n-bit addition circuit, there­
fore, is accomplished by designing a total of n single-bit addition cir­
cuits. In order to allow n single-bit adders to be connected together
to form an n-bit adder, the single-bit adder stages need to be full
adders (adders with three inputs). The least significant carry-in is
tied to a 0, and each of the remaining carry-in inputs is tied to the
carry-out of the previous single-bit addition.
As an introduction to adder design, we shall first consider a half
adder (adder with only two inputs, addend and augend). The resul­
tant circuit will have application in the design of a full adder. In
fact, an n-bit adder circuit using n- 1 full adders and one half adder
also can be designed.

5.2.1 Half Adder (HA) The half adder (HA) unit is a simple multiple-output combina­
tional circuit used for adding two bits without a carry-in. The truth
table for the two inputs, Ai and Bi, and the output sum, Si' and
carry-out, Gi , are shown in Figure 5.2.

FIGURE 5.2 Half Adder: [a]

Block Diagram and [hJ Truth
Table. Augend Addend Ai Bi Cj Sj
0 0 0 0
HA 0 1 0 1
1 0 0 1
1 1 1 0
Carry-out Sum

[a] [b]

Using a Karnaugh map, the equations for the sum, Sj, and carry,
Gi , are as follows:
S·I = A·ll + A·~· = A· lD
I I rut W B· I t [5.1]
C·t = A·B·
I I [5.2]

Ther;e are several ways to implement these functions. Figure 5.3

 5.2 Binary Adders 133

FIGURE 5.3 HA Circuit: [a] illustrates four ways to implement the HA functions. Circuits using
Using AND and X-OR Gates, [b] [a] an AND and an X-OR gate, [b] NOR gates, [cJ NAND gates,
Using NOR Gates, [c] Using and [dJ multiplexers are shown.
NAND Gates, and [d] Using
MUXs.

A
C; Cj

B;

A
Bj D-Si Sj

[a] [bJ

Bj 0 B; Bj

Sj
Aj 1-of-2
10 MUX E

Cj

Cj S;
[c] [d]

5.2.2 Full Adder (FA) Afoll adder (FA) is a three-input, two-output logic circuit that
adds two binary digits, Aj and Bi, and a carry-in from the i - I bit
position, Gi - 1• The block diagram and the corresponding truth
table are shown in Figures 5.4[a-b]. The K-maps for the sum bit, Si,
and the carry-out, Gil are constructed from the truth table and
shown in Figure 5.4[c]. The equations for the sum and carry-out
can then be obtained from the K-maps as follows:
Si(Ai,Bj,Gi - 1) = AjB,ci-l + AJ3,Ci-l + A;B,Ci - 1
. + AJ3jGi - 1 [5.3]

= :4',,+ 'B j + Gi- 1 + Ai + Bi + Vj- 1 + Ai + Bi

+ £:';-1 + Ai + Bi + Gj - 1 [5.4]

and
Gi(Aj,Bj,Gj_l) = AiBi + A,ci-l + B,ci-l [5.5]
= Aj + Bi + Ai + Ci - 1 + Bi + Ci - 1 [5.6]
134 CHAPTER FIVE Design of Combinational Circuits

FIGURE 5.4 Full Adder: [a] Inputs Outputs

Block Diagram, [6] Truth Table,
Ai 8; Ci - Ci S;
and [c] K-Maps for the Carry-Out 1

and the Sum. 0 0 0 0 0

0 0 1 0 1
FA Carry-in
0 1 0 0 1
0 1 1 1 0
1 0 0 0 1
1 0 1 1 0
1 1 0 1 0
1 1 1 1 1
i

[a] [b]

A;8i Ai8 j
Ci - 1 00 Ci- 1 00 01 11 10
o O· 0 o 0 IT] 0 IT]
1 0 lIT] 0 IT] 0

[c]

There are several ways to implement the FA equations. The

direct implementation of Equations [5.3] and [5.5] leads to FA cir­
cuits using only NAND gates; If Equations [5.4] and [5.6] are used,
the equivalent NOR circuits may be obtained. Figures 5.5[a~]
respectively show the typical FA circuits using only NAND and
only NOR gates. Each of these circuits requires a total of 12 gates
and 31 gate inputs. However, by making use of the bridging tech­
nique, the number of gates can be reduced.
Note that the K-map for the sum output is irreducible, produc­
ing a cumbersome circuit. Note also that the K-map for Sj is that of
a three-input X-OR function. A review of Figure 4.36 indicates that
the sum equation may be reduced to
Si = Ai EB Bi EB Ci~l [5.7]

Examining the carry-out K-map reveals that it could be bridged

with an X-OR function as follows (see Chapter 4, Problem 16):
c· =
I (A·I l1:\
W 1- 1 +
B·)C·
I A·~·
P, [5.8]

Now recall that Ai EB Bj is the sum output and A/3i is the carry­
 5.2 Binary Adders 135

FIGURE 5.5 FA Circuit: [a]

Using Only NAND Gates and [b]
Using Only NOR Gates.

A
fj.
- I
Ci- 1

Ai
Bi

Ci - 1

(a]

(b]
out for an HA. Thus. the HAs designed in the last section may be
used for realizing an FA. Expressing the FA equations in terms of
the HA equations, Equations [5.1] and [5.2], gives
Si = Si(HA) E11 Ci- 1 [S.9]
Ci = Si(HA) • Ci - 1 + Ci(HA) [S.10]
where Si(HA) and Ci(HA) are the sum and carry-out of the HA. Note
also that Equation [5.9] involves an X-OR operation between the
carry-in and the HA sum. If the sum output of the HA and the
carry-in are fed into a second HA, the final sum output will be the
136 CHAPTER FIVE Design of Combinational Circuits

FA sum, Si' In addition, if the carry-outs from both of the HAs are
ORed together, the FA carry-out, Gi, is obtained. Figure 5.6[a]
shows the FA circuit using two HAs and an OR gate. The circuit
corresponding to Equations [5.7] and [5.8] also may be imple­
mented using only NAND gates or only NOR gates. The resultant
NAND equivalent circuit is shown in Figure 5.6[b]. This NAND
circuit requires only nine NAND gates and a total of 18 gate inputs.
We have designed an FA circuit using three fewer gates and 13
fewer gate inputs than would be necessary had we not made use of
the X-OR K-map structure and bridging.

FIGURE 5.6 FA Circuits: [a]

Using HAs and [b1 Using Only
NAND Gates.

Ai Addend Sum Addend Sum Si

HA HA I
I

Bi Augend Carry Augend Carry

Ci

Ci - 1
[aJ

Ci _ 1- - - - - - - - I - - - - - - - - - - - l
D - - - - - - Ci

[b]

We have completed the design of half and full adders. These

devices are also commercially available in the form of MSI devices.
Four FAs usually are connected and are commercially available in
an MSI four-bit adder Ie. Using a commercially available four-bit
adder, two quantities, A:0~lAo and BsBzBIBo, can be added. The
resultant sum is SSS~lSO' and the carry-out is from the As, Bs, and
G2 addition. Figure 5.7 demonstrates the connection of n FAs to
 5.2 Binary Adders 137

FIGURE 5.7 n-Bit Parallel

Adder Circuit: [a] Using n FAs
and [b] Using n- 1 FAs and One
HA.

AI B, Ao Bo

X Y

FAn- 2 Cj FAa Ci 0

Co S Co S S Co S

Co

SI
[a]

Ao Bo

FAn I Cj FAn- 2 Cj FA2 Ci HAo

C, Co

So
[b]
X = augend Co carry-out

Y = addend S sum

Cj carry-in

make an n-bit adder.· The delay of this n-bit ripple adder is n!:J.
FIGURE 5.8 Multiplication of where !f:. is the propagation delay of the carry-out of a single FA.
Two Three·Bit Numbers.
This delay is accumulated when the carry into the multi-bit adder
has to propagate through all.ofthe FAs to get to the final carry-out.
B2 B, Bo
Consequently, this delay beComes more and more significant as n
A2 . A, Ao
becomes larger.
AoB2 AoB, AoBo Multiplication of binary numbers makes use of addition just
A,B2 AlB, A,Bo as multiplication of decimal numbers does. The multiplication of
A2B2 A2B1 A2Bo two three-bit numbers, A = A~ lAo and B == Br/3I Bo, is symbolicilly
obtained as shown in Figure 5.8, where P5P.,;P3P'lflPO forms the prod­
138 CHAPTER FIVE Design of Combinational Circuits

uct. Note that for the multiplication of two n-bit numbers, the product
has the possibility of 2n bits.
Later in the text sequential design techniques will be presented
that will allow designing a multiplier that uses a repetitive algo­
rithm for multiplication. It is possible, however, to design a combi­
national circuit using FAs that will pedorm the multiplication of
two binary numbers by pedorming the sum of the three partial
products of two numbers. This combinational circuit must be able
to add columns of bits. Example 5.1 will demonstrate how FAs can
be used to do similar functions.

EXAMPLE 5.] SOLUTION

Design a circuit using FAs that may Let the column of numbers be llo, ah a2, a3, a4, and as. The sum if all were 1
be used for adding a column of six would add up to 110, so there will be three outputs: S2, Sh and So. How- .
single-bit numbers. ever, an FA may add up to only three single bits. Two FAs may be used to
obtain two partial sums of the six bits as follows:

where XI and X2 are the carry-outs and YI andyz are the respective sums. As
a next step the least significant bits,YI and Y2, may be added as follows:

YI
Y2
o

where X3 and So are the carry-out and the sum, respectively. Finally, Xb X2,
and X3 could be fed into a fourth FA to yield the carry-out, S2, and the sum,
Sil as follows:

The complete circuit, therefore, would require four FAs. The resulting cir­
cuit is obtained as shown
. . in Figure 5.9.
 5.2 Binary Adders 139

FIGURE 5.9 as a4 a2 81

1
B

FA C, a3 FA C, 80

S
X1 Yl

A B

FA Cj FA Cj 0
• - % ';";;;-:;.'~

Co S S
So

A augend Co carry-out
B == addend S == sum
Cj == carry-in
rsmr L 11

Another example of the application of adders in digital systems is

given in Example 5.2. This example also demonstrates how a com­
puter that is designed to handle binary quantities of n bits can per­
form operations on 2n-bit quantities. In programming, such an
operation is commonly called a multiple-precision operation.

EXAMPLE 5.2 SOLUTION

Use four-bit multipliers and four-bit . FIGURE 5.10

binary adders to design a circuitfor
multiplying two eight-bit numbers.
The block diagrams of the /
/4 /
/4
multiplier and adder units are .
provided in Figure 5.10. The four­ Multiplier Multiplicand Augend Addend
bit multipliers are assumed to be 4-bit FA
1/
ROMs. The two four-bit quantities 4-bit multiplier ROM ,­ Carrry-out Carry-in /
to be multiplied make up an eight­
bit address. Each ROM storage Product V Sum
location is the eight-bit product of /

the two four-bit quantities that /

v4
make up its address.
14() CHAPTER FIVE Design of Combinational Circuits

FIGURES.II

4 4 4

y x Y

4-bit
4-bit 4-bit 4-bit
multiplier ROM
multiplier ROM multiplier ROM multiplier ROM

Z
o 8

CO- 3

V U V

4·bit FA

w W

4 4
4 AO- 3
A4- 7 4
U V U V
U

Co 4-bit FA Cj Co 4-bit FA Ci Co 4-bit FA Cj DO- 3

w w W

4 4 4

P8-11 PO- 3

x =:: multiplier v =::addend

Y multiplicand W sum
Z =:: product C; =:: carry-in
U augend Co = carry-out

A good approach to any design problem is to break the given problem into
several simpler problems. This procedure is necessary in this example i.p.
order to make the problem fit the devices that are provided. Consider the
mqItiplication of two eight-bit numbers, Xl and X2, each consisting of a
 5.3 Binary Subtracters 141

least significant four bits, Li , and a most significant four bits, Mi' The eight­
bit number can then be expressed as the sum of the two four-bit parts:

x I = 24(M·)! +L !

where Mi is shifted to the left four places (multiplied by 24) before being
added to Li. The product, P, may now be expressed as

P = [24(Ml) + Ld[24(M2) + L2]

= 28(MtM2) + 24(M\L2 + M~\) + (L)L2)

Each of these four partial products may be obtained using four four-bit
multiplier units. The four multiplier units would respectively have (a) Ml
and M2, (b) MI and ~, (c) M2 and LI , and (d) Ll and ~ as inputs. This
configuration would result in a total of four eight-bit outputs: AJ BJ CJ and
D, respectively. Since two eight-bit quantities are being multiplied, a 16­
bit product is expected. Note also that D should be added to the sum of B
and C that have been shilled to the left four places, and this in tum should
be added to A that have been shilled to the left eight places. A network of
six four-bit FAs may be employed, as shown in Figure 5.11, to obtain the
16-bit sum of the shilled partial products. The final product is obtained by
adding the partial products. Care must be taken to connect the partial
products at the correct bit positions relative to their power of2. The shill­
ing is implicit in the interconnection pattern of the adder modules.
Note that a single ROM for multiplying two eight-bit numbers would
require a total of 1,048,576 bits, that is, 16 bits of address and 16 bits in
each location, to store the 16-bit product, or 216 X 16 bits. Each ROM
that we used in this example had a size of only 2048 bits, giving a total of
8192 bits. This design would probably be less expensive but would be
. slower due to the time required to perform the additions. Such time and
money trade-otfs will be a typical design decision that must be made by
every engmeer.

5.3 Binary Subtracters Many arithmetic circuits also require a unit for subtraction. A sub­
tracter circuit could be designed from scratch. However, recall from
Chapter 1 that subtraction also is possible by adding the comple­
ment of the subtrahend to the minuend. Consequently, rather than
designing a straightforward subtracter, a multi-bit subtracter can
be made using complement arithmetic. This design would involve
the use of a multi-bit parallel adder circuit that isfed with the com­
plemented subtrahend and the minuend.
In order to accoinplodate the multi-bit parallel adder to the
requirement of our preseIlt desigy{, certain modification is necessary
for complementing the subtrahend. If each bit of the subtrahend is
individually complemented, the corresponding 1's complement will
be obtained. A 2's complement could be formed by adding a one to
the LSB of the corresponding 1'8 complement, that is, by making
the carry-in to the LSB position of the adder a 1. To perform the
142 CHAPTER FIVE Design of Combinational Circuits

complementing of each bit, we can use an X-OR gate. An X-OR

has a useful characteristic that makes it perform as a programmable
inverter. Consider a two-input X-OR gate for which one input is a
bit of the subtrahend and the other input is tied to a select line, E.
When E = 1, the output will be the complement of the subtrahend
bit. VVhen E =0, the output will be the uncomplemented bit, since
from the X-OR truth table,
IffiY=Y
OffiY=Y
The property of an X-OR function provides the possibility of
having both an adder and a subtracter out of the same circuit. Such
a circuit is shown in Figure 5.12 where the enable, E, allows the cir­
cuit to add or to subtract by taking the 1's complement and adding
a one to the LSB. When E = 0, the circuit adds with a carry-in of
0, and when E = 1, the circuit complements B and adds that to A
with a carry-in of 1.

FIGURE 5.12 Four-Bit Adder/

Subtracter Unit.

~----------~~----------~~----------~~----'---E

B1 Bo

Ao

x X X Y X

FA Cj FA Ci FA Ci FA Ci

S S Co S Co S
C2 C1 Co
C3 S3 S2 Sl So
X = augend Ci carry-in
y addend Co carry-out
S sum

EXAMPLE 5.3 SOLUTION

Obtain a combinational circuit for From Section 1.4 we know that the LSB of a number always remains
realizing the 2's complement of a unchanged when obtaining the 2's complement. If the LSB, llo, is 1, then
five-bit binary number, the next higher bit, a" is inverted. An X-OR gate in the fonn of flo ED al
pi

5.4 Carry Look-Ahead (CLA) Adders 143

may be used to accomplish this. The next higher bit, ~, is complemented if

either al or flo is 1. This line of argument could be carried out for all of the
remaining bits. Such logical tests can be implemented by peIforming flo
+ al + . . . + am-I> which would indicate whether or not a 1 is present in
the least significant m bits.
It is apparent, therefore, that X-OR and OR gates can be used to real­
ize the necessary circuit for performing 2's complement of a number. The
OR gates would perform tests and the X-OR gates would complement bits
if necessary. The resulting circuit for a five-bit number a4a3a~laO is
obtained as shown in Figure 5.13.

FIGURE 5.13

2's complement of the input

5.4 Carry Look~Ahead The particular multi-bit parallel adder circuit developed in the pre­
vious section is sometimes referred to as a ripple adder because a
(CLA) Adders carry from one unit of the adder may have to ripple through several
units before the sum is obtained. Such ripple adders have also been
used to form either 2's complement, or l's complement sign-and­
magnitude binary adder/subtracters. The performance of a ripple
adder/subtracter, however, is limited by the time required for the
carries to ripple through all of the stages of the circuit. For such
devices the maximum delay is directly proportional to the number
of FA units.
One particular method of speeding up the combinational addi­
tion process is known as carry look-ahead (CLA). In Figure 5.14 it may
be seen that the carry-out is the same as the carry-in as long as one
144 CHAPTER FIVE Design of Combinational Circuits

FIGURE 5.14 FA Conftgurations

Resulting in a Carry-Out.

o o

x y

o
Co S Co S

o o o
X = augend Co carry-out
Y = addend S sum
C, = carry-in

of the other two inputs is a 1. Also, the carry-out is always a 1 inde­

pendent of the carry-in when both of the other inputs are Is, and 0
if both are O. Consequently, two useful functions can be defined: the
carry-propagate, Pi, and the carry-generate, Gi
p·=A·ffiB·
1 I \Il I [5.11]
G·I = A·' B· I I [5.12]

where Ajand Bi are the addend and augend, respectively, of the ith
full adder.
The FA equations, Equations [5.7] and [5.8], can then be rewrit­
ten as
s· I
= p. \Il
1
ffi C· 1­
I [5.13]
C·I = G· + Pc.
I I 1­ I [5.14]

For a four-bit adder the carries for the various stages are as follows:

Co = Go + POC- I [5.15]
CI = GI + PICO
= G1 + P1GO+ PIPOC- 1 [5.16]
C2 = G2 + P2Cr
= P2 + P2G1 + P2P1GO+ P~IPOC-I [5.171

and
C3 = G3 + P3C2
I
= G3 + P3G2 + P~2Gl + P~2PIGO + P3P2PIPOC-I [5.18]
 5.4 Carry look-Ahead (CLA) Adders 145

while the sum bits are

So = Po EB C_ I [5.19]
SI = PI EB Co [5.20]
S2 = P2 EB CI [5.21]

and
[522]

In Equations [5.15-18] the carry-out for each of the stages is

dependent only on the initial carry, C_ b and the corresponding
propagate and generate functions. The equation for C3 can be
implemented with a two-gate level circuit. Since each of the gener­
ate and propagate functions can be expressed in terms of the two
data bits, C3 is available after two gate delays, resulting in a fast
addition process. The block diagram of a four-bit CLA adder is
shown in Figure 5.15. It consists of three sections, each having four
subunits. The PG j section generates the carry-propagate and carry­
generate functions. The CLi section intakes the outputs of the previ­
ous section and generates the carries. Finally, the SUi section gener­
ates the sum bits. However, the carry units are different for every
bit, and their complexity increases as they move further away from
the LSB.
The internal hardware for the four-bit CLA adder is shown in
Figure 5.16. Each propagation-generation unit requires five NAND
gates, each sum unit requires four NAND gates, and the n-bit carry
section requires a total of (n 2 + 5n)12 NAND gates. A four-bit CLA
adder, therefore, requires a total of 54 NAND gates and involves a
total of eight units of gate delay. In comparison, addition in a four­
bit ripple adder requires 12 units of gate delay. The four-bit CLA,
therefore, cuts down the time factor by about one-third. Similarly, a
64-bit CLA adder requires almost five times as many NAND gates
as a 64-bit ripple adder, but reduces the propagation delay by a
factor of 17. It follows, therefore, that the CLA adder will provide a
faster addition time, especially when the number of bits is higher.
An inventory of the n-bit CLA adder reveals that the sum and
the carry subunits require a total of9n two-input NAND gates. The
carry section, however, requires (2n + 1) two-input NAND gates
and (n + 3 - m) m-input NAND gates for 3 m n + 1. There­
fore, from a practical standpoint C}:.A for too large n turns out to be
i

~t
~r .
quite problematic. It was seen in Chapter 3 that NAND gates with
too many inputs are hard to come by. This limitation is com­
pounded by the fact that the carry-in, C_ b must drive a total of n
I + 1 gates. In addition, the propagate functions will be subjected to
~Si'
I
c/,·
a fan-out requirement on the order of (n + 1)2/4. All of these
~ ~:.". . . .•.
together result in a serious fan-in problem for n too large.
,
~~:~
146 CHAFFER FIVE Design of Combinational Circuits

FIGURE 5.15 Four-Bit CLA

Adder: [a] Block Diagram and [6]
Internal Circuitry.

CLA

SO-3

[a]

Xo Yo

11 11 11 11 Propagate­
PG 3 PG 2 PG , PG o generate
section

P3 G3 P2 G2 PI G1 Po Go

i
1_

IIII
CL3 CL2 CL 1 CLo Carry
look-ahead
section

C3 C2 C1 Co
-

I~ Sum
sU 3 I-+- sU 2 I+- su, ~ suo section
Cout
l 1 1 1
So
[b]

. Commercially available four-bit adders perform an internal

CLA for a four-bit add. These outputs are available and labeled as
P for propagate and G for generate. These outputs when used with
another MSI device called a look-ahead carry generator provide signifi­
cant acceleration in the add operation, particularly for large num­
 5.4 Carry Look-Ahead (CIA) Adders 147

FIGURE 5.16 [a] Single

Propagation-Generation Unitt [b]
Single Sum Unitt and [c] Carry
Units of a Four-Bit CLA Adder.

AI
P;
8;

~
/
G;
[a]

Si

[b]

eP:~
Go >-r>J }-e o

P3
P2
P1 P1

Po
Po
C 1­ C- 1
P3
P2
P1 P1
C1 Go

Go

G1 P3

P2
C3
G1
P2
P1
Po P3
C- 1
G2
P2
P1
Go G3
C2
P2
G1

G2

[c)
148 CHAYI'ER FIVE Design of Combinational Circuits

hers of bits. Figure 5.17 shows the connections for performing the
addition of two 16-bit numbers with CLA.
Carry-propagate and carry-generate functions for more than
four bits can be derived continuing the same process used for deriv­
FIGURE 5.17 Hi·Bit Addition ing C3 and S3' Circuit complexity makes such implementation
with CLA Modules. impractical except for special-purpose, high-speed requirements.

1
Addend Augend
1 1 Addend Augend
1 1 Addend
1
Augend
1
Addend Augend
1

Sum4-bit fast adder r-- Sum4-bit fast adder ,...- Sum4-bit fast adder
r- Sum4-bit fast adder r-

P G Cj Co P G Cj Co P G Ci Co P G CI _ Co
I I I T I
+
51 6 5 12-5 15
5 8 -5 11
I 5c 57 50-53

r-

I I I i r
P3 G3 P2 G2 P1 G1 Po Go Cj

Look-ahead carry generator

P G C2 C1 Co
I I

T
I
For use with more than 16 bits

EXAMPLE 5.4 SOLUTION

Design an eight-bit fast adder 'J11e modified carry equations for an eight-bit fast adder are derived as
where the fan-in and fan-out follows:
problems are avoided by allowing
the carries to ripple through after Co = Go + C-1Po
the addition of every four bits. C1 = G1 + GoPl + C-1PoPl
C2 = G2 + G1P2 + GoPIP2 + C,,-lPoPIP2
C3 = G3 + G2PS + G1P2P3 + GoPI P2P3 + C- 1PoPI P2P3
C4 = G4 + C3P4
95 = G5 + G~5 + C3P~5
C6 = G6 + GsP6 + G~5P6 + C~~5P6
C7 = G7 + G~7 + G5P~7 + G~5P~7 + C3P~5P~7
 5.5 Code Converters 149

The first four equations are similar to Equations [5.15-18]. The last four
equations have a similar form, but C3 is treated as the carry-in to the fifth
bit. The resultant circuit may now be obtained, as shown in Figure 5.18, by
FIGURES.I8 having two separate units for the carry section .

.Addend
. Augend

BY{B ..... --.

T
I ±t I
I 1

PG 7 PG 6 PG s PG 4 PG 3 PG 2 PG I PG o

.'. P7 /G
7
,
,~6 G6 Ps IG s P4 fG 4 P3 IG 3 P2 G2 PI /G 1 Po IGo

I I .!
If rI f{ Tf 11 [{ J1 11
i

CL3 CL 2 CL I CLo CL3 CL 2 CL I CLo

I
I I ~ J

C7 C6 Cs C4 C3 C2 Cl CO

I
IP 7 I
IP6 IPs

I
P
I4 r 3
lP 2 lP I
lPo

SU 7 SU6 SUs SU 4 SU 3 SU 2 SUI SUo

i
i

I I I I I I I So
I

5.5 Code Converters It was pointed out in Chapter 1 that many different binary codes
exist that are used in various digital subsystems. Sometimes it is nec­
essary to transfer data from one subsystem to another. Code converter

i
circuits are required to convert one form of binary code to another.
Many of these converters use combinational logic, and there are
many that use sequential logic as well.
~; The following examples will illustrate the combinational tech­

~I t.',
niques in the design of various code converters. We shall consider
several conversion schemes: Gray-to-binary, binary-to-Gray,
i' :~
'j~,:'

. binary-to-BCD, and BCD-to-binary.

,.
:~
150 CHAPTER FIVE Design of Combinational Circuits

EXAMPLE 5.5 SOLUTION

Design a circuit for converting a The Gray code is a reflected binary code such that the Gray code for one
five-bit Gray code into its binary number differs from the next number in only one bit position. In binary
equivalent. space the codes are one unit distance apart. The truth table of Figure 5.19
shows the corresponding Gray and binary numbers.

FICURE5.J9
Gray Binary

G4 G3 G2 G1 Go B4 B3 B2 B1 Bo

0 0 0 0 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 1
0 0 0 1 1 0 0 0 1 0
0 0 0 1 0 0 0 0 1 1
0 0 1 1 0 0 0 1 0 0
0 0 1 1 1 0 0 1 0 1
0 0 1 0 1 0 0 1 1 0
0 0 1 0 0 0 0 1 1 1
0 1 1 0 0 0 1 0 0 0
0 1 1 0 1 0 1 0 0 1
0 1 1 1 1 0 1 0 1 0
0 1 1 1 0 0 1 0 1 1
0 1 0 1 0 0 1 1 0 0
0 1 0 1 1 0 1 1 0 1
0 1 0 0 1 0 1 1 1 0
0 1 0 0 0 0 1 1 1 1
1 1 0 0 0 1 0 0 0 0
1 1 0 0 1 1 0 0 0 1
1 1 0 1 1 1 0 0 1 0
1 1 0 1 0 ,­ 0 0 1 1
1 1 1 1 0 1 0 1 0 0
1 1 1 1 1 1 0 1 0 1
1 1 1 0 1 1 0 1 1 0
1 1 1 0 0 1 0 1 1 1
1 0 1 0 0 1 1 0 0 0
1 0 1 0 1 1 1 0 0 1
1 0 1 1 1 1 1 0 1 0
1 0 1 1 0 1 1 0 1 1
1 0 0 1 0 1 1 1 0 0
1 0 0 1 1 1 1 1 0 1
1 0 0 0 1 1 1 1 1. 0
1 0 0 0 0 1 1 1 1 1

The next step in the design process normally would be to produce five
five-variable K-maps for detennining B4, B3, B2, Bb and Bo as functions of
Col, G3, G2, Gh and Go- Note, however, that G4 and B4 are exactly alike. In
 5.5 Code Converters 151

addition, B3 is seen to be a 1 when only one of G4 and G3 is a 1; B2 is a 1

when only an odd number of G4 , G3 , and G2 are 1; BJ is a 1 when an odd
number of G4, G'j, Gz, and G1 are 1; and, finally, Bo is a 1 when an odd
number of the five Gray code inputs are 1. These observations eliminate
the need for the K-map minimization scheme, although as a general rule,
K-maps should be used and will always give a correct, if not minimum,
solution. As in this particular case, however, careful observations may at
times reduce complex problems to simpler ones. With our prior experience
the application of X-ORs to this problem should become obvious. The
binary outputs are obtained as follows:

B4 = G4

B3 = G3 E9 G4

B2 = G2 E9 G3 E!1 G4

B I = G1 E!1 Gz E9 G3 E9 G4

Bo = Go E9 G1 E9 G2 E9 G3 E!1 G4

Four two-input X-OR gates may be cascaded, as shown in Figure 5.20, to

realize the desired Gray-to-binary conversion circuit. This design could be
extended to an n-bit converter by cascading an additional (n - 5) two­
input X-OR gates.

FIGURE 5.20

::~9-r5-=i ==::
}---+----- 82

G1 - - - - f
'----
GO _____________ ~~BO

EXAMPLE 5.6 SOLUTION

Design an n-bit code converter for The truth table of Example 5.5 could be used again to solve for Go through
converting binary to the equivalent G4 in tenns of Bo through B4• Checking for patterns in the bit relationships
Gray code. to avoid going through a lengthy minimization process, we note that Go is a
1 only when either Bo or BI is a 1. Likewise, G1 is a 1 when either Bi or B2 is
152 CHAPTER FIVE Design of Combinational Circuits

a 1; G2 is a 1 when either B2 or B3 is a 1; and, similarly, Gn - J is a 1 when

Bn - 1 is a 1. Using our knowledge ofX-ORs we obtain

G3 = B3 EEl B4
G2 = B2 EEl B3
G1 = Bl EEl B2
Go = Bo EEl BJ
The logic circuit of an n-bit binary-to-Gray converter is obtained as shown
in Figure 5.21.

FIGURE 5.21

EXAMPLE 5.7 SOLUTION

.
Design a four-bit module that could The partially complete truth table for the binary-to-BCD conversion is
be used to obtain an n-bit binary­ obtained as shown in Figure 5.22. It can be seen that the LSB of both the
to-BCD conversion circuit. binary and the BCD numbers are the same. We could, therefore, design a
circuit that will convert the remaining binary bits to the corresponding
BCD bit: DJJ.fJ-Pl' A close examination of the truth table reveals that

(BJ33BzBl)
DJJ~-Pl =.

1
(BJ3~zBl + 0011) otherwise

The regular design ofthe conversion module would consist of obtaining the
minimized Boolean expressions for D4, D3 , D z, and Dl from the correspond­
ing K-maps as shown in Figure 5.23. In the K-map, however, the output
 5.5 Code Converters 153

FIGURE 522 Binary BCD

84 83 82 8 1 80 0 7 0 6 Os 0 4 0 3 O2 0 1 Do

0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 0 0 0 1
0 0 0 1 0 0 0 0 0 0 0 1 0
0 0 0 1 1 0 0 0 0 0 0 1 1
0 0 1 0 0 0 0 0 0 0 1 0 0
0 0 1 0 1 0 0 0 0 0 1 0 1
0 0 1 1 0 0 0 0 0 0 1 1 0
0 0 1 1 1 0 0 0 0 0 1 1 1
0 1 0 0 0 0 0 0 0 1 0 0 0
0 1 0 0 1 0 0 0 0 1 0 0 1
0 1 0 1 0 0 0 0 1 0 0 0 0
0 1 0 1 1 0 0 0 1 0 0 0 1
0 1 1 0 0 0 0 0 1 0 0 1 0
0 1 1 0 1 0 0 0 1 0 0 1 1
0 1 1 1 0 0 0 0 1 0 1 0 0
0 1 1 1 1 0 0 0 1 0 1 0 1
1 0 0 0 0 0 0 0 1 0 1 1 0
1 0 0 0 1 0 0 0 1 0 1 1 1
1 0 0 1 0 0 0 0 1 1 0 0 0
1 0 0 1 1 0 0 0 1 1 0 0 1

FIGURE 523 8 28 1 8 28 1
83 11 10 83 00 01 11 10
00 01
00 0 0 0 0 00 0 0 0 0
.--
01 0 1 1 1 01 1 0 0 0

11 - 1- - - 11 -
i--
- - -

10 1 1 - - '10 0 1 - -

. "': . .
.- - 8 28 1 8 28 1 ­
83 00 01 11 10 83 00 01 11 10

00 0 0
- 1 1 00 0 111 1\ 0

01 0 0 1 0 01 0 0 0
-1

11
r---
- - -
r--
- 11
r---
- - - --
I=-
10 1
I--
0
'--
- -
i....--i
10 1
r---
0 - -
"'--
154 CHAPTER FIVE Des' of Combinational Circuits

corresponding to the inputs 10-15 may be considered as don't-cares. The

equations for the BCD digits are found to be

D4 = B3 Bl + B3 B2 + B4
D3 = B.lJl + Bsl12Bl
D2 = B2BI + B4Bl + B3B2
DI = B~3Bl + B3 B2Bj + B~l

The implementation of the conversion module, as shown in Figure 5.24, is

now reasonably straightforward using the preceding equations.

FIGURES.24

I I I
The equations found for D4 , D3, D 2, and Dl are the equations that
would be implemented in a commercial binary-to-BCD Ie. Another solu­
tion using an adder would involve designing a circuit that outputs a 1
whenever B.lJ3B2Bl > 0100. The circuit output is tied to the least signifi­
cant two addend inputs of a foUr-bit adder while B4, B3, B2, and Bl are tied
to the corresonding augend inputs. The sum outputs of the four-bit adder
unit, as shown in Figure 5.25, would yield the desired BCD output.

FIGURES.2S

4-bit adder

Before the module designed in Example 5.7 is used for conver­

sions of more than five bits, a thorough understanding of the inter­
relationship between a binary number and its BCD equivalent
 5.5 Code Converters 155

number is necessary. The decimal value of an n-bit binary number

was given by Equation [1.4] as
(Nl)\O = ((... (((a n-l)2 + an-2)2 + an-3)2 + ... + a2)2
+ at)2 + ao
°
where each of the coefficients, aj, is either or 1. This nested multi­
plication by two can be carried out by shifting the binary number
(0.an-tan-zan-3 ... a3a2atao) to the left n times. As these bits are
shifted left, each group of four consecutive bits, beginning with the
binary point, represents a decade. Within each decade each left-shift
represents a multiplication by two. However, if a bit passes from the
MSB of one decade to the LSB of the next higher decade, its value
increases from 8 to 10 only (instead of 16). Figure 5.26 shows the
, nine possible BCD digits, their values after being shifted left without
correction, and their values after being corrected and then shifted.
Note that for values less than or equal to 0100, the shifted value
gives the correct BCD digit. If the BCD digit is greater than 01 00,
the shifted number can be corrected by adding a binary 0110 to the
shifted bits. An equivalent correction can also be made by adding
0011 to the BCD digit prior to the shifting. The device already
designed in Example 5.7 adds 0011 to the four-bit input if the four
bits represent a value greater than binary four.

