Industrial Power Transformers -- Operation and maintenance [part 3]

Home | Articles | Forum | Glossary | Books

AMAZON multi-meters discounts AMAZON oscilloscope discounts


Basic theory

One definition of noise is 'an unpleasant or unwanted sound.' 'Sound' is the sensation at the ear which is the result of a disturbance in the air in which an elementary portion of the air transfers momentum to an adjacent elementary portion, so giving that elementary portion motion. A vibrating solid object sets the air in contact with it in motion and thus starts a 'wave' in the air. Any movement of a solid object may cause sound provided that the intensity and frequency are such that the ear can detect it.

Thus any piece of machinery which vibrates radiates acoustical energy.

Sound power is the rate at which energy is radiated (energy per unit time). Sound intensity is the rate of energy flow at a point, that is, through a unit area.

To completely describe this flow rate the direction of flow must be included.

Sound intensity is thus a vector quantity. Sound pressure is the scalar equivalent quantity, having only magnitude. Normal microphones are only capable of measuring sound pressure, but this is sufficient for the majority of transformer noise measurement situations.

A sound source radiates power. What we hear is the sound pressure, but it is caused by the sound power emitted from the source. The sound pressure that we hear or measure with a microphone is dependent on the distance from the source and the acoustic environment (or sound field) in which sound waves are present. By measuring sound pressure we cannot necessarily quantify how much noise a machine makes. We have to find the sound power because this quantity is more or less independent of the environment and is the unique descriptor of the 'noisiness' of a sound source.

Sound propagation in air can be likened to ripples on a pond. The ripples spread out uniformly in all directions, decreasing in amplitude as they move further from the source. This is only true when there are no objects in the sound path. With an obstacle in the sound path, part of the sound will be 'reflected,' part 'absorbed' and the remainder will be transmitted through the object. How much sound is reflected, absorbed or transmitted depends on the properties of the object, its size, and the wavelength of the sound. In order to be able to predict or modify sound pressure levels at any position away from a 'vibrating' machine's surface, it is therefore necessary to know both its sound power and its surrounding environmental properties.

Noise emission by transformers in operation is inevitable. It can give rise to complaints which, for various reasons, are difficult to resolve. The two main problems are: first, distribution transformers are normally located closer to houses or offices than are other types of equipment; and, second, since they operate throughout the 24 hours of every day, the noise continues during the night when it is most noticeable.

In approaching the noise problem it is therefore essential to consider not only the engineering aspects, but also to remember that noise is a subjective phenomenon involving the vagaries of human nature.

The subjective nature of noise

The subjective nature of noise is underlined by the standard definition in IEC 60050-801: 1994 Glossary of electrotechnical terms which states that it is 'sound which is undesired by the recipient.' It is thus easy to see how people at a party can enjoy it, while neighbors wishing to sleep find it both disturbing and annoying. It also shows why some sounds such as the dripping of a tap can be classified as noise, especially since intermittent sounds are usually more annoying than continuous ones.

Fortunately, transformer noise is not only continuous, but also largely confined to the medium range of audio frequencies, which are the least disagree able to the human ear. The absence of inherently objectionable features means that the annoyance value of transformer noise is roughly proportional to its apparent loudness. A good starting point for tackling the problem is therefore to determine the apparent loudness of the noise emitted by transformers of different types and sizes.

Methods of measuring noise

The measurement of noise is by no means as simple as that of physical or electrical quantities. Loudness, like annoyance, is a subjective sensation dependent to a large extent on the characteristics of the human ear. It must therefore be dealt with on a statistical basis, and research in this field has shown that the loudness figure allocated to a given sound by a panel of average observers is a reasonably well-defined function of its sound pressure and frequency.

Since sound pressure and frequency are the objective characteristics measured by a sound level meter, it is possible to obtain a rating proportional to the loudness of a sound from the appropriate meter readings. A sound level meter is illustrated in FIG. 10, while a more comprehensive analyzing meter is shown in FIG. 11.

FIG. 10 Sound level meter (Brüel & Kjaer)

FIG. 11 Sound level meter with octave band filter (Brüel & Kjaer)

FIG. 12 Equal loudness curves (Robinson and Dadson)

To enable meter readings to be correlated with loudness values, a quantitative picture of the response of the human ear to different sounds must be available. Standardized loudness curves from BS 3383 (now replaced by ISO 226) are reproduced in FIG. 12. They show how the sensitivity of hearing of the average person varies with changes in both the frequency and pressure of the sound. Sensitivity decreases towards the low and high limits of the audio frequency range, so that sounds falling outside the band from approximately 16 Hz to 16 kHz are inaudible to most human observers.

