Skip to main content

Full text of "Photometry of the gas-filled lamp"

See other formats


NATIONAL INSTITUTE OF STANDARDS & 

TECEtiOLOGY 

Research Information Center 

Gaithersbmg, MD 20899 



PHOTOMETRY OF THE GAS-FILLED LAMP 



By G. W. Middlekauff and J. F. Skogland 



CONTENTS 

Page. 

I. Introduction 587 

II. New variables and photometric difficulties 588 

III. Experimental investigation of the variables 589 

1. Lamps, apparatus, and methods 589 

2. Preliminary measurements 590 

3. Current variation at constant speed with different voltages 590 

4. Current variation at constant voltage with different speeds 592 

5. Voltage variation at constant current with different speeds 594 

6. Candlepower variation at constant voltage with different speeds. . 595 

IV. Law of the variations 596 

V. Two photometric methods 597 

1. Rotating lamp method 597 

2 . Total flux method 598 

VI. Possible errors introduced by rotation 599 

VII. Investigation of the cause of the variations observed 600 

VIII. A probable explanation of the cause of the variations observed 601 

IX. Summary 602 

I. INTRODUCTION 

In a recent paper ^ the authors have shown that for all vacuum 
tungsten lamps, within a wide range of wattage, the voltage- 
current-candlepower characteristics are the same regardless of 
the make or method of manufacture. When this investigation 
was extended to include gas-filled lamps, it was soon discovered 
that before consistant and reproducible results could be obtained 
on the photometer, new and unexpected variables, not present 
in the vacuum lamp, had to be carefully considered. It is with 
an experimental study of these variables that this paper is par- 
ticularly concerned. 

^ A preliminary paper on this subject was read at a meeting of the scientific staff, Bureau of Standards, 
Nov. 20, 1914, and afterwards published in the Electrical World, Dec. 26, 1914, p. 1248. 

^Middlekauff and Skogland, Characteristic Equations of Tungsten Filament Lamps and Their Applica- 
tion in Heterochromatic Photometry, this Bulletin, 11, p. 483 (Scientific Paper No. 238). 



588 Bulletin of the Bureau of Standards [voi. 12 

II. NEW VARIABLES AND PHOTOMETRIC DIFFICULTIES 

In the -gas-filled lamp the filament is coiled into the form of a 
closely wound spiral or helix, the diameter of which, though 
small, is large in comparison with the filament itself. Hence, as 
viewed from certain directions perpendicular to the axis of the 
lamp, one section of this helical filament may, more or less com- 
pletely, cut off the light from another section. In addition to 
this cause of irregularity in the light distribution, there is usually 
a lack of symmetry in the arrangement of the filament as a whole 
about the vertical axis of the lamp. Consequently, the light on 
the photometer screen flickers excessively when the lamp is 
rotated at ordinary speeds. Hence, without auxiliary apparatus, 
accurate measurements of mean horizontal candlepower are 
practically impossible; the weight of the lamps, especially of 
the larger units, the necessity of using mercury in the rotator, 
etc., preventing a sufficient reduction of the flicker by the use of 
high speeds of rotation. Fortunately, this difficulty is readily 
overcome by placing back of the lamp two mirrors ^ inclined to 
each other so that the photometer screen is illuminated simul- 
taneously by the direct light from the lamp and the reflected light 
from the mirrors. AVith this arrangement very low speeds of 
rotation may be used. 

Although this expedient obviates the flicker difficulty, it does 
not eliminate the most serious trouble caused by rotation. It 
was found that at constant voltage both current and candlepower 
have different values when the lamp is rotating than when it 
is stationary, the current changing in one direction and the candle- 
power always in the opposite direction. In other words, there is 
a change in the operating efficiency of the lamp although the 
voltage remains the same.^ It was found, further, that the 
values of current and candlepower may be either increased 
or decreased by rotation, depending upon the speed, and by 
amounts sufficiently great to affect seriously the result of candle- 
power and efficiency determinations. 

