ScopeThis guide provides a simple introduction to the theory of how light flicker is produced andhow it is perceived. It includes a description of the international standards that define howvoltage fluctuation should be measured and what levels are needed to reduce light flickerto acceptable levels. There are also sections describing what you can do to prevent orreduce the impact of voltage fluctuation.

IntroductionOne of the many aspects of power quality that can affect electricity users is voltagefluctuation that causes lights to flicker. Whenever the electrical load changes, the supplyvoltage is affected proportionately. Most people have seen this occur in their house; whenthe refrigerator or the furnace starts, some of the lights may dim. If a large enough changeoccurs, such as the start-up of a large industrial motor, lights can dim or brighten, not onlyfor that customer, but all over town. Normally the customer whose load is changing is themost affected, but other customers are all affected to some degree, depending on howlarge the load change is and how close they are to the changing load.

If large voltage changes occur in rapid succession, light levels will vary, and when thevariation becomes large enough to be noticeable or annoying, the effect is called light flicker.Experience has shown that most people are tolerant of an occasional dip in the lights, butwhen flicker is frequent or continuous, they begin to complain.

How flicker is perceived

The response of the eye

Like other human senses, the eye has amazing capabilities. Without conscious thought, oureyes can adapt to light levels varying by a factor of 10,000 from bright sunlight to faintstarlight. Despite that enormous dynamic range, we can detect variations in brightness ofless than 1%. Even more impressive, dedicated parts of the brain filter the incominginformation, removing background clutter and extracting the most important features. Oureyes are also extraordinarily sensitive to rapid change. As the light level gradually drops by afactor of 100 at dusk, we may be almost unaware of the change, but a 1% step change inambient light level due to a sudden change in voltage will almost certainly attract ourattention.

Fundamental constraints on the response of the eye limit our perception of very rapidevents. The mechanisms of converting light into nerve impulses take a finite time to occur,so the brain has averaging mechanisms that smooth out the time delay between nerveimpulses, presenting us with a constant picture, despite the “bucket brigade” nature of the

Power QualityA guide to voltage fluctuation and light flicker

www.bchydro.com

incoming signals. As a result of this automaticsmoothing, pictures displayed in rapid successionappear continuous, an effect that movies andtelevision use to good advantage.

Gaps shorter than about 50 milliseconds (1/20 ofa second) are filled in, as our processing systemignores them as it would variations in nerveimpulses, and averages them out.

We are, therefore, quite insensitive to changesthat occur at frequencies above about 20 cyclesper second. It is no coincidence that ourtelevision and electric lights run at 60 cycles persecond, fast enough that they do not appear toflicker.

The net result of the balance between averagingand pupil adaptation is that our sensitivity tochanging light levels increases the faster theychange, up to the point where the finiteresponse time of the eye and the automaticsmoothing mechanism begin to come into effect.This begins to occur at about 9 cycles persecond, the frequency at which we are mostsensitive to light flicker.

The effect of voltage on light output

There is one more level of complication involvedwhen we consider the effect of variations insupply voltage. Since we are concerned with theeffect of light flicker caused by voltage changes,we must consider not only the perceptionprocess, but also the conversion of electricity intolight, and the way that the output level of alamp depends on the power supply voltage.Different lamp types (incandescent, fluorescent,arc lamps) respond differently, but in general thelight output is proportional to the powerconsumption, although the power is notnecessarily proportional to the voltage. In ballastedlamps, such as arc lamps and fluorescents, theballast stabilizes the power, which is thereforemore or less proportional to the voltage.

Incandescent lamps act like resistors, so thepower is roughly proportional to the square ofthe voltage (actually, the dependence is changedslightly by the fact that the resistance increasesas the filament temperature increases).

As a result, incandescent lamps tend to be themost sensitive to changing voltage, and so flickercalculations are based on incandescent lamps asa worst case.

An incandescent lamp produces light by passingelectric current through a tungsten filament untilit glows white-hot. We tend to think that whenwe turn on the switch, the light comes oninstantly, but in fact it takes around a tenth of asecond for the filament to heat up completely.Like any other object, it has a certain amount ofthermal inertia, although because it is small, thetime it takes to heat up is short.