FIGURE 5.26 Shifted BCD BCD Digits Shifted BCD Digits Corrected and
Digits. BCD Digits Left One Bit Then Shifted Left One Bit
0000 00000 00000
0001 00010 00010
0010 00100 00100
0011 00110 00110
0100 01000 01000
0101 01010 10000
0110 01100 10010
0111 01110 10100
1000 10000 1 0110
1001 10010 11000

In order to provide an understanding of the use of multiple

binary-to-BCD converters in converting binary digits of more than
five bits, an example involving seven bits will be solved with expla­
nations of each step. The steps will lead us to an algorithm that can
be used for the conversion of any n-bit number. The converter mod­
ule we designed is used to correct a BCD digit prior to the necessary
shifting mechanism. The process of shifting bits is achieved by
means of hard-wiring between separate stages of converter modules.
A separate stage of converter module would be necessary for each of
the shift operations.
156 CHAPTER FIVE Design of Combinational Circuits

EXAMPLE 5.8 SOLUTION

Convert 12710 = 11111112 to BCD. Step 1. Correct-Shift: Note that for the first two left-shifts of 0.1111111 2 no
converter module would be necessary. The third shift, however, results in an
integer larger than four. Therefore, the integer needs to be corrected before fur­
ther shifting of bits:
0111.1111 X 24 uncorrected
1010 . 1111 X 24 corrected BCD prior to shift
3
1 0101 . 111 X 2 hybrid number after shift

Mer completing Step 1 we have a hybrid number; the integer portion is in

BCD while the fractional portion is in binary. The value of this hybird
number is
[15 + (7/8)] X 23 = 12710

Step 2. Correct-Shift: The four bits on the immediate left of the binary point
could be larger than 0100, and in the present case they are. A new correct-shift
operation is required, therefore:
1 0101 . 111 X 23 from previous operation
1 1000 . 111 X 23 corrected BCD prior to shift
11 0001 . 11 X 22 hybrid number after shift

Also,
[31 + (3/4)] X 22 = 12710

Keep in mind that the shift operation just performed brought in a.bit from
the right of the binary point to be included in the BCD portion. This move
is valid since our four-bit input binary-to-BCD converter actually converts
five bits, the four connected and the one immediately to the right of those
connected.
Prior to shifting, we again add 0011 to those BCD digits that could be
greater than 0100. The left-most BCD digit can at most be 0011, so no cor­
rection prior to shifting is necessary.
Step 3. Correct-Shift:
11 0001 . 11 X 22 from previous stet>
11 0001 . 11 X 22 corrected BCD prior to shift
110 0011 . 1 X 21 hybrid nuinber after shift

The hybrid number giveS [63 + 112] X 21 = 127 10­

Step 4. The next correction prior to shifting requires two devices since there
are sufficient bits in both BCD, digits to have values greater than 0100. Again,
the left-most binary-to-BCD has one input coruiected to 0 to allow for the pOs­
sibility of a four-bit output with only three bits in:
 5.5 Code Converters 157

01100011 . 1 X 21 from previous step

1001 0011 . 1 X 21 corrected BCD prior to shift
10010 0111 .0 hybrid nwnber after shift

The process is now complete.

You would discover after reviewing the preceding steps that the bit posi­
tions can remain constant provided the converter modules are moved right
to effect a left-shift. The process gives the circuit of Figure 5.27.

FIGURE 5.27

Binary-to-BCD

Binary-la-BCD

04 03 O2 0,

I I
Binary-to-BCD

Binary-to-BCD Binary-to-BCD

Consider the pattern in the circuit of Figure 5.27. The conversion

of an n-bit binary number to BCD can be performed using the con-·
verter designed in Example 5.7 as follows:
1. Add a 0 to the left of the MSB position of the binary
number. Connect to the inputs of a four-bit converter mod­
ule a 0 and the three MSBs of the binary number.
158 CHAPTER FIVE Design of Combinational Circuits

2. Take the three least significant processed bits and the most
significant unprocessed bit as inputs to a converter module.
The most significant processed bit (MSPB), which is the MSB
of the left-most converter module, is considered in the next
step.
3. If there are three MSPBs, connect a 0 to the MSB position
of a converter module and the three MSPBs to the remain­
ing three inputs; otherwise carry the MSPBs from above
operations to the next leveL
4. Once a converter module is added to a level, the number of
converter modules remains the same until three MSPBs
accumulate, at which time another is added.
5. This process is continued until the LSB is the only unpro­
cessed bit.

EXAMPLES.9 SOLU110N
Use several of the modules designed The logic circuit necessary for obtaining the conversion of this II-bit
in Example 5.7 for converting the binary number follows directly from the previous discussion of the n-bit
binary number 110101101002 to its binary-to-BCD conversion algorithm. The resulting circuit configuration is
BCD equivalent. shown in Figure 5.28. After every three levels, the number of modules per
level increases by one. The output of the network is 1716BCD as expected.

EXAMPLE S.10 SOLUTION

Design a four-bit module suitable Since the LSBs of the binary and BCD numbers are equal, the design will
for the BCD-to-binary conversion of consider the four MSBs of the five-bit BCD value. The design involves
five bits. undoing the conversion of binary-to-BCD numbers considered in the pre­
ceding few examples. Intuitively we might assert that this will involve shift­
ing right and then correcting the resultant values by subtracting 0011.
Accordingly, the combinational module that could be used for this algo­
rithm has the following characteristics:

otherwise
1
j

 FIGURE 5.28 o o o 0

Binary-la-BCD

04 03 O2 01

1 0 10 11
B4 B3 B2 B1

Binary-to-BCD

04 03 O2 01

OJO J1 1
B4 B3 B2 Bl

Binary-la-BCD

Binary-to-BCD Binary-to-BCD

04 03 O2 01 04 03 O2 01

1 0 0 \0 10 0 1 1
B4 B3 B2 B1 B4 B3 B2 Bl

Binary-to-BCD Binary-to-BCD

04 03 O2 01 04 03 O2 01

o 0 10
\0 1 0 1 0
B4 B3 B2 Bl B4 B3 B2 B1

Binary-to-BCD Binary-la-BCD

Binary-to-BCD Binary-to-BCD Binary-to-BCD

04 03 O2 01 04 03 O2 01

o 1 0 0 \0 \0 1 0

Binary-to-BCD Binary-to-BCD Binary-to-BCD

I
~t
o
'----y---' '----.,y~--' '~--'Y--~
o

:';,!••.
;;:"

:p,;:
1 7 6

fi
.7,:·
.~
160 CHAPTER FIVE Design of Combinational Circuits

FIGURE 5.29
The truth table for such a module is obtained as shown in Figure 5.29.
The don't-ca.res in the binary table occur for six different input values that
I will never appear as uncorrected BCD digits. For example, a 0101 will never
0 4 0 3 O2 0 1 84 8 3 8 2 81 occur since that would imply the presence of an unlikely BCD number, either
0 0 0
1010 or 1011, prior to the shift-right operation. Examination of the truth table
0 0 0 0 0
indicates that BCD digits equal to or less than 0100 require no modification.
0 0 0 1 0 0 0 1
However, for BCD digits 8 through 12 a value of 3 must be subtracted for
0 0 1 0 0 0 1 0
correction. G:mstructing and minimizing the corresponding K-maps for B4, B3,
0 0 1 1 0 0 1 1
B2, and B1gives the following Boolean equations:
0 1 0 0 0 1 0 0
0 1 0 1 - - -
0 1 1 0 - - -
0 1 1 1 - - BI = DlJ4 + D)D4 = D) ED D4
1 0 0 0 0 1 0 1
B2 = D2D4 + DID~4 + D)D2
1 0 0 1 0 1 1 0
1 0 1 0 0 1 1 1 B3 = D2D4 + D)D2D4 + D)D3D"
1 0 1 1 1 0 0 0 B4 = D3D4 + D)D2D3
1 1 0 0 1 0 0 1
1 1 0 1 - -
1 1 1 0 - - - The circuit for the BCD-to-binary module may now be obtained as shown
1 1 1 1 -
in Figure 5.30.

FIGURE 5.30

01 ~

/~D 81

~~
V I \
1
.....-­ r- f-)
~
/~
BCD-to-Binary

1
~ \~L
I

/'

'-­ I
 5.5 Code Converters 161

EXAMPLE 5.11 SOLUTION

Use the module designed in As mentioned previously, this process should be the reverse of the binary­
Example 5.10 to periorm the to-BCD conversion scheme developed earlier.
conversion of 12710 and
consequently develop an algorithm Step 1. Shift the BCD quantity to the right by one bit, which will replace
for use inann-bit BCD-to-binary one bit to the right of the decimal. We are considering a hybrid number again.
conversion. Note that The digits to the right of the radix point will be binary and those to the left
BCD. The process of shifting results in uncorrected BCD values. The BCD-to­
12710 = 000100100111 BCD
binary converter is used to correct these uncorrected results:

1001 0011 . 1 X 21 uncorrected BCD

01100011 . 1 X 21 corrected BCD

The hybrid number is [63 + (112)] X 2 = 127 10­

Step 2. Note that the left-most 0011 needs no correction:

0011 0001 . 11 X 22 uncorrected BCD

0011 0001 . 11 X 22 corrected BCD

The hybrid number still gives [31 + (3/4)] X 4 = 127 10,

Step 3. Correct-Shift:
0001 1000. 111 X 23 uncorrected BCD
0001 0101 . 111 X 23 corrected BCD

And, [15 + (718)] X 23 = 127 10,

Step 4. At this point we have five bits of BCD left to convert to binary:

1010. 1111 X 24 uncorrected BCD

0111 . 1111 X 24 corrected BCD

And, [7 + (15/16)] X 24 = 127 10,

The resultant value 11111112 is indeed equal to 000100100111BCJ)' The
resulting circuit using the converter is obtained as shown in Figure 5.31.
An algorithm now may be developed for extending the process just
examined to the conversion of n-bit BCD numbers:

Step 1. Starting at the right end, skip the LSB and connect four-input,
BCD-to-binary converter modules to all remaining bits. If two or less bits
would be connected to the left-most converter, leave it off and extend the bits to
the next leveL If all bits at a level are connected to a converter module, the
MSB of the output will be zero and should not be carried to the next level.

Step 2. Skip the least significant processed bit at each level and reassign bits
162 CHAPTER FWE Design of Combinational Circuits

as directed in Step L Continue until the MSB (processed or unprocessed) from

an upper level is included as the most significant input to a converter. Remem­
ber not to bring down bits that would be zeros from upper MSB positions (see
Figure 5.31).

FIGURE 5.31 0 0 0 0
I
I I I I I ! I I
0 0 O 0
4 3 2 1 0 4 0 3 O2 0 1

BCD-to-Binary' BCD-to-Binary

84 83 82 81 84 83 82 81

BCD-to-Binary

84 83 82 81

BCD-to-Binary

BCD-to-Binary

5.6 BCD Arithmetic Even though most computers use regular binary numbers for their
arithmetic operations, some special-purpose computers and calcula­
Circuits tors operate in the decimal number system using BCD. BCD-based
systems require morem~mory to store information because of less
efficient coding and complex arithmetic circuitry. However, the
final results in these systems do not have to be decoded prior to dis­
play as decimal digits. BCD arithmetic is usually complicated by
the fact that some of the sums or differences are invalid. When two
BCD numbers are added on a binary adder, it is possible to obtain
16 different sUms, of which six are undefined. In addition, when
 5.6 BCD Arithmetic Circuits 163

there is a carry-out, four additional conditions turn out to he unde­

fined as well. Figure 5.32 shows the to-he-corrected and the corre­
sponding corrected BCD sums and carry. The largest uncorrected
sum could be 10011, corresponding to the sum of two BCD nines
and a carry-in of one. Note that the sums that are greater thaI'). 1001
are all undefined and are in need of correction.

FIGURE.5.32 Table for the

Uncorrected and Corrected BCD Uncorrected Corrected
Sums.
C3 S; S2 S; S~ C3 S3 S2 S1 So

~1
0 0 0 0 0 0 0 0 0
... . . . . .. . . . No correction needed
0 1 0 0 1 0 1 0 0
0 1 0 1 0 1 0 0 0 0
0 1 0 1 1 1 0 0 0 1
0 1 1 0 0 1 0 0 1 0
0 1 1 0 1 1 0 0
1 1
0 1 1 1 0 1 0 1 0 0
0 1 1 1 1 1 0 1 0 1
1 0 0 0 0 1 0 1 1 0
1 0 0 0 1 1 0 1 1 1
1 0 0 1 0 1 1 0 0 0
1 0 0 1 1 1 1 0 0 1

A BCD adder could be designed using the design techniques we

have already studied. Such an adder circuit would have a. total of
nine inputs and five outputs. Of the nine inputs, four inputs would
be for the augend, another four for the addend, and the ninth input
would he the carry-in. Of the five outputs, one would he for the
carry-out and the remaining four would he for the sums. Therefore,
the truth table would consist of 29 = 512 different input combina­
tions, many of which would lead to don't-care conditions. The two
obvious design choices for such a circuit would he

L Make use of a ROM or PAL or PLA.

2. Minimize the function using the Q-M technique and then
generate the circuit using combinational gates.

In either case, the d(,!Sign would he too involved to pursue further.

The easiest route to a BCD adder design is to base its operation
a
on four-bit FA module. If the sum of the corresponding nine
inputs exceeds nine, a carry is generated and the sum is corrected
by means of a correction circuit. It should be noted from Figure
5.32 that the corrected sum is obtainable from the uncorrected sum
simply by adding a (0110)2 to it. Furthermore, it can he seen that
C3 becomes 1 only when correction becomes necessary. The condi­
164 CHAPTER FIVE Design of Combinational Circuits

tion for the generation of a carry-out can be found, therefore, from

-the minimization of the K-map for C'j shown in Figure 5.33. The
Boolean expression for the corrected carry-out becomes
[523]

FIGURE 5.33 K.Map for the

Corrected Carry Bit.
S3 C3
- r-­
0 0 1 0 1 - - -
0 0 1 0 1 - - -

0 0 1 1 1 - - -

0 0 1 1 1 - - -
'-­
. _.--.,----'

S2

An analysis of our previous discussion would reveal that the

single-decade BCD adder would involve the following subunits:

a four-bit FA for addition,

a carry decoder circuit,
a second four-bit FA circuit for correction.

An adder made in this way is shown in Figure 5.34. The sum bits of
the first adder unit are introduced at the augend inputs of the sec­
ondadder unit. When the corrected carry-out, C3, is a 0, theuncor­
rected sum remains unchanged. When it is aI, the uncorrected sum
is added to 0110 to give the corrected sum at the output of the sec­
ond four-bit FA.
At this time the designer may be concerned with the amount of
worst-case propagation delays; The propagation delays basically
arise .from the two FA subunits used in this circuit. A close examina­
tion of the circuit, however, reveals that the correction does not
need all four of the single-bit FAs. So never needs any correction. An
improved version of the BCD adder may be obtained by having
only two HAs and on~ FA in the correction unit. This improved cir­
cuit is shown in Figure 5.35. Another approach for designing BCD
adders would be first to convert the BCD numbers into binary, add
the resulting binary numbers, and then transform the binary sum
into the correct BCD sum. This design requires both postcorrection
and preconversion circuitries and, therefore, is more expensive than
the one already designed.
 5.6 BCD Arithmetic Circuits 165

FIGURE 5.34 BCD Adder

Circuit.

4-bit FA

Sum
S3 So
s~ s;

Addend

4-bitFA

Next consider the design of a BCD subtracter. Subtraction is

more complex than BCD addition, since the possibility of having a
negative difference exists. Recall from earlier observations made in
Sections 1.4, 1.5, and 5.3 that there are at least two ways to handle
negative BCD numbers: either by 1O's complement or by 9's com­
plement. The technique using 10's complement would result in a
circuit similar to that of Figure 5.12. The scheme using 9's comple­
ment, however, would require an end-round-carry into the system.
The BCD subtracter using 9'8 complement co1J.ld be designed using
a BCD adder and a 9'8 complementer circuit. In order to obtain the
9's complementer circuit, consider the truth table in Figure 5.36.
Using the table ofFigure 5.36, we would obtain
0 3 = r;.r;.~ = II + 12 + 13
O2 = II E9 12
0 1 = II
0 0 = To
166 CHAPTER FIVE Design of Combinational Circuits

FIGURE 5.35 Improved BCD

Adder Circuit.

Addend

4-bit FA

Sum
FIGURE 5.36 Truth Table for So
the BCD 9's Complement.
S;

BCD Input 9's Complement

13 12 11 10 0 3 O2 0 1 0 0

0 0 0 0 1 0 0 1
0 0 0 1 1 0 0 0
0 0 1 0 0 1 1 1
0 0 1 1 0 1 1 0
0 1 0 0 0 1 0 1
0 1 0 1 0 1 0 0
0 1 1 0 0 0 1 1
0 1 1 1 0 0 1 0
1 0 0 0 0 0 0 1
1 0 0 1 0 0 0 0
Carry-out S3 Sl So

A BCD 9's complementer circuit using the preceding equations is

shown in Figure 5.37[a]. Figure 5.37[b] shows how a four-bit adder
may be used to achieve the same goal. The latter circuit performs
the l's complement of the BCD digits and then adds to it the 2's
complement of six (1010).

FIGURE 5.37 9's Complementer:

[a1 Using Simple Gates and [b1
Using a Four-Bit FA

11-~-.--;D- 0,
4-bit FA

'-------- 01

10 -----1[:>---- 00 '-----v-----'
9's complement

[a] [b]
 5.6 BCD Aritlunetic Circuits 167

Once the 9's complementer is designed, the rest of the steps are
obvious. If X-OR gates are used in place of the inverters of Figure
5.37[b], a combined BCD adder/subtracter unit may result. The
BCD adder/subtracter unit so designed is shown in Figure 5.38.
Each decade unit consists of three four-bit FAs. The first one works

FIGURE 5.38 A Single-Decade

BCD Adder/Subtracter Unit.

Augend/Minuend Addend/Subtrahend

.-------/-+---+--*---+-.--+-.......­ SUB/ADD

,I ['---_----.1 I I Jr------II 1
Augend Addend
- Co 4-bit FA
Sum

Augend Addend From the

r-- Co 4-bit FA Cj 1 - - - - - - next lower
Sum decade
position

To the
next higher
decade
position
,'-­

1 I
Augend Addend
4-bit FA Cj I (
Sum ..L
I-I
L - - - - - , - - - I, -I,,---'---'-1 ----I

Sum/Difference
168 CHAPTER FIVE Design of Combinational Circuits

as the 9's complementer unit, the second as the adder unit, and the
third works as part of the correction circuit. When the SUBI ADD
input is a 0, the unit performs simple addition as the complementer
unit leaves the addend unchanged. When SUBI ADD is a 1, the
subtrahend is complemented and is then added to the minuend.
Consequently, the sum appears at the final output when the SUBI
ADD input is a 0 and the difference is obtained otherwise. When
several such units are cascaded to make a multi-decade BCD
adderlsubtracter, the carry-out from the most significant decade
unit is fed into the carry-in of the least significant decade unit. This
arrangement guarantees the inclusion of an end-around-carry to
satisfy the need of the 10's complement system.

EXAMPLE 5.12 SOLUTION

Use the XS3 code for designing a As you can see from the truth table in Figure 5.39, the 9's complement of
BCD adder/subtracter unit. an XS3 number may be obtained simply by taking its 1's complement.
Consequently, it is easier to obtain the 9's complement of an XS3 than that
of its BCD equivalent.
It would be worthwhile to review Example 1.17 at this time. Notice that
correction would be necessary for obtaining the true XS3 sum. Corre­
sponding to the expected decimal numbers 3 and 7, the sum yielded 1001
and 1101, respectively, and no carry-out. However, 1000 and 0000 were
respectively obtained along with a carry-out in place of decimal numbers 8
and O. It can now be concluded that there are two simple rules that must
be followed while adding numbers in the XS3 code:
a. If a carry is produced, 0011 is added.
b. If a carry is not produced, 0011 is subtracted. This subtraction is
usually done by adding the 2's complement of 0011 (1101) and
ignoring any carry if so produced.

These rules should be better understood by examining the following cases.

Case I. 4710 011110102
FIGURE 5.39 + 3410 + 01100111 2
81 to 11100001 2
9's Complement + 1101 00 112 correction
Decimal XS3 ofXS3
1 1011 0100 XS3
0 0011 1100
(8 1)10
1 0100 1011
2 0101 1010 Case II. 011110102
3 ! 0110 1001 + 100110002 9's complement
4 0111 1000 1000100102
5 1000 0111 l "1 2
6 1001 0110
00110011 2 correction
7 1010 0101
8 1011 0100 0100 0110
9 1100 0011 (1 3ho
 5.6 BCD Arithmetic Circuits 169

Case III. 3410 0110 0111 2

- 47 10 + 10000101 2 9'8 complement

- 13 10 1110 11002
1101 1101 2 correction

1011 1001 2
-+ 0100 0110 2 = - (13)10

A positive result is indicated when adding the 9's complement results in a

carry-out of the MSB. This carry-out is used as the end-around-carry. Fig­
ure 5.40 shows a BCD adder/subtracter circuit that has incorporated the
rules of XS3 addition and subtraction.

BCD BCD
Augend/Minuend Addend/Subtrahend

XS3 XS3
converter converter

To
next +----+-+--+--+--------------+-t--.--t--1f----t--t---.-- SUB/ADD
higher

decade

Carry to
next higher Augend Addend Carry-in from
decade
- - . - - - f Co 4-bitFA next lower
Sum decade

Augend Addend
4-bit FA

~_r~~--!-._-+_~--_r~f__------INVERT

~----------~v~-----------
Sum/Difference
170 CHAPTER FIVE Design of Combinational Circuits

When the SUB/ADD input is a 0, the addend arrives at the first adder
along with the augend. Otherwise, the complement of the subtrahend is
added to the minuend. The output of the first adder is added to either 0011
or 1101, depending on whether or not a carry is generated. The INVERT
input is activated when there is no end-around-carry, and in that event the
output is complemented to yield the final result.

5.7 Arithmetic Logic The examples presented in previous sections have shown some of
the multi-bit arithmetic functions that are used in digital systems.
Unit (ALU) Many times it may as well be necessary to perform bit-by-bit logic
operations between two multi-bit operands. A multi-function circuit
that can operate on groups of bits, therefore, proves to be extremely
advantageous in many complex digital systems. Combinational
design techniques generally are used to design such multi-fonction cir­
cuits, otherwise known as arithmetic logic units (ALUs). The various
operations are usually selected by means of several control or select
lines.
The various logic operations are realized by routing the inputs to
circuits that perform various logic functions and using a MUX to
select anyone of the possible logic operations. There are commer­
cially available ALUs that operate on two four-bit values with 32
arithmetic and 16 logic operations selectable by the combination of
four control inputs, a mode selector, and a carry-in. In this section
the design of a relatively less complex ALU will be considered to
demonstrate the process and some typical functions.
Consider an ALU with two four-bit inputs, A3A2AIAo and
B3B2BIBo. Let us fIrst consider bit-by-bit logic operations of several
types. Between any two inputs, X and Y, there could be a total of 16
different types oflogic outputs: 0, 1,.X, Y, X, Y, X + 1', X + 1', X
+ Y, Xl', Xl', xY, X + Y, XY, X EB Y, and X EB Y. All of these
outputs are achievable by means of logic gates.
Figure 5.41[a] shows a logic circuit consisting of a l-of-8 MUX
and only eight logic gates. This circuit is able to perform up to eight
different logic operations between its inputs, Ai and B,. A specillc
operation is selected by means of the three control inputs: S2, Sh
°
and So. The enable input, S3, is set to a to enable the MUX in all
of these eight cases. The truth table in Figure 5.41 [b] lists the possi­
ble logic operations and the corresponding control conditions. For
example, when control input is O1QO, the D4 input of the l-of-8
MUX is activated, and consequently AtBi becomes available at the
MUX output. This l-of-8 MUX could have been replaced with a
l-of-16 MUX for realizing 16 different logic operations. In that
case, however, an additional control input would be necessary.
The arithmetic section may be designed around a four-bit ripple
or CLA adder circuit. If the carry-in is utilized, it is possible to
obtain a larger number of arithmetic operations for the. same
 5.7 Arithmetic Logic Unit (ALU) 171

FIGURE 5.41 Logic Section of

an ALU: [a] Circuit and [b]
Function Table.

S3

Enable
A; 00

B; 01

O2

03

1-of-8 T
MUX
04 f F;

!
Control Inputs Output

S3 S2 S1 So Fi

0 0 0 0 Ai
0 0 0 1 Bi
0 0 1 0 A;Bi
0 0 1 1 Ai+ B;
0 1 0 0 A;B;
0 1 0 1 A;+Bi
0 1 1 0 Aj(t)B j
0 1 1 1 AEBB;
[a] [b]

number of control inputs. For every control condition the arithmetic

output when Cj = 1 is always one greater than the corresponding
output when q = O. Figure 5,42[a] shows an arithmetic circuit cor­
responding to a single-bit input. It may be used to obtain a total of
16 different arithmetic operations. This circuit consists of a 1-of-8
MUX, a full adder, -and five logic gates. In order to differentiate
,;1~
~~.,
between the arithmetic and the logic operations, 3'; is used to select

I
arithmetic operations. The output of the MUX unit is fed into one
of the adder inputs while another of the FA inputs is tied to Ai
ilf~ directly.
Ii ~~:~
Figure 5.42[b] lists all of the arithmetic operations and the corre­
sponding control inputs needed for operating this arithmetic unit.

1
-2":
-,.;."< The MUX output carries into the adder either a function of both Aj
<:·········
"
---
~~

~\ .

172 CHAPTER FIVE Design of Combinational Circuits

FIGURE 5.42 Arithmetic Unit of

an ALU: [a] Circuit and [b]
Function Table.

E
DO

~C> 01

0 O2

1 .03

~ 1-of-8 fp-
04 MUX
f

L
05
~ -

1 L
06

07 ~
Xi Yi
12 11 10
From
FA! Cj +-­ FA -
IS2 ISl Iso i 1

-
Co Sf

(8] To FA i + 1 I ~
Control Fi

S3 S2 Sl So Cj=O Cj = 1

1 0 0 0 Ai+Bi A;+ Bj + 1
1 0 0 1 . A;+ H; A+ 8 j + 1
1 0 1 0 A Aj + 1
1 0 1 1 A; ­ 1 A
1 1 0 0 Aj+AiBi A+ABi+ 1
1 1 0 1 A+Ail
.f
J" I A+AiBi+ 1
1 1 1 0 Ai + (Ai + Bj) Ai + (A + Bj ) + 1
1 1 1 1 Ai + (Ai + 8d A + (Ai + B;) + 1

[b]
 5.8 Decoders and Encoders 173

and Bi, or a 0, or a 1. For example, when the control input is 1100,

the D4 input of the MUX is activated, which introduces Aj, Aft;,
and the carry-in as inputs to the FA. Thus the sum becomes Ai plus
Afti in the absence of a carry-in, and Ai plus AiBi plus 1 otherwise.
Consequently, four l-of-8 MUXs, twenty (5 X 4) discrete gates,
and a four-bit FA may be connected to fonn an aritlunetic unit for
processing two four-bit binary inputs. Note when the control input
is 1011 and q = 0, the four bits of A are added to a string of four
Is. This operation is equivalent to adding A to the 2's complement
of 0001.
Both of these units, aritlunetic and logic, may now be combined
together to make an ALU, as shown by the block diagram in Figure
5.43. The two basic units could be internally ORed together to pro­
duce the ALU output. The ALU processes four bits of A and four
bits of B simultaneously. If S'j = 0, one of the eight logic operations
would be realized, and when S'j = 1, one of the 16 arithmetic oper­
ations would be available. Consequently the ALU consists of four
single-bit arithmetic units and four single-bit logic units, and it
could perfonn a total of 24 different operations. If the word size
exceeds four bits, several four-bit ALUs may be cascaded by tying
the carry-out of one ALU to the carry-in of the next ALU.

FIGURE 5.43 Block Diagram of

a Four-Bit ALU. 4/
I A
4/
4/ F I
/ B
ALU
4/ 1/
$3-S0
I Co I

1/
I
r Ci

5.8 Decoders and Very often in digital systems it is necessary to convert one code to
another. The process that detennines what character, dIgit, or
Encoders
number a code represents is called decoding. A decoder is an integral
part of this process. It is a specially organized combinational circuit
that -translates a code to a more useful or meaningful fonn. In this
section we shall look at just a few of the many decoding functions.
One of the frequently used decoders is a BCD-to-seven-segment
decoder. This particular type of decoder accepts as inputs BCD and
provides outputs to drive a seven-segment LED display device, as
shown in Figure 5.44, in order to decode bits into readable digits.
The decimal inputs to a circuit are first changed to equivalent
binary fonn by means of BCD-to-binary converter modules for the
desired binary operation. The resultant binary output is finally
reconverted back to equivalent BCD output and is usually dis­
174 CHAPTER FIVE Design of Combinational Circuits

FIGURE 5.44 Seven-Segment played by means of seven-segment display devices. These devices
Display Device. are used often in calculators.
The display device consists of seven light-emitting segments that
represent each of the ten decimal numbers when activated in suit­
able combination. As an example, segments a,j, g, c, and d have to
be illuminated to represent a 5. There are two choices for represent­
ing a 1: eitherf and e or band c. The normal decimal code for these
indicators is shown in Figure 5.45. A Boolean expression corre­
sponding to each segment of the display may now be found. The
Boolean equations for the illumination of display segments may be
obtained as follows:
a(D,C,B,A) = (D + C + B + /1) + (D + 'C + B + A)
+ (D + r: + B + A)
b(D,C,B,A) = (D + C + B + /1) + (D + r: + B + A)
+ (D + r: + B + A)
c(D,C,B,A) = (D + C + B + A) + (D + C + B + A)

d(D, C,B,A) = (D + C + B + A) + (D + r: + B + A)
+ (D + r: + B + A) + 7J5 + C + B + 11)
e(D,B,C,A) = (D + C + B + 11) + (D + 'C + B + A)
+ (D + r: + B + A) + (D + C + 71 + A)
+ (D + C + B + A)

f(D,C,B,A) = (D + C + B + A) + (D + C + B + A)
+ (D + C + B+ A)
g(D,C,B,A) = (D + C + B + A) + (D + C + B + A)
+ (D + r: + B + 11)

FIGURE 5.45 Truth Table for

Segments
Seven-Segment Decoding
Function. Number a b c d e f 9
o DCBA 1 1 1 1 1 1 0
1 DCBA 0 1 1 0 0 O. 0
2 DCBA 1 1 01 1 0 1
3 DCBA 1 1 1 1 0 0 1
4 DCBA 0 1 1 0 0 1 1
5 DCBA 1 0 1 1 0 1 ,
6 OCBA
7 [5CBA ,
0 0 1 1 1 1 1

8 DCBA
1 1 0 0 0 0
1 1 1 1 1 1 ,
9 DCBA 1 1 1 0 0 1 1
 5.8 Decoders and Encoders 175

These Boolean expressions for the segments lead to the implementa­

tion of the desired decoder circuit as shown in Figure 5.46. The
NOR circuit requires 16 gates and a total of 59 gate inputs while
the NAND circuit requires 17 gates and a total of 87 gate inputs.

FIGURE 5.4{) BCD-to-Seven­

SegmentDecoder Circuit: [a]
Using NOR Gates and [b] Using
NAND Gates.

:§J-x o
c---\

;D-x. A - - - r ._ __'"

b
C -----\
A ~-__i. _______

~:E[}-x,
d
C-----\
A
raj

CiL}-D
_ B
A
.. /0
::8<-__)-"
~B )-,. ~i3 )-'5

~§ )-" :f3 )-"

~ 131-__)-" ~B r',

~B )-'9
176 CHAPTER FIVE Design of Combinational Circuits

We shall consider next a slightly different decoder scheme that

accepts n bits of binary information and converts them to up to 2n
unique outputs. For example, two bits of binary information would
result in four unique output values. One such decoder scheme is
shown by the circuit of Figure 5.47. This decoder is commonly
known as a 2-4 line decoder. If the enable input E is 0, the decoder
is enabled, and when E is 1 the decoder is disabled. When E is 0, all
outputs are 1 except for the line corresponding to the decimal value
of the input bits. For example, when AB = 00, only Do is 0. Such a
scheme would allow us to activate four different circuits by means of
the four decoder outputs. In that case the operations of these four
peripheral circuits could be controlled by means of two select
inputs, A and B.

FIGURE 5.47 2-4 Line Decoder:

[a] Block Diagram and [b] Circuit.

Do

E ~~o!

DO 10

A 'I 2-4 DIP

line O2
B 10 decoder D2 E

D3 P D3

[a] [b]

With proper connections a demultiplexer circuit can be made

from a decoder. A demultiplexer (DMUX) receives data on a single
entry line and outputs this data on one of its many output lines. Fig­
ure 5.48 shows a 1-4 line DMUX circuit where the decoder inputs
are treated as the DMUX selects. The enable input E is connected
to the input data, which appear at the output specified by the val­
ues of A and B.
There are times when several decoder/DMUX circuits may be
cascaded together to form a larger decoder/DMUX. Figure 5.49
shows how two 2-4 line decoders are combined by means of their
 5.8 Decoders and Encoders 177

FIGURE 5.48 1-4 Line

Demultiplexer: [a] Block Diagram
and [b] Circuit.

Do
E

Do
01
1-4 01
E line
DMUX O2
I, O2

03

10 03
A B
[a] [b]

enable inputs. When C = 1, only the top decoder is enabled and

the lower one is disabled. When C = 0, the top decoder is disabled
and the bottom decoder is enabled. The combined circuit functions
as a 3-8 line decoder.
The decoders that have been discussed thus far are often con­
structed using transistors in AND formation. The number of transistors
used in each of the gates is approximately equal to the number of
inputs to each gate, and the number of gates present is on the order of
the number of decoder outputs. Therefore, the number of transistors
required in a decoder circuit increases as 2n with increasing inputs.
O:msequendy designers would like to see a decoder scheme whose sum
of gates and gate inputs is reasonably small. Figure 5.50 shows a par­
ticular configuration for a 4-16 line decoder that uses a reduced
number of transistors. Such a configuration is commonly known as the
tree-type decoding network. An examination would show that 64 tran­
sistors are needed to fabricate such a circuit. OJmparatively, a regular
4-16 line decoder designed similarly to that shown in Figure 5.47
would require a total of 72 transistors.
The decoder network also may use an innovative scheme called
the balanced decodingScheme, illustrated in Figure 5.51. It requires
only 56 transistors. The significance of this improvement becomes
more important as larger decoders are considered. The regular
decoder network of Figure 5.47 is still the fastest because it involves
only two stages of NAND gates. The inclusion of an enable input
in a 4-16 line decoder, however, would involve an additional 18
transistors.
178 CHAPTER FIVE Design of Combinational Circuits

FIGURE 5.49 3-8 Line Decoder

Using Two 2-4 Line Decoder
Units. Do ~

A /1 2-4 Dl p--
line
decoder D2 p-
B 10

D3 b -
E

c .~

Do J-

- 11 2-4 Dl J-
line
decoder
10 D2 0--

D3 P-
E
)

An encoder is· a combinational circuit that accepts a digit on its

inputs and converts it to a coded output. In fact, an encoder
reverses the function of a decoder. As an example, let us consider
the design ofa decimal-to-BCD encoder. The device has a total of
10 inputs-one for each decimal digit-and four outputs to represent
the corresponding BCD numbers. These decimal inputs could possi­
bly be the keys on a hand-held calculator.
The truth table for the encoder is listed in Figure 5.52[a] from
which the following expreSsions may readily be obtained:
A=I+3+5'+7+9

B=2+3+6+7

C=4+5+6+7

D=8+9

The resulting circuit is very straightforward, as shown in Figure

5.52[b].
 5.8 Decoders and Encoders 179

FIGURE 5.50 4·16 Line Tree­

Type Decoder: [a] Block Diagram
and [b] Circuit.

00
0 1 P- -
O2 P- 00
0 3 P-
0 4 P-
- -
-
~
13 Os I-

4-16 06
- 12 0 7 P-
line
- 11 decoder 0 8 P- I-

~
09
- 10 0 10
0 11 P-
0 12 P- ~
0 13 P-

~
0 14
0 15
~

(a]

r-
)- -
~

I-
~

--
}-~
-r= ,.......
-
I-

-
08
r- f--

-rt= ~ ~

'+;":
"1;
)- '-
"
.... i--

-~ '--

I- '--

..

---<
.t l - i--

I--

I
~

I...--
[b1
180 CHAPTER FIVE Design of Combinational Circuits

FIGURE 5.51 4w16 Line

Balanced Decoder Circuit.

