Electrical noise and mitigation

Home | Articles | Forum | Glossary | Books



In this section, we will learn about noise in electrical circuits, the reasons for their generation, types of noise and mitigation. We will cover shielding as a means of noise control and the role played by grounding and how properly designed grounding can reduce noise. We will learn about zero signal reference grids for noise-prone installations.

We will briefly deal with the subject of harmonics and how they affect power and electronic equipment and about ways of controlling them.

Definition of electrical noise and measures for noise reduction

Noise, or interference, can be defined as undesirable electrical signals, which distort or interfere with an original (or desired) signal. Noise could be transient (temporary) or constant.

Unpredictable transient noise is caused , for example, by lightning. Constant noise can be due to the predictable 50 or 60 Hz AC 'hum' from power circuits or harmonic multiples of power frequency close to the data communications cable. This unpredictability makes the design of a data communications system quite challenging.

Noise can be generated from within the system itself (internal noise) or from an outside source (external noise). Examples of these types of noise are:

Internal noise

• Thermal noise (due to electron movement within the electrical circuits)

• Imperfections (in the electrical design).

External noise

• Natural origins (electrostatic interference and electrical storms)

• Electromagnetic interference (EMI) - from currents in cables

• Radio frequency interference (RFI) - from radio systems radiating signals

• Cross talk (from other cables separated by a small distance).

From a general point of view, there must be three contributing factors before an electrical noise problem can exist. These are:

1. A source of electrical noise

2. A mechanism coupling the source to the affected circuit

3. A circuit conveying the sensitive communication signals.

Typical sources of noise are devices, which produce quick changes (spikes) in voltage or current or harmonics, such as:

• Large electrical motors being switched on

• Fluorescent lighting tubes

• Solid-state converters or drive systems

• Lightning strikes

• High-voltage surges due to electrical faults

• Welding equipment.

FIG. 1 shows a typical noise waveform and how it looks when superimposed on the power source voltage waveform.

FIG. 1 Noise signal (top) and noise over AC power (bottom)

Electrical systems are prone to such noise due to various reasons. As discussed in the previous section, lightning and switching surges are two of these. These surges produce high but very short duration of distortions of the voltage wave. Another common example is 'notching', which appears in circuits using silicon-controlled rectifiers (power thyristors). The switching of these devices causes sharp inverted spikes during commutation (transfer of conduction from one phase arm to the next). FIG. 2 shows the typical waveform with this type of disturbance.

Harmonics in supply system is yet another form of disturbance. This subject will be reviewed in detail later in the section. A typical waveform with harmonic components is shown in FIG. 3. Switching of large loads in power circuits to which automatic data processing (ADP) loads are connected can also cause disturbances. Similarly, faults in power systems can cause voltage disturbances. All these distortions and disturbances can find their way to sensitive electronic equipment through the power supply mains connection and cause problems. Apart from these directly communicated disturbances, sparks and arcing generated in power-switching devices and high-frequency harmonic current components can produce electromagnetic interference (EMI) in signal circuits, which will require to be properly shielded or screened to avoid interference. FIG. 4 shows diagrammatically the reasons for noise from the equipment within a facility.

FIG. 2 Waveform distorted by notching

FIG. 3 Waveform distorted by harmonics

FIG. 4 Noise emanating from electrical systems within a facility

The following general principles are applicable for reducing the effects of electrical noise:

• Physical segregation of noise sources from noise-sensitive equipment

• Electrical segregation

• Harmonic current control

• Avoiding ground loops which are a major cause of noise propagation (including measures such as zero signal reference grid, explained later in this section)

• Shielding/screening of noise sources and noise-susceptible equipment including use of shielded/twisted pair conductors.

How are sensitive circuits affected by noise?

Noise is only important if it’s measured in relation to the communication signal, which carries the data or information. Electronic receiving circuits for digital communications have a broad voltage range, which determines whether a signal is binary bit '1' or '0'. The noise voltage has to be high enough to take the signal voltage outside these limits for errors to occur.

The power and logic voltages of present day devices have been drastically reduced and at the same time, the speed of these devices has increased with propagation times now being measured in picoseconds. While the speed of the equipment has gone up and the voltage sensitivity has gone down, the noise conditions coming from the power supply side have not reduced at all. The best illustration that can be given of this condition is to consider where the signal voltage has been and what is happening to it compared to the noise voltage (see FIG. 5). In years gone by, signal voltages may have been 30 V or more but since then have steadily been decreasing. As long as the signal voltage was high and the noise voltage was only 1 V, then we had what most instrument engineers would call a very high signal to noise ratio, 30:1.

