670

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 29, NO. 3, MAYIJUNE 1993

Harmonics from Compact Fluorescent Lamps

Rudolph R. Verderber, Fellow, ZEEE, Oliver C . Morse, and William R. Alling
Abstract- Measured characteristics of compact fluorescent lamps (CFLs) and incandescent lamps with diode-type devices have low power factor (47 to 67%) and/or high total harmonic distortion (100%THD). There is little effect on a buildings power quality if the CFLs comprise less than 10% of the buildings load. Even if the CFL is as high as 26% of the buildings load, the voltage distortionis less than 5%.Circuits and their estimated costs are presented to correct for power factor and suppress the harmonics.

to reduce the harmonic content of these lamps are described and discussed in terms of their cost and applicability for CFLs. 11. EXPERIMENTAL New CFLs were obtained and burned in for 100 hr. Lamps were operated at 60 Hz with either magnetic ballasts or at high frequency with electronic ballasts. The integrating sphere was calibrated with a standard NBS lamp. The lamps were positioned base up when burned in and measured to assure that the lamps cold spot and mercury reservoir coincided. The lamps operated in this position provided maximum light output and efficacy [l]. The CFL characteristic measured included the light output, input power, voltage, line current, and harmonics (fundamental up to the 32nd). In addition to the CFLs, incandescent lamp systems that had diodes and triacs in series were also measured. The building simulated assumed i) total initial building load was 100 kVA at 0.85 lagging power factor, ii) lighting was 50% of the total load at unity power factor with no harmonics (incandescent), and iii) the CFL replacements have a THD of 115% (0.60 PF); the CFLs voltage and current are in phase and are four times more efficacious than the incandescent lamps.
111. RESULTS

I. INTRODUCTION

HIS PAPER presents some data on the performance of compact fluorescent lamps (CFLs) and other types of devices for use in incandescent sockets to improve the efficacy or reduce energy. These systems have low power and factor andor high total harmonic distortion (THD), their impacts on the power quality is of primary concern. Utilities have also expressed concern since they have been actively recommending their use in their demand-side management (DSM) programs through giveaways and rebates. At present, no serious power quality problems have been reported with the use of CFLs. One reason may be that these lamp systems presently comprise a very small portion of the lamp market and are usually only a small portion of a buildings load. However, if their use continues to escalate such that they are a major portion of a buildings load power, quality may suffer. In fact, the poor power factor and/or high harmonics might limit the use of CFLs as an efficacious replacement for the incandescent lamp. Their potential impact on power quality of the electrical load is described for a building where the lighting is initially 50% of the load and resistive (100% power factor and no harmonics). The lighting load is gradually refitted with a CFL operated at high frequency, and the change in power quality is determined. Although the CFL lamps are generally cost effective, their high initial cost compared with the incandescent lamp has limited their acceptance. Thus, manufacturers have introduced the simplest ballast (magnetic and electronic) designs. These low-cost designs employ neither power factor corrections nor harmonic filters. Several circuit approaches that could be used
Paper IUSD 92-5, approved by the Production and Application of Light Committee of the IEEE Industry Applications Society for presentation at the 1991 Industry Applications Society Annual Meeting, Dearbom, MI, September 28-October 4. This work was supported by the Assistant Secretary for Conservation and Renewable Energy, Office of Buildings and Community Systems, Building Equipment Division of the U.S. Department of Energy under Contract DE-AC03-76F00098. Manuscript released for publication September 9, 1992. R. R. Verderber and 0. C. Morse are with Lawrence Berkeley Laboratories, University of California, Berkeley, CA 94720. W. R. Alling is with Diablo Scientific Laboratories, Danville, CA 94526. IEEE Log Number 9208781.

Lump Pe$ormance
Tables I and I1 list the electrical and optical data collected for the incandescent lamp systems and for the CFLs, respectively. The incandescent lamp with the diode will have high even harmonics because of asymmetry (dc component), whereas the lamp with the triac had high uneven harmonics (no dc component). The two incandescent lamps with diodes of opposite polarity are a resistive load. Table I1 shows that all of the CFL tested have low power factors (between 47 and 67%). The low power factor of the lamps operated at high frequency is primarily due to high total harmonic content with some leading phase component. One should notice that the ballast factor determines the lamps light output and varies from a low of 62% to over 114%. It is essential for the lighting designer to know a ballasts ballast factor for the CFL in order to obtain the specified illuminance. The data shows CFLs operated at high frequency are sometimes twice as efficacious as CFLs operated at 60 Hz. Building Model Table I11 list the power quality of the building when sockets are refitted with CFLs. When the CFL replacements are about

