obtained by drilling a hole through the center of a hand-hole cover and adding a small stuffing box made from the gland of a valve. In the operation of this instrument the flow of water in any tube of the boiler in which the circulation indicator might be placed causes the wheel to rotate at a rate of speed proportional to the rate of flow. By placing the receiver to the ear a click is heard for each revolution, and one revolution of the indicator means the passage of approximately 1 foot of water. As the speed of rotation in all trials up to date has not been too high for an observer to count the clicks, no difficulty was encountered in keeping a record of the rate of flow in any boiler tube under observation under varying conditions of operation of the boiler. Several instruments were built and tried before success was attained with this one. The instrument has lately been used to work an automatic counter through a couple of telegraph relays. FIG. 75.-Result of observations with circulation counter (propeller in back of boiler, middle tube, third row from bottom); test 379. Revolutions per minute calculated from averages of two readings of revolutions per fifteen seconds. For information obtained see figs. 7 (p. 16) and 75 and pages 163 and 164. Fig. 75 shows the effect of cleaning fires, and of firing, on the speed of water circulation in a tube of the boiler, as measured relatively by the circulation indicator illustrated in fig. 74 and described on pages 160-2. The circulation is prompt in its changes and the values obtained vary considerably. The readings were taken by recording the number of revolutions in a fifteen-second interval opposite time figures, and from such data the temporary rate in revolutions per minute was calculated and plotted on this chart. The curves of fig. 7 (p. 16) are based on readings taken for several days with the circulation indicator. The data from which the two curves were plotted were obtained by counting the total number of revolutions of the indicator for each half-hourly period, and by calculating the percentage of builder's rated horsepower of boiler developed for each such period, respectively. The small circles give the positions of points and the numbers in the large circles near by give the number of half-hours fulfilling the coordinate values of the points. After plotting, the points were averaged in value in both horizontal and vertical strips, each point being included for averaging as many times as indicated by the number in the circle near by. Thus the two curves were determined. It will be noticed that they are very close together, indicating the reliability of the method of working up the data. The important point is that the circulation rapidly drops behind the amount of steam made (per cent of rated capacity developed), especially at high rates of working. Thus at 70 per cent of rated capacity the average speed of rotation of the indicator was 80 revolutions per minute. At 105 per cent of rated capacity the rate of revolution was 102, whereas to be proportional it should have been 120; the speed of circulation fell about 15 per cent short.. This result is reasonable when we consider that, so far as one can make any speculations, the circulating forces are perhaps roughly proportional to the amount of steam which is generated and entrained with the rising water, whereas the frictional resistance to circulation is perhaps proportional to the square of the average velocity of circulation. This failure of circulation to keep up proportionally with demands on it must decrease the efficiency of the boiler at higher rates of working, by allowing a proportionally larger percentage of the water-heating surface to be covered with steam bubbles, thus virtually reducing the heating surface. That this condition does supervene is indicated on a number of charts, for instance those shown in figs. 5 (p. 14) and 8 (p. 17), which should be noted in this connection. At a later date the circulation indicator was put in the middle tube of the lowest row of tubes, this being one of the tubes inclosed in clay tiles except at the rear end. The revolutions per minute for various capacities are given in the following table: Readings of circulation indicator showing revolutions per minute for various capacities. The number of revolutions at any capacity is approximately three times as great as the number shown in fig. 7 for the second row of tubes just above. In an earlier experiment the same circulation indicator was placed in the third row of tubes from the top of the boiler, and it was found that the rate of revolution was very slow indeed. This result indicates that the bottom row of tubes is doing far more work than any other row, and that as we go from the bottom row up the amount of work done decreases very rapidly. The probability that the bottom row of tubes absorbs