Do
r--

D1
r- l -

-
-

I-

rL
-
f ~ - J

r - l~ J
L
./

r
r--
D8

-
r ~ -
L
J ~ -
-
D11
10 r
L

'-

'---

'---
 5.8 Decoders and Encoders 181

FIGURE 5.52 Decimal-to-BCD BCD

Encoder: [aJ Truth Table and [b] 9 --------t----4
Circuit. Decimal Input DCBA
8 ----+----1
o
0 0000 '--­
1 0001 7----~~-~
6 ---H-i'--l
2 0010 5 ---_...I-H---I c
3 0011 4 ---t-t-H'r-L---""
4 0100
5 0101 B
0110 3 ---'-+-1--1--1
6
2 --t-f-I--If-L_--""
7 0111
8 1000
9
A
1001
1-----"-_
[a] [b]

EXAMPLE 5.13 SOLUTION

Consider the seven-segment display This design may be accomplished by modifying the already-designed
device of Figure 5.44. Design a BCD-to-LED display of Figure 5.46. Each unit accepts four BCD inputs
scheme to store up to five BCD and outputs seven LED outputs, as shown by the block diagram of Figure
integers such that the left-most 5.53. The circuit of Figure 5.53 requires an added feature that suppresses
BCD zeros are not displayed. For the display of leading zeros. First it must be determined if the MSB is a O.
example, 00932 should be displayed If it is, the next significant bit is tested, and so on. The first nonzero bit
as only 932. stops further testing. Accordingly, one may design a circuit for the ith bit if
it is a zero or not, as shown by the block diagram of Figure 5.54.

FIGURE 5.53
o

C
----l BCD-to-
7,-segment
a
b
c
d I I
LED e
decoder

I
B
f

A
9
I

A high TDS input to the decoder implies that Tftis Decoder needs to he
§.earched. A high NDS output similarly implies that the Next Decoder needs to he
Searched as well. The NDS output of one input may be introduced to the
~ext unit on the right as the IDS input, and so on. This feature may be
accomplished by adding a circuit such as shown in Figure 5.55 to the
already available circuit.
182 CHAPTER FIVE Design of Combinational Circuits

FIGURE 5.54 DeB A

1 1

TDS - - + I BCD-to-7· segment \---.. NOS

decoder

I-I
I-I

FIGURE 5.55 C D
A B X > - - - - - - NOS
TDS-----~-~

""-_ _ _ _ _ T
o all the 10 first-stage
NAND gates of Figure 5.46[b]

When the BCD digit corresponds to a 0 and the TDS input is high, the
above circuit disables the decode circuit and suppresses the display. At the
same time the resulting NDS becomes a 1, resulting in further search for
zeros if there are any. If the BCD digit is not a 0, the .decoder.is..not dis­
abled and further search is abandoned. The overall five-bit circuit is
FIGURE 5.56 obtained as shown in Figure 5.56.

BCD Inputs

f4 f4 f4 f4 f4
Decoder Decoder Decoder Decoder Decoder
- TDS NOS ­ TDS NOS - TDS NOS ,....-­ TDS NOS r - ­ TDS NOS ­
17 17 f7 17 17

,,
I-I I
-

I
-,,
-
-,-, I-I
1
-
 5.8 Decoders and Encoders 183

EXAMPLE 5.14 SOLUTION

Design a four-input priority encoder The block diagram for such a device may be as shown in Figure 5.57. As a
such that when two inputs, D j and beginning step, the corresponding truth table needs to be known. The truth
Dj , ard:tigh simultaneously, Di has table for such a device is easily obtained as shown in Figure 5.58. The
priority over Dj when i > j. The don't-cares are introduced under input columns whenever appropriate.
encoder produces a binary output The Boolean equations for the outputs are obtained directly as follows:
code corresponding to the input
that has the highest priority. fo = Ds + DID2
It = D2 + Ds
The four-input priority encoder circuit is obtained accordingly, as shown
in Figure 5.59. The request indicator, M, shows whether or not any of the
four inputs are active.

FIGURE 5.57 Request

indicator

00
FIGURE 5.58 to
01
Inputs
O2
t1
0 0 0 1 O2 0 3 t1 to 03
1 0 0 0 0 0
E
1 0 0 0 1
- - 1 0 1 0
- -
"
"

- 1
·"~'i:";":":"
1 1
Enable

FIGURE 5.59

00
to
01
0'
2,

11
D3
Enable
184 CHAPTER FIVE Design of Combinational Circuits

5.9 Error-Control Errors may occur as digital codes are transmitted from one system
or subsystem to another. There is always a possibility, albeit small,
Circuits that a random-noise pulse will change a zero to a one or a one to a
zero. It is possible, however, to code the data so that the occurrence
of an error can be detected after the data have been received. The
simplest approach is to add an extra bit, called a parity bit, to each
of the number codes. If the coded data including the parity bit have
an even number of ones, the code is said to have even parity. If the
coded word including the parity bit has an odd number of ones, the
code is said to have odd parity. Prior to sending a code, the number of
ones are counted and the parity bit is set to make the number of
ones odd or even as determined by the parity scheme chosen. At the
receiving end, a check is made to see how many ones are present in
the coded word. If odd parity is used and an even number of ones
are received, it implies that an error has occurred. However, the
proposed parity bit scheme cannot detect the occurrence of an even
number of errors. For situations where the probability of multiple
errors is high, a more sophisticated coding scheme must be used. In
computers or communications equipment the possibility of random
noise causing changes in more than one bit is low.
In order to check for or generate the proper parity bit in a given
code, it is necessary to determine whether an odd or even number of
ones are present. An X-OR gate functions in such a way that the
output of an even number of ones is always a 0, and the output of
an odd number of ones is always a 1. As an example, the circuit of
Figure 5.60 makes use of these gates to generate an even parity bit
for normal BCD input and to check for a possible error that may
have been caused during transmission. The parity generator circuit
at the source end examines the contents of the four data lines and
.accordingly generates a parity bit so that the encoded message (five
bits in all) has even parity. At the receiver end the parity-checking
circuit determines if an error has occurred or not. A high output at
the parity-checking circuit indicates the occurrence of an error dur­
ing transmission. This circuit could be made suitable for odd parity
by simply replacing the final X-OR gates of both generator and
checker circuits with X-NOR gates.
Several other schemes are also available, especially for coding
decimal digits. The most common of these are the 2-out-of-5 code
and 2-out-of-7 code. They are listed along with parity-coded BCD
in the table of Figure 5.61. The 2-out-of-7 code is also known as the
biquinary code. The zeroth through the sixth bit have positional
weights ofO, 1,2,3,4,0, and 5 respectively. In both of these m-out­
of-n codes, there are m ones and (n - m) zeros. The advantages of
these schemes are understood by comparing the different permissi­
ble codes. Two codes are said to be at a distance p if the codes differ
from each other in p locations. Clearly~ each code in an m-out-of-n
 5.9 Error-Control Circuits 185

FIGURE 5.60 Even Parity

Generator-Checker Circuit.

Source Receiver
Ds---*----------------------------------+---------------------------Ds
D4 D4
~ .;' ~

~ ~

Generator
Checker

FIGURE 5.61 Some Codes with

BCD with
BCD with

Error Control. Decimal Even Parity

Odd Parity
2-out-of-5 2-out-of·7
0 00000 00001 00011 0100Q01
1 00011 00010 00101 0100010
2 00101 00100 00110 0100100
3 00110 00111 01001 0101000
4 01001 01000 01010 0110000
5 01010 01011 01100 1000001
6 01100 01101 10001 1000010
7 01111 01110 10010 1000100
8 10001 10000 10100 1001000
9 10010 . 10011 11000 1010000

scheme is at least distance two away from the next code. Conse­
quently, these codes can be used to detect single errors.
To correct k errors the minimum distance between two code
words must not be smaller than 2k + 1. A total of k errors would
produce an error word· k distance away from the correct code word.
To be able to correct this error, no other Ie errors should be able to
produce this same error word. The error word, therefore, should be
at a distance at least k + 1 from any other code word. Accordingly,
the minimum distance between two code words should be 2k + 1.
A minimum <iistance of two provides single-error detectability; any
single error moves the code closer to where it was than to any other
186 CHAPTER FIVE Design of Combinational Circuits

possible code. A circuit could be designed to move it back to its cor­

rect position. Of course, this minimum distance code could also be
used instead for double-error detection. A minimum distance of four
will provide both single-error correction plus double-error detec­
tion. A minimum distance of five would allow double-error correc­
tion. Next we will examine a code for which both error detection
and error correction are straightforward.
One of the most useful codes is the Hamming code. This scheme not
only provides for the detection of an error, but also locates the bit
position in error so that it may be readily corrected. A block dia­
gram of the implementation of this scheme is shown in Figure 5.62.
The mbits of data are encoded with p parity bits before transmis­
sion. The received word is tested by a checking circuit to see if any
FIGURE 5.62 Block Diagram of error has occurred. The decoder then locates the exact position of
the Hamming Code. error and, accordingly, the data are corrected by a corrector circuit.

rri +p m+p m
~ Encod~r Channel , Corrector ~

11m +p
12f!

P
Checking circuit Decoder t-­

The Hamming code uses multiple parity bits placed at specific

locations of the coded word. The general rules for the generation of
parity bits are summarized as follows:

1. If m is the number of information bits, then the number of

parity bits, p, is equal to the smallest integer value of p that
satisfies 2P m + p + 1.
2. Parity bits are placed at the locations 1,2,4,8, 16, and so
on, of the coded word. Information bits are placed in order
at locations 3,5,6, 7,9, 10, and so on.
3. Each parity bit individually takes care of only a few bits of
the coded word. To determine the bits of the coded word
that are checked by a parity bit, every bit position is
expressed in binary. A parity bit would check those bit

j
positions, including itself, that have a 1 in the same loca­
 5.9 Error-Control Circuits 187

tion of their binary representation as in the binary repre­

sentation of the parity bit.

Example 5.15 illustrates the coding mechanisms involved.

EXAMPLES. IS SOLUTION
Determine the Hamming-coded The number of bits is m = 5; therefore, p = 4. This coding would result in
word for the,m,essage 10 10 1 \iSing a nine-bit coded word. The corresponding message bits (10101) are posi­
even parity. tioned respectively in locations 3, 5, 6, 7, and 9. These locations are speci­
fied in the table of Figure 5.63 as M I , M 2, M s, M'I> and Ms respectively. To
determine the exact value of each parity bit, each of the position designa­
tions is expressed at first in binary. The parity bits are generated from the
following observations:

a. PI checks bit positions 1, 3, 5, 7, and 9 since all of these locations

have a 1 in the LSB of their binary representations. The message
bits present at four of these locations are, respectively, 1,0,0, and
1. The parity bit at location 0001, therefore, should be a 0 to
maintain an even parity.
h. P2 checks bit positions 2, 3, 6, and 7 because they all have a 1 at
the same location in their binary representations. Bits 3, 6, and 7
house, respectively, a 1, 1, and O. Therefore, P2 must be a O.
c. Ps checks bit positions 4, 5, 6, and 7. Bits 5, 6, and 7 house,
respectively, 0, 1, and 0, which requires that Ps be a 1.
d. P4 checks bit positions 8 and 9 and should be a 1 to maintain an
even parity.

Therefore, the coded word is 001101011.

Bit Designation Pl P2 Ml P3 M2 M3 M4 P4 Ms
Bit Position 1 2 3 4 5 6 7 8 9
0001 0010 0011 0100 0101 0110 0111 1000 1001

Message Bits 1 0 1 0 1
I
Parity Bits 0 0 1 1
I

As mentioned earlier, the Hamming code also provides a means

to detect and correct a single error. The general detection algorithm
188 CHAPTER FIVE Design of Combinational Circuits

consists of the following steps:

step 1. Check parity on each parity bit Pn and the bits for which it
provides parity.
Step 2. If the test indicates the preservation of assumed parity, a 0
is assigned to the test result. A failed test is indicated by a 1.
Step 3. The binary number formed by the score of parity tests

indicates the location of the bit in error.

Example 5.16 illustrates the detection mechanisms.

EXAMPLE 5.16 SOLUTION

Detennine if any bit is in error in The bit position table, as shown in Figure 5.64, is first prepared and then
the coded word 0011 01111. The the coded bits are placed in their proper places. The following observations
message was coded with even can be made regarding the coded word:
assumption.
PI checks bits 1,3,5, 7, and 9. Consequently the
first test fails since there are three Is-..:; 1 (LSB)
P2 checks bits 2, 3, 6, and 7. This test also
fails since there are three 1s -..:; 1
P3 checks bits 4, 5, 6, and 7. This test also
fails since there are three 1s -..:; 1

P4 checks bits 8 and 9. This is a good check

since there are two Is -..:; 0 (MSB)

The test score is 0111. The bit in error is the seventh bit (0111); therefore,
the seventh bit is changed to O. Therefore, the correct coded word should
FIGURE 5.64 be 001101011. This result agrees with the earlier findings of Example 5.15.

Bit Designation P1 P2 M1 P3 M2 M3 M4 P4 Ms
Bit Position 1 2 3 4 5 6 7 8 9
0001 0010 0011 0100 0101 0110 0111 1000 1001
Received
Message 0 0 1. 1 0 1 1 1 1

The hardware implementations of the Hamming code scheme

are relatively easy to achieve. Consider as an example a logic circuit
for processing five bits of data, just as in the last two examples. The
parity bit generator circuit is to generate four additional bits: Pi) P2,
 5.9 Error-Control Circuits 189

P3, and P4' The rules for the generation of parity bits, as stated ear­
lier, may be used to obtain

PI = Ml EB M2 EB M4 EB Ms

P2 = Ml EB M3 EB M4

P3 = M2 EB M3 EB M4

P4 = Ms

The generator circuit may be implemented with X-OR gates. The

resultant circuit is shown in Figure 5.65. Correspondingly, for an
odd parity assumption the final X-OR gates of all stages need to be
replaced with X-NOR gates and M5 inverted to generate P4 • It is
important to note that the minimum distance between two code
FIGURE 5.65 Hamming-Coded words is three because one change in the data bits produces at least
Message Generator. two changes in the parity bits.

Using similar reasoning the parity-checking circuit may also be

designed. The parity of each parity bit and its corresponding data
bits are tested. The Boolean "equations for the nine-input parity-test­
ing circuit are readily obtained as follows:

C1 = Y1 EB Y3 EB Y5 EB Y7 EB Y9

C2 = Y2 EB Y3 EB Y6 EB Y7

C3 = Y4 EB Ys EB Y6 EB Y7

C4 = Ya EB Y9

190 CHAPTER FIVE Design of Combinational Circuits

where Yb Y2, ••• , Ya, and Yg correspond, respectively, to Ph P2, Mb

Ps, M21 Ms, M4, P4, and Ms of the generator circuit. The parity­
checking circuit is again realized using X-OR gates and is shown in
Figure 5.66[a). The value C4C3C2C1 points to the bit in error and

FIGURE 5.66 [a1 Parity­

Checking Circuit and [b1
Correction Circuit.

[)-C2

~---------------------------------------~
[a]

Corrected
data
bits

I
 Problems 191

may be used to correct the bit. Our interest is in recovering the cor­
rected data bits. Making use of the X-OR programmable inverter
function, the equations for the corrected bits are obtained as
follows:
D\ = (C4 ' C3 • G2 • G1) EB Y3

D2 = (C4 • G3 • C2 • G1) EB Y5

D3 = (V4 • G3 • G2 • VI) EB Y6

D4 = (C4 • G3 • Gz • G1) EB Y7

D5 = (G4 ' ('3 • V2 • G1) EB Yg

For example, if the parity test results are 0111, then the fourth data
bit (i.e., the seventh bit of the coded word) is in error. The bit in
error is corrected by complementing Y7• All of the other bits remain
unchanged. The corresponding error-correcting circuit is shown in
Figure 5.66[ b].

5.10 SUlDlllary In this chapter different aspects of combinational circuit design

were presented. Primary emphasis was placed on designing smaller
and manageable modules, several of which were then cascaded
together to realize a robust system. In particular, the design and
working principles of binary and nonbinary adders/ subtracters,
code converters, decoders, encoders, and various error-correcting
circuits were explored. Many of these devices will be used time and
again throughout the rest of this text for developing more advanced
concepts.

Problems 1. Design an FA circuit using logic gates suitable for adding two
bits of addend, two bits of augend, and carry-in input.
2. Obtain a single-bit FA using only MUXs.
3. Design a single-bit FA using only NOR gates.
4. Use the FAs designed in Problem 1 to perform addition of six­
bit numbers. Show the configuration of the setup for adding
(llOllOh and (OOOOlOh.
5. Design a four-bit FA using combinational logic.
6. Design a four-bit FA using ROM technology.
7. Use bridging to implement a standard full subtracter circuit
(three inputs and two outputs) using X-OR gates.
8. Verify Equations [5.4] and [5.6].
9. Design a circuit for dividing a Jour-bit number by a four-bit
number.
10. The following message needs to be transmitted using the
Hamming code under even parity assumption. Determine the
192 CHAPTER FIVE Design of Combinational Circuits

parity bits and the order in which the coded message will be
sent. The to-be-coded message is 1010111001011. Show the
corresponding circuit.
11. Design a half subtracter circuit using (a) only NOR gates and
(b) only MUXs.
12. The Hamming-coded message received under odd parity
assumption is 1010 111001011. Determine if the message has
any error and write out the correct message bits only. Obtain
the corresponding correction circuit.
13. Design a full subtracter using half subtracter modules.
14. Using only a four-bit binary adder, design decimal code con­
verters for the following conversions:
a. 8-4-2-1 to XS3
b. XS3 to BCD
c. XS6 to XS3
d. BCD to XS3
15. Remove the combinational FAs from the drcuit of Example
5.2 and replace these with equivalent ROMs. Show the ROM
logic for one of these units and determine the total ROM size
needed for the complete circuit.
16. Design a combinational circuit capable of comparing two
eight-bit binary integers (without sign bits) X and Y. The out­
put Z should be a 1 whenever X > Y.
17. Design a controllable, dual-purpose, four-bit converter that
converts binary to Gray and also Gray to binary.
18. Use the module of Example 5.7 for obtaining the following
conversIOns:
a. 15-bit
b. 20-bit
c. 25-bit
Justify your designs using exemplary nontrivial binary inputs.
19. Use the module of Example 5.10 for obtaining the following
conversIOns:
a. 15-bit
b. 20-bit
c. 25-bit
Justify your designs using exemplary nontrivial BCD inputs.
20. Design an adderlsubtracterusing cascaded ALUs. Show how
it works when adding and when subtracting if A = 84 and B
= 32. Repeat the problem using base-16 equivalents of the
numbers.
21. Design a 12-bit FA in which carries are allowed to ripple after
the first six bits of addition.
 Problems 193

22. Show how the ALU can be used to (a) subtract one from and
(b) add one to a number. Show the setup if the number is 76 10,
23. Show how a 3-8 line decoder could be used to generatej(A,B,
C) = ~m(0,1,3,5).
24. Design a logic circuit that multiplies an input decimal digit
(in BCD) by five. The output is also in BCD form. Show that
the outputs can be obtained from the input lines without using
any logic gates.
25. Implement the FA circuit of Problem 1 using MUXs.
26. Obtain the most minimal circuit that squares a three-bit
binary number.
27. Design a special-purpose unit using FAs (a) for adding 12
single-bit binary numbers and (b) for adding 17 single-bit
binary numbers.
28. Design a four-bit CLA circuit where the propagate function is
defined as p.I = A·I + B·I instead of p.I = A·1 &\I7 B·I' How does the
current design differ from the one discussed in Section 5.4?
29. Use the techniques considered in Section 5.4 to obtain a four­
bit fast subtracter.
30. Obtain the CLA carry equations when n > 13 and show that
the maximum fan-out is dependent on variable P(n-2)/2 and is
equal to {[(n + 1)2/4] + 2} for odd n. Also show that for even
n, the maximum fan-out is dependent on both p(nl2)-t and
P{nI2) and is equal to {[n(n + 2)]/4 + 2}.
31. Design an n-bit binary comparator circuit to test if an n-bit
number A is equal to, larger than, or smaller than a second n­
bit humber B. The problem could be broken into one unit of a
half comparator module and n - 1 units of full comparator
modules, as shown in Figure 5.Pl. Each of the modules gives
out two outputs: Gn and Ln , such that
a. Gn = 1 and Ln = 0 if An > Bn
b. Gn = 0 and Ln = 1 if An < Bn
c. When An = Bm then Gn = Gn - t and Ln = L n- t for a full
comparator and Gn = Ln = 0 for a half comparator.

FIGURES.PI

A, B, Ao Bo

1 G3 G2
1 G,
1 ...
Go
Fe L3 Fe L2 Fe L, Fe Lo
He
194 CHAPTER FIVE Design of Combinational Circuits

Describe the working principles of the design and show how

you could improve upon this design.
32. Design an n-bit comparator module different from that of
Problem 31 such that the comparison process begins from the
MSB and moves toward the LSB until the final decision is
made.
33. Obtain the circuit for a 16-input, 4-output priority encoder.
34. Design a serial-to-parallel converter circuit that routes a long
sequence of binary digits into four different output lines as
specified by external control signals.
35. Design a BCD adder circuit that first converts the BCD num­
bers into binary, then adds the resulting binary numbers, and
fmally converts the binary sum to the correct BCD sum.

Agrawal, D. P. "Negabinary carry look ahead adder and fast multiplier."

Suggested Readings
Elect. Lett. voL 10 (1974): 312.
Arazi, B. "An electrooptical adder." Proc. IEEE. vol. 73 (1985): 162.
Bartee, T. C. Digital Computer Fundamentals. New York: McGraw-Hill,
1985.
Benedek, M. "Developing large binary to BCD conversion structures."
IEEE Trans. Camp. vol. C-26 (1977): 688.
Bin Nun, M. A., and Woodward, M. E. "Halfadders modulo-2 using read­
only memories." Elect. Lett. voL 10 (1974): 213.
Bywater, R E. H. Hardware/Software Design oj Digital Systems. Englewood
Cliffs, N.].: Prentice-Hall International, 1981.
Daws, D. C., and Jones, E. V. "Hardware-efficient bit sequential adders
and multipliers using mode-controlled logic." Elect. Lett. vol. 16 (1980):
434.
Feldstein, A., and Goodman, R "Loss of significance in floating point sub­
traction and addition." IEEE Trans. Compo vol. C-31 (1982):328.
Floyd, T. L. Digital Fundamentals. 2d ed. Columbus, Ohio: Charles E. Mer­
rill, 1982.
Gaitanis, N. "Single error correcting and multiple unidirectional error
detecting cyclic AN arithmetic codes." Elect. Lett. vol. 20 (1984): 638.
Greer, C. R, and Thompson, R. A. "Combinational logic design with
decoders." IEEE Trans. Camp. vol. C-27 (1978): 86~.
Hamming, R W. Coding and Information Theory. Englewood Cliffs, N.J.:
Prentice-Hall, 1980.
Lai, H. c., and Muroga, S. "Minimum parallel binary adders with
NOR(NAND) gates." IEEE Trans. Camp. vol. C-28 (1979): 648.
Langdon, G. G., Jr., and Tang, C. K. "Concurrent error detection for
group look-ahead binary adders." IBM J. Res. & Dev. vol. 14 (1970):
563.
Ling, H. "High-speed binary adder." IBM J. Res. & Dev. vol. 25 (1981):
156.
Liu, T. K.; Hohulin, K. R.; Shiau, L. E.; and Muroga, S. "Optimal one­
bit full adders with different types of gates." IEEE Trans. Camp. vol. C­
23 (1974): 69.
Suggested Readings 195

Majerski, S. "On determination of optimal distributions of carry skips in

adders." IEEE Trans. Elect. Compo voL EC-16 (1967): 45.
Peterson, W. W. "Error correcting codes." Sci. Am. voL 215 (1962): 96.
Schmookler, M. S. "Design oflarge ALUs using multiple PLA Macros."
IBM]. Res. & Dev. vol. 24 (1980): 2.
Wakerley, J. Error Detecting Codes, Self-Checking Circuits and Applications.
Amsterdam: North-Holland, 1978.
Sequential
Devices_______ -----J

6.1 Introduction
A circuit is known as combinational as long as its steady-state out­
puts depend only on its current inputs. If, on the other hand, the
present value of the outputs are dep<mdent on both the present val­
ues of the inputs and the past values of the inputs, the circuit is con­
sidered to be a sequential circuit. One of the important applications of
digital techniques is where digital signals are received and inter­
preted by the system, and control outputs are generated in accor­
dance with the sequence in which the input signals are received.
Therefore, such systems require circuits that respond to the past his­
tory of the inputs. In general, sequential circuits have the capability
of storing information. Consequently, sequential circuits find wide
application in digital systems as counters, registers, control logic,
memories, and other complex functions.
The most common sequential circuit is the flip-flop. A jlil-flop
(FF) is an electronic device that has two stable states. One state is
. assigned the logic 1 value arid the other the logic 0 value. The out­
put of the FF can assume either of the stable states based on input
events, and the output can be checked to determine what event
occurred in the pair. There are a number of FFs in common usage
in digital circuits, and they differ from one another in the number of
inputs they have and in the manner in which the binary state is
affected by the inputs. The possible changes in the FF outputs gen­
erally have a direct correspondence to the frequency with which the
input is changing value. However, there is a type of sequential cir­
cuit memory, device, known as a monostahle multivibrator, that pro­
duces circuit output independent of the input frequency. This chap­
ter iritroduces the logical behavior and control bf various types of
FFs. Mer studying this chapter, you should be able to:
o UnderStand the design and working principles oflatches;
o Understand the design and working principles ofFFs;
196
 6.2 ~tches 197

o Understand the design and working principles of the

monostable multivibrator;

o Understand the importance and significance of sequen­

tial circuits in generaL

6.2 Latches A latch is a bistable circuit that is the fundamental building block of
a flip-flop. The latch is basically a combinational circuit that has
one of its outputs fed back as an input. It can be realized from an
OR gate with its complemented output fed back as one of its inputs,
as shown in Figure 6.1. We have considered /);.t to be the lumped gate
delay (total of all propagation delays) of the gates that are used. If
the input II is held at logic 1, the OR output results in a 1 and,
therefore, the complemented output, O2, is a O. This O2 output is fed
back to the OR gate after a time equal to the lumped delay. As a
result the output of the OR gate is held at logic 1. Once the OR
output is set to this· condition, the gate output will remain in this
same state. This phenomenon is commonly known as the latching
effict. Consequently, this circuit could be used for the storage oflogic
1. This latching condition will prevail until the feedback path is
broken.

FIGURE 6.1 Latch for Storing

Logic 1.

The NOT gate of Figure 6.1 may now be replaced by an AND

gate to provide for the storage of logic O. Such provision is known as
the unlatching of the gate, which is illustrated in Figure 6.2. This cir­
cuit, however, is able to store both logic 1 and O. Each of the two
latch outputs, 0 1 and O2, is a logical complement of the other. If one
holds 12 input to logic 0, the feedback route between the input and
the output of the OR gate will be logically broken and the OR gate

FIGURE 6.2 AND-OR Latch.

)--4_-0,

198 CHAPTER SIX Sequential Devices

will return to its initial condition. When the input 12 is a 0, the out­
put O2 is maintained at logic 0, and when I, is held at logic 1, the
output 0 1 remains at logic 1.
The latching concept developed in this section will be used next
to come up with a standardized latch suitable for subsequent devel­
opment of FFs. An FF has the capability of storing a single binary
bit of information. When the values stored in the FFs change, we
say that the sequential circuit changes state. Generally, however, an
FF should have two outputs, called Qand Q, that are complements
of each other. The characteristic table of Figure 6.3 details the per­
tinent working principles of one such basic latch unit, where t is
used to denote the time variable and 6.t is the short time duration
between a change in the input and a possible change in the output.
The interval 6.t is equivalent to the lumped delay of the circuit. The
two inputs S (set) and R (reset) are used to control the output based
on the current state of the output. If R = 0 and S = 1, the output is
turned on if not already on. If R = 1 and S = 0, the output is
turned off if not already off. When S = R~= 0, no output change
occurs. The to-be-designed latch circuit, however, manifests an
undesirable condition when both inputs go to 1 simultaneously.
When S = R = 1, the two outputs, Qand Q, would no longer be
complements of each other. In addition, the behavior of the latch
would become unpredictable once the inputs returned to 0, Conse­
quently, the simultaneous existence of S = R = 1 is forbidden.

FIGURE 6.3 Characteristic Inputs

Table for a Basic Latch. . Output
R(t) S(t) Q(t) Q(t + Clt)

0 0
0 0
1 1
0 1
0 1
1 1
0 0
1 0
1 0
0 -
1 1
1

The circuit may be determined accordingly, using the K-map of

Figure 6.4[a]. The equation for Q(t + 6.t) is obtained as follows:
Qit + 6.t) = Set) + R(!jQ(t)
= S(t) . R(t)Q(t) [6.1] .
 6.2 Latches 199

FIGURE 6.4 RS Latch: [a] K­

map and [b] Circuit.

R(t)S(t)
o(t) 00 01 11 10

0 0 1 - 0
R---iL-----'"
1 1 1 - 0 o---e--Q
5---------------1
O(t + Llt)
[a] [b]

This equation, known also as the next-state equation, states that after a
short time, b.t, the new value of Qis determined by the values of Q,
R, and S at time t. The corresponding circuit is shown in Figure
6.4[b]. If Rand S values change at time t, a new value of Qwill
result b.t time later. b.t time in this circuit is the total gate delay of
the two NAND gates. The output of NAND gate 1 is traditionally
known as the Qoutput since the outputs of two NAND gates are
complements of each other.
FIGURE 6.5 Revised RS Latch The revised characteristic table of Figure 6.5 shows the corre­
Characteristic Table. sponding characteristics of the RS latch (also called reset-set latch)
where the don't-care of the forbidden state is assumed to be equal to
OCt) R(t) O(t + Llt) a O. The equation for Q(t + b.t) may also be obtained as follows:
0 0 S(t) Q(t + b.t) = Q(t)R(t)S(t) + Q(t)R(t)

1
0 0
1
= R(t)[Q(t) + Q(t)S(t)]

1 0
1 1 0 = R(t)[Q(t) + S(t)]

~ "
= R(t) + [Q(t) + S(t)] [6.2]

This NOR form of latch could also be derived by grouping the

zeros of the K-map of Figure 6.4[a]. The two circuits, NAND and
NOR latches, are respectively known as RS and SR flip-flops or
more commonly as latches. The corresponding latch circuits are
FIGURE 6.6 SR Latch: [a] Block obtained as shown in Figure 6.6.
Diagram, [b] NAND Circuit, and
[c] NOR Circuit.

S o 0-'-0
S---.o.
0--....... 0

R R R---J
"----­
[al [b] (c]
200 CHAPTER SIX Sequential Devices

Note that the latch outputs are always complements of each

other: When Qis aI, Qis 0, and when Qis a 0, Qis 1. The forbid­
den state occurs when S = R = 1 at the same time. As long as S
and R are both set at 1, Qand Qare forced to be at the same logic
value simultaneously, thus violating the basic complementary
nature of the outputs.

EXAMPLE 6.1 SOLUTION

Obtain the response of a periodic FIGURE 6.7
square wave of period 8 units, when
it is fed into the sequential circuit of
Figure 6.7. Assume that the NAND
gate has a total delay of 1 unit
X

The output, Z, is given by

Z(t + at) = X(t)Z(t) = X(t) + Z(t)

Consequently, the timing diagram is obtained as shown in Figure 6.8.

Whenever X changes from a 0 to aI, the output Z starts oscillating with a
period of 2M (2 units in this case). However, when X changes from a 1 to a
0, the output Z becomes 1 after a time delay of at.

FIGURE 6.8
X
~I 4
I
8 12
L
16

Z
0 5 6 7 8 13141516

1
Y
0
4 5 6 7 12131415

Time---+
ililmISt

The SR latch described earlier has its time delay lumped

together. However, there is another form of the latch model known
commonly as the distributed gate delay model. Consider, for example,
the NAND latch of Figure 6.6[b] where the gates numbered 1 and 2
are assumed to have gate delays tl and t2, respectively. Accordingly,
Q(t + t1) = Set) • Q(t) = Set) + Q(t) [6.3]
Q(t + t2) = R(t) • Q(t) = R(t) + Q(t) [6.4]
 6.2 Latches 201

These two equations now may be combined to yield

Q(t + tl + t2) = R(t + [I) • Q(t + t l )
= R(t + t1) [S(t) + Q(t)] [6.5]

Therefore,
Q(t + t\ + t2) = R(t + tl)[S(t) + Q(t)] .[6.6]

Assuming equal gate delays for a total of dt delay, the resulting

equation becomes

-( t + :2
Q(t + dt) = R dt) [S(t) + Q(t)] [6.7]

The timing diagrams of Figure 6.9 show the behavior pattern of

a NAND latch corresponding to the distributed gate delay modeL
For simplicity it has been assumed that both of the gates have 1 unit
length of gate delay and also that Qand Qat time t = 0 are, respec­
tively, 0 and 1. This timing diagram shows that the latch works as
intended. However, under various input conditions the latch may
have problems. Example 6.2 will illustrate one of these input condi­
tions and its consequences.

FIGURE 6.9 Timing Diagram of 1

a NAND Latch. S
0
6 9
R
0
14 18

Q
0
7 16

Q
0
8 15

Time ..

SOLUTION
Obtain the timing diagram for the FIGURE 6.10
latch of Figure 6.10 when the input
II changes from 1 to 0 for a
duration much shorter than the
total gate delay.

For this example the gate delays for the NAND and NOT gates are chosen
202 CHAPTER SIX Sequential Devices

to be 3 units and 2 units, respectively. Accordingly, the timing diagrams of

Figure 6.11 are obtained. The output is oscillatory in nature and is not
latched to any fIxed value.

FIGURE 6.11
11
0

O2
;~ n
8 9
n
13 14
n

18 19

01
;1 U
10 11
LJr---u
15 16 20 21

Time ----+

6.3 Clocked SR In the last section we introduced circuits for the SR FF. For depend­
able operation of such devices one must attempt to prevent tran­
Flip-Flop sient pulses from appearing on either of the inputs. It is advanta­
geous to control the times when the SR FF output is allowed to
change by means of an additional input. This additional signal is
commonly called a clock. The clock pulses (CK) can be periodic or a
set of random pulses. Almost always, however, they are periodic.
The purpose of the clock input is to force the FF to remain in its
rest (or hold) state while changes occur on the set and reset inputs.
CK is set to logic 1 once the inputs have settled. The NAND and
NOR latches with clock input are shown in Figure 6.12. In order to
operate these devices effectively, the following conditions must be
met:
1. The FF inputs should be allowed to change only when CK
= 0.
2. The clock input should be long enough so that the outputs
will be able to reach steady states.
3. The condition S = R = 1 must not be allowed to occur
when CK is equal to logic 1. For proper operation, there­
fore, S(t)R(t) should always equal zero.

It can be seen that the circuit action can occur only when the CK
signal is high. When CK = 0, the FF outputs do not change. The S
and R inputs may, however, be simultaneously high when the clock
is absent since the FF will be inhibited. The overall functioning of
the gated SR FF is illustrated by the characteristic table of Figure
6.13. Note in the timing diagram shown in Figure 6.14 that it is
 6.3 Clocked SR Flip-Flop 203

FIGURE 6.12 Clocked SR FF: S

[al Logic Symbol, [b] NAND Gate S Q
Q
Circuit, and [el NOR Gate Circuit.

