RADIATED and CONDUCTED NOISE
Switching mode power supplies, light dimmers, computer networking systems, poor connections that arc, and other "accidental transmitters" that either switch or spark can create considerable RF energy on wiring. It is helpful, when attempting to reduce such noise, that we understand how the noise travels from the source into the receiving system.
Definition of a Source
Let's consider the problematic device a "thing in a black-box", and call it the "source". It doesn't matter if it is a computer, the spark in an electric fence, a light dimmer, or a doorbell transformer. The smallest area we can isolate creating the unwanted energy will be the source.
Radiation from a Source
There is very little radiation from a source. As a matter of fact, even a very powerful single-terminal source can't transmit at any distance. In order to be an effective radiator, the source must have two or more connections to the outside world. The multiple connections (two or more) could be the two (or three) wires in a power cord, or a insertion point in a single wire. It could also be a connection to two totally different sources. Let's look at examples of each.
Consider a poor slice in a conductor carrying more than a few volts can cause a RFI generating arc. Two of the most common examples are an electric fence or power line.
The noise results when an insulated area with an electrical open circuit at low voltages breaks down as sufficient voltage appears across the gap. This ionizes air in the gap, and the resulting plasma conducts heavily. Such arcs can be particularly noisy at radio frequencies, because resistance can fall faster than voltage increases creating a negative-resistance. Negative resistances created in spark plasma were actually used as the amplifying element in crude oscillators. The plasma excited resonant circuits in the very first "single-frequency" transmitters, and at times those transmitters spanned distances of thousands of miles to very poor receivers using what were very poor antennas by today's standards!
Each conductor direction leaving the arc forms half of a large antenna. The entire system, from the source out in both directions, radiates. There isn't any effective way to cure or substantially reduce the unwanted radiation except by stopping the arc. Poor slices can become quieter in heavy rain, and can be broken up by wind or wire movement.
This type of arc excites the conductor entering the splice from one direction as one terminal, and the wire exiting the splice as the other terminal! The coupling mechanism is like a dipole feedpoint, and is very efficient at radiating over very wide frequency ranges.
Arcing to Another Conductor
A poor insulator or a close-spaced gap can break down and generate noise.
Once again, the noise results when an insulated area with an electrical open circuit at lower voltages breaks down when sufficient voltage appears across the gap. This ionizes air in the gap, and the resulting plasma conducts heavily. Such arcs can be particularly noisy at higher radio frequencies, because resistance can fall faster than voltage increases creating a negative-resistance.
Shunting-arcs generally have poor low-frequency energy, because the gaps are almost always connected to a very high impedance at low frequencies. This makes a very poor antenna connection for the arc, and reduces low-frequency response. If the gap had a path to a low resistance, the plasma would form a dead-short in the path. This would quickly cause a catastrophic system failure in systems able to supply sufficient current.
Non-arcing RF Generators
Many devices contain high-speed switching systems. These devices include but are not limited to televisions and monitors, computers, lighting systems with dimmers, and low-voltage lighting systems using switch-mode power supplies.
These non-arcing sources are similar to arcing sources, in that they require at least two terminal paths to become effective radiators. They generally are frequency selective or frequency periodic, and produce a broad buzzing signal that drifts around.
No matter what source is at work, we can be sure more than one conductor is at work in coupling from the source. If we disturb that path by isolating one conductor with a high RF impedance, we can reduce the interference. But by far the most effective way to remove or reduce interference, other than by removing the actual offending source, is by "shorting" the RF path with a bypass capacitor. As we will see later, this is almost always much more effective than adding series impedance.
Common Noise Sources
Arcing Powerline Insulator or Hardware
The bell-insulator in a power line is probably the most common noise source. Older bell insulators has two interlocking but well-insulated metal posts. These metal posts have considerable area, because of the required mechanical strength. This results in considerable capacitance, in the order of a few dozen picofarads or greater. The ends of each bell terminal are held by a loose-fitting pin assembly. This pin assembly is the root of most powerline noise problems.
Bell insulators are often misused by power line crews. They are sometimes used to terminate short spans between poles, where the correct insulator would be a rigid post type or a polymer insulated fiberglass rod. The bell insulator, used in a short span, often is installed without sufficient tension. The lack of tension allows corrosion to set-up on the loose-fitting pins, and the capacitance of the insulator creates a voltage divider. Since the pin capacitive reactance is very high compared to the insulators leakage resistance and capacitance, a very large voltage can appear across the pin. This is true even when the pin is connected to grounded hardware! These loose short spans are called "slack spans", and bells should NEVER be used on slack spans! Slack spans with bell insulators generally become quieter during damp weather or in or just after rain, and can often be broken up by wire movement.