3 This Bulletin, 2, p. 416; 1906. 

* This general effect was observed about the same time also at the Electrical Testing Laboratories, New 
York, and described by Dr. C. H. Sharp, Trans. Ilium. Eng. Soc, 9, p. 1021; 1914. 



sk%fanr^'] Photometry of the Gas-Filled Lamp 589 

The gas-filled lamp therefore presents a new problem in pho- 
tometry, involving not only greater difficulties on account of 
greater color difference over those encountered in photometering 
vacuum lamps but also the difficulty of dealing with variables 
which, if not properly considered, may introduce considerable 
uncertainty in the results. 

Below are given some experimental results obtained in a sys- 
tematic study of the lamp and from these the law of the varia- 
tions is deduced. From a consideration of this law a practical 
method of measuring mean horizontal candlepower free from 
the errors caused by rotation is discovered and an explanation 
of the cause of the variations observed is offered. 

III. EXPERIMENTAL INVESTIGATION OF THE VARIABLES 

1. LAMPS, APPARATUS, AND METHODS 

The lamps used in the principal experiments described below 
were of the nitrogen-filled 450- watt series and 7 50- watt and 
looo-watt multiple types, all having round bulbs with the fila- 
ments centrally mounted. The number of anchor wires in the 
different lamps varied from 5 to 9. Additional experiments 
were made with a number of 200-watt series and 750-watt mul- 
tiple lamps \vith straight side bulbs, the former having one anchor 
wire, the latter five. Other experiments were made with an 
ordinary vacuum lamp filled with nitrogen and also with a vacuum 
lamp having a coiled filament such as used in the gas-fidled lamp. 

Measurements of voltage and current were made simultaneously 
b}^ means of two potentionieters. The rotator was supplied 
with mercury-cup connections and two sockets so that the lamp 
could be mounted either tip up or tip down. The rotating 
mechanism was under perfect control at every speed, which 
was determined by a specially designed indicator, and the cur- 
rent to the lamp was supplied from a storage battery, thus insur- 
ing perfectly steady working conditions. In the candlepower 
measurements use was made of the two -mirror arrangement 
described above. 

A few rough preliminary tests indicated that for any change 
in speed the percentage change in candlepower was at least 



590 Bulletin of the Bureau of Standards [Voi. 12 

several times the percentage change in current but always in 
the opposite direction. As current can be measured with con- 
siderably higher accuracy than candlepower, such measurements 
furnish a convenient method of determining approximately the 
variation of the candlepower. It was therefore decided that 
before making careful determinations of candlepower, it would 
be advisable to make a series of measurements of voltage, cur- 
rent, and speed, varying only one of these quantities at a time. 

2. PRELIMINARY MEASUREMENTS 

The general effect of rotation on current and power consump- 
tion with voltage constant was first determined in three series 
of fairly accurate measurements made on a ii6-volt, 750-watt, 
short-necked lamp at 90, 105, and 116 volts, respectively. In 
each series current readings were taken at different predeter- 
mined speeds, a reading with the lamp stationary being taken 
as a check between the changes of speed. During these measure- 
ments the lamp was used first tip up and then tip down. With 
the lamp in either position and the speed above 40 revolutions 
per minute the current was less than with the lamp stationary 
and decreased with increasing speed, the effect with tip up being 
about twice as great as with tip down. Other lamps of the same 
size and of different sizes (450 watts to 1000 watts) and of different 
manufacture were tested in the same manner and similar results 
were obtained in every case. 

In these preliminary measurements it was learned that very 
consistent and reproducible current values could be obtained 
provided the speed of rotation is alccurately known and properly 
controlled; otherwise apparently inconsistent and erratic results 
were to be expected. Hence, in all subsequent tests the speed 
was as carefully measured and controlled as the voltage and 
current. 

3. CURRENT VARIATION AT CONSTANT SPEED WITH DIFFERENT 

VOLTAGES 

The next step taken was to determine with great care the 
changes in current due to rotation at a specified speed at different 
values of voltage through a wide range. This test was applied 



Middlekauff.'] 
Skogland. J 



Photometry of the Gas-Filled Lamp 



591 



to two lamps, the 7 50- watt lamp mentioned above and a 1000- 
watt lamp, both rated at 116 volts. Readings of current corre- 
sponding to a chosen voltage were first made with the lamp sta- 
tionary, then with it rotating at a speed of 150 rpm. In this way 