CK CK

Q
R Q
R
[a] [b]

R
Q

CK

S
[c]

FIGURE 6.13 Characteristic Inputs Outputs

Table for the Clocked SR FF.
CK(t) S(t) R(t) Mode Q(t + At) Q(t + lit)
0 - - No action Q(t) Q(t)
Jl.. 0 0 Hold Q(t) O(t)
Jl.. 0 1 Reset 0 1
Jl.. 1 0 Set 1 0
Jl.. 1 1 Invalid - -

FIGURE 6.14 Timing Diagram

of a Clocked,SRFF. CK
o
S
o
R
0-----'

Q
o
necessary to consider the circuit only at the time CK changes from
low to high to see if the output changes.
It is now appropriate to introduce several operational character­
istics that are, commonly associated with the FF usages. Figure 6.15
shows some of these specifications, of which setup and hold time are
the most important ones. The setup time, ts , is the time· necessary for
204 CHAPTER SIX Sequential Devices

FIGURE 6.15 Timing

Characteristics under Worst-Case
Condition for Maximum Clock
Frequency Determination.

1 t
,... s -H
I+- th-+l
---'"-----~---___
1
14 tNS
, I , 1
I 1
I
--,-~
,
I I
1"'4--~1------
I I
tp .11
I

I 14 I TcK I

~
I

Inputs must be stable Inputs must be stable

the input data to stabilize before the triggering edge of the clock Its
value is extremely critical since it manifests itself either by ignoring
actions or by resulting in partial transient outputs, commonly
referred to as partial set and partial reset outputs. Consequently, it is
. possible to begin a set or reset mode, causing the output to start to
change, but to withdraw back to its initial state. In some cases the
output might even end up in a metastable state in which the FF is
neither set nor reset. Again, the hold time, th, is the time necessary for
the data to remain stabilized beyond the triggering edge of the
clock This is also a critical parameter in determining the correct
behavior of a FF.
The maximum allowable clock frequency for an FF is usually
determined from a knowledge'of setup time; hold time; FF propa­
gation delay, tp; and propagation delay of the next-state decoder,
tNS' The maximum clock frequencY,!cK, under worst-case condition
is obtained from

[6.8]

The constraint of Equation [6.8] must be met when using any FF,
integrated or not.

EXAMPLE 6.3 SOLUTION

Obtain Q(t + Ilt) as a function of FIGURE 6.16
the inputs and Q(t) in the circuit of
Figure 6.16. x---\
S Q.

CK

Y --..._---------i R Q
 6.4 JK Flip-Flop 205

FIGURE 6.17
From Equation [6.1] the next-state equation follows as

Q(t + at) = Set) + R(t)Q(t)

X(t) Y(t) Ott) O(t + M)
0 0 0 1 In this circuit Set) = X(t) + yet) = X(t) • Yet) and R(t) = Yet). Note also
0 0 1 1 that S(t)R(t) = X(t) • Yet) • yet) = O. Hence, Set) and R(t) are not simulta­
0 1 0 0 neously equal to 1 and, therefore, the circuit meets all conditions necessary
0 1 1 0 for a perfect operation. Consequently, we may obtain
1 0 0 0 Qit + at) = X(t)Y(t) + Y(t)Qit)
1 0 1 1
1 1 0 0 The table, as shown in Figure 6.17, lists all possible combinations of the
1 1 1 0 inputs and the corresponding outputs.

6.4· J~~Flip- Flop We saw in the last section that the clocked SR FF has an indetermi­
nate state. When using clocked SR FFs the designer is required to be
/ cautious about the FF inputs. This troublesome restriction can be
removed by modifying the SR FF; the refmed FF is known as theJK
FF. This modification involves feeding the outputs of the FF back
into the inputs of the circuit shown in Figure 6.12[a]. The resulting
circuit, its block diagram, and its functional behavior are shown in
Figures 6.l8[a-c].

FIGURE 6.18 JK FF: [a] Block

Diagram, [b] Logic Circuit, and [c] J-----l
Characteristic Table.
J 0
CK
CK
K----I 2
K 0

[a] [b]

Mode J(t) K(t) Ott) O(t + At) 1

0 0 I
Hold 0 0
1 1 I

0 0 I
Reset 0 1
1 0
0 1
Set 1 0
1 1
I
i 0 1
Toggle 1 1
1 0

Ie]
206 CHAPTER SIX Sequential Devices

Even when the] and K inputs are both 1, the outputs of NAND
gates 1 and 2 cannot simultaneously be O. With Q = 0, NAND gate
2 outputs ai, and when Q = 1, NAND gate 1 outputs a 1. Conse­
quently, the input restriction of the SR FF is automatically elimi­
nated. The additional feedback provides for an additional switching
mode, called toggle, to the FF. The characteristic table of Figure
6. 18[c] describes in detail the actions of the FF. The next-state equa­
tion may accordingly be obtained as follows:
Q(t + /).t) = ](t)OJ.t) + K(t)Q(t) [6.9]

If] = 1 and K = 0, the FF is set to an output of 1 ifnot already set.

Similarly when] = 0 and K = 1, the FF resets to 0 if not already
reset. If] = K = 1, the FF output is complemented (toggled), and
when} = K = 0, no change takes place.
In spite of many advantages the]K FF still has a serious limita­
tion, which is illustrated in Figure 6.19 and should be understood
by the designer. When the clock input goes high, the FF responds
according to the] and K inputs. The flowchart of Figure 6.19[a]
shows the desired circuit operation and the consequence of having a
clock pulse that is too long. If the Q output changes before the ter­
mination of the clock input, then the input conditions to NAND
gates 1 and 2 change again, and this leads to subsequent change in
the Q output. As a consequence, Q may be indeterminate at the ter­
mination of the clock input. Such a possibility exists as long as /).t is
less than TCK' It is desired that only one FF change occur during
each clock input. This may be easily accomplished by maintaining
the clock width much smaller than the total delay time.

FIGURE 6.19 [a]JK FF

Flowchart and [b] Pulse Width
Problem inJK FF.

Compu1eJK
inputs

1
J
No 0

K
0 I-t--TcK---+1
I
CK 1
0 I
1
Change FF Q
0 nJI
outputs
_____I At I+-­
[a] [b] /
 6.4 JK Flip-Flop 207

Another way to eliminate the problem caused by a clock pulse

width that is too long is to design the FFs to respond to only transi­
tions of the clock, either 1 -? 0 or 0 -? 1. Edge-triggered FFs are
provided with a pulse-narrowing circuit, as shown in Figure 6.20.
As you can see, there is a small delay on one of the NAND gate
inputs so that the inverted clock pulse arrives at the gate input a
couple of nanoseconds later than the true clock pulse. This results in
an output spike of an extremely small time duration at the very
beginning of the clock pulse. This narrow pulse is then used for the
clock input of theJK FF, eliminating the necessity for narrow clock
pulses. It is appropriate to consider the effect this narrow pulse
might have on the FF actions. Depending on the parameters of the
gates that are being used in the circuit of Figure 6.20, the resulting
pulse could be too narrow to trigger an FF. In such an event more
than one, but only an odd number of, NOT gates could be used in
place of the first NOT gate. Again the number of NOT gates should
not be too large, because a clock input that is too wide might result.

FIGURE 6.20 Pulse-Narrowing

Circuit. CK
To the CK
input Of the FF

Edge-triggered devices are of two types. Positive edge-triggered

devices respond when the clock input makes the transition 0 -? 1,
and the negative edge-triggered devices respond when the clock
input makes the transition 1 -? O. Figure 6.21 shows the logic sym­
bols for both positive edge-triggered and negative edge-triggered SR
and JK FFs where the arrowhead input corresponds to the CK
input. Another way to achieve edge triggering is to use a special FF
known as the master-slave FF that appears to trigger only on the
clock edge.

FIGURE 6.21 Logic Symbols for SR JK SR JK

Edge-Triggered FFs. S Q
J Q S Q J Q

R Q K Q R Q K

Positive edge-triggered Negative edge-triggered

208 CHAPTER SIX Sequential Devices

EXAMPLE 6.4 SOLUTION

Obtain the timing diagram for the FIGURE 6.22

sequential circuit shown in Figure

6.22 for at least six clock cycles.
Assume that QI(0)Q2(0) = 00. I
~ J, 0, ,-­ r-- J2 O2 ,---J

-c~
!

'--- ­ K, 0, K2 O2

CK

Figure 6.23 shows the timing diagram that is obtained readily by making
use of the function table of aJK FF. It may be seen that within two cycles
QI is set and Q2 is reset. The waveform will not change until Ql is reset
externally.

FIGURE 6.23
CK

O,------.....J

EXAMPLE 6.5 SOLUTION

Obtain the response of the circuit of FIGURE 6.24
Figure 6.24, where each of the gates
is assumed to have 1 unit of gate
delay. The input x remains high for
a duration longer than 4 units.

The input, x, is assumed to be 5 units wide. The timing diagram is then

obtained as shown in Figure 6.25. This circuit locates the trailing edge of
the input pulse. Note that the circuit of Figure 6.20 functions likewise but
locates the leading edge of an input.
 6.5 Master-Slave Flip-Flop 209

FIGURE 6.25
x
:1 3 8

:1 4 9

<I f - - - _ _ _- - - l n'----__
9 10
Time ,.

6.5 Master-Slave The master-slave concept is introduced into the FF circuitry to

eliminate the requirement for limiting the clock width of a value
Flip..Flop determined by the circuit gate delays. This type of FF is composed
of two sections: the master section and the slave section. This device
is dependent not on the synchronous clocking of both units, but
rather on their alternate turn-on and turn-off characteristics. The
logic circuit of an SR master-slave FF is shown in Figure 6.26. It
consists of a master FF, a slave FF, and an inverter for achieving
out-of-phase clocking of the two units.

FIGURE 6.26 Master-Slave SR Master Slave

FF.
S SM OM Ss Os o

CK i'>

R RM OM Rs Os

~
V
As a result of the presence of the inverter, the master unit is
turned on. and the slave unit is turned off when CK = 1. When CK
= 0, the master unit is turned off and the slave unit is turned on.
The circuit works as follows: for all inputs of Sand R, except when S
= R, QM = Sand QM = R when CK = 1. At this time the slave
unit remains turned off. When the clock input goes to 0, Qs = QM,
Qs = QM, and the master uI1it is turned off.
210 CHAPTER SIX Sequential Devices

The timing diagram shown in Figure 6.27 illustrates the

sequence of operations that takes place in a master-slave SR FF.
The overall master-slave outputs appear to change at the negative
edge of the clock input. However; there are many Ie FFs that are
the positive edge-triggered type. The master-slave cascading may
be accomplished for any FF by similar introduction of an inverter
between the two sections. As another example, Figure 6.28 shows
the logic diagram of a master-slaveJK FF. This is .slightly different
from that of a master-slave SR FF in that the outputs of NAND
gates 7 and 8 are introduced as inputs to NAND gates 2 and 1,
respectively. We have learned from Figure 6.19 that the Qand Q
outputs might change several times during a wide clock pulse, lead­
ing to an unpredictable FF condition.. Similar clock inputs would
still cause the master outputs, QM and QM, to change; but the slave
outputs, Qs and Qs, would not change becauSe the inverted clock
pulse disables the slave section. The values for J and K are still
determined by the preclock values of Qs and Qs. When the clock
input to the master section goes low, the clock input to the slave sec­
tion goes high, transferring QM and QM to the slave section. We
have thus eliminated the problem of clock pulses that are too wide
by the master-slave principle. A representative timing diagram for
the master-slaveJK FF is shown in Figure 6.29.

FIGURE 6.27 Timing Diagram CK 1

of a Master-Slave SR FF. 0

1
S
0
I
I
I
I
I
I
l
t I I I
1 I I L L
I
R
0 ~ I
I
I I
I
I
I I
1 I
I
OM
0 I I I

1
Os
0 I l
1
0
0 I I

A master-slaveJK FF is also not without problems. The master

section of the FF is vulnerable during the period when the clock is
high and, therefore, may be set or reset by appropriate changes of
the input. This results in "Is and Os catching" problems. When Qs
..:... 0, Qs = 1, and the clock is high and while still high, the J input
becomes high, QM is set, and consequently Qs "catches" a 1 on the
trailing edge of the clock input. Again when Qs = 1, Qs = 0, and
 6.5 Master-Slave Flip-Flop 211

FIGURE 6.28 Master-Slave JK

FF.

CK--~------------~

7 D---e--+--OS

8 o--+--+--Os
K-,.,-------1 6

FIGURE 629 Master-SlaveJK

FF Timing Diagram. CK~
J ----'

K ----------'

Os---------.....J

K becomes a 1 after the clock has already become high, QM is reset,

and Qs "catches" a 0 on the trailing edge of the clock input. It is
important, therefore, to make sure that no such input changes can
gain entry into the FF.
Often an edge-triggered JK FF is also provided with two addi­
tional control inputs: preset and clear. The preset (PR) and clear
(eLR) inputs allow initializing the FF to either a set (Q = 1) or a

~It.:~'.:
reset (Q = 0) condition. Addition of these two control inputs
requires alteration of only the slave section of the FF. Figure 6.30
~.
..
~?:
shows the logic diagram and the corresponding slave section of the
. ~:.:
.'<~: FF that allows preset and clear inputs. Throughout this text both
preset and clear inputs are considered to be active when low. Often

!I~·t.·
.~..
.~( . ..
.'.
these two FF control inputs are not labeled in the FF logic diagram .
In such cases preset and clear inputs are always indicated by verti­

,;;.

~r·
212 CHAPTER SIX Sequential Devices

FIGURE 6.30 Complete Master­

SlaveJK FF: [a] Logic Diagram
and [b1 Slave Circuit.

D----+--Os
PR
J 0

K 0
CLR

CK

[a] [b]

cal inputs (with a bubble) respectively at the top and bottom of the
corresponding FF logic diagram.

6.6 Delay and Trigger There are two other types of flip-flops that are commonly used: the
delay (D) and the trigger (T) FFs. Unlike those in the previous sec­
Flip-Flops tions, these two FFs have only one control input line besides the
clock (excluding set and preset). Both of these FFs can be realized
by externally manipulating the inputs of aJK FF.
Often it is necessary to have a sequential device that simply
retains the input data value between clock pulses. The D FF per­
forms this function. The FF output follows the FF input whenever a
clock pulse is 1 and holds the value the input had when the clock
changed to O. The logic diagram and the characteristic table for a
D FF are shown in Figures 6.31[a-b]. A comparison of this charac­
teristic table with that for theJK FF (Figure 6.18[c]) reveals that a
J
D FF is realizable from a JK FF by making K = and using J as
FIGURE 6.31 D FF: [a] Logic the D input, as illustrated in Figure 6.31 [c]. The next-state equation
Diagram, [b] Characteristic Truth of the D FF is given by
Table, and [c] Circuit
Implementation. Q(t + dt) = D(t) [6.10]

D 0
D(t) O(t) O(t + At) D- - - - I t - - - - - - I J 0
0 0
0
1 0
0 1 '------I K
1
1 1

[a] [b] [c]

 6.6 Delay and Trigger Flip-Flops 213

If severe restrictions placed on the clock input of an SR FF pose

no problem, the SR FF can also be used to produce a D FF. The
resulting circuit is shown in Figure 6.32[a]. Figure 6.32[b] shows a
slight variation of the circuit of Figure 6.32( a] where advantage is
taken of the special properties of NAND gates to eliminate one gate
and still retain the characteristics of a D FF.

o--~I--Q
CK--~---------~
'---_..-'
[b]

The T (trigger) FF, often called a toggle FF, has a single input
that causes the output to change each time a pulse occurs at the
input. The output remains unchanged as long as T = O. The logic
diagram and the characteristic table for a T FF are shown in
Figures 6.33[a-b]. It should be noted that theJK FF has this mode
available. The JK FF can be reorganized for realizing a T FF, as
shown in Figure 6.33[c]. As long as both the T input -and the CK
input are high, the FF output will change. Its next -state equation,
FIGURE 6.33 TFF: [a] Logic therefore, is obtained as follows:
Diagram, [b] Characteristic Table,
and [c]Circuitlmplementation. Q( t + 6.t) = Q( t) ffi T( t) [6.11]

T Q T(t) Q(t) Q(t + M) T---e----j J Q

0 0
0
1 1
5 1
0 1 '-'----i K
1 0

A different version of the T FF involves a one-input device. Both

J and K inputs of the JK FF are tied to a 1 to realize this unclocked
214 CHAPTER SIX Sequential Devices

T FF. The input data are then introduced at the original clock
input The corresponding circuit for the unclocked T FF is shown in
Figure 6.34.

FIGURE 6.34 Unclocked T FF:

[a] Block Diagram and [b] Logic o
Circuit.
o J

o K o

The unclocked T FFs are very important but are not made com­
mercially. The logic usually is obtained using aJK FF as shown in
Figure 6.34. One may even obtain this function from aD FF. In fact
it is also easy to transform an unclocked T FF back to a JK FF.
Such a conversion circuit is shown in Figure 6.35.

FIGURE 6.35 Conversion of an

Unclocked T FF to a Regular JK
FF. J-----t

CK O---CI>T

K-----i

EXAMPLE 6.6 SOLUTION

Analyze the circuit of Figure 6.36. FIGURE 6.36

n----ID 01----..-­
y-----l

CK Ol---e-­
 6.7 Monostable Flip-Flop 215

The next-state equation of this circuit is as follows:

Q( t + !J.t) = D(t)

= Y(t)Q(t) + X(t)Q(t)

A close examination of this equation reveals that it is very similar to Equa­

tion [6.9] in that Y acts like the J input and X acts like the K input. In
other words, this circuit functions exactly like aJK FF.
fn

6.7 Monostable The monostable FF, also known as a one-shot, is an edge-triggered

device used for producing output pulses of a duration independent
Flip-~Iop of the input frequency. It produces an output pulse of specified
"

width that is initiated by an input trigger signal. After a specified

period of time the output returns to its quiescent state. The pulse
duration is determined by the parameters of the resistor-capacitor
network located external to the one-shot.
One-shots are of two types: nonretriggerable and retriggerable.
In the nonretriggerable one-shot, if the device receives two succes­
sive trigger pulses of separation t less than the width Ot of the output
pulse generated by a single trigger, the second trigger input is
ignored by the device. The retriggerable one-shot would be acti­
vated by the second trigger pulse, resulting in an output pulse of
width approximately t + Ot. By applying a succession of trigger
pulses separated by t < Ot, the output of a retriggerable one-shot
could be maintained high (logic 1) as long as desired.
A one-shot may be designed using basic logic gates and a resis­
tor-capacitor network. However, it is more convenient to use an Ie
one-shot because they are widely available and relatively inexpen­
FIGURE 6.37 Retriggerable sive. Figure 6.37 shows the logic diagram and the function table of a
One-Shot FF: [a] Logic Diagram standard retriggerable one-shot. The four inputs, Ah A2 , Bl , and B2 ,
and [b] Trigger Conditions. are available to provide flexibility of operation. The capacitor, C,

c R
rl( v Vee =5V
Al A2 81 82 T
)­ 1->0 1 1 1 0->1
Q 1 1->0 1 1 0->1
-
'­

)r > -
0 -
0

0
0->1
0->1
1
1
1
0->1
0->1
0->1 0->1
Q
0 1 o -> 1 0->1
[a] [b]
/
216 CHAPTER SIX Sequential Devices

and the resistor, R, are external to the Ie one-shot and are used to
control the duration of the output pulse. Note that the trigger pulse,
T, is given by Al • A2 • BI • B2. The triggering conditions, as shown
in Figure 6.37[ b], cause T to change from a 0 to a 1.
The duration of the output pulse, lJt, is determined by the resis­
tor-capacitor network. Adjustable resistors and I or capacitors may
be used to trim the output pulse to the desired width. In general, 5t
is given by
& = f(R,C) [6.12]

wheref(R,C) is a function of the resistor and capacitor. The manu­

facturer provides the exact numerical relationship or curves, giving
the output pulse width as a function of the timing resistors and
capacitors. The minimum output pulse usually is realized using no
external capacitor. Note, however, that there will be some stray
capacitance existing between the terminals even in the absence of
the external capacitor.
The retriggerable one-shot may be transformed into a nonretrig­
gerable one-shot by feeding the Qoutput as one of the NAND gate
inputs, say, A2, while the other input A1 is treated as the only trig­
gering input. The remaining two inputs, BI and B2, should be tied
to a 1. if
It is advisable to use monostable FFs only when no other solution
can be found. Circuits with a number of monostables are very diffi­
cult to troubleshoot. Monostables can be falsely triggered by noise
in the power supply voltage, causing serious circuit malfunctions.

6.8 Sequential Circuits The general form ,of a sequential circuit .is shown .in}i'igure 6.38.
The circuit in consideration has p inputs, q outputs, and r FFs used
as memory. The combinational part of the circuit monitors the

FIGURE 6.38 Block Diagram of

a General Sequential Circuit. Xl
Inputs Combinational Zl } Outputs
} X --'-'- - . J Zq
p circuits 1-----'-'--+

r+~ __ ___ ~ ~-

FFcontrol
CK input - 1 - - - + - - - - - e
1,--1---- equations
(
/
01
FF

outputs

{: Memory
circuits

Or
\.. /
 6.8 Sequential Circuits 217

input values, A;-, checks the FF states, Q.b and computes the FF con­
trol variables to assure that the next initiating action causes the cor­
rect changes to be made in the FF values. In addition the combina­
tional part also computes the correct outputs, Z" for the circuit.
Thus the current inputs and the previous-state information stored in
the circuit's memory (FFs) are used to generate the current outputs
and to determine the next state in the sequential circuits. The clock
input is used only in clocked sequential circuits (the predominant
type of sequential circuit).
The memory part of the circuit may be provided by using bist­
able devices such as FFs, relays, magnetic devices, switches, and so
on. The most commonly used bistable device, however, is the FF.
The control characteristics of various FFs are summarized in Figure
6.39, which provides the FF excitation inputs necessary to cause
change in the FF output, (2: For example, the output of aJK FF can
be changed from 1 to 0 by setting K = 1 while the J input could be
tied either to a 1 or to a O. The corresponding state transitions
between Q = 0 and Q = 1 for each of the four FFs are shown in
Figure 6.40 where the conditions for transitions are indicated next

FIGURE6:J9 FF Control
Characteristics. Q(t) Q(t + at) J K D T
0 0 0 0 0 0
0 1 1 0 1
0 0 1 1 0 1
0 0 0

FIGURE 6.40 Transition

Diagrams: [a] SR FF, [b]JK FF,
[c] DEF¥and,[dJ TFF.'

SR 10 J= 1 K=O

R=O J=O­

SR 01 K =1
[a] [b]

T=1

D 0 T=1
[c] [d]
 218 CHAYI'ER SIX Sequential Devices

to the transition lines. These characteristics will play an extremely

important role in the design of complex sequential circuits.
One of the advantages of sequential systems over purely combi­
national systems is that circuit savings may be possible through the
repetitive use of the same logic circuit. Some examples are multi-bit
adder/subtracter circuits, multi-bit comparator circuits, and multi­
bit code-converter circuits. It was shown in Chapter 5 that n differ­
ent combinational circuit units are needed to accomplish an n-bit
parallel operation. However, the price we will pay for using the
same logic circuit repetitively is in the circuit operation speed.
Consider the four-bit ripple adder circuit shown in Figure 6.41.
This addition operation is called parallel since two four-bit numbers
are fed as inputs simultaneously to the adder circuit, and after the
gate delays, the resultant bits become available simultaneously.
However, addition can also be implemented serially. A sequential cir­
cuit that can perform such an operation is illustrated in Figure 6.42.
The process can be started by introducing Ao and Bo to the FA. The
resultant sum bit is stored in a multi-bit storage device, made up of
several FFs and called a register, and the carry-out is stored in an FF
and fed back into the FA. The process is repeated until all of the
bits have been considered. The final sum would consist of the last
carry-out and the sum bits stored in the register (see Problem 1 at
the end of the chapter). The characteristics of the registers and how
the input bits are fed sequentially to the adder will be considered in
a later chapter. For the time being it will suffice to say that the reg­
ister is a storage area where the bits can be moved slowly to the
right as the storage process continues.
However, in spite of all their merits sequential systems are not
devoid of deficiencies. Some of these disadvantages are very critical
to the functioning of the system; therefore, they must be considered
very carefully. Consider a very simple circuit such as the one shown
in Figure 6.43. When the input I is a 0, the output 0 becomes aJ.
This output is then fed back again to the NAND gate. As a conse­
FIGURE 6.41 Four-Bit Parallel
Adder Circuit. quence, the values of the output disagree with that of the input,

A2 82 Ao 80

Xi Y,

1
Xi Yi I
Xi i Xi Yi
Y I
FA Ci FA Ci ' FA Ci

Co Si Co Si Co S;
1
-

Co
 6.8 Sequential Circuits 219

FIGURE 6.42 Primitive Block

Diagram for the Parallel Adder.

Addend Augend
FA Ci ....

Co Sum

Cj S;

I Delay I --1 Sum storage

FIGURE 6.43 Indeterminate resulting in an indetenninate feedback system. Such problems,

Feedback Circuit. however, may be avoided by eliminating the conditions of continu­
ing oscillations.
Next consider the combinational circuit of Figure 6.44[a] where
the gates numbered 1, 2, and 3, respectively, have gate delays of
~tb ~t2' and flts. Assume that at time to the inputs are A = B = C
= 1, and at time t > to the input A changes from 1 to O. Assume
further that fltl < ~t2' The resulting timing diagram is illustrated
in Figure 6.44[ b]. It is apparent that the combinational circuit out­
put results in a transient error pulse for a duration of M2 - ~tl'
This error pulse is small but not negligible. If the output of this cir­
cuit is introduced as the clock, preset, or clear inputs to an FF, we
can expect to see additional problems in the sequential circuit.
However, errors will not occur until the pulse width exceeds the
time required to trigger the FF. These and other problems are tack­
FIGURE6.44: .[a] Combinational

Circuit -and·[b] Its Corresponding

led by taking proper precautions either during the design or during

Timing Diagram When B = C the operation of a sequential system.
=1.

to
to + Ml
1 I I
AI
o I
I I
I Ito + M2
'1 01 I I
A
I I I
B
1
I I
I
A '3 '2 0 I I, to + Ml + M3
I I
C '3 01 I
I
I
n I
{s] {b] I to + M2 + M3
220 CHAYrER SIX Sequential Devices

There are three different types of sequential circuits. They are

classified according to the characteristics of the inputs and memory
types.
Synchronous sequential circuits: Synchronous sequential circuits
involve FF action that occurs in synchronization with the clock
input. The values of the external variables that control the FF
states may change while the clock is not present. All transients
due to the previous clock must have disappeared prior to the
next clock for correct circuit action.
Pulse-mode circuits: The input variables of pulse-mode circuits
can have only mutually exclusive pulses. In addition, the FFs
that are used are not clocked since no clock is present in this
mode.
Fundamental-mode circuits: Fundamental-mode circuits involve
level inputs and asynchronous memory devices. The FFs change
state whenever an input variable logic level changes.
While the first type listed, also known as a clocked sequential cir­
cuit, is synchronous, the other two are asynchronous in character.
However, synchronous sequential circuits account for the over­
whelming majority of sequential circuits.

6.9 Summary In this chapter the concept of a sequential circuit was introduced.
The design and working principles of latches, various FFs, and the
monostable multivibrator were discussed. Particular emphasis was
placed on the various practical limitations that these devices have.
Finally, the possibility of having different classes of sequential cir­
cuits was explored. The design and the characteristics of these
sequential systems will be presented respectively. in the next three
chapters.

Problems 1. The FA receives two external inputs X and Y; the third input
Z comes from the output of a D FF as shown in Figure 6.P 1.
The carry-out is transferred to the FF at every clock pulse.

FIGURE6.Pl x s
y FA
z­ -
Co

Q D

Q <' CK
 Problems 221

The S output represents the sum. Obtain the state equations

for S and Co.
2. Draw the timing diagram for the given input signal and cir­
cuit of Figure 6.P2. Assume the starting value of Q2Ql = 00.

FIGURE6:P2

~. J1 01 J2 O2
I
--c > "'-V'
i
1 K1 01 K2 Q2

CK

3. What sequence should repeat for the sequential circuit of Fig­

ure 6.P3 for the following initial inputs:
a. QSQ2~ = 001 b. QSQ2Ql = 100

FIGURE6.P3

CK~---------------+--------------~

4. Obtain a T FF from a D FF.

5. Explain the behavior of the circuit of Figure 6.P4.

FIGURE6.P4 A--~

l--------I D 0
B--I---\

C---r----------------------------G> o

6. A sequential circuit has two inputs, X and Y, and one output,

Z, such that
222 CHAPTER SIX Sequential Devices

11 = XQ2 + YQ2

12 = XQl

Kl = XfQ2

K2 = Xf + Ql

Z = XYQl + XYQ2

Obtain the logic diagram and state equations.

7. Find the output and state sequences for the circuit of Figure
6.P5 if the initial state is Q = 0 and the input sequence is
a. x= 101101100 b. x= 111011101

FIGURE6.P5 r­

x ~ ,~

V \
r- c-­ ) l
f

.­

- Q S

< i
CK

'---­ Q R

8. Show the detailed working of the circuit of Figure 6.35.

9. Repeat Example 6.2 when the NAND gate has a delay of 4
units and the NOT gate has a delay of3 units.
10. Repeat Example 6.2 when the gate delays are lumped
together.
11. Repeat Example 6.2 when the NAND gate has a delay of 4
units and the NOT gate has a delay of 3 units, but assume a
lumped model for the circuit.
12. Comment on the behaviors of the two circuits of Figure 6.P6.
13. Analyze the circuit of Figure 6.P7.
14. Obtain a sequential system for performing a multi-bit, BCD­
to-binary conversion. Describe the general working principles
of your circuit. _
 Suggested Readings 223

FIGURE6.P6

D Q L Q
A I

-
>T
CI(_ .
B ~~7= ""'
/'"'
I

Q
I
[a] . [bJ

FIGURE6.P7
A-~ Q

T
B--~------------~

15. Obtain a sequential system for performing a multi-bit, binary­

to-BCD conversion. Describe the general working principles of
your circuit.
16. Obtain a sequential system for performing a multi-bit, Gray­
to-binary conversion. Describe the general working principles
of your circuit.
17. Obtain a sequential system for performing a multi-bit, binary­
to-Gray conversion. Describe the general working principles of
your circuit.
18. Obtain a sequential syst~m for performing a multi-bit com­
parison between two numbers. Describe the general working
principles of your circuit.

Almaini, A. E. A. "Sequential machine implementations using universal

Suggested Readings
. logic modules." IEEE Trans. Compo vol. C-27 (1978): 951.
Beraru, J. "A one-step process for obtaining flip-flop input logic equa­
tions/' Compo Des.vol. 6 (1967): 58.
Chen x., and Hurst, S. L. "A comparison of universal-logie-module real­
izations and their application in the synthesis of combinatorial and
sequential logic networks." IEEE Trans. Compo vol. C-31 (1982): 140.
Fleischhammer, W., and Dortok, O. "The anomalous behavior of flip-flops
in synchronizer circuits." IEEE Trans. Compo vol. C-28 (1979): 273.
Fletcher, W. I. An Engineering Approach to Digital Design. Englewood Cliffs,
N.J.: Prentice-Hall, 1980.
Floyd, T. L. Digital Fundamentals. 2d ed. Columbus, Ohio: Charles E. Mer­
rill, 1982.
224 CHAFTER SIX Sequential Devices

Gaitanis, N., and Halatsis, C. "Negative failsafe sequential circuits." Elect.

Lett. voL 16 (1980): 615.
Huffman, D. A. "The synthesis of sequential switching circuits." ] Frank.
Inst. vol. 257 (1954): 275.
Joseph, J "On easily diagnosable sequential machines." IEEE Trans.
Compo vol. C-27 (1978): 159.
Kline, R. Strnctured Digital Design Including MSI/ LSI Components and Micro­
processors. Englewood Cliffs, N.J.: Prentice-Hall, 1983.
Manning, F. B., and Fenichel, R. R. "Synchronous counters constructed
entirely ofJ-K flip-flops." IEEE Trans. Compo voL C-25 (1976): 300.
Mealy, G. H. "A method for synthesizing sequential circuits." Bell Syst.
Tech.] vol. 34 (1955): 1045.
Nanda, N. K., and Bennetts, R. G. "Reconvergence phenomenon in syn­
chronous sequential circuits." Elect. Lett. vol. 16 (1980): 303.
Noe,P.S., and Rhyne, V. T. "Optimum state assignment for D flip-flop."
IEEE Trans. Compo vol. C-25 (1976): 306.
Taub, D. M. "Hardware method of synchronizing processes without using
a clock." Elect. Lett. vol. 19 (1983): 772.
Unger, S. H. "Self-synchronizing circuits and nonfundamental mode oper­
ation." IEEE Trans. Comp. vol. C-26 (1977): 278.
Voith, R. P. "Minimum universal logic module sequential circiuts with
decoders." IEEE Trans. Compo vol. C-26 (1977): 1032.
Wilkens, E. J "Realization of sequential machines using random access
memory." IEEE Trans. Compo vol. C-27 (1978): 429.
Witte, H.-H., and Moustakas, S. "Simple clock extraction circuit using a
self-sustaining monostable multivibrator output signal." Elect. Lett. vol.
19 (1983): 897.
226 CHAPTER SEVEN Design of Synchronous Sequential Circuits

already encountered some examples of two-state state diagrams in

Figure 6.40.
FIGURE 7.1 State Diagram. The state diagram in Figure 7.1 represents a synchronous
circuit with three states, A, B, and C, and an input variable, x.
x/a In each state it is necessary for the circuit to be able to deter­
mine which state it is in and what the current value of x is, and
then to set up the FF inputs such that the correct state is
x/a entered when the clock input occurs. The arrows connecting the
states represent the occurrence of a clock input, and the vari­
ables aJ ngside the arrows show the input condition that causes
that pai..h to be followed.
If the circuit is currently in state A and x = 1, the circuit will
remain in state A when the clock occurs (x/O). If x = 0 the circuit
enters into state B when the clock occurs (x/O). In both of these
cases the circuit yields an output of O. The output value and the
input condition are both indicated next to the corresponding transi­
tion paths. If the circuit is in state B and x = 1 when the clock
occurs, the circuit returns to state A (x/O). However, if x = 0 when
the clock occurs, the state C is entered, resulting in an output of 1
(XI 1). And finally the circuit moves coincident with the clock from
state C to states A and B, respectively, when x = 0 and x = 1. In
either case the output remains O. The state diagram must show each
of the states of the circuit and all conditions necessary for entering
or exiting the states. In this case two transition lines leave each of
the states.
The implementation of a sequential circuit with n states will
require m FFs where 2m > n. The outputs of these FFs are called the
state variables and are used to identify which state the circuit is in. An
additional design tool that contains the same information as the
state diagram in tabular form is the state table. The state table of the
system shown in the state diagram of Figure 7.1 is given in Figure
FIGURE 7.2 State Table 7.2. It can be seen that the outputs are associated with the transition
Corresponding to the Diagram. of paths only and are not functions of any transition states. Circuits
Figure 7.1. such as this are generally known as Mealy-type machines. The
Mealy outputs, in most cases, are pulses coincident with the input
pulse causing the state transition. An alternate output type, called
Next State (NS)
Output (Z) the Moore output, is associated with the present state only. The gen­
Present eral forms of the Mealy and Moore circuits are shown in Figure 7.3.
State (PS) x=O x 1
These circuits take their names, respectively, from G. H. Mealy and
A B,O A,a E. f. Moore, two of the most famous pioneers in sequential design.
B C,1 A,O The outputs from Moore-type circuits are independent of the
C A,O B,O inputs. The Moore outputs change their values only when the states
change because of a change of the inputs .
. There are many systems that possess both Mealy and Moore out­
puts; in other words, some outputs are conditional on both the
inputs and the state of the circuit, while others are dependent only
on the state of the circuit. Note, however, that the Mealy output is
 j

Design of CHAPTER
SEVEN
Synchronous

Sequential Circuits-----'

7.1 Introduction
In this chapter we will examine clocked sequential citcuits. These
circuits will employ combinational circuits and flip-flops. All circuit
action will take place under the control of a periodic sequence of
pulses called a clock. Each clock pulse will permit the circuit to
either remain in its present state (present set of FF values) or move
to another state (a new set of FF values). The advantage of clocked
sequential circuits is that glitches that occur due to the imperfect
nature of the logic devices will have no effect. This is possible only if
we choose the clock period such that all glitches due to multiple
delay paths end before the FFs encounter future changes.
The synthesis of sequential circuits consists of obtaining a table
or diagram for the time sequence of inputs, outputs, and internal
states. Boolean expressions are then derived by incorporating the
behavior patterns of FF memory elements. In the following sections
we will introduce these design sequences along with several syn­
chronous sequential circuit examples. After studying this chapter,
you should be able to:
o Obtain a state diagram for a synchronous sequential

machine;

o Eliminate redundant states;

o Realize a sequential circuit from the state table;
o Differentiate between Mealy and Moore circuits.

