Lighting Res. Techno!' 22(2) 103-109 (1990) Printed in Great Brit'lin

Summary The temporal modulation of light from halophosphate, triphosphor and multibandfluorescent lamps (controlled by a conventional choke circuit) was measured as a function ofwavelength. Within each category, all lamps had similar functions for peak-peak modulation. Atthe short-wavelength end of the visible spectrum all lamps showed a modul.ation near 100%.Halophosphate and multiband lamps had a low modulation at the long-wavelength end of thespectrum and gave the lowest overall modulation. Certain deluxe lamps had a modulation greaterthan 80% throughout the spectrum. The modulation of photopic energy, and energy transducedby the photoreceptors was calculated. Triphosphor lamps gave greater modulation thanhalophosphate, the lowest modulation being from warrn.:white halophosphate lamps.

Modulation of light from fluorescent lamps

A j Wilkins BSc DPhil and C ClarkMedical Research Council Applied Psychology Unit, IS Chaucer Road, Cambridge CB2 2EF

Received 26 October 1989, in final form 22 January 1990

1 Introduction

Fluorescent lamps usually pulsate in brightness twice witheach cycle of the AC electricity supply (for example, inEurope at a frequency of 100 Hz). The rapid pulsations arenot perceived as flicker but they are nevertheless resolvedby subcortical visual structures in man(l,Z) and animals(3l.Lamps can now be controlled by high-frequency circuitswhich remove most of the 100 Hz modulation. The incidenceof eye-strain and headaches in offices lit by the high-frequency lighting is less than half that in offices lit con-ventionally(4l.

Fluorescent lamps can be classified according to phosphorcomposition: namely, halophosphate lamps, triphosphorlamps, deluxe lamps of various kinds and the recently intro-duced multi-band lamps. The modulation of light variesfrom one type of lamp to another because of differences inthe persistence or afterglow of the phosphors.

The following study is divided into two parts. In the first,the modulation for a variety of lamps has been measured asa function of wavelength in order to facilitate the devel-opment of a lamp with more persistent phosphors (since itwill be some time before all installations are changed to high-frequency ballast). In the second part, the physiologicaleffectiveness of' modulation from halophosphate andtriphosphor lamps are compared, to assist those who mightwish to choose lamps with minimum pulsation from thosethat are currently available.

_~__-.MQQ!IJation

2.1 Method

A fluorescent lamp was mounted in a 4 or 5 ft (1.2 m or1.5 m) batten and controlled by conventional 240 V 50 Hzrated lagging circuit using a 4.uF power factor correctingcapacjtor operated from the domestic 240 volt AC supply. Apipe with a matt black inner surface was used to direct lightfrom a circular section of the surface of the lamp to theentrance slit of a monochromator. The pipe had an innerdiameter of 17.5 mm and its open end was covered by anultraviolet filter (to eliminate any possibility of second-orderdiffraction) and placed against the glass surface of the hori-zontally mounted lamp, at a point equidistant from the ends.The length of the black tube was sufficient to restrict the

entry of the light to an angle less than that appropriate forthe aperture of the monochromator (fl3.5), so that nearlyall the incoming light was captured by the mirror surfaces.An estimate of the contribution from scattered light wasobtained as follows. The luminance of the inner surface ofthe matt black pipe was measured with a spot photometerat an oblique angle similar to that at which scattered lightwhich had been reflected from the inner surface of the tubecould enter the monochromator slit. The luminance was lessthan 8% of that of the surface of the lamp. At this obliqueangle much ofthe reflection was specular: and partly becauseof the specular component the variation in reflectance withwavelength was less than 10% between 400 and 700 nm. Thescattered light was further attenuated wi~hin the mono-chromator. The fact that it had a negligible contribution inbroadening the bandwidth of the monochromator can bejudged from the measurements of bandwidth obtained; seebelow. It was therefore unnecessary to image the source ontothe monochromalOr slit. A silicon photodiode (RS 303-674)was mounted against the exit slit of the monochromatorwithout any interposed optics; The efficiency of the mono-chromator (Jobin-Yvon HlO DV) was greater at the short-wavelength end of the visible spectrum where the sensitivityof the photodiode was lowest. The current from the photo-diode was amplified by a current-voltage converter with again of lO volts per microamp. The output voltage wasfiltered by an amialiasing filter and sampled ·at a frequencyof 4.3 kHz by a 12-bit analogue-digital converter. The fre-quency response of the photodiode-amplifier assembly wassuch that the OUtput was within 3 dB of DC levels up to afrequency of 500 Hz and above 2 kHz was attenuated bymore than 12 dB.

