RESEARCH DEPARTMENT

A DETERMINATION OF SUBJECTIVE WHITE

UNDER FOUR CONDITIONS OF ADAPTATION

Report No. T.OII-5

Seri al No. 19511-/8

W. N. Sp r050n, M. A. --(w. Proctor Wi 1 50n)

Section

1

2

3

4

5

6

Re~ort No. T.045

A DETERMINATION OY SUBJECTIVE WHITE

UNDER YOUR CONDITIONS OY ADAPTATION

Title

SUMMARY •

INTRODUCTION

METHODS.

2.1 Method I: Using- Tricolorimeter •

2.1.1 2.1.2

Apparatus and Calibration Experimental Procedure

2.2 Method 11 • • • •

2.2.1

2.2.2

Apparatus and Calibration

Experimental Method •

DIRECT COMPARISON OF METHODS I.AND II •

COMPARISON WITH OTHER WORKERS ••

CONCLUSIONS • . .'. . . RECOMMENDATION

APPENDIX

REFERENCES

Page

1

1

2

2

2 3

6

6

7

13

13

14

14

15

16

, CONFIDENTIAL

Report No. T.045

February 1954 Serial No. 1954/8

A DETERMINATION OF SUBJECTIVE WHITE

UNDER FOUR CONDITIONS OF ADAPTATION

SUMMARY

Subjective white has been determined for twelve colour-normal observers under four conditions of adaptation. Two methods were used and a comparison is made between the results. Further comparisons are made between the results of the present investigation and those of other workers. The report concludes with a recommendation for subjective white for television viewing with a tungsten adapting stimulus.

1. INTRODUCTION.

The purpose of the experiments described in this report was to find the area on the chromaticity diagram which is accepted as white by the statistical average of colour-normal observers, and to determine how this is affected by the type of adaptation, including the condition without surround adaptation. The white point is of considerable importance in a colour television system, since it deter­mines the colour balance of the picture. Further, it is essential that e~ual voltages into the grid (or grids) of the cathode ray tube corresponding to the red, green, and blue signals shall produce a satisfactory white. If this is not so, then the brightness control will produce changes of hue as well as changes of brightness and saturation.

A specification (1,2) already exists for signal white and it might be considered that there is no need for a separate investigation. However, the con­ditions under which the signal white specification holds may well be somewhat different from those which hold for viewing a colour television picture. More specifically, the angular size of field, the brightness level (luminance) and the surround adaptation are likely to be sufficiently different to justify a separate investigation. Further, the re~uirement for a signal colour is certainty of recognition in comparison with a few highly saturated colours such as blue, green, yellow, and red. Slight tendencies towards red, green or blue, which would not affect the rec~nition of a signal colour, might be rather undesirable in the white of a colour" television picture.

2

The American literature on colour television makes frequent reference to white, but usually avoids defining it. However, a recent paper (3) on the C.B.B. system does give some white points which have been used in colour television, and specific comments about these will be made later.

2. METHODS.

Two methods were used; in the first the observers were given control of the knobs on an apparatus for producing a wide range of colours and asked to make white; in the second method the observer was asked to name a succession of colours which were in the neighbourhood of the white point. The two methods do not corn­p~etely agree in their final results but this fact is not regarded as being without interest - particularly as it has a bearing on the much debated point in subjective experiments, whether it is a proper procedure to give the observer control of the knobs.

2.1 Method I: Using Tricolorimeter.

2.1.1 Apparatus and Calibration.

The app~ratus consists of a tricolorimeter and subsidiary projector, Fig. 1. The tricolorimeter itself houses three projectors, one with a red filter, a second with a green filter, and the third with a blue filter, in their respective light paths. These are arranged to illuminate the sa.me circular portion of a sheet of opal glass which serves as a back-projection screen. The subsidiary projector is used to provide a surround field of either tungsten or daylight type of illumination at a brightness of about one tenth of that of the central field, but it supplies no light to the central field itself. The area of the surround is approximately three times that of the central field, a figure recommended by Hopkinson, Mackenzie, and Nixon (4) as being the minimum for the surround field to have adequate effect. The purpose of the adapting fields is to simulate the three main conditions under which a colour television picture may be viewed - namely, diffused daylight, diffused tungsten light or in an otherwise completely dark room. A fourth condition will be added in Method II.

The colour produced by the tricolorimeter is controlled by the individual voltages of the three lamps - these being adjusted by three 'Variac' transformers at the front of the tricolorimeter. This method of adjusting the luminance of the indiv­idual contributions has certain disadvantages in calibrating the instrument but was adopted after more elegant arrangements had, for various reasons, been rejected. As far as the observers are concerned, there is no disadvantage in controlling the volt ages of the lamps as opposed to other methods of attenuating the light output.

The method of calibration was as follows: the colour filters were removed from the front of the projectors and for each projector data were taken with a photo­electric tristimulus colorimeter (5) of the relation between the voltage on the lamp

r

2/~ I

FIG.I ADDITIVE TRICOLORIMETER

GENERAL LAYOUT - METHOD I

t- "-1 I

CE.NTRAL FIELD

SURROUND FIELD.

9"

...-.------.-------,-_----r--------r-_I 0·46 ,~

PHOTOELECTRIC 0·4SI-----------I-----------+-----------+--------...,----+-- TRISTIMULUS---1

FIG.2 CHECK ON CALIBRATION OF ADDITIVE TRICOLORIMETER

PORTION OF C.I£. CHART SHOWING BLACK- BODY LOCUS & DATA ON MATCHINGS.

OBSERVER WNS

--tooBG

COLORIMETER MEASUREMENTS.

0·40~------------------~-----------------+-----------------+_-----~'-~3~O~OO~·~k+_----~ ________ ~ IMAltH

3500' k

4000'K

O·35~------------------~--------------~~+-----------------+-------------------~--------------~

O·30~--------------~--+-------------------4---------------4--------------~--------1

O·29~ __________________ ~ _________________ ~ ____________ ~ ________________ ~ ____________ ~

0·25 0·30 0·35 0'40 0-45 0·49 COEFFICIENT OF X

3

on the one hand and the chromaticity and relative brightness of the opal screen on the other. As expected, the chromaticities were very near to the black-body locus, and the variations in colour temperature from one lamp to another for a given voltage were only slight. The light outputs varied rather more and had to be treated individually.

