In the case of very long sights, even this method failed, because the circle could only be read to half-minutes, which was too coarse a graduation to give a good result, and in these cases the method used was one of repetition. The wire being brought to the base of the staff as before, and a first reading of the vertical circle taken, the wire was brought to the top of the staff by the tangent screw, then to the bottom again by altering the levelling screws slightly, again to the top by the tangent screw, and so on, three or four times, and then a second reading was taken on the vertical circle. The slight alteration of level had no sensible influence on the result, and it is obvious that by automatically summing up, say, four intercepts in this way, a very much more accurate value of the subtense angle was obtained than was possible from a single measurement. In practice I found it was best to carry in the waistcoat pocket a card giving the distance corresponding to any number of minutes of difference of reading after a four-fold repetition, and it was quite practicable to measure up to three kilometres of distance within one hundred metres of the truth; as this only represented a millimetre on the sheet, and as, moreover, errors were not cumulative, owing to the independent fixation of each successive station by re-section, the accuracy was all that could be desired, and the rapidity of measurement was very great. In this long distance type of tacheometry, finely graduated staves were of no use; the form of staff employed was a broad-faced one, fifteen centimetres wide, bearing fifty-centimetre divisions painted alternately black and white right across the whole breadth of the staff.[77] An average of about six or eight conspicuous points having been telemetrically fixed from a station, the detail was sketched in around them, and other more distant points were at the same time fixed by plane-table intersections from several stations. At the stations the pencil sketching of relief was by form-lines which were subsequently replaced by hachure-rendering when inking up the sheets in camp.