The lamp was run for 20 minutes before measurements weretaken every 5 nm benveen 380 and 720 nm. 0.5 mm entranceand exit slits on the monochromator provided a pass bandof :!::2.5 nm at halfheighl. The pass band was estimated withthe mercury lines of a tubular lamp from which the phosphorcoating had been omit tcd (kindly supplied by Thorn Light-ing). This lamp was also used to calibrate the monochromalOrscale, so that minor errors could be corrected. Additionalreadings were taken when the light output reached a localmaximum, lhal is, at the principal mercury lines (405, 436,546 and sn nm).A computer algorithm calculated the time-averaged voltageand divided the periodic waveform into cycles, each cycle

© 1990 The Chartered InstilUtio~ of Building Services Engineers ]03

A J Wilkins and C Clark

Figure l' (a) The measurements of light output as a function of time, assampled by a 12-bit analogue-digital convener at a frequency of 4 kHz,showing the light from one cycle of the SO Hz hC electricity supply. Thisrecording was taken at a position near one end of the tubular lamp so as toexaggerate the SO Hz component for the purposes of illustration. 1hecomponent can be seen as a difference in the height of the twO peaks. (b)Schematic to show the four measurements taken for each cycle of the hCsupply. The average peak voltages of 10 such cycles were obtained. Thevoltage obtained with no light was subtracted from the averages. Thepeak-peak modulation was taken as (V] - V.)i(V 3 + V.) x 100. The SO Hzcomponent was estimated as [(V, - V.) - (VI - V 2)]/[(V, - V,) +(V, - V 2)] x 100%.

beginning when the voltage of a sample first exceeded theaverage. The maximum and minimum voltages for eachcycle were then obtained. Twenty cycles were analysed inthis way and the data divided into ten pairs of successivecycles oflight output, each pair resulting from one completecycle of the 50 Hz AC electricity supply. The maxima andminima for the ten pairs of cycles were then averaged,providing four voltages, two maxima and two minima percycle of the AC supply. The difference between the largestmaximum and smallest minimum was arbitrarily used asan estimate of the peak-peak modulation. Modulation wasdefined as:

(lmax .,- lmin)/(lmax + lmin) x 100%The difference between the larger maximum and smallerminimum was subtracted from the difference between thesmaller maximum and larger minimum to provide anesti-mate of the (small) contribution from modulation at 50 Hz.Figure 1 illustrates the calculations.

(a) 2.2 Results

The peak-ta-peak modulation of halophosphate, triphos-phor, various deluxe lamps and the new multi-band lampsis shown as a function of wavelength in Figures 2-5. Onlydata for the range 400-700 DID are shown because the lightoutput outside this range was insufficient to provide reliableestimates of modulation. Between 400 and 709 DID repeatedmeasurements gave estimates of modulation that were gen-erally within ±1% of those shown.

Figure 2 shows the modulation from halophosphate lamps.Note that the curves show only the proportion of variation,and give no indication of the amount of light emitted at eachwavelength. The estimate of modulation at the mercury linesdepends pardy on the pass-band of the monochromator andwould have been larger had the pass-band been narrower.Note that the curves from the various lamps are all verysimilar, showing little modulation at the red end of thespectrum because of the long-persistence phosphors. Thewarm white lamp shows the lowest modulation at this endof the spectrum.

The triphosphor lamps shown in Figure 3 have a highermodulation at the red end of the spectrum, although themodulation is lower in the range 470-510 nm. The first·twolamps listed in the legend are filled with krypton and thethird and fourth with argon. The gas filling has little or noeffeci:, perhaps because none of the krypton lamps showedstriations when they were under examination.

Other lamps (Figure 4) can show considerable modulationthroughout the spectrum. Imerestingly, these are usuallylamps with a good colour rendering index and relatively fullspectrum_ The recently developed multi-band lamps (Figure5) are unusual in having good colour rendering and lowmodulation.

Measurements were made from the centre of the lamps andthe contribution from 50 Hz modulation was less than 3%for all·lamps and all wavelengths, with the exception of theTrue-light lamp for which it varied with wavelength from5% to 7%.

2.3 Discussio1l

The modulation curves in Figures 2, 3 and 5 are all similarwithin type, regardless of the manufacturer or the nature ofthe gas within the lamp. The differences between types are

smallermin

largermax

largermin

smallermax

(b)

time1----1

lOms

o

average

light

Halophosphate lamps

Triphosphor lamps

100 ---,

zrl kUl~Thorn Poly lux 4000

80. Philips 36W/84Phi lips 3?W/84

,~

Phlhps .10W/8~0- v~ •c 60

r\.})V~~j2en::> 40"00E t;.~l .. .