Most engineers would say you have no problem distinguishing the signal as long as you have such a high signal to noise ratio. As the electronic equipment industry advanced, the signal strength went down further, below 10 and then below 5. Today we are fighting 1-, 2- and 3 V signals and still finding ourselves with 1, 2 and 3 V of electrical noise. When this takes place for brief periods of time, the noise signal may be larger than the actual signal. The sensors within the sensitive equipment turn and try to run on the noise signal itself as the predominant voltage.

FIG. 5 Relative magnitudes of signal and noise (then and now)

When this takes place, a parity check or a security check signal is sent out from the sensitive equipment asking if this particular voltage is one of the voltages the sensor should recognize.

Usually, this check fails when it’s a noise voltage rather than the proper signal that it should be looking at and the equipment shuts down because it has no signal. In other words, the equipment self-protects when there is no signal to keep it operating. When the signal to noise ratio has fallen from a positive direction to a negative direction, the equipment interprets that as the need to turn off so this it won’t be running on sporadic signals.

In the top portion of FIG. 5, a 20-30-V logic signal is well in excess of the noise that is occurring between the on and off digital signal flow. In the bottom picture, however, the noise has raised its head above the area of the logic signal which has now dropped significantly into the 3-5 V range and perhaps even lower. You will also notice that the difference between the upper and lower pictures in the graph shows the speed with which the signal was transmitted. In the upper graph, the ons and offs are relatively slow, evidenced by the large spaces between the traces. In the lower graph, the trace is now much faster. There are many more ons and offs jammed into the same space and as such, the erratic noise behavior may now interfere with the actual transmission.

The ratio of the signal voltage to the noise voltage determines the strength of the signal in relation to the noise. This 'signal to noise ratio' (SNR) is important in assessing how well the communication system will operate. In data communications, the signal voltage is relatively stable and is determined by the voltage at the source (transmitter) and the volt drop along the line due to the cable resistance (size and length). The SNR is therefore a measure of the interference on the communication link.

The SNR is usually expressed in decibels (dB), which is the logarithmic ratio of the signal voltage (S) to noise voltage (N). S N SNR 10log dB () =

An SNR of 20 dB is considered low (bad), while an SNR of 60 dB is considered high (good). The higher the SNR, the easier it’s to provide acceptable performance with simpler circuitry and cheaper cabling.

In data communications, a more relevant performance measurement of the link is the bit error rate (BER). This is a measure of the number of successful bits received compared to bits that are in error. A BER of 10-6 means that one bit in a million will be in error and is considered poor performance on a bulk data communications system with high data rates.

A BER of 10-12 (one error bit in a million million) is considered to be very good. Over industrial systems, with low data requirements, a BER of 10-4 could be quite acceptable.

There is a relationship between SNR and BER. As the SNR increases, the error rate drops off rapidly as is shown in FIG. 6. Most of the communications systems start to provide reasonably good BERs when the SNR is above 20 dB.

Frequency analysis of noise

Another useful way of evaluating the effects of noise is to examine its frequency spectrum. Noise can be seen to fall into three groups:

1. Wideband noise

2. Impulse noise

4. Frequency-specific noise.

The three groups are shown in the simplified frequency domain as well as the conventional time domain. In this way, we can appreciate the signal's changing properties as well as viewing the amplitude in the customary time domain.

FIG. 6 Relationship between the bit error rate and the signal to noise ratio

Wideband noise contains numerous frequency components and amplitude values. These are depicted in the time domain graph shown in FIG. 7 and in the frequency domain graph shown in FIG. 8.

In the frequency domain, the energy components of wideband noise extend over a wide range of frequencies (frequency spectrum).

FIG. 7 Time domain plot of wideband noise

FIG. 8 Frequency domain plot of wideband noise

Wideband noise will often result in the occasional loss or corruption of a data bit. This occurs at times when the noise signal amplitude is large enough to confuse the system into making a wrong decision on what digital information or character was received. Encoding techniques such as parity checking and block character checking (BCC) are important for wideband error detection so that the receiver can determine when an error has occurred.

Impulse noise is best described as a burst of noise, which may last for a duration of say up to 20 ms. It appears in the time domain as indicated in FIG. 9.

FIG. 10 illustrates the frequency domain of this type of noise. It affects a wide bandwidth with decreasing amplitude vs frequency.

Impulse noise is brought about by the transient disturbances in electrical activity such as when an electric motor starts up, or from switching elements within telephone exchanges. Impulse noise swamps the desired signal, thus corrupting a string of data bits.

As a result of this effect, synchronization may be lost or the character framing may be disrupted. Noise of this nature usually results in garbled data making messages difficult to decipher. Cyclic redundancy checking (CRC) error detection techniques may be required to detect such corruption.

FIG. 9 Time domain plot of impulse noise

FIG. 10 Frequency domain plot of impulse noise

Although more damaging than wideband noise, impulse noise is generally less frequent. The time and frequency domain plots for impulse noise will vary depending on the actual shape of the pulse. Pulse shapes may be square, trapezoid, triangular or sine For example.

