Introduction to Combinatorial Logic

Objectives:
1. Ability to read simple logic diagrams
2. Ability to design simple combinatorial circuits
3. To introduce basic building blocks of more complex systems
Materials:
1. Samples of logic technologies
2. Projectable of logic axioms
3. Circuit Sandbox for demos - including prebuilt circuits - Inverter, Two Input
Gates, 4 Input NAND, A xor BC (five parts), 4 bit adder, decoder, demux, mux
4. Setup with scope, etc. to demonstrate physical properties of gates (needed for
final session devoted to this topic)
I. Introduction
A. From physics we learn that all the complexity of the universe is built up
from combinations of certain fundamental building blocks: electrons,
protons, neutrons (or quarks if you will.) In like manner, all the
complexity of digital computers is built up from fundamental building
blocks called GATES.
1. A gate is an electronic circuit having one or more inputs and one output,
each of which can be in one of two states, commonly called 0 and 1.
Further, the output at any time is some function of the inputs.
2. Gates can be realized in many different ways, but we will say little
about this. We should note that the generations of computer hardware
are largely distinguished by the technology used to realize gates: relays
(0); vacuum tubes (1); transistors (2); integrated circuits (3); very large
scale integration (4). With all pre-IC technologies, a gate would consist
of a number of discrete components wired together on a circuit board
(cf sample transistor board); with IC technologies, one or more gates
are completely contained on a silicon chip. With todays technology,
we may have one chip containing millions of gates.

B. For our purposes, we will think of a gate as a black box having one
or more inputs plus a single output.
1. The voltage at the output of the gate is some function of the input
voltages.
2. Example: a two-input gate
Vin1
Vout = f(Vin1, Vin2)
Vin2

3. Gates are designed so that they respond to two discrete voltages one of which is associated with the value 0 (false) and the other
with the value 1 (true). With the kinds of gates we will be be using
in lab (TTL and CMOS), the value 0 (as input or output) will
normally be represented by a voltage level close to 0. The value 1
will normally be represented by a voltage level of about 5 volts. (In
more recent computers, the there has a been a trend to use a lower
power supply voltage - and hence lower value for 1 - to conserve
power - e.g. 3.3 to as low as 1.2 Volts).
4. In actual practice, any gate technology will accept a RANGE of
voltages for each logic value. For example, with TTL gates of the
kind we will be using in lab, the following rules hold:
a) Any voltage between 0 and 0.8 Volts is regarded as low, and is
interpreted as a zero.
b) Any voltage between 2 and 5 volts is regarded as high, and is
interpreted as a one.
c) Any voltage between 0.8 and 2 volts, or if an input is not
connected at all, constitutes an undefined input. The output of
any gate which has any undefined input can be 0, 1, or
undefined itself. (This may be represented by the symbol "X")

d) Any voltage less than 0 or greater than 5 volts is illegal, and

may physically destroy the gate.
e) This turns out to be a fundamental distinction between digital
and analog devices. Because any value in a certain range is
interpreted as a particular discrete value (0 or 1), digital systems
have a high tolerance for electrical noise. On the other hand,
when digital systems fail, they tend to fail catastrophically.
(Contrast the signal on an old VCR tape versus a DVD disk;
consider also the effect of a scratch on a DVD.)
5. The output of a gate whose inputs are all defined will normally
always lie in either the defined 0 or 1 range.
a) However if the outputs of two or more gates are connected,
with at least one trying to output a 0 and the other a 1, then the
output is itself is undefined.
b) The exception to this statement is that there is a type of gate
called a tri-state gate whose output can be in one of 3 states: 0,
1, or floating (denoted by the symbol "Z"). If the outputs of
two or more tri-state gates are connected, and only one is
actually output a 0 or 1 (the rest are floating), then the value of
the output signal will be the one specified by the non-floating
gate. (We will see examples of this later in the course.)
6. Actually, the convention we have been describing is one of two possible
options, known as positive logic. It is also possible to adopt the
convention of letting a low voltage (between 0 and 0.8 volts) represent a
1, and a high voltage (between 2 and 5 volts) represent a 0. This yields
negative logic - but we won't discuss this further now - however, we will
run across examples where this is used in later labs.
7. To keep our discussion simple, we normally describe the inputs
and outputs of a gate as being 0 or 1 (or sometimes L = low or H
= high, rather than worrying about the specific voltages.

C. We will represent the behavior of a gate by a table of combinations or truth

table which shows the output of the gate for each possible combination of
inputs. For example, a 2 input AND gate has truth table:
A
Y

Note: we are using a nondescript box.