7.2 State Diagrams and The functional interrelationship that exists among the input, the
output, the present state, and the next state is best illustrated by the
State Tables state diagram or the state table. The state diagram is a graphical rep­
resentation of a sequential circuit in which the states are represented
by circles and transitions between states shown by arrows. We have

225
 7.2 State Diagrams and State Tables 227

FIGURE 7.3 General Model of

Sequential Machines: [aJ Mealy lnputs:
Next-state
and [b] Moore. decoder .....
t' ..• . ..
r--4
Memory

L;+
. Output
decoder
<4­

L..:..,

t .. ·1 Mealy
outputs
raj

Inputs:
Next-state
decoder

Memory

Output • Moore
decoder • outputs

[b]

easily convertible to equivalent Moore outputs and vice versa.

(More about this conversion will be said in one of the worked-out
examples.) Figure 7.4[a] shows the format for a state diagram where
the Moore-type outputs are circled along with the corresponding
present states. Figure 7.4[bJ shows the state table corresponding to

.FIGURE 7.4 Moore Model for a

Sequential Circuit.

NS
PS x=o x 1 Z

A C B 0
B C A 1
C 0 C 0
0 A A 1

raj [bJ
228 CHAPTER SEVEN Design of Synchronous Sequential Circuits

the state diagram of part [a]. It is important to point out that both
Mealy- and Moore-type circuits are equally applicable to both syn­
chronous and asynchronous circuits; and the minimum number of
external inputs to anyone of these circuits is one. For a synchronous
circuit, that one input must be the system clock.

EXAMPLE7.i SOLUTION

Obtain the state diagram of a The realization of the state diagram for the controller is very straightfor­
controller for a serial machine that ward, as shown in Figure 7.5. This follows from Rule 2(b) of Section 1.4.
performs the 2's complement The 2's complement of a number is obtained by complementing all bits to
operation (see Example 5.3 for the the left of the least significant 1 in that number. State A takes care of the
equivalent parallel scheme). situation when none of the serial inputs are changed, whereas state B corre­
sponds to the changing (1 's complement) of inputs. The controller remains
in state A as long as the low-order Os of the input are encountered. The first
input of 1 moves the machine to state B so that all subsequent inputs are
complemented. To begin a new conversion, the machine needs to be reset
back to state A (indicated by the broken line).

FIGURE 7.5

0/0

EXAMPLE 7.2 SOLUTION

Obtain the state table for a The state diagram for this machine is obtained as shown in Figure 7.6. The
-synchronous sequential machine machine resides in state A as long as the sequence does not begin. This situ­
that deteCts a 01 sequence. The ation would include two distinct cases: either (a) the machine is yet to see a
detection of sequence sets the single bit, or (b) the machine has so far examined a 1 or a string of 1s.
output, Z = 1, which is reset only However, once the first bit, 0, of either sequence, 01 or 00, has been
by a 00 input sequence. detected, the machine moves to state B. Finally, state C is reached if either
(a) the complete sequence, 01, has been located, or (b) the resetting

FIGURE 7.6 0/0 1/1

1/0 1/1

0/1
0/0
 7.2 State Diagrams and State Tables 229

sequence, 00, is yet to begin. The state table of the machine readily follows
from the state diagram. It is shown in Figure 7.7.

FIGURE 7.7
NS,Z
PS x 0 x 1
A B,O A,O
B B,O C,l
C B,1 C,1

EXAMPLE 7.3 SOLUTION

Obtain the Moore equivalent state FIGURE 7.8
table for the Mealy machine of
Figure 7.8.
NS, Z
PS x==O x 1
A C,O AO
B B,O AO
C 0,1 C,l
0 0,0 B,O
E C,1 A,O
FIGURE 7.9
The state table of Figure 7.8 is a Mealy type since the outputs are not ass0­
NS, Z ciated only with the states. It is the desired goal of this problem to associate
PS x==O x==l each of these outputs with a particular state. It may be seen that the two
A A,O
next states, C and D, are associated with two different outputs, 0 and 1.
C',O
The next states, A and B, are associated always with an output of 0 only.
B, . "}j,,, ,.. ~B,0 ,~"O
Accordingly, four new states, C, C', D', and D", may be introduced to
,

C' , 0,1 C',l

replace the states C and D. The states C and D' correspond, respectively,
C' 0,1 C",l
to states C and D when the output is a O. Similarly, the states C' and D"
0 0,0 B,O
correspond, respectively, to states C and D when the output is a 1. The
E C',l AO state table obtained after the introduction of only C and C' is shown in
Figure 7.9. Next, D' and D" may be included to obtain the state table as
shown in Figure 7.10.

FIGURE {.fO NS, Z

PS x=O x 1
A C',O AO
B 8,0 AO
C' 0",1 C",l
C' 0",1 C',1
0' 0',0 B,O
Oil 0',0 B,O
E C',1 AO
230 CHAPTER SEVEN Design of Synchronous Sequential Circuits

At this time each of the states has only one output associated with itself.
Accordingly one could obtain the equivalent Moore machine as shown in
Figure' 7.11. TIlls table has been constructed in a way so that it resembles
the format of Figure 7.4[b].

J7I(;l1~ 7.11 NS
PS x=O x= 1 Z
A C' A 0
B B A 0
C' 0" C" 0
C" 0" C" 1
0' 0' B 0
0" 0' B 1
E C" A -

Note in Example 7.3 that the Moore machine of Figure 7.11 is

equivalent to the original Mealy machine of Figure 7.8 only in the
sense that its output appears as pulses. The outputs that occur in (:'
and D" are pulses that are high (1) for a full clock period. In the
Mealy circuit the pulses are high (1) while the clock is high. More­
over, two additional states were necessary to make this conversion
complete. In fact, the Moore equivalent machine generally consists
of more states than the corresponding Mealy machine.

7.3 Equivalent States

When constructing a state diagram from the word statement of a
design problem, a state that is identical to another state may inad­
vertently be included. The redundant state or states will increase
the number of total states and may require the addition of another
FF, making the circij.it more expensive. Redundant states also
decrease -the number of don't-cares or unused states and thus
increase the overall complexity of the circuit equations. In addition,
fault diagnostic techniques used for failure analysis are based on the
assumption that no redundant states exist. One of the design goals,
therefore, is to eliminate all redundant states from the state diagram
and/or table.
Sometimes redundant states are obvious. In Figure 7.12 a state
diagram and a state table are given that include several redundant
states. Two states are defined as equivalent if they have identical out­
puts and make transitions to the same states for a given control vari­
able value. In Figure 7.12 no state can be equivalent to state D
because it is the only state with an output of 1. We can make the
statement that state E is equivalent to state F only if state A is
equivalent to B and state F is equivalent to state G. Again, states F
and G are equivalent only if state G is equivalent to state E and also
 7.3 Equivalent States 231

FIGURE 7.12 Sequential Circuit

with Redundant States: [a] State
Diagram and [b] State Table.

NS
I PS x=O x=1 Z
A B C 0
B A C 0
C 0 C 0
0 0 E 1
E A F 0
F B· G 0
G A E 0

[a] [b]

if state A is equivalent to B. This line of reasoning becomes

extremely confusing for large systems.
A systematic way of looking for redundant states usually is
accomplished by means of a new tool called an implication table. The
numbers of rows and columns in this table are both equal to n -1
where n is the number of states in the to-be-reduced state table. An
;'""
implication table that is used to locate the possible redundant states
of Figure 7.12[ b] is illustrated in Figure 7.13[a]. This table provides
a bookkeeping technique that allows a systematic way to find states
that are equivalent. The algorithm employed for the equivalency
search in the implication table is as follows:
1. All states except the very first one are used as row labels of
the table, and all states except the very last one are used as
column labels.
2. The entries at each of the cells are made by comparing the
next-state columns of the two states that are uSed to iden­
tify the cell in question. No entries are made in a cell if it
corresponds to states that have the same output and the
same next states for all control variables. If the equivalency
of the two states is dependent on whether or not the states P
and Q are equivalent, then the pair (PO) is entered in the
corresponding celL A cross "X" is placed over those cells for
which outputs of the two states are different for anyone
input condition.
232 CHAPTER SEVEN Design of Synchronous Sequential Circuits

FIGURE 7.13 State Reduction

by Implication: [a-6] Implication
Tables and [e] Equivalence
Partition Table.

B AB
B AB

C C

D D

E E

AB AB
F CG F
/ ,eG FG FG

AB AB AB
G CE EF G EF )
CEA ;' EG EG
A B C D F A B C D E F
E?
[a] [b]

F (FG)

E (EFG)

D (EFG)

C (EFG)

B (EFG)

A (AB)(EFG)

(AB)(C)(D)(EFG)

(e]

3. The next step involves elimination of as many cells as possi­

ble based on the respective cell entries. If the entries in any
cell include at least one pair of states that. are none­
quivalent (i.e., the cell corresponding to that pair has
already been crossed out), then that cell is eliminated by
marking a X on it.
4. Successive passes are then made through the entire table to
determine if any more states should be crossed out to indi­
cate nonequivalency. The cell ha~ng an entry PQ should
be crossed out only if the particular cells corresponding to
the labels P and Q have already been eliminated. This pro­
 7.3 Equivalent States 233

cess of elimination is continued until no other cells can be

eliminated.
5. Redundancy is then determined by txamining the surviv­
ing cells of the implication table. Each surviving cell corre­
sponds to an equivalency condition between the two states
that are used to label that cell.

The role of the preceding algorithm will become meaningful when

we use it to investigate the state table of Figure 7.12[b].
The entries in the table of Figure 7.l3[a] are made by comparing
each of the states with the others. For example, CG is entered in the
cell corresponding to A and F since the equivalency of these two
states is dependent on the equivalency of states C and G. Likewise,
both CG and AB are entered in the cell corresponding to Band F
since these two states would be equivalent only if both (a) C and G
and (b) A and B are equivalent. Similar reasonings are made in
completing the remaining cells.
Referring to Figure 7.13[a] and the state table of Figure 7.12[b],
D is the only state that has an output of 1, so it cannot be equivalent
to any other state. Therefore, all cells that have either a row or a
column designated by D are crossed out. Next all the cells that have
an entry composed of D are crossed out. As a consequence· we find
that all cells corresponding to 'the row and column designated by C
also have been eliminated. This indicates that state C, like D, is dif­
ferent from all other states as welL We next cross out all cells that
have at least one pair of entries involving C. Continuing this pro­
cess, we come up with the table of Figure 7.13[b]. Only four surviv­
ing cells are left in the implication table.
An equivalence partition table, as shown in Figure 7.13[c], is next
obtained by listing all horiwntallabels (in the reverse order) as its
row labels. A check is now made of each column of the final impli­
cation table (Figure 7.13[b]), from right to left, making note of the
cells that have not been eliminated. The row and column labels of
the surviving cell form an equivalent pair that is noted in the equiv­
alent column identifier in the partition table. Equivalent pairs such
as (PQ) and (QR) imply the presence of a larger group of equivalent
states (PQR). For each column of the implication table an entry is
,"-'
''''''~?' ';- "
made in the equivalence partition table, provided there is at least
one surviving cell in that column. The entries from the previous
rows of the partition table are entered as long as there is no repeti­
tion of the entry. In the example in question, the cell corresponding
to G and F is not crossed out, and therefore the equivalent pair (FG)
is listed next to F in the equivalence partition table. There are two
cells corresponding to rows F and G and column E that result in
(FE) and (EG) or the equivalent group (EFG). Therefore, we write
(EFG) next to E, and since (EFG) already contains (FG), nothing
234 CHAPTER SEVEN Design of Synchronous Sequential Circuits

from the previous row is repeated here. Since columns B, C, and D

have no surviving cells, the entry (EFG) is repeated three times. No
new equivalencies result until we come to consider A, for which the
entry (AB) is added to the list. The last line of the partition table
reveals that there are two sets of equivalent states: (AB) and (EFG).
., This implies that state A is equivalent to state B and states E, F, and
G are equivalent to each other. Consequently, the redundant states
are now eliminated by considering only one state from each equiva­
lent group. Tbe resultant state diagram and the state table are
shown in Figure 7.14 where B, F, and G have been removed. The
state B has been replaced by A and states F and G have been
replaced with E.

FIGURE 7.14 [0] State Diagram

with No Redundancy and [b] State
Table with No Redundancy.
NS
PS x=o x=1 Z
A A C 0
C 0 C 0
0 0 E 1
E A E 0

[a] [b]

I
7.4 State Assignments State assignment is the process of adopting a binary coding scheme for
the symbolic states of the state table so that it is possible for the cir­
cuit to remember which state it is in. Each bit in the code represents
the output of an FF and is called a state variable. For n number of
states, a total of m FFs will be necessary such that m is the smallest
integer satisfying the relationship 2m > n. Any unique assignment is
valid; however, it is always better if an attempt is made to assign_
codes in such a way that the number of cases where more than one
bit in a code changes when states change is kept to a minimum.
When the symbolic states are replaced with the binary coding
scheme, a binary state table, commonJy known as the transition table,
results.
FIGURE 7.15 Transition Table The state diagram of Figure 7.1 has three unique states, and,
for the Circuit of Figure 7.1. therefore, at least two FFs are necessary for designing the corre­
sponding logic circuit. This implies that of the four different codes­
PS NS 00,01; 10 and ll-onJy three can be used. For this example, if one
0 10 2 x=o x=1 chooses to assign A = 00, B = 01, and C = 11, the resulting transi­
00 01,0 00,0 tion table of Figure 7.15 is obtained.
We may see that the present state QIQ2 = 01, upon receiving
@ U,1 00,0
11 01,0
the input x = 0, moves to the next state 11 with the resultant out­
00,0
10 --,­ ---,­ put of 1. During this transition the ~·value changes from 0 to 1, as
shown highlighted in the table. In the same present state when x =
 7.5 Excitation Maps 235

1,Q2 changes from 1 to a O. In either case the binary state changes

only one of its bits. However, the present state QI Q2 = 11 is differ­
ent from the others, because when x = 0 both of the FF bits need to
change from 1 to O. There are situations where such a condition
could cause problems, as we shall discover later.

EXAMPLE 7.4 SOLUTION

Obtain a transition table for the There are four states, and, therefore, only two FFs need to be considered.
sequential:machine of Figure The transition table results when A = 00, C = 01, D = 11, and E = 10,
7.l4[a]. as shown in Figure 7.16. For the state assignments made, the transition
table shows that there are no transitions for which both state variables
change. Consequently, this assignment of states is considered to be very
good.

FIGUKE 7.16 PS NS
0,0 2 X 0 x 1 Z
00 00 01 0
01 11 01 0
11 11 10 1
10 00 10 0
..

7.5 Excitation Maps Up to this point when considering FFs we have been concerned
with how they respond to various inputs. We will now encounter the
design problem. of determining their inputs such that the proper
values are present to cause the next state to result when the clock
input occurs. This input control is accomplished by deriving the
respective excitation equations. The output equations and the state
variable excitation equations are derived separately, as shown in
Figures 7.17[a-h]. The FF input maps are usually called excitation
maps.,> ;!~J.i-)
FIGUKE 7.17 [a] Output Table 00
1 2 00
1 2

and [b] Excitation Maps. x 00 01 11 10 X 00 01 11 10 X 00 01 11 10

0 0 1 0 -- 0 0­ e -1 -- 0 1­ -0 -1 -

1 0 0­ 0 -- 1 0­ 0­ -1 - 1 0­ -1 -0 -­

z
[a] [b]
,,_., ,

The combinational circ~it corresponding to the transition table

shown in Figure 7.15 has three variables-x, Qll and Q2-that gen­
erate the FF inputs. Ql and Q2 represent the current state of the cir­
236 CHAPTER SEVEN Design of Synchronous Sequential Circuits

cuit, and x is the external control input that in combination with

the present state determines which action is to occur. The excitation
table entries are the values of FF inputs that would cause the transi­
tion to next-state variables when the clock input occurs. We have
chosenJK FFs, as an example only, for generating the exdtations.
One particular transition is emphasized for illustration both in the
transition table and in the corresponding excitation map of Figure
7.17. The transition of Qll as shown in Figure 7.15, from 0 to 1
requires the FF input conditionJl = 1 and Kl = -. This informa­
tion is entered in the corresponding cell, x = 0 and ~ Q2 = 01, of
the excitation map. Continuing this procedure, the excitation and
output equations may be obtained from the K-maps of Figure 7.17:

Jl = XQ2

Kl = 1

J2 = x

K2 = Q,x + QIX = Ql EB x

Z = Q,Q2X' CK

Note that the clock input, CK, is ANDed with Ql Q~ to produce the
FIGURE 7.18 Circuit desired output of a synchronous machine. The resultant sequential
Implementation of the Example. circuit is obtained as shown in Figure 7.18.

J, 0, z
x

CK------~--------.-------------------------~----------~

/'

f-,!
I

0 f Consider a circuit similar to that of Figure 7.1 but having a

Mealy output, Zll and a Moore output, Z2, as shown in Figure
7.19[a]. There is to be an output coincident with the clock input
when the circuit moves from state B to state A, and another output
whenever the circuit is in state C. For the Zl output, assuming nega­
tive edge-triggered FFs, the circuit must be in &tate 01 and both x
and clock must occur. The Mealy output is given by Zl = ~Q2X'
CK. If the FFs being Used for the state variables were positive edge­
triggered, the Zl output would occur in the state following 01 since
the positive edge of the clock would move the circuit into the next
state. The next state's state variables, 00, and the inputs would be
ANDed to form Zl in this case. However, the Moore output is a 1
 7.6 Design Algorithm 237

FIGURE 7.19 Circuit with

Mealy and Moore Outputs: [a]
State Diagram, [b] Transition
Table, [c] Output Tables, and [dJ
Circuit.

"

PS NS,Z1
0 10 2 x=O X= 1 Z2
A 00 01,0 00,0 0
B 01 11,0 00,1 0
C 11 00,0 01,0 1
- 10 , ,

[a] [b]

o 1 00
1 2
O2 0 1 X 00 01 11 10
0 0 - 0 0 0 0 -

1 0 1 1 0 1 0 -

[c]

x ~J-~
Y
i }­ J1 0 1 f--­ LJ2 O2 '---4
~i)
I
V
1
r---'­ )'

K1 01 ­
r1.
h: [>

K2 O2 ......

CK Z1

[d]

only when the circuit is in state C. Figures 7.l9[bJc] show the steps
involved in obtaining the final circuit of Figure 7.19[d].

7.6 Design Algorithm We have examined the steps of the design of a synchronous sequen­
tial machine in the last few sections. Figure 7.20 gives a comprehen­
sive flowchart of an algorithm for the design.of sequential machines.
238 CHAPTER SEVEN Design of Synchronous Sequential Circuits

FIGURE 7.20 Sequential Circuit

Design Flowchart.

State diagram and table

formulation

Redundance elimination

No

Has
the machine been
reduced enough?

Yes

FF selection and state

assignment

Excitation equation and

circuit implementation

The algorithm can be summarized by the following steps:

Step 1. Obtain the state diagram from the word statement of the
problem.
Step 2. Obtain the state table from the state diagram.
Step 3. Eliminate the redundant states.
Step 4. Make state assignments.
Step 5. Determine the type of FFs to use and obtain the corre­
sponding excitation maps.

Step 6. Determine the output and FF equations.

Step 7. Construct the logic circuit.

 7.6 Design Algorithm 239

These design steps often lead to a rather lengthy process that var­
ies from problem to problem. The following examples illustrate the
implementation of the sequential design algorithm.

EXAMPLE 7.5 SOLUTION

Design a two-bit clocked sequential Steps /-2. We can visualize such a device as the one that receives clock
counter circuit that counts clock pulses as input. Each time a clock pulse is received, the counter should
pulses. count up. No control variable is needed; only the occurrence of the clock
pulse is necessary for a state change. We shall assume further that the states
change at the trailing edge of the clock input. The counter could be
designed such that the outputs go through the sequence 00 ~ 01 ~ 10 ~
11 and repeat. Since the counter is only a two-bit device, it would have to
reset at the fourth clock. The state diagram and the corresponding state
table for the counter are obtained as shown in Figure 7.21. We might
choose to have the outputs directly from the FF outputs, in which case such
outputs would be classed as Moore type.

FIGURE 7.21

PS NS Z1 Z2
A B 00
B C 01
C D 10
D ,A 11

[a] [b]

Step 3. The four states-A, B, C, and D-are all different since they stand
for completely different events. Consequently, we may conclude without
any doubt that none of these are redundant states.

Step 4. Making state assignments is an important step for at least one

reason. The assignments of the states are crucial in determining the sim­
plicity, or for that matter complexity, of the resultant circuit. The output
circuits can be eliminated totally if the state assignments for the states A, B,
C, and D are made the same as the corresponding M~re outputs of each
state. Accordingly, the chosen assignments are A = 00, B = 01, C = 10,
andD = 11.

Step 5. The number of FFs are indeed two. This fact was also given in
the initial word statement of the problem. We might choose JK FFs, for
example. Consequently both the transition table and excitation map are
obtained as shown in Figure 7.22.
24{) CHAPTER SEVEN Design of Synchronous Sequential Circuits

FIGURE 7.22 PS
0,02 NS Z,Z2
0, 0,
00 01 00 O2 0 1 O2 0 1
01 10 01 "
:"<,
1~ 1­
0 0­ -0 V
10 11 10
11 00 11 1 1­ -.1 -'--1 -1

[a1 [b1

Step 6. The excitation maps are minimized to give

Jl = Q2

KI = Q2 /

J2 = 1 V

K2 = 1

The outputs are easily realizable directly from Figure 7.22[aJ. They are as
follows:
ZI = QIQ2 + QIQ2 = Q,(Q2 + Q2) = Ql
Z2 = ~Q2 + QIQ2 = Q2(~ + Ql) = Q2
Step 7. The resulting circuit obtained from these excitation and output
equations is shown in Figure 7.23.

FIGURE 7.23

J, 0, f­
1
J2 02 J
r-c :>
K1 0, f­
K2 O2 r-­

CK

EXAMPLE 7.6 SOLUTION

Repeat the design of Example 7.5 Since the FF state assignments are different from the corresponding Moore
by assigning A = 00, B = 01, C outputs, ZI and Z2, the circuit will require additional gates. The corre­
= 11, and D = 10. Construct the sponding transition table, theJK excitation maps, and the output map are
corresponding timing diagram as shown in Figure 7.24. The corresponding Boolean equations are obtained
well. as follows:
 7.6 Design Algorithm 241

1l = Q2
K, = Q2
12 = Q,
K2 = Q,
Z, = Ql
FIGURE 7.24 Zz = Q,Q'l + Q,Q2 = QI EB Q2

PS
0 10 2 NS Z1 Z2
01 01 01
00 01 00 O2 0
l--_~
02 0 1 O2 0 1
01 11 01
11 10 10 o 0­ -1 0 1­ 0­ 0 00 11
10 00 11
-0 1 -0 -1 1 01 10

[a1 [b)

The resulting logic circuit diagram is obtained as shown in Figure 7.25. It

is obvious that the circuit in Example 7.5 is simpler. The timing diagram of
Figure 7.26 illustrates the operation of the circuit of Figure 7.25.

FIGURE 7.25

 242 CHAPTER SEVEN Design of Synchronous Sequential Circuits

/ EXAMPLE 7.7 SOLUTION

oJ

Complete the design of a clocked The state diagram consists of five states, A-E. States B, C, D, and E repre­
sequential circuit that recognizes sent, respectively, the occurrence of the first, second, third, and fourth bits
the input sequence 1010, including of the sequence 1010. State E has a Moore output Z = 1 indicating the
overlapping such that for input x completion of a sequence. A subsequent input of 0 would move the circuit
= 00101001010101110 the to state~, which indicates the input is out of sequence. An input of 1 while
corresponding output Z is in E moves the circuit from state E to state D since sequence overlapping is
00000100001010000. allowed. The corresponding state diagram showing the transitions for each
value of x is provided in Figure 7.27.

FIGURE 7.27
f\

:j x ()
?L "

Upon power-up, the circuit begins from state A. As long as the string of
input is devoid of 1 (i.e., the first bit of a 1010 sequence), the circuit
remains at this be~nning state. Once a 1 h~ been located the circuit
moves to state B, indicating that the first bit has already been detected.
Subsequent detection of 0, 1, and 0, in that order, would amount to mov­
ing the circuit to states C, D, apd E, respectively. Once the state E is
reached, the circuit gives an output indicating the completion of a 1010
FIGURE 7.28
sequence. However, while at state B if the circuit detects a 1, the circuit re­
enters state B. This is due to the possibility that the most recently observed
1 might be the beginning of a 1010 sequence. For similar reasons, the
NS detection of a 1 at state D causes the circuit to move to state B also. Again
PS x=O x=1 Z
-
at state C, a detection of 0 eliminates the possibility of having the desired
sequence, 1010. So, the circuit resets back to state A. Likewise, the circuit
A A B 0
resets from state E to state A if it locates a O. However, an interesting case
B C B 0
happens when the circuit is at state E and it has just detected a 1. This
C A D 0
time the circuit moves back to state D. This is due to the fact that detection
D E B 0
of a 1 at state E is equivalent to detecting the third bit of a newer 1010
E A D 1
! sequence.
The problem involves five states requiring three FFs. Three of the eight
possible states will remain unused. The state table and transition table cor­
PS NS
responding to the arbitrary assignments of A = 000, B = 001, C = 011, D
0 102 03 X=O X= 1 Z = 111, and E = 101 are shown in Figure 7.28. The'output and excitation
000 000 001 0 maps corresponding to the use ofiK FFs may now be obtained as shown in
001 011 001 0 Figure 7.29. Proper grouping of the K-map cells results in the following
011 000 111 0
equations:
111
101
101
000
001
111
0
1
Z= QIQ2' CK J

il = Q~

7.6 Design Algorithm 243

FIGURE 7.29 \
) ('"
~! .j .­
0 1 '" '..<"""'" 01 01
' j '~''---.-'----r--.....~
'., ~": ~
-
J-' 0- -- -- --
~ t­
O- 0­
I
. r·
I O.,. :!
) ! 0-
c' -,
-- -- 0- 1­
],
~i
0 - - - 0- 1- -1 -0 -0 -0 -0 -0
0

0 0 0 1
o 0- 0- -0 -1 -0 -1 -0 -1
If',
\

O2 O2
J2 K2 J3 K3

Kl = Q~ + Q2 X = Q2 EI1 X
12 = QIX + QIQ3X
K2 = X + Ql
13 = x
K3 = QIQ~ + QIQ2X = (Ql EI1 Q2)X
The sequential circuit of the 1010 sequence detector is obtained using the
above equations and is shown in Figure 7.30[a]. The timing diagram of
Figure 7.30[ b] shows the relationship between the output, the clock, and
the input. '

FIGURE 7.3Oa

W-U-====-- z

.:. :"-.,­

CK--~---+-r--------~--~------+----~

[a]
 244 CHAPTER SEVEN Design of Synchronous Sequential Circuits

In the examples considered thus far we have used JK FFs and

gates. D FFs and other logic devices may also be used. The excita­
tion table for all of the FFs.is given in Figure 6.39. In Examples 7.8
and 7.9, D FFs are used.

\j
EXAMPLE 7.8 SOLUTION
Obtain a scale-of-seven up-counter, FIGURE 7.31
as shown in the state diagram of
Figure 7.31, using D FFs and PLA.
Assume that the counter is tied to a
seven-segment display device.

The state table of the counter may be obtained as shown in Figure 7.32.

FIGURE 7.32 PS
0 30 20 1 NS

000 001
001 010
010 all
011 f'--jOO
100 -­ 101

\
101
110
110
P9o.
/ ,! 1,
 7.6 Design Algorithm 245

The excitation K-maps corresponding to D FFs are next obtained as shovm

in Figure 7.33. The Boolean equations, therefore, are as follows:
D3 = Q2Q3 + QIQ2'­
D2 = Q\Q2 + QIQ2Q3
FIGURE 7.33 DI = QIQ2 + QIQ3

0 302 030 2 (h02

01 00 01 11 10 0 1 00 01 11 10 0 1 00 01 11 10

0 0 0 0 '1 0 0 1 0 0 0 1 1 0 1

1 0 I, 1 ::,) 1 1 f) 0 - (1" 1 0 0 - 0
-" r--J' ---~

0,

The PLA circuit configuration follows as'shovm in Figure 7.34. The dots in
the intersection matrix correspond to either an OR or an AND operation.
The segment allocation for the LED display has already been defined in
Example 4.8.

FIGURE 7.34

6543210
I

~ -" -­ ~-

r----- (, GJ ~
_ \1/'7\.-1 *' 0,
01
,,­
f- .. - 0,
, ,­

L \
\ -­ /
\/ (\"
p- Ol f ­
------
;­
V '­ 1-­ - -L
;
__--1-·­
~i
1­ -" O2 O2 f -
!
l-'
.I OR
i

-
r> O2 f -
POt ]-\'1 +7 ,
Q
RV'T 1 + Lr
(­ !, _ -', i -, Y 03
-_,-­
03 f -
S
T
U
V > 03 f -
CK,­
 246 CHAPTER SEVEN Design of Synchronous Sequential Circuits

EXAMPLE 7.9 SOLUTION

Using D FFs and assorted gates, The MSB of both numbers may be fed as inputs to the comparator. Two
design a sequential comparator that outputs, Z, and Z2, may be assumed to accompany the circuit. If X > Y,
is to determine which of the two then Zl = 1; if X < Y, then Z2 = 1; and if X = Y, then Z, = Z2 = o.
multi-bit numbers, X and Y, of The state diagram of the comparator is obtained as shown in Figure
equal length is larger. 7.35, where Xi and Yj are the ith bits of X and Y, respectively. The transi­
tiontable may then be obtained as shown in Figure 7.36. For simplicity,

FIGURE 7.35
XiV, =

Q
o
FIGURE 7.36
NS

PS XiYi 00 01 11 10 Zl Z2

00 A ofCfA '13
() (.
A 00
\0 B !O 8 ;'B ·;8 ';8 10
-~5----:------~\~ __ _ ! In
,v . C ~. r C.rC
i"I: '~.
"" !C .JJC 01

--'1--,..~
XI(QO\ \.
01 LI __ \1 . !o
'\()
,,
0 1\'0 V'i \0
i
M, I (.
\\ ,I I0 I
we may assign A = 00, B = 10, and C = 01. These assignments would
l 0 \ to v \v i 0
allow us to derive circuit outputs directly from the FFs.'That this is possible

0 I \6 ! \
\
\01 Of 0)
!
i
I
(~
v
would become obvious by comparing Examples 7.5 and 7.6. Accordingly,
the excitation equations are given by

I f \ .,-- II Dl XJ',Q2 + Q;

=

-- -­
,--,.,-,,-- ."~-.
I
..
I
I'
'­

D2 = XiY,Q! + Q2

p~
1 f sLIG /
 7.7 Incompletely Specified Diagrams 247

The resultant circuit for the sequential comparator may now be readily
obtained. The circuit is illustrated in Figure 7.37.

FIGURE 7.37

X;----i o 0,

Ci \
\ £\
, '",
y;--_.-i

o
11 o

I
.
\! :i
o

CK------~-------------~.

7.7 Incompletely All problems considered thus far in this chapter have been completely
specifud; that is, all next-state and output values were completely
Specified Diagrams defined in their state diagrams and state tables. In this section we
shall consider state diagrams and/or state tables that are termed
incompletely specified. Such sequential circuits include don't-care out­
puts in their respective state tables and state diagrams. These cir­
cuits have an added advantage over the completely specified cir­
cuits since the presence of don't-cares may contribute to simpler
Boolean expressions.
The minimization process of state tables that contain don't-cares
is tedious and requires special consideration. Implication tables are
used for removing state redundancies, but the steps involved are dif­
ferent from those for the completely speCified state tables (see Sec­
tion 7.3 for details). The steps for incompletely specified state tables .
involve the following variations:
1. The entries in the implication table are made exactly as
before, but a don't-care in the output is considered to be a
1 or a 0 depending on whether this particular choice aids
the formation of an equivalent group.
/
248 CHAPTER SEVEN Design of Synchronous Sequential Circuits

2. Once the successive passes and crossing out of the cells have
been completed, the designer should make entries in the
equivalence partition table as before. However, it must be
understood that two equivalent pairs like (AB) and (AC)
do not automatically imply the existence of a larger equiv­
alent group (ABC) unless there exists an equivalent group
(BC) or (BCX). This extra condition is necessary because
the don't-cares may have been treated as both 0 and 1
under different conditions.
3. The maximum number of states in the reduced circuit is
equal either to the number of sets of maximal compatibles
or to the number of states in the original circuit, whichever
is less.
4. A closure table is obtained by considering.the maximal com­
patibles as states and grouping their next states under
respective input columns. The reduced state table is con­
structed by renaming the sets of maximal compatibles.
Note, however, that the resulting reduced state table might
still be incompletely specified.

The application of these rules is illustrated in Example 7.10.

EXAMPLE 7.10 SOLUTION

Obtain the reduced state table for FIGURE 7.38

the sequential machine shown in

Figure 7.38.
NS,Z

PS x 0 x 1

A A­, B,1
B G,­ 0,0
C B,1
B,-!

0 A,1
B,­
E C.-f
.
A-
F F-'­ C,+

.
' !