As we have seen, moderately fast changes arethe most visible, and the filament time constantis close enough to the frequencies of interestthat it must be taken into account. Thermalinertia provides another smoothing mechanismthat reduces the impact of higher-frequencychanges and slows down step changes.

The thermal inertia depends on the size andshape of the filament, and so this term dictates aslightly different response curve in Europe, where220-volt lamps are used, from North America,where the standard is 120 volts.

Figure 1 shows the light output of anincandescent lamp when the 60 Hz power isturned on. After the filament warms up, the lightoutput has a substantial amount of flicker. Notethat because the filament heats equally withpositive and negative voltages, the flickerfrequency is doubled to 120 Hz, too fast to bevisible to the human eye.

2

Light

Voltage

0 10 20 30 40 50 60 70 80 90 100 time (ms)

Figure 1. The light output of an incandescentlamp during starting

The threshold of perceptibility

All of these factors taken together determinehow likely a given level of flicker is to be noticedby an observer. It is difficult to classify levels ofvisibility; the only quantitative measure is todetermine the threshold of perceptibility. This canbe done in a scientific way by presentingobservers with a light source that has variableamounts of flicker and asking them to press abutton or make a signal when the light seems toflicker. Different individuals have differentsensitivities, so trials with a number of observersare required to achieve repeatable results. Thethreshold of perception is defined as the amountof flicker that is perceptible to 50% of observers.

These experiments have been performed innumerous studies, and a standard flickerperceptibility curve, shown in Figure 2 below, hasbeen drawn based on the results. The curveshows the percentage voltage variation at thethreshold of perception for different flickershapes and repetition frequencies.

Figure 2. Flicker perception threshold vsfrequency

The tests can be done with two types of flicker:smooth sinusoidal variations, or square waveswith sharp step changes, as shown in Figure 3.As you might expect, smooth variations aremuch less perceptible, especially at lowfrequencies, where the adaptation of the eyetends to conceal the changes.

At frequencies near the peak sensitivity (8.8 Hz),voltage variations as small as 0.2% can beperceived. At higher frequencies, the thermalinertia of the lamp filament and the averaging ofthe eye mean that larger changes must occurbefore they are visible. For square variations, theresponse at medium to high frequencies is quitesimilar to that for sine waves, but a little moresensitive. At low frequency, step changes appearas independent events, and the threshold ofperceptibility levels off at a constant value of0.5%.

3

Smooth change

Step change

Flicker Perception Threshold

0 5 10 15 20 25 Frequency (Hz)

3%

2%

1%

0%

Square wave

Sine wave

0 50 100 150 200 Time (ms)

140

135

130

125

120

115

110

105

100

150

100

50

0

-50

-100

-1500 50 100 150 200 Time (ms)

Figure 3. Test voltage waveforms for flickerperception

How flicker is produced

Voltage fluctuation basics

Voltage fluctuation, or variation in the voltage atthe electrical outlet, can be caused by events atmany different points in the power distributionsystem. For most consumers, power generatedby many large generators comes to themthrough a high-voltage electrical transmissionnetwork, or grid. Power flows through this gridat voltages around 100 or 200 kV to a substation,where the voltage is reduced by a transformer toa lower voltage, typically 12 to 25 kV. It thenflows through an underground or overheaddistribution system until it reaches a distributiontransformer, where it is reduced further to theconsumer voltage (typically 120/240 V in asingle-phase service or 120/208 V or 347/600 Vin a three-phase service). The power then flowsthrough a service conductor to the customer‘smeter and distribution panel and through thebuilding electrical system to the outlet or lightfixture.

The voltage at the outlet is determined by twofactors: the generator output voltage and thevoltage drop, or loss, in the transmission anddistribution system.

Rapid variations in the generator output voltageoccur very infrequently. BC Hydro managesvoltages at different points in the grid tomaintain maximum efficiency and proper flow ofpower, and changes to voltage and power floware carried out slowly in a controlled manner.Generators are equipped with automatic voltageregulators to maintain output voltage levelsdespite changes in load, and BC Hydro hasvoltage regulators in its distribution system.