105 
104 

103 
103 
101 
100 

ee 

es 

97 









1 
























A 
































\ 






























\ 


N 




i^ 




























"^ 


J 






31 rpm 


Tip Down 






1 T 


1 — *i 










^v^ 




_^ 




IE 


^ ipta 


Tip Down 




\—. ^ 




Tip Up __ 










- > 


K^l 




















A 


/ 




A 


^ 


























/ 


/ 
























/ 




/ 


























[ 


J 




























' 


A 




























> ( 


/ 




























b 














laap At 1000-lfatt, 116- Volt 
Lamp B: 750-iratt, ll&-Volt 






f 




















V 





























































10 
Volts 



90 100 110 130 130 140 150 



Fig. I. — Ratio of current with lamp rotating at constant speed to current with lamp 

stationary 

the changes in current for each lamp at different voltages over a 
range from 0.5 volt to 100 volts were obtained. 

The results for both lamps rotated tip up are shown by the 
lowest curve of Fig. i, values obtained for the 7 50- watt lamp 



592 Bulletin of the Bureau of Standards [Voi. 12 

being represented by crosses and those for the looo-watt lamp by 
circles. The behavior of the current in these two lamps is thus 
shown to be practically identical at every value of voltage except 
from 3 volts to 8 volts, which is a critical region. It is probable 
that even within this region these curves would be found to coin- 
cide also were measurements made at a sufficiently large number 
of neighboring points. The upward indentation in the curve at 
this point is as real and reproducible as any other part of the 
curve. 

Upon rotating these lamps tip down, it was found that for the 
same speed the percentage change in current for all voltages was 
only about half as great as when the lamps were rotating tip up. 
The values for the looo-watt lamp tip down are given by the 
middle curve of Fig. i.. This curve is traced through all values 
obtained except one in the neighborhood of 5 volts. This value 
indicates that an indentation would probably have been found in 
this curve also had sufficient time been taken to pass by small 
voltage steps through this region. It might be stated that, in 
order to obtain sufficiently steady conditions for accurate meas- 
mements of crurent in this region, a period of at least 10 minutes 
was required at each point, as small variations of speed resulted 
in relatively large changes in current. These two curves are 
similar in form, both becoming asymptotic to the axis of abscissas 
at about i per cent and 0.5 per cent, respectively, below the current 
obser^^ed for the lamp stationary. That is, for this speed, at about 
85 per cent of the normal voltage, the current reaches a value 
which from that point on bears a practically constant ratio to 
the value at zero speed. 

4. CURRENT VARIATIONS AT CONSTANT VOLTAGE WITH DIFFERENT 

SPEEDS 

In order to investigate further the behavior of the current in 
the critical region above mentioned, as well as at other points, 
measurements of current at constant voltage, with variable speed, 
in succession at different values of voltage were made on the 
looo-watt lamp through a considerable range extending as low as 
0.5 volt. The speed was varied from 5 rpm to 200 rpm, the lamp 
being rotated tip down. The results of this test are shown in 



Middlekauff,! 
Skogland. J 



Photometry of the Gas-Filled Lamp 



593 



Fig. 2. The important fact indicated by these curves is that as 
the speed is increased from zero the current, at constant voltage, 
first increases to a maximum and then decreases to the stationary- 
value, and continues decreasing below the stationary value as the 
speed is farther increased. (By stationary value is meant the 
value obtained with the lamp stationary.) 

In order to confirm this change in sign of the current variation 
the 7 50- watt lamp was tested tip down while rotating at a speed 



99 

98 

97 

96 

95 

g 94 
u 



93 

































































^ 


^^ 
















00 Vol 


t8 














—0 1 




\ 


s 






























V 


■®s 




■Ci 






— a-_ 




30 Vc 


Ite 














\ 


X 

































X 


\. 




X 


X^ 
























N 


\ 




s 


^0.! 


' Volt 






















N 


X 






























\ 


\ 


1 Vol 


I: 





































20 40 60 80 
Pevolutlone Par Minute 



100 130 140 160 180 200 330 340 



380 300 



Fig. 2. — Ratio of current with lamp rotating tip down at different speeds to current with 
lamp stationary; voltage constant at different valves 

of 2 1 rpm. The results of this test are shown -by the uppermost 
curve of Fig. i . It will be noted that this ctuve is of practically 
the same form as the other two curves of this figure, but that it 
is turned in the opposite direction and lies entirely above the loo 
per cent or stationary values. That is, at this speed of rotation 
the current is higher than at zero speed for all values of voltage. 
Hence, correct qualitative results may be obtained at any prac- 
tical voltage. 