In general, the narrower and steeper a pulse, the more energy is placed in the higher frequency regions.

Frequency-specific noise is characterized by a constant frequency, but its amplitude may vary depending on how far the communication system is from the noise source, the amplitude of the noise signal and the shielding techniques used.

This noise group is typical of AC power systems ( FIG. 11 and 8.12) and can be reduced by separating the data communication system from the power source. As this form of noise has a predictable frequency spectrum, noise resistance is easier to implement within the system design.

Filters are typically used to reduce this to an acceptable level.

FIG. 11 Time domain plot of constant frequency noise

FIG. 12 Frequency domain plot of constant frequency noise

5 Categories of noise

An electrical noise falls under one of the following categories: transverse mode or common mode. Transverse-mode noise is a disturbance, which appears between two active conductors (phase or neutral) in an electrical system. Such a noise is therefore measurable between two line conductors or between a line conductor and neutral. This is usually having its genesis from within the power system ( FIG. 13).

FIG. 13 Transverse-mode noise

Common-mode noise, on the other hand, appears simultaneously in each active conductor and therefore cannot be measured like a transverse-mode noise. It usually involves the ground conductor and originates from some external disturbance ( FIG. 14).

FIG. 14 Common-mode noise

Disturbances from other equipment in the same distribution system

An important factor to be taken note of in dealing with electrical system generated noise is electrical segregation of noise-producing equipment and noise-sensitive equipment. FIG. 15 illustrates this principle. In case A, the 'noisy' AC units and noise-sensitive ADP loads share a common power supply system. Frequent starts of AC compressors could cause voltage fluctuations, which will be communicated to ADP power units and can translate as noise in ADP units' electronic circuits. In case B, a separation of circuits has been achieved by employing different sub-circuits for AC loads and ADP loads but this may not have much impact as far as noise is concerned since the sources are shared.

In case C, a two-winding transformer has been introduced in the ADP circuit feeder.

This will act as a cushion for the noise due to the inherent inductance of the transformer, which won’t allow steep noise fronts to pass through. In case D, two separate transformers feed the AC loads and ADP loads with transfer-switching provision. The two-winding transformer has been retained. Obviously, D is the best case solution but expensive. In some situations, it may not be feasible to implement too. C will, however, provide an acceptable solution without being quite as expensive as D and can be retrofitted easily where required.

FIG. 15 Segregation of noisy loads

Earth loop as a cause of noise

As we have seen in earlier sections, two different equipment with a communication cable between them and each of the panels connected to a local ground point form an earth loop, which can give rise to noise currents in the communication cable. A typical building electrical system with multiple earth points is shown in FIG. 16. Note how each panel/equipment in the distribution system is connected to ground at the nearest convenient point of the building grounding system. Note how two sensitive equipment units (shown in the upper right of the diagram as EDP devices) are connected to ground points A and B with the grounding system's inherent impedance shown between them.

The EDP devices have a communication cable running between them with the ends of the cable screen connected to the EDP panel's enclosure. Any stray current in the ground system between A and B will cause a noise voltage between points A and B, which in turn can drive a current through the cable screen that can couple as a noise through the communication cable conductors.

FIG. 17 shows how a noise can originate in the electrical power supply system. In this case, the HVAC motor winding acts as a capacitance between the electrical system and the motor's grounded enclosure. Whenever the motor starts, this capacitance sends a pulse of current through the insulation into the motor frame, which is grounded through the metallic conduit carrying the cable, leads, to the motor. The random ground connections between this conduit and other grounded metal parts act like a ground loop and create an inter-cabinet potential difference between two sensitive equipment (EDP units 1 and 2). This can cause noise pulses to flow into the serial data cable connecting the two systems, resulting in data errors.

FIG. 16 Earth connections in building electrical distribution systems causing ground loops

FIG. 17 Starting of HVAC motor gives rise to noise due to ground loops

The ways in which noise can enter a signal cable and its control

Electrical noise occurs or is transmitted into a signal cable system in the following ways:

• Galvanic (direct electrical contact)

• Electrostatic coupling

• Electromagnetic induction

• Radio frequency interference (RFI).

If two signal channels within a single data cable share the same signal reference conductor (common return path), the voltage drop caused by one channel's signal in the reference conductor can appear as a noise in the other channel and will result in interference. This is called galvanic noise.

Electrostatic noise is one, which is transmitted through various capacitances present in the system such as between wires within a cable, between power and signal cables, between wires to ground (as we saw in the HVAC motor example) or between two windings of a transformer. These capacitances present low-impedance paths when noise voltages of high frequency are present. Thus noise can jump across apparently non conducting paths and create a disturbance in signal/data circuits.

