peak power delivered to a load, expressed
in watts. It is simply the maximum power
at any instant in time taken during the to-
tal time period being considered. This term
is often used when a device or application
is subjected to time-varying signals such as
pulse modulation.

peak power

the maximum value of the

square of the absolute value of a signal over
the duration of the signal; i.e., for the signal

x(t), max

|x(t)|

. The maximum instanta-

neous power of a signal.

peak signal to noise ratio (PSNR)

objec-

tive distortion measure of the difference be-
tween a discrete time signal

, defined over

[1 . . . n], and a degraded, restored or other-
wise processed version of the signal

ˆx

, de-

fined as

P SNR = 10log

max

i=1

− ˆx

)

PSNR differs from the standard signal to
noise ratio by using the square of the sig-
nal’s maximum value rather than its variance
in the numerator. In image coding, this value
is often assumed to be 255, for 8-bit images.

peaking generator

(1) a utility generating

unit, typically driven by a gas-fired turbine,
available to rapidly come on line when the
system demand reaches its highest levels.

(2) a generator used by a plant to reduce

the peak demand drawn in a given period of
time (typically during a one-month interval).

peaking unit

a generator used only to sup-

ply peak periods of electric power demand.

Pearl Street Station

the first investor-

owned electric utility plant, started by Edison
in 1882 in New York City. It provided low-
voltage, DC electric service to 85 customers
with an electric load of 400 lamps.

PEB

See

post-exposure bake

PEC

See

perfect electric conductor

pel

a picture element that has been en-

coded as black or white, with no gray scale
in between.

pellicle

a thin, transparent membrane

placed above and/or below a photomask to
protect the photomask from particulate con-
tamination. Particles on the pellicle are sig-
nificantly out of focus, and thus have a much
reduced chance of impacting image quality.

2000 by CRC Press LLC

penalty function

See

cost function

penstock

a water tube that feeds the tur-

bine. It is used when the slope is too steep
for using an open canal.

pentode

vacuum tube with five active

electrodes:

cathode, control grid, screen

grid, suppressor grid, plate.

per-unit system

a dimensionless system

for expressing each quantity in terms of a
fractional part of a “base” value, often the
nominal or rated value of the system. Typical
electrical calculations require four base quan-
tities (voltage, current, impedance, and ap-
parent power), any two of which may be cho-
sen arbitrarily. The per-unit system greatly
simplifies calculations in electrical systems
containing transformers with non-unity turns
ratios, making the voltage differences trans-
parent.

percent impedance

the per-unit impedance

expressed as a percentage on a certain MVA
and voltage base.

percent system

a variation of the per-unit

system in which the ratios expressing system
quantities are expressed as a percentage of
the base quantity.

perceptron

one of the earliest neural algo-

rithms demonstrating recognition and learn-
ing ability. Since the output of neuron is bi-
nary, the neuron can classify an input into
two classes,

A and B. The perceptron model

containing a single neuron is expressed as
follows:

z = 1, if

> T and z = 0,

< T where z is the output, x

an input,

is an element of the intercon-

nection weight matrix, and

T is a general-

ized threshold. If the interconnection weight
matrix is known, an input vector

) can be

classified into

A or B according to the result

z. On the other hand, the interconnection

weight matrix can be formed if a set of in-
put vectors

) and their desired outputs z

are known. The process of forming the inter-

connection weight matrix from known input–
output pairs is called learning. The percep-
tron learning is as follows. First, the weight
w

and the threshold

T are set to small ran-

dom values. Then the output

z is calculated

using a set of known input. The calculated

is compared with the desired output

t. The

change of weight is then calculated as fol-
lows:

δw

= η(t − z)x

where

η is the gain

term that controls the learning rate, which is
between 0 and 1. The learning rule deter-
mines that the corrected weight is

= w

(old) + δw

The process is repeated until

(t −z) = 0. The

whole learning process is completed after all
input–output pairs have been tested with the
network. If the combination of inputs are lin-
early separable, after a finite number of steps,
the iteration is completed with the correct in-
terconnection weight matrix. Various optical
systems have been proposed to implement the
perceptron, including correlators.

The original perceptron was a feedfor-

ward network of linear threshold units with
two layers of weights, only one layer of
which (the output layer) was trainable, the
other layer having fixed values.

perceptron convergence procedure

a su-

pervised learning technique developed for
the original perceptron. If the output

unit

i in the output layer is in error, its in-

put weights,

, are adjusted according to

= h(t

− y

, where

is the target

output for unit

i and h is a positive constant

(often taken to be unity);

is the output of

unit

j in the previous layer which is multi-

plied by weight

and then fed to unit

i in

the output layer.

perceptual coding

involves the coding of

the contextual information of the image fea-
tures by observing the minimum perceptual
levels of the human observer.

perfect code

a t-error correcting forward

error control block code in which the num-
ber of nonzero syndromes exactly equals the

2000 by CRC Press LLC

number of error patterns of t or fewer errors.
Hamming codes and Golay codes are the only
linear nontrivial perfect codes.

perfect electric conductor (PEC)

a con-

ductor that has infinite conductivity or zero
resistivity.

perfect reconstruction

the condition

states that the output of a filter bank is a de-
layed version of the corresponding input. In
other words, there are no aliasing and (phase
and magnitude) distortions for the output of
this filter bank.