0093-9994/93$03.00 0 1993 IEEE

~-

VERDERBER et al.: HARMONICS FROM COMPACT FLUORESCENT LAMPS

611

TABLE I LAMPS INCANDESCENTTYPE

IOOW POWER (W) CURRENT (A) POWER FACTOR LIGHT OUT (Im) EFFICACY(lm/W) HARMONICS (%) 2nd 3rd 41h 5th 7th 9th THD
101.3 0.85 1.OO 1740 17.2

lOOW DIODE 85.5 1.03 0.70 780 9.2

(2) lOOW (2) DIODE 175 1.46 1.oo 1584 9.7

1oow TRlAC 88.5 0.81 0.92 1290 14.5

lo

THD 115%-

10

20

30

40

50

0.1 2.9 0.1 2.0 1.1 0.4 3.7

39.5 3.9 10.5 1.5 1.2 0.5 41.0

0.1 2.5 0.1 1.8 1.4 0.6 3.4

1.8 22.2 1.8 14.2 9.1 5.6 29.0

CFL LightingLoad (96)

Fig. 1. Voltage distortion versus percent CFL load.

Two 100 watt lamps with diodes in reverse polarity

TABLE Il COMPAC~ FLOURESCENT LAMPS

miz
15W z w o POWER0 9 5 9.3 19.7 2 . . 03 CURREIWA) 0.17 0 1 0 4 0 3 .6 .0 .3 P O W E R FACTOR 0 4 0.50 0 4 051 .7 .1 LIGHTb) 380 320 560 1040 RATEDLOam) 400 4 0 0 9 0 0 1200 BALLAST FAmOR 0 9 0 8 0.62 0.87 .5 .0 EFF.(LMIw) 40 3 4 2 8 5 1
7w

MGHFREOUENCY
2ow 2. 03 0.34 05 .0 1030
1200

7W

1w 1 10.7 01 .3 0.67 610

MO

15W 1. 43 01 .8 06 .5 930
900

18W 17.3 03 .1 0.47 1260 1100

mw 2ow
1. 78 02 .3 06 .4 1040 1200 2. 33 032 06 .1 1370 1200

08 .6 5 1 15.3 26 . 22 . 09 . 1. 57

1 0 1 0 1 1 0.87 1 1 .2 .3 .5 .4 57 65 7 3 5 9 5 9
7. 94 48.5 1. 84 1. 35
100

HARMONICS(96)
3rd

5h t 7h t 9th

THD

81 . 23 . 11 . 06 . 81 .

86 . 2.6 13 . 08 . 91 .

1. 78 19 . 13 . 02 . 1. 80

1. 24 23 . 2.2 1.1 1. 25

8. 13 5. 08 2. 06 16.2
106

95.7 86.3 7. 26 5. 81 162

61.5 35.1 4. 08 3. 51
98

79.1 4. 89 30.5 32.1 114

Fig. 2. Typical voltage and current waveform for an IBM computer power supply unit. The sine wave is the input line voltage waveform, and the spiked wave is the input current. This produces extremely high harmonic distortion, and the measurement will normally measure a power factor of 0.4-0.6.The types of loses associated with this type of distortioncan be significantly greater than the losses associated with the 0.5 power factor.

TABLE III 100-KVA BUILDING SIMULATION LOAD

10 TOTAL HARMONICS (%) TOTAL POWER FACTOR % CFL LOAD VOLTAGE DISTORTION (Oh)
1.7

ROFlT WITH W P A C T F U LO 25 50
4.5 9.8

100
23.5

0.84
1.5

0.82 4.1

0.78 9.4

0.65
26.3

0.3

0.7

1.5

4.4

25%, the building power -..ctor is re-xed by a,mt 5%. When all of the CFL's are installed (26.3% of the building load), the voltage distortion at the service entrance is still within the IEEE-5 19 recommended limits (5%). Fig. 1 is a plot showing the impact of a harmonic source of 115 and 55% on voltage distortion as a function of its relative load. For a load with a THD just above loo%, the 5% voltage distortion limit is exceeded when its relative load to the system is about 30%.