Later we will see that there is a special
shape for certain types of gate.

B
Inputs
B
A
0
0
0
1
1
0
1
1

Output
Y
0
0
0
1

1. Note that for a gate with 1 input there are 4 possible truth tables,
representing:
- a gate whose output is always 0, regardless of input
- a gate whose output is always 1, regardless of input
- a gate whose output is the same as its input
- a gate whose output is the opposite of its input:
A

Input
A
0
1

Output
Y
1
0

Of these, only the last two are of any interest at all, and he last one
is most useful. (The first two are sometimes called stuck at 0 and
stuck at 1, and typically arise from a gate being defective for
some reason.)

a) The gate whose output is the same as its input is often used in
situations where it is necessary to amplify a signal to drive a
number of gates. Such a device is called a BUFFER:
symbol:
b) The gate whose output is the opposite of its input is called an
INVERTER. Two symbols can be used for this function:
or
Note use of the "bubble" to symbolize inversion of the logic
sense. In the first case, the symbolism is that the output of the
gate is the opposite of that from a simple buffer. In the second
case, the symbolism is that the output of the gate is the same as
that of a simple buffer whose input has been inverted.
Obviously, the two symbols describe two ways of looking at
the SAME BEHAVIOR, and so can be used interchangeably to
describe the same physical circuit.
DEMONSTRATE (File Inverter)
2. For a gate with 2 inputs, there would be 16 possible truth tables.
(For each of the 4 combinations of the 2 inputs, we could choose
one of two outputs. Thus, there are 24 possible truth tables.
Again, only a limited number are of interest - we defer discussion
until later.)
3. In general, for a gate with n inputs, how many truth tables are
possible?
There are 22n possible truth tables - e.g. for 3 inputs there are 223
= 28 = 256 possible truth tables.

II. Boolean Algebra

A. To describe the behavior of networks of gates, we use a notational
system called boolean algebra (or switching algebra). In this system,
we have only two values, 0 and 1, and three primitive operations:
1. Inversion or complementation, written or A'. (Note - in these
notes I will use the symbol A' because it's easier to type! The book
consistently uses - pronounced "A-bar"):
A

A'

0
1
1

2. Addition or Logical or, written A + B or A B (I'll use + in these

notes, which is also the notation the book uses)
A

A+ B

0
0
1
1

0
1
0
1

0
1
1
1

Note: 1 + 1 = 1!

3. Multiplication or Logical and, written A B or A * B or A B or

AB (I'll use in these notes. The book doesn't use any special
symbol - just writes something like AB):
A
0
0
1
1

B
0
1
0
1

A B
0
0
0
1

B. As in any algebraic system, more complex operations can be built up

out of the primitive operations, with operator precedence rules or
parentheses for grouping - e.g:
6

A' + AB = (A') + (AB) - truth table:

A'

AB

A'+AB

0
0
1
1

0
1
0
1

1
1
0
0

0
0
0
1

1
1
0
1

C. As in other algebras, there are certain properties that govern the behavior
of more complicated expressions, such as associativity, commutativity, and
distributivity. However, whereas these are axioms for most algebraic
systems, for boolean algebra they are theorems since any one of them can
be proved from the definitions above by exhaustion (perfect induction).
Here are some of the key properties:
(Note: there are several others that appear in the book but are omitted here)
PROJECT`
A, B, C:
Identity:

A 1 = 1 A=A
A+ 0 =0 +A=A

Null Element:

A 0 = 0 A= 0
A+ 1 =1 +A= 1

Idempotence:

A+A=A
AA=A

Involution:

(A')' = A

Complements:

A A' = 0
A + A' = 1

Commutativity:

A+ B = B +A
A B = BA

Associativity:

A + (B + C) = (A + B) + C
A (B C) = (A B) C

Distributivity:

A (B + C) = (A B) + (A C)
A + (B C) = (A + B) (A + C) !!