The colour filters were measured on a Unicam spectrophotometer and their tristimulus values and luminance factors were computed for colour temperatures of 2,7000 K, 2,5000 K, and 2,2500 K. This range of colour temperatures was sufficient for the present purposes, and with the da~a obtained from the photoelectric tristimulus colorimeter, curves of tristimulus values, X, Y and Z against voltage were plotted for each colour projector. Although rated at 115 volts, the lamps were used at a voltage not exceeding 100 in order to increase the stability of performance and permanence of the calibration.

As a check on the accuracy of this slightly involved calibration, two colour­normal observers each produced a match with a white surround illuminated by the sub­sidiary projector, the chromaticity of which was measured with the photoelectric tristimulus colorimeter. The accuracy of the tristimulus colorimeter in the neighbourhood of its own calibrating point 2,700o K is such that the chromaticity of the surround illumination was known to better than ±O~OO2 in both x and y values. The results also included intentional mismatches in the sense of producing a white that was slightly but definitely too blue, too yellow, too red, etc., and these points are all shown in Fig. 2. It will be noticed that the two observers produce different matches but that the true (i.e. objectively measured) chromaticity lies between the matches given by the two observers. This was considered to show that the curves of X, Y,and Z against voltage were substantially correct.

2.1.2 Experimental Procedure.

After the tricolorimeter had been calibrated, the measurements proper were started. The procedure adopted with each observer was as follows: with the surround illumination switched on, the blue component of the central field was approximately adjusted to a certain arbitrary value (actually that corresponding to about 80 volts across the blue projector lamp) and the observer was then asked to produce what he considered to be white by adjusting the red and green controls. When the observer was satisfied, the voltage readings of the three lamps were noted down by the person conducting the experiment, who then turned down the red and blue controls to zero, leaving the green control where the observer had just set it. He was asked to repeat his production of white - this time using the red and blue controls only. Once more the voltages were noted when the observer was satisfied with his result. Finally, the green and blue controls were turned to zero and a third setting asked for, this

4'

time with the red control fixed. In this way three settings were made for one state of adaptation - using two of the controls at a time.

The type of adaptation was then changed (i.e. from tungsten to daylight or vice versa) and three more settings of white were made. Finally, the surround illumination was switched off, and, in an otherwise dark room, the observer made three further settings of white.

The chromaticities corresponding to the nine settings were calculated from the voltage readings using the calibration curves, and the results were recorded on the standard C.I.E. chart. Table I and Fig. 3 show the detailed results for one observer and serve to illustrate the method described above.

Twelve colour-normal observers were used (in addition to a number of persons with slightly or markedly abnormal colour vision whose results are not included in this report). The results for tungsten type adaptation are shown in Fig. 4, while Fig. 5 gives the results for daylight adaptation and Fig. 6 shows the effect with no adapting stimulus. Fig. 7(a) combines all three conditions of adapting stimulus and also shows the mean chromaticity values. Fig. 7(b) shows the ellipses of variation* as between successive settings for all observers, after ignoring the differences in mean chromaticity values from one observer to another.

Some general conclusions may be stated from Figs. 7(a) and (b):

(1) Observers show greater sensitivity to changes in the green~magenta direction than along the black-body locus.

(2) The type of adaptation does have a definite influence on the sort of white selected; thus the whites selected with tungsten adaptation lie in a region corresponding to lower colour temperature. The results with daylight and dark adaptation are very similar ~ the mean chromaticity points being relatively clbse.

(3) The variation ellipses (showing the 50 per cent. contour) also show the feature mentioned in (1ro They maintain an approximate parallelism with the black~body locus. The ellipse for the condition of no surround adaptation is noticeably more nearly circular. We may note in passing that these ellipses have a different direction from those reported by MacAdam (6). Although the conditions of our experiment are somewhat different from MacAdam's, it is slightly surprising to find that his ellipses intersect the black~body locus at a considerable angle.

One feature of the results shown in Fig. 7 gave cause for considerable doubt. The departure from the black-body locus was difficult ~ if not impossible - to under~ stand. The adapting stimuli were slightly greenish, owing to the heat~absorbing glass used in the subsidiary projector, and that might account for the tungsten and daylight adaptation results showing a bias to the green side of the black-body locus. However,

* See Appendix for method of calculation.

0·4 6

0·45

FIG.3 ONE OBSERVE~S RESULTS

TUNGSTEN .-----~ X ADAPTING

FOR THREE STATES OF ADAPTATION. --- ~ .....-- A 2S00'K --iY /"....... . /

~ 27000K

.,./ / /'

/" // CO'k /' 2/ /

DAYLIGHT {- / ~500'K ADAPTING !"'X-\ // STIMULUS ~ -- \ 3 / --- ,!,..." f- - ...... ,_ ---S 4000'K

/ . ) /'

/ /'

\1 / /

'-./

0'35 / 4800'1(

7 • THE 9 SETTINGS ARE SHOWN:

8 1,2,~. WITH TUNGSTEN ADAPTATION. • 4,5,6. WITH DAYLIGHT ADAPTATION.

6OOO'k 7.8,9, WITH NO ADAPTING STIMULUS.

9 6500'K •

7fXXJ' K

0·30 /-'1(

0'29 . 0·25 0·30 0·35 0'40 0-45 0·"

COEFFICIENT OF X

6

0·4 S

0·40

>­..... e

0'35

0·30

0'29 025

-

FIG.4 TUNGSTEN ADAPTATION.

MEAN SETTINGS OF TWELVE OBSERVERS EACH POINT REPRESENTS THE MEAN OF THREE SETTINGS.

• CBBW •

HASP

• • GCW

RFV

• .JG ROAM 4000 0 K

• DC

• • ER SJl .. // PHCl

, /./ 4800°1( ,/ - I ;:r:.

