Requirements to be Met.-The requisites for an automatic flush-tank are: 1, Certainty of action; 2, rapidity of discharge; 3, simplicity of construction; 4, ease of inspection of all its parts; 5, durability; 6, economy of cost and maintenance.

Strange as it may appear, there are flush-tank syphons, sold in considerable numbers, which cannot possibly be made to work under the usual conditions imposed by the requirements for flushing sewers.

Rapidity of discharge is next in importance to certainty of action. The sewer pipe should be filled for some distance in order to get the proper benefit from the flush.

In simplicity of construction the syphon tanks are superior to the valve tanks, and as durability is likely to depend upon simplicity of construction, the syphon tanks will, in general, be most durable.

Complicated mechanism is undesirable for use in a flush-tank, which must work automatically, and often for a long time without inspection. It is not at all uncommon to find that devices which look well on paper fail utterly when put to the test of actual service.

Quantity of Water Required.-An erroneous idea prevails as to the quantity of water required for flushing sewers by the use of automatic flush tanks. A properly designed system for a city of 10,000 inhabitants ordinarily requires from twenty to fifty flush-tanks, of a capacity of about 150 to 300 gallons, discharging daily, or at most twice a day. The maximum amount of water required is about two per cent. of the water supply. This momentary discharge does not sensibly occupy the capacity of the main sewers further down the line, being, as before stated, but a very small percentage of the ultimate discharge. An equally efficient flushing by a constant stream, applied directly and without the intervention of a flush tank would require an amount of

water materially encroaching upon the capacity of the main sewers, and would be inadmissible under ordinary conditions of water supply, on the score of economy.

Rapidity of Discharge.-Flush-tanks are ordinarily adjusted to discharge automatically once in each twenty-four hours. Their capacity of discharge should equal or exceed that of the pipe into which they empty.

Experiments made by the writer with a flush-tank of 7,000 gallons capacity, having a ten-inch outlet, opening into an eight-inch sewer, demonstrated that with the minimum grades indicated in Table XV, there was no danger of gorging the sewer at a distance of one or two hundred feet from the flush-tank, although the hydraulic head was seven feet and the capacity of the tank was sufficient to fill the sewer for a distance of 2,682 feet. At a distance of 600 feet the flow, as observed in a man-hole, did not fill the sewer.

In the case of flush tanks as ordinarily constructed the tank can hardly discharge too rapidly.

The rate of discharge from the tank should at least equal the capacity of the sewer when the flow has acquired the velocity due to its inclination. A sewer six inches in diameter, laid at a grade of five-tenths per hundred, discharges, when full, at the rate of 215 gallons per minute. The conditions above named would therefore require the tank to discharge at the rate of 100 gallons in 28 seconds or less.

Experimental Data. Some valuable experiments were made upon the effect of flush tanks in the sewers in Washington, D. C., by Asa E. Phillips, Assistant Engineer, District of Columbia, and the results were presented by him in a paper read before the American Society Municipal Improvements, in October, 1898.

The following extracts are taken from his paper:

"No formula has been proposed for the volume of water required for different grades and sizes, and the only rule known to have been used appears to be of little value. This important detail is determined by individual judgment, generally unsupported by investigation or experience, so that the common practice has varied within a large range of values, while tanks of uniform size usually have been constructed regardless of differences in the size or gradient of the sewers to be flushed. The uncertainty as to the precise effect of the flush and the complex conditions as to contributing population, rate of water consumption, etc., have been justly considered a bar to any precision in this respect. Recent discussion of the subject, however, has tended to establish certain limitations in the use of flushing devices which should lead to improvement in the general practice.

"The work to be accomplished by the flush is the removal at regular and frequent intervals of solid matter flushed into the sewer from house laterals, and there stranded because of the shallow depth of flow and sluggish current and its carriage down the line to a point where the depth and velocity are sufficient to insure removal to the ultimate point of discharge. The efficiency of the flushing device in performing this work is not well understood. But little is known of the effect under the varying conditions encountered, especially for widely different grades and at considerable distance from the dead end. It is generally considered, however, that the effect diminishes very rapidly as the distance increases, and becomes almost imperceptible 600 or 700 feet from the tank, but, so far as can be ascertained, this has been observed only in cases of flush of small volume on flat grades, or where the depth of ordinary flow was considerable. Of the effect of discharge of 600 gallons or more, such as were used in the cases to follow, there appears to be no recorded observations, so that no comparison with those already published can be made. Several grade conditions have been selected for the purpose of illustrating the effect of the flush under such circumstances, and an attempt made to indicate by means of diagrams the different results obtained. For these no special accuracy is claimed, but in the few cases given the differences are sufficiently marked as to suggest certain conclusions therefrom.