Demorgan's
Theorems:

(A + B)' = A' B'

(A B)' = A' + B'
8

D.As an illustration of proof by perfect induction, we prove the first form

of Demorgan's theorem:
A

A+B (A+B)'

A'

B'

A'B'

0
0
1
1

0
1
0
1

0
1
1
1

1
1
0
0

1
0
1
0

1
0
0
0

1
0
0
0

Notice that, for all possible values of A and B, the (A+B)' and A'B'
columns are the same - implying that (A+B)' = A'B' for all A and B
III. Realization of Boolean Functions - Basic Types of Gates
A. Crucial to the design of computer systems is the ability to realize
circuits composed of gates whose output corresponds to some boolean
function of its inputs. To do this, we usually use certain types of gates
as building blocks, with more complex functions realized by
combinations of gates in which the output of one gate becomes an
input to another.
B. While we could theoretically implement any truth table as a hardware
primitive, most designs are built around a limited number of basic
functions implemented in hardware. We have met one already, the
inverter. Two-input functions of interest include:
1. 2-input AND:
A
Y
B

Y=A B

2. 2-input OR:
A
Y

Y=A+B

3. 2-input NAND:
A
Y
B

Y = (A B)'

4. 2-input NOR:
A
Y
B

Y = (A + B)'

DEMONSTRATE each of the above using Circuit Sandbox. (File

Two-Input Gates)
5. The latter two are of particular interest, for two reasons:
a) Most of the transistor circuits used in realizing gates have the
effect of inverting the signal passing through them. Thus, NAND
and NOR can generally be realized more simply than AND or OR.
b) NAND and NOR are logically complete. Given either one, it
is possible to realize any boolean function. Examples
(1)

Invert realized using NAND:

Y = (A A)' = A'
(2) AND realized using NAND:
A
Y
B

Y = ((A B)')' = A B

10

(3) OR realized using NAND:

A

Y = (A' B')' = A + B
(by DeMorgan's theorem)

DEMONSTRATE EACH OF THE ABOVE (Create on screen)

6. For any particular gate technology, usually one of the latter two
will be more easily realized. In the case of the TTL technology we
will be using in lab, this is NAND. For consistency, we will work
most of our examples for the rest of this unit using NAND.
7. Note that DeMorgan's theorem and the above construction suggests an
alternate symbol for NAND:
A
Y
B

Y = A' + B' = (A B)'

This symbol says that the output of a NAND gate is the same as that of
an or gate presented with the inverse of the inputs presented to the
NAND. (The first symbol said that the output of a NAND gate is the
inverse of that of an AND gate presented with the same inputs.)
Note that both symbols describe exactly the same gate - a NAND gate
may be drawn either way. By convention, the symbol that is chosen is
the one that most clearly represents the intended use of the gate - e.g. the
above diagram for an OR implemented with NANDS is more
conventionally drawn as follows:

11

OR realized using NAND and inverters:

C. Two other functions of two inputs are of some interest, though less
than the previous four
1. 2-input XOR:
A
Y

Y=A B

Truth table:

0
0
1
1

0
1
0
1

0
1
1
0

DEMONSTRATE (using file Two Input Gates)

2. 2-input XNOR (Coincidence)
A
Y

Y = (A B)'

Truth table:

0
0
1
1

0
1
0
1

1
0
0
1

Note that A XNOR B is true iff A = B. (XNOR and XOR are

complements)

12

3. Both can be realized using other types of gates. For example, the
following is a realization of XOR using NAND - assuming that
both the inputs and their complements are available. (If not, two
inverters would also be needed - however, for reasons we will see
later, it is often that case that both the uncomplemented and
complemented form of each input are readily available).
A
B
Y
A
B

Z = (A'B)'' + (AB')'' = A'B + AB' = A B

DEMONSTRATE (Create on screen)
D. We often find gates with more than two inputs, but they are usually
extensions of the types of 2-input gates we have been looking at. For
example, 4-input NAND:
A
Z

B
C
D

Truth Table:

13

E.

A B C D Z
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1

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

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

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

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

1. DEMONSTRATE using Circuit Sandbox (File 4 Input NAND)

2. Observe that, interestingly, this circuit outputs a zero just when its
four inputs are all 1. The fact that transistor circuits naturally
inject a level of inversion into their computation - making NAND
easier to implement than AND - leads to the use of negative logic
(true = 0, false = 1) in some situations. In this case, we could think
of our gate as a sort of and gate whose inputs use positive logic
and whose output uses negative logic !

