1From 1969 until 1985 several Byzantine hydraulic machines, mostly mills, were excavated in Hanghaus 2 of Ephesos by the Österreichisches Archäologisches Institut (ÖAI). Our stone saw, being one of these machines, was discovered in 1983 in room WT 2 (fig. 1). The co-operation of Österreichisches Archäologisches Institut and Ôsterreichische Akademie der Wissenschaften, Institut fiir Kulturgeschichte der Antike (ÖAW), and Römisch-Germanisches Zentralmuseum Mainz (RGZM) on the other hand, allows the investigation of these important archaeological finds1

Sources and approaches of interpretation

2The interpretation is based on the documentation of 1983 as well as a three weeks campaign in August 2006. Wooden parts were not found, only a few metal parts remained from WT 2. The metal artefacts in this room were apparently robbed. Although investigations continue, the presented results serve as a base for discussion2. Water technology and mechanical engineering experienced only few fundamental innovations until the 18^th/19^th century. Thus, the understanding of the machine is improved using analogies to technological history of the modern times.

Stone cutting machines: necessary components of antique stone production

3There are three further references known for water-driven stone saws. Ausonius mentions them in his poem Mosella (371 A. D.) working at the banks of the river Ruwer at Trier (Weis 1997, p. 51-52). In the southern Kryptoportikus of the Artemis temple of Gerasa, Jordania (6^th or 7^th A. D.) there existed a hydraulic stone cutting machine with two saws similar to the Ephesos machine but with four blades for each saw (Seigne 2000, 2006 and this volume). The earliest evidence comes from Turkey in the form of a relief on a sarcophagus lid from Hierapolis (3^rd, maybe 2^nd century A. D.) that gives a detailed representation of a double saw with water supply, water wheel, transmissions, connecting rods, stone saws and sawing frames (Grewe in print and this volume, Ritti in print). The evidence at Ephesos is important in helping us to determine how several millions square meters of polished marble slabs were produced in Roman and Byzantine times. It is proposed that the large scale production of polished slabs and opus sectile was possible only by employing water-driven mechanical stone saws.

The Ephesos stone cutting machine

Localization, building and date

4The sawing room WT 2 (fig. 1 and 2) belongs to several Byzantine constructions in Hanghaus 2 and is the lower one of a row of hydraulic mills, similar to Barbegal in France. WT 2 lies in the northwest of the Hanghaus, thus close to the ancient city, and its position means that transportation of the stones and slabs was minimised. The room is not rectangular, its maximum length and width are approximately 13 by 11 m. The southern part of the sawing room is divided and here is the polishing track, where the slabs were finished. The machine occupies an area of 3 by 8,5 m. Findings under the occupation level date the construction into the late 6. to 7. century A. D (Ladstätter in print). South of the sawing room there is a further room, WT 1 where a second polishing track existed as well as polishing places for the production of smaller marble objects. Capitals were also manufactured here. WT 1 and 2 together are a workshop area, in which almost all activities of contemporary stone handicraft occurred.

Water supply (aqueduct, water channels and regulation)

5The old aqueduct still running above (Thür 1995) was used for the mills of Hanghaus 2. In the Hanghaus the headrace channel ran on top of the walls of the buildings and passed downwards through a slot in the top of the wall (fig. 2), striking the water-wheel driving the saws. The slot as well as the wall below are heavily covered with sinter. The slot is scarcely 3 m above the bottom of the channel in which the water wheel ran. Stopping and starting the machine was done by regulation of the water supply, but there is no evidence to show how the water was diverted away from the wheel. Maybe the saw and another mill in the neighbour room were supplied with water alternatively, using a switch.

Drive (wheel channel, water-wheel, wheel-shaft and bearings)