"The Park street line is the first of these. This sewer is 12 inches in diameter, about 1,870 feet in length, and has a uniform grade throughout of 9 inches per 100 feet. Preliminary examination discovered slightly unfavorable conditions for experimental work, such as an uneven grade, rough joints, and poor alignment in places, all of which affected the observations and results to some extent. It may now be stated that, with the exception of the Chapin street line, the sewers cited are old and generally possess these irregularities, but, excepting slight silt accumulation at points distant from the basin, they were found to be very clean. The presence of even a small amount of silt in

the invert, particularly at man-holes, undoubtedly affected the flow and was the chief source of error in the observations.

"The tank on Park street was found to have an effective capacity of 84 cubic feet, or about 630 gallons, and discharged through an 8-inch syphon in the mean time of forty-two seconds. No attempt was made to determine the velocity at the point of discharge, but this data would indicate an approximate mean velocity of 6 feet per second. Observations of the flush were simultaneously taken at all of the man-holes, and the depths of flow were recorded at intervals of fifteen seconds or less. These depths were referred to the time of syphon discharge, and the diagrams of the flush have been constructed from this datum.

"The first diagram, (Fig. 1, Plate XXVI), shows the form of the flush wave as taken at five consecutive man-holes within 1000 feet of the tank. The lower four man-holes were not platted because of the confusion of lines that would result, but the data are given in Table XXII. This diagram and tabulation show how well the depth of flow is maintained for very long distances. One thousand feet from the dead end the flush is very efficient, and at a distance of nearly 2,000 feet appears to be quite effective. This large radius of effect is doubtless due to the volume of water used, as published data for smaller discharges indicate that a tank of half this capacity would have a very greatly diminished influence.

TABLE XXII.-Park Street Sewer.

CAPACITY OF TANK, 84 cubic feet=630 GAL.; TIME OF DISCHARGE, 42 SEC.

"The second diagram for this line, (Fig. 2, Plate XXVI), shows the computed velocity curve for the ordinary flow, the assumed velocity to prevent sedimentation and the accelerated velocity due to the flush. These curves are not

supposed to be precise, but they illustrate the purpose of the flushing device, and to some extent the degrees of effectiveness required. The curve of normal flow shows the very low velocity along the upper portion of the line, and its gradual increase approaching the required velocity at the lower end, while the flush curve shows a corresponding high velocity at the upper end and its rate of fall toward the 21⁄2-foot-per-second velocity, where in theory at least it would seem that the two curves meet at a common tangent point. In this sewer, it may be noted, the normal flow does not attain a rate of 22 feet per second, and probably would not for a distance of 2,000 feet. It is also to be observed, however, that the flush would seem to maintain this velocity for a distance probably as great; so that, so far as these observations go, they indicate that for this very long line and flat grade the flush tank is efficient and of the proper size.

"The second series, taken on one of the Connecticut avenue sewers, is chiefly interesting as showing the effect produced on a varying and decreasing gradient.

TABLE XXIII.-Connecticut Avenue Sewer.

CAPACITY OF TANK, 82 CUBIC FEET-615 GAL.; TIME OF DISCHARGE, 45 SEC.

"The diagram, (Fig. 3, Plate XXVI), illustrates the diminished velocity and enlarged area of section of the flush wave, as recorded at man-hole No. 3, showing the marked effect of the diminished rate of grade in the portion of the sewer immediately above. Unfortunately this could not be further observed, owing to a change in gradient from this point. It probably indicates, however, that for a line of varying slope the minimum should be considered in fixing the capacity of the tank.

"The third series was taken on the Chapin street sewer, which has the reversed conditions of a varying but rapidly increasing gradient. The observed effect of the flush is very clearly shown by the diagram, (Fig. 4, Plate XXVI). The rapid run-off and greatly reduced area of section toward the lower end of the sewer indicate the very high velocity which such steep slopes must produce. In this case the minimum grade being 1 per cent., and that for a length of only