14

IV.Realizing Any Boolean Function Using A Network of Gates

A. Observe that, since we have gates that are directly equivalent to the
primitives of boolean algebra, we can convert any boolean expression
into a network of gates that realizes it.
Example: Y = A (B C)
DEMONSTRATE (File A xor BC)
B. Actually, we can realize any boolean function using just AND and OR
gates. Observe that any boolean function can be written in sum-ofproducts form by using its truth table.
Example: the above
A B C

0
0
0
0
1
1
1
1

0
0
0
1
1
1
1
0

0
0
1
1
0
0
1
1

0
1
0
1
0
1
0
1

1. Find each row in the truth table where the output is 1. This will
contribute one term to the sum. Thus, this function can be
represented as a sum of four products:
2. Each product will contain each of the inputs either plain or
complemented. An input will appear plain if the row in question
contains a value of 1 for that input, and complemented if it
contains a 0. Thus, the fourth row in the truth table (the first
containing a 1 output) gives rise to the term A'BC.
A term that is the product of the true or complemented form of all
the inputs is called a minterm.
15

3. A function equivalent to our original function is therefore

Y = A'BC + AB'C' + AB'C + ABC'
a) Note that each term is 1 for exactly one combination of inputs
and 0 for all others.
b) Each term therefore gives us one of the 1's of the function.
c) This form of expression is called sum-of-products form.
d) This can then be realized by using a two-level AND-OR
network:
DEMONSTRATE - Second part of File A xor BC
This realization is called a two-level AND-OR realization, since
(1) The input must pass through two gate levels before arriving
at the output. This is of some significance, since all gates
have a certain finite propagation delay. Note that, apart
from building a special-purpose gate, two logic levels is the
best one can do for most functions.
(2) It is called an AND-OR realization because the first level
consists of AND gates, and the second level is a single OR
gate.
C. A boolean function in sum of products form can also be realized by
using just one type of gate: the NAND gate.
1. Realize each term by using a NAND gate with as many inputs as
there are inputs to the overall function. Each gate input goes either
to one of the function inputs or its complement as discussed above.
The NAND gate's output is 0 just when the term itself is 1.
(Negative logic)
2. Sum all of the terms using a NAND gate (represented - for clarity in its alternate form) with as many inputs as there are terms.

16

DEMONSTRATE - NAND version of the above (Third part of file)

Note that the only difference between this and the previous
realization is adding "bubbles" to turn the ANDs and the ORs into
NANDs.
3. We call this circuit a two-level NAND-NAND realization.
D. While the method outlined above gives a realization that cannot
generally be improved on in terms of number of levels, it is not
optimal in terms of number of gates.
1. Observe that our sum of products form can be simplified by the
properties of boolean algebra as follows:
A'BC+AB'C'+AB'C+ABC' = A'BC + A(B'C' + B'C + BC')
= A'BC + A(B'C' + B'C' + B'C + BC')
= A'BC + A(B'C' + B'C + B'C' + BC')
= A'BC + A(B'(C'+C) + C'(B'+B))
= A'BC + A(B'1 + C'1)
= A'BC + A(B' + C')
= A'BC + AB' + AC'
2. This yields a simplified two-level NAND-NAND realization
DEMONSTRATE - Fourth part of file
3. This simplification would result in a considerable cost savings in
construction.
a) The original circuit needed:
4 3-input NAND gates
1 4-input NAND gate
b) The new circuit requires:
2 3-input gates
2 2-input gates

17

This involves both fewer gates and simpler gates, and so would
be easier to build from SSI chips (as we will do in lab) or on a
VLSI integrated circuit.
E. For functions of four variables or less, this simplification can be done
simply by the use of a Karnaugh map.
1. Example, for the above:
AB/C

00
01
11
10

0
1
0
1

0
0
1
1

(Note order of rows!)

2. Example for four variables:

Y = A'BC'D' + A'B'C' + A'C'D + AB'CD + AB'C' + AB'D'
AB/CD

00 01

11

00
01
11
10

1
1
0
1

0
0
0
1

1
1
0
1

0
0
0
1

10

Simplified form: A'C' + AB'

3. In using a Karnaugh map, we cover all of the 1's by using maximal
subcubes of one of the following forms:
8s completely covering two adjacent rows or two adjacent
columns
4's completely covering one row or one column
squares of 4 - including ones that wrap around top and bottom or
left and right, and ones that involve the four corners
18

2s that cover two adjacent squares horizontally or vertically including ones that wrap around top and bottom or left and right
4. The method works because the values are ordered in such a way that any
two horizontally or vertically adjacent entries differ in exactly one input,
whose value is irrelevant if both squares contain a 1.
5. In setting up functions to be minimized using a Karnaugh map, it is often
helpful to use a different notation for the function. Basically, we
represent each 1 of the function by a decimal number that is the
equivalent of the binary number representing the values of the inputs.
a) Example: Y = A'B'CD + AB'C'D can be written
Y(A,B,C,D) = (3, 9)
because A'B'CD corresponds to an input pattern of 0011 =
decimal 3 and AB'C'D corresponds to an input pattern of 1001
= decimal 9
b) In doing this, one must be sure to include all the variables in
each term.
Example: Y = A'CD + ABC
must be written as Y = A'B'CD + A'BCD + ABCD' + ABCD
which yields (3, 7, 14, 15)
F. Often, in designing a logic circuit, it happens that certain input
patterns are known to not be possible. Thus, the output produced for
these patterns is irrelevant. We refer to this as a don't care condition.
1. Example: a 7-segment decoder translates the binary representation of
a decimal digit into 7 outputs corresponding to the 7 segments of an
LED display. Consider just the output for the bottom bar:

19

This needs to be on when the input is 0, 2, 3, 5, 6, or 8, and off when

the input is 1, 4, 7, or 9. We don't care what its state is for inputs
10-15, since these don't correspond to any decimal digit and thus
presumably cannot occur. Thus, our function for the bottom bar is
Y(A3,A2,A1,A0) = (0, 2, 3, 5, 6, 8) + d(10, 11, 12, 13, 14, 15)
(where d stands for dont care)
2. In constructing the map, we put 1's for the 1's of the function, d's
for the don't cares, and 0's for everything else.
Example:
A3 A2 / A1 A0 00

01 11 10

00
01
11
10

0
1
d
0

1
0
d
1

1
0
d
d

1
1
d
d

3. In constructing a covering, we must cover ALL the one's, but only

cover d's which help us to construct bigger sub-cubes. For
example, in the above it would be convenient to cover four of the
d's, while leaving two uncovered. This means that the circuit will
output 1 for inputs 10, 11, 13, and 14; but 0 for inputs 12 and 15.
But since those input patterns do not represent legal decimal digits,
we really dont care what the circuit would do in those cases!
A3 A2 / A1 A0 00

01 11 10

00
01
11
10

0
1
d
0

1
0
d
1

1
0
d
d

1
1
d
d
20

4. (If time) do another segment as an example

G. When working with NOR logic, it is convenient to use a mirror-image
technique:
1. Working directly with the truth table, we can represent function as a
product of sums, derived by looking at the zeroes of truth table. Each
input appears in ordinary form if the row contains a zero;
complemented form if it contains a 1. Example: our earlier example
Y = A (B C). We look for the zeroes of the truth table:
A B C Y
0
0
0
0
1
1
1
1

0
0
1
1
0
0
1
1

0
1
0
1
0
1
0
1

0
0
0
1
1
1
1
0

This yields a product of sums form in which each factor is zero for just
one row of the truth table; of course, the overall function is zero just
when any factor is zero, and one when all the factors are one:
Y = (A + B + C)(A + B + C')(A + B' + C)(A' + B' + C')
2. We can realize with a product of sums with one NOR gate for each
factor plus one NOR (drawn in dual form) to multiply all factors.
DEMONSTRATE: Fifth part of A xor BC file
3. We can also use Karnaugh-map minimization, we covering the
ZEROES, and put the variable in the expression in straight form if
it is a zero and complemented if it is a one.
21

Example: map for the above:

AB/C 0 1
00
01
11
10

0
0
1
1

0
1
0
1

Simplified expression is (A + C)(A + B)(A' + B' + C')

V. More Complex Building Blocks
A. The individual gates we have been designing with thus far were first
made available (in the 1960's) on chips which were classified as Small
Scale Integration (SSI) devices - but they are still available in this
form today. They typically consist of some number of gates of the
same type in a single 14-pin package. In each case, the determining
factor on the number of gates is the number of available pins (12
after allowing two for power and ground). It is possible to obtain, on
one 14-pin chip, any of the following:
1. 6 inverters or buffers
2. 4 2-input gates
3. 3 3-input gates
4. 2 4-input gates
5. 1 8-input gate
Note that these devices really make very inefficient use of
integrated circuit technology; even a chip produced using a very
low-end manufacturing process could contain many more gates
than this.