The microphone of sound measuring instruments is in effect a transducer for measuring sound pressures, which are normally expressed in newtons/meter^2 or pascals. Since the sensitivity of the human ear falls off in a roughly logarithmic fashion with increasing sound pressure, it is usual to calibrate instruments for measuring sound levels on a logarithmic scale, graduated in decibels, or dB.

The scale uses as base an r.m.s. pressure level of 20 µPa, which is approximately the threshold of hearing of an acute ear at 1000 Hz. Thus noise having an r.m.s. pressure level of d pascals (or d newtons/meter^2 ) would be said to have a sound pressure level of 20 log10 d/0.00002 dB. The decibel scale is used for the ordinate of FIG. 12, each 20 dB rise in sound level representing a ten fold increase in sound pressure.

The curves of FIG. 12 represent equal loudness contours for a pure note under free-field conditions. They show that the average human ear will ascribe equal loudness to pure notes of sound level 78 dB at 30 Hz, 51 dB at 100 Hz, 40 dB at 1000 Hz 34 dB at 3000 Hz. 40 dB at 6000 Hz and 47 dB at 10 000 Hz.

Thus at 30 Hz the ear is 38 dB less sensitive than at 1000 Hz and so on.

The loudness level of any pure tone is numerically equal to the decibel rating of the 1000 Hz note appearing to the equally loud. From this definition, it follows that the loudness level of any 1000 Hz tone is equal to the decibel rating. At other frequencies this does not hold, as the figures in the previous paragraph show.

Determining loudness

The equal-loudness curves show how the sensitivity of the ear varies with frequency, but do not indicate how the ear responds to changes in sound pressure level. For this purpose the sone scale of loudness has been standardized. The reference point of this scale is taken arbitrarily as a loudness of 1 sone for a level of 40 phons, that is 40 dB at 1000 Hz. It has been found that each rise or fall of 10 phons in loudness level corresponds to a doubling or halving respectively of the loudness (FIG. 13).

The sone scale is linear, so that a noise having a loudness of 2A sones sounds twice as loud as a noise of A sones. It should be noted that the noise emitted by two similar sources does not sound twice as loud as the noise emit ted by each source separately. The sound pressure level is increased only by 3 dB and the apparent loudness by about one-quarter.

Sound measuring instruments

The equal loudness contours shown in FIG. 12 were used to derive simple weighting networks built into instruments for measuring sound level.

FIG. 13 Relation between loudness and loudness level.

Sound level is defined as the weighted sound pressure level. The construction of a sound level meter is shown diagrammatically in FIG. 14. Historically, A, B and C weighting networks were intended to simulate the response of the ear at low, medium and high sound levels, respectively. However, extensive tests have shown that in many cases the A-weighted sound level is found to correlate best with subjective noise ratings and is now used almost exclusively. Although C weighting is retained in more comprehensive meters, B weighting has fallen into disuse.

FIG. 14 Block diagram for sound level meter

The meter illustrated in FIG. 11 and shown in block form in FIG. 14 offers A and C weightings at the touch of a button, and also a linear (unweighted)

option for frequency analysis purposes and where the actual sound pressure level is to be measured.

The microphone used in a sound level meter is non-directional and the A-weighted frequency characteristic and dynamic response of the meter closely follow that of the human ear. As the range of the ear is around 140 dB, while the meter illustrated has a linear 30 dB scale, attenuators are necessary to cover the full measuring range required. The range switch is adjusted until a convenient scale reading is obtained, and the sound level of the noise is then the sum of the meter reading and of the attenuator setting.

If noise is fluctuating very rapidly, the meter response may not be fast enough to reach the actual level of a noise peak before it has subsided again.

The meter illustrated will, however, measure and display the maximum r.m.s. level of a noise event at the touch of a button.

A sound level meter effectively sums up a given noise in terms of a single decibel value. Although sufficient for many requirements, this yields little information as to the character of the noise, as it represents only its magnitude.

To determine the character of noise, a frequency spectrum must be measured by means of an audio frequency analyzer such as that illustrated in FIG. 15.