It is only when a major event, such as atransmission line outage due to a lightning strikeor mechanical failure, occurs, or when systemdemand is greater than the combined availablegeneration capacity, that the regulators areunable to maintain the voltage. Under theseconditions there may be a temporary undervoltageor overvoltage until the condition is corrected, or,in the worst case, the system may becomeunstable, causing some areas to lose power

completely and other areas to separate from thegrid. Fortunately, this type of event is extremelyrare – even one or two events a year isconsidered unacceptable – and corrective actionis taken to prevent future repetitions.

Most voltage variations are caused by changes inthe voltage drop in the distribution system. Eachcomponent of the system has losses associatedwith it, and produces a voltage drop that isapproximately proportional to the current flowingthrough it. The ratio of voltage drop to currentflow is called the impedance, and canconveniently be expressed as the percentagedrop in voltage at rated current. Transformerimpedances vary; for example, an impedance of10% means that the output voltage of thetransformer will be 10% lower when it is fullyloaded than when it has no load.

The output voltage of a 120 V distributiontransformer with 10% impedance might varyfrom 126 V at no load to 114 V at rated load.Distribution lines, service conductors and housewiring each produce a voltage drop as well,typically around 3% each at full load. For theseimpedances, the total voltage drop from the gridto the outlet, if every component were loaded toits maximum limit, would be 19%.

In practice, most of the distribution systemcomponents are normally loaded to around 50%of their rating, and because the total distributionload is the sum of many customer loads, thevoltage drop is usually fairly stable, in the rangeof 5% to 10%. Individual customer circuit loadsvary from 0 to 100% of rated load, but even in asingle house it is quite unlikely that all theappliances and lights would be on at the sametime, so the total load of a single customermight vary from 10% to 90%.

Loads change all the time, through manualswitching, thermostats or changes in motorloads. When you push a piece of wood throughyour table saw, the current increases substantiallyas it tries to keep the blade spinning against theincreased resistance, and the current drops againwhen the resistance ceases. In the same way,industrial loads such as wood chippers orconveyer belts draw more or less current,

4

depending on the motor loading. Any time aload changes, it affects the voltages throughoutthe electrical system, although most of thechanges are so small as to be undetectable.

Let us take as an example your 3/4 hp, 120 Vtable saw, plugged into an outlet in the work-shop. When the saw is switched on, it probablydraws about 30 A, or twice the current rating ofthe circuit, for a second or so, then drops backto about 3 A at idle. The 30 A momentary cur-rent will increase the current draw throughoutthe system.

The magnitude of the current increase in thedistribution line is reduced by the distributiontransformer ratio. For a 14.4 kV line, the ratio is120:1, and the load current surge will be 0.25 Aat 14.4 kV.

If we assume the substation transformer is ratedfor 1,000 amps with a 10% impedance, this willcreate a voltage drop of (0.25/1,000) x 10% =0.0025% at the transformer output. A typicaldistribution line one mile long might have animpedance of 3% at a rated current of 300 A,for a further voltage drop of (0.25/300) x 3% =0.0025% at the distribution transformer. It isclear that the impact of the load change on theoverall power system is undetectable.

Figure 4. Total voltage dip at different points

As we get closer to the load, the effects becomegreater. A typical small pole-mounted residentialdistribution transformer is rated 100 A at120/240 V, and if we assume 10% impedance,the starting surge will create a voltage drop ofabout (30/100) x 10% = 3%.

Several houses probably share the transformer,and each of them will see a momentary 3% dipin voltage, large enough that your neighbours’lights will flicker.

The service conductor to a house usually hasabout 2% impedance at 100 A, so all the lightsin your house will see an additional dip of 0.6%,for a total of 3.6%. Number 14 wire has a ratingof 15 A and an impedance of about 0.06% perfoot, so if there is a lamp plugged into the sameoutlet as the saw, 30 feet from the panel, thewiring will drop the lamp voltage by a further3.6%, for a total dip of 7.2%. Figure 4 illustratesthe voltage drops in this example.