37703°— 16 8 



594 Bulletin of the Bureau of Standards • [Voi. 12 

Upon carefully testing all available 450-watt to looo-watt 
lamps, including both series and multiple types in roimd bulbs 
and one 750-watt lamp in a straight-side bulb, it was found that 
in every lamp while rotating with the tip down the current at con- 
stant voltage returned to the stationary value at a speed between 
33 rpm and 40 rpm. This held true through a considerable range 
of voltage, extending above and below the normal, or rated, value. 
The series lamps included in this test had the same form of moimt 
and practically the same arrangement of the filament as the mul- 
tiple lamps. 

For 200-watt, 6.6-ampere lamps, with V-shaped filament and a 
single anchor wire, it was found necessary to use a speed of from 
140 to 160 rpm to obtain the stationary value of the current, the 
value of the ciu-rent being greater than the stationary value up to 
the speed mentioned. It appears, therefore, that the smaller the 
number of loops and corresponding anchor wires in the lamp the 
greater the speed that is required to give the stationary value of 
the current. 

Although there was a considerable difference in the speeds 
required for lamps having different numbers of loops in the fila- 
ment, it was possible to find for every lamp tested a particular 
speed at which the cinrrent returned to the stationary value, the 
speed being in every case such as can be used in practice. 

5. VOLTAGE VARIATION AT CONSTANT CURRENT WITH DIFFERENT 

SPEEDS 

The effect of rotation on voltage was determined by testing a 
few of the 450-watt lamps, tip down, with the current held con- 
stant at 6.6 amperes and the speed varied over a wide range, 
beginning at 5 rp^m as the lower limit. For every lamp it was 
foimd that as the speed was increased from zero the voltage first 
decreased to a minimum and then increased to the stationary 
value and continued increasing above the stationary value as 
the speed was further increased. In other words, the effect of 
rotation upon the voltage at constant current was opposite in 
sign to the effect upon the current in the same lamps at constant 
voltage. 



MtlS""^'] Photometry of the Gas-Filled Lamp 595 

Since the variable quantity (i. e., voltage or current) in either 
case returns to its stationary value, it is evident that the speed at 
which this occurs, and therefore at which the power consumption 
is the same as when the lamp is stationary, may be determined by 
holding either quantity (voltage or current) constant and observ- 
ing the variation of the other as the speed is gradually increased 
from zero. Principally as a matter of convenience, therefore, all 
subsequent tests were made at constant voltage. 

6. CANDLEPOWER VARIATION AT CONSTANT VOLTAGE WITH DIFFERENT 

SPEEDS 

The effect of rotation on candlepower was very strikingly shown 
by the following experiment: The voltage was so adjusted that 
the filament when stationary was just visible in a darkened room. 
Upon rotating the lamp at a speed of about 1 80 rpm it was found 
that, although the current decreased in value, the intensity of the 
light largely increased, the effect being greater with the tip up 
than with the tip down, thus corroborating the preliminary meas- 
urements of ctirrent and candlepower at higher voltages. With 
the lamp in either position and either rotating or stationary the 
upper loops of the filament glowed more brightly than the lower 
ones, thereby indicating a considerable difference in temperature 
between closely neighboring regions of the gas. 

Since there was a speed at which the power consumption was 
the same as when the lamp was stationary, it was considered 
probable that at this same speed the candlepower also would return 
to the stationary value. This led to a series of simultaneous meas- 
urements of ctirrent and relative values of the candlepower, with 
voltage constant and speed varied over a wide range. 

Results obtained in this way with the 7 50- watt lamp operated 
at 90 volts are shown by the curves of Fig. 3. All values of current 
and candlepower, respectively, are expressed in percentages of 
the value of each obtained with the lamp stationary and the tip 
down. In determining the candlepower with the lamp stationary, 
readings were taken at intervals of 7.5° about the vertical axis, 
there being a total of 48 determinations. 