Electromagnetic interference (EMI) is caused when the flux lines of a strong magnetic field produced by a power conductor cut other nearby conductors and cause induced voltages to appear across them. When signal cables are involved in the EMI process, this causes a noise in signal circuits. This is aggravated when harmonic currents are present in the system. Higher order harmonics have much higher frequencies than the normal AC wave and result in interference particularly in communication circuits.

Radio frequency interference involves coupling of noise through radio frequency interference. We will now describe these in some detail.

Galvanic coupling (or common impedance coupling)

For situations where two or more electrical circuits share common conductors, there can be some coupling between different circuits with deleterious effects on the connected circuits. Essentially, this means that the signal current from one circuit proceeds back along the common conductor resulting in an error voltage along the return bus, which affects all the other signals. The error voltage is due to the capacitance, inductance and resistance in the return wire. This situation is shown in FIG. 18.

Obviously, the quickest way to reduce the effects of impedance coupling is to minimize the impedance of the return wire. The best solution is to use a balanced circuit with separate returns for each individual signal shown in FIG. 19.

Electrostatic or capacitive coupling

This form of coupling is proportional to the capacitance between the noise source and the signal wires. The magnitude of the interference depends on the rate of change of the noise voltage and the capacitance between the noise circuit and the signal circuit.

In FIG. 20, the noise voltage is coupled into the communication signal wires through the two capacitors C1 and C2, and a noise voltage is produced across the resistances in the circuit.

The size of the noise (or error) voltage in the signal wires is proportional to the:

• Inverse of the distance of noise voltage from each of the signal wires

• Length (and hence impedance) of the signal wires into which the noise is induced

• Amplitude (or strength) of the noise voltage

• Frequency of the noise voltage.

There are four methods for reducing the noise induced by electrostatic coupling. They are:

1. Shielding of the signal wires

2. Separating from the source of the noise

3. Reducing the amplitude of the noise voltage (and possibly the frequency)

4. Twisting of the signal wires.

FIG. 18 Impedance coupling

FIG. 19 Impedance coupling eliminated with balanced circuit

FIG. 20 Electrostatic coupling

FIG. 21 indicates the situation that occurs when an electrostatic shield is installed around the signal wires. The currents generated by the noise voltages prefer to flow down the lower-impedance path of the shield rather than the signal wires. If one of the signal wires and the shield are tied to the earth at one point, which ensures that the shield and the signal wires are at an identical potential, then reduced signal current flows between the signal wires and the shield.

FIG. 21 Shield to minimize electrostatic coupling

Note: The shield must be of a low-resistance material such as aluminum or copper. For a loosely braided copper shield (85% braid coverage) the screening factor is about 100 times or 20 dB that is, C3 and C4 are about 1/100 C1 or C2. For a low-resistance multi-layered screen, this screening factor can be 35 dB or 3000 times.

Twisting of the signal wires provides a slight improvement in the induced noise voltage by ensuring that C1 and C2 are closer together in value; thus ensuring that any noise voltages induced in the signal wires tend to cancel one another out.

Note: Provision of a shield by a cable manufacturer ensures that the capacitance between the shield and the wires is equal in value (thus eliminating any noise voltages by cancellation).

Magnetic or inductive coupling

This depends on the rate of change of the noise current and the mutual inductance between the noise system and the signal wires. Expressed slightly differently, the degree of noise induced by magnetic coupling will depend on the:

• Magnitude of the noise current

• Frequency of the noise current

• Area enclosed by the signal wires (through which the noise current magnetic flux cuts)

• Inverse of the distance from the disturbing noise source to the signal wires.

The effect of magnetic coupling is shown in FIG. 22.

The easiest way of reducing the noise voltage caused by magnetic coupling is to twist the signal conductors. This results in lower noise due to the smaller area for each loop.

This means less magnetic flux to cut through the loop and consequently a lower induced noise voltage. In addition, the noise voltage that is induced in each loop tends to cancel out the noise voltages from the next sequential loop. Hence an even number of loops will tend to have the noise voltages canceling each other out. It’s assumed that the noise voltage is induced in equal magnitudes in each signal wire due to the twisting of the wires giving a similar separation distance from the noise voltage (see FIG. 23).

FIG. 22 Magnetic coupling

FIG. 23 Twisting of wires to reduce magnetic coupling

The second approach is to use a magnetic shield around the signal wires (refer FIG. 24). The magnetic flux generated from the noise currents induces small eddy currents in the magnetic shield. These eddy currents then create an opposing magnetic flux Ø1 to the original flux Ø2. This means a lesser flux (Ø2 - Ø1) reaches our circuit!

FIG. 24 Use of magnetic shield to reduce magnetic coupling

Note: The magnetic shield does not require earthing. It works merely by being present.