As the sine-wave peaks, the corrosion or oxides in the joint of the loose-fitting pin break down and arc. This causes an unintentional electrical noise to be generated, and the two terminals for the transmitter become the power line conductor and the grounded pole or hardware on the pole. With a wooden pole, the entire upper area of the can be excited with leakage currents from leakage capacitance and leakage resistances. Even if an insulator isn't arcing, a loose metal staple over a grounded wire or other poorly connected metal-to-metal joints will arc. Even a very tiny arc can excite the ground wire running down the pole and the power line wires as two (or more) terminals for the source.
To a much lesser extent transformers, lightning arrestors, fuse assemblies, and disconnects can have internal arcs.
Trees and foliage may also contact HV power lines and "burn" against the wire. Generally foliage noises are not extremely loud sources at low frequencies because one terminal of the source is through the tree, but occasionally they can be strong at lower frequencies.
As a general rule, but not always:
Older electric fences had a few-second timed off-and-on buzz. Newer fences have a "popping" noise, because they charge the fence in "tics". Fence problems are similar to power line problems, except of course the hardware is much smaller! Most severe problems are related to poor insulators, bad splices, or loose hardware.
Computers, TV's, and Switching Systems
In older times, TV sets created problems from sweep circuits. The horizontal sweep circuit operates with a saw tooth shaped waveform that was rich in harmonics. Early video systems overpowered the horizontal sweep system, and used the extra energy to supply high tension for the CRT. This resulted in many point-to-point wires carrying high power harmonic-rich 15kHz RF energy. A typical 25-inch television would use about 15 watts of the horizontal sweep system power for sweep, and perhaps 30 watts for powering the CRT second anode!
Wasted energy from the 50-75 watt sweep system was often coupled to the outside world a push-pull signal between the antenna leads and the power cord.
Modern computers and TV's mostly create noise-problems from internal switching power supplies, rather than sweep systems. The sweep systems, if ever even required, now generally only power the sweep. The HV often comes from separate HV power supplies.
Switch mode power supplies, whether in a TV, computer, telephone, video, or lighting system are rapidly becoming the most common source of modern RFI problems. Most of this has to do with poor testing techniques, lack of good standards, poor enforcement of the few poor standards we have, and a lack of skill or knowledge in power supply and device design and installation.
Fortunately most problems can be corrected by external filtering, but it generally needs to be done at the offending device.
Curing RF Egress or Ingress
Once the outside world is reached, the push-pull nature of the source can appear as a common-mode signal following a group of wires for many miles. We can think of this as one very long wire with the earth as the return path. The signal either directly radiates into our antennas, or it excites our antennas via common mode paths along cable shields.
While power lines and high-tension electric fences must be cured by removing the arc at the source, lower voltage systems can almost always be cured through proper external bypassing.
As a first response, we often like to throw a few ferrite beads at the problem. In more sophisticated approach, we might use multiple-turn series inductors. What these approaches miss is the great difficulty in obtaining adequate series impedance. When we consider the series impedance has to be totally ISOLATED for each conductor, and further consider core saturation from low-frequency operating currents we quickly realize series impedance is NOT the most effective method. It is helpful and may reduce RFI to acceptable levels in mild cases, but it certainly is an ineffective cure. Even if we are successful at one frequency range, it becomes virtually impossible to have a high series impedance over very wide frequency ranges.
The best approach is to add some RF series impedance and to augment it with excellent bypassing of all leads entering or leaving a device to one common point. That common point generally is connected to a safety ground, available in the USA at the third ground terminal of every modern-code electrical outlet.
Coaxial or audio cables should all leave the device through a bulkhead plate with all shields grounded. The power cord should pass through that bulkhead, and be bypassed to the bulkhead with properly selected UL/CSA/VDE approved line-bypass capacitors. Any chokes should be installed between the source device and the bulkhead. When this happens, the device becomes an isolated entity all by itself. It can not excite outside world conductors with unwanted RF.
It is indeed fortunate that the very same things that cure RFI often lead to greatly improved lightning protection.