The curves for the lamp with the tip down are the result of 
two independent series of measurements made on different days, 



596 



Bulletin of the Bureau of Standards 



[Vol. 12 



values obtained in the two series being represented by crosses 
and circles, respectivel3^ The agreement of these two series 
shows that current and candlepower values are reproducible to 
a high degree of precision. 

The ratio of the ordinates to which these curves were drawn 
was purposely so chosen as to clearly represent to the eye the 
constant relation of candlepower to current throughout the 
whole range of speed employed. 





113 


s 








u 




S, 


108 


\ 




100,3 


104 


100.0 


100 


99.7 


96 




t 


99.4 




















99.1 










g 




o 


98.8 

















i 




1 

* Tip np 






















^ 




^ 
























^ 


^c^^ 


vet>o*^^ 






Down 














^ 


^ 


•?e 


,0..^ 






? "*^ 










^ 




^ 


^^ 




^ 




















Nftr- 




% 


V 






























S 


^ 


"^ 


. ^ 




























N 






^ 


? Tl 


p Dowi 


1 






















\ 


^^t 




























\ 


N 
































»■■■ Tip Up 

' 1 









30 40 60 80 100 120 140 160 180 200 
Pevolutiona Per Minute 



30 340 250 230 300 



Fig. 3. — Ratio of current and candlepower with lamp rotating tip up and also tip down 
to current and candlepower, respectively, with lamp stationary tip down 

IV. LAW OF THE VARIATIONS 

The results of these experiments may be summarized as fol- 
lows, in which the voltage is to be regarded as constant: 

1. With the lamp stationary, both current and candlepower 
have higher values when the lamp is mounted tip up than when 
it is mounted tip down. 

2. Rotation of the lamp causes both ciurent and candlepower* 
to change in value but in opposite directions, the change in both 
being about twice as great when the tip is up as when it is down. 



Middiekauff,-^ Photometry of the Gas-Filled Lamp 597 

3. As the speed is increased from zero the current first increases 
to a maximum and then returns to the value it has when the 
lamp is stationary and continues decreasing below this value as 
the speed is further increased. The candlepower varies in the 
opposite direction. 

4. For each position of the lamp there is a particular speed at 
which both current and candlepower have each the same value 
as when the lamp is stationary, the speed at which this occurs 
being practically the same for all lamps having the same number 
of loops in the filament. For lamps having different forms of 
filament mounting the speed for the above condition varies from 
lamp to lamp, being highest for those having the smallest number 
of loops. 

V. TWO PHOTOMETRIC METHODS 

1. ROTATING LAMP METHOD 

From a consideration of the above results it is evident that if 
the voltage be held constant and the lamp be rotated at the speed 
at which the current has the same value as when the lamp is sta- 
tionary an acciu"ate determination of mean horizontal candlepower 
may be made and the same efficiency rating given to the lamp as it 
would have if operated stationary at the same voltage. It is fortu- 
nate that in the position in which gas-filled lamps are most generally 
used in practice — that is, tip down — the speed for the above con- 
dition is somewhat higher and therefore more practical for most 
lamps than the speed for the same condition with the tip up. 

Although the above method has its limitations which are 
stated below, it fulfills all the requirements for determining the 
voltage-ciurent-candlepower characteristics of gas-filled lamps 
by the use of an ordinary bar photometer outfit. With a knowl- 
edge of the reduction factor, the watts per candle of the lamp 
may be determined for any given voltage, and then with the 
lamp stationary and with this voltage as a starting point the 
variations of current, candlepower, and watts per candle with 
voltage are as readily determined for a gas-filled lamp as for one 
of the vacuum type. In watts per candle determinations the 
factor expressing the relation of mean spherical to horizontal 



59^ Bulletin of the Bureau of Standards [Voi. 12 

candlepower in the given direction must, of course, be applied 
to the candlepower readings. 

For groiips of lamps having the same form of mounting, and 
therefore the same reduction factor, the method of rotation is 
perfectly satisfactory in determining relative candlepower and 
efficiency ratings, and practically the same speed may be used 
for all lamps of the group. However, as stated above, this 
method presupposes a knowledge of the reduction factor, which 
for gas -filled lamps in general, as now made in the present more 
or less experimental stage of manufacture, varies considerably 
from lamp to lamp, principally on account of the great variety 
of methods of mounting the filament. 