G G,­ G-

The implication table is as obtained as shown in Figure 7.39. The intersec­

tion of A and B is crossed out, which indicates that tf and B are unequal.
This is because when x = 1, the next states result in different outputs.
More boxes are crossed out as repeated passes are made through the table.
To start, the cells with the entry AB are crossed out. At the end of such
search the equivalence partition table is obtained as shown in Figure 7.40.
For better understanding of the operations, consider row B. In row B there
are seven groups-(BG), (BE), (BD), (BG), (GG), (EG), and (DFG)-that
redu,ce to four possible groups of three states: (BEG), (DFG), (BGG), and
-

7.7 Incompletely Specified Diagrams 249

FIGURE 7.39

_.
',-',
E

1~')
!­
" . F
f-)

G BG OG BG CG

A B C o E F

FIGURE 7.4() F (FG)

E (EG)(FG)

o (OG)(OF)(EG)(FG)
o (EG)(OFG) Using step 2
C (CG)(EG)(OFG)
B (BG)(BE)(BO)(BC)(CG)(EG)(OFG)
B (BEG)(OFG)(BCG)(BOG) Using step 2

FIGURE 7.41 A (AO)(AF)(AG)(BEG)(OFG)(BCG)(BOG)

A (AOFG)(BEG)(BCG)(BOG) Using step 2

NS,Z
(BDG). (BG), (BE), and (EG) yield (BEG); (BG), (BG), and (GG) yield
PS x=o x=1
(BGG); and (BG), (BD), and (DFG) yield (BDG). Note that (BG) has been
AOFG AFG,1 BCG,1 used in all three determinations, and such manipulations are valid. In the
BEG CG,­ AOG,O final form we haveJi.ve-possible sets renamed as follows:
-r.~(
BCG BG,1 BOG,O
P = (ADFG)
BOG AG,1 BOG,O
Q = (BEG)
R = (BGG)
FIGURE..7.42
S = (BDG)

NS,Z
The closure table can now be constructed as shown in Figure 7.41. In the
first row of the closure table, AFG are the next states for'states ADFG when
PS x 0 x 1 x = O. For x = 1, BGG are the next states for states ADFG. Now we may
construct the reduced state table by using the variables P, Q, R, and S. The
P P,1 R,1
table can be organized as shown in Figure 7.42. Corresponding to x = 0,
Q R,­ P,O
state R could move to anyone of the three states, Q, R, and S. This is
R QR$,1 $,0
because (BO) is present in all of those three sets of compatibles. Note that
$ P,1 $,0
the reduced table is still incompletely specified.
250 CHAPTER SEVEN Design of Synchronous Sequential Circuits

7.8 Ideal State In the previous sections state assignments were made arbitrarily
with no consideration of the consequences. It will be seen that the
Assignments combinational circuit complexity is different for different sets of
state assignments. The number of possible state assignments for any
given problem is impressive. For n present states and p flip-flops,
there are 2P!/[n!(2P - n)!] ways of selecting n out of the 212 possible
combinations. For each of these ways there are n! permutations of
assigning the n combinations to the n states. Again, for each of these
assignments there are 2P ways of interchanging logic 0 and logic 1
and there are pl ways of interchanging the FFs. Consequently, there
may be a total of[(2P - 1)!]/[(2P - n)!p!] unique assignments. For
example, the number of unique assignments for a nine-state system
can be calculated to be 10,810,800.
The optimum state assignment is one that reduces the amount of
combinational logic of a sequential system when compared to other
assignments. Many different approaches to this state assignment
problem have been developed. The complexity and cost of the cir­
cuit will differ for different combinations of state assignments. The
identification of the best state assignments has been the subject of a
considerable amount of research. We can attempt to locate the best
set by generating those output and excitation tables that allow the
formation of large clusterings of ones. Use of the following guide­
lines will probably result in simpler circuits:
1. Adjacent assignments should be given to those states that
have the same next state for any given input.
2. Two or more states that are the next states of the same
state, under adjacent inputs, should be given adjacent
assignments.
3.. States that have the same output for a given input should
be given adjacent assignments.

The term adjacent assignments means that the states appear next to
each other on the mapped representation of the state table. The
assignment guidelines work best with D and JK FFs. These rules
usually lead to a good, but not necessarily to the optimum, solution.
It may not always be possible to satisfY all of the guidelines at the
same time. In case of conflicts, rule 1 is preferable. An attempt
should be made to satisfY the maximum number of suggested adja­
cencies. However, remember that an ideal state assignment may not
always reduce the cost, and it is true also that the cost of the devices
is often an insignificant part of the overall cost of a digital system.

7.9 Summary
In this chapter all aspects of the design of a synchronous sequential
circuit were considered. It is possible to design numerous types of
digital systems using synchronous sequential design. However, there
 Problems 251

are many digital systems that are of the asynchronous type as well.
We will investigate the nature of both pulse-mode and fundamen­
tal-mode circuits prior to elaborating an additional application of
sequential circuits. Chapter 10 presents a variety of such applica­
tions that include sequential circuits of all three types and some
involving combinations of all three. Next, in Chapter 8, we shall
consider pulse-mode sequential circuits.

Problems 1. Design a three-bit counter that counts up when a control vari­

able E = 0, and counts down when E = 1.
2. Design a four-bit binary up-counter usingJK FFs.
3. Design a synchronous sequential circuit using SR FFs that
results in an output of 1 whenever each the foHowing
sequences occurs:
a. 0001 e. 10010
b. 0101 f. 11011
c. 1101 g. 10011
d. 1011 h. 110110
4. Repeat Problem 3 using JK FFs.
5. Repeat Problem 3 using T FFs.
6. Asswne a two-bit binary counter that counts up when A = 1
and B = 0; counts down when A = a and B = 1; halts when
A = a and B = 0; and is forbidden to operate when A = B
= 1. Obtain the state diagram and the JK equations.
7. Obtain the equivalent Mealy state table from the machine of
Figure 7.14[b].
8. Obtain the equivalent Mealy circuit for the circuit of Figure
7.14[b].
9. Given the state tables of Figure 7.Pl, find the logic equations
and logic diagrams for each table usingJK FFs.

FIGURE7.PI
NS,Z

PS x=O x 1
A A,O C,O
8 0,1 A.O NS,Z
C F,O F,O
0 E,1 8,0 PS x=O ' x= 1

E G,1 G,O A A,O 8,0

F C,O C,O 8 C.O 8,0
G 8,1 H,O C 0,0 8.0
H H,O C,O 0 A,1 8,0

[a] [b]
252 CHAPTER SEVEN Design of Synchronous Sequential Circuits

10. Repeat Problem 9 using SR FFs.

11. Repeat Problem 9 using D FFs.
12. Design a synchronous sequence detector that produces an out­
put of 1 whenever anyone of the sequences 1100, 1010, and
1001 occurs. The circuit resets to its initial state after a 1 has
been generated.
13. Find the Moore equivalent circuits of the two machines given
in Problem 9. UseJKFFs.
14. Analyze the sequential circuit shown in Figure 7.P2. Obtain
the state equations and the state diagram.

FIGURE7.P2
,.­
}~ z
x \

/
) )­
L02 J, ----<

<
0, K, i--­
CK
O2 Jz :-­

<
02 K2

15. Construct the state diagram for the following synchronous

equations:
Dl = XQ2 + Q.l~ + XQIQ.2
D2 = XQ2 + XQ.IQ2
z = Q2' CK

16. Design a BCD counter with (a)JK FFs and (b) D FFs.
17. Design a four-bit Gray code up-counter using (a) JK FFs and
(b) D FFs.
18. Design counters that follow each of the following binary
sequences. For example, "0, 1, 3, 2, 5, 7, 4, and repeat"
implies that the counter repeats the sequence 0,1,3,2,5, 7, 4,
0,1,3,2,5,7,4,0,1,3,2,5,7,4,0,1 and so on.
a. 0, 1,3,2,5, 7, 4, and repeat. Use SR FFs.
b. 0, 1,3, 2,6,4,5, and repeat. Use TFFs.
c. 0,2,4,6, 1, and repeat. UseJKFFs.
 Problems 253

19. A synchronous sequential circuit is shown with its state dia­

gram in Figure 7.P3. Determine the state diagram for the cir­
cuit if the primary and secondary variables are interchanged
as shown in Figure 7.P4.

X X1 Z1 Z
Comb.
logic 1/0
X2 Z2
01"1 0/1

0 1/0
0

o CK

FIGURE7.P4
r-­ X1 Z1 ~

x X2 Z2 Z

'---­ 0 0­

0 < ~CK

20. Obtain the reduced state machine from the state tables shown
in Figure 7.PS. Obtain the corresponding sequential circuits
FIGURE ,7.P5 using D FFs.

NS, Z NS,Z'

x 1 NS,Z
PS x 0 x 1 PS x=O
PS x 0 x=1
A F,O 0,1 A C,O B,1
B C,1 F,1 B C,1 A,1 B,1
A , E,O
C F,1 B,1 C E,O B,1 B F,O 0,1
0 E,1 G,1 0 F,O A,O C E,O B,1
E A,1 0,1 E A,O G,1 0 F,O B,O
F G,O B,1 F 0,1 C,1 E C,O F,1
G A,O 0,1 G E,1 C,1 F B,O C,O

[a] [b] [cl

254 CHAPTER SEVEN Design of Synchronous Sequential Circuits

21. Repeat Problem 20 usingJKFFs.

22. Repeat Problem 20 using SR FFs.

Almaini, A E. A "Sequential machine implementations using universal

Suggested Readings
logic modules." IEEE Trans. Compo voL C-27 (1978): 951.
Armstrong, D. B. "A programmed algorithm for assigning codes to sequen­
tial machines." IRE Trans. Elect. Compo vol. EC-ll (1962): 466.
Dunworth, A., and Hartog; H. V. "An efficient state minimization algo­
rithm for some special classes of incompletely specified sequential
machines." IEEE Trans. Compo vol. C-28 (1979): 531.
Fojo,]. M., and Barreiro, F. D. "A method for reducing the number of
input variables to synchronous sequential circuits." IEEE Trans. Compo
voL C-24 (1975): 1029.
.Hartmanis,]. "On the state assignment problem for sequential machines
I." IRE Trans. Elect. Compo vol. EC-1O (1961): 157.
joseph, ]. "On easily diagnosable sequential machines." IEEE Trans.
Compo vol. C-27 (1978): 159.
. McCluskey, E. ]., and Unger, S. H. "A note on the number of internal
variable assignments for sequential switching circuits." IRE Trans. Elect.
Compo voL EC-8 (1959): 439.
Nagle, H. T.,jr.; Carroll, B. D.; and Irwin,j. D. An Introduction to Computer
Logic. Englewood Cliffs, N.].: Prentice-Hall, 1975.
Nanda, N. K., and Bennetts, R. G. "Reconvergence phenomenon in syn­
chronous sequential circuits." Elect. Lett. vol. 16 (1980): 303.
Noe, P. S., and Rhyne, V. T. "Optimum state assignment for D flip-flop."
IEEE Trans. Compo vol. C-25 (1976): 306.
Pattavina, A, and Trigila, S. "Combined use of finite-state machines and
Petri nets for modelling communicating processes." Elect. Lett. vol. 20
(1984): 915.
Paull, M. C., and Unger, S. H. "Minimizing the number of states in
. incompletely .specified sequential switching functions." IEEE Trans.
Elect. Compo voL EC-8 (1959): 356.
Rao, C. V. S., and Biswas, N. N. "Minimization of incompletely specified
. sequential machines." IEEE Trans. Comp. vol. C-24 (1975): 1089.
Stearns, R. E., and Hartmanis,]. "On· the state assignment problems for
sequential machines II." IRE Trans. Elect. Compo voL EC-1O (1961): 593.
Unger, S. H. "Self-synchronizing circuits and nonfundamental mode oper­
ation." IEEE Trans. Compo voL C-26 (1977): 278.
Waxman, j., and Rootenberg, ]. "Logic circuit for cyclic detection in a
state diagram." IEEE Trans. Compo vol. C-26 (1977): 303.
Yamamoto, M. "A method for minimizing incompletely specified sequen­
tial machines." IEEE Trans. Compo vol. C-29 (1980): 732.
Yang, C. C. "Closure partition method for minimizing incomplete sequen­
tial machines." IEEE Trans. Compo vol. C-22 (1973): 1109.
Introduction to CHAPTER
TElV
Counters, R~gisters, and
Register Transfer Language

10.1 Introduction
With the study of flip-flops and sequential circuits behind us,the
study of counters and registers will be a natural and straigntfurward
extension. Counters and registers are essential to the design of
advanced circuits found in digital computers. Counters are employed
to keep track of a sequence of events, and registers are used to store
and manipulate data that contribute to all or many of these events.
In other words, most of the robust digital systems would have two
major functional units: a unit where the manipulations are con­
ducted and a unit that is used for regulating the events (if the first
unit. Registers and associated logic subunits help to make the first
unit, and counters could be used for running the second unit.
Therefore, without an understanding of flip-flops, counters, and
registers, design of digital systems would be impossible.
Counters are particularly common in the control and arithmetic
units of processors, where they are used to keep track of the
sequence of instructions in a program, to distribute the sequence of
timing signals, for frequency division for causing time delays, for
counting, and a host of other similar operations. Counters may
count in binary or in nonbinary fashion. They are commercially
available in a large variety of mediuin-scale integrated devices. The
basic operational characteristic of a counter is sequential; for every.
present state there is a well-defined next state. The design of a
counter involves designing combinational logic that decodes the
present state and enables entry into the next state of~e counting
sequence. Counters are generally classified into two groups: syn­
chronous and asynchronous. A synchronous counter qas all FFs 'change
state synchronously with the clock input whether a periodic clock or
an aperiodic pulse occurs. An asynchronous (or ripple) counter is made
up of FFs that do not change simultaneously with the clock input.
Another application for FFs is for storing bits of information.
When FFs are configured to store multi-bit information, they are

296
 10.2 Synchronous Binary Counters 297

referred to as registers. Registers are classified according to the way

information bits are stored and retrieved. If data are stored and
removed at either end of a multi-bit register, one bit at a time, the
register is referred to as a serial or shift register. However, if all bits of
the word are stored or retrieved simultaneously, the register is
referred to as a parallel register.
Another area that needs to be investigated is how a digital sys­
tem, however small it may be, is built, integrated, and run. The
control unit of a system can be designed using the methods that
were developed in the last three chapters. However, these sequential
design techniques are inadequate for the representation of subsys­
'-~----- terns that are used strictly in the manipulation of data. The tool
that has been found useful in accomplishing this representation is
known as the register transfer language (RTL). RTL helps to translate
a specification mechanically into its hardware realization.
The beginning of this chapter is devoted to the development and
study of various counters and registers. This discussion is then grad­
ually expanded to include the basics of RTL. Finally, RTL is used
in the design of complete digital systems. After studying this chap­
ter, you should be able to:
o Design and analyze both synchronous and asynchro­
nous counters;
o Design and analyze serial, parallel, and hybrid
registers;
o Design and analyze systems that have counters and
registers;
o Translate complex operations into equivalent RTL
sequences;
o Use RTL in the design of data and control units.
--~

10.2 Synchronous
Synchronous counters are distinguished from asynchronous (or rip­
pie) counters in that the clock pulses in synchronous count~rs initi­
Binary Cpunters
ate changes in the FFs used in the counter. The simplest possible
counter is a single-bit counter that alternates between two states, 0
and 1. A toggle FF using a singleJK FF, with both inputs tied to 1
(J = K = 1), will function as a single-bivcounter alternating
between the two states with the occurrence of each clock. The out­
put of the FF has a frequency that is one-half the clock frequency.
A two-bit binary up-counter with four states was already
designed in Example 7.5. Such a counter consists of two JK FFs
whose states Q2QI could be assumed to move in sequence through
00,01, 10, 11,00,01, and so on. The correspondingJ and K inputs
of the two FFs are given by

JI = KI = 1

J2 = K2 = Ql

298 CHAPTER TEN Introduction to Counters, Registers, and R TL

Note that these equations are slightly different from those given in
Example 7.5. The positions of the FFs have been reversed and out­
put equations are abandoned altogether. We can take the outputs
directly from the FFs.
We shall now attempt to synthesize a three-bit binary up-counter
of the nonterminal type. With every clock input the counter moves
to the next higher state. Consider the FF outputs to correspond to
the present state. The first step in the design sequence of a sequen­
tial circuit is to begin with a state diagram and a state table fol­
lowed by the assignment of states. The state diagram, the state
table, and excitation maps of a three-bit counter are shown in Fig­
ure 10.1.
The excitation maps of Figure lO.1[c] may be used to obtain the
i K equations as follows:
il = Kl = 1
i2 = K2 =Ql

i3 = K3 = QIQ2

The resulting circuit diagram is shown in Figure 10.2.

We learned in Chapter 5 that many of the complex combina­
tional designs are realized using heuristic techniques. This simplifi­
cation is more true when modularity is evident in the system. Quite
similarly, many sequential design problems can be accomplished
without going beyond state tables. A close examination of the state
table of Figure lO.1[b] reveals the presence of a certain degree of
regularity. Note that QI changes with every clock pulse and, in gen­
eral, Qi changes state if all less significant bits are 1. Similar conclu­
sions about the regularity of counter design can be made by inspect­
ing the respective i K equations of the three counters that we have
considered thus far. An inspection of the i and K equations leads us
to the conclusion that, based on a regular pattern, these equations
can be extended to i and K equations for the nth bit of a multi-bit
up-counter as follows:

in = Qn-lQn-2' .. Q3Q2Ql = Qn-"Jn-l

~ = Qn-lQn-2'" Q3Q2Ql = Qn-1Kn- 1

\
There are two different ways of connecting the inputs to succes­
sive FFs based on the two forms of the equation for the nth term.
Both are illustrated in Figure 10.3. Part [a] of the figure shows a
configuration where the FF outputs are combined in parallel. The
propagation delay at the input of each FF is the same for illl stages.
However, the fan-in to the AND gate and the fan-out of each FF
increase as the number of counter stages is increased. Figure lO.3[b]
shows an equivalent configuration using the second form of the in
and ~ equations. The fan-in of the AND gates is always two; how­
 10.2 Synchronous Binary Counters 299

FIGURE 10.1 Three-Bit Binary

Up-Counter: [a] State Diagram,
[b] State Table, and [c] Excitation
Maps.

PS
0 30 20 1 NS
000 001
001 010
010 011
n on' 100
100 101
101 110
110 111
111 ~o\
; I \
I \
[a] [b]
(')

ex" ""'v I

'nil ' 0 f r r I}
1 - 1 0 1 - - 0 0 (1 ! 0

1 - - 1 0 1 - - - -
'l=J -

- 1 1 - - - 1 oI r-~"
1,

- 1, 1 - - - 1 oI o 0 1
L~~.
0

[c]

FIGURE 10.2 Logic Circuit of a

Three-Bit Up-Counter.

CLOCK--4---------~--------------~

f
,/
 300 CHAYI'ER TEN Introduction to Counters, Registers, and RlL

FIGURE 10.3 Configurations for

the] and K Inputs of an n-Bit
Counter: [aJ Parallel and [bJ
Serial.

On-l
[a]

ever, the propagation delay to the nth FF increases as the number of

counter stages is increased. Note that the first method allows faster
clocking and 'counting but results in fan-in and fan-out problems for
large counters.
Figure 10.4 shows a synchronous n-bit binary up-counter using
JK FFs and two-input AND gates. Two control signals are added in
this circuit, CLEAR and COUNT. A high on the CLEAR input
resets the counter; the counter remains reset until the CLEAR sig­
nal is withdrawn. The COUNT signal is used to disable the clock
pulses. The designer can use this to block the clock input and hold
FIGURE 10.4 n-nit Synchronous any nonzero count state. If the preset (PR) inputs are effectively
Up-Counter.

On-2 ·On-3
On-l

CLEAR KnCLROn

~------- ... --~--~

COUNT
CLOCK
~~--------------~~-------------- ... --------~
10.2 Synchronous Binary Counters 301

used, one may even be able to set the counter to its maximum count
state.
There are many occasions in a digital system when a down­
counter is required. A binary number is set into the counter that
then counts toward zero as the clock pluses occur. These counters
can be designed in the same way as up-counters. The equations for
the down-counters are also seen to have regularities. The i and K
equations for an n-bit binary down-counter are obtained as follows:

il = Kl = 1

i2 = K2 = Ql

i3 = K3 = QlQ2 = i2Q2 = K2Q2

By comparing these equations with those of the up-counters, we

note that the resulting circuit is similar in nature. The in and Kr.
inputs are taken from the two-input AND gate, whose inputs are
Qn-l and Kn- l. Note in the case of the up-counters the correspond­
ing AND inputs were Qn-l andin-l'
For some of the applications a counter may be required to count
up or down. One such application could be a counter device that
keeps track of total cars inside a parking garage. As each car enters
the garage, the counter counts up; as each car leaves, the counter
counts down. This is a more complex design than that of all coun­
ters considered so far since it requires at least one control signal, E,
to determine the direction of the count. We may assume that when
E = 1 the circuit counts up, and when E = 0 the circuit counts
down. One way to synthesize such a circuit would be to follow the
standard steps for the design of sequential circuits (see Chapter 7,
Problem 6, for example). However, we can combine the equations
for up- and down-counters to derive the respective i and K equa­
tions of an n-bit, up-down counter as follows:

il = Kl = 1
i2 = K2 = EQI + EQI
i3 = K3 = EQIQ2 + EQl~

The excitation function of the up-counter is ANDed with E, and

that of the down-counter is ANDed with E, and, finally, the two
corresponding composite functions are ORed to obtain the i and K
equations. Consequendy, when E = 1 these equations reduce to
those of an up-counter, and when E = 0 the equations reduce to
302 CHAPTER TEN Introduction to Counters, Registers, and RTL

FIGURE 10.5 Four-Bit Binary those of a down-counter. The implementation of a four-bit, up­
Up-Down Counter. down counter is shown in Figure 10.5.

E-...---,

CLEAR

CLOCK
As pointed out earlier, a counter can be designed to count in a
nonbinary manner as well. Two examples of nonbinary counters
are a BCD decade counter and a Gray code counter. In the former,
the counter counts 0000 through 1001 and then resets back to 0000.
A four-bit Gray code counter, on the other hand, counts 0000, 0001,
0011,0010,0110,0111,0101,0100, 1100, 1101, 1111, 1110, lOlD,
1011, 1001, and 1000 in that order and then resets to 0000 before
resuming count-up operation again. Example 10.1 illustrates the
design of a BCD decade counter.

EXAMPLE 10.1 SOLUTION

Obtain the J and K equations for a The design steps, followed in the usual sequence, consist of the state dia­
BCD up-counter. gram (Figure 10.6), the state and transition table (Figure 10.7), and excita-

FIGURE 10.6

n
 10.2 Synchronous Binary Counters 303

FIGURE 10.7 PS I

0 40 3 0 2 0, NS J4 K4 J3 K3 J 2 K2 J 1 K,

0000 0001 0­ 0­ 0­ 1
0001 0010 0­ 0­ 1­ -1
0010 0011 0­ 0­ 0 1­
0011 0100 0­ 1­ -1 -1
0100 0101 0­ -0 0­ 1­
0101 0110 0­ -0 1 -1
0110 0111 0­ -0 0 1­
0111 1000 1­ -1 -1 -1
1000 1001 -0 0­ 0­ 1­
1001 0000 -1 0­ 0­ -1

,.-,,~.

tion maps (Figure 10.8). The resulting J and K equations are obtained
from Figure 10.8 as follows:

Jl = Kl = 1
J2 = QlQ4
K2 = Ql
Js = Ks = Q,Q2
J4 = Q,Q2Q3
K4 = Q,

FIGURE 10.8
~(i)
O2 '" <- \
~6;~~~.,? .', ,';
. ! "

0 0 0 0 0 0 1 0 0 1 - - c./J 1 ~p - 1

0 0 1 0 - - - - 0 1 - - .' ~
- - - -
- - - - - - - - - - - - - - - -
- - - - 0 0 -- - 0 0 - - 1 1 - - L

- - - - - - 1 0
0 0 1 0 - - 1 0 1

- - - - - - - - 1

- - - - - - - -
304 CHAPTER TEN Introduction to Counters, Registers, and RTL

Steps similar to those in Example 10.1 can be used to determine

the circuit of a BCD down-counter. Such an exercise gives the1 and
K equations for a BCD down-counter as follows:

1. = K. = 1

12 = Q.(Q3 + Q4)

K2 = Q.

13 Q.Q4

=
K3 = Q.Q2

14 = K4 = Q.Q2Q3

The design of the BCD counter considered in Example 10.1 is a

classical one. It is not necessary for a designer to go through all of
these cumbersome steps if an already working circuit can be modi­
fied to suit the requirement of the new design. For example, con­
sider a four-bit binary up-counter, also known as the modulo-16 (or
divide-by-16) counter. A synchronous BCD counter works exactly
as the already designed four-bit, modulo-16 counter until state 9
(Q4Q3Q2Ql = 1001) is reached. The four-bit counter advances to
1010, while the BCD counter starts all over at 0000. Assuming that
a four-bit counter is to be modified to a BCD counter, it is necessary
to reset FFs when they are about ready to go to 1010 otherwise. The
modification process requires circuit changes to satisfy the following
conditions when the four-bit up-counter reaches count 1001:

Qll the LSB, should become a 0,

Q4, the MSB, should become a 0,
Q2 should be prevented from becoming a 1.

An inspection of the circuit of Figure 10.4 reveals that the least

significant FF is always in a toggle mode, which implies that Q; will
change to a 0 by itself. Note also that the MSB, Q4, of the four-bit
counter changes only when the count reaches either 0111 or 1111.
This change happens because the inputs 14 and Kt are both held to
a 0 during the other counts. Consequently, it is necessary to supply
Kt with a 1 instead of a 0 when the count reaches 1001. This can be
accomplished if Ql is fed directly to the K4 inpl1t and 14 is discon­
nected from the 14 input. The next consideration is to prevent Q2
from switching back to 1 once the count reaches 1001. This action is
accomplished by supplying12 and K2 with the ANDed output of Ql
and Q4' This modification does not cause any problem because Q4 is
a 1 until the count reaches 1000. And once the count reaches 1000,
Q2 need not be turned on in a BCD counter. In summary,
 10.2 Synchronous Binary Counters 305

Ql should be fed to K4,

K4and14 should not be connected,
QIQ4 should be fed to12 and K2.

FIGURE 10.9 Synchronous BCD Consequently, the BCD counter may be obtained by modifying the
Up-Counter. modulo-16 counter circuit, as shown in Figure 10.9.

l l
1-4~ J 1 PR 0 1 ) J 2 PR O2 t - ) l J3 PR 0 3 I - ~J' JR O,
-< >­ r- iC~ r--< > rC ~
-
r­ K1CLR Q1 '-K2CLR02 -K3CLR03 r-- t ­
K4CLR04 ' ­
\) \

CLOCK ­
COUNT ­

CLEAR

Synchronous circuits requiring irregular count sequences, such as

the Gray counter mentioned earlier, are usually designed using the
familiar step-by-step technique. A count sequence is said to be irreg­
ular if it is not magnitude ordered. In general, it is harder to come
up with an irregular counter by modifying a regular counter.
Examples 10.2 and 10.3 relate to counters with irregular count
sequences.

EXAMPLE 10.2 SOLUTION

Obtain the state table for the FIGURE 10.10

counter ~hgwn in Figure 10.10,

starting from the count 000. The

MSB of the count is at Q3'

CLOCK~----~------------~~--r-----~
306 CHAPTER TEN Introduction to Counters, Registers, and RTL

This circuit can be analyzed by assuming a present state and using our
knowledge ofJK FF operation. Corresponding to each of the present states,
the next state is found by determining the corresponding J and K values of
each FF. WhenJK = 00, the corresponding Q remains unchanged; when
JK = 01, Qis reset; whenJK = 10, Qisset; and whenJK = 11, Qtoggles.
The resulting state table is obtained as shown in Figure 10.11. The circuit
has an irregular state sequence 0, 1,3,4,6, and repeat. Consequently this
circuit is an irregular counter.

FIGURE 10.11

PS I J3 K3 J2 K2 J 1 K1 NS

000 00 01 10 001
001 00 11 10 011
011 11 11 11 100
100 00 11 00 110
110 11 11 01 000
I

EXAMPLE 10.3 SOLUTION

Obtain a synchronous counter that The state table and the JK excitations are obtained as shown in Figure
produces the count sequence 0, 2, 4, 10.12. The easiest way to solve this problem would be to consider the pre­
3, 6, 7, 0, .... sent and corresponding next states. The JK excitations necessary for the
corresponding transitions are then determined for each of the cases. Note
that states 1 and 5 don't occur in the count sequence. Consequently, we
assume the corresponding next states and the JK excitations to be don't­
cares. This assumption should help in the minimization step. The J and K
Boolean equations are obtained using K-maps. They are as follows:

FIGURE 10.12 !
PS NS
I I
0 3 0 2 0 1 0 30 2 0 1 ! J3 K3 J2 K2 J 1K1

000 010 0- 1- 0­
001 -- -- -- --
010 100 1- -1 0­
".

011. '1.1.0. .J- -.0, -~.

100 011 -1 1- 1­
101 --- -- -- --
110 111 I -0 -0 1­
111 000
1 -1 -1 -1
 to. 3 Asynchronous Binary Counters 307

1, = Q3
K, = 1
12 = 1
K2 = Ql EB Q3
13 = Q2
K3 = QI + Q2

The resulting counter is obtained as shown in Figure 10.13

FIGURE 10.13

CLOCK--~~--+-------------~---+----------~

CLEAR

It is appropriate at this point to offer a word of caution. Since

there is are various cascadable multi-mode binary and nonbinary
counters available, a designer might be tempted to use them in his
or her design as a starting chip. Before doing so, the designer must
thoroughly examine the circuit specifications and diagrams. If JK
FFs are used, the familiar 1's-catching problem (see Section 6.5 for
details) may crop up here. More specifically, if the clock input is not
at the proper level, erroneous operation might result if a particular
mode is changed. There may be times, therefore, when a counter
may have to be designed from scratch.

10.3:t\synchronous
All of the counters considered thus far employed synchronous cir­
cuits, that is, the FF actions were synchronized with the clock pulse.
Binary Counters .
The advantage of a synchronous counter is that tall bits of a count
change simultaneously except for slight differences in FF delays. If a
specific count is being decoded, all bits are' available at the same
time, eliminating momentary errors at the decode output. We shall
now introduce asynchronous counters, also known as ripple coun­
ters. The FF clock inputs in a ripple counter are not tied together.
308 CHAPTER TEN Introduction to Counters, Registers, and RTL

In fact, the clock inputs are cascaded from output to input (almost
like the rippling of carries in a ripple adder). They are used to
reduce the amount of control logic required to construct a binary
counting sequence. Asynchronous counters have limitations but also
provide less expensive counter options for those cases where their
limitations will not affect the circuit. The AND gates between FFs
in the synchronous binary counter design may be eliminated by
observing the counter state table. The LSB needs to be changed in
every present-state to next-state transition. In all bit locations, Qi
changes each time 0-1 makes ~ transition from a 1 to a O.
A negative edge-triggeredJK FF in a toggle mode changes state
each time the signal connected to the clock input makes a 1 ~ 0
transition. An asynchronous counter using T-configuredJK FFs has
its least significant FF activated by the system clock. The 1 ~ 0
transitions of that FF may be used as the trigger (clock input) signal
for the next most significant FF. This process of triggering is contin­
FIGURE 10.14 Four-Bit Ripple ued for as many bits as desired.
Counter: [a] Circuit and [b] The logic circuit of a four-bit ripple counter is shown in Figure
Timing Diagrams. 1O.14[a], where four T FFs are cascaded together. The output of

SET----I

CLEAR

[a]

01

O2
0 01 1 I 0 01 1 .I 0 01 1 I 0 01 1L
03
0 0 0 01 1 I 0 0 0 0/ 1L

0 0 0 0 0 0 0 01 1L
04
[b]

 10.3 Asynchronous Binary Counters 309

each FF provides the clock signal for the next FF. The timing dia­
gram without delays for the four-bit ripple counter is shown in Fig­
ure 1O.14[b]. Examination of the timing diagrams shows that the
frequency of the Q4 pulse is one-sixteenth of the frequency of the
input pulse x. Each stage of the counter divides the frequency of the
preceding stage by two.
Counters like the one shown in Figure 10.14[ a] are simple in con­
cept, but have at least two disadvantages: a forced regular binary
sequence and speed. The first disadvantage is not so much of a seri­
ous handicap, but the speed is. In reality, the rippling effect
through the FFs causes delay between each count that is propor­
tional to the number of FFs in the counting chain. Consider the tim­
ing diagram of Figure 10.15 that shows the situation existing in the
.: .­ counter when the count is 1111. Ql does not change to 0 coincident
with the trailing edge of the sixteenth x-pulse until time tp the prop­
agation delay of each FF. The same is generally true for the syn­
chronous counter, but for the asynchronous counter Q2, Q3, and Q4
change at times 2tjJ 3~, and 4~, respectively, 'Deyond the negative
edge of the sixteenth x-pulse. For an n-bit ripple counter to reach a
valid count before the next clock pulse, T > n~, where T is the
period of the input pulse. If the final count is all that is of interest,
the condition T > ~ is all that must be met. In this situation,
changes in the LSB are constantly rippling to higher-order bits.
After the last pulse is input, it will be n~ before the final count can
be correctly read. Note that in Figure 10.15 the counter does not go
through the transition 1111 --? 0000. Instead the counter passes
through the state transition sequence 1111 --? 1110 --? 1100 --? 1000
--? 0000. These transitions occur in rapid succession but they result
in undesired transient conditions that might cause further problems

FIGURE 10.15 Timing

Consequences of a Four-Bit X--...J
Asynchronous Counter During
tllt --1> 0000 Transition. 01--------1---..

I
02-------L-~1~
I
I
I
I
03--------I-~1~-.
I
I
I I
04-------~~1--~~1-.
I I
I I
, I
I I I I
I t + 2f( I f + 4tf
I I
t+ If t +3f(
310 CHAPTER TEN Introduction to Counters, Registers, and RlL

in a circuit that is being driven by such a counter. This must be

borne in mind when such counters are used in generating a control­
ler of a complex system.
If the circuit of Figure 1O.14[a] is modified so that Qll (2) and
Q3, are used respectively as the clock inputs for the second, third,
and fourth T FFs, then the circuit will operate as a four-bit ripple
down-counter. The ripple counter of Figure 1O.14[a] may also be
modified to generate a resettable counter. Say, for example, we are
interested in generating only BCD counts. Such a counter is shown
in Figure 10.16. The circuit operates as a modulo-16 ripple counter,
but when the state 1010 is reached the counter resets immediately.
This resetting is accomplished by means of a combinational decod­
ing scheme that is tied with the reset inputs of every FF. The
FIGURE 10.16 Asynchronous
counter stays at the count of 1010 for the decoding gate delay plus
BCD Up-Counter: [a] Circuit and the reset input-to-output delay.

[6] Timing Diagram. The ripple counter of Figure 10.16 exhibits other transient

COUNT----.---------.---------+---------,

CLOCK - - - + 0

CLEAR
. raj

CLOCK

I
I I I I I
01 II I II I I
II II II II
"
O2
::,
II
::1
II
::,
II
::1
II
I:~
II
III
'"
!II
III
'"
III '"
III
III
III
1111
1111
III I
1111 1111

03
III III III III 1111
III Jill III 1111 III I
III 1111 III 1111 1111

04
111
111
1111
1111
III
III
11111
III! lit
III
III 1111 III lUll 1111
III 1111 III 11111 1111
III 1111 III 11111 III I
III 1111 III 11111 1111
{bJ
»

to.3 Asynchronous Binary Counters 311

behavior that needs to be closely examined. The transient behavior

of the counter is listed as follows:
0001-') 0010 (ideal), 0001-') 0000-') 0010 (actual),
0011-') 0100 (ideal), 0011-') 0010-') 0000 -') 0100 (actual),
0101-') 0110 (ideal), 0101-') 0100-') 0110 (actual),
0111-') l000{ideal),OI11-')OllO-')Ol00 -') 0000 -') 1000
. '':''.­

(actual),
1001-') 0000 (ideal), 1001-') 1000-') 1010-') 0000 (actual).

The worst-case transient occurs during the transition 0111 -') 1000.
Since an intermediate count is to be decoded, the clock period must
be longer than the 0111 -') 1000 propagation delay.
Another alternative to the BCD counter design involves feeding
the clock input of a modulo-5 counter with the output of a single T
FF. The combination of a modulo-2 and modulo-5 counter results
in a modulo-l0 counter. Such cascading of one counter with
another should be pursued whenever possible. Some of the counter
designs that we have considered thus far have demonstrated how to
reset a counter once a specific count has been reached. There are
cases where a different approach may be necessary. It is always pos­
sible to preset a counter to a specific count by means of a Jam-entry
scheme, as shown in Figure 10.17. The bit that is to replace the old
value at the Qi location is fed directly to the corresponding entry
point, Xi' The new bit is transferred to the FF output when the clock
pulse occurs. The counter will begin counting from the preset count

FIGURE 10.17 Jam-Entry

Scheme 'for Presetting the ith FF.

FF;

CLOCK
input -.._ _- - J
312 CHAPTER TEN Introduction to Counters, Registers, and R1L

as further clock pulses arrive at the input. This is possible only in

counters made up of independent FFs.
Because of the problems of asynchronous counters, they should
be used with a certain degree of caution. The designer must be
aware of their limitations. The next section introduces the charac­
teristics of IC counters. In a subsequent section (Section 10.9), more
counter configurations will be introduced that exhibit several
advantages over the synchronous and asynchronous counters.