20

0

400 450 500 550 0380 G50 700

wavelenglh (nf'~)

700650

GEe warm whiteOsram whiteThorn 35 white

() Philips 33 cool white+ Thorn cool while

. ~-~~oI-

t • aM.~

500 550 600

wavelength (nm)

450O+~~,.....~~--r-"--'-~r-r...,-',--'-""",~~""",,--'-~-.--l

400

100

80

---cf!.'

c 60.2co::> 40-00E

20

Figure 2 Modulation from halophosphate lamps as a function of wave-length.

Figure 3 Modulation from triphosphor lamps. The first (WO listed in theinset contain krypton gas and third "nd fourth contain argon.

104 Lighting Research and Technology

Fluorescent lamp modulation

Miscellaneous lamps la}

100%

c(I) ca

500nm A(I)0) Eco0 0 550nm>.~ 600nm0-

0%700650600550

Dura-test True-liteThorn Kolor-riteThorn Northlight

500450

---1:J-

•a

O+-........~,..,....,..-,.-.-..-.-.-.,.-.,-,-.,-,....,.... ..........,...........-.-.......-,-.......--.--i400

20

40

~o

c.2.

A J Wilkins and C Clark

repeated until one cycle containing exactly 128 samples wasobtained. The corresponding 128 samples from the first ADCchannel provided' the voltage fluctuation from the photo-diode through one cycle of the electricity supply. ,.

Once the lamp had run for 20 minutes, measurements weremade every 5 nm between 400 and 700 nm. Various halo-phosphate and triphosphor lamps were compared, includingthose wi'th high and low correlated colour temperature.

On some lamps there were low-amplitude, high-frequencycomponents of the light output that were not strongly phase-locked to the electricity supply, and which can be seen asripple' on the waveforms in Figure 6. The data for at least5 cycles were obtained and averaged, greatly reducingthese components, and increasing signal/noise ratio at theextremes of the visible spectrum where the signal wasweak.

Initially data were obtained for halophosphate andtriphosphor lamps with similar correlated colour tem-perature (4000 K). The available narrow and wide lampversions were obtained from two major manufacturers. Thequoted characteristics of correlated colour temperature, ren-dering index and colour coordinates for these lamps (shownin Table 1) closely resemble the characteristics of either CIElamp type F2 (listed as 4230 K, 64 and 0.3721, 0.3751) ortype F1I (listed as 4000 K, 83, and 0.3805, 0.3769).

Measurements were then taken from lamps with lower colourtemperature: both halophosphate and triphosphor. Thecharacteristics of these lamps are shown in Table 1.

3.2 Results

Figure 6 shows the voltage for selected wavelengths (500,550 and 600 nm) as a function of time, each curve expressedas a percentage of the time-averaged voltage (represented bythe horizontal line). The data are for one halophosphateand one triphosphor lamp. The change in phase lag withwavelength is obvious. (The curves may be used to providea rough estimate of the (reduced) modulation that wouldoccur were the lamp run on an electricity supply with afrequency of 60 Hz, although the interactions between thephosphors may differ at this frequency in a way that wouldbe difficult to derive analytically).

Every 5 nm between 400 and 700 nm the time-averagedvoltage from the photodiode was taken as equivalent to thepower at that wavelength, as given from the spectral powerdistributionJor the appropriate CIE lamp type. From. thiswas calculated the varIation in power with time. Figure 7shows the calculated spectral power distributions for thelamps over half of one cycle of the electricity supply, begin-ning when the light from the mercury direct emission wasminimal and subsequently at intervals of 2.5 ms.

From the spectral power distributions for one completecycle of the supply was calculated the variation in photopicluminance over time and these data are shown in Figure8. Figure 8 also shows, for comparison, the variation inluminance with time measured directly using a V).-correctedphotodiode (Hagner) connected to the same data acquisitionand analysis system. The calculated and measured curveswere superimposed, although the trace from direct measure-

Table 1 Characteristics of various halophosphate and triphosphor lamps

Cl E category F2 F3 F4 FIl F12_._-_ .._- _._--_._-.•.- -_._--..._....-. .-.... __.._-_._...-.

J

...... "3 ::>0 ...... 00 0

t'0

0 2 ... 0 0"" .".