This instrument is essentially a variable filter which suppresses noise components at all frequencies outside the desired band. As it is tuned over the audio band, any marked amount of noise at a particular frequency is clearly demonstrated by a sharp rise in the meter reading. From the readings obtained, a continuous spectrum can be derived.

Where less discrimination is acceptable, a filter of wider bandwidth may be used to sum up all components of the noise in a certain frequency range. The most common bandwidth is one octave, although one-third octave filters are also used for more precise applications. The mid-band frequencies are inter nationally standardized; for octave band filters they are 31.5, 63, 125, 250, 500, 1000, 2000, 4000 Hz and higher for precision (Class 1) grade meters. The 250 Hz band, for example, spans the octave 180-360 Hz.

The day-to-day performance of a sound level meter is usually checked periodically using a calibrator. The latter produces an accurately known sound level against which the meter can be set up. To ensure that the calibration is not affected by extraneous noise, the calibrator is usually fitted over the micro phone to form a closed cavity. This not only greatly reduces ambient noise, but also ensures that the source to microphone spacing is exactly the same at every calibration.

FIG. 15 Audio frequency analyzer (Brüel & Kjaer)

Sound level measurements of transformers

As explained above, in making measurements of noise at a particular point in space using a microphone, the quantity measured is the sound pressure level.

The quantity is expressed in decibels, usually with an A scale weighting, and abbreviated as dBA. For many years transformer users and manufacturers quantified the noise produced by a transformer in terms of these microphone readings to provide an average surface sound pressure level or average surface noise level which was an average of sound pressure level readings taken at approximately 1 m intervals around its perimeter at a distance of 0.3 m from the tank surface. As a means of comparing the noise produced by individual transformers this provided a fairly satisfactory method of making an assessment. Clearly, a transformer with an average surface noise level (usually simply termed 'noise level') of 65 dBA was quieter than one having a noise level of 70 dBA. However, with recent environmental requirements demanding low noise levels, it has become necessary to be able to predict the sound pressure level at a distance of, say, 100 m from the substation. It is therefore essential to know the sound power level of the transformer(s). This is expressed in terms of the integral of sound pressure over a hemispherical surface having the transformer at its center. The units of measurement remain decibels. This approach has the benefit of allowing the noise contribution from the transformer to be assessed at any distance and the contributions from different sources to be added (applying an inverse square law to the distance and adding logarithmically), and is now the preferred method by noise specialists for expressing transformer noise levels. There is, unfortunately, confusion between the two quantities which is not helped by the fact that both are measured in the same units. Many transformer users still specify average surface noise level when procuring a transformer or expect the sound power level to be the same in numerical terms as the average surface noise level. In fact, in numerical terms the sound power level is likely to be around 20 dB greater than the average surface sound pressure level. The actual relationship will be derived below.

In the UK noise measurements were formerly made in accordance with EN 60551 Determination of transformer and reactor sound levels. In 2001, the standard covering transformer noise measurement was incorporated into the EN 60076 series of standards with the issue of Part 10 of that document.

EN 60076-10 introduced further changes into transformer noise measurement practice in addition to the change in emphasis from sound pressure levels to sound power levels identified above. The first of these changes followed as a result of the significant reductions that have been made in recent years in transformer core noise with the result that noise produced due to the flow of load current in transformer windings, which had hitherto generally been ignored when specifying and measuring noise levels, can no longer automatically be regarded as negligible. A method of determining the noise contribution due to the flow of load current is therefore included. The second change is to allow an alternative method of deducing the sound power of a transformer based on the measurement of sound intensity rather than sound pressure. It is claimed that this alternative method is less susceptible to errors due to environmental factors such as external sound sources, reflections and standing waves.

When the sound pressure method of measurement is used, EN 60076-10 recommends that the test environment should represent as near as possible free-field conditions, that is it should be essentially undisturbed by reflections from objects and the environment boundaries but, recognizing that such an ideal environment cannot generally be found within a transformer factory, the method requires that an environmental correction based on the size of the transformer radiating envelope in relation to the boundary surface enclosing the test bay is calculated and subtracted from the average A-weighted sound pressure level obtained from the test measurements.

If the sound intensity method of measurement is to be used, the standard again recommends that an environment essentially undisturbed by reflections from nearby objects and environment boundaries should be provided, but states that the method allows accurate determinations to be made with up to two reflecting walls so long as these are at least 1.2 m from the radiating envelope of the test object or three reflecting walls at least 1.8 m distant.