10.4 I C Counters In the first three sections of this chapter we have used both classical
and heuristic design techiques to design counters. Similar multi-bit
counters are available in IC form. A typical four-bit binary counter
module with inputs and outputs is shown in Figure 10.18. We are
.already familiar with most of its features. The different inputs are
described as follows:
A, B, C, D: These inputs are used for presetting the counter to an
irutial value. It has been assumed that A is the LSB and D is the
MSB.
QA, QB, Qc, QD: These are the FF outputs of the counter. QA
corresponds to the LSB and QD corresponds to the MSB.
CARRY-OUT (CO): This output becomes a 1 when the count
equals 1111 and the ENABLE control input is a 1. CARRY­
OUT is equivalent to QAQsQcQD • ENABLE.
LOAD: This control input is used to load A, B, C, and D values
into the counter. When -the LOAD input is a low (0), the values
are either loaded immediately if loading is asynchronous or
loaded at the next clock pulse in the synchronously operated
counter.
CLEAR (CLR): This control input when set to a 0 causes the
counter to be cleared. The counter is cleared immediately if it
has an asynchronous clear. In a counter with synchronous clear
the output changes coincident with the next clock pulse.
ENABLE (E): This input must be high for the counter to count.
CLOCK (CK): The 1 ~ 0 transitions of this input are counted
by the counter when the ENABLE input is a 1.

FIGURE 10.18 Four-Bit Binary

Counter Module.

CARRY­

OUT

 lOA I C Counters 313

The four-bit IC counters may be cascaded together, as shown in

Figure 10.19, to form counters of any bit length. When counter # 1
has a count of 1111 and its ENABLE is a I, the CARRY-OUT is a
1, which activates the second counter. When the next clock occurs,
counter # 1 resets to 0000 and counter #2 counts up to 0001 and
the CARRY-OUT of counter # 1 is reset to a O. The eight-bit
counter output then becomes 0001 0000. During the next 15 pulses,
~ .- - ­ counter # 2 would remain qisabled and counter # 1 would keep on
counting upward. Again when counter # 1 reaches count 1111,
counter #2 counts up to 0010. And this process continues as more
FIGURE 10.19 Eight-Bit IC inputs appear at counter # 1. Each time the setup of Figure 10.19
Counter.. receives a total of 16 clock pulses, the # 2 counter counts up by 1.

-..._+<1 LOAD
--e-aCLR CARRY­ CARRY­
#1
-----iE OUT OUT

Input

The LOAD and CLR controls can be manipulated in many dif­

ferent ways to obtain many count variations. Example lOA illus,,:
trates such manipulations of a binary IC counter mode.

EXAMPlJE·l0.4 SOLUTION
Design a counter using the module One of the ways to accomplish this design is to make use of the CARRY­
of Figure 10.18 that outputs a 1 OUT output, which is 1 when the count reaches 1111. We can load the
each time six counts have been counter With an ipi~al value that will cause an output to occur five pulses
received. later. The CARRY-OUT is then used to reload the initial value. We begin
from 1010, and when the counter reaches 1111 the:: CARRY-OUT would
give a 1. Note that the first output upon turning the power on may not
even be 1010. The CARRY-OUT either may become 1 before six pulses
have occurred or may require up to 15 pUlses. The number of pulses is
dependent on what value the counter assumes upon power-up. The state
diagram of the required sequence js shown in Figure 10.20.
314 CHAPTER TEN Introduction to Counters, Registers, and RTL

FIGURE 1020

Successive outputs should occur after every six pulses. The circuit con­
figuration using an IC counter with a synchronous load is obtained as
shown in Figure 10.21. When the counter reaches 1111 the CARRY-OUT
becomes a 1, causing a low at the LOAD input. This low forces the counter
to begin again from the 1010 state.

FIGURE 1021

CARRY­
OUT
Input
'----+IloZ

An alternate solution to this problem may be obtained by making use of

the CLR input. In this .case we begin from 0000, and once the count
reaches 0101 an output is made to Occur, and at the same time the counter
is reset. A count of 0101 is decoded using external logic, and the counter is
. then made to reset using the synchronous CLR. The flfStpulse may require
as many. as seven pulses or none, depending on what the counter state is
after power-up. In the worSt case the beginning state could be 0110 and
the counter would be reset once the count reaches 1101. The state diagram
and the corresponding circuit configuration are obtained as shown in Fig­
ure 10.22. Note that the output z will be a 1 whenever the count is 0101 or
1101. The count will be 1101 only if the counter shows a coUnt larger than
0101 at power-up.

FIGURE 1022[a]

[a]
 to.5 Basic Serial Shift Registers 315

FIGURE 1022[6]

----(1 LOAD

CARRY­
OUT
Input

Z--.....-t

[b]

Given a modulo-m counter to count n pulses, where n < m, and then to

restart the count, it is thus advisable to follow either of the following
schemes:

a. Use the CARRY-OUT and load m- n into the preset inputs. The
CARRY-OUT will be 1 every n pulses.
b. Decode a count of n - 1 and use the output of the decode gate to
clear the counter. The decode gate will be active every n clock
pulses.

10.5 Basic Serial Shift The shift register is one of the most extensively used functional
devices in digital systems. A shift register consists of a group of FFs
Registers connected so that each FF transfers its bit of infonnation to the
adjacent FF coincident with with each clock pulse. In other words,
shift registers store bits of infonnation, behaving like temporary
memory, and upon external command shift those infonnation bits
one position to either right or left, depending on the design of the
device.
The action of a right-shift register whose shift-right serial input is
tied to a 1 is illustrated in Figure 10.23. The bits shifting out of the
right-most FF are lost. With each clock input the bits move one
position to the right while a 1 moves in at the MSB. After 11 clock
pulses, the data in the register prior to shiftiIl:g are replaced by a
string of Is. A quite useful application of shift-right registers
requires a connection of the right-most FF output to the input of the
left-most FF. In that case the LSB is not lost b.ut appears at the
MSB. Consequently, after two clock pulses, for example,
00101100101 would be replaced by 01001011001. Such a shift-right
register is known as a circulate-right regiSter. In the event these same
data were stored in a shift-left register and the MSB output was
connected to the LSB input, the data would be 10110010100 after
two clock pulses. This latter type of register is known as a circulate­
left register.
 316 CHAPTER TEN Introduction to Counters, Registers, and R'IT.

FIGURE 10.23 Shift-Right After Clock Bit Pattern

Register Action.
0 00101100101
1 10010110010
2 11001011001
3 11100101100
4 11110010110
5 11111001011

10 11111111110
11 11111111111

A typical stage, il, in a multi-bit serial shift-right register can be

designed using the design procedures described in the earlier chap­
ters. Figure 10.24 shows the state diagram of the FF concerned. The
FF state reflects the current content of that position of the shift regis­
ter. If the current state of the nth bit of the shift register is a 0 and
that of the (n + l)th bit is a 0, the nth bit remains unchanged when
) the clock occurs. Similarly, Qn does not change if the present states
of both Qn and Qn+l are 1. For the other cases, Qn changes and
takes the value of Qn+l' Figures 10.23 and 10.24 illustrate the
action of serial shift-right registers and make it obvious that the
function of each of the FFs is governed by the same next-state equa­
tiorts. This observation reduces the design of a multiple-bit shift reg­
ister to the design of a single stage. For n bits, nsuch stages are cas­
caded.
The circuit action described by thestate diagram in Figure 10.24
is that of a D FF. This can be verified by comparing the state dia­
gram of Figure 10.24 with that of Figure 6.40[c]. Multi-bit, shift­
right registers can be built using edge-triggered D FFs or JK FFs
connected as D FFs. Such shift-right registers are shown in Figure
10.25 where SRI is the entry poiIit for shift-right input and SRO is
the exit point for shift-right output. While using D FFs~ the Qoutput
of each stage is connected to the D input of the succeeding stage. If
JK FFs are used, the Qand Qoutputs of each stage are connected to

FIGURE 10.24 State Diagram

for the Qn Bit of a Multi.Bit, Shift·
Right Register. Qn +l

Qn +l
 10.5 Basic Serial Shift Registers 317

FIGURE 10.25 Four-Bit, Shift­

Right Register: [a] Using D FFs
and [b] UsingJK FFs.

SET

SRI - - - - - - - - 1 0 3 PR 031----1

HOLD
CLOCK----;
CLROo SRO

CLEAR---!

[a]

SET

S R I - - - - -.......----iJ3 PR 031----IJ2 PR 021----IJ 1 PR 0 11----1

HOLD
CLOCK----;
L-_-"

CLEAR

the J and Kinputs of the next stage. The first JK FF is modified to

that of a D FF by supplying the data directly to the J input and the
complement of data to the corresponding K input. Note that the
remaining FFs do not have an inverter between their J and K inputs
.since each of the J inputs is already a complement of the corre­
sponding K input. A high at the CLR input would reset the register
FFs, and similarly a high at the SET input would place a 1 at each
of the FFs. A high at the HOLD input would disable the FFs and
this could be used for storing the bits indefinitely. SET, CLR, and
HOLD inputs must be fed with 0 for operating the register in serial
mode.
The shift registers of Figure 10.25 are classed as serial-in, serial­
out, shift-right registers. Similar design techinques may be used, to
318 CHAFI'ER TEN Introduction to Counters, Registers, and RTL

obtain a shift-left register. In fact, both shift-right and shift-left

capabilities may be combined to obtain a bidirectional serial-in,
serial-out shift register. A controlled shift-left register is shown in
Figure 10.26 that has an additional control input 8LE that deter­
mines what it does on the next clock pulse. The 8LI is the entry
point for shift-left input and 8LO is the exit point for shift-left output.
When the shift-left enable, 8LE, is low, the FF output is fed back to its
data input. In this way, digital information bits may be restored
indefinitely. It is interesting to see how the HOLD input has been
eliminated in this case. The clock inputs are still allowed to excite
the FFs. However, in the previous case, as shown in Figure 10.25,
the clock input was·notallowed into the FFs;Again when the8LE
control is high, the serial input sets up the right-most FF, Qo sets up
the second FF, ~ sets up the third FF, and so on. With 8LE high,
FIGURE 10.26 Controlled Shift. the circuit functions as a shift-left register. The serial output is
~ft Register. obtained out of the left-most FF.,r-

SLO ~---------r--~----------+---'---------~--~~--~SLE

su

SET

~--------~---'----------~--~--------~~--'--------CLOCK

CLEAR

The shift-left and shift-right registers can be combined to obtain

the bidirectional shift register of Figure 10.27. This register is simi­
lar in design to that of an up-down counter where an up-counter is
combined with a down-counter. The direction of shift is controlled
by the inputs 81£ and SRE. SLE and SRE control inputs cannot
be 1 simultaneously, except when HOLD = 1, so they are mutu­
 10.6 Parallel-Load Shift Registers 319

FIGURE 10.27 Bidirectional

Shift Register.

SET
SAE
... -.--------+-~~------~~~~------~~
SLE
r + - + - - - - - - - - - - / - - - - t - l - - - - - - - - + - . . - 1 - + - - - - - - - 1 - From (n - 2)th
stage

CLocK'::....:"<...;:.·,-''-~
HOLl)
(n + 1)th stage nth stage (n - 1)th stage

ally exclusive. When the SRE input is high, each of the FFs loads
the respective Qoutput of the FF on its immediate left. When the
SLE input is high, each of the FFs loads the Qoutput of the FF on
its immediate right. When both SLE and SRE are 0, the FFs are all
reset. Note this time how the CLR control has been eliminated.
When HOLD = 0 the register functions in its serial mode, and
when HOLD = 1 the old data bits are restored. We might decide
to eliminate one of the two shift controls. In that event we may
decide to keep SRE only by making sure that SLE has been
replaced with the complement of SRE. The register would function
as a shift-right type when SRE = 1 and as a shift-left type when
SRE = O. But this arrangement causes a problem if we need to
"'; -:
reset the FFs at any time. This problem could be solved, however,
by feeding the complement of a CLR control to each of the FF
resets.
An important application of shift registers is their role in arith­
metic operations. A binary number can be multiplied by 2 by shift­
ing the number one bit to the left and divided by 2 by shifting the
register content one bit to the right. As we will see later, the bits
shifted in at one end and out at the other end are not unimportant;
they are used in arithmetic operations in many instances.

10.6 Parallel.Load An n-bit, serial-load shift register requires n clock pulses to load an
n-bit word. A parallel-load shift register, in comparison, loads all infor­
Shift Registers mation bits simultaneously. Both serial-in and parallel-load shift
registers have specific applications in digital systems. A parallel-in,
serial-out shift register using master-sh~ve SR flip-flops is shown in
Figure 10.28. The parallel data are loaded using the jam-entry
 320 CHAPTER TEN Introduction to Counters, Registers, and RlL

FIGURE 10.28 Three-Bit,

. Parallel-In, Serial-Out Shift
Register.

SET
Do

~----r-~----~+-----------r-~----~~----------+---~~E

00

CLEAR------1

scheme that was discussed in Section 10.3. When the enable signal
E is high, the data are loaded into the register in parallel. Again, if
E is low, the Qoutput of the FF of every stage is shifted to the right
by means of the combinational gates. In either case the HOLD con­
trol must be held low. Parallel-in, serial-out shift registers allow
accepting data n bits at a time on n lines and then sending them one
bit after another on one line. This is a standard mode of communi­
cation between digital systems.
At the receiving end of two digital systems communicating over a
single data line, it is necessary to collect n bits and then transfer
them in parallel to the receiving system that is designed to handle n
FIGURE 10.29 Four-Bit, Serial­ bits simultaneously. Figure 10.29 shows the logic circuit of such a
In, Parallel-Out Shift Register.

SET­ ~
V
SRI 8 3 PR 0 3 8 2 PR 02 8 1 PR 01 So PR 0 0
- t-­

y~ >- rC> -<:> rO r>

R3 '03 R2 '02 R1 'OJ Ro '00 r- r - ­
CLR CLR CLR CLR
) )

HOLD

CLOC K
-~ ,

D3 D2 D1 Do
CLEAR­
~
V
 10.7 Universal·Shit~ Registers 321

serial-in, parallel-out register. The serial data are entered to the S

input of the left-most FF while the data are transferred in parallel
from the Qoutputs. The register is organized in exactly the same
way as that of Figure 10.25[b]; however, all of its Q outputs are
available, which is not the case for all shift registers. Both parallel
loading access features are included in the four-bit, parallel-in, par­
allel-out register shown in Figure 10.30. The CLR, SET, and
HOLD inputs are set low for normal operation. When the LOAD is
low, the shift register performs the shift-right operation. When the
FIGURE 10.30 Four-Bit, LOAD is high, the inputs 13, 12, Ib and 10 are loaded in parallel into
Paralle.-In, Parallel-Out Shift the register coincident with the next clock pulse. The outputs 0 3, O2,
Register~.. Ob and 00 are available in parallel from the Qoutput of the FFs.

SRO
0, 00
SET
PR PR
..-------1 So 00

0,
CLR

CLEAR

LOAD'~,~~-r-------------'-r------------~~------------~

•
10.7 Universal· Shift A universal-shift register is a versatile shift register that has capabilities
for parallel loading, parallel outputs, bidirectional shifting, and
Registers bidirectional serial input and output. In other words, it is capable of
operating in all of the register modes described previously. There
could be two different ways to realize a universal-shift register:
either by modifying a parallel-in, parallel-out shift register or by
building one from scratch.
322 CHAPTER TEN Introduction to Counters, Registers, and RTL

The circuit of Figure 1O.31[a] shows the logic diagram of a four­

. bit, parallel-in, parallel-out shift-right register. Its internal circuitry
and functions are exactly like those of the circuit of Figure 10.30. It
has five entry points for the data inputs-SRI, 13, 12) III and l o-four
outputs-03 through Oo-and five control inputs-SET, HOLD,
CLR, CK, and LOAD. The register is capable of either serial or
parallel entry and serial and parallel output. Note that in the cir­
cuit of Figure 10.30 the normal serial mode is only shift-right. Such
devices are also available commercially. When· LOAD is low the
FIGURE 10.31 Block Diagram register works as a shift-right register, and when LOAD is high it
of Figure 10.30: [a] in Shift-Right works as a parallel-in register.
Mode and [6] in Shift-Left Mode.

SLO

03 O2 0, 00 03 O2 0, 00
SET - SET
HOLD 4-bit, - HOLD 4-bit,
parallel-in, parallel-in,
CLR parallel-out, - CLR parallel-out,
shift-right shift-right
CK register --0 ~CK register

LOAD - LOAD
SRI 13 12 I, 10 SRI 13 12 I, 10
I
)
SLI
[8] [b]

The circuit of Figure 10.30 can be externally modified to perform

a lett-shift operation, as shown in Figure 1O.31[b]. This modification
is accomplished by selecting the parallel-mode, LOAD = 1, and by
connecting each parallel input to the output of the FF on the imme­
diate right. This form of shifting is called a wired shift since the shift­
ing action is accomplished by external connections.
An n-bit, universal-shift register can be designed by means of n D
FFs and n 1-of-4 multiplexers. The ith bit of suCh a universal regis­
ter is shown in Figure 10.32. Now all that remains is to cascade n
such units. When AB = 00 the register perfornis a parallel-in oper­
ation,. when AB = 01· the register restores the old value, when AB
= 10 the register performs a right-shift operation, and when AB =
11 the register performs a left-shift operation..
 10.8 Shift Registers as Counters 323

FIGURE 10.32 itb Bit of a SET

Universal-Shift Register.

,----10; Oi

CLOCK----~--a

CLEAR

f f
1-of-4 MUX E
-------1/0

10.8 Shift Registers as Shift register ICs are used at times to generate counts or controlled
sequences. As a result registers are used extensively in multiple
Counters address coding, parity bit generators, and random bit generators.
The output of each stage and its complement mtlSt be accessible for
these applications. These register outputs are used to drive combi­
national (eedback logic, as shown in Figur~ 10.33. The feedback
logic determines the next state of Qn. In the case of a bidirectional
register, the feedback logic controls shift-left and shift-right signals
and sets up a 1 or 0 to the appropriate SLI and SRI input.
The state diagram of a four-bit shift register with the}3 input of
FIGU~ 10.33 Feedback Shift­ th~ input FF available is shown in Figure 10.34. If the shift register
Right Register Configuration. is initially in the state-Q3Q2QlQo = 1001, then. there are two possi­

Feedback logic

,.
"

.., f:' ..
;,1" .
In On In- 1 On-l Jn~ 0 11-2 ... - JO 00
•

Y r-C >

Kn On
rC t>

Kn­ 1 On­ 1
rC >
Kn -i5n­ 2
r<: >
... - r-- Ko 00

CLOCK
 324 CHAPTER TEN Introduction to Counters, Registers, and R TL

FIGURE 10.34 Universal State

. Diagram for a Four-Bit Feedback
Register.

ble next states. These are 0100 if the 13 input is a 0, or 1100 if the 13
input is a 1. These values correspond to a shift-right operation. All
possible internal states of the register and all possible transitions
between the states are considered'in this state diagram.
In order to design a counter or sequence generator, the designer
selects the desired sequence of states on the universal state diagram.
Based on the desired sequence the feedback logic is designed so that
the register will cycle through the selected sequence of states. Exam­
ple 10.5 illustrates the technique.

EXAMPLE 10.5 SOLUTION

Design a decade sequence generator The state transitions and the corresponding feedback logic condition are
using a shift register that follows the obtained as shown in Figure 10.35. The resulting K-map for the feedback
sequence function13 is obtained as shown in Figure 10.36. The feedback function is

So ~ Sa ~ S12 ~ S'4 ~ SIS ~ S7 13 = Q, ~ + Q2~ + Q3Q2Ql

~ SII ~ Ss ~ S2 ~ Si ~ So
The feedback logic is realized using combinational logic. It is then fed to
the 13 input of the shift-right register, as shown in Figure 10.37, to obtain
the desired sequence generator. Note that the nonsequence states lead
eventually to the sequence as follows:
 to.8 Shift Registers as Counters 325

FIGURE 10.35 FIGURE 10.36

Transitions
( ""\
PS .. NS J3
1 - 1 1
0000 1000
1

1000 1100
1
0 0 - -
1100 1110
1

1110 1111
1
- 1 0 0
1111 0111 0
0 - 1 -
0111 1011
1

1011 0101 0
0101 DOlO 0
0010 0001 0
0001 0000 0
I

FIGURE 10.37

rC
r(
~

'-­

l(
c

J3 03 J2 O2 J1 01 Jo 00

v~
. ,:

t> r< [> - - r t'> Lc I'>

K3 03 K2 O2 K1 01 Ko 00

CLOCK

Consider, for example, that}3 is a 0 whenever the present register state is at

S3 (= 0011). Therefore, the system would shift from state S3 to SI (= 0001)
as per the state diagram of Figure 10.34. If the sequence generator in this
example enters an unused state due to a glitch or on power-up, it will
return to the decade sequence after a maximum of two clock pulses.
326 CHAPTER TEN Introduction to Counters, Registers, and RTL

10.9 Counter and There are hundreds of applications of counters and shift registers.
They are used extensively in computers. In general, digital com­
Register Applications puters process numbers by repeated arithmetic and logic opera­
tions. Execution of a specific instruction usually involves moving the
instruction and data between registers. The data are operated on by
the ALU as they are transferred between registers. These transfer
sequences are in turn controlled by sequential circuits. In particu­
lar, registers provide the means for the storage of bits as they are
being processed. On the other hand, counters keep track of the next
memory location and count the intervals in the sequences that con­
trol these complex operations. In this section we shall consider only
a few of their many important applications.
The operations in digital computers are performed in parallel in
most cases since this is a faster mode of operation. In comparison,
serial operations are slower but require less complicated and less
expensive circuits. Consider the add function. In Chapter 5 the
design of parallel addition circuits was examined in detail. The
techniques developed were reasonably fast but they involved very
complex circuitry. Frequently, the designer must make a trade-off
between time and the number of components.
The add operation can also be performed by loading the addend
and augend into two serial shift registers and shifting one bit from
each register into a single-bit FA, as shown in the block diagram of
Figure 10.38. The carry-out of the FA is stored in a D FF and fed
back as the carry-in to the FA to be added to the next pair of signifi­
FIGURE 10.38 Serial Adder
Configuration.

SRI
- ADD SRE A SRO
GLOG K ~
Xi Sj ;---

Yi FA
---to Ci - 1 Ci ;---
SRI
SRE B SRO
~
~
PR
~ 0 0
•
Q ,
y
v--
GLR
~

J f-- CLEAR
~'"
 10.9 Counter and Register Applications 327

cant bits from the shift registers. The sum bit is shifted into the shift
register containing the augend as the augend bits are continually
being shifted out to the right.
Initially the shift registers A and B hold the augend and addend
and the D FF is cleared. The summation is achieved by connecting
each pair of bits, through shifting, together with the previous carry­
out into the FA circuit and by transferring the sum bits serially, into
the register A. The ADD command starts and stops the operation.
When ADD is high the registers perform a shift-right operation at
each clock, and when ADD is low the registers maintain a hold
mode. In the next chapter we shall consider every aspect of how to
design such a serial adder circuit. In the meantime we will develop
other relevant concepts.
Operations in digital systems are controlled by a sequence of tim­
ing pulses. The control unit in a serial computer must generate a
signal that remains high for a number of pulses equal to the number
of bits in the shift registers. For example, the serial adder system of
Figure 10.38 requires a control signal, ADD, for its operation. Fig­
FIGURE 10.39 Generation of ure 10.39[a] shows a control circuit that generates a signal that
Timing,Sequences: [a] Circuit and remains high for a period of 16 clock periods. The four-bit IC
[6] Timing Diagram. counter and the SR FF are initially CLEARed. The BEGIN signal

BEGIN S PR Q

CLOCK
I I I I
..t>
A B C D HALT
1--< LOAD R CLR Q
CLEAR ­ ~ ,.. CLR
V .­ E
IC counter CARRY­
OUT
-r t>CK
QA 08 Oc OD

[a]

CLOCK

BEGIN -.Ji'-_____________________

HALT---------------------------------------~

Q-----' L
[b]
328 CHAPTER TEN Introduction to Counters, Registers, and R1L

sets the SR FF, which in turn enables the counter. The FF output Q
remains high. for 16 pulses, as shown in the accompanying timing
diagram of Figure 1O.39[ b]. When the counter reaches count 1111,
HALT is activated, which in turn resets the FF. The BEGIN signal
is synchronized with the clock and is made to stay on for one clock
period. It could be made to stay on for a longer period; however, if
it is made to last for more than 15 clock periods, the circuit will not
function as expected. This HALT signal might be used in another
similar circuit to generate a BEGIN pulse.
In a parallel mode of operation, a single pulse is generally used
to specifY the time at which an operation should be executed. Shift

registers may be used to realize such a timing circuit when con­

FIGURE 10.40 Four-Bit Ring nected as a ring counter. A shift-right register used as a ring counter is

Counter: [a] Circuit and [6] . shown in Figure 10.40. A feedback path is provided from the serial

Timing Diagram. output to the serial input of the shift register. A shift register con-

CLEAR

HOLD
[a]

CLOCK
0 LrL
03
0

O2
0 L
01
0

00
0
[b1
10.9 Counter and Register Applications 329

nected as such circulates the register contents. Initially the CLEAR

input is set to a 1. This presets the right-most FF to a 1 and clears
the others. The starting output word, therefore, is 0001, and as the
clock pulses occur the output word becomes 1000,0100,0010,0001,
and so on. Only four unique states are possible. This circuit would
allow timing four sequential operations by tying each of the four
operation's initiating lines to one bit of the ring counter. Each oper­
ation would then be active one-fourth of the time.
A four-bit binary counter has 16 counting states, whereas the
ring counter has only four states. Thus the ring counter makes an
extremely uneconomical use of FFs, which is more true for systems
requiring a large number of timing signals. An excellent alternative
to using an uneconomical ring counter is to use an n-bit binary
counter and an n-to-2T1 line decoder. This combination is often
referred to as a Moebius or Johnson counter (no relation to the coau­
thor). It also involves a shift-right register like that of a ring counter,
but it is connected in a switch-tail configuration.
As stated earlier, an n-bit ring counter provides only n distin­
guishable states. The number of states can be doubled if the shift
register is connected in a switch-tail configuration, as shown in the
four-bit Johnson counter of Figure 10.41. Here Qo instead of ~ is
fed back as the D3 input. The register shifts one bit to the right with
every clock pulse, and at the same time the complement value of
the fourth FF is transferred to the left-most FF. Consequently, this
would result in eight different counting states: 0000, 1000, 1100,
1110, 1111, 0111, 0011, and 0001. These eight states must be
decoded to give eight distinct timing sequences. Unlike the previous
circuit, the least significant FF need not be preset. For minimum
chip count, a ring counter or a Johnson counter is the best choice
for a timing circuit.
We have emphasized that counters are often used for sequencing
various arithmetic and!or logic operations. And in most instances
these operations are exe~uted only if certain conditions are met. Fig­
ure 10.42 shows a four-state controller that can control four distinct
operations. The sequencer circuit consists of a two-bit synchronous
counter, a l-of-4 MUX, and a 2-4 line decoder. Each of these four
operations is activated by a low on the respective sequencer output.
The sequencer output, Dm will become low during its allocated time
only if Cn (representing the corresponding condition) is a 1. The
counter is first CLEARed. Consequently, it S'I:ARTs at the 00
address when the function Fo is performed provided the condition
Co is met. The function corresponds to a specific operation, and the
condition Co may be the result of one or several test results. When
the counter reaches the 01 value the function Fl is executed, pro­
vided that the condition C1 is met. The synchronous count allows
the operations to be executed only at regular intervals. Once the
sequencer is STARTed, the test conditions determine whether or
330 CHAPTER TEN Introduction to Counters, Registers, and R 11..

FIGURE 10.41 Four-Bit Johnson

Counter: [a]· Circuit and [6]
Timing Diagram.

I-----IDo PR 0 0

0,
CLOCK
CLEAR

HOLD
[a]

CLOCK
0

03
0 I
O2
0

0,
0

00
0
[b]

not the function is selected by the decoder. If the condition is not

met, the decoder remains disabled, and as a result the correspond­
ing function is not executed. The significance of a sequencer circuit,
such as that of Figure 10.42, becomes meaningful only when it is
allowed to control a complex digital system. Prior to the sequencer
implementation, concepts ofRTL (Register Transfer Language)
will be introduced in Section 10.11. .
 10.10 Bus Concept 331

FIGURE 10.42 Two-Bit

Operation Sequencer.

j I I I
00 0 1 O2 0 3
10
1-01-4 MUX EO--
11
f 7
START ~ I
I
~ J 1 . . ,. 0 1 J2 O2 ~

CLOCK ~V -C j')

'--
Kl 0, .-- K2 O2
!)

CLEAR -~~
V
' - - - - 11
2-4 decoder EO--
10
00 0, D2 03

I I I I

10.10 ·Btls Concept

Digital systems have large numbers of registers, and as part of the
computational process it is often required to transfer data from one
register to another. Consider the case of m registers with n bits in
each. To allow direct inter-register transfer it would be necessary to
have a total of (m!)n data paths, which would require an awesome
number of wires running between registers. Such a data transfer
problem is solved by a group of wires called a bus. The bus concept
is analogoUs to a mass transport system where each commuter waits
in line until the transport becomes available (in this analogy the bus
can carry only one passenger). For a parallel transfer of n bits the
bus consists of only n lines, as shown in Figure 1043[a].
Each of the 1-of-8 MUXs is equipped with eight input lines,
three select lines, and one output line. The least significant inputs of
each MUX are connected to the respective FFs of register A. The
FF A7 is connected to the first MUX, A6 is connected to the second
MUX, As to the third MUX, and so on. The next significant input
of each MUX is connected to the respective FFs of register B. This
process is continued until all eight registers are connected. Figure
332 CHAPTER TEN Introduction to Counters, Registers, and R11.

FIGURE 10.43 [a] Register-to­

Bus Transfer Circuit for Eight
Registers, [b] Block Diagram of
Figure 1O.43[a], and [c] Bus-to­
Register Transfer Circuit for Eight
Registers.

A
[7J6 5 4[3 211loJ I 0 7 -0 ;
II ... I

B7 17
J~7 ~
r~71
H7
~6 J~61 ~6 r16
C6 E6 G6 C s Es s
r
~s J~s ~sT~s
G C4
r r
£4

~4 ~4 ! ~4 ~4
G4

1-01-8 MUX 1-01-8 MUX 1-01-8 MUX 1-01-8 MUX 1-01-8 MUX
START
~~ E I E E E E

" 12 11 10 f 12 I, 10 f 12 I, 10 f 12 I, 10 f
I I I I
x .1 1 1 1 .
y
z
SELECT -
.
I I
8-line bus
[aJ

TIT l Bus line

[b]

Y'----I/l 3-8 decoder .Eo---o::. START

z' 10

[c]
 10.10 Bus Concept 333

1O.43[a] shows the necessary connections. When the select lines, x,y,
and z, are all low the least significant input to each of the MUXs is
selected, and consequently the bus is loaded with the contents of
register A. Each combination of the x, y, and z inputs selects the
contents of a particular register and the contents are then attached
to the bus. The simplified block diagram is shown in Figure
1O.43[b]. It is possible to design a bus system without the MUXs if
the registers have tri-state outputs or are connected to the bus using
tri-state buffers. Finally, the contents of the bus are required to
reach a certain destination register. This requirement is accom­
plished by the circuit arrangement of Figure 1O.43[c] where the
select lines x' ,y', and z' determine the particular destination register
by means of a 3-8 line decoder. Upon receiving the proper select
inputs, the contents of the bus are loaded into the selected register.
Another example of the use of a bus involves a simple memory
device. Registers often are assembled together to form a larger stor­
age array. This arrangement of registers is referred to as a scratch­
pad memory~ Figure 10.44[a] shows an arrangement of registers that
can store up to four four-bit words. The device consists of four regis­
ters that in tum have four FFs each. When the WRITE ENABLE
(WE) is low, the four data inputs, 10, Ib 12, and 13, are routed to a
particular location of each register as specified by the entries in the
WRITE SELECT (WS) lines. A 00 on the WS lines will store the
input bits in the respective Oth cell of the registers. Similarly, 01, 10,
and 11 applied at the WS lines would respectively select the first,
second, and third bit of each register. Stored data from the scratch­
pad memory may be retrieved through the four output lines by
applying a low to the READ ENABLE (RE) and necessary address
bits to the READ SELECT (RS) lines. Figure 10.44[b] shows the
logic diagram of the memory formed using gated D latches.
Scratch-pad memory, although very fast, is not extensively used.
FIGURE 10.44 I6-Bit Scratch­
It is not particularly suitable for LSI because too many pin-outs
Pad Memory: [a] Block Diagram

WRITE LOGIC 1---+1 READ LOGIC

00

l
10

I,
Data Data
12 lOUtputs
13
WRITE
WE
ENABLE

READ ENABLE
~ ~
[a] WRITE SELECT READ SELECT
334 CHAPTER TEN Introduction to Counters, Registers, and RTL

FIGURE 10.44 (Continued) 16·

. Bit Scratch-Pad Memory: [b]
Circuit Showing D Latch at the
ith Row andjth Column and the
ith Output.

O;o---r---------­
RS,
RS o
WE
R
O;,---r--""'"

RS,

RS o

Oi2---r--""'"

AS,

RS o

0 ; 3 - -l ---""
RS, 0---'
RS o

°"" i(= row #) "" 3

°"" j(=: column #} "" 3

R {WSl for j = 0,1 AS lines: RS" ASo

WS, for j = 2,3 WS lines: WS" WSo

p ={WSo for j = 0,2

WSo for j = 1,3
[b]

make such a chip economically unattractive. Again, being

fabricated from several gates I FFs, scratch~pad uses significant
power and chip area. We shall look at some of the alternative mem­
ory sources in Chapter 12.
The applications of shift registers and counters are limited only
by the. imagination of the designer. Their importance will become
more obvious as the design of digital systems continues in this and
subsequent chapters.

EXAMPLE 10.6 SOLUTION

Design a four-bit register capable of One of the procedures for performing the 2's complement operation (see
perfonning the 2~s complement Section 1.4 for details) requires that all bits on the left of the least signifi­
operation on its content. cant nonzero bit be complemented. Using this scheme, a register can be
built for loading the numbers in parallel and then performing the 2's com­
plement operation using additional logic.
 10.10 Bus Concept 335

Define a control input, LOAD, to be used for selecting the parallel load
operation, and its complement, LOAD, for selecting the 2's complement
operation. The loading excitation corresponding to a register consisting of
only T FFs is given by

T; = LOAD' Ii + LOAD' (0-1 + 0-2 + ... + Qo) for i > 0

where Ii is the parallel data input to the ith FF. The first term corresponds
to loading the parallel input and the remaining terms correspond to per­
forming a complement operation if at least one of the less significant bits is
a 1. The equation is valid as long as the FFs have been CLEARed initially.
Note that in this complementing scheme Qo remains unchanged.
Accordingly, the excitation equation for the least significant FF is given by

To = LOAD' 10 + LOAD (Qo) 0

The overall excitation equation for the ith FF is now obtained as follows:

T; = LOAD 0Ii + LOAD (0-1 + (0-2 + (0-3

0

+ (0 0 • + (Qo + 0) ...))))