High-permeability steel makes best magnetic shields for special applications. However, galvanized steel conduit makes a quite effective shield. 8.8.4 Radio frequency radiation

The noise voltages induced by electrostatic and inductive coupling (discussed above) are manifestations of the near field effect, which is electromagnetic radiation close to the source of the noise. This sort of interference is often difficult to eliminate and requires close attention of grounding of the adjacent electrical circuit, and the earth connection is only effective for circuits in close proximity to the electromagnetic radiation. The effects of electromagnetic radiation can be neglected unless the field strength exceeds 1 V/m. This can be calculated by the formula:

0.173 power

Field strength distance =

…where field strength is in volt/meter, power is in kilowatt and distance is in kilometer.

The two most commonly used mechanisms to minimize electromagnetic radiation are:

1. Proper shielding (iron)

2. Capacitors to shunt the noise voltages to earth.

Any incompletely shielded conductors will perform as a receiving aerial for the radio signal and hence care should be taken to ensure good shielding of any exposed wiring.

More about shielding

It’s important that electrostatic shielding is only earthed at one point. More than one earth point will cause circulating currents. The shield should be insulated to prevent inadvertent contact with multiple points, which behave as earth points resulting in circulating currents. The shield should never be left floating because this would tend to allow capacitive coupling, rendering the shield useless.

Two useful techniques for isolating one circuit from the other are by the use of opto isolation as shown in FIG. 25, and transformer coupling as shown in FIG. 26.

FIG. 25 Opto-isolation of two circuits

FIG. 26 Transformer coupling

Although opto-isolation does isolate one circuit from another, it does not prevent noise or interference being transmitted from one circuit to another. Transformer coupling can be preferable to optical isolation when there are very high speed transients in one circuit. There is some capacitive coupling between the LED and the base of the transistor which in the opto-coupler can allow these types of transients to penetrate one circuit from another. This is not the case with transformer coupling.

1. Good shielding performance ratios

The use of some form of low-resistance material covering the signal conductors is considered good shielding practice for reducing electrostatic coupling. When comparing shielding with no protection, this reduction can vary from copper braid (85% coverage), which returns a noise reduction ratio of 100:1 to aluminum Mylar tape, with drain wire, with a ratio of 6000:1.

Twisting the wires to reduce inductive coupling reduces the noise (in comparison to no twisting) by ratios varying from 14:1 (for four-inch lay) to 141:1 (for one-inch lay). In comparison, putting parallel (untwisted) wires into steel conduit only gives a noise reduction of 22:1.

On very sensitive circuits with high levels of magnetic and electrostatic coupling, the approach is to use coaxial cables. Double-shielded cable can give good results for very sensitive circuits.

Note: With double shielding, the outer shield could be earthed at multiple points to minimize radio frequency circulating loops. This distance should be set at intervals of less than 1/8th the wavelength of the radio frequency noise.

2. Cable ducting or raceways as magnetic shield

These are useful in providing a level of attenuation of electric and magnetic fields. These FIG.s are valid for a frequency of 60 Hz for magnetic fields and 100 kHz for electric fields. Typical screening factors are:

• For 5 cm (2 in.) aluminum conduit with 0.154 in. thickness

- Magnetic fields 1.5:1

- Electric fields 8000:1

• Galvanized steel conduit (5 cm (2 in.), wall thickness 0.154 in.)

- Magnetic fields 40:1

- Electric fields 2000:1

3. Cable spacing as a means of noise mitigation

In situations where there are a large number of cables varying in voltage and current levels, the IEEE 518 - 1982 standard has developed a useful set of tables indicating separation distances for the various classes of cables. There are four classification levels of susceptibility for cables. Susceptibility, in this context, is understood to be an indication of how well the signal circuit can differentiate between the undesirable noise and required signal. It follows that a data communication physical standard such as RS-232E would have a high susceptibility, and a 1000-V, 200-A AC cable has a low susceptibility.

The four susceptibility levels defined by the IEEE 518 - 1982 standard are briefly:

1. Level 1 - high: This is defined as analog signals less than 50 V and digital signals less than 15 V. This would include digital logic buses and telephone circuits. Data communication cables fall into this category.

2. Level 2 - medium: This category includes analog signals greater than 50 V and switching circuits.

3. Level 3 - low: This includes switching signals greater than 50 V and analog signals greater than 50 V. Currents less than 20 A are also included in this category.

4. Level 4 - power: This includes voltages in the range 0-1000 V and currents in the range 20-800 A. This applies to both AC and DC circuits.