6The machine was obviously driven by a wooden water-wheel (fig. 3, n. A), but no remains of it have been found. However, measurements of the building allow a reconstruction. Under the head-race inlet slot there are two large rectangular beam holes a distance of 65 cm apart (fig. 3, n. 5 and fig. 4, n. 5). Their lower edges lie approx. 80 cm above the floor. Between these beam holes there is the wheel-pit channel, 50 cm wide and fairly deep (fig. 2, n. 6 and fig. 3, n. 6). The wheel channel is 8,5 m long and is open for 3,5 m and further downstream is covered by a stable stone setting of larger spoliae, about 80 cm high (fig. 2; fig. 3, n. 7 and fig. 4, n. 7). The north channel served for the outlet of the tail-race water (fig. 2; fig. 3, n. 8 and fig. 4, n. 8). Two 5,5 m long parallel beams were supported at their western ends by the stone setting and at their eastern ends by the wall cavities (fig. 4, n. B). The two bearings (fig. 3, n. C) of the water-wheel shaft (fig. 3, n. D) were fixed onto and supported by the two parallel beams. Thus, the height of the shaft over the wheel channel was approximately 1,6 m, which determines that the diameter of the wheel could not exceed 3,2 m. Allowing for underside clearance in the wheel-pit the diameter of the water-wheel was close to 2,8 m having a circumference of approx. 8,8 m. Within the channel the maximum width of the wheel was almost 50 cm. The power of the postulated wheel can not exceed a few KW, possibly corresponding to the power of a very small moped. Further computations are necessary3.

Running deviations within the sawing cut (tilting, twisting and distorting)

7A stone saw can be subjected to several different forces; it can be twisted and tilted (Rôder 1971, fig. 62) or the saw blade itself can be distorted. Tilting for example causes an unwanted convex cut but other forces are balanced to some extent by the guiding offered by an advanced sawing cut. If the friction becomes too great the machine will stop. Overheating and breaking of the saw blades should be impossible.

Fig. 3. Remaining elements of the machine: 1, 2: Blocks to be cut. 3: Water supply by a channel on top of the wall. 4: Slot in the wall to feed the wheel with water. 5: Beam holes for supporting the wheel-shaft bearings 6: Wheel channel. 7: Spoliae providing the support for the wheel-shaft beams. 8: Water outlet. 9: Post holes for the suspension frame (all four of them covered by reconstructed elements). 10: Beam holes for the suspension frame.
- Reconstruction: A: Water wheel. B: Shaft beams. C: Wheel-shaft bearings. D: Shaft of the water-wheel. Ε Crank-pin. F: Drive-plate. G: Connecting rods. Η Cutting blades. I: Cutting frames. J: Posts for the suspension. K: Long beams of suspension. L: Short beams of suspension. M: Suspension pulleys. N: Pulleys. O: Counterweights.

Fig. 4. View along the running direction of the water wheel of the Byzantine hydraulic stone cutting machine from Ephesos (TR). 1 and 2: Blocks to be cut. 1. 1 and 2. 1: Complete, finished cuts. 1. 3 and 2. 3: Unfinished cuts. 5: Beam holes for the shaft beams. 6: Elongation of the wheel channel (collecting a part of the used flushing water). 7: Spoliae providing the support for the shaft beams. 8: Water outlet. 9: Post holes for the suspension frame. -Photo: ÖAI 1983

Power transmission to the saw (crank and connecting rod)

8At each end of the wheel-shaft the saw frame is driven by a crank (fig. 3, n. E/F) that converts the circular motion into linear4. The peripheral velocity of our 2. 8 m diameter water-wheel is 1,5 m/s (Speidel 1912, p. 121) which results in a rotational speed of about 5,7 s for one revolution. Converted into the linear cutting motion this gives a time of approximately 3 s per any horizontal moving back or forth.

9Running distance and thus speed depended on the diameter of the proposed drive-plate (fig. 3, n. F) which is unknown. Modern stone saws possess e. g. drive-plates of approximately 20 cm diameter (Ludwig 1995, p. 261), a bigger diameter would cause a tilting of the saw. Thus we have a running distance of 20 cm and a speed of well 7 cm per second which, although quite slow for a stone saw, they lie within an acceptable range.

10The power transmission from the crank to the saw frame was made by a wooden connecting rod (fig. 3, n. G). The longer this rod, the less the sawing cut was tilted. The 4,3 m distance between wheel-shaft and cutted stones gives the largest possible length of the connecting rod. It is possible that the Ephesos machine may have had a transmission construction as suggested by the Hierapolis sarcophagus relief which would result in a different machine speed (Grewe in print and this volume, Ritti in print). However, archaeological evidence for this is missing at Ephesos.