Sound pressure measurements are made using a type 1 sound level meter complying with EN 60651. A check of the meter using a calibrated noise source should be made before and after the measurement sequence.

If using the sound intensity method, measurements are made with a Class 1 sound intensity meter complying with EN 61043. This instrument measures sound intensity by the use of a pair of pressure sensing microphones separated by a fixed distance so as to provide the directional element. The method of calibration of this instrument is described in EN ISO 9614-1.

The methods of making measurements are essentially similar regardless of which system is used. It is first necessary to determine the principal radiating surface. This is the surface defined by the vertical projection of a string con tour encircling the transformer from floor level up to the height of the tank cover. The string contour is to include all cooling equipment attached to the tank, tank stiffeners, cable boxes, tapchanger, jacking and transport lugs, etc., but exclude any forced air cooling auxiliaries, bushings, oil pipework, valves or any projection above the tank cover height.

Separate free-standing coolers mounted at a distance 3m from the transformer are treated as separate sound sources, the principal radiating surface being that enclosed within a string contour encircling the equipment but excluding conservators, framework, pipework valves and other secondary devices.

For transformers with a tank height less than 2.5 m, measurements are taken at half the tank height. For transformers with a tank height equal to or greater than 2.5 m, measurements are taken at one-third and two-thirds of the tank height. Measuring points around the tank perimeter are to be spaced not more than 1 m apart. For transformers having no forced cooling, or with forced cooling equipment mounted on a separate structure at least 3 m distant from the main tank, or for dry-type transformers installed within enclosures, the micro phone is placed at a distance of approximately 0.3 m from the principal radiating surface (see FIG. 16).

For measurements made on separate free-standing coolers up to a height of 4m (excluding conservators) measurements are taken at half the cooler height.

For coolers with a height 4m, measurements are taken at one-third and two thirds of the cooler height.

Sound pressure level measurements are taken at no-load and all readings are recorded using the A-weighting. If an octave band analysis is required the linear response is used. The transformer is excited on its principal tapping at rated voltage and frequency, but preliminary check tests may be made to see if there is any significant variation of noise between different tapping positions.

If it is required to perform load-current measurements these are carried out in the same way as the open-circuit measurements except that one set of winding terminals are short circuited and a sinusoidal voltage equal to impedance volts, to circulate short circuit current, is applied to the other terminals. This assumes that exciting the core with impedance volts does not produce any significant core noise.

To enable a decision to be made as to whether load-current measurements should be made the standard gives a formula for obtaining a rough estimate of their magnitude as

(eqn. 1)

where LWA, IN is the A-weighted sound power level of the transformer at rated current, rated frequency and impedance voltage

Sr is the rated power in MVA

Sp is the reference power (1 MVA)

It will be seen that if, for example Sr is 100 MVA, substituting the appropriate value in Eq. (eqn. 1) gives a value of 75 dBA for the approximate rated current A-weighted sound power. The standard goes on to suggest that if the guaranteed total sound power level for the transformer including the core noise is more than 8 dB greater than this approximate rated current contribution, then there is no merit in performing a formal load-current measurement.

FIG. 16 Plan view of transformer with tank-mounted radiators showing principal sound radiating surface and measurement contours for self-cooled and forced-cooled ratings.

For separate cooling structures mounted at least 3 m from the main tank a series of measurements are taken at a distance of 2m from the principal radiating surface with pumps and fans running but with the transformer de-energized.

FIG. 17 shows the location of the principal radiating surface and the microphone positions.

The average A-weighted background noise level is calculated before and after the tests. Initial and final background sound pressure levels should not differ by more than 3 dB, and the higher background level must be at least 8 dB FIG. 17 Typical microphone positions for noise measurement on forced air cooling auxiliaries mounted on a separate structure spaced not less than 3 m away from the principal radiating surface of the transformer tank lower than the uncorrected average measured on the energized transformer, but if the uncorrected average for the transformer is less than guarantee it may be deemed to have met the guarantee regardless of the background measurement.

The average surface sound pressure level is then generally computed by taking a simple arithmetic average of the series of measurements taken around the perimeter of the equipment as described above. Strictly speaking how ever, the average should be logarithmic but provided the range of values does not exceed 5 dB, taking an arithmetic average will give rise to an error of no greater than 0.7 dB. A true average is given by the expression:

(eqn. 2)

where LpA is the A-weighted surface sound pressure level in decibels


is the A-weighted sound pressure level at the ith measuring position in decibels

N is the total number of measuring positions

The corrected average A-weighted sound pressure level, LpA, can then be calculated by subtracting the environmental correction from the average value in accordance with the expression below

(eqn. 3) where K is the environmental correction to take account of test location.