power factor. There are passive methods and active methods. Simple power factor correction can be accomplished using a capacitor to shift phase angle until the line current and voltage are in phase, i.e., power factor= 1.0. The electronic correction of power factor andor suppression of harmonic distortion is more complex. The front end of electronic fluorescent ballasts convert the alternating line current to dc. This is accomplished using a full-wave rectifier bridge followed by a filter capacitor. The bridge conducts during the period of the cycle when the line current exceeds the capacitor voltage. The effect is that current is drawn from the power source over a very short time, as shown in Fig. 2. The line current is no longer sinusoidal, which results in the generation of harmonics. The amount of harmonic distortion is generally expressed as a percentage of the fundamental 60-Hz current. European Standards (IEC 555-2) express the amount of harmonic distortion in absolute terms as the maximum permissible current allowed at each harmonic frequency. Conventional Magnetic Ballasts Most CFL's operated at 60 Hz employ uncorrected magnetic ballasts and suffer from a poor lagging power factor as shown in Table I. In this case, a capacitor is added to the circuit to correct the power factor. A more unique circuit 121 developed for the Navy's 20-W F20 T-12 lamp is shown in Fig. 3. This circuit not only corrects the power factor but reduced

Iv. METHOD OF
Power Factor

CORRECTION

Power factor correction and suppression of harmonic distortion go hand in hand. There are two methods for correcting

672

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 29, NO. 3, MAYIJUNE 1993

Cl
II

INDUCTOR

Fig. 5. Passive power factor correction ll. Fig. 3. Power factor correction for magnetic ballast.

U I

Fig. 4. Passive power factor correction I.

Fig. 6. Passive power factor correction III.

the third harmonic to less than 3% for the 20-W F20 T-12 fluorescent lamp. This circuit could also be applied to the CFLs. Although the slight extra size might limit applications for the CFL, with adaptors that fit into the Edison sockets, there are many suitable hard-wired CFL applications where the additional size poses no restraints.
Fig. 7. Passive power factor correction IV.

Electronic Ballasts With electronic ballasts, there are more methods of correcting power factor and suppressing harmonic distortion. The obvious method of solving the harmonic problem is the elimination of the filter capacitor that causes the problem. Unfortunately, without the filter capacitor, the unfiltered dc is 100% modulated at 120 Hz. This modulated wave is undesirable for fluorescent systems, reducing system efficacy and lamp life while increasing the lamps flicker. The latter will be the source of stroboscopic and subliminal flicker effects. At present, most CFLs operated at high frequency employ no filtering, as evidenced by their large harmonic content. Passive Correction Methods Five methods of passive correction are described. Fig. 4 shows a method using a tuned series LC network before the input bridge (L1, Cl). The value of L1 and C1 is calculated to be between 2.5 to 3.0 times the line frequency. C, is the filter capacitor and can provide nearly ripple-free output, depending on the size of the capacitor. This circuit corrects the power factor and provides limited harmonic distortion improvements to meet the current U.S. recommendations but not the current European standards. Placing the inductor before the diode bridge provides the power supply with considerable protection against line spikes and transients. Fig. 5 shows a parallel-tuned resonant circuit similar to the 1 circuit shown in Fig. 4. L1, C are tuned to the third harmonic, and this circuit provides good power factor correction and limited harmonic distortion suppression. The main advantage of this circuit is that of cost and size and the attenuation of the third harmonic. The major disadvantages include virtually no attenuation of the higher harmonics and a slightly more challenging task of building L1 with acceptable levels of audible noise.

Ic-3

Fig. 8. Passive power factor correction V.

Fig. 6 is a circuit similar to the circuit shown in Fig. 4 with C1 placed after the bridge, which allows a smaller and cheaper magnetically biased inductor. This circuit provides good power factor correction but limited improvement in harmonic distortion. Fig. 7 is a method of correction sometimes known as the valley fill circuit. It has adequate power factor correction and moderate amounts of harmonic distortion suppression. Capacitors C1, Cz are charged in series to the peak line voltage such that each will have one half the peak line voltage across it. When the line voltage drops below the 50% point, the capacitors discharge, filling the valley and preventing the voltage from going to zero. The load must be able to tolerate or compensate for a ripple of about 50%. The cost of this circuit is reasonable, and the harmonic distortion levels are reduced by about 75%. Further improvements in the harmonic distortion are possible by putting an inductor in series with Dz. addition, a pair of input inductors or a balum should In be used for transient protection and RFI suppression. Fig. 8 shows another variation of the previous correction methods that allows for a substantial reduction in harmonic energy. The method is also resonant at the third harmonic, but here, the extra harmonic energy flows into the load, and it is possible to use smaller components to accomplish the same end result. This circuit has the potential of meeting most European standards at reasonable cost.
L1,

