An introduction to Communication Methods Across Power
Distribution Systems and Their Effects.
By Samuel Nyall Stearley
Abstract—This Paper is intended as an introduction to
methods that use existing power distribution systems to
communicate with devices. It will go over the other effect
that such communications have, namely the creation of
noise and the effect that such noise has on other forms of
communication. There is an emphasis on the modeling of
this noise.
Index Terms—Broad band over power lines, Ripple
Control, Load control.
I. INTRODUCTION
This paper is a discussion of methods and issues whereby
power lines are simultaneously used to deliver power and
information to their target systems. The level of information
can vary from complex data to simpler Boolean signals. And
the data being transmitted can move from a single source to
single or multiple destinations. This paper will go into the
details of signal injection, their effectiveness and noise
consequences.
II. DYNAMIC LOAD CONTROL
One of the simplest forms of communication over a power
system is that of adding a ripple voltage. The purpose of this
signal is the control of critical loads that are being powered.
The ripple voltage is an indicator that tells these loads to
switch on or off. This is often used as a safety feature; if too
much current is being drawn and the source will soon give out,
the signal can be sent telling the loads to shutdown. Thus
giving the load the chance to shutdown safely.
There are several methods available for load control.
Switch over meters are dumb terminals by themselves and
cannot be controlled dynamically. They have been used in the
US before World War II. Or instead of injecting a ripple the
frequency of the power is reduced from sixty to fifty hertz. A
third popular method in the US used for large scale load
management systems is the use of a separate radio controller.
1
All these methods are considered to be a “direct control” of
the loads and are used to control loads that have large storage
characteristics. This means that they do not require a steady
power source to continue functioning. Examples include
Samuel Stearley is a student of Electrical Engineering at
Calvin College, Grand Rapids, MI, 49546, US; email:
sstear70@calvin.edu
heaters (space or water) washing machines, driers, etc. These
type of devices do not need to be constantly supplying heat
and can afford to take a break every few seconds. [1]
Through the use of ripple control, certain German cities
have been able to keep their energy use sustained at an almost
constant draw. This means that more power is being used
during what were previously slump periods. So more power is
being consumed overall and that means higher sales. Used in
such a manner, up to 30% of the costs of installing ripple
control can be derived back. But there are other advantages
like the control of meters. [1]
Signal injection is a critical issue and must be done
properly. The signal must be added in a synchronous manner
at the appropriate timing with the existing power. It is often
done at the medium voltage level (10 to 35kV) to all three
phases; however it can be done at other levels as desired. The
higher voltage levels are distributed to farther reaches, and if
the signal needs to go farther then injection should be done at
these higher levels.
There are two methods for injecting data into a power
network. The network can be intercepted serially. With this
method power supply can be briefly stopped and data is sent
instead. The advantage of this method is that data does not
jump across transformers onto the higher voltage level.
However adding hardware in series with existing power lines
can be dangerous. If the data transmitter fails it could bring
down the entire network. [1]
The second method is to add the signal in parallel with the
power. Meaning it is layered on top of the existing three
phase sinusoidal wave. The disadvantage of this is that the
data can get through the transformer to the higher voltage
lines. For high frequency signals this is minimized as the
impedance the transformer applies to the signal is proportional
to frequency (ZInductor = w*l*j) and prevents power being
delivered to the high voltage lines. If the data will be
transmitted above 210 Hz then parallel distribution is fine, else
serial is used.
In practice most frequencies used fall between 110 to 750
hertz. Particular values used depend on the system. It is
possible that resonant effects can occur with power factor
correction capacitors and help propagate the signal. However
in general resonance with series impedance is something that
needs to be avoided. Frequency of injection also affects the
current that makes its way into the high voltage supply
network, and from this network into neighboring networks.
The chosen frequency should not overlap with possible
harmonic frequencies that could exist in the network, else a
fake signal could be detected as the actual signal.[1]
Receivers involve a band pass filter that checks for the
desired frequency. To prevent accidental detection the desired
frequency must be maintained for a set period of time before
the system judges the signal is valid. Standard ripple receivers
also feature three relays for guiding power about but custom
ripple receivers exists for specific applications, like the street
lighting receiver shown in Fig 1.
Figure 1
Figure 2
II. DIGITAL COMMUNICATION
A. Networking the home
A form of residential network communication method being
promoted currently by the “Home Plug Alliance” is deemed
the HomePlug specification. The idea of interfacing to the
power network via the house plugs to provide networking
capability is not a new idea and has been done in the past. But
these past advances have not lived up to expectations and did
not perform as well as hoped. Recently filtering technology
has taken a large enough leap to allow a newer generation to
arrive.