K is generally of the order of 2-5 dB depending on the volume of the test bay in relation to the size of the transformer.

A similar process is used to give the average sound intensity level if sound intensity measurements have been taken except that there is no environmental factor to be subtracted.

Calculation of sound power level The sound power level can be calculated using the sound pressure levels deter mined above by computing the effective area for the measurement surface according to the relevant method of measurement and relating this to the standard measurement surface, which is 1 m2. The A-weighted sound power level is thus

LWA is the A-weighted sound power level in decibels with respect to 10_12

W S is the area of the measurement surface, in square meters with respect to S0 _ 1 m2.

The measurement surface S then has the following values:

For self-cooled transformers, or forced cooled transformers with the forced cooling equipment unenergized, and measurements made at 0.3 m from the principal radiating surface:

S _ 1.25hlm (eqn. 4)

where h is the height in meters of the transformer tank

lm is the length in meters of the contour along which measurements were made

1.25 is the empirical factor to take account of the sound energy radiated by the upper part of the transformer over which no measurements were made

For forced cooled transformers with forced cooling equipment also energized or for measurements on separate free-standing cooling structures:

S_ (h _ 2)lm (eqn. 5)

where 2 is the measurement distance in meters

h is the height in meters of the transformer tank or of the cooling equipment, including fans (see FIG. 18)

FIG. 18 Cooler with forced air cooling auxiliaries showing boundaries of principal radiating surface.

Addition and subtraction of sound power values

Sound power levels can be added and subtracted to provide sound powers for cooling equipment alone when a value has been obtained for transformer and cooler or to add the sound power contributions from iron circuit (no-load measurement) and rated current measurements, but this must be done logarithmically. For example to determine, LWA0, the sound power level of the cooling equipment:

(eqn. 6)

where LWA1 is the sound power of the transformer and cooler

LWA2 is the sound power of the transformer or, the combined A-weighted sound power due to core and rated current,

(eqn. 7)

where LWA,UN is the A-weighted sound power level of the transformer at rated voltage on open circuit

LWA, IN is the A-weighted sound power level at rated current

Interpretation of transformer noise

The point has already been made that in order to obtain accurate measurement of transformers noise these should ideally be made in free-field conditions, and this requirement always presents problems in the investigation of noise related issues. As identified at the start of this section, a great deal of valuable information can be gained by making measurements of average surface noise level, and provided these measurements are always made in the same or similar environment, it is feasible to identify, and to a significant extent quantify, those design and construction measures that will enable a reduction in noise level to be obtained. A free-field environment, or its equivalent, is how ever necessary if absolute measurements are to be obtained.

FIG. 19 Transformer undergoing a noise test in an anechoic chamber (ABB Power T&D Ltd.).

The closest approximation that can generally be obtained to a free-field environment is an anechoic chamber having walls that create zero reflection of incident sound pressure. Such chambers were widely used in early experimentation in connection with transformer noise and FIG. 19 shows a distribution transformer undergoing a noise test in an anechoic chamber. As stated earlier, because distribution transformers can often be located closer to housing than larger transformers, these units were the first to be investigated with the object of reducing noise. There is a limit to the size of such chambers, however, and these cannot normally be provided for large high-voltage transformers (although the automotive industry does use very large chambers). This gave a reason for the development of measuring techniques involving the use of correction factors applied to measurements taken in the normal test bay.

A typical analysis of transformer noise is reproduced in FIG. 20, which can be considered as a composite graph of a large number of readings. In this diagram, the ordinates indicate the magnitude of the various individual constituents of the noise whose frequencies represent the abscissae. The most striking point is the strength of the component at 100 Hz or twice the normal operating frequency of the transformer. Consideration of magnetostrictive strain in the transformer core reveals that magnetostriction can be expected to produce a longitudinal vibration in the laminations at just this measured frequency.

Unfortunately, the magnetostrictive strain is not truly sinusoidal in character, which leads to the introduction of the harmonics seen in FIG. 20.

