The Hobson shield followed Greathead's as closely as any other, in having a diaphragm with closable doors, but a modification of Beach's sharpened horizontal shelves was also used. However, these functioned more as working platforms than supports for the earth. The machine was 21-1/2 feet in diameter, an unprecedented size and almost twice that of Greathead's current one. It was driven by 24 hydraulic rams. Throughout the entire preliminary consideration of the project there was a marked sense of caution that amounted to what seems an almost total lack of confidence in success. Commencement of the work from vertical shafts was planned so that if the tunnel itself failed, no expenditure would have been made for approach work. In April 1888, the shafts were started near both riverbanks, but before reaching proper depth the almost fluid clay and silt flowed up faster than it could be excavated and this plan was abandoned. After this second inauspicious start, long open approach cuts were made and the work finally began. The portals were established in the cuts, several thousand feet back from each bank and there the tunneling itself began. The portions under the shore were driven without air. When the banks were reached, brick bulkheads containing air locks were built across the opening and the section beneath the river, about 3,710 feet long, driven under air pressure of 10 to 28 pounds above atmosphere. For most of the way, the clay was firm and there was little air leakage. It was found that horses could not survive in the compressed air, and so mules were used under the river. In the firm clay, excavation was carried on several feet in front of the shield, as shown in the model (fig. 42). About twelve miners worked at the face. However, in certain strata the clay encountered was so fluid that the shield could be simply driven forward by the rams, causing the muck to flow in at the door openings without excavation. After each advance, the rams were retracted and a ring of iron lining segments built up, as in the Tower Subway. Here, for the first time, an "erector arm" was used for placing the segments, which weighed about half a ton. In all respects, the work advanced with wonderful facility and lack of operational difficulty. Considering the large area, no subaqueous tunnel had ever been driven with such speed. The average monthly progress for the American and Canadian headings totaled 455 feet, and at top efficiency 10 rings or a length of 15.3 feet could be set in a 24-hour day in each heading. The 6,000 feet of tunnel was driven in just a year; the two shields met vis-a-vis in August of 1890.