Home power networks vary considerably depending on the
regional wiring codes, age, and size of the house. Extensive
testing has been done to verify that modern in house
networking is universally viable for the consumer. For
example it is not a good idea to create consumer electronics
that people in New York can use, but not people in Texas. [2]
The results showed that speeds of 1.5 Mbps where produced
in ninety-eight percent of connections tested. (Out of 6690)
Seventy-seven percent of these connections where able to
achieve five Mbps transfer speeds. This test involved simple
communication between two devices. More advanced testing
among five devices. showed that ninety-three percent of
houses where able to sustain talking speeds of three Mbps.
While these tests left out houses found in the Midwest states
they are quite convincing that home networking via the power
network is feasible. [7]
B. Internet Delivery
Of significant interest lately and a hotly debated political
issue is the emerging technology of internet providers
providing broad band internet connections through power
lines. However it generates a large amount of interference
with Ham radios (among other things) so many Hobbyists are
up in arms over this potential threat.
The idea is that the majority of the internet transfer is done
through the normal means. However the “last mile” delivery
whereby an individual’s home is connected to the internet is
done via the power lines (Fig 2). The data is traveling at such
a high rate that it cannot get through the inductors of the
power transformers so the transmission cannot be entirely
done via the high, medium and low transmission lines.
The FCC regulations that govern unintentional emitters are
partly to blame for the cause of the scandal. They require that
current-carrier systems be tested in three “typical” real world
installations. The definition of “typical” is not standardized,
or that there even exists three distinct kinds of “typical”
applications.
However this works out to the advantage of a company
developing hardware. If passing the FCC’s rules were as
simple as doing white lab coat experiments to measure
generated electromagnetic fields, then PLC hardware would
not be legal to use. If the FCC judged PLCs based on
emissions then a double standard would result. There would
be one standard that was lenient and allowed PLC operation,
while another would disallow other class of devices.
When doing lab coat experiments the standard is when at a
distance of 30 meters from the unintentional emitter (whose
frequencies are below 30 MHz) it should not create an electric
field greater than thirty micro volts per meter. The testing is
done at a bandwidth of 9 kHz.
However this does not provide enough protection of ham
radio signals. A 30 uV/m field will create a noise on a half
wave dipole of -86.4 dBW which is equivalent to 338 micro
volts. However when comparing 338 uV to existing radio
signal transmissios (Which are much smaller by design) the
noise is 16 dB, this is very harmful. [8]
This is not to say that the “thirty micro volt per meter when
at a distance of thirty meters” rule is invalid, but that it was
designed for the analysis of point sources. With Power lines
the source is as long as the carrying part of the wire. For areas
that Broadband over Power lines is implemented, the
interference can blanket wide neighborhoods or entire cities.
[8]
The discussion of noise so far has only addressed one
component and does not consider further complications with
the geometry of the power networks. There are issues with the
power lines being designed for operation at sixty Hz. And
there are issues with the buildings the power lines interface to.
The electrical wirings of buildings result in a good
transmission line, but there are other aspects involved. For
example when a light switch is opened (ie the light is turned
off) the result is an antenna configuration that radiates nicely.
And there might be many other loads the building is connected
to that also radiate.
Furthermore there are odd loopholes in the FCC regulations
that are used inappropriately. When doing actual testing it is
permitted to do measurements at three meters and extrapolate
to thirty meters. In actuality this underestimates the noise by
20 dB. The FCC rules do specify conditions in using this
approximation, like when measuring at a distance of thirty
meters is impractical. But for power lines it simplicity to
setup equipment thirty meters from a power line.
The modeling and comparison of E field noise produced by
a high frequency to that of a low frequency is important in
understanding what is occurring. For a case study a twenty
meter two line power transmission cable delivers power to a
house. One of the lines has a 120 volt AC source at 60 Hz
while the other is ground. The lines are made out of copper
with a radius of 1 centimeter. Using transmission line
modeling methods described in Fig 3. The voltage at any
point z in the line at any time t can be found.
Figure 3
With the voltage in the transmission line, the E-field internal
to the line may be found. If the voltage did not vary with the
length of the transmission line in the z direction then the Efield would be zero. From the E-field the electric flux density
is derived. The electric flux density is multiplied by a
differential cross sectional area to find the differential charge
it contains. Then from the differential charge a differential
external E-field to a point P can be calculated. After
Integrating all the differential E fields the total E field is
found. These dimensions and what is being integrated are
described in fig 4.