The IEEE 518 also provides for three different situations when calculating the separation distance required between the various levels of susceptibilities. In considering the specific case where one cable is a high-susceptibility cable and the other cable has a varying susceptibility, the required separation distance would vary as follows:

• Both cables contained in a separate tray

- Level 1 to level 2-30 mm

- Level 1 to level 3-160 mm

- Level 1 to level 4-670 mm

• One cable contained in a tray and the other in conduit

- Level 1 to level 2-30 mm

- Level 1 to level 3-110 mm

- Level 1 to level 4-460 mm

• Both cables contained in separate conduit

- Level 1 to level 2-30 mm

- Level 1 to level 3-80 mm

- Level 1 to level 4 -310 mm.

The FIG.s are approximate as the original standard is quoted in inches.

A few words need to be said about the construction of the trays and conduits. It’s expected that the trays are manufactured from metal and be firmly earthed with complete continuity throughout the length of the tray. The trays should also be fully covered preventing the possibility of any area being without shielding.

Briefly galvanic noise can easily be avoided by refraining from the use of a shared signal reference conductor, in other words, keeping the two signal channels galvanically separate so that no interference takes place.

Electromagnetic induction can be minimized in several ways. One way is to put the source of electromagnetic flux within a metallic enclosure, a magnetic screen. Such a screen restricts the flow of magnetic flux from going beyond its periphery so that it cannot interfere with external conductors. A similar screen around the receptor of EMI can mitigate noise by not allowing flux lines inside its enclosure but to take a path along the plane of its surface. Physical separation between the noise source and the receptor will also reduce magnetic coupling and therefore the interference.

Twisting of signal conductors is another way to reduce EMI. The polarity of induced voltage will be reversed at each twist along the length of the signal cable and will cancel out the noise voltage. These are called twisted pair cables.

Electrostatic interference can be prevented or at least minimized by the use of shields. A shield is usually made of a highly conductive material such as copper, which is placed in the path of coupling. An example is the use of a shield, which is placed around a signal conductor. When a noise voltage tries to flow across the capacitance separating two conductors, say a power and a signal conductor (actually through the insulation of the conductors), it encounters the conducting screen, which is connected to ground. The result is that the noise is diverted to ground through the shield rather than flowing through the higher impedance path to the other conductor.

If the shield is not of a high conductive material, the flow of the diverted current through the shield can cause a local rise of voltage in the shield, which can cause part of the noise current to flow through the capacitance between the shield and the second conductor.

4. Optical cables

The best method is of course to use signal cables of optical type, which are immune to all forms of electrical noise. Their use is very common in communication cables and for network conductors for supervisory control and data acquisition (SCADA) systems found in major electrical installations where noise is inherent in the environment. Most industrial controls such as distributed control systems in power and various process industries prefer use of fiber optic conductors as their data highway.

All these methods are routinely applied in practice as noise reduction measures. We will discuss about one of the important components used in power distribution systems for sensitive equipment viz., the shielded isolation transformer.

Shielded isolation transformer

A shielded transformer is a two-winding transformer, usually delta-star connected and serves the following purposes:

• Voltage transformation from the distribution voltage to the equipment's utilization voltage.

• Converting a 3-wire input power to a 4-wire output thereby deriving a separate stable neutral for the power supply wiring going to sensitive equipment.

• Keeping third and its multiple harmonics away from sensitive equipment by allowing their free circulation in the delta winding.

• Softening of high-frequency noise from the input side by the natural inductance of the transformer, particularly true for higher frequency of noise for which the reactance becomes more as the frequency increases.

• Providing an electrostatic shield between the primary and the secondary windings thus avoiding transfer of surge/impulse voltages passing through inter-winding capacitance.

FIG. 27 shows the principle involved in a shielded transformer. The construction of the transformer is such that the magnetic core forms the innermost layer, followed by the secondary winding, the electrostatic shield made of a conducting material (usually copper) and finally the primary winding. FIG. 28 shows this detail. It can be seen that the high-frequency surge is conducted to ground through the capacitance between the primary winding (on the left) and the shield, which is connected to ground. Besides the shield, the magnetic core, the neutral of the secondary winding and the grounding wire from the electronic equipment are all terminated to a ground bar, which in turn, is connected to the power supply ground/building ground. It’s also important that the primary wiring to and secondary wiring from the isolation transformer are routed through separate trays/conduits. If this is not done, the inter-cable capacitances may come into play negating the very purpose of the transformer.

FIG. 27 Principle of a shielded two winding transformer

FIG. 29 shows the proper way for an isolation transformer to be wired. Note that the AC power supply wiring and the secondary wiring from the transformer are taken through separate conduits. Also, the common ground connection of the isolation transformer serves as the reference ground for the sensitive loads. The AC system ground electrode connection is taken through a separate metal conduit. If these methods are not followed and wiring/earth connections are done incorrectly, noise problems may persist in spite of the isolation transformer.

