Considering the double-window sector-disc of Figure 3, it was found that whether the windows were opened to their full angular width of 90 degrees, or were nearly closed, the vanishingflicker speeds and frequencies for given maximum cyclic illumination on the target were substantially the same. In the former case there would of course be equal successive intervals of light and shadow; while in the latter, there would be shorter intervals of light followed by longer intervals of shadow. The mean illumination would be proportional to the angular aperture of the window. Consequently, the vanishing-flicker frequency

FIG. 5-CURVES OF FREQUENCIES AT WHICH FLICKERING, IN DIFFERENT INTENSITIES OF ILLUMINATION, APPEARED TO VANISH

does not seem to depend upon the mean illumination of the target nor upon the shape of the wave of cyclic illumination, but only upon the maximum and minimum illumination; or the limits of illumination in each cycle. At frequencies above the vanishing-flicker frequency, or with steady apparent illumination, the latter was found to indicate the true average value in the photometer, within limits of observational error; or to follow Talbot's law of sector-disc light reduction.*

* Talbot's Law as Applied to the Rotating Sectored Disc, by E. P. Hyde, Bulletin of Bureau of Standards, Vol. II, No. 1, 1906.

All of the observations described in this paper were made with the target stationary and with the observer's eye at rest; that is, with the image on the retina fixed in position. If the target was set in motion, the flickering at once became more noticeable. When the vanishing-flicker frequency was reached, it was usually possible to restore the flicker temporarily by directing the eye quickly from point to point over the target. If the target had been in motion, the speeds of sector-disc rotation necessary for vanishing flicker would have greatly increased.

Measurements were then made with a reduced range of flicker. The range of flicker may be defined as the ratio of the difference between the cyclic maximum and minimum illuminations upon the target, to the maximum cyclic illumination. In all of the above-mentioned observations, the range of flicker was 100 per cent; or the illumination on the target was alternately maximum and zero, with each window and blank of the rotating sector-disc, respectively. In the case of the ordinary alternatingcurrent carbon-arc lamps, the range of flicker is known to be large. With incandescent lamps operating on alternating-current circuits, the range of flicker depends upon the thermal capacity of the filament; that is, upon the specific heat of the filament and its mass per unit of surface. For a given alternating-current frequency within the flicker range, such as 25 cycles of current per second, giving rise to 50 cycles of flicker per second, the range of flicker is small for low-voltage, large candle-power, coarsefilament lamps, but may be large for high-voltage, small-candlepower, fine-filament lamps. According to the observations of Mr. J. T. Morris,* in a 220-volt, 5-cp carbon-incandescent lamp, with a filament of no doubt very small cross-section, the range of flicker was 63.9 per cent of the mean illumination, which corresponds to 51.5 per cent of the maximum, as here defined.

In order to reduce the range of flicker, the target was illuminated by two stationary incandescent lamps, side by side. One

* Experiments on Carbon, Osmium and Tantalum Lamps, by J. T. Morris, The Electrician, Vol. LVIII, No. 1491, p. 318, Dec. 14, 1906. His paper quotes a research by MM. Girard and Magnol in Bulletin Société Internationale des Electriciens, Vol. VI, pp. 29-32, as stating that the ratios of difference between maximum and minimum candle-power to mean candle-power of carbon incandescent lamps on 25-cycle circuits were observed to be 15 per cent with a 110-volt, 32-cp lamp; 20 per cent with a 110-volt, 10-cp lamp. and 53 per cent with a 110-volt, 5-cp lamp.

was placed behind the rotating sector-disc so as to be cyclically obscured thereby. The other was left unobscured. By varying the sizes or candle-powers of these two lamps, the range of flicker. could be readily adjusted for different series of tests. The results are indicated in Figure 5. Curve D corresponds to 44 per cent flicker (or 56 per cent of the total illumination kept steady), curve E to 33 per cent flicker, curve F to 7.5 per cent flicker and curve G to 3.3 per cent flicker. With ranges of less than 7.5 per cent, the flicker ceased to be disagreeable to the eye, especially as the frequency increased. The abscissæ of these curves are the various maximum illuminations normally incident upon. the target. The observations were difficult to make in the case of the 3.3 per cent range of flicker, and contained considerable discrepancies between the estimates of the three different observers.

The lowest range of flicker which could be recognized with certainty was 1.4 per cent, and the most sensitive flicker frequency for making it apparent was a low frequency in the neighborhood of 2.5 cycles per second. This is not far from the flicker frequency in the usual observations of mean horizontal candlepower with an incandescent lamp rotating about a vertical axis at 180 revolutions per minute, or 3 revolutions per second, and ordinarily 2 cycles of flicker per revolution; that is, 6 flicker cycles per second.

The curves D, E, F, G show that the vanishing-flicker frequency increases with the range of flicker, but in nothing like the same proportion.

The highest vanishing-flicker frequency that could be produced was found when looking directly at a 75-cp incandescing lamp through the rotating sector-disc at a distance from eye to filament of about 50 centimetres. This frequency had a mean. value of 66 cycles per second, with 100 per cent range. It was substantially the same for any angular aperture of the sector windows that permitted all of the filament to be seen at once.

Experiments with the open arc between vertical carbon electrodes and with alternating-current supply, showed that flickering, on stationary targets with stationary eye, did not entirely disappear until the alternating-current frequency was 60 cycles per second. At this frequency the flicker frequency of the arc column

would be 120 cycles per second, which is far in excess of any vanishing-flicker frequency with sector-disc interruption of incandescent lamps, as found in these tests. The anode glow will, however, only have a flicker frequency of 60 cycles per second, or that of the supply current; since the anode will alternately change from one carbon to the other once in each alternatingcurrent cycle. It would seem probable, therefore, that the light from an alternating-current arc lamp possesses two flicker frequencies; namely, one equal to that of the current and affecting the light emitted by the tips of the electrodes, and the other of twice this frequency and affecting the light emitted by the vapor of the arc, or arc stream. The sensation of flicker due to a stationary illumination-image on the retina might then be expected not to disappear entirely until both flickers exceeded the vanishing limit. If this reasoning is correct, the possibility is suggested of having less visible flickering with flaming arcs than with ordinary enclosed arcs at low alternating-current frequencies; because in the flaming arc, the light emitted from the electrodes is relatively so much reduced with respect to the light emitted by the arc stream. This question has not, however, been investigated in the experiments here reported.

The authors desire to acknowledge the assistance of Mr. S. R. Crosse in the observations here reported.

In conclusion, the following deductions appear to be warranted by the observations:

I. The frequency of flicker at which flicker ceases to be visible is approximately the same for different observers with normal sight.

2. The maximum frequency of vanishing flicker with stationary retinal image (stationary target and stationary eye) is in the neighborhood of 66 cycles per second.

3. Vanishing-flicker frequency increases in all cases when the illumination on the target is increased, but in nothing like the same proportion. The increase of vanishing-flicker frequency is rapid for increasing illuminations below 0.5 foot-candle and 100 per cent range; and increases but slowly for illuminations exceeding I foot-candle.

*Lombardi and Melazzo, Stroboscopic Observations on the Alternating-Current Arc. Transactions of the International Electrical Congress, St. Louis, 1954, Vol. II, Figure 2, opposite p. 805.

4. Vanishing-flicker frequencies are less with colored targets than with white targets, for equal illuminations and sizes of retinal image.

5. The vanishing-flicker frequency does not depend upon the mean illumination on the target; or at least only to a relatively small degree. It depends on the maximum and minimum cyclic illuminations.

6. The vanishing-flicker frequency does not depend appreciably upon the wave-shape of flicker; that is, upon the manner in which the illumination varies in passing between the maximum and minimum cyclic values.

7. The vanishing-flicker frequency increases somewhat with the area of the target, for a given surface quality of the latter, distance from the eye, and incident illumination.

8. The vanishing-flicker frequency increases somewhat as the target is approached to the eye, for a given size of target and intensity of incident illumination.

9. The last two foregoing deductions may be jointly expressed by saying that the larger the area of retina stimulated. by flicker, the higher is the vanishing-flicker frequency; but in nothing like the same proportion.

IO. The greater the range of flicker, i. e., the ratio of difference in cyclic illumination to the maximum, the greater the vanishing-flicker frequency; but in nothing like the same proportion.

II. The smallest range of flicker that was found to be recognizable with certainty was 1.4 per cent, observable only at a low frequency.

12. The most sensitive flicker frequency for small ranges of flicker was in the neighborhood of 2.5 cycles per second. 13. Flickering ceased to be objectionable with ranges less than 7.5 per cent.

14. The conditions favoring disagreeable flickering with stationary retinal images (fixed target and fixed eye) are powerful illuminations, large flicker ranges, bright surfaces of large area, and low flicker frequency.

15. The conditions tending to produce unobjectionable flickering with stationary retinal images, are feeble illuminations, small ranges of flicker, small targets and dark colors of reflecting surfaces, with high frequencies of flicker.