22

B. Very early in the history of integrated circuit logic, it was realized

that one could obtain a much higher level of integration by building an
interconnected network of gates on a chip, with pins needed only for
overall input to and output from the network. A number of basic
functions were identified that prove to be highly useful, and chips
implementing them came to be known as Medium Scale Integration
(MSI) devices.
C. Today, of course, whole CPU's are implemented on a single chip,
often containing millions of gates. However, the basic building
blocks developed early in the history of IC technology remain useful
as part of the overall design of a more complex system. We are going
to learn about a few of them now - not because they are terribly useful
today in their own right, but because they continue to serve as basic
building blocks of more complex systems.
VI.Decoders, Demultiplexers, and Multiplexers
A. A decoder is a circuit that has n inputs and 2n outputs (for some small
integer n - typically <= 4 with discrete chips). At any given time,
exactly one of the outputs is active, as selected by the inputs.
Example: a 1 out of 8 decoder has 3 inputs and 8 outputs. Based on
the value of the inputs, exactly one of the outputs is active.
1. A typical application is in specifying which one of a group of
similar devices is to respond on a particular operation.
Example: A disk controller is used to control 4 disk drives. When a
command is sent to the controller, 2 bits of the command are used
to specify which of the 4 drives is to perform the operation.
00 = drive 0
01 = drive 1
10 = drive 2
11 = drive 3

23

A one out of 4 decoder could process these two bits to select one
of the 4 drives:
drive
select
bits in
command

drive
drive
drive
drive

0
1
2
3

DEMONSTRATE using Circuit Sandbox file Decoder

2. One place decoders are often used is in memory systems to select a
specific chip based on particular bits of the address. We will see
examples of this much later in the course.
3. A typical decoder circuit, built using NAND gates.

24

a) Note that the selected output is low (0), and the other three outputs
are high (1). This is an example of negative logic. Decoders are
typically designed this way because a level of inversion is inherent
in the typical transistor circuits used to implement gates.
b) Devices that are designed to be used with decoders (eg memory
chips) often have active low enable inputs as a result.
B. A circuit that is very similar to a decoder is a demultiplexer.
1. A demultiplexer has one additional input (often called enable).
When enable is active, the device functions just like a decoder; but
when enable is inactive, none of the outputs is selected.
2. One place where demultiplexers are useful is when chaining
decoders together to produce a bigger decoder - e.g. a 1 out of 16
decoder (4 selection inputs, 16 outputs) might be made using 5 1
out of 4 demultiplexers - with one used to select which of the other
four is enabled, based on two bits of the selection value, with the
other two bits going to all four of the other demultiplexers.
3. A demultiplexer uses the same circuit as a decoder, but with one
extra input to each gate - called "enable".

25

DEMONSTRATE using Circuit Sandbox - file demux

C. The final special building block we will consider - and one we will see
quite a bit of when we discuss the internals of a CPU- is a device
called a multiplexer.
1. A multiplexer (MUX) has n selection inputs and 2n data inputs (for
some small integer n - typically 2-5), and one output. It selects one
of the 2n data inputs to appear on its output, as determined by its n
selection inputs.
2. One typical application in CPU's arises when a group of bits in an
instruction is used to select one of several possible registers to
provide data for an operation.

26

Example: In the Z80, quite a number of instructions use a 3-bit

field to select one of 8 registers (B, C, D, ...) to provide data. For
example, there is a family of 8-bit add instructions that use the
value of 3 bits in the instruction to determine which register gets
added to A:
1 0 0 0 0 r r r - where r r r designates which register to add:
000
001
010
011
100
101
110

add contents of B
add contents of C
add contents of D
add contents of E
add contents of H
add contents of L
add contents of a memory location

This can be implemented by using one MUX per bit, with the
selection inputs tied to the appropriate field of the instruction
register.
to adder input7

to adder input6

B7

(from bits
6..2)

to adder input1

to adder input0

B6

B1

B0

C7

C6

C1

C0

D7

D6

D1

D0

E7

E6

E1

E0

H7

H6

H1

H0

L7

L6

L1

L0

mem7

mem6

mem1

mem0

A7

A6

A1

A0

(to bits
6..2)

register select bits in instruction

27

3. A multiplexer could be built using a decoder with n selection

inputs, plus 2n and gates and one or gate. In fact, though, a
multiplexer is usually built by folding the and gates of the decoder
with the final layer of and gates.
data
input 0
data
input 1

data
output

data
input 2
data
input 3

DEMONSTRATION using Circuit Sandbox (File MUX)

4. One interesting application for a multiplexer is this: any n-input
boolean function can be realized using an n selection-line MUX.
The method is this:
a) The inputs (which we will call A0..An-1) are applied to the
selection lines of the MUX.