Consequently, if at least one of the bits on the right is a 1, the bit in ques­
tion is complemented. The equation may be used to obtain the register cir­
FIGURE 10.45
cuit shown in Figure 10.45. The register should be LOADed with the data

CLEAR

01 00
PR CLR
T1

'.~'. f

HOLD

~------------~~------------~~------------~~~-- LOAD
336 CHAPTER TEN Introduction to Counters, Registers, and R 11.,

only after the FFs have been CLEARed. The corresponding 2's comple­
ment i~ then obtained by supplying a low LOAD input. Note that a high
HOLD could be used at any time to keep the register content unchanged
indefinitely.
Pl. r 1! =

EXAMPLE 10.7 SOLUTION

Design a digital wristwatch that A detailed examination of the problem reveals that this device can be
displays the month, day, and time made to function in the prescribed way in several steps:
accurate up to 1 second using
BCD-to-seven-segment LED
Step 1. Sixty of the 60 Hz clock pulses may be counted in sequence to indi­
cate 1 second.
display devices. All display except
for seconds should be adjustable by Step 2. Sixty seconds may be counted in sequence to indicate 1 minute.
the corresponding external
ADJUST switches. Assume that the Step 3. Sixty minutes may be counted in sequence to indicate 1 hour.
ADJUST switches can be used only
Step 4. Twelve hours may be counted in sequence to indicate one-half day.
if the external DISABLE switch has
been turned on. The external Step 5. Two sequences of 12 hours may be counted to indicate 1 day.
BEGIN switch may be used to
resume the operation. Assume Step 6. Days are counted, and once the count equals the maximum number
further that the MINUTE (MIN) of days in a given month, the month count should be made to go up by 1 and
ADJUST automatically resets the the day count should be set to 1.
seconds and you have a 60 Hz Step 7. At the completion of 12 months, the month count should be set to 1.
quartz crystal to run your device.
From the steps listed it appears that this problem can be solved by inter­
connecting several counters and decoders. A list of the components needed
for the wristwatch includes the following:

scale-of-sixty counters,
up-counter that counts 1 through 12,
binary-ta-BCD converters,
decoding circuit to determine the last day of a month,
multiplexers,
TFF,
seven-segment display devices.

Steps 1-3 can be realized using three modulo-60 counters. Two modulo-12
counters could be used to implement Steps 4-7. Steps 5 and 6 can be
implemented, respectively, using a TFF and last-day decoder circuit. BCD
converters and seven-segment displays will be used for the purpose of dis­
play, and MUXs will be used for routing the data. •
The IC counter module of Figure 10.18 will be used as the basic unit for
producing the counter modules a and b as shown, respectively, in Figures
10.16 and 10.17.
Internally, module a consists of two four-bit IC counters casca.ded
together. The carry-out of the first is made to enable the second counter
(see Figure 10.19 for a similar circuit). When the count reaches 59.
j
j
 10.10 Bus Concept 337

FIGURE 10.46

-b-
,:
r-c LOAD A B C 0 L--a LOAD A' B' C' 0'

ro CLR ,.. CLR

IC counter IC counter
Inputs { E Co E
08 Oe 00 ~>CKOA Os O'e 00
[ k >CKOA

?
Outputs

- Za\
...,r-' i---

(111011), zero count is LOADed. However, module b consists of a single

four-bit IC counter that is made to LOAD 0001 every time the count
reaches 1100.
Module c was designed already in Example 5.7 (Figure 5.24). The only
other module that remains to be designed is the decoding circuit for identi­
tying the last day of a month. In terms of their lengths, the months may be
classifed into four groups, respectively having a total of 28, 2~, 30, and 31
days. The month of February has 28 days normally but has 29 days in a
leap year. The module d may be designed accordingly, as shown in Figure
10,48, using a l-of-4 MUX and two four-bit IC counters. The end-of­
month is decided upon by the four values of 4B. The MUX output
becomes a 1 at the end of 28, 29, 30, and 31 input pulses when AB is 00,
01, 10, and 11, respectively. The decoding circuit is set to 1 each time an
end-of-month has been located.
, The modUles may now be assembled to yield the wrist watch using the
steps described earlier. In &tep 1 the 60 Hz oscillator output is introduced
into module a, as shown in Figure 10.49. A high SEC output indicates that
an integral multiple of 60 input pulses has been counted. This is followed
by Step 2, as shown in Figure 10.50, where a BEGIN input would allow
another module a to count SEC pulses. For every integral multiple of 60
338 CHAPTER TEN Introduction to Counters, Registers, and RTL

FIGURE 10.47

A
I B C 0
1
-()
LOAD
"'CLR IC counter Co !---­
Inputs { E
~CK
~ OA 08 Oc OD

X/(r Outputs

y ~
V~
lb

SEC pulses, a high will be generated at the minute output, MIN. The out­
puts of this module a are then converted to equivalent BCD numbers by
means of three binary-to-BCD converter modules as shown in the circuit.
For an explanation of this multi-bit, binary-to-BCD conversion scheme,
review Example 5.B.
The BCD output is displayed by meatlS of BCD-to-seven-segment
LED display devices. The display will become 00 whenever the MIN
ADJUST input is activated. This allows for the fact that the second display
is cleared automatically whenever the minute display needs to be adjusted.
Note, however, that the DISABLE input must always be activated prior to
MIN ADJUST operation. The resulting low at M </utput would be used
for all of the remaining ADJUST operations.
In Step 3, as shown by the circuit of Figure 10.51, the output of Figure
10.50 is fed into another module a. This configuration is made possible by
means of a 1-of-2 MuX selectable by M. A high M would cause the circuit
to count MIN pulses. In comparison, a low M would allow the adjustments
of MIN counts. The HOUR output becomes a 1 corresponding to every
integral multiple of 60 MIN pulses.
 10.10 Bus Concept 339

FIGURE 10.48

---oCLR

------IE d

- - - 0 CK

1
-
A 8 C D A' B' C' D'
J.-{) LOAD L-o LOAD

C CLR
CLR
IC counter IC counter Co '---­
E Co E

-a )CK'
,--<: ~CK
OA Os Oc OD OA O's Oc OD

(/)

l' :;
(/) 0­
.­

7?r
:; ::J
0­
C
o
ld
r- Do 1­ e.­
l- Dl of­
D2 4 f f '"' ,l' -- ~

I­
II TI
D3 MUX I I tLl
11 12
A
I
Y Y
B
' .. ­
Y -

FIGURE 10.49
-_+_---o CLR
E a SEC
60Hz
oscillator
340 CHAPTER TEN Introduction to Counters, Registers, and RlL

FIGURE 10.50

M
~
MIN ADJUS T

}-~ CLR
BEGI ~ 0
J
60 Hz .­
'=-c > 1- E a Za - MIN
DISA ~ --C >CK
K Q
Os OA OD Oc Os OA
0

SEC
I
8 4 83
I 8I
82 1

04 03 O2 0 1
I l
84 83 82 8 1

04 03 O2 01
I I
84 83 82 8 1
0
04 03 O2 01
I
I I

I-I
II
; ;
II
FIGURE 10.51

1--0 CLR
MIN ­ Do
ADJUST 1-of-2
1- E a Za - HOUR

MUX r "'>CK
v O'S OA OD Oc Os OA
­
MIN 0 1 Jo
I
I
84 83
I 8I
82 1
M
04 03 O2 0 1
I I
84 83 82 8 1
04 03 02 01
«
I I
84 83 82 8 1

I II
04 03
I I
O2 0 1

1-'
0 I ,
 10.10 Bus Concept 341

The circuitry in Step 4 is constructed using a module b, as shown in Fig­

ure 10.52. Its working principle is very similar to that of Figure 10.51. The
112-DAY output becomes a 1 corresponding to every integral multiple of
12 HOUR pulses.
The circuit of Figure 10.53 corresponds to both Steps 5 and 6. The
counting circuit goes up by a 1 corresponding to every other 112-DAY
pulse. Every time the end-of-month is located by means of AB inputs, the
circuit ofmodule d is set to day 1. The MONTH output becomes high each
time the end-of-month is identified.

FIGURE 10.52

HOUR 00
ADJUST b Zb 1f2-DAY
1-of-2
MUX 1 CK
OD Oc 0 8 OA
HOUR 0 1 10

I-I
n U
"

FIGURE 10.53

DAY ­ 1-C CLR

ADJUST 00
1-of-2 1 ­E
MUX ,,, "'VCK d Zd ­ MONTH
-
..-­ 0110 A 8 Oil OD Oc 0 8 OA
From month
~ I decoder
{ circuit of
6 0
Figure 10.54
...... 0
T< b--1f2-DAY
I
84 83 82 81

- Q 04 03 O2 0 1 •
I 84 8j 8 2 8 1
0

I I
04 03 O2 0 1

,,
i

I-I I-I
I ,
342 CHAPTER TEN 1n.troduction to Counters, Registers, and RTL

In the last step, as shown in Figure 10.54, the MONTH output is fed
into a module h. The FF outputs, Qo, Qc. Qs, and ~. and the external
input, LEAP, are decoded to generate A and B. These outputs, A and B,
are fed into modul~ d of Figure 10.53. Note that LEAP is the extra input
required for indicating that the current year is a leap year. This input
needs to be entered prior to February 29. The month decoder circuit of
Figure 10.54 examines the four FF outputs of module h. For the months
January, March, May, July, August, October, and December, both A and
B are set high. For all other months, except February, A = 1 and B = O.
In February A = B = 0 if LEAP = 0, and A = 0 and B = 1 if LEAP =
1. Consequently, the decoder has the following Boolean equations:

A = Qo + Qc + ~
FIGURE 10.54
B = Qo EB ~ + ~Qc' LEAP

1--<: CLR
MONTH - Do
ADJUST 1- E b lb r - ­
1-of-2
MUX 1 ¢ G>CK
Qo Qc Os OA
MONTH ­ 0,
10 0
I
M
I
84 83 8 2 8,
I 1 I I
Month A t-- To circuit
decoder of Figure
04 03 O2 0, } 10.53
0 0 circuit 8 t-­
I r l I
I-I i
'-I
, I LEAP
LI

The design is now complete. After counting the chips involved, let's call
this a table dock. Note that BCD counters instead of binary counters could
have been used in this design, which would have reduced the total chip
count. However, the incentives behind this particular design were to
demonstrate the use of binary counters, to demonstrate the use of binary­
to-BCD converter modules, and to demonstrate the thought processes
involved in digital circuit designs.

10.11 Register Transfer The diversity of register types and applications points to the need
for concise language to describe the flow of infdimation (and pro­
Language Operations cessing enroute) of bits between registers. The most commonly used
way to attain this goal is by an informal scheme called register trans­
fer language, RTL, which was introduced first by 1. S. Reed. A thor­
ough investigation would reveal that a complete register transfer
description is made up to two units: data and control. The data unit
consists of registers, data paths, and logic necessary to implement a
10.11 Register Transfer Language Operations 343

set of register transfers. The control unit, on the other hand, generates
necessary signals in a specific sequence to regulate the register
transfers within the data unit. The RTL scheme has the ability to
specify the hardware involved in both units.
The simplest of all RTL operations is represented by P ~ Q,
which indicates that the data in register P are replaced by the data
in register Q. It is also understood that both of these registers have
the same number of bits. Such an operation can be completed dur­
ing a single clock period, and thus it corresponds to a single-state
transition of a sequential machine. One elock period, referred to as
the cycle time, may be taken as the basic unit of time at the MSI
complexity leveL For uniformity, the following standardization is
essential:

1. The contents of the registers should be denoted by one or

more letters with the first always in uppercase. A concate­
nated register is the result of joining two or more registers
in a string (represented by the register symbols separated
only by commas) so that the LSB of the first-mentioned
register is one bit to the left of the MSB of the second one,
and so on.
2. Transfer between the registers will be considered parallel.
In other words, all of the bits 'will transfer at the same
instant of time.
3. The bits of each register shall be numbered from right to
left. Ao and An - 1 respectively represent the LSB and the
MSB of an n-bit register named A. Note also that A1,4 rep­
resents the bits 1 and 4 of register A, A1-4 represents bits 1
through 4 of register A, and AM represents the subset of bits,
M, of the register A.

Table 10.1 lists some of the most important RTL examples that
include arithmetic, bit-by-bit logic, shift, rotate, scale, and condi­
tional operations. In order to differentiate between the arithmetic
and the logic operations, the following convention is maintained.
The arithmetic addition is represented by a + symbol, the logical
OR operation by a V symbol, and the logical AND operation by a
1\ symboL The shift, rotate, and scale operations are generally rep­
resented by two lowercase letters. For shift and rot~te, the first letter
indicates the type of operation (r for rotate and s for shift) and the
second letter indicates the particular direction (r for right and. I for
left). Furthermore, in the rotate operation the LSB and the MSB
are considered to be adjacent. For all shift operations a 0 will be
assumed to occupy the vacant bit. For the scale operations sel indi­
cates scale left and scr indicates scale right.
344 CHAPTER TEN Introduction to Counters, Registers, and RTL

TABLE 10.1 Examples of RTL

Type of Register Bits after
Operation Meaning Operation
General
A3 -E- Az Bit 2 of A to bit 3 of A A = 11110
A3 -E- B,4 Bit 4 of B to bit 3 of A A = 11110
AI _3 -E- BI ­ 3 Bits 1 through 3 of B to bits 1 A = llOOO
through 3 of A
Bits 1 and 4 of B to bits 1 and A = 10100
40fA
Groups of bit Z of B to bits 1 A = llOOO
through 3 of A
Arithmetic
B-E-O Clear B B = 00000
A-E-B+C Sum of Band C to A A = 11001
A-E-B-C Difference B - C to A A = 10111
C-E-C+l Increment C by 1 C = 00010
Logic
A-E-BI\C Bit-by-bit AND result of B A = 00000
and Cto A
A -E-B V C4 OR operation result of B with A = llOOO
bit 4 of Cto A
C-E-V Complement C C = 11110
B-E-B+ 1 2's complement of B B = 01000
B-E-AE]jC X-OR operation result of A B = 10111
and C to B
Serial
B -E-sr B Shift right B one bit B = 01100
B -E- sl B Shift left B one bit B = 10000
B -E-sr2 B Shift right B two bits B = 00110
B-E-rrB Rotate right B one bit B = 01100
B -E- rl2 B Rotate left two bits B = OOOll
B -E-scr B Scale B one bit (shift right B=11100
with sign bit unchanged)
B -E- sel B Scale B one bit (shift left with B = 10000
sign bit unchanged)
Shift right concatenated B BJC = 0011000000
and C two bits
Conditional
IF(B4 = 1) If bit 4 of B is ai, then C is C = 00000
C-E-O cleared
IF (B > C) If B is greater than or equal B = 00000
B -E- 0, C1 -E- 1 to C, then B is cleared and C C = 00011
is set to 1

Initial values: A
--
= 10110, B = 11000 and C = 00001.
z

The operations as described in Table 10.1 may now be combined

to write complex functions or a sequence of operations. The control
conditions are included along with the operation state to distinguish
one set of executions from another. Recall from the circuits of Fig­
ure 10.43 that the execution of an operation and the transfer of data
 10.11 Register Transfer Language Operations 345

are usually regulated by one or several control conditions. For

example, the loading of the contents of A into the bus may be
expressed as follows:
x. ji . Z: bus ~ A;

The control condition xjiz is separated from the corresponding oper­

ation by the H:" sign, while the sign ";" indicates the end of the
operation. This RTL statement indicates that when xjiz = 1, the
content of register A should be tran,sferred to the bus. When more
than one operation, A ~ Band B ~ sr B, are to be performed
under the same control condition S = 1, the operation is expressed
as follows:
S: A ~ B ; B ~ sr B ;

The operations that are performed simultaneously follow the same

control condition. Accordingly, when S is true, B is transferred to A
and if) also restored after being shifted right one bit. Similarly, the
RTL operations describing the function of the circuit of Example
10.6 may be summarized as follows:
CLEAR: R ~O;

LOAD: R~I;

LOAD: R ~ R + 1 ;

where R is the register receiving the input bits from register 1. These
three RT~ statements describe the action of a significant amount of
hardware. In this and the next chapter we shali be discussing sev­
eral of the complex circuits and their correlation with the corre­
sponding R TL statements. Such one-to-one correspondence will
make us appreciate the simplification that results from the use of
RTL.
Consider the circuit shown in Figure 10.55. Here we have two
registers, A and B, having four D FFs each. For simplicity the FFs

FIGURE 10.55 Realization ora

Register Copying Operation. ! !
V 03 V O2
!
V 01 Do
A
'0 3 0 3 O2 O2 01 01 00 00

6 6 ~
-
6
x

V 03 V O2 V 01 V Do
B

03 0 3 O2 O2 0 1 01 00 00
346 CHAPTER TEN Introduction to Counters, Registers, and RTL

within the registers are not internally cascaded together. The FF

outputs of register A are connected to the respective D inputs of reg­
ister B. Corresponding to the trailing edge of X input, the contents
of register A are loaded into register B. This hardware operation
will be represented by the following RTL statement:
X: B ~A ;

This could also be written as follows:

X: B3 ~A3, B2 ~A2' Bl ~Ab Bo·~Ao;

Both of these operations are equivalent; but we would prefer to use

the first form since it is more concise.
Note that a transfer operation is really a copying operation
where the contents of the source register remain unaltered. This
type of R TL operation is the most common, but there are other pos­
sibilities. Consider the four situations of Figure 10.56. In each of
these cases each of the registers is of four-bit length.
In Figure 10.56[a] the Qoutputs of register A are connected to
the respective D inputs of register B. A clock input P would result in
the following transfer:
P: B ~A;

Figure 10.56[0] shows register A where its MSB is connected to a O.

Each of the ~ outputs of register A, except for the LSB, is fed to the
Dn - 1 input of the same register. A pulse at R would cause the fol­
FIGURE 10.56 Realization of: lowing transfer:
[a] a Complement Transfer
R: A ~sr A;
Operation, [6] a Shift-Right
Operation.

Do
A

0,
SRI

Do
P
A
03 0, 0, 00 00
8 SRO
03 03 O2 O2 0, 0, 00 00
[a] [b]
 10.11 Register Transfer Language Operations 347

FIGURE 10.56 (Continued)

Realization of: [c] a Rotate-Right
Operation, and [d] a Logical
Operation and Transfer.

I s
..~~
I VA I V6 6
V
V D3 D2 Dl Do
A
>~h 0 3 O2 O2 01 0 1 00 0 0
I I
.,~, -c
[c) ,.

Do Do
A B

[d)

Similarly, Figures lO.56[c-d] respectively corresp~nd to the follow­

ing register transfers:
s: A ~ rr A ;

T:C~AAB;

Note that in all of these valid RTL expressions, the source and des­
tination registers have the same number of bits.
 348 CHAPTER TEN Introduction to Counters, Registers, and R TL

... EXAMPLE 10.8 SOLUTION

Design a typical stage for perfonn­ You might decide to use JK FFs for the design of register A. Register B
ing the following operations: doesn't need to be designed because it is no different than a regular register
with parallel outputs. The JK excitations necessary to turn on the ith FF
Tl:A~O;
for perfonning the required operations are obtained as follows.
T2:A ~A VB;
A+-O is possible when Ki = Tl
T3: A ~A 1\ B;
A +- A V B is possible whenJ; = T2 ' B; and K; =0
T4 : A ~A;
A ~ A 1\ B is possible when J; = 0 and K; = T3 ' B;
where A and B are two multi-bit is possible when Ji = Ki = T4
registers of equal bit size.
The ,corresponding circuit is obtained as shown in Figure lO.57.

FIGURE 10.57 T,-----.

T2 - - - t - - - ;
)--------1 J; Oi
Ai

CLOCK

~-r----r-----------~

10.12 RTL
The importance ofRTL to describe the internal operations of a dig­
ital system is primarily due to the flexibility with which a design
Applications
can be described and the direct way the data and control circuitry
can be realized from RTL statements. Consider, for example, a sys­
tem with a four-bit input, I, a four-bit output, 0, and three four-bit
registers, X, Y, and Z, in which the following algorithm is to be
implemented:
A: X ~I;
B: Z ~s12 X;

C:Y~X;

D: Y ~ Y /\ Z;
E: IF (Y3 = 1) 0 ~ 0 ELSE 0 ~ Y;
10.12 RTL Applications 349

The algorithm begins at state A and continues through state E.

The input, I, is shifted left two bits and stored in Z, and the comple­
mented input is stored in Y. This operation is followed by a bit-by­
bit AND operation between the contents of Y and Z, and the result
is stored subsequently in Y. If Y3 = 1, only 0000 is placed on the
output lines; otherwise, the contents of Y are placed on the output
lines. In brief, this particular algorithm includes various logic, shift,
and conditional transfers.
We shall now attempt to develop the hardware necessary from
this RTL algorithm. Each of these RTL statements has two signifi­
cant parts. In each statement, the right side of the ~ sign specifies
the signals that must be generated for either storage in registers or
transfer to output lines. Correspondingly, the left side of ~ specifies
the destination registers or output lines. For this example X can be
obtained by taking lines from the Qoutputs of register X No logic
gates would be necessary for the shifting operation, as evident from
Figure 1O.56[bJ. However, four AND gates would be necessary in
state D. Figure 10.58 shows the connections necessary to develop the
required signals, I, X, sl2x:. 1', Y /\ Z, and O.
The next step is to feed the generated signals to the respective
destination register or output lines. Figure 10.59 shows the involved
connections. The complete RTL algorithm requires five clock peri­
ods. Corresponding to the first RTL statement, the inputs, I, are
loaded to the D inputs of register X at the trailing edge ofclock A, a
transfer pulse fed to the clock input of register X during the allo­
cated time for the first RTL statement. Similarly, at the trailing
edge of clock B, AlAoOO are loaded into register Z. This is equivalent
to the transfer of A after it has been shifted left twice. The third
RTL statement requires that X be transferred to Y. This could have
been possible simply by loading X to register Y at the trailing edge
of clock C. However, it can be seen that the fourth RTL statement
also includes a load operation involving register Y. To allow for
these load operations, two separate ports of AND gates are used to
select the inputs to Y, and a port of OR gates is used to combine
them. The Qoutputs of register X are transferred throllgh one of the
AND ports by CL, a level signal that is high during the period
available for the third R TL statement. Similarly, the corresponding
outputs of both Y and Z are transferred through the other AND
port by DL. This AND port functions as the bit-by-bit AND logic
circuit and also as the select port for register Y. The, outputs of these
two AND ports are loaded into register Y at the trailing edge of
clocks C and D, respectively. The conditional transfer of the fifth
RTL statement is realized when EL is high by ANDing each of the
Y bits with Y3 and EL. It is appropriate to point out that when data
are loaded into a register, they remain there until the power is
turned off or other data are loaded. However, when data are placed
350 CHAPTER TEN Introduction to Counters, Registers, and R TL

FIGURE 10.58 Preliminary Step

for the Development of Signals: [a]
1, [b] X and sl2 X, [c] Yand Y A Z,
and [d] O.

o0

~
s/2 X

0, 00
y Z
03 O2 O2 0, 0, 00 0 0

~
YAZ o
[c] [d]

on the output lines, they remain there only during the steps for
which they are valid.
The next step in the design is to regulate this algorithm by gener­
ating A, B, CL, C, DL, D, and EL signals in the proper order as
specified by the algorithm. To run the ~ystem, a synchronizing sys­
FIGURE 10.59 Complete Data
Unit.

1
-
A B
1 T
03
c
\ O2 V 01 \i Do
(')
V 03 O2
b0
V 1 Do
x Z
03 03 O2 O2 01 0 1 00 Oc 03 0 3 O2 O2 01 01 00 00
L

r-­
...
CL OL
1 ---­
1 t[

Y
,--v i",--v ",--v I'LL

0
) ()
V 03 V O2 V 01 Do
y -

03 0 3 O2 O2 01 0 1 00 00

EL
-­ A A

,
351
352 CHAPTER TEN Introduction to Counters, Registers, and R TL

tern clock signal is necessary) as shown by the timing diagram of

Figure 10.60. The time during which each step is valid is indicated
by the level signals. When a level signal is ANDed with the system
clock, the corresponding transfer clock pulse is generated. Even
though AL and BL are not used for any of the operations in Figure
10.59, they are considered necessary in the control unit since they
are used to generate A and B clocks, respectively. However, there is
no need to generate the E clock signal.
We have already seen in Section 10.9 how various register orga­
nizations can be used to produce an operation sequencer. This par­
ticular algoritlun requires only a five-bit ring counter, as shown in
Figure 10.61. Using CLR input, theLSB of this ring counter is set
and the remaining FFs are reset asynchronously. Consequently the
FF outputs start generating the respective level signals in the correct
order. The corresponding transfer clock pulses are realized by
ANDing the level signals with the system clock.
It is now appropriate to consider several practical aspects about
the functioning of this control unit. In order to correctly change Qs
from a 0 to a 1, the CLR operation must be done in such a way as
to assure a full clock period for the first R TL statement. The circuit
might be expected to HALT this algorithm at the end of the algo-

FIGURE 10.60 Timing Diagram System

of the Control Unit. clock
o I I
I
I I
I
I
I
I

r-ti!
I

AL 1

o _---J. I
l l
~I-,:--:--i-+I-t-I- ' - - - l - - - - - ­
I I I I I

1 I I I I I

A
o n~~:~~~l-~:~-7_------­
I

I I. : :

BL
o-----~r-il~~~:--L:~--+_--------­
I : I I

B _
o _______~H I: ! ~~I~I--lro---+-------­
I I I

CL
~----------~r-1l~I~l~-~------­
1
I I
,I
I

C
o n~~:~--~-----­
I I

! !
DL
o------------~~~--~----­ I I

I I

o
o--------------~n~--~------­ I

I I

1
EL o Il~ _____
 10.13 Summary 353

FIGURE 10.61 Controller for the

Data Unit of Figure 10.59.

A B C CL o OL EL

, "r r-

L L
l ~ A A A A
- 0 5 PR 05 . D4 PR 0 4 0 3 PR 0 3 O2PR O2 0, PR 0, Do PR 0 0 f­
-:.cIo '.
"

,-,t '
~ > ClR Q5 rc I> ClR04 rC :> ClR 03 rC > ClR 02 -<: :> ClR 01 r<: > ClR Oo
~ (
System
clock I
ClEAR-
'"
V~

rithm, which could be accomplished by ANDing the system clock

with Ql prior to supplying it to various points of the circuit. This
arrangement will disable the system clock when QI = 1. EL
becomes high and stops the control circuit froni repeating the algo­
rithm. Consequently the outputs would be available indefinitely. In
many of the arithmetic circuits of the next chapter we will consider
the HALt operation in more detail.
Throughout this section we have preferred to use negative edge­
triggered FFs, primarily because transfer clock pulses are easily
derived by ANDing the corresponding level signal with the system
clock. Generating the transfer clock pulses would not be as easy if
positive edge-triggered devices were used. Thus in the event a
designer is faced with Using a positive edge-triggered device, several
modifications are necessary. These modifications involve inverting
the system clock before feeding it to the clock input of positive edge­
triggered devices. Furthermore, NAND gates rather than AND
gates should be used for generating the transfer cJock pulses.

10.13 Summary
In covering the application of what is known as the traditional
sequential machines, our studies moved through synchronous and
asynchronous counters; serial, parallel, and mixed-mode registers;
and operation sequencers. An understanding of these functional
units was subsequently applied to the development ofRTL (register
354 CHAnER TEN Introduction to Counters, Registers, and RTL

transfer language). This unique tool was then implemented in the

design of both data and controller units. An understanding of the
concepts presented is necessary, for much of the remaining material
in this book is dependent on understanding this chapter.

Problems
1. Determine the state diagram of the three-bit programmed
counter whose excitation equations are given as follows:

Dl = Q3Ql + Q2Ql

D2 = Q2(~ + Q3)

D3 = Q3

2. Design a three-bit, Gray-code counter using (a) D FFs and (b)

JKFFs.
3. Obtain a four-bit presettable asynchronous counter. Discuss
its operation and significant characteristics.
4. Design a modulo-12 counter using a modulo-3 and a modulo­
4 counter. Discuss its functions and characteristics.
5. The registers of Figure 10.38 store six bits each. A holds
010101 and B holds 001010. List the binary values in A and Q
after each shift. Assume that Q was cleared initially.
6. You are given D FFs and several l-of-4 MUXs and nothing
else. Obtain the logic diagram for a three-bit register that is
able to do the following: hold the present data, shift right, shift
left, and load new data in parallel.
7. Design a four-bit circulate-right shift register using (a) D FFs
and (b) JK FFs.
8. VerifY the equations of the four-bit down-counter given in
Section 10.2.
9-. Design a four-bit shift register using JK FFs and a minimum
number of assorted gates that performs the shift-left operation.
10. Explain how the unused sequences of Example 10.5 are
obtained.
11. Design a divide-by-2048 counter using specific four-bit binary
counters. The counter consists of a single FF, ~, followed by
three cascaded FFs that form a divide-by-8 counter. It has two
inputs, A and B, and four standard FF outputs, ~ through
0J. The two reset inputs, Rl and R2, clear the FFs when both
are high. The counter counts in sequence when at least one of
the reset inputs is low.
12. Obtain a circuit of as few FFs as possible to sequence 16 differ­
ent operations.
13. Consider the truth table for a four-bit, Gray code-to-four-bit
 Problems 355

binary equivalent conversion. It can be seen that if the MSB

of the codes is disregarded, the numbers 8 through 15 will be
found to be mirror images of the numbers 0 through 7. Design
a parallel Gray code-to-parallel and serial binary converter
using a four-bit shift register and only one FF. Explain the
detailed functioning of the circuit.
14. Draw the logic diagram of a four-bit register with clockedJK
FFs having control inputs for the increment, complement, and
parallel transfer micro-operations. Show how the 2's comple­
ment can be implemented in this register.
15. Design a two-bit counter that counts up when control variable
C is a 1 and counts down when C is a O. No counting occurs
when the control variable D is a O.
16. Using a four-bit binary counter with synchronous clear and
asynchronous load and clock action on the leading edge, com­
plete the necessary circuit to make a two-digit BCD counter.
Make it as hardware-efficient as possible.
17. You have available a four-bit adder and a large assortment of
gates, counters, multiplexers, and decoders. Design and draw
the circuit for a device that will multiply a two-bit (plus sign)
sign-magnitude quantity by 3. Note: You don't have to use an
adder.
18. Using the four-bit adder and any additional circuits you want,
design a circuit that will multiply a number X2X1XO (assume
signed magnitude with the sign bit handled elsewhere) by 2.5.
The result is to be rounded to the next highest integer if the
product results in a fractional part.
19. Design the circuit whose block diagram is shown in Figure
1O.P1, using decoders, counters, MUXs, FFs, and so on. The
circuit has as inputs a clock and four select lines. The select
lines determine the number by which the input clock fre­
quency is divided. The ou~put is a clock divided by the select
line value. If F is the frequency of the input clock and D is the
binary value of the select lines, the output frequency,J, is FI
(D + 1).

FIGURE lO.Pl
CLOCK (F) - - - t > i CLOCK (0: 1)

r o
20. Design a two-bit counter that counts up one count at a time
when C = 0 and counts up two counts per clock when C = 1.
356 CHAPTER TEN Introduction to Counters, Registers, and R TL

21. Using a four-bit universal-shift register, design a 12-bit, serial­

in, parallel-out, left-shift register.
22. Using a four-bit universal-shift register, design a 12-bit, paral­
lel-in, serial-out, right-shift register.
23. Using a four-bit universal-shift register, design a 16-bit, uni­
versal-shift register.
24. Use four-bit binary counters in parallel and assorted gates to
realize a divide-by-39 counter. Show the circuit configuration.
25. Design a typical stage that implements the following logic
microoperations:
a. Ql: A ~A V 13 b. PI: A ~A V B

Q2: A ~A 1\ B P2: A ~A 1\ 13

Q3: A ~A VB P3: A ~A V Ii

Q4: A ~A 1\ B P4 : A ~A 1\ B

Ps: A ~A EB B
26. Describe the function and characteristics of the counter circuit
of Figure 10.P2.

FIGURE lO.P2

J1 01

Kl 01

27. Obtain the universal state diagram for a four-bit, shift-left

feedback register.
28. Obtain the universal state diagram for a four-bit bidirectional
feedback register.
29. Design the additional circuitry that would be necessary in the
table clock design of Example 10.7 so that the external LEAP
input would not be needed. The table clock circuitry wou'ld
have to take into consideration the consequence of a leap year
on its own.
30. Design a wristwatch using BCD counters instead of biliary
counters. Comment on this design after comparing it with that

! 32.
of Example 10.7.
31. Design the complete data and control units for the algorithm
of Problem 25a.
Design the complete data and control units for the algorithm
of Problem 25b. _
 Suggested Readings 357

33. Using a seven-segment decoder and a display driver, design

an electronic die as illustrated in Figure lO.P3. With equal
probability the die should display the digits 1 to 6 on the
seven-segment display when a switch is depressed and
released. You may use a 1 MHz oscillator and a make/break
push-button switch.

FIGURE lO.P3 +5 V

1kH
,
Die
7
_I
1- ~
... '
• •

Awwal, A. A. S., and Karim, M. A. "A digital pseudo-random number

Suggested Readings generator scheme." Paper presented at the Proceedings of the Twenty­
Eighth Midwest Symposium for Circuits and Systems, Louisville, Ken­
tucky, August 1985.
Barbacci, M. R. "A comparison of register transfer languages for describ­
ing computers and digital systems." IEEE Trans. Compo vol. C-24
(1975): 137.
Bell, C. G.; Eggert, J. L.; Garson, J.; and Williams, P. "The description
and use of register transfer modules (RTM's)." IEEE Trans. Compo voL
C-21 (1972): 495.
Berndt, H. "Functional microprogramming as a logic design aid." IEEE
Trans. Compo voL C-19 (1970): 902.
Chambers, W. G., and Jennings, S. M. "Linear equivalence of certain
BRM shift-register sequences." Elect. Lett. vol. 20 (1984): 1018.
Dadda, L. "Composite parallel counters." IEEE Trans. Compo vol. C-29
(1980): 942.
David, R. "Testing by feedback shift register." IEEE Trans. Compo vol. C­
29 (1980): 668.
.
Divan, D. M.; Hancock, G. C.; Hope, G. S.; and Burton, T. H.
~

1:'.

"Microprogrammable sequential controller." lEE Proc. E., Compo & Dig.

Tech. vol. 131 (1984): 201.
Dormido, S., and Canto, M. A. "Synthesis of generalized parallel coun­
ters." IEEE Trans. Compo vol. C-30 (1981): 699~ .
Dormido, S., and Canto, M. A. "An upper bound for the synthesis of gen­
eralized parallel counters." IEEE Trans. Compo vol. C-31 (1982): 802.
Downing, C. P. "Proposal for a digital pseudorandom number generator."
Elect. Lett. vol. 20 (1984): 435.
Hayes, J. P. Digital System Design and Microprocessors. New York: McGraw­
Hill, 1984.
Hill, F. J., and Peterson, G. R. Digital Systems: Hardware Organization and
Design. 2d Ed. New York: Wiley, 1978.
358 CHAPTER TEN Introduction to Counters, Registers, and RTL

Hill, F.]., and Peterson, G. R. Digital Logic and Microprocessors. New York:
Wiley, 1984.
Kline, R. Structured Digital Design Including MSI! LSI Components and Micro­
processors. Englewood Cliffs, N.J.: Prentice-Hall, 1983.
La, H. Y.; Lu,]. H; and Aoki, Y. "Programmable variable-rate up/down
counter for generating binary logaritluns." lEE Proc. E'J Compo & Dig.
Tech. vol. 131 (1984): 125.
Manning, F. B., and Fenichel, R. R. "Synchronous counters constructed
entirely of J-K flip-flops." IEEE Trans. Compo vol. C-25 (1976): 300.
Rhyne, V. T. "Serial binary-to-decimal and decimal-to-binary conver­
sion." IEEE Trans. Compo vol. C-19 (1970): 80B.
Swartzlander, E. E. "Parallel counters." IEEE Trans. Compo vol. C-22
(1973): 1021.
Tien, P. S. "Sequential counter design techniques." Compo Des. vol. 10
(1971): 49.
Vasanthavada, N. S. "Group parity prediction scheme for concurrent test­
ing of linear feedback shift registers." Elect. Lett. vol. 21 (1985): 67.
Winkel, n, and Prosser, F. The Art if Digital Design-An Introduction to Top~
Doum Design. Englewood Cliffs, N.J.: Prentice-Hall, 1980.