This illustrates how voltage variations propagatethrough the system. The closer you are to avarying load, and the farther from the substation,the more you will be affected. Most flicker andvoltage variation problems affect the customerwhose load is changing, or his or her immediateneighbours. The exception, and the most difficultcase to deal with, is when a commercial orindustrial load has frequent changes that arelarge enough to affect the voltage on thedistribution line, and therefore to impact all thecustomers sharing that line. As the exampleillustrates, this occurs only with large loadchanges. Given the distribution system describedabove, to reach the threshold of perception(about 0.5% change for an isolated event), adistribution current surge of about 25 A wouldbe required. For a three-phase load, this isequivalent to a load change of about 1 MW, orthe starting current on a 300 hp motor.

Common flicker sources

Isolated events such as motor starts and processshutdowns can cause changes in voltage, butthese isolated events are not classified as flicker.The public does not generally see these events asa major annoyance, as long as the voltagechanges are infrequent and not large enough tocause problems other than minor perceptiblelight flicker.

The most difficult sources of repetitive flicker arelarge loads that fluctuate at frequencies in thevicinity of the peak sensitivity region, near 10 cycles

5

0.0025 0.005

3.0

3.6

7.2

Substation Line Transformer Panel Outlet

8

7

6

5

4

3

2

1

0

per second. The classic example is an arcfurnace, which has a large random variation inload current due to the nature of the arc. Thevariations are erratic, but often include substantialcomponents in the range of 1 to 20 cycles persecond. For a furnace located in an urban areawith a strong distribution system, flicker maybecome a concern at load levels around 1 MWor more. Any arc furnace located far from asubstation on a relatively weak line could be asource of flicker and should be checked forpossible flicker problems.

Another fairly common flicker source is a woodchipper that chops wood into chips for pulp andpaper or for waste wood disposal. These machinesconsist of a chipping wheel driven by a largemotor. As the wood is fed in at intervals, themotor current changes as the chipping wheel isloaded or unloaded, and affects the voltage asshown in Figure 5.

The fluctuation rate depends on the feed materialand mechanism, and may produce flicker in thesensitive frequency range. Since they are mostlyrural, many of these plants are located at theend of weak distribution lines, and are thereforepossible flicker sources, even if the load is only inthe range of 100 hp or so.

Figure 5. Wood chipper voltage fluctuations

Electric heating, either process heating or spaceheating, is a potential source of flicker, but inmost cases this heating is controlled in a ratherslow cycle by a thermostat or process controller,and the changes are infrequent enough to avoidthe most sensitive frequency range. If the load is

large and the impedance is high enough,however, electric heating can be a source ofannoying flicker, especially in the same building.Particular care should be taken with electronicthermostats, since they can have very fast controlcycles. Controllers with cycle times of a fewseconds or less should be avoided, since theirload cycling can lead to high levels of local flicker.

Periodic variations at higher frequencies that areclose together can produce flicker at their “beat”,or difference, frequency, like two guitar stringstuned close together. This effect is primarily aconcern with multiple adjustable speed drivesrunning at slightly different frequencies, say 50and 55 Hz.

On the source side, diesel generators cancontribute to voltage fluctuations, particularly ifthere are variations in the fuel supply.

These are just a few examples – other load typescan also produce flicker if the conditions areright. The key thing to look for is frequent loadchanges of substantial magnitude in relation tothe distribution line capability. Whenever thistype of load is on line, the potential for flicker isthere.

Types of voltage variation

Flicker sources in the power system may generatesmooth changes or step changes, and anythingin between, and the shape has an impact onflicker perceptibility. A load such as a woodchipper is a good example of a square waveflicker source. As each piece of wood feeds intothe chipper, the power demand increasessuddenly when the chipper bites into the wood,and then drops quickly back to no load when thepiece is consumed.

Loads using modern adjustable speed drivecontrols, on the other hand, tend to resemblethe sine wave variation, since most drives haveadjustable ramping limits that produce a smooth,controlled acceleration and deceleration tominimize stress on the equipment as well asflicker. In many cases, the flicker shape is not aregular periodic change, and a complicatedmathematical calculation is required to estimatethe threshold of perceptibility.