;:;:..... :2 ....., VI ....., 'T 00 ccx 3: ~ ~ ....., x 3 x ~ ~ ~::l - ::l ::l ::l §!'" "0 "0 ~ ~ ~ g ;;:, 0 >-::l 0 0 "'" ~ "0 "0 ~c: :;. c:Lamp brand u u .... ....., u u

'" '" .~ '"'n

jE 0. ~ .!?- E E 0. ~C ..c S? ~ .-~ ..c ..c ~ E:e.. f- e.. e.. :--_._.._.... --_·_---__---_0- _____._...___ .._______.._... ._... -_. ---_._-_.Rated power (W) 36 36 40 40 36 36 ·H) .>0 36 40 .;0Outer diameter (mm) 25.2 25.8 37.9 37.5 25.9 25.2 3; 25.2 25.8 37.5 25.5Correlated colourtemperature (K) 4200 4000 ~200 4000 3500 2900 2Y1.lO 4000 4000 ·WOO 2950Colour rendering index 58 ·61- 58 62 58 51 5 j ~5 85 ·S5 S5

AppearanceCool Cuol C'.)01 Cool White \'('arm \,'arm Coolt Coolt Coolt \X-'armwhile whilt" while white ",hil,' white

Quoted colour coordinatesx 0.372 0.379 0.372 0.379 0.412 0.440 0.440 0.380 0.380 O.3i

(a)

_ ......0.-__........... _

E

'Ec

~::J..

C>C>

Fluorescent lamp modulation

. ....I I I I

400 500 600 700

wavelength (nm)

(b) 'E-:-'E

ic

.,L.~::J..

C>

"=C>

• -

...J • ! ei. -

I400

Vol. 22 No. 2 (1990)

I500

wavelength

I600

(nm)

I700

Figure 7 Spectral power distri-butions computed from CIE dataand shown as a function of timeover half of one cycle of the elec-tricity.supply, beginning when lht'light from the mercury direclemission was minimal (lOp histo-gram) and subsequently at inter·vals of 2.5 ms. The horizomal axesrange from 400--700 nm in biD'5 nm wide. (a) F2 halophosphalc'lamp (Thorn Pluslux 4000); (b.Fll triphosphor lamp (ThornPolylux 4000).

107

A J Wilkins and C Clark

10ms

10ms

Figure 8 Variation in luminance wirh time as calculared from data suchas rhose in Figure 7, and as measured directly using a VA-correctedphotodiode. Ca) F2 halophosphate lamp (Thorn Pluslux 4000); Cb) Fllrriphosphor lamp (Philips colour 84). The trace wirh high-frequency rippleis rhat measured with rhe corrected phorodiode and is based on one sweepwith no averaging. The ripple was nev::r grearer rhan rhar shown in Ca) andwas usually less than that in Cb).

approximated a sine wave with no rectification. In terms ofFourier analysis, the power in a sine wave is greater thanthat in a rectified sine wave of similar peak-peak amplitude.Partly for this reason the luminance modulation for the (typeF2) halophosphate lamps had a power in the fundamentalFourier component (at 100 Hz) only.about 70% of that forthe triphosphor lamps.

The flicker indexeS) (area of curve above f the mean/areabelow) for the F2 and F11lamp data shown in Figure 7 was21.0% for the halophosphate lamp as compared with 31.2%for the triphosphor. The standard deviation of the samplevoltages expressed as a percentage of the mean (anothermeasure of variation) was 17.4% for the halophosphate lampand 24.8% for the triphosphor.

Using VA, the scotopic luminance was calculated as a func-tion of time and gave a peak-:-peak modulation of about60% for the halophosphate lamps but only 40% for thetriphosphor lamps (see Table 1).

Figure 9 shows the variation in colour coordinates with timefor two "mps. The chromaticity coordinates published bythe manufacturers are quoted in Table I. The values arebased upon the time-averaged power spectra. Figure 9 showsthe variation in colour coordinates with time for two lamps.The time-averages of these values are the same as the chroma-ticity coordinates quoted in Table 1 provided a wei~htingfor luminance is applied according to the centre of gravitylaw of colour mixture (see Hunt(6) pp 59-60).

From the x,y, z values the light absorbed by the three classesof cones was calculated using the equations described bvHunt(6) (p 147). The peak-peak modulation in the long-,medium- and short-wavelength cones is listed as R, G, andB respectively in Table 1. Hunt's mode(6; descrihes twochromatic channels, one red-green opponent and the otherblue-yellow. The modulation for these channels is also gi venin Table 1. (The achromatic channel described by the modelgave values similar to those for photopic luminance.)