6000°1(

6500°1(

7('fJ(J° I(

/-Ok .---------- f-,.,. .',.,---,----,

-----0·30 0·35

COEFFICIENT OF X

,.. ---"- ,-

-----1-------,-,-,------

X ADAPTING STIMULUS

• ~ WNS

~: 2700'K

MG

~O'k 3500°1(

0·40 0·45 0·41

ro~ 6

045

0·40

0·35

0·30

0·29 0·2S

FIG.5 DAYLIGHT ADAPTATION

MEAN SETTINGS OF TWELVE OBSERVERS. EACH POINT REPRESENTS THE MEAN OF THREE SETTINGS.

• • Cr-BW MG

• WNS • XADAPTIHG RFV STIMULUS

• .04000°1( PH Cl JG

• GCW • HASP

ROAM

• • / 04aOO°1( DC STl

• ER

60000 k

6500°1(

7r:J:IJ° K

fooo'k 0·30 0·35

COEFFICIENT OF X

"

~ ~ 2700' I(

CO'k 3500° I(

0-040 0-45 0·.

0·46

0· ... 5

FIG.6 NO ADAPTATION.

MEAN SETTINGS OF ELEVEN OBSERVERS EACH POINT REPRESENTS THE MEAN OF THREE SETTINGS

0'40

>­.... o

!Z ... !d .... .... ... o u

0':55

• DC

0·3 0

0'2 9

PHCL

• ER

0'25

• ROAM

• SJL

• GCW

I ,._-----" ...

• • CaBW HASP

<I;

• RFV ~/

,-4800°1( ./

,/ .'

6000°1(

6500°,t<

7000' K

/-01( .

0-30 0·35 COEFFICIENT OF X

-----l------:::::- 2500° I(

.-----------2700'1(

~----3000' I(

/~50·1( .. //

MG " /////-/

V"'--:5500°1(

K .JG

040 045 O·

________ -------r------.-----r-------,---, r 046 "

0·45~----------------~------------------+_----------------_+------------------+_------------~

FIG 7(a) AREAS ON CHROMATICITY DIAGRAM COVERED BY SETTINGS OF TWELVE OBSERVERS FOR

THREE CONDITIONS OF ADAPTATION.

MEAN CHROMATICITY FOR TUNGSTEN ADAPTATION

---- -- ./ '\//

TUNGSTEN X AOAPTION

STIMULUS

0-40~------------~--_+----~~=-------~~----~----~~--~--------~~--~~~------------~

MEAN CHROMATICITY FOR DAYLIGHT

>- ADAPTATION L>. o I-Z !!! v ~ \L &AI

8

MEAN CHROMATICITY FOR NO ADAPTATION

60000K

6.500'1C

7000'"

r- NO ADAPTATION .... ,'-_ .. DAYLlGHT ADAPTAnON -'-,TUNGSTEN ADAPTATION

O'30~------------~~-r------------------~----------------~------------------~------------~

O·29L-________________ ~ __________________ ~ ________________ ~ __________________ ~ ____________ __J

0·2.5 0-30 0·3.5 0·40 0·45 0-49 COEFFICIENT OF X

O·46r-------------------.--------------------,r-------------------~-------------------.---------------,

O·4S~-------------------r------------------~~------------------4_------------------_+--------------_4

FIG.7(b) ELLIPSES OF VARIATION BASED ON

CONS1STENCY OF SETTINGS.

.----x 0-40r--------------------r------------------~~~~----~~~~--+_--------~~--~~~--------------~

>­lA. o t­:z ~ ~ lA. lA. UJ

8

O·~sr_--------------~~t-~~~--------~~t_------------------t_------------------t_------------~

60000 K

__ NO ADAPTATION ,'-_, DAYLIGHT ADAPTATION ~-'-.TUNGSTEN ADAPTATION

THE ELLIPSES SHOWN INCLUDE SO °/0 OF THE SETTINGS.

0-30r---------------~--~------------------~~------------------~------------------~--------------~

l 0-29

0'2S 0·30 COEFFIClEHT OF X

0·40 0·45 0-49

5

TABLE I

TYPICAL RESUL'rS lI'OR ONE OBSERVER

Obser- Volt-ages (Mean Values) ver on X Y Z Totals x y z x y I!I GCW Lamps

63'6 4'65 2'03 -87'2 1'30 4'16 0'30

*81'7 0'30 0'28 2'87 ~ --- --

Q 6'25 6'47 3°17 15'89 0'393 0'407 0"199 0 -- -- --

.r< 61'0 3'98 1°72 -"!is *87'0 1'28 4'12 0'30 +' 86'8 0"88 0'38 3'80 ~ -- -- --re!

5'64 6'22 4'10 15'96 0°353 0'300 0'257 0"370 0'392 0'237 -< -- -- --s:1 <1) *61°0 3'98 1'72 -+' 83"3 1'09 3"54 0'26 CD

J 85"5 0°36 0"36 3'55 -- -- --

8 5'43 5'62 3"81 14"86 0'365 0°378 0'256 -- -- -- -'

50'8 1'80 0'77 -74"7 0'74 2'37 0'17

*81'2 0'30 0'27 2'80 -- -- -- ...

Q 2'84 3'41 2'97 9'22 0°308 0'370 0'322 -- -- --0

.r< 50"3 1'71 0'74 -~ *74'7 0'74 2'37 0'17 +' 81'6 0'30 0"28 2'86' ~ -- -- --'d 2''75 3"39 3'OS 9'17 0'300' 0'370 0"330 >- 0"300 0";:367 0'333 ...: -- -- --~ *50'3 1'71 0'74 -"ao 75'3 0'78 2"44 0'17 -rl M 83'5 0"33 0'31 3'19 I» al -- -- --A 2"B2 3'49 3"36 9'67 0'292 0'361 0'347 -- -- -- -,

45"8 1'.10 0'41 -68'2 0'50 1'58 0'11

*81'0 0'30 0'27 2'75 - -- -- ~

1"00 2'26 2'86 7'02 0"271 0°322 0'407 -- -- --~ 46'0 1"12 0'42 -0 *68'2 0"50 1'58 0°11 'M

+' 79'0 0'25 0'25 2"46 ro +' -- -- --rc' 1',87 2"25 2'57 6'69 0'280 0'336 0"384 >- 0'281 0'334 0"385 et

"C -- --<t

*46'0 1'12 0'42 , -0 67'7 0'48 1'51 0'10 z

76'9 0°22 0'22 2'16 , -- -- --1'82 2'15 2°26 6'23 0'292 0'345 0'363 -- -- -- --

The voltages are given in the order red, green, blue each time.