6

Voltage

Current

V125

120

115

A400

300

2000 20 40 60 80 100 120 s

Figure 6 shows an example of an irregular RMS*voltage variation that was measured when anirrigation pump started on a rural distributionline. This sort of variation needs a moreadvanced treatment than simply looking it up onthe perceptibility threshold curve.

Figure 6. An irregular motor starting voltagetransient

Non-repetitive events

In the standards, it is clear that flicker refers onlyto repetitive variations in the power supplyvoltage. Occasional individual events such asmotor starts, utility circuit switching andequipment failures or trips, if they are notrepeated on a regular basis, are specificallyexcluded from the category of flicker.

This is not to say that occasional voltage changesdo not affect power quality; they should certainlybe monitored and kept down to tolerable levels.Such events should be treated in the same way asother disturbances like undervoltage, overvoltage,transients and outages. The most common eventof this type is the starting current surge for aninfrequently started motor. Guidelines are availablefor the maximum voltage dip on motor starting.In a new motor installation, the voltage dipshould be calculated from the starting currentand system impedance, and if the guidelines willbe exceeded, a soft-start mechanism, such asreduced-voltage starting, should be applied.

Flicker measurement

Early flicker measurement

In the early days of the power system,measurement and analysis tools were primitive.Problems with flicker occurred from time to time,and engineers of the day developed empiricalguidelines as to what levels of flicker weretolerable and what levels would likely lead tocomplaints. In keeping with the tools at theirdisposal, the parameters they used weresimplified.

Figure 7. Typical rule-of-thumb flicker limit curves

The guidelines were based on the number ofvoltage changes per minute (or per second orper hour) and the percentage voltage change.

Curves were drawn to show the allowable num-ber of changes of a given magnitude. Figure 7shows some typical examples of this type ofcurve. In this figure, the curve labelled “IEEE141” is the “Borderline of irritation” curve fromFigure 3-8 of IEEE 141. The “Urban limit” and“Rural limit” are curves historically used by BC Hydro‘s distribution department. The“Threshold of perception” curve is taken fromthe curves shown in the IEC flicker standards.

This system, while simple in concept, has severalobvious shortcomings. Clearly there are someapproximations made in simplifying a series ofevents (especially complex ones like those shownin Figure 6) into a single magnitude and repetitionrate. As we have seen, there are also significantdifferences in perception for step changes asopposed to gradual changes, so the slope of thetransition should also be taken into consideration.

7

120

118

116

114

112

110

108

1060 10 20 30 40 50 60 70 Time (seconds)

* RMS (root means square) – refers to the most commonmathematical method of defining the effective voltage orcurrent of an AC wave.

IEEE 141

9

8

7

6

5

4

3

2

1

00.01 0.1 1 10 100 1,000 10,000 Changes per minute

Rural limit

Urban limit

Threshold of perception

Another concern is how to combine differentcomponents, for example, if you have a largechange every 10 seconds and a smaller changeevery second. There is also the question of howto deal with episodic flicker, where you may haverapid changes for a few minutes once an hour,or once a day.

Figure 8. The IEC flicker meter

The IEC flicker meter

To address some of these issues, the InternationalElectrotechnical Commission (IEC) introducedstandard IEC 868, Flickermeter functional anddesign specifications, in 1986. The standarddefined the design and performance of ameasuring instrument for flicker. The intent wasfor the meter to be used at the lighting locationof interest.

A block diagram is shown in Figure 8. The IECflickermeter monitors a single-phase voltagesignal, which is passed through rectifier, filterand conditioning circuits to produce a signalproportional to the RMS voltage.

This RMS voltage signal is further processed tofilter out the base level and measure the amountof voltage variation that is occurring.

The voltage variation signal is then passedthrough a series of filters that approximate theperceptibility curve shown in Figure 2. Theresulting output signal is called the instantaneousflicker sensation P, and has a value of 1.0 at thethreshold of perceptibility, whatever the flickerfrequency. This signal can then be fed torecording, averaging and statistical analysiscircuits. The flickermeter standard did not definethe statistical analysis to be done, but provided acapability for statistical analysis.

