to the vibration of the spring which measures P, and are caused by
variations of the coefficient of friction of the dynamometer. By making
P very much smaller than in the usual form of the dynamometer, any
errors in determining it have much less influence on the measurement of
the work absorbed. We may go further. The cord may be taken over four
pulleys; in that case a variation of 20 per cent. in the frictional
coefficient only alters the total friction on the pulleys 1¼ percent. P
is now so insignificant compared with Q that an error in determining it
is of comparatively little consequence.

[Illustration: FIG. 5]

The dynamometer is now more powerful in absorbing work than in the form
Fig. 3. As to the practical construction of the brake, the author thinks
that simple wires for the flexible bands, lying in V grooves in the
pulleys, of no great acuteness, would give the greatest resistance with
the least variation of the coefficient of friction; the heat developed
being in that case neutralized by a jet of water on the pulley. It would
be quite possible with a pulley of say 3 feet diameter, and running at
50 feet of surface velocity per second, to have a sufficiently flexible
wire, capable of carrying 100 lb. as the greater load, Q. Now with these
proportions a brake of the form in Fig. 3 would, with a probable value
of the coefficient of friction, absorb 6 horse power. With a brake in
the form Fig. 4, 8.2 horse power would be absorbed; and with a brake in
the form Fig. 5, 8.8 horse power would be absorbed. But since it would
be easy to have two, three, or more wires side by side, each carrying
its load of 100 lb., large amounts of horsepower could be conveniently
absorbed and measured.

* * * * *