* The asterisk indicates which of the three lamps was pre-set.

b

6

the results with no adapting stimulus showed the same effect. Two possible explanations were considered: (1) calibration error affecting the higher colour temperature region of the colour diagram; and (2) non-uniformity of the colorimeter field. Explanation (1) was thought unlikely but not impossible, as the calibration is not too straightforward. Explanation (2) involves a feature of the apparatus not mentioned previously. When attempting to produce a white, the observer becomes aware that parts of the periphery of the tricolorimeter field have a different colour. With most of the field near to white, these regions of the periphery are noticeably pinkish. The fact that the results show an overall green trend may well be caused by a desire to avoid this pinkish periphery. Incidentally, some effort was made to eliminate this non-uniformity, but it appeared impossible to do this without completely altering the design of the tricolorimeter.

In addition to the disturbing greenish trend of the results, many observers were producing whites corresponding to very high colour temperatures of 7,0000 to 8,0000K. This, once more, seemed doubtful, although if the observer produces a very bluish white the experimenter must accept it without query.

However, another approach to the problem was thought out as an independent check of the results obtained using the additive tricolorimeter. This will be described as Method II.

~.2 Method II.

2.2.1 Apparatus and Calibration.

The tricolorimeter is a rather inefficient method of making colours if one is only interested in producing near-whites. A more direct approach would be to use a single projector and modify its luminous flux to the required extent by filters of desaturated green or magenta colour when examining points perpendicular to the black­body locus, or by the use of colour-temperature-raising or colour-temperature-lowering filters to proceed along the black-body locus in either direction. This more direct method of making near-whites was used in the second method. Its calibration is also more direct and somewhat brighter screen illumination is possible. Thus, with an Aldis 750 projector using a 250-watt lamp, on a stabilised mains supply - a range of 25 near-whites of brightness about 10 ft.-lamberts on a flashed opal screen of effective size 9·2 in. by 8 in. was obtained. The additive tricolorimeter produced whites of only 2i-3 ft.-lamberts on a circular patch of about 6i in. diameter (of which the central 5; in. was used} with three lOO-watt lamps.

The Aldis projector uses two thicknesses of Chance ON20 heat-absorbing glass and consequently gives a definitely greenish white. The chromaticity of this light, including the effect of the flashed opal screen, was accurately measured by the photo-electric tristimulus colorimeter. A hypothetical filter characteristic shown in Fig. 8, Curve A was empirically determined, which gave very nearly the same chromaticity with Illuminant A as was directly measured for the projector. To cancel the marked green trend, a Chromex MG8 filter was found to be necessary, and this was checked both by

2·0.-----------------r-----------------.-----------------~_,

1·8~----------------r---~~+---------~----------------+-~

'C'-l-

~1·6r_----------------r_~r_----+_------+_----------------+-~ z i: ;:) ....J ....J

~ o ~1·4r_----------------r_~------~~----+_----------------+-~

Cl: o l­t.)

~

zl·2r-----------------~------------~--+_----------------+-~ o ~ 11)

i: 11) z -c

PROJECTOR + OPAL SCREEN.

'+----~~--------_4-A

~I.O~----------------r_~~~~~------t_~====;=========~~B~

PROJECTOR + OPAL SCREEN + CHROHEX HG.8

O·8r-----------------r-----------------+_----------------+-~

0·7~ __ ~ ____________ L_ ________________ ~ ________________ ~_J

400 soo 600 700 720 WAVElENGTH m}l

FIG.S HYPOTHETICAL GREEN FILTER WHICH GIVES SAME COLOUR AS THE ALDIS PROJECTOR.

o.461-----I----..:...--II-----~---------,------

O.4Sr------r------+-------+-------t-----~

~ .... o

FIG.9 GRID OF CHROMATICITY POINTS USED IN METHOD 2.

_-PROJECTOR CHROMATICITY CALCULATED FROM FIG.8A.

CHROMATICITY CALCULATED FROM FIG.BB.

0·35 r------T-----\-;~LF~~~---::::~rl-------~~----~

6500'K

7000' K 21

G MEASURED

CH ROMATICITIES

O.30r---------------L-~~~~'k~----------------1-------------------~------------------J-----------____ J

I O.2~~.2~5~---------------oO~.300-----------------rO~.3~5----------------~O~.4~O~----------------OJ.4-5--------------0J.49 COEFFICIENT OF X

7

direct measurement, and from calculations with the hypothetical filter characteristic and the measured transmission curve of the Chromex MG8 filter, The combined effect of these two curves is also shown in Fig. 8, Curve D. The agreement between the measured and calculated values is better than 0'002 when allowance* is made for a slight rise in colour temperature as between the calculated and measured values. The colour-temperature-raising filters used were qhromex 0'29 CTR, 0'55 CTR, 0'75 CTR, and 0'97 CTR. These enabled the colour temperature of the illuminant to be raised, in steps, up to approximately 5,0000 K. Whether this would be sufficiently blue' had yet to be proved, but it did in fact turn out to be regarded as blue by the observers under almost all conditions of adaptation. A grid of chromaticity points as shown in Fig. 9 and detailed in Table 11 was obtained by the abovementioned colour-teMperature-raising filters in combination with Chromex 4MG, SMG, 12MG, and 16MG magenta filters. The neutral filters mentioned in Table 11 are necessary to prevent considerable changes in luminance on changing from one colour combination to another. The neutrals were tested for their neutrality (i.e. absence of colour bias) and found to produce shifts which were small compared with the spacings of the chromaticity points shown in Fig. 9. The chromaticitiesof points 5, 10, 15, and 25 were directly measured, the chromaticities of points 1, 3, 11, 12, 13, 14, 21, and 23 were calculated from spectrophotometric data on the filters obtained with the Unicam spectrophotometer and the remainder of the 25 points interpolated.