2. TOTAL FLUX METHOD 

During life tests of vacuum lamps the blackening of the bulb in 
all directions is about proportional to the distribution of the light, 
and hence a measure of the reduction in light during life may be 
obtained by observing the reduction in the mean horizontal candle- 
powder. In the gas -filled lamp, however, the greater proportion of 
the blackening occurs at the top of the bulb, the volatilized material 
thrown off from the filament being carried upward by the gas, and 
hence a true measure of the reduction in light during the life of 
the lamp can not be obtained by mean horizontal candlepower 
measurements alone but by determinations of the total flux or 
mean spherical candlepower. This can be most conveniently done 
by means of an integrating photometer, such as the Ulbricht sphere. 
The advantage of the sphere over other forms of integrating pho- 
tometers, such as the Matthews instrument, in measuring gas -filled 
lamps is that the lamps are measured while stationary and thus 
the complications arising from rotation are avoided. 

For the determination of the voltage-current-candlepower char- 
acteristics the sphere has no particular advantage over the method 
of rotation and has the possible disadvantage in that if the coating 
of the sphere is not perfectly nonselective the ratio of the candle- 
power at different voltages may be more or less affected by a 
change in the absorption of the light at the different colors. 



fkofa^'"'^'] Photometry of the Gas-Filled Lamp 599 

VI. POSSIBLE ERRORS INTRODUCED BY ROTATION 

The accompanying table shows the errors possible in efficiency 
determinations and the consequent errors in life values introduced 
by measuring candlepower and watts while the lamp is rotating. 
The values given are the result of photometric measurements made 
at different speeds on a 750-watt, 115-volt, nitrogen-filled lamp 
having nine anchor wires supporting the filament. 

Data on Errors in Efficiency and Life Values 



Position of lamp 



Speed in 
rpm 



Rated volts 
for 0.70 
watt per 
candle, 
lamp 
rotating 



Watts per 
candle at 
rated volts; 
lamp sta- 
tionary, tip 
down 



Per cent 

error in 

rated 



Per cent 

error in 

rated watts 

per candle 



Per cent 
life at rated 
volts; lamp 
stationary, 

tip down 



Tip down. 



Tip up. 




35 

70 
120 



115.0 
115.0 
113.5 
112.8 
109.9 

113.8 
113.8 
110.1 
107.4 
104.4 



0.700 
0.700 
0.718 
0.725 
0.760 

0.713 
0.713 

0.757 
0.793 
0.837 



0.0 
0.0 
1.3 
1.9 
4.4 

1.0 
1.0 

4.3 
6.6 
9.2 



0.0 
0.0 
2.6 
3.7 
8.6 

1.9 
1.9 
8.1 
13.3 
19.6 



100 
100 
121 
131 
184 

115 
115 
178 
252 
375 



On account of the increased efficiency of this lamp caused by 
rotation, for example, at 70 rpm tip down, it was found to have 
a specific consumption of 0.7 watt per candle at 113.5 volts 
instead of at 115 volts as found at zero speed with the tip down. 
If, therefore, the lamp had been photometered at the factory 
while rotating at this speed, it would have been rated at 113.5 
volts and would have been operated on the life rack at 0.718 
watt per candle instead of 0.7 watt per candle and would have 
had a life 21 per cent greater than would be expected from the 
rating. If, however, it had been photometered with the tip up 
at 180 rpm (as is sometimes done in practice), it would have 
been rated as having a specific consumation of 0.7 watt per 
candle at 104.4 volts. If put upon the life rack at this voltage, 
it would operate at 0.837 watt per candle instead of 0.7 watt per 
candle and give a life 275 per cent in excess of normal. In other 



6oo Bulletin of the Bureau of Standards [voi. 12 

words, a long-life performance is fictitious at the given specific 
consumption rating, or the consumption rating is fictitious for 
the given life. This holds true if the specific consumption is 
determined at any speed above that at which the current and 
candlepower have the same value as when the lamp is stationary 
with the tip down. For this lamp the particular speed is 35 
rpm. For speeds below this value with the tip down, the specific 
consumption rating would be too high and the life value would 
be lower than expected from the rating. 