VERDERBER

et al.: HARMONICS FROM COMPACT FLUORESCENT LAMPS

673

of life. Their effects on power quality has been described in [4]. However, the results show that if these lamps are made with diodes of opposite polarity in a random manner, the poor power quality aspect will be greatly reduced.
I

I l l
TO CONTROL L O G I C

CFL In most commercial buildings where incandescent lamps are a small portion of the lighting load power, quality problems are unlikely if retrofitted with CFLs. In the above example of a building, the voltage distortion at the service entrance met the IEEE-5 19 voltage distortion when all the incandescent lamps were replaced. Another concern is the high harmonic triplens for the CFL since they will increase the neutral currents in three-phase electrical distribution systems. Even if both the transformer and neutral wire are properly sized to carry the additional load, the 12R line and transformer losses will be increased. Care must be exercised to assure that these circuits do not become unbalanced by placing all the CFL retrofits on a single three-phase branch, further increasing the neutral currents. Today, there are no power factor or harmonic standards for CF1 systems. In fact, there are no harmonic standards for any gas discharge lamp system. The IEEE-519 Standard is only concerned with the harmonics for a complete electrical distribution system and not any particular component. There is a consensus by the ANSI Fluorescent Lamp and Ballast Committee for the 4- and 8-ft fluorescent ballasts that recommends limiting THD to 32% and the third harmonic to 27%. Some utilities have offered rebates only for electronic ballasts that limit their THD to 20% or less. The electronic ballast manufacturers have responded to this challenge by introducing a low-harmonic version at a slight increase in cost. In the future, a similar approach may be taken by the utilities for CFLs or they may consider providing higher rebates for CFLs with lower harmonics as an incentive to the end user for the added cost of this product. Cost and Other Considerations Table IV describes the relative cost and merits of implementing each of the corrective scenarios previously described. Although there can be wide variations in cost between different manufacturers, the authors (WRA) experience has been that the simple power factor correction method can be implemented for a direct labor and material cost of between $0.50 and $0.75 in 1991 dollars. The active schemes can be implemented for between $2.00 and $2.50. The reader should bear in mind that the specific costs are highly dependent on purchasing volumes, degree of factory automation, and learning curve experience. VI. SUMMARY The measured performance of CFLs and some rectified incandescent lamps will cause considerable harmonic distortion of the input power. This includes both the electronically and magnetically ballasted CFLs with poor power factors and/or high harmonics. The model of a buildings electrical load indicate that there is little cause for concern for power quality problems when the CFLs are less than 25% of the buildings

Fig. 9. Active power factor correction I.

~ ~ C O N T RL~OL~ I C

Fig. 10. Active power factor correction II.

Active Correction Methods For the higher wattage CFLs, the size and weight of the passive corrective components increases significantly, and corrections using active devices become more economically feasible. Fig. 9 shows a circuit of correcting power factor with active devices. Although the circuit looks like a conventional switching regulator, the novelty lies in the fact that C1 is very small, and the line frequency ripple is filtered by a large capacitor C,. The output voltage is controlled by the duty cycle of the switching frequency, but unlike other circuits, the duty cycle is held constant over each half cycle and changes only when the input ripple voltage is held to zero. Fig. 10 shows the same implementatiop of a fly-back or boost method of correcting power factor, ,which is being used by many power supply companies whysell in the European market. Again, the output capacitof/is large and the input capacitor small. For further information on active power factor correction, see U.S. patents 4 737 636, 4 277 728 and U.K. patents 2 024 544 and 2 124 045. Since these active circuits for power factor correction and harmonic distortion suppression can add signific$t cost, they most likely are not used for the low-wattage CFLs until they are imposed to regulations. V. DISCUSSION Incandescent Lamps The semiconductor devices in series with incandescent lamps are used in two ways. There are retrofits (diode button types) that are placed in sockets or incorporated in a standard incandescent lamp in which the life of the lamp is greatly extended. As shown in Table I, the efficacy is also greatly reduced and becomes economic effective only if the labor to change a lamp is high [3]. Some special incandescent lamps are designed to be operated with a diode in series. The diode reduces the 120-V input voltage to about 85 V. This allows smaller size, lower resistant filaments that operate at a higher current (increasing filament temperature), permitting better optical control and higher efficacies with little sacrifice