2.2.2 Experimental Method.

The general layout is shown in Fig. 10 and will be seen to be similar in many ways to that used for Method I. The same projector was used to provide the surround illumination, although in this part an additional 2 MG filter was added to cancel the

* The actual value by which adjustments were made was e mir~s.

TABLE II

LIST OF COLOUR FILTER COMBINATIONS USED IN METHOD 11

MG CTR Neutral MG CTR Neutral MG C'1'R Neutral MG erR Neutral MG CTR Neutral

(1) (2) (3) (4) (5) - 0'-97 0'34 - 0'75 0'40 - 0'55 0'54 - 0'29 0'76 - - 0'90

(6) (7) (8) (9) (10) 4 0'97 0'25 4 0'75 0'34 4 0'55 0'54, 4 0'29 0'66 4 - 0'90

(11) (12) (13) (14) (15) 8 0'97 0'145 8 0'75 0'25 8 0'55 0'34 8 0'29 0'54 8 - 0'76

(16) (17) (18) (19) (20) 12 0'97 - l2 0'75 0'25 12 0'55 0'34 12 0'29 0'54 12 - 0''76

(21) (22) (23) (~) (25)

16 0'97 - 16 0'75 0'145 16 0'55 0'25 16 0'29 0'40 16 - 0'66

METAL MASK WITH WHITE CARD ON FRONT SURFACE.

I I

FIG.IQ

9"

APPEARANCE OF FIELD

COLOUR FILTERS

GENERAL LAYOUT-METHOD 2

CENTRAL FIELD.

SURROUND FIELD.

MAIN PROJECTOR.

»

8

slight green trend of light from this projector. The observers were seated about 3 feet from the screen and colours were presented to them in random order. These they were asked to name in accordance with a restricted list, viz., white, red, green, blue, magenta, cyan, and yellow; or, expressed in a little more detail, if the observer did not consider the colour to be white he was asked to name the direction in which it departed from white, using the tricolours and the complementary tricolours. For those observers to whom some of the names were strange or uncertain, a demonstration of the fully saturated colours was given before the test started.

It will be recollected that this method was designed primarily as a cross­first method, the most doubtful point being the departure from the black­

As a first check on this, the condition of no adaptation was used and a check on the body locUS. restricted number of colour combinations presented to the observers, viz. 1, 3, 5, 11, 13, 15, 21, 23, 25. Each combination was presented twice, although the observer was not informed that this was going to be done. The same twelve observers* were used as in Method I, and the individual consistency of naming was immediately found to be good. Further, there was a much greater measure of agreement as to which of the colour com­binations was to be called white than might have been expected. The general analysis of the individual answers is shown in Table III and graphically in Fig. 11 which·also gives the results of Method I. The assessment of white for each observer is based partly on their direct naming of a particular combination as white, and also on their judgments of green-magenta, blue-yellow, red-cyan. Hence it is possible to assess the white point even when the observer does not name any colour combination as white. Fig. 11 gives the individual white points of the twelve observers, seven of whom called combination 13 Fig. 9 white both times it occurred in the random sequence in which the colours were presented. Other observers preferred a bluer white but there was complete agreement on both showings that combination 11 was bluish. This contrasts !3omewhat with the whites chosen when the observers had control of the knobs of the additive tricolorimeter. Further, there is no question of any of the combinations 1, 3, or 5** being named white, i.e., no detectable tendency to choose a point on the green side of the black-body locus. Nor again are combinations 11, 13, or 15 described as magenta, which must occur if any observer's' subjective white point lies midway between the lines given by 1, 3, 5 and 11, 13, 15. The mean value of the white point for the twelve observers is just' over 4, ()(X)O K, individual preferences ranging from 3,9000 K to 4,700o K. Any tendency to green must be small, and certainly less than half the mean value of green noted in the first experiment.

* With one exception, LR replaced WNS for Method II.

** Deta1le~ inspection of Table III will show that this is not quite. true. Combination No. 1 on being presented first in the sequence was called White by five observers. When No. 1 was ~resented a second time later on in the sequence, no observer called it White. It lS assumed that comparative inexperience (there was no trial run) or lack of proper adaptation at the beginning of the experiment is the reason for the original namings not being consistent with the remainder.

1·46r-------------------T--------------------r------------------~------------------~r-------------~

·4Sr-------------------+-------------------~------------------~------------------~r_------------~

FIG.l1 METHOD 2. NO ADAPTATION, TWELVE 08SERVERS

RESULTS OF METHOD I ALSO SHOWN FOR COMPARISON.

2700' K

·40~------------------+-------~~~~~--_+------~--~~----~------~~~~~~~~--------------~

MEAN VALUE METHOD I

6S00'K

7000' K

R

THE LETTERS AT EACH GRID POINT REPRESENT AVERAGE NAMINGS (SEE TABLE 3). INITIALS GIVE OBSERVERS? PLACINGS OF WHITE.

1·30 I----------------....L....--+-------------------+------------------~--------------------t__-----------I

['29

0·25 0·30 0'3S 0-40 0-45

9

TABLE III

No Adapting Stimulus

NUMBER OF OBSERVERS

Colour General Reaction Combination W R Y G C B M No. White Red Yellow Green Cyan Blue Magenta (12 Observers)

1 5 3 4 cBW 3 9 cB

3 1 4 5 2 GC 8 4 CG

5 9 3 gy 10 2 gY

11 12 B 12 B

13 11 1 W 8 1 3 yW

15 3 9 rY 7 5 YR

21 3 9 bM 2 10 bM

23 1 5 1 5 RIM 4 8 RH

25 9 1 2 mR 12 - R

Sequence used in the experiment: 1, 25, 13, 15, 3, 21, 23, 11, 5, 21, 13, 5, 15, 1, 23, 3, 25, 11.