Measurements and calculations were then obtained for halo-phosphate lamps with relatively low correlated colour tem-perature, for which the CIE spectral power distributions F3and F4 are appropriate. As can be seen from Table 1, themodulation is lower for the 'warm white' than for 'white'or 'cool white' halophosphate lamps because of a greatercontribution of red light from the persistent phosphor. Nosuch change with colour temperature occurred for thetriphosphor lamps. The Fl2 triphosphor lamp with a colourtemperature of 3000 K had a modulation similar to that ofthe triphosphor lamps with a colour temperature of 4000 K.

As can be seen from Table 1, lamps with large diameter hada slightly greater modulation than those with small.

3.3 Discussion

The light from the triphosphor lamps had a greater photopicluminance modulation than the light from the halophospharelamps of similar correlated colour temperature. Warm whitehalophosphate lamps had the lowest modulation.

The modulation in light energy captured by the !ong-.medium- and short-wavelength photoreceprors was greaterfor the triphosphor lamps than for the halophosphate. Ingeneral, therefore, the findings suggest that triphosphorlamps have greater physiologically effective modulation thanhalophosphate lamps. It remains to be derermined whetherthe differences between the lamps are physiologicallysignificant. For example, it is not yet known whether, in

0%

C roU

:ii Ec_'E 0:::l~

ment sometimes showed a high-frequency ripple, most pro-nounced for the lamp shown in Figure 8(a). As can beseen from Table 1, tlre-direcr measurements"af luminancemodulation (lmax - lmin)/(lmax + lmin) were well within I% ofthe calculated valucs, indicating (a) that the choice of spectralpower distributions was appropriate, and (b) that the ripplehad a negligible effect, even though the direct measurementsdiffered from those calculated from the spectral power distri-butions in that they were not averaged over several cycles ofthe supply voltage.

Not only was the peak-peak modulation in luminancegreater for the triphosphor lamps (type FIl) than for thehalophosphate (type F2), the variation in luminance withtime shown in Figure 8 had a different shape. The variationfor the' halophosphate lamps approximatcd a rectified sinewave whereas the variation for the triphosphor lamps

100%

0%

·c roU

:ii EcE 0:J ';f!. •

100%

(a)

(b)

108 Lighting Research and Technology

(a)

1

x coordinate

(bl

x coordinate ]

~jj

F luorescenl lamp modulation

installations fitted with low-frequency ballast, changing fromtriphosphor to halophosphate lamps would effect a reductionin headaches and other symptoms.

Over the last two decades, improvements in colour rendering.have been achieved at the expense of greater modulation,a trend only recently reversed. Lamps with good colourrendering have been associated with greater apparent 'visualclarity' than lamps with relatively poor folour rending andsimilar illuminance. For example, Boyce17) found that whenKolor-rite lamps illuminated an office scene they requiredan illuminance 0.7 times that for white halophosphate lampsin order to produce a match in which the lighting from bothtypes of lamps was judged equally satisfactory. Any aversiveeffects of flicker might be expected to increase with illu-minance and perhaps offset improvements in clarity, andthis might be one (of several) reasons why Kolor-rite lampswere judged equally satisfactory at lower illuminances.

It might prove possible to create a new lamp by combiningphosphors that are more persistent than those currentlyused. It is doubtful whether such a lamp would have ashigh an efficacy or as good maintenance properties as thosecurrently available. However, any consideration of econ-omics should, cif course, include the contribution toemployees' health. A 5-10% improvement in productivitydwarfs all other economic parameters.,:8

Acknowledgements

We thank Dr P R Boyce (Electricity Council ResearchCentre), L Bedocs and M Fuller (Thorn Lighting), J Procterand E Glenny (Philips Lighting) and D Alger (Jobin Yvon)for advice, R Edwards for assistance with the apparatus andL Bedocs, E Glenny, R Hunt, M Perry, P Ramby, R Wealeand two referees for their comments on an earlier version ofthe manuscript. The tubular lamp used for calibration waskindly supplied by Thorn Lighting. Thorn Lighting, PhilipsLighting and Duro-Test International generously providedlamps for measurement. We thank the Medical ResearchCouncil and the Health and Safety Executive for support.

References

figure 9 Variation in colour coordinates with time. (,a) F2 halophosphatelamp (Thorn PlusJux 40001; Cb) FJ I triphosphor lamp (Thorn Polvlux4000'.

y coordinate

10ms

lI

-~1-<

iI

Brindley G S Beats produced by simultaneous stimulatioll of ,hehuman eye with inrerminem lighi and interminem or alternating elec-tric .:urrem]. I'hysiol. ,London} 164(1) J57-167 (1962)

2 Greenhouse D S, Berman S Nl, B3iley I Land Raasch T \V :1bs/rlwsof Annuol Meeting Association R,sc