28

b) Each possible combination of A0..An-1 selects one of the 2n

data inputs of the MUX and routes it to the output. We connect
the corresponding MUX input to 1 if the truth table shows a 1
output for that row, and to 0 if it shows a 0 output:
EXAMPLE: Realize the following truth table using a MUX.
(This is the same example function we have used in a number
of places)
A B C

0
0
0
0
1
1
1
1

0
0
0
1
1
1
1
0

0
0
1
1
0
0
1
1

0
1
0
1
0
1
0
1

0
0
0
1
1
1
1
A B C

c) Actually, it turns out one can also realize an arbitrary function

of n input variables using an n-1 selection input 2n-1 data input)
MUX plus an inverter. (This is left as an exercise to the
student)
29

VII.Full Adders
A. One example of such a building-block is called a full adder.
Note that the boolean or operation, represented by +, is not the same as
addition. For example, 1 + 1 = 1 if + means or, but 10 if it means add.
B. To perform arithmetic addition, we use a building block known as a
full adder. This has two outputs (sum plus carry) and three inputs
(two bits to be added, plus a carry in from the previous bit). The
symbol that is typically used is:
A
B
Cin

Cout

Which implements the following truth table:

A

Cin

Cout

0
0
0
0
0
(The sum is 1 if an odd number of
0
0
1
1
0
inputs are 1; the carry is 1 if at
0
1
0
1
0
least two of the inputs are 1)
0
1
1
0
1
1
0
0
1
0
1
0
1
0
1
1
1
0
0
1
1
1
1
1
1
C. The following equations describe the two outputs as functions of the
inputs:
= A B Cin

Cout = AB + (A + B)Cin

(+ = or)

However, the latter is generally implemented instead as

Cout = AB + (A B)Cin
(which is equivalent) because we've already computed and dont
need to use an extra gate to compute + instead.

30

D. This can be implemented by the following network of gates:

E. Full adders can be interconnected to produce adders that handle numbers of

any given number of bits. For example, the following shows how four fulladders could be interconnected to yield an adder that adds two 4-bit numbers
to produce a 5-bit sum (or 4 bits plus carry out if you prefer). This sort of
interconnection is called a ripple-carry adder (so-called because, in the worst
case, a carry generated at stage 0 may ripple through stages 1, 2, 3 - e.g.
when adding 1011 + 0101.
A3
B
3

Cout (overall)

A2
B2

A1
B1

A0
B0
0

0
31

DEMONSTRATE using Circuit Sandbox - file 4 bit adder

VIII.The Physical Characteristics of Gates
A. In our discussion thus far, we have treated gates as black boxes - i.e.
idealized devices whose inputs and outputs are one of two discrete
values, and whose output at any time is purely a function of its inputs
at that time. In reality, of course, gates are physical devices, and their
actual behavior differs a bit from the ideal behavior we have
discussed. We want to look briefly at a few of the ways that actual
physical characteristic impact on functionality.
B. Noise Immunity
1. We have talked as if gate inputs and outputs are simply 0 and 1. In
reality, the input to a gate is a voltage, with any voltage in a certain
range being treated as 0, and any voltage in a different range being
treated as 1, and any other voltage being prohibited.
Example: TTL gates of the sort we are using in lab:
0 .. 0.8 volts = 0
2.0 .. 5.0 volts = 1
2. In reality, there is always a certain amount of electrical noise in
the circuits of which a gate is a part.
a) Some of this comes from the gates themselves. Ideally, when a
gate switches from - say - 0 to 1, its output should look like
this:

In reality, its output typically looks more like this:

32

This behavior (known as ringing) is a consequence of the

physical transitions occurring inside the transistors from which
the gate is constructed.
b) Moreover, electrical noise is often induced by other nearby
gates. When a nearby gate changes state, a gate that is not
changing may still exhibit some fluctuation in its output.

c) Also, other nearby devices - especially those that spark (e.g.

motors), radio signals, etc. can also result in fluctuations in the
actual voltage on a wire connecting two gates.
3. As long as the electrical noise does not cause the output of a gate
to move outside the range that is reliably recognized as 0 or 1 (as
the case may be), it should have no effect on other gates. The
safety margin is known as the noise immunity of the gate.
Example: TTL gates are typically designed so that their zero
output is guaranteed to be no greater than 0.4 volt. Since another
gate will recognize any voltage less than 0.8 as a zero, there is a
margin for error of 0.4 volts. As long as electrical noise does not
cause the apparent output to rise more than this, the circuit will
behave reliably.
Noise becomes more of a problem as supply voltage drops.
4. However, if there is too much electrical noise present (due to
inattentiveness to certain issues in design), the circuit will not be
reliable. This is why digital systems typically incorporate various
kinds of metal shielding to keep out ambient signals, and why
33