VII. INVESTIGATION OF THE CAUSE OF THE VARIATIONS 

OBSERVED 

The following experiments were made with a view to finding 
the cause of the variations observed in this type of lamp: 

1. To determine the effect of increased external ventilation of 
the bulb caused by rotation, a draft of air was driven against a 
looo-watt lamp by means of an electric fan and a series of meas- 
urements of current at constant voltage were made at different 
speeds with " air on " and " air off. " It was found that at every 
speed the cooling effect of the air caused a small but measurable 
increase in the current but that this increase was very small in 
comparison with the decrease in ciurent caused by the rotation of 
the lamp. Hence, the decrease in current can not be attributed 
to an increase in external ventilation of the bulb by rotation. 

2. Another test was made to determine whether or not the 
earth 's magnetic field had a noticeable effect in producing a coun- 
ter electromotive force, thus changing the effective voltage on the 
lamp. This was done by pacing the lamp in the middle of a 
solenoid having a field strength of about 500 gausses, and the 
current in the lamp was observed when the current in the solenoid 
was thrown on and off, both with the lamp stationary and rotating. 
As no change was detected it was concluded that the much weaker 
field of the earth certainly did not produce a counter emf of suffi- 
cient magnitude to account for any part of the changes noted. 

3. To determine the effect of possible distortion of the filament 
upon the candlepower in the horizontal plane, candlepower meas- 
urements at various angles with the vertical were made upon the 



skSgiand!'^'\ Pkotometry of the Gas-Filled Lomp 60 1 

1 000- watt lamp (lamp A, figs, i and 2) rotating with the tip 
down, first at a speed of 45 rpm and then at a speed of 200 rpm. 
There was practically the same percentage increase in candle- 
power at every angle when the lamp was rotating at the higher 
speed — that is, the spherical reduction factor, which was about 
0.86, was not appreciably changed by rotation. Hence distortion 
of the filament was not the cause of the changes observed. 

4. A 1 00- watt vacuum tungsten lamp having a coiled filament 
similar to that of the gas-filled lamp, and similarly mounted in the 
bulb, when rotated exhibited none of the changes observed in the 
case of the gas-filled lamp. The same was found to be true for a 
500-watt vacuum tungsten lamp having an ordinary (not coiled) 
filament moimted in the usual manner. On the other hand an 
ordinary 250-watt lamp when filled with nitrogen exhibited all the 
variations of the coiled filament gas-filled, lamp, the ciurent at 
constant voltage returning to the stationary value at a speed of 
about 100 rpm. It was therefore concluded that the gas alone 
must be the disturbing element which causes the changes observed 
in the gas-filled lamp. 

VIII. A PROBABLE EXPLANATION OF THE CAUSE OF THE 
VARIATIONS OBSERVED 

The results of the above tests show quite conclusively that 
practically the whole of the effect observed in the change in cur- 
rent, candlepower, and specific consumption arises from a change 
in the convection currents of the gas within the bulb. At low 
speeds, as seen above, there is an increase in the current over the 
stationary value for all voltages, the greater percentage increase 
being at the low voltages. At high speeds there is a decrease in 
current as compared with the stationary value for all voltages, 
the greatest percentage decrease being also at low voltages. 

When the lamp is stationary there is a considerable difference in 
temperature between the gas at the top and bottom of the bulb. 
Very low speeds of rotation disturb the steady current of hot gas 
which is rising in the center and flowing down at a distance from 
the axis. The result is an increase in heat convection by the gas, 
and consequently a lowering of the temperature of the filament. 



6o2 Bulletin of the Bureau of Standards [Voi.iz 

As the filament has a positive temperature-resistance coefficient, 
the resistance decreases and consequently the current increases at 
constant voltage. At high values of voltage the mean temperature 
of the gas is higher than at low voltage, consequently the tempera- 
ture of the filament is reduced in a lesser degree by rotation and 
the change in current, though positive, is less at high than at low 
voltage. 

On the other hand, when the lamp is rotated at high speeds the 
cooler gas at the bottom of the bulb is thrown out by centrifugal 
force, which varies as the square of the speed, and the hot gas 
tends to remain near the center, thus considerably retarding the 
convection currents which cool the filament. The result is an 
increase in the resistance of the filament as a whole and a decrease 
in the value of the current. At high voltages the mean tempera- 
tinre of the gas is higher than at low voltage and the retarding 
effect due to centrifugal force is decreased; hence the change in 
current due to rotation is less than at low voltages. 

At low speeds, as seen above, the temperature of the filament 
is lowered and consequently the candlepower is decreased. At 
high speeds the temperature of the filament is raised, and hence 
there is an increase in the candlepower. 

IX. SUMMARY 

The new high efficiency gas-filled lamp introduces variables not 
hitherto encountered in the photometry of incandescent electric 
lamps. On account of the comparative broadness of the filament 
spiral and the dissymmetry of the filament mounting, there is 
considerable irregularity in the distribution of the light about 
the vertical axis. Consequently, when the lamp is rotated, as is 
commonly done in rating lamps at the factory, the light as seen 
in the photometer flickers so excessively as to render accurate 
measurements of candlepower practically impossible without the 
use of auxiliary apparatus. However, as is sometimes done, if 
two mirrors inclined to each other be placed back of the lamp, 
the flickering is so much reduced as to permit accurate candle- 
power measurements even at very low speeds of rotation. 



¥k^gianr^'] Pkotometry of the Gas- Filled Lamp 603 

But this expedient does not eliminate the most serious trouble 
caused by rotation. It was found that at constant voltage both 
the current consumed and the candlepower are different when the 
lamp is rotating than when it is stationary, the current changing 
in one direction and the candlepower always in the opposite 
direction; that is, there is a change in the operating efficiency of 
the lamp. Furthermore, this change in efficiency may be either 
positive or negative, depending upon the speed, and it is about 
twice as great when the lamp is rotating tip up as when it is 
rotating tip down. 

Fortunately, from the standpoint of photometry, there is for 
each lamp in either position a particular speed at which the current 
and the candlepower have the same values, respectively, as when 
the lamp is stationary. Hence, with the lamp rotating at this 
speed its candlepower can be measured with accuracy in spite of 
its rotation. Th^ speed for the above condition is practically the 
same for all lamps having the same number of loops in the fila- 
ment; but for lamps having different forms of filament mounting 
it varies from lamp to lamp, being greatest for those having the 
smallest number of loops in the filament. 

If the above precaution as to speed adjustment is not observed 
and lamps are r^^ted while rotating at speeds ordinarily used in pho- 
tometering vacuum lamps, the errors which enter may amount to as 
much as I to 2 per cent in current, or watts, in one direction, and as 
much as 1 5 to 20 per cent in candlepower in the opposite direction. 
Hence the voltage f oimd for a desired operating efficiency may be 
so much in error as to give a lamp on test at this rated voltage a 
fictitious life value three or four times as large as the lamp would 
give if it were operated stationary at a voltage corresponding to 
that efficiency which during the rating was only apparent. That 
is, the lamp maybe given credit for a much longer life than it really 
deserves. On the other hand, the speed may be such as to cause 
errors in the opposite direction resulting in a lamp life much 
shorter than would be expected from the apparent efficiency 
rating. 

Another peculiarity of the gas-filled lamp is that while it burns 
the blackening occurs, not all over the bulb in approximate pro- 



6o4 Bulletin of the Bureau of Standards [Voi. 12] 

portion to the light distribution as in the vacuum lamp, but princi- 
pally at the top of the bulb because the volatilized material is carried 
upward by the gas. Hence in making a life test a true measure 
of the reduction in total light diu-ing the life of the lamp can 
not be obtained, in the usual manner, by mean horizontal can- 
dlepower measurements, but by determinations of the total flux 
or mean spherical candlepower. This is accomplished most 
rapidly and conveniently by means of an integrating photometer, 
such as the Ulbricht sphere, in which the lamp is measured sta- 
tionary, and thus all the complications arising from rotation are 
entirely avoided. 

As to the cause of the variations observed in candlepower and 
efficiency when the lamp is rotated, it is concluded from the results 
of a number of special tests that the whole effect is produced by a 
change in the convection currents of the gas, a consequent varia- 
tion in the temperature distribution in the bulb, resulting in a 
change in the resistance, and therefore a variation in the current 
and candlepower of the lamp. 

The authors acknowledge their indebtedness to Dr. E. B. Rosa 
for his kindly interest and valuable suggestions, and to H. B. 
Sinelnick for efficient assistance in the laboratory work. 

Washington, January 25, 191 5.