FIG. 20 Typical analysis of noise emitted by transformers

Deviation from a 'square-law' magnetostrictive characteristic would result in even harmonics (at 200, 400, 600 Hz, etc.), while the different values of magnetostrictive strain for increasing and decreasing flux densities - a pseudo-hysteresis effect - lead to the introduction of odd harmonics (at 300, 500, 700 Hz, etc.). Reference to FIG. 12 indicates that the sensitivity of the ear to noise increases rapidly at frequencies above 100 Hz. On the 40 phon contour, it requires an increase of 12 dB in intensity to make a sound at 100 Hz appear as loud as one at 1000 Hz. The harmonics in a transformer noise may thus have a substantial effect on an observer even though their level is 10 dB or more lower than that of the 100 Hz fundamental.

Although longitudinal vibration is the natural consequence of magnetostriction, the need to restrain the laminations by clamping also leads to transverse vibrations, this effect being illustrated in FIG. 21. Measurements taken on this effect suggest that transverse vibrations contribute roughly as much sound energy to the total noise as do the longitudinal vibrations. As already pointed out, two similar sources sound about 25 percent louder than one. By the same token, complete elimination of the transverse vibration would reduce the loudness of the transformer noise by only about a ? fth. Although valuable, this reduction, even if technically and economically possible, is insignificant com pared with the halving of the loudness which can be achieved by a reduction of 10 dB in the noise level of both longitudinal and transverse vibrations.

FIG. 21 Vibrations de to magnetostriction: (a) longitudinal and (b) transverse

The other main source of noise from the transformer core is due to alternating attractive and repulsive forces between the laminations caused by flux transfer across the air gaps at the leg to yoke and inter-yoke joints. These forces can be reduced by special building and design techniques of which the best known and most widely used is the step-lap form of construction described in Section 4.1.

FIG. 20 covers typical transformers incorporating cold-rolled laminated cores operating at flux densities between 1.6 and 1.8 Tesla. Even variations of 10 percent in flux density have been shown to produce changes of noise level of the order of only about 2 dB, although the character of the noise may vary appreciably. From this it will be apparent that it is most uneconomic to obtain a reduction in noise level by the employment of low flux densities. This is perhaps demonstrated best by reference to experience with cold-rolled steel.

To make the optimum use of these newer materials, it is necessary to operate them at flux densities of 1.65-1.85 Tesla when operating conditions permit this. While this higher flux density tends to lead to a higher noise level for a given size of core, current results suggest that the difference is quite small for a given transformer rating, due to the smaller core made possible by the use of the higher flux density material. Considerable work is being undertaken to obtain even quieter operation by suitable treatment of the raw material and by particularly careful assembly of the finished core laminations.

Turning to other possible sources of noise emitted by a transformer, as indicated earlier, the forces present between the individual conductors in the winding when the transformer is loaded must be considered. These forces are, however, of a sinusoidal nature so that any vibration consists of a fundamental at 100 Hz with negligible harmonics. The fundamental is thus effectively dwarfed by the much greater 100 Hz fundamental generated by the core, while there are no harmonics to add to the annoyance value. Acoustic measurements confirm this conclusion by showing that, except on very large transformers the noise level increases by no more than 2 dB (15 percent rise in loudness) from no load to full load. Any variation is in fact attributable more to changes in flux density than to variations in the forces in the windings.

The other major source of noise is the transformer cooler. Fans produce noise in the frequency range 500-2000 Hz, a band to which the ear is more sensitive than it is to the 100 Hz fundamental produced by the core. The pre dominant frequencies are dependent on many factors including speed, number of blades and blade profile. Sound power level is dependent upon number and size of fans as well as speed and, for many forced-cooled transformers, the cooler can prove to be a significantly greater source of noise than the transformer itself. An example is the transformer shown in plan view in FIG. 16.

This is a 40 MVA ONAN rated 132/33 kV transformer with tank-mounted radiators having an emergency ODAF rating, using two pumps and eight fans, of 80 MVA. By careful attention to core design and use of modern HiB steel, an average surface sound pressure level of only 47 dBA, corrected for back ground, has been achieved at the ONAN rating with a sound power level of 66 dBA. However, with all pumps and fans running for the emergency ODAF rating, the average surface sound pressure level is increased to 60 dBA and the sound power level to 82 dBA.

These comments on transformer noise assume the absence of resonance in any part of the unit. Normally the minimum natural frequency of the core and windings lies in the region of 1000 Hz. FIG. 20 indicates that the exciting forces are very low at this or higher frequencies. Accordingly, it can confidently be expected that the unfortunate effects associated with resonance will be avoided. The natural frequency of the tank or fittings being lower, resonance of these is much more likely to occur, since the vibrations of the core can be transmitted by the oil to the tank. If any part of the structure has a natural frequency at or near 100, 200, 300, 400 Hz, etc., the result will be an amplification of noise at that particular frequency.