Each colour presented twice: reactions shown separately in above table.

Coding used in General Reaction column: One capital letter only means that all observers agreed on naming or at most'one dissentient only. A small letter followed by a capital means that two or three observers give the reply indicated by the small letter, the remainder giving the naming indicated by the capital letter. Two capJtal letiters means roughly equalnamings but with the latter letter pre­dominating. Two capitals separated by an oblique stroke mean exactly equal numbers for the two namings.

It was not originally intended to repeat the whole of the experiment by this second method, but ac a much greater level of consistenc1was being achieved, it seemed imperative to do so. This was done with the full set' of 25 colour combinations for both tungsten and daylight ac~ptation. Repeats were not used, as this would have involved extending each session from a quarter of an hour to half an hour for each type of adaptation. However, attention was paid to internal consistency of naming, and most observers were found to possess a high level of self-consistency.

10

TABLE IV

Tungsten Adaptation

Colour General Reaction Combination W R Y G C B M (12 Observers) No.

1 9 3 bC 2 3 9 gC 3 7 5 CG 4 1 10 1 G 5 2 10 yG

6 2 10 cB 7 3 9 cB 8 8 4 cG 9 11 1 W

10 6 6 Y/G

11 12 B 12 10 ~ mE 13 2 10 wB 14 9(11) 2(1) 1(0) rW(W) 15 3 ·2 7 wrY

16 10 2 mE 17 4 8 bM 18 2 3 1 6 wrM 19 7 4 1 RW 20 1 9 2 yR

21 3 9 bM 22 2 10 rM 23 12 M 24 9 3 mR 25 11 1 R

Sequence used in the experiment: 23, 19, 25, 13, 1, 2, 4, 9, 12, 15, 17, 6, 22, 8, 14, 21, 20, 11, 5, 16, 3, 18, 24, 10, 7, 14.

No. 14 was presented twice and bracketed figures are second namings.

The results for tungsten adaptation are summarised in Table IV and shown graphically in Fig. 12. The higher level of consistency between observers is again obvious in contradistinction to the results of the first methods and there is only a slight tendency of the group as a whole towards a greenish white - so slight that it is probably of no statistical significance. Evaluated as a (simple) moment diagram, the mean white is 0~08 of a step towards the green. (The step being a change in filtering of 4 MG.) This is less than one-tenth the excess green content of the white chosen by the group as a white when using the first method. The mean value now has a colour temperature of 3,400o K with individual preferences varying from 3,1500 K to 3,7000 K.

Fig. 13 and Table V show the results for daylight adaptation. The mean white point will be seen to correspond to a colour temperature of 4,0500 K with individual preferences ranging from 3,4000 K to 4,4000 K. The mean tendendency to­wards a greenish white is 0,125 of a step, and is thus again small compared with the

'O·46.------------r-----------r---------,-----------,r-------

O.45~---------------_4-----~----------_+------------------_r------------------;-------------__.

FIG.l2 METHOD 2. TUNGSTEN ADAPTATION

TWELVE OBSERVERS RESULTS OF METHOD I SHOWN FOR COMPARISON

METHOD 1---191

ADAPTING STIMULUS

·35~--------~~----4r-~~~r_-\~~---~~_i---------_t-------_;

6S00'K

THE LETTERS AT EACH GRID POINT REPRESENT AVERAGE NAMINGS (sEE TABLE 4). A NUMBER GIVES THE NUMBER OF OBSERVERS WHO CAlLED THE

,30~----------~~_+------------~-------------_1-----------------~

29L-_________ ~~ ________ ~~ __________ ~ ______________ ~~--------~~~

0·25 0·30 0·35 0'40 0-45 0·49 COEFFICIENT OF X

0·~6~------------------~------------------~------------------~------------------~------------~

O·~S~------------------~------------------+-------------------+-------------------+-------------~

FIG. 13 METHOD 2 DAYLIGHT ADAPTATION

TWELVE OBSERVERS RESULTS OF METHOD I ALSO SHOWN FOR COMPARISON

O'~O~------------------~----~~--~~~--+-----~~~~------+-~T---~~~~~~~--------------~

>-u.. 0

~ R

~ !::! u.. u.. I.&J 0 u

0·35

M

0·30~--------------L---r-------------------+-------------------+-------------------+----------------1

O·29~. __________________ ~ __________________ ~ __________________ ~ __________________ ~ ______________ ~

0·25 0'30 0·35 0-40 0-4S COEFFICIENT OF X

11

TABLE V

Daylight Adaptation

Colour General Reaction Combination W R Y G C B M (12 Observers) No.

1 8 4 BC 2 6 6 C/G 3 10 2 cG 4 3 9 yG 5 12 Y

6 3 9 cB 7 5(0) 1(8) 1(2) 5(2) B/W(cG) 8 8 4 CG 9 3 8 1 wY

10 1 11 Y

11 11 1 B 12 5 4 3 BW 13 11 1 W 14 5 2 5 Y/W 15 5 7 RY

16 1 4 7 BM 17 2 10 bM 18 1 11 M 19 12 R 20 12 R

21 12 M 22 2 10 rM 23 4 8 RM 24 7 5 MR 25 12 R

Sequence used in the experiment: 7, 10, 24, 18, 3, 16, 5, 11, 20, 21, 14, 8, 23, 19, 25, 13, 1, 2, 4, 9, 12, 15, 17,:'., 22, 7.

No. 7 was shown twice; second namings shown in brackets. in the series, No. 7 was called White by five observers; much more consistent with the other namings.

On being presented first the repeat gave a result

results of Method I. The chromaticity of the daylight adapting stimulus corresponded to a colour temperature of about 4,3000 K. This is a very yellowish type of daylight, and, for completeness, a blue type of daylight (roughly north sky light) was used as a fourth condition of adaptation. It is not considered ve~y likely that the ambient lighting reaching a colour television receiver in a room with the usual type of furnishings would be as blue as north sky light, but the daylight surround adaptation used previously may be a little too low in colour temperature to represent the average sort of ambient daylight. A further calibration point was hence considered to be desirable.

