The ordinary methods of grinding cylindrical surfaces with emery, are not sufficiently exact for the production of these rollers, as the leading source of error in all grinding processes, namely, the unequal distribution of the abrading powder arising from the absence of control, allows the loose emery to accumulate upon the lowest points, and, consequently, after a certain approach to accuracy has been attained, the further continuance of the grinding leads to the depreciation of the surface by the continual introduction of new errors. The impossibility of producing by these means large cylindrical rollers, sometimes required to be as much as 6 feet long and 18 inches diameter, with sufficient accuracy to press uniformly a single sheet of the thinnest paper, has led, after numerous tedious and expensive experiments, to the final abandonment of all abrading powder, and the required accuracy of contact is attained by the simple friction of the surfaces of the rollers rubbing upon each other, plain water being plentifully supplied to lubricate the surfaces, and prevent their heating and tearing each other.

The rollers are first turned as truly cylindrical as possible in the lathe, and tested for parallelism by carefully measuring the circumference at various parts, with a thin copper wire wrapped around the cylinder, a more exact test than gaging the diameter, the journals of the cylinder are turned at the same time, in order to ensure their being concentric.

The rollers are next mounted on their own bearings, in a frame similar to that in which they are to be employed, and their surfaces are carefully adjusted to each other, the bearings of the one roller being fixed, and those of the other placed under the control of a screw adjustment, that admits of the rollers being closed upon each other so that the highest points alone just touch. The rollers are now examined to ascertain whether they fit each other tolerably well throughout their length, as when both rollers have been turned in the same lathe, they will in all probability possess the same general error, or both be either concave or convex in the direction of their length. Most generally long rollers will be turned slightly concave, from the slide of the lathe being more worn in the middle by short works, and the two rollers when placed in contact will show double the amount of error, which if considerable is sometimes reduced by-grinding each roller separately, with a lead grinder supplied with emery and mounted on the end of a lever, that is used to press the grinder in contact with the surface of the cylinder, much the same as in polishing large turned works.

When the errors are so far reduced, that they cannot be detected by the line of light between the cylinders, they are put in revolution, and the one roller marked slightly with a piece of chalk applied at intervals of a few inches, the revolution of the rollers transfers the chalk lines from the one roller to the other, at those parts where they touch, which shows at a glance the highest parts. The points thus indicated are successively reduced with the grinder until the rollers fit each other sufficiently well to transfer all the lines with tolerable regularity, which indicates a moderate approach to general truth, but by no means sufficient for the purpose, as numerous minute errors will still remain that cannot be detected by the chalk lines.

The rollers are now carefully adjusted so that their highest points alone touch each other, and the rollers are driven at different velocities, by separate straps leading to pulleys fixed on the axis of the rollers, which revolve in the same direction, so that the two surfaces in contact meet and pass each other in opposite directions, and the velocities being different, the relative positions of the rollers are continually changing, and it is only after many revolutions that the same points again come in contact.

The friction of the two surfaces causes them mutually, although slowly, to abrade each other, and a constant stream of water is directed upon the rollers, to lubricate their surfaces and prevent them from heating. The latter is a point of considerable importance, as should the rollers become unequally heated from their surfaces being dry, or from too high a velocity being employed, the surfaces would not only be liable to tear, but the irregular expansion of the metal would continually introduce new errors, and the true cylindrical form could not result.

Attention is required to keep the rollers in equal contact with each other at the high points throughout their length, and as these are gradually reduced, the rollers are slightly closed upon each other to bring the next series of high points in contact. It being considered that if the rollers were firmly pressed upon each other, they would bo more liable to copy their mutual irregularities, and also that the pressure would be liable to cause the one roller to follow the path of the other, or be driven by their surface contact, notwithstanding the action of the belts on the driving pulleys.

As the surfaces approach nearer to perfection, the length of contact is gradually increased, and proportionately greater care is required in the adjustment of the rollers, to prevent the friction becoming so great as to tear the surfaces, or cause increase of temperature. The process is tedious, and requires to be continued for several days, until the contact of the surfaces is as perfect as possible, throughout the length of the rollers, in every position in which they are brought, by the continual change of their relative positions.

A very smooth and polished surface is produced in this manner by the use of water alone, but for those rollers required to possess a still smoother surface, Messrs. Hopkinson and Cope, (from whose practice the foregoing particulars have been derived,) have adopted the use of oil instead of water for the last finish, and the smoothness of surface thus produced leaves little room for improvement.