Figure 4
The calculations are now given.
The principle of
superposition is used to separately examine a 120 volt source
at a frequency of 60 hertz and a 1 volt signal at 3.5 mega
Hertz. The formulas used to gauge the transmission line
parameters are different at high frequencies than they are at
lower frequencies and they are changed for each of the
frequencies considered. For calculation sake the final
transmission line delivering power to a home will be a twenty
meter copper cable with a radius of one centimeter. It will be
composed of two lines, a ground line and power line.
At the beginning the initial inputs are given.
Lcable  20
7
Rcable  0.01
7
Seperation cable  .021
c  5.8 10
 air  4 10
air  1.005
v 0  120
  60 2 
air  1.5 10
Knowing the system parameters the electric length () of
the transmission line are found. See equation 5.
 12
At this point the low frequency model (equations 1 to 4)
must be used to estimate the parameters R, L, G, and C (Fig
2) of the two line system.
Finally the fully developed formula for Voltage in the
transmission line at a given point and time is written in
equation 6.
Fig xyz is a three dimensional plot of this voltage. Time is
on one axis and distance down the transmission line is on the
other. Fig 5 displays the sixty hertz source voltage is able to
maintain a magnitude of a 120 volts down the entire
transmission line.
Fig 6demonstrates the amount of E-field noise the sixty
hertz, 120 volts source creates at a point thirty meters from
the halfway point of the line. The normal source voltage
creates an E-field signal that is less than five nanovolts/meter.
Figure 5
The fact that voltage barely varies down the
transmission line is going to have a significant effect. The Efield inside the transmission line is the derivative of the
voltage (Equation 7)
Figure 6
This derivative expands into a much larger expression.
The electric flux density comes from multiplying the Efield by 0 (Formula 9). This electric flux density is
multiplied by a differential surface area as described in fig 3
to get the charge it contains. Finally this differential charge
divided by a 4R^2 is integrated the length of the
transmission line to calculate the E-field external of the
transmission line. (Formulas 10 -12) This E-field is
symmetric about the transmission line. It is this E-field that
will have an impact on antennas and other devices: Note that
z no longer refers to distance down the transmission line, but
the axis drawn in fig 3.
Having calculated the noise from the power line the E-field
noise that results from a 3.5 MHZ wave placed on top of the
power signal needs to be found for comparison. The process
is similar to what has already been done with a few
exceptions. The first being that the high frequency model
(Formulas 13-16) must be used for estimating the C, R, L and
G parameters of the system. From this point on variables
will have a ‘_2’ appended to their names.
From the transmission line parameters to the electric length
in equation 17.
A graph of the magnitude of this E-field (Fig xyz) shows a
signal that peaks at 3 micro volts per meter. The goal was to
get a 30 micro volt per meter signal that is known to be
produced. However a factor of ten can be explained away as
the result of the initial parameters of the system.
From these constants the voltage as function of time and
position can be graphed. The bold line represents voltage at
a distance of zero into the transmission line, it is part of a one
volt sinusoid. As the voltage moves into the transmission
line the peak of this sinusoid can be seen to drop down.
Figure 4
Figure 3
The internal E-field (equation 18) that results from this
voltage is of a significantly larger magnitude than the internal
E-field created by the sixty hertz source.
This three micro volt per meter E-field is six hundred times
larger than the E-field produced by the sixty hertz sinusoid.
Furthermore if there was an antenna that was tuned for this
frequency sitting in a thirty micro volt per meter field it
would pick up 2.2 nano-watts of power, when real signals
deliver pico-watts of power.
However it seems that the FCC is ignoring these issues in
the push to get high speed internet access to the masses.
[1] HD-Comsys & Itd tim, “Ripple Control Systems,” http://www.tel.hr/hdcitd/Engleski/RCS.htm. 2004.
[2] Home Plug Power Alliance, “Home Plug Field Test Results,” 2003
[3] Ed Hare. “Calculated Impact on PLC Stations Operating in the Amateur
Radio Service,” 2002.
Samuel S. Author was born in Missouri in 1982. He is currently a senior
EE student and will graduate in 2004 with a BSE with an emphasis in
Electrical Engineering from Calvin College, Grand Rapids Mi..
He has previously worked for Delphi Automotive where he maintained
and developed software than modeled how parts of an engine’s valve train
operate.
With the internal E-field, Internal Electric Flux and the
external E-field are found in equations 19 -22.