In connection with the subject of the behaviour of principals under test load, some consideration may here be given to the disputed question as to how far keys and cotters, or screwed connections, are advantageous or otherwise to the braced structure.

Custom has, to a large extent, sanctioned the employment of keys, cotters, or coupling screws in the main ties of roof principals of all spans, apparently with the idea that such adjustments are desirable or necessary in the final regulation of the exact span between end connections or shoes.

It is perhaps doubtful whether the ordinary methods of careful setting out and good templet work do not, in the yard, secure an amount of practical accuracy sufficient for all such requirements, when the actual dimensions of the work to which the roof is to be attached have been, as they should be, accurately ascertained and rigidly adhered to in construction.

It may again be argued that the use of keys, cotters, or coupling screws may lead to considerable modifications in the stresses in the members thus connected, when the adjustable connection has been used in such a way as to over- or under- strain the members affected.

A case of this kind came under the writer's notice during the testing of a roof principal of the type shown in Fig. 345. Keyed and Cottered connections were used at the points A, B, C, D, E, F, G, and H.

Upon the application of the test load, which was suspended from the points J, K, L, M, N, 0, P, the results of observation showed evident traces of undue stress in the members PB and GJ, a considerable proportion of the tensile stress which should by theory have passed through the members AB and GH having been diverted through PB and GJ, with the result of deformation in the top boom at P and J, these bars having to be reset to their original curve, while some elongation of rivet holes in connections at P and J showed the direction and intensity of stress.

It was considered these results were probably due to an in-u dicious use of the hammer by the workmen employed in erecting the principal, who found the keys and cotters, to their thinking, too loose.

Incidentally, the result appeared also to point to the desirability, in designing roof trusses of this type, of avoiding such an inclination of the end braces in relation to the main tie as would, by accidental misfit, lead to serious modification of the calculated stresses.

In concluding this chapter, some brief reference may be made to the subject of setting-out roof principals of considerable span, and for illustration the type of truss shown in Fig. 345, and already referred to, may be taken.

This truss was used to cover an area unequally divided into two spans, portion of which was rectangular, and the remainder on a curve.

The actual distance between the seatings of the principals on the bed-stone course of the main walls of the building was ascertained trigonometrically by the use of a theodolite having a metal wall-tripod, and from a measured base laid out on the top of the wall itself. By these means the exact value of all the varying spans was obtained from the masonry work as built.

The upper member of the truss was a segment of a circle, and the centres of eye-bar pins in the tension member, or main tie, were set out also on a circular curve of larger radius.

Fig. 345.

The dividing up of the upper and lower chords having been determined, the exact lengths of the braces were calculated trigono-metrically.

The entire truss was then carefully set out on a wood platform or scribing floor, the centres of all geometrical intersections being marked by gun-metal wood screws, having a small conical hole drilled in the nick in the centre of the screw head. These screws could be set with great accuracy by slight taps of the hammer while being screwed in, and were not subject to after-disturbance. The points of the trammel bars being adjusted to the conical holes in screw heads, the exact dimensions, say centre to centre of eye-bar heads, were accurately transferred from the setting-out floor to the drilling machine, and the important fiducial points being in metal, wear and tear were reduced to a minimum, and mechanical accuracy was maintained to the last.

The mode in which the erected work finally came together was proof of the soundness of the methods employed.

Such a setting-out floor as above described should be under cover, as if exposed to the weather the accuracy of the work in large structures may suffer, and a straight line drawn obliquely across a series of parallel boards with line and pencil one week, may the next week exhibit a lamentable falling off from its previous accuracy.