circuit boards typically incorporate devices known as a capacitors

at the power pins of chips to minimize interaction between chips
through the power supply.
5. Digital systems designed for use in environments where there is a
large amount of electrical noise (e.g. systems exposed to nuclear
radiation, systems that control electrical devices such as motors)
require special hardening to prevent problems. Absent this, a
system that performed reliably in design may fail in surprising
ways in the field.
C. Fanout
1. We have designed networks of gates in which the output of one
gate may be connected to the input(s) of some number of other
gates. However, the number of gates one gate may drive is not
unlimited.
2. The term fanout refers to the maximum number of other gates of
the same type one output is designed to drive. For example, the
TTL gates we are using in lab are typically have a fanout of 10 one output of a particular gate may be connected to a maximum of
10 inputs of other gates.
3. Exceeding the design fanout can result in a reduction in the noise
margin; exceeding it sufficiently far can produce unreliable
behavior. (I.e. a gate may be outputting a zero, but other gates may
see the output as undefined and may interpret it as one instead.)
4. What does one do if a single output must drive more inputs than
the fanout allows?
For example, what if we are using gates designed for a fanout of 10
in a circuit where one output has to drive 32 inputs?

34

5. This is a case where we need to make use of a buffer - a gate

whose output is identical to its input:

What we do is have the gate in question drive some number of

buffers, each of whose output drives (up to) the maximum number
of inputs.
inputs of 8 gates
output
of some
gate

inputs of 8 gates
inputs of 8 gates
inputs of 8 gates

D. Propagation Delay
1. In our ideal description of gates, we assumed that the output at any
point in time reflects the inputs at that same time. In reality, as
physical devices, all gates exhibit a finite propagation delay, such
that when the input changes, the corresponding change at the
output occurs some amount of time later.
Example: the typical propagation delay for TTL gates of the sort
we are using in lab is about 10 ns.
2. Of course, when gates are interconnected in such a way that the
output of one gate is connected to the input of another, the overall
propagation delay for a network is the sum of the propagation
delays down the longest (in terms of time) path from input to
output in the network.
Example: consider the full adder we looked at earlier:

35

Suppose that the propagation delays for the and and or gates are
0.1 ns each, while that for the xors is 0.2 ns. (XOR requires a
more complex circuit). Then the delay from A or B to Cout is
max(0.1 + 0.1 ns, 0.2 + 0.1 + 0.1 ns) = max (0.2, 0.4 ns) = 0.4 ns.
3. This delay is very important; it is, in fact, one of the key factors in
determining how fast a digital system can run.
Example: Suppose a CPU is designed to run at 2 GHz. Then the
worst-case propagation delay down any path must be less than
1 / (2 GHz) = 1 / (2 x 109 sec-1) = 0.5 x 10-9 sec = 0.5 ns. If all
gates in the system have the same propagation delay, and the the
longest path in the system involves 10 gates, then an individual
gates propagation delay must be less that 0.05 ns.
4. As you know, CPU clock speeds have increased dramatically in
your life time (by a factor of better than 1000 : 1). This has been
made possible, in part, by a dramatic decrease in propagation delay
due to technological improvements. (There have also been
reductions due to techniques that have shortened the longest path)
Example: One could not build a 2 GHz CPU using TTL gates of
the sort we are using in lab, since the typical propagation delay for

36

a gate is about 10 ns - about 200 times as great as what is needed

for such a clock rate!
a) One key factor has been using processes that reduce the feature
size - the size of the individual components on the chip.
(1) Propagation delay is much less if a gate is designed only to
drive other gates on the same chip, as opposed to sending a
signal to another chip. (The issue is not so much the
physical distance, but rather the strength of the signal
needed to ensure sufficient noise immunity.)
(2) Physically smaller circuits can change state more quickly.
b) Another factor has been the use of lower voltages, which
reduces the swing between low and high, and, in particular,
the time needed to charge/discharge capacitance on the chip.
Of course, this comes at the cost of a reduction in noise
immunity, which calls for other compensating measures to
minimize extraneous signals.
5. While clock speed increases arising from propagation delay
decreases have been dramatic in the past few decades, there is
strong evidence that we havehit a wall.
Cf an Dr. Dobb's Journal entitled The Free Lunch is
Over (March 2005). The essential thesis is that techniques for
increasing clock speeds that have worked in the past are beginning
to hit fundamental limits, and that future increases in speed will
largely come through various forms of concurrency, rather than
improvements to raw speed. That seems to be the case today with
the rise of multicore processors. We will spend quite a bit of time
on this later in the course.

37