Statistical measures of flicker

In 1991 the IEC issued standard 868-0,Flickermeter part 0: Evaluation of flicker severity,which defined the statistical analysis and criteriafor maximum acceptable flicker levels for a 230 V,50 Hz system. In this document, a statisticalanalysis based on a ten-minute sample is used tocalculate the parameter Pst, the short-termperceptibility index. This index is calculated froma combination of five percentile values, i.e., the Pvalue exceeded 50%, 10%, 3%, 1% or 0.1% ofthe time during the ten minutes. In this case, avalue of 1.0 for the Pst index represents the levelat which flicker is seen as annoying by mostobservers. Below that level, there may be timeswhen flicker is perceptible, but it should be rareenough that it is not annoying. In cases wherethe flicker has a large long-term variability,sampling can be done over an appropriate longinterval, and the series of Pst values can beaveraged using a cube law average (i.e., cube,then average, then take the cube root). Thisensures that occasional intervals of annoyingflicker are given appropriate weighting.

Present-day standards

In 1994 the IEC reorganized the flicker standardsinto IEC 61000-3 and 61000-4-15. The basiccontent of the standards has not changed, butthe standards have been updated and thenumbering has been revised to conform to theoverall IEC numbering scheme and fit in withother power quality standards. The IEEE has twostandards that deal with flicker, IEEE 141 andIEEE 519, but they are moving to adopt standardscompatible with IEC. IEEE task force P1453 has

8

Block 1: Voltage adapter

Block 2: Squaring

Block 3: Filters

Block 4: Square and smooth

Block 5: Statistics

Block 5: Statistics

developed a flickermeter definition appropriateto 120 V, 60 Hz systems, and is currently workingon adopting a standard compatible with the IEC.The IEC is working on changes to incorporate120 V, 60 Hz systems.

Flicker source identification

When light flicker occurs, one of the challengesis to identify the source of the flicker so thatcorrective action can be taken.

In some cases there may be only a single largefluctuating load, and the source of the flicker willbe obvious, but in other cases there may bemultiple large loads connected to a line on whichflicker is observed. In this case, conventional flickermeasurement techniques have little to offer, butadvanced methods can provide a solution.

If a monitor that can measure currents as well asvoltages is installed, the current drawn by eachload can be monitored along with the line voltage.

After recording one or two major load changes,the line voltage can be plotted against loadcurrent. The source impedance can then becalculated, i.e., the change in voltage per amp ofload change. Once this is done, the voltagefluctuation caused by that customer can becalculated from the load current variations. Bycomparing the calculated fluctuation with themeasured fluctuation, a direct measurement canbe made of the percentage of the total fluctuationdue to the customer being monitored, as shownin Figure 9.

In a complicated system, each customer can bemonitored either in turn or simultaneously, andthe contribution of each load to the flicker canbe determined, so that appropriate measures canbe taken.

This technique provides an additional benefit:proposed changes to the supply or the load canbe evaluated. Load current variations multipliedby the impedance provide an estimate of theresulting flicker levels.

Changes to the supply system will change thesource impedance in a predictable way, and theimpact of those changes on the voltage

fluctuation can be predicted, taking into accountmeasured load current variations. Recordingcurrents along with voltages gives the systemdesigner a whole new set of tools to use indealing with flicker. Figure 9 shows some samplevoltage and current waveforms and how theycan be used to separate internal from externaldisturbances.

Figure 9. Flicker source identification waveforms

What you can do to reduce flickerAs described in the section on how flicker isproduced, load changes have the greatest impacton the voltage in their immediate vicinity. Inconsequence, the first electricity customer tosuffer from voltage flicker is often the customerwhose load is causing the flicker. As a rule ofthumb, for typical source impedance at thecustomer panel in the range of 5% to 10%, aload change of about 10% of the panel ratingwill cause visible light flicker. For a residentialcustomer with a 100 A service, load changes ofaround 10 A, or 1 kW, would qualify, while forindustrial or commercial customers the levelincreases in proportion to their total load. Ifrepetitive load changes of this magnitude occur,you are likely to suffer annoying flicker levels.The visible flicker level is determined by the mag-nitude of load changes as well as the sourceimpedance in common between the changingload and the lighting circuit. If you are sufferingfrom local flicker, therefore, you have three basicstrategies to choose from: reducing the loadchanges, reducing the source impedance, ordecoupling the load from lighting circuits.