perfect shuffle

interconnects that con-

nect sources 1

, 2, 3, 4, 5, 6, 7, 8 to detectors

, 5, 2, 6, 3, 7, 4, 8, respectively. The op-

eration divides 1

, 2, 3, 4, 5, 6, 7, 8 into two

equal parts 1

, 2, 3, 4 and 5, 6, 7, 8 and then

interleaves them. The size of the array must
be 2

. The array returns to its original or-

der after

n operations. When the option to

exchange pairs of neighboring elements is
added to a perfect shuffle network, any arbi-
trary permutation of the elements is achiev-
able.

performability

the probability that a sys-

tem is performing at or above some level of
performance,

L, at the instant of time t.

periodic convolution

a type of convo-

lution that involves two periodic sequences.
Its calculation is slightly different from that
of discrete linear convolution in that it only
takes summations of the products within one
period instead of taking summations for all
possible products. Circular convolution is
an operation applied to two finite-length se-
quences. In order to make its result equal
to that of the linear convolution, the two se-
quences have to be zero-padded appropri-
ately. These two zero-padded sequences can
then serve as periods to formulate two peri-
odic sequences. At this time, the result of
the periodic convolution is equal to circular
convolution.

periodic coordination

coordination pro-

cess occurring when a sequence of control
decisions is required over a given time in-
terval and these decisions are directly made
by the local units but the operation of these
units is periodically adjusted by the coordina-
tor; an example is operation of an industrial
process controller with several independent
regulators, which at specified time instants
are provided with new parameters or additive
compensation signals so as to achieve better
results from the overall point of view, for ex-
ample, to achieve better transient responses
to the changes of the free inputs.

periodic signal

a continuous time signal

f (t) with period T ≥ 0 such that

f (t) = f (t + T ) for all − ∞ < t < ∞ .

The smallest such

T is its fundamental pe-

riod. For example, the signal

f (t) = A sin(ωt + φ)

is periodic with fundamental period 2

π/ω.

Let

T be a positive integer. A discrete

time signal

f [k] is periodic with period T if

f [k] = f [k + nT ] for all integers k.

The smallest such

T is its fundamental pe-

riod. For example,

f [k] = cos[

k] is pe-

riodic with period 4.

On the other hand,

f [k] = cos(k) is not periodic, as a positive
integer

T does not exist to satisfy the defini-

tion.

If a signal is not periodic, then it is aperi-

odic.

periodic structure

structure consisting of

successive identically similar sections, sim-
ilarly oriented, the electrical properties of
each section not being uniform throughout.
Note that the periodicity is in space and not in
time. The analysis of infinite periodic struc-
ture is significantly simplified by the Flo-
quet’s theorem. See also

Floquet’s theorem

periodic waveform

phrase used to de-

scribe a waveform that repeats itself in a uni-
form, periodic manner. Mathematically, for

2000 by CRC Press LLC

the case of a continuous-time waveform, this
characteristic is often expressed as

x(t) =

x(t ± kT ), which implies that the waveform
described by the function

x(t) takes on the

same value for any increment of the

kT ,

where

k is any integer and the characteris-

tic value

T > 0, describes the fundamental

period of

x(t). For the case of the discrete-

time waveform, we write

x(n) = x(t ± kN),

which implies that the waveform

x(n) takes

on the same value for any increment of sam-
ple number

kN, where k is any integer and the

characteristic value integer

N > 0 describes

the fundamental period of

x(n).

peripheral

an ancillary device to a com-

puter that generally provides input/output ca-
pabilities.

peripheral adapter

a device used to con-

nect a peripheral device to the main com-
puter; sometimes called an I/O card, I/O con-
troller, or peripheral controller.

peripheral control unit

a device used to

connect a peripheral device to the main com-
puter; sometimes called an I/O card, I/O con-
troller, or peripheral controller.

peripheral controller

See

peripheral

control unit

peripheral device

a physical mechanism

attached to a computer that can be used to
store output from the computer, provide input
to the computer, or do both. See also

I/O

device

peripheral processor

a computer that

controls I/O communications and data trans-
fers to peripheral devices. It is capable of
executing programs much like a main com-
puter. See also

I/O channel

peripheral transfer

a data exchange be-

tween a peripheral device and the main com-
puter.

peripheral unit

a physical mechanism

attached to a computer that can be used to
store output from the computer, provide in-
put to the computer, or do both. See also

I/O

device

permalloy

a family of ferromagnetic al-

loys consisting of iron, nickel, and molyb-
denum that saturate at moderate flux density
levels and have a low coercive force.

permanent fault

a fault that remains in

existence indefinitely if no corrective actions
are taken.

permanent magnet (PM)

a magnet that

produces an external magnetic field by virtue
of the alignment of domains inside the ma-
terial and retains its magnetism after being
subjected to demagnetizing fields.

permanent magnet AC motor

a generic

term used to describe both permanent magnet
synchronous motors and brushless DC mo-
tors.

permanent magnet brushless DC machine

a machine that is similar in structure to a per-

manent magnet synchronous machine, con-
taining armature windings on the stator and
permanent magnets on the rotor. The per-
manent magnet brushless DC machine, how-
ever, is characterized by a trapezoidal flux
density distribution in the airgap instead of
the sinusoidal distribution of the synchronous
machine. In operation, a DC voltage is ap-
plied sequentially to the stator coils to create
a rotating field that pulls the rotor with it. To
correctly operate, the brushless DC machine
requires sensors to determine the rotor posi-
tion so that the proper stator phases may be
excited.

permanent magnet DC machine

a DC

machine in which the field excitation in the
stator is provided by permanent magnets in-
stead of electromagnets.

2000 by CRC Press LLC

Content .. 129 130 131 132 ..