FIG. 28 Construction of a shielded two-winding transformer

FIG. 29 Wiring/earthing of a shielded two-winding transformer

Avoidance of earth loop

We discussed earlier in this section about the earth loop being a primary mechanism of noise injection into sensitive signal circuits. One of the important noise mitigation measures is therefore the avoidance of ground loops altogether. We have also seen in the previous sections that while keeping a separate ground for the sensitive equipment may resolve noise issues, it’s an unsatisfactory solution from the safety point of view.

The correct approach is therefore to keep a common electronic ground but bond it firmly with the power system ground at the source point. FIG. 30 shows an installation with a ground loop problem.

FIG. 30 Ground loop problem

Here, the main computer system (bottom) and its user terminal are shown connected to the power circuit (including ground wiring) at two different points. A communication cable runs between the computer and its terminal. A ground loop is thus formed with the length of communication cable and the ground wire acting together in series.

FIG. 31 shows one way in which this loop can be tackled, by bringing the two power and ground connections together to outlets at a single point.

This arrangement may not be feasible or practical to adopt. What is really possible is to introduce additional impedance in the ground loop so that the high-frequency noise prefers to take another low-impedance path and diverts itself away from the communication path. This is the principle behind the use of a longitudinal (or 'balun') transformer. FIG. 32 demonstrates the action of this method.

FIG. 31 Solution to the ground loop problem

FIG. 32 Use of a 'balun' transformer for noise mitigation

Use of insulated ground (IG) receptacle

The IG receptacles are used in situations where we wish to avoid the mixing of sensitive equipment ground and the building power system ground at all points except the power source (say, the secondary of the shielded isolation transformer) thus avoiding ground loops from forming. FIG. 33 shows such a receptacle. The receptacle frame has a separate ground connection, which is bonded to the general ground system through the metallic conduit to ensure safe conditions. But the grounding wire from the sensitive equipment is an insulating wire, which runs through the conduit directly to the ground point of the source. FIG. 34 illustrates such a connection.

FIG. 33 An exploded view of IG receptacle

FIG. 34 Grounding while using an IG receptacle

Zero signal reference grid and signal transport ground plane

From the foregoing, it will be clear that correct ground connection is a key factor for error-free operation of sensitive equipment and elimination of ground loops to the best possible extent is of extreme importance.

A practical way in which the above can be achieved is by using the support structures of the raised floor (which are common in computer installations and control rooms) as a ground grid called the zero signal reference grid (ZSRG). The grid is formed by the support structures of the raised floor usually arranged as 2 ft square tiles. Copper conductor of #4 AWG size is clamped to the structures forming a grid. All signal grounds of the sensitive equipment and enclosures of the equipment are connected to this grid by short grounding leads. The grid itself is connected to the power ground through more than one conductor. It’s ideal to place the isolation transformer also on this grid and connect the secondary neutral point to the reference grid. FIG. 35 shows the construction of a ZSRG installation.

FIG. 35 Zero signal reference ground using a cavity floor

When communication cables are used to interconnect two sensitive equipment, use of a signal transport ground plane (STGP) is recommended. This is a copper foil or a GI sheet on which the communication cable is placed so that it’s shielded from electrostatic transfer of noise. Metallic cable trays on which a cable is placed and clamped close to it can also serve as an STGP. Within the same room the STGP can be bonded to the ZSRG at one or more points. When a cable runs between installations in different parts of a building, it will be necessary to have individual ZSRGs in each area and also ensure that the STGP is bonded at either end to these grids. In case balun transformers are also used on the signal cable, noise will be further reduced. FIG. 36 shows such an installation.

Use of ZSRG has an added bonus too. It provides numerous parallel grounding paths and thus avoids resonance situation. Resonance happens when the length of a ground lead coincides with quarter wavelength of the noise frequency (or ¾, 1¼, etc.) causing the earth lead to act as an open circuit to these frequencies. With multiple ground paths, this is unlikely to happen.

FIG. 36 Use of signal transport ground plane

A judicious mix of ZSRG, shielded transformer and STGP used appropriately will go a long way in avoiding grounding related noise problems.

Raised floor-supporting structure as a signal reference grid:

• Bolted down stringers (struts between supporting posts) assure low electrical resistance joints.

• Isolation from building steel except via computer system earthing conductors, conductor to computer systems central earthing point and to power source earth.

• Ideal floor height for crawling access is 30 in. less than 18 in. restricts airflow.

For larger computer rooms, install firewall separation barriers to confine fire and Halon extinguishing gas.

Self-resonance effect in grounding/bonding conductors

When dealing with grounding of power system conductors, we are concerned with the ground circuit resistance. But when dealing with circuits with high-frequency signals, it becomes necessary to consider the impedance of the grounding conductor. The grounding electrode conductors (the conductors that connect a system to the ground electrode) exhibit distributed capacitance and inductance in the length of the conductor. A particular conductor may resonate at a certain frequency (or it multiples) and may thus behave like an open-circuited conductor (refer FIG. 37).