Noise reduction on site

Control of the noise emitted by a transformer rests almost entirely with the manufacturer, who must Endeavour to achieve the customer's specified requirements wherever possible. A certain amount of noise is, however, inevitable and, if it proves offensive, must be dealt with by the purchaser who can do much to ensure acceptance of the transformer long before it is delivered.

Typical average sound levels of a range of transformers are given in FIG. 22. They should, of course, be compared with any test figures for the actual transformer to be installed, as soon as any figures become available. The levels quoted in FIG. 22 will, however, provide a reasonable basis for preparatory action. The reduction of noise level with distance must be allowed for.

Doubling the distance from a point source of noise means that a given amount of sound must be spread over 4 times the area. From this cause alone, doubling of distance results in a 6 dB fall in sound level. In practice, scattering combined with the absorption by the air itself ensures that the noise reduction is greater, particularly at the higher frequencies.

FIG. 22 Typical transformer average surface noise levels FIG. 23 Curves showing measured attenuation of transformer noise with distance

FIG. 23 shows measured values of attenuation with distance for typical transformer ratings. Assuming that there is no screening between the transformer and a given building, these curves enable the noise level outside the building to be computed.

Normally it is not necessary to reduce transformer noise in the vicinity of residential buildings to such a level that it is inaudible. Experience suggests that the transformer noise will be acceptable if it is not audible inside a bed room of the nearest house at night time when a small window of the room is open. Under these conditions the transformer noise level outside the house can be considered as the permitted maximum transformer noise at this position.

Provided the sound level meter reading taken outside the house is not more than 2 dB above the bedroom background level, both being measured at the A weighting, the acceptable noise level inside will not be exceeded. From Table 2 which also gives the calculated equivalent 'phon' values of the transformer noise as obtained from the typical composition and equal loudness contours and tone summation curve, it will be seen that three types of background level have been given, and these are considered to be representative of conditions existing at night in neighborhoods of the kinds referred to.

Table 2 Acceptable maximum noise levels outside dwellings [From CIGRE Paper No 108 (1956) Transformer noise limitation. Brownsey. Glever and Harper.]

Using the values from Table 2 as a basis, it is possible to determine whether the noise level within nearby bedrooms will be acceptable if the transformer is sited at various alternative locations. One or other location may well ensure that no householder is subjected to an unduly high noise level.

Failing this, the investigation may still show the minimum attenuation necessary to bring the noise level down to an acceptable level. The most appropriate method of achieving this object can then be selected and work put in hand immediately, so that the site is ready when the transformer is delivered.

Provided the noise level resulting from transformer operation is below that given in the above table, conditions should be satisfactory and no corrective action is necessary. In fact, with well-designed transformers, acoustic conditions will normally be satisfactory under urban conditions at all points beyond 15 m from the transformer for a rating of 200 kVA and 25 m for a rating of 500 kVA. Assuming that bedroom windows do not face directly on the transformer, it is possible to decrease these distances by about two-thirds.


In urban areas, it is normally impracticable to site transformers more than 100 m from the nearest dwelling. In this case, transformers with ratings in excess of about 60 MVA will probably need to be provided with some form of attenuation giving a noise reduction of between 10 and 25 dB.

The most obvious method of attenuation is by the provision of a suitable barrier between the transformer and the listener. The simplest form of barrier is a screening wall, the effectiveness of which will vary with height and density as well as with the frequency of the noise. The attenuation of a 100 Hz noise by a 6m wall will not normally exceed 10 dB outside the immediate 'shadow' cast by the wall.

Such attenuation just reaches the bottom of the range cited but some slight further attenuation can be achieved by judicious use of absorbent material.

This treatment may result in an attenuation as small as 2 dB and will seldom give a figure in excess of 6 dB. While absorbent material may give some relief on existing installations or may make a single wall shielding a transformer in one direction more effective, it will not usually provide a complete solution where an untreated screen wall is itself unsatisfactory.