Sawing procedure (blade, abrasive and flushing)

11The sawing of harder rocks is essentially a form of grinding where the blade of a stone saw does not possess teeth, except when cutting very soft stones. Until the invention of blades with segments equipped with diamonds, cutting was performed by steel blades, with quartz or steel sand being added as abrasives (Ludwig 1995, p. 259-269). It is assumed that in the Ephesos machine, abrasives were probably applied by hand (fig. 5), as still was the case with some 20^th century stone-sawing machines. The blade ground the abrasive into the rock thus dividing the stone. As has been observed in some Roman sawing cuts, the blades became thinner during sawing (Roder 1971, p. 306). Thus, in Antiquity the saw blades could have been made of iron or soft non-ferrous metal, such as copper or bronze using abrasive such as quartz sand or carborundum emery5.

12The abrasive required water lubrication to be effective, which also reduced friction and rinsed the sawn particles away from the cut. Small wooden gutters would have ensured continuous flushing (fig. 5). The flushing water was taken from a water pipeline at the north wall of WT 2 (fig. 5). Probably two small columns supported the gutters. The heavy abrasive will have settled down close to the cutted stone, so its recuperation was possible.

Saw (blades, cut order, number of blades, sawing frame, suspension, lowering and block moving)

13The length of the saw blades (fig. 3, n. H) can be determined from the measures of the sawn pieces (fig. 2, fig. 3, n. 1/2 and fig. 4, n. 1/2), whose greatest length measures 2,6 m. If we add the double crank shaft length of 10 cm as well as a necessary freewheel of about 10 cm at each end, we arrive at a blade length of approximately three meters. The blades were only a few millimeters thick and their height was about 10 cm.

14The remains of two completed cuts (fig. 4, n. 1. 1 and 2. 1) show that large slabs were manufactured and that the sawing began in the inside close to the wheel channel and step by step was moved towards the outside of the stone blocks (fig. 4, n. 1. 2/1. 3 and 2. 2/2. 3). As on every block there are two parallel unfinished cuts that all have been stopped at the same depth (fig. 4, n. 1. 2/1. 3 and 2. 2/2. 3), the machine had two blades on each side-altogether four. Also, both of the two blocks were cut at the same time. In modern terms our saw is a „double two-bladed gate. “The Gerasa saw was a hydraulic machine with two stone blocks sawed at the same time, as well as it is shown by the antique sarcophagus lid from Hierapolis.

15For the suspension of a saw blade a vertical H-shaped clamping frame is possible. The relief of Hierapolis shows such frames, apparently with only one blade each. In stone technology this is called a “cutting-off saw”. With employment of several blades this construction is pushed to its limits by the increased blade numbers. In modern stone-cutting technology a rectangular, horizontal frame with parallel clamped sheets is used (Ludwig 1995, p. 259-269), which is called a “full gate”. In this frame (fig. 3, n. I) the centre of gravity lies slightly above the sawing cut so that the tendency of the saw to deviate from its course is much reduced. How the blades (fig. 3, n. H) were fixed is unknown. In Ephesos, there were two seperate sawing frames, both of them running at the same time, and if only one was being used, the second ran empty. For positioning a new stone block, the frame on that side was disengaged whilst the machine continued working at the other side. The saw blades would have been positioned in the center of the frame, with the connecting rod on the same axis to avoid side forces, twisting and distortion or “bucking” of the frame. Thus the connecting rod, saw blades and axle center of the frame were in a straight line together with the sawing cuts. The saw frames were about 3 m long and 1,1 m broad.

16The frameworks were suspended in an arrangement that could be described as similar to two coupled pendulums. For a good example of a modern reconstruction see Warneke (1997). Whilst not working, the vertical lines of the suspension as well as the horizontal lines of the sawing frames form a rectangle. In operation, thus the frame moving back and forth, the lines describe a parallelogram with its center of gravity moving back and forth. So, the saw blades permanently dance up and down. With a relatively long suspension, this vertical movement is negligible and is limited to the points of reversal during the oscillating motion. This slight lifting allowed the abrasive to get under the saw blades, so that an oscillating suspension is considered very necessary. However, although the sawing frame oscillates, the saw blades slide almost horizontally through the stone.

17At Ephesos there is evidence to suggest how the saw frames were suspended. Above the holes for the shaft beams there are two rows of small beam holes. The upper row (fig. 3, n. 10) is high enough for a 2,6 m long frame suspension (fig. 3, n. K). These holes are almost in line with the four holes (fig. 3, n. 9 and fig. 4, n. 9) for the posts (fig. 3, n. J), which carried the long beams of the suspension. Since the two sawing frames had to be hung on all of their four corners, two additional shorter beams were necessarily (fig. 3, n. L).

18Suspension ