voltage and several readings are taken. The lamps to be standardized are then mounted successively on the same carriage and readings are made. At the end the standard is again measured. If the lamps that are compared with the standard are of the same nominal candle-power as the standard, the left-hand carriage is kept at the same position on the bar. If the candle-power is different, the left-hand carriage is placed so that the balance comes at about the same place on the bar. This method of measurement, which is a substitution method, has many advantages in accurate work. It eliminates all dissymmetry on the two sides of the photometer, and thus avoids the necessity of reversing the photometer screen. In making the measurements of the effect of reflection from the walls the right side of the photometer was entirely protected from reflected light from the walls by large screens, which nearly surrounded that half of the photometer bar, so that when the left side of the Lummer-Brodhun sight-box was closed by a shutter no illumination was visible in the eyepiece. The velvet-covered diaphragms and the lamp on the left side of the photometer, on the contrary, were left exposed to the reflected light from the walls, as in the ordinary method of measurement, except that the lamp on that side was not burning. Therefore the only light that penetrated to the screen was that which, reflected from the walls, fell upon the lamp and black velvet diaphragms and was reflected by them to the screen. In order that the walls should be illuminated to the same extent as in the ordinary measurements, a 16 c. p. lamp was burned near the position of the comparison lamp, and another lamp was burned near the position of the standard lamp, but so shielded in each case that no direct light could reach the photometer. The lamp on the left near the unburned standard was changed from time to time from a 32 c. p. lamp to a 16 c. p. lamp, depending upon whether a 32 c. p. unburned lamp was in the left socket at the distance from the screen at which a 32 c. p. lamp is usually placed or a 16 c. p. lamp was in the socket at the position used with a 16 c. p. lamp. These precautions were taken in order to secure as nearly as possible working conditions. Under these circumstances it was perfectly plain, on looking into the eyepiece of the photometer, that the left side of the sight-field was slightly illuminated, and it only remained to measure this illumination under the different conditions likely to occur. To this end a 1 c. p. lamp, which had been placed on the carriage on the right in the position' of the comparison lamp, was brought to incandescence, and the light from this lamp was cut down by absorption strips of smoked glass having known coefficients of transmission. Thus successive absorption strips were added until a balance was obtained, and the number of strips and the positions of the three carriages were recorded. This method of procedure was followed through a great number of measurements, the conditions being changed from time to time until all the conditions that are actually encountered were included in order to determine whether any error due to the walls could possibly be introduced in any comparisons that are ever made. What was actually measured each time of course was the illumination of the screen due to the reflected light, and it was merely a matter of computation to determine what the candle-power at any distance would be in order to produce the illumination found. I shall not give the results of all the measurements that were made, but merely indicate the results for two typical cases which show clearly the order of magnitude of the effect. First, from the direct measurement of the illumination that was produced when a 32 c. p. standard (not burning) was placed at the position at which a 32 c. p. lamp is generally placed in standard measurements, and another 32 c. p. lamp was illuminating the walls, calculation showed that the illumination on the screen, due to the reflection from the walls, was that which would be produced by a source of 0.003 candle at the position of the 32 c. p. standard. In a similar way the correction to a 16 c. p. lamp was found to be about 0.001 or 0.002 candle. Thus we see that not only would the relative error in comparing a 16 c. p. lamp with a 32 c. p. lamp be negligible, but that the absolute error for either one would be negligible in the most refined measurements, being of the order of magnitude of 1 part in 10,000. To be sure, the determination of this correction may be considerably in error, but since the correction is so very small it could be increased to several times its value and still be beyond recognition. All of these measurements were made with the windows covered with heavy curtains, the only sources of illumination of the walls being lamps similar to ones that would ordinarily illuminate the walls when making photometric measurements. When the curtains were raised the effect was greatly increased, so that the error in a 32 c. p. lamp was several tenths of a candle, varying with the brightness of the sky and with the distribution of objects about the room. Approximately the same relative error was found for the 16 c. p. lamp, so that in comparative measurements the errors in general would be negligible; but since the errors of measurement of these corrections were necessarily large, and since we found the corrections to vary considerably when an observer moved in front of a window, it is desirable to exclude daylight by curtains. In conclusion, I desire to express my indebtedness to Mr. F. E. Cady for valuable assistance in making the observations. INFLUENCE OF WAVE FORM ON THE RATE OF INTEGRATING INDUCTION METERS. By E. B. ROSA, M. G. Lloyd, and C. E. REID. We give in this paper the results obtained with five integrating induction wattmeters, on which we have made a large number of tests, although further work remains to be done. These results may therefore be regarded as preliminary, illustrating the methods employed and the results obtained when changes are made in the wave form by altering the magnitude or phase of the harmonics present. Two of the meters employed were sent to the Bureau of Standards for test by the makers. The others were meters which we happened to have in the laboratory when the tests were undertaken. The following is a list of the meters: No. 1, Stanley (magnetic suspension type), 50 amperes. No. 4, General Electric (1902 House type), 25 amperes. No. 5, Siemens & Halske, 25 amperes. All the meters are made for 60 cycles, single phase. The first four are American instruments; the last is of German make. Each meter was tested at full load and at 110 volts, and at approximately unity power factor. In order to determine the effect on the rate of an induction meter due to varying the wave form, it is necessary to eliminate carefully any effects due to variation in the temperature of the meter or changes in the frequency of current, or other alterations in the conditions of the meter or circuit. In most cases the effect of a moderate distortion of the wave is small, and unless all measurements are made with great care the effects looked for may be masked by other effects or by errors of measurement. The meters were tested alternately with current of sine wave form and with a distorted wave, the distortion being produced by adding a harmonic of three times the frequency FIG. 1.-Showing the resultant of combining the fundamental and harmonic of three times the frequency and 25 per cent of the magnitude of the fundamental, giving first a peaked wave and second (when the phase of the harmonic is reversed) a flat or dimpled wave. and harmonic are of sine-wave form. Both fundamental FIG. 2.-Showing the resultant of a fundamental and a harmonic, as in fig. 1, except that the phase of the harmonic has been shifted 30° by changing the coupling 5o. FIG. 3. Showing the resultant of a fundamental and a harmonic as in fig. 1, except that the phase of the harmonic has been shifted 60°. FIG. 4. Showing the resultant of a fundamental and a harmonic as in fig. 1, except that the phase of the harmonic has been shifted 90° by changing the coupling 15°. The wave form for "peak" and "flat" are here alike, except that the steeper side is in advance in the "peak" and the more gradual slope is in advance on the "flat." "Peak" and "flat" are conventional terms, indicating the phase of the harmonic. If the coupling were shifted 15° more, the "flat" curve would become peaked. |