Fig. 14. The results with this fourth adapting stimulus are shown in Table VI and For the first time, combination No. 11 was not called blue by almost all

observers; on this occasion eight still classed it as blue and the remaining four as white. However, even with this bluish adapting stimulus, no observer classed No. 11

)·46r-------------------r-------------------,--------------------r------------------~---------------

1·4S~------------------~--------------~--~------------------~-------------------+--------------~

FIG.14 METHOD 2. NORTH SKY ADAPTATION

TWELVE OBSERVERS.

GY

2700'K

)·40~------------------+-------------~~--~------~--~~-----+~------~~--~~~--------------__4

~

.... ::>

-z: .., ~ .... .... .., ::> :J

1'35

bC

ADAPTING STIMULUS

65000 K

70C1JO K

8OOO0 k

R

M

D·30~--------------~--+-------------------~-------------------+------------------_4----------------1

O·29~ __________________ ~ __________________ -L __________________ -L __________________ ~ ______________ __J

l 0·25 0·30 0·35 0'40 0·45 COEFFICIENT OF X

jP

tnz

12

as yellow, i.e., no one required a white of higher colour temperature than that given by combination No. 11. The mean value of white for the whole group is now 4,460o K, with individual preferences ranging from 3,800o K to 5,300o K. One observer selected point No. 7 as white and gave judgments of neighbouring points completely consistent with this choice. Apart from t~s one observer, the choices were either along the central (black-body) line or with only a slight trend towards the green. The mean value for the group of observers shows a trend towards the green of 0221 step, rather more than the daylight (4,300o K) stimulus, but still of small magnitude.

It will have been noticed that as the colour temperature of the surround is increased, the white selected by the observers has risen in colour temperature. This is summarised in Fig. 15, which shows colour temperature of subjective white against the colour temperature of the surround illumination. The range of colour temperature used in this work is considered adequate for all colour television viewing in the home,

TABLE VI

North Sky Adaptation

Colour General Reaction Combination W R Y G C B M No. (12 Observers)

1 1 9 2 bC 2 12 G 3 1 11 G 4 3 9 yG 5 7 5 GY

6 3 1 4 4 wC/B 7 3 3 2 4 ~B 8 3 9 9 12 Y

10 11 1 Y

11 4 8 WE 12 8 1 1 1 2 mW 13 9 2 1 rW 14 2 3 7 wrY 15 5 7 RY

16 2 10 bM 17 2 10 rM 18 12 R 19 10 2 yR 20 9 3 yR

21 ~ 12 M

22 1 11 M 23 6 6 RIM 24 12 R 25 11 1 R

Sequence used in the experiment: 15, 17, 6, 22, 8, 14, 21, 20, 11, 5, 16, 3, 18, 24, 10, 7, 23, 19, 25, 13, 1, 2, 9, 4, 12.

No repeats.

1 • .4,0;'-1 ----------r---------r---------,-'---------T-------

580 0

,... '" on 1&.1 z: .... :::c !:l! a:: .a -I .. .... o

....

500

~ 400 CL

1&.1 .... :::c :.

0

0

I

.. UPPER LIMIT FIG.l5 SIGNAL WHITE

EFFECT OF COLOUR TEMPERATURE OF SURROllfD ON MEAN WHITE POINT.

~ ~ ADAPTING STIMULUS

..-

z: c 1&.1 E

~ ~

lOO 0 I-LOWER LIMIT

SIGNAL WHITE

1\ 260 2400 3000 4000 5000

COLOUR TEMPERATURE OF ADAPTING STIMULUS oK (/ft-t 8R/GHTNESS)

.-e-

6000 6800

13

although a purely academic investigation might continue with both higher and lower colour temperatures. From general considerations, it is thought that the curve in Fig. 15 must be'S' shaped, although the central approximately straight portion is all

that is apparent from this work.

3. DIRECT COMPARISON OF METHODS I AND lI.

The fact that Method I gives a much greater spread of chromaticity points is not difficult to understand, since the adjustment of two controls may be a rather difficult task for persons unskilled in colour work, and certainly more difficult than just naming a colour. That the mean values should be significantly and markedly different is more difficult to understand. To check that this really is a genuine effect and not merely a calibration or other error, the two methods were set up side by 'side and each observer was asked to set up a white with the additive tricolorimeter (under conditions of no surround adaptation). Having taken as long as he chose over this and being as satisfied with the results as the slight non-uniformity of the tri­colorimeter field permits, he was' then shown some of the combinations used in Method 11 and asked to give preferences as between the tricolorimeter white of his own making and the other one (both at the same brightness level). From the preferences expressed it became clear that the difference between the methods is genuine and that Method 11 gives the more reliable result. It would app,ear that here we have an experiment whe::e it is unwise to give the observer control over the variables and a procedure of nammg'samples selected in a random order is the proper method to adopt.

4. COMPARISON WITH OTHER WORKERS.

Two specifications for signal white have alrea~7 been mentioned, and Fig. 16 , shows the results of the present work (Method 11) together with areas on the chromaticity diagram corresponding to these specifications. B.S. 1376 seems to be a slightly less critical specification than the one given by Gibson, Haupt, and Keegan, but the areas are rather similar. The present work fits in quite well with the American paper and seems to suggest that although the purpose for which the Lunar White specification was issued was rather different, nevertheless it provides a satisfactory specification of white for colour television receivers. The extra region in the blue from 5,5000 K, to 7,6000 K given by B.S. 1376 would not appear to be required for our purposes.

Another specification for white, in this case for the luminescence emitted by a cathode-ray tube phosphor used in monochrome television, is quoted by Sheldon (7). This specification gives an area on the chromaticity diagram which extends from 5,0000 to 9,0000

K for a tube with a clear glass envelope. This range of colour tempera:-ures is ap~reciably more blue than the colour temperatures of any of the white~found in the present investigation.

American colour television workers seem to have used Illuminant C as a white point to some considerable extent. A recent paper on the C.B.S. system gives Illuminant C for

0.46r---------------~----------------~-----------------r----------------,--------------

0.45~---------------r----------------+---------------~--------~~~--~------------J

FIG.16 COMPARI50N OF PRE5ENT WORK

WITH 51GNAL WHITE 5TANDARD5.