Noise and vibration from large transformers will also be transmitted via the ground. Ground-borne vibration can cause adjacent structures to vibrate which may then amplify and retransmit the noise. These effects can be reduced by placing the transformer on anti-vibration mountings -- strips of rubber or other resilient material, usually 80 mm wide and 40 mm thick. The number of strips and the spacing of these is arranged to ensure that the loading is optimized as near as possible for the material. They may be simply set out perpendicular to the major dimension of the tank base with an even spacing or in a more elaborate pattern as, for example, in FIG. 24 which aims to provide a more even loading taking into account irregularities in the plinth and the tank base. Whichever arrangement is used, the openings around the perimeter of the tank base should be closed, otherwise the spaces between the pads can provide resonant chambers for amplification of the sound.

FIG. 24 Typical arrangement of anti-vibration mountings for 60 MVA transformer

FIG. 25(a) Typical acoustic enclosure for 1.32 kV transformer

It is generally necessary to provide concrete or steel acoustic enclosures for large generator transformers, since these can be a source of high levels of off site noise. For transmission transformers and grid bulk-supplies transformers the need for noise reduction measures will depend on the location of the sub station, but consideration should be given to specifying that these are at least made suitable for the future fitting of an acoustic enclosure should this be found to be necessary after the transformer has entered service. Such provision also allows for the transformer to be installed at alternative sites, some of which might be more environmentally sensitive than others. Provision mainly involves the installation of bushings on extended turrets which will pass through the structure of the enclosure. A typical acoustic enclosure capable of producing an attenuation of around 25-30 dB is shown in FIG. 25.

The need for enclosures of this type has tended to lessen in recent years in view of the steady improvements in noise reduction measures adopted by transformer manufacturers and, of course, installation within an acoustic enclosure has the major disadvantage that a separate free-standing cooler bank must be provided outside the enclosure. This adds to the overall costs of the transformer, and any fans associated with the cooler will probably contribute considerably more to the off-site noise anyway than would the transformer itself, as will be evident from the typical cooler noise levels quoted above. Fan noise can be reduced by the provision of attenuators in the form of inlet and outlet ducts. These add considerably to the size of the cooler, since to be effective they must have a length of the order of one or 2 times the diameter of the fans on either side of the fan. FIG. 26 shows a large induced draught transformer oil cooler with fans installed within inlet and outlet noise attenuators. The cooler has a total of 18 fans provided to dissipate over 3.3 MW of losses from a large quadrature booster. The attenuators enable a noise level of 60 dBA to be obtained at a distance of 2m from the cooler.

Harmonic content of noise

Provided openings are located at the optimum position, the attenuation given by any reasonable structure will normally be above that necessary to give tolerable conditions in nearby houses. An important factor in this connection is the relatively great attenuation of the harmonic content of the noise, which has been shown earlier in this section to have a nuisance value out of all proportion to its magnitude.

FIG. 25(b) Typical details of acoustic enclosure.

Once such a structure has been erected, adequate maintenance is of the utmost importance if the initial attenuation is to be maintained. Any openings and doors should be checked frequently, as gaps may develop to a sufficient size to permit considerable escape of sound energy.

It is often advisable to compare a frequency spectrum of the noise emitted by the transformer to be installed against the spectrum of the background noise at the proposed location. In the values quoted earlier, typical frequency spectra have been assumed. Any marked deviations of the actual noises from the character attributed to them can lead to a considerable reduction in the masking power of the background noise. For this reason alone, a frequency analysis on site is valuable, even if it is compared only with an average spectrum for transformer noise, such as that given in FIG. 20.

FIG. 26 Induced draft cooler for 2000 MVA quadrature booster.

The cooler has five sections for the shunt unit and four for the series unit. These are designed to dissipate 1900 and 1425 kW, respectively, with one section of each out of service. A noise level of 60 dBA at 2 m from the cooler was obtained by means of inlet and outlet attenuators on each of the fans (GEA Spiro-Gills Ltd.)

Location of transformer

Topographical features of the site should be exploited to the full in order to reduce noise. Where possible the transformer should be located in the prevailing down-wind direction from houses. Existing walls and mounds should, if possible, be kept between dwellings and the transformer. Natural hollows can sometimes be used to increase the effective height of screen walls, as can artificially constructed pits.

Cultivated shrubs and trees form only an ineffective barrier to noise as sound attenuation is largely determined by the mass of the barrier. In some cases where the smaller ratings of distribution transformers are installed, the psycho logical effect, however, may be sufficient to avoid a complaint simply because the transformer becomes hidden by the trees and is therefore not visible.

Top of Page

PREV. | Next | Guide Index | HOME