It’s therefore advisable that grounding circuits of such systems be connected using multiple conductors with different lengths so that the combined grounding system does not resonate as a whole for any frequency. The ZSRG thus fulfills this need by providing multiple ground paths of differing lengths so that the ground path always has low impedance for any signal frequency (refer FIG. 38).

FIG. 37 Impedance variation of a typical grounding electrode conductor

FIG. 38 Effective grounding by ZSRG

Harmonics in electrical systems

The subject of harmonics is not directly related to this course, except that it’s also a contributory factor in electrical noise. It also causes several other problems in power circuit components such as motors, transformers and capacitor banks.

A load that is purely resistive has the same wave shapes for voltage and current. Both are normally pure sinusoids. Most induction motors fed directly from AC mains also behave in a similar manner except that they draw some reactive load as well. The current waveform is still sinusoidal (see FIG. 39). However, the current waveform gets distorted when power electronic devices are introduced in the system to control the speed of motors. These devices chop off part of the AC waveform using thyristors or power transistors, which are used as static switches.

FIG. 39 Voltage and current waveforms of an induction motor

Such altered waveforms may be mathematically analyzed using Fourier transforms as a combination of vectors of the power frequency (50/60 Hz) and others whose frequency is a multiple of the power frequency. The power frequency component is called the fundamental and higher multiples are called harmonics. It should be remembered that all electrical generators produce only voltage at fundamental frequency. But there has to be a source if a harmonic current has to flow. It’s therefore construed theoretically that all harmonic-producing loads are current sources of harmonics. These sources drive harmonic currents through the rest of the system consisting of the source as well as other loads connected to it. These currents flowing through the different impedances of the system appear as harmonic voltages. It’s usual for the voltage waveform of such a system to appear distorted. Also, the harmonic currents flowing through the other loads of the system give rise to several abnormalities (refer FIG. 40 for the effects harmonics have on different system components).

Effects of harmonics:

Capacitors Amplify harmonics on electrical distribution system Electrical wiring Phase and neutral conductors undersized Engine generators Transferring capability and operation disrupted Induction motors May fail prematurely due to fifth harmonic Metering Inaccurate measurement of power Over-current protection Breaker and fuse nuisance tripping Sensitive electronic loads Voltage drop between neutral and earth Transformers Decreased efficiency and overheating Uninterruptible power systems Line and load interaction

FIG. 40 Effects harmonics have on different system components

Let us see how these shunt filters function. We can use a computer to show what happens as harmonics are filtered from a distorted wave. The example chosen is a 120° square wave current with a 10° commutation time; a typical line current waveform for a DC motor drive and for many AC drives. Here is the square wave before any filtering. The distortion factor is 26% not too pretty a waveform ( FIG. 41a). Now let us take out the fifth harmonic.

This may not look a whole lot better, but the distortion factor is down from 26 to 18%, so things are improving ( FIG. 41b). Now let us take out the seventh as well.

Things are actually looking better now. We can see the sine wave starting to emerge. Distortion factor is down to 11% ( FIG. 41c). Next, we take out the eleventh.

Still no beauty queen, but the distortion factor is now only 8% ( FIG. 41d). Let us add in the final element and remove the thirteenth harmonic.

This is our final current waveform ( FIG. 41e). The distortion factor is 6%, so we are putting a reasonable current into the utility. Of course, the significance of this current waveform to the voltage distortion would depend on the source impedance and the current level.

FIG. 41 Reduction of harmonics by filters

Higher-frequency harmonics can be propagated by the power conductors acting as antennae and appear as induced noise voltages in nearby signal circuits.

It’s not possible to prevent harmonic currents altogether. But they can be prevented from flowing through the entire system by providing a separate low-impedance path for them. This is done by the use of adequately rated series tuned circuits consisting of a reactor and capacitor, which have equal impedance at a specific harmonic frequency.

Several such tuned banks (one for each harmonic frequency) will be needed to totally divert all harmonics away from the system. However , for practical reasons, only a few of the lower order harmonics with larger magnitudes are filtered out, which is adequate to provide substantial reduction of harmonic content.

FIG. 41 shows how a filter might remove the high-frequency components and how the wave shape might appear as the removal takes place. A full treatment of this subject is beyond the scope of this guide.


In this section, we have dealt with electrical noise in detail and the ways in which noise finds a path into sensitive signal circuits. We learnt the various methods by which noise can be reduced by avoiding shields, by separating the cabling, by using shielding transformers, by eliminating earth loops and by using zero signal reference grounds and signal transport ground planes. We also briefly dealt with the generation of harmonics and how they can be filtered.

Top of Page

PREV.   NEXT   Guide Index HOME