674

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 29, NO. 3, MAYNUNE 1993

TABLE Tv COSTS AND

CIRCUIT
Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10
SIZES OF CORRECTIVE

CIRCUITS
COMPLEXITY
Minor Minor Minor Medium Minor Major Major

COST
Lowest

SIZE
Largest Large Large

Low
Low Moderate Low Highest Highest

Large Large
Small Small

Oliver C. Morse received the B.E. degree from Yale University in 1955, the M.S. degree from the University of New Hampshire in 1962, and the Ph.D. degree from the University of California, Berkeley, in 1970, all in electrical engineering. After graduating from Yale, he designed temperature control systems at MIT. He then worked on missile acoustics at White Sands Missile Range in New Mexico. Later, he was Deputy Scientific Leader at the South Pole Station in Antarctica, where he studied the ionosphere and VLF radiation. In 1973, he joined the staff at the University of California as a plant engineer. He presently works on advanced lighting systems research at Lawrence Berkeley Laboratory.

total load. Several reasonable low-cost passive circuits that can improve the power factor as well as suppress the harmonic distortion have been described. The use of active circuits will be more effective but more costly. The active methods are best considered for the larger wattage CFLs (above 30 W) and for special applications where minimum harmonics are required.
REFERENCES

[I] M. J. Siminovitch, F. M. Rubinstein, and R. E. Whiteman, Thermally efficient compact fluorescent lamps, in Proc. IEEE-IAS Ann. Con$. (Seattle, WA), Oct. 7-12, 1990. [2] L. J. Perper, Lamp ballast with near unity power factor and low harmonic content, US. Patent 4 538 094, Aug. 27, 1985. [3] R. R. Verderber and 0. Morse, Cost effectiveness of long life incandescent lamps and energy buttons, ECM, vol. 80, no. 55, 1981. [4] T. E. Anderson, N. W. Miller and A. J. Henderson, Effects of incandescent lamps with series diodes on power distribution systems, J. IES, vol. 17, no. 3, Winter 1988.

Rudolph R. Verderber (F93) received the B.S. degree from Syracuse University, the M.S. degree from Florida State University, and the Ph.D. degree from the University of Lancaster, all in physics. He has been a staff scientist at Lawrence Berkeley Laboratory since 1977, where he has been assigned to the Lighting Systems Research Group. His research involves the development, demonstration and techtremester of advanced lighting equipment and concepts. Specific projects have included electronic ballasts, compact fluorescent lamps, lighting management systems, and the thermal performance of fixtures. Dr. Verderber is a fellow in the Illuminating Engineering Society. He was the Golden Gate Section President of the Energy Management Committee in 1990. He is also the Papers Chairman for the IEEE Industry Applications Societys Production and Application of Light Committee. He is with the Professional Advisory Group and the Advanced Lighting Professional Advisory Group for the California Energy Commission.

William R. Alling was employed as a Design Engineer at ACF Industries, which later became part of Singer Simulation Systems. At Singer, he shared responsibility for the design, development, and field service functions for a variety of aircraft and fire-control simulation projects. At Singer, he later advanced to become the National Sales and Marketing Director for Singer Traffic Controls, which applied early microcomputer technology to the control of motor vehicles. He then was employed as Director of New Product Development and head of LearAvia, which manufactures aviation related electronics equipment. He has been active in the field of electronic ballasts and lighting control systems for the past 18 years. He is a co-founder of Luminoptics, which is one of the pioneering electronic ballast companies, where he was the principle shareholder and Executive Vice President and was responsible for the definition, development, and production of all solid-state systems for commercial and industrial lighting applications. In 1980, he was elected President of the Corporation and, in late 1981, sold the Company to Universal Manufacturing Corporation, which is one of the worlds largest merchant suppliers of ballasts for lighting, where he remained as a Vice President for approximately two years. During his tenure at Luminoptics, he developed many of the concepts of using controlable output electronic ballasts and computerized lighting control systems in a commercial environment. He is an active participant in the ANSI standards making organization and served as President of the Diablo Section of the Illuminating Engineering Society. Currently, he is President of LMP Corporation, Diablo Scientific Laboratories Ltd. (DSL) and Stevens Management Corporation, which develops highefficiency power supplies with good power quality characteristics including very high power factor with low levels of distortion.