9

External disturbances

Voltage

Current

0 20 40 60 80 100 120 Time (s)

121

120

119

118

114

116

Reducing the load changes

Flicker magnitude is influenced by threecharacteristics of the load changes: the magnitude,the speed of the change, and the repetition rate.You may be able to control any or all of these toreduce the impact of the flicker. It is sometimespossible to split up a large load, thus reducingthe magnitude and/or speed of the change. Forexample, a large compressed air bank might havethree motors, all controlled by the same pressureswitch so that they start and stop together.

By simply changing to three independent pres-sure switches, set at slightly different points, thestarts and stops can be staggered, substantiallyreducing the resulting flicker. Similar strategiescan be used for many loads like heaters,refrigerators, pumps and compressors, which arenot time-critical. In some cases it may be difficultor costly to break up the load, but in other casesa small investment in controls may provide asubstantial benefit in flicker reduction. Some ofthese strategies can pay off in conservation aswell, running only a portion of the load unlessthe full power is needed, and thus reducing thelosses.

Other options include using soft-starters,reduced-voltage starting of motors or installingvariable speed drives with slow starting ramps.With a manufacturing process, it may be possibleto slow or stagger the process start throughsimple program changes.

Reducing the source impedance

For most low-voltage equipment, the majority ofthe source impedance is in the service drop, step-down transformer and low-voltage wiring.

If long distances are involved, the voltage drop inthe low-voltage wiring can be substantial, and youmight consider increasing the conductor size,which will not only reduce fluctuation on loadssharing those conductors, but will also reducelosses and result in some energy savings.

The transformer impedance depends mainly onthe size of the transformer, so the higher thetransformer rating, the less flicker a connectedload will produce on the secondary side of the

transformer. In some cases it may be appropriateto change transformers, or switch loads betweentransformers, to minimize flicker.

In general, the higher the voltage, the lower theeffective impedance, so there may be caseswhere increasing distribution or end-use voltagewill provide an effective remedy.

Reducing the coupling

The flicker induced in a light source by a changingload depends on the impedance that the twoloads share. Often the simplest solution to localflicker, or even to area flicker, is to change thesource connections for the lighting load.Industrial and commercial sites often have severalsupply transformers, which allows for someflexibility. If the site has one large fluctuating load,feeding that load from a dedicated transformer,or at least removing all lighting loads from thattransformer, can eliminate or greatly reduce flicker.The worst case of lighting loads connected tothe same circuit as the fluctuating load shoulddefinitely be avoided, and lighting circuits shouldbe powered from a flicker-free circuit or trans-former whenever possible. Another possibility ischanging phases, since most lighting is singlephase. Simply changing phases to a phase with asmaller variable load may reduce flicker problemssubstantially in some cases.

Lighting changes

Occasionally, flicker can be perceived because thetype of lighting used is especially sensitive tovoltage fluctuations. If other remedies are veryexpensive and the area of flickering lights is limited,then changing the type of lighting could beexplored. Perhaps the most sensitive type oflighting is incandescent lighting on a commonthyristor-based dimmer switch. If reduced light-ing levels are desired, sometimes staged orblocked fluorescent lighting can achieve thesame effect as incandescent light dimming.

Active countermeasures

Recent advances in power electronics have led tothe availability of several different brands of staticvoltage compensators (SVC). These devices usesolid-state switching of inductors or capacitors to

10

rapidly compensate for changing line load, thusstabilizing the distribution voltage. The principle isthe same as with switchable line compensationcapacitors or tap changers, but they operate quicklyenough to reduce voltage flicker substantially.

These compensators are expensive, but they mayprovide a cost-effective alternative to upgradinglines when simpler remedies fail. Figure 10 showsan example. In addition to reducing flicker, staticvoltage compensators have been reported toimprove productivity of an arc furnace by 15% to17% by supplying energy to the arc so thatquicker, more efficient melts occur. This increasein productivity typically pays for the SVC in lessthan two years.

A number of large electrical equipmentcompanies manufacture flicker compensators. Atpresent these include the ABB Statcom, theMitsubishi D-Statcom, and the Siemens Dstatcom.Experience with these devices is limited, but anumber of prototype installations have been inservice for up to a few years.