so large a portion of the total heat, absorbed mostly on account of conduction through the clay tiles and radiation to the exposed portion of the tubes in the rear over the hot brickwork, makes it easy to realize that the efficiency of the boiler as a heat absorber may well rise far more rapidly with increasing furnace temperature than is indicated by the equation for heat absorption from the gases due to convection only, as developed on pages 129 and 130. C-SHAPED v. FLAT-BOTTOMED TUBE TILES. In boilers of the Heine furnace type the bottom row of tubes is incased in clay tiles. Such an installation is advantageous when coals break down, evolving gases difficult to burn, for the clay tiles serve a far better cause in protecting the burning gases from sudden cooling, and in acting as a small heat reservoir, than in protecting the tubes from burning. The first tiles used by the testing plant were of a C shape inside and out, and although they were made of excellent material they were so easily damaged by fire tools, and cracked so often from sudden temperature changes, that a new set had to be put in every two or three months. Later, both boilers were fitted with tiles which are flat on the bottom, giving the furnace roof the appearance of a ceiling. They are only 15 per cent heavier than the old C tiles, and could easily be made as light as the old tiles by cutting off the upper outside corners, above the tubes. The new tiles have withstood hot fires for over a year and are still good. They are among the most satisfactory parts of the boiler equipment. The difficulty referred to has been experienced at several power houses, and overcome in the same way, with equal satisfaction. ORSAT TOTALS. The sum of the percentages of CO2, O2, and CO of an Orsat gas analysis is not at all constant. The sum is usually low at the start of a test and gradually rises during the first two hours. It also varies from day to day and from coal to coal. The first explanation suggested was that the available hydrogen of the coal burned to water that condensed before the gas sample arrived at the Orsat measuring tube, so that out of every hundred volumes of air which entered the furnace one or two volumes might well be missing in the Orsat measuring tube. On this assumption, when the same quantity of air is used to burn various fuels those highest in available hydrogen should give the lowest totals of Orsat analyses. To test this reasoning, a tabulation was made, headed "Classification of average flue-gas totals on basis of per cent of available hydrogen in dry coal." Many tests were grouped, first according to pounds of dry chimney gases per pound of "combustible." The tests of each of these groups were then reclassified according to per cent of available hydrogen in dry coal. For each subgroup the average was found for the Orsat totals of CO2, O2, and CO. A glance along the horizontal rows of flue-gas totals shows but little relation between available hydrogen and flue-gas totals. This failure to connect per cent of available hydrogen with shortage in Orsat totals suggests that other be at work. causes may Perhaps the occurrence of low totals in the first part of a test is due to the fact that an oxygen molecule may combine with carbon and form two CO molecules, which will occupy twice the space, thereby making the Orsat totals higher. During the first part of a test the CO is low, and later it rises. CO and available hydrogen have opposite effects on the totals. So it is, after all, not to be expected that either taken alone will show anything. More work will be done on this problem. Classification of average flue-gas totals on basis of per cent of available hydrogen in dry coal. Classification of average flue-gas totals on basis of per cent of available hydrogen in dry coal-Continued. PER CENT OF CO IN COMBUSTION CHAMBER. Fig. 14 (p. 23) shows that when combustion-chamber temperatures are high, the per cent of CO in the rear of the combustion chamber is high and the totals of CO2, O2, and CO are low. This concurrence of high CO and low totals of Orsat analyses has often been noticed. As a converse test, the classification tabulated below was made on per cent of CO as a basis, to obtain the average of the Orsat totals. They are practically constant until the CO becomes high, when they drop at a comparatively rapid rate. This relation is of the same nature as that indicated in fig. 14 (p. 23). No explanation is given here, as several possible ones were found to be doubtful on investigation. Classification of Orsat totals on basis of volumetric per cent of CO in combustion chamber (tests 318-382). Per cent of CO. 0 to 0.10 to 0.20 to 0.30 to 0.40 to 0.50 to 0.60 to 0.70 to 0.80 to 0.90 to Over The following table was constructed to ascertain whether such conditions of poor combustion as permitted a high percentage of CO in |