~ 27000K 0.40~---------------r----------------+-~==~~~~---+~~~--~~~~~------------J

>­

'"" o I­Z .... Q La. La. .... o ..,

\........-::~_;:7''''--",.~~''OOOOy /'

.~~~/'/ TELEVISION WHITE ~~P' /' ~~AAKnM /

SIGNAL WHITE GI8S0N. HAUPT & KEEGAN. J.O.S.A. DEC. 1945.

PROC.I.R.E / D.lS j-__________________ rO~T.~I~J·~~~~~sP~'~~8~O~OO~K----------~~-+-------------------+---------------J

'·50 1-------L:........:.:~~-----::.,L--+--------I---------I---------1

"Z9 ~---------t.30-------L~--,d~---------:~-------~--------.J 0·25 O·lO 0·l5 0·"0 .. 0'45 0'49 COEFFICIENT or: It

h

14

some sets of primaries but goes on to describe 'a more satisfactory white' which turns out to have chromaticity co~ordinates of (x = 00'330 y = 0,341 z = 02329)*. Goldmark and his co-authors are not very specific in their comments as to what precisely is white; they content themselves with the comparison of whites given by two sets of primaries. This more satisfactory white is definitely blue by comparison with our own results, and is not the optimum point for which to aim under any conditions of surround illumination likely to be encountered in practice.

5. CONCLUSIONS.

Subjective white has been determined for twelve colour-normal observers under four conditions of adaptation. Two methods were used, and of these, that in which the observer does not have control of the variables is shown to be the more reliable and consistent. A slight tendency towards a greenish white appears to exist, but it is certainly of small magnitude. The type of adaptation has a definite influence on the observers' choice of white, and numerical data on this are given (Fig. 15). The results of the present work come within the range given by Signal White specifications and suggest that the whites reported in the American colour television literature are definitely too blue.

6. RECOMMENDATION.

The optimum white has been shown to be appreciably affected by the adapting field, and it seems impossible to satisfy the requirement for white when viewing with tungsten ambient lighting simultaneously with the requirement when viewing with daylight ambient lighting (or no ambient lighting). It is probably legitimate to assume that most television viewing (80 per cent. or more) is done in the evenings. An enquiry conducted by Audience Research found that, only 12 per cent. of viewers use no ambient lighting. For evening viewing 88 per cent. of the television public use tungsten ambient lighting while viewing. Fig. 17 shows a recommended white area on the C.I.E. diagram based on the results of tungsten adaptation. It is centred on the mean given by Method 11 and gives an area corresponding to the 50 per cent. ellipse for the average colour-normal observer as found in Method I.

The equation for the recommended white is as follows:

336OX2 ~ 779204X + 10,224~ - 4,339~44y - 8,976xy + 687:191 = 0

This is an ellipse of semi major and minor axes of lengths 0:033 and 0:010 with the major axis parallel to the black-body locus and the centre at the co-ordinates x = 0:405, y = 0:390, i.e. 3,5000 K.

* These values are not given in the C.B.S. paper but follow from the statements of chromaticity co-ordinates of the primaries and their relative luminosities.

O~6'-----------------'-----------------'-----------------'-----------------'--------------'

O·+5r-----------------r-----------------+-----------------~----------------~------------_1

FIG.17 RECOMMENDED WHITE FOR VIEWING WITH

TUNGSTEN AMBIENT LIGHTING.

O·40r-----------------r-----------------+-----------------~~~+7~~nr~~~------------_;

>­... o l­X ~ u u: u.. w o ....

O·~s~----------------r-------------~L-+_----------------+_----------------~------------~

60000K

6S000K

0·30~------------~--r-----------------+-----------------+_----------------1_------------_4

O·29L-________________ ~ ________________ ~ ________________ ~ ________________ ~ ____________ ~

o·a 0·30 0·3S 0·40 COEFFIC.IENT OF X

15

APPENDIX

Method of Calculating the Ellipses of Variation (Method I).

Each observer made three settings for a given adaptation condition: let the chromaticities corresponding to these three readings be

Let the mean chromaticity be x y,

thus x =i(x1 + x 2 + x 3 )

y =~(Yl +' Y2 + Y3)

Let the deviations from the mean be

xl - x etc.

Yl - Y etc.

For a given state of adaptation we have 12 sets of results

of I1xl ~Yl" ~Y3 (one for each observer)

By considering only the 11 values we are intentionally ignoring any differences in the mean values (such as are shown in Figs. 4, 5, and 6) but are considering the amounts by which different observers vary from their own means. From the complete set of values of 11 we calculate the average degree of consistency of an observer in repeating his own settings. The method is as follows:

From the 36 values of 11 x!::.. y,

we find ~ (I). x)2

=

where N = number of values of 11 x

=

16

p= ~(~x ~y)

1'2(l1x)2 ox ~(~y)2

where~x~y are standard deviations and p is the coefficient of correlation.

The ellipse of variation is given by

where a is determined by the probability of including a known fraction of the total results.

a is obtained from the equation

P 1 ~ exp {- 2(1 ~ P 2~} where p is the fraction of the results included within the ellipse. cent. ellipse this reduces to

REFERENCES

1~ B.S. 1376: 1947, Colours of Light Signals.

For the 50 per

2~ Gibson, K.S., Haupt, G.W. and Keegan, H.J., J. Opt. Soc. !mer., 35/772/1945.

3~ Goldmark, P.C., Christensen, J.W. and Reeves, J.J., Froc. Inst. Radio Engrs, 39/1288/1951.

4.. Hopkinson, Mackenzie and Nixon, Photo Journal, 918/2/1951.

50. W.N. Sproson, B.B.C. Research Dept. Report No. T.034.

6;. MacAdam, D.L., J. Opt. Soc. !mer., 32/247/1941.

7,. Sheldon, J.L., J. Soc. Mot. Pict. and Tel. Engrs, 56/65/1951.