Figure 10. The effect of active compensation

Planning ahead to avoid flicker

The most cost-effective way of dealing with flickeris to avoid it in the first place. Anytime largevariable loads are added to a distribution line,there is a potential for flicker. A new load of thismagnitude usually requires consultation with BC Hydro prior to installation, and that is the idealtime to identify and deal with potential flickerproblems. The first level of screening is to look atthe magnitude of voltage change that can occur dueto load variation. This is determined by multiplyingthe load change by the source impedance.

If the maximum voltage change (excludinginfrequent events) is less than 0.2%, it is unlikelythat any visible flicker will occur. If large voltagechanges can occur, the situation should bereviewed by BC Hydro staff or consultants todetermine more accurately whether flickerproblems are likely.

If so, remedial measures can be implemented,and the costs can be properly assessed againstthe new load.

Following this process, you can make informeddecisions about the cost and benefits of the newload, and, with BC Hydro, you can devise themost effective means of avoiding flickerproblems.

11

Difference

Voltage

Current

0 20 40 60 80 100 120 Time (s)

130

128

126

124

122

120

118

116

114

112

Compensator turned on

Current

Voltage dips reduced & smoothed

ReferencesA number of publications are availablecontaining more detailed information on flicker.The most definitive information is in thestandards listed below. There are also severaltextbooks and handbooks available through theIEEE web site. The article “Voltage and LampFlicker Issues,” available on the web, contains anexcellent technical review of the issues involvedin applying the IEC standards in a North Americancontext. Anyone interested in becoming involvedin the process should visit the IEEE task forcewebsite and contact a member of IEEE task forceP1453 for further information.

Websites

IEC home page: www.iec.ch

IEEE home page: www.ieee.org

IEEE task force home page:grouper.ieee.org/groups/1453

Handbooks

IESNA Lighting Handbook – 9th Edition,Illuminating Engineering Society, 2000

Articles

S.M. Halpin, R. Bergeron, T. Blooming, R.F. Burch,L.E. Conrad, T.S. Key, “Voltage and Lamp FlickerIssues: Should the IEEE Adopt the IECApproach?”,http://grouper.ieee.org/groups/1453/drpaper.html

M. Sakulin and T.S. Key, “UIE/IEC Flicker Standardfor Use in North America: Measuring Techniquesand Practical Applications”, Proceedings ofPQA’97, March 1997, Columbus OH.

G.F. Reed, M. Takeda, J.E. Greaf, T. Aritsuka, T.Matsumoto, Y. Hamasaki, Y. Yonehata, A.P.Sidell, R.E. Chervus, “Application of a 5 MVA,4.16 kV D-Statcom system for voltage flickercompensation at Seattle Iron & Metals”, IEEESummer Power Meeting, July 2000

AcknowledgementThis brochure was adapted from material originallypublished by the Power Quality Service Center,Portland, Oregon, and is used under licence.

Legal noticeThis work has been prepared by Performance EnergyPartnership, Inc. (PEP) and Ferraro, Oliver & Associates,Inc. (FOA) under contract with Northwest Power QualityService Center (NWPQSC) and modified by BC Hydro.Neither PEP, FOA, NWPQSC, BC Hydro nor any personon behalf of any of them makes any warranty withregard to this material, including, but not limited to,the implied warranties of merchantability and fitnessfor a particular purpose. Neither BC Hydro norNorthwest Power Quality Service Center (including itsdealers or distributors) shall be liable to the purchaseror any other person or entity with respect to anyliability, loss or damage caused or alleged to be causeddirectly or indirectly by this manual. Northwest PowerQuality Service Center and BC Hydro shall not be liablefor errors contained herein or for incidentalconsequences damages in connection with thefurnishing, performance or use of this material.

12

For more information contact:

Your Key Account ManagerPhone: Lower Mainland – 604 453-6400, other areas – 1 866 453-6400Fax: 604 453-6280

A04-587h March 2005This brochure was printed on an environmentally friendly paper, made from100% post-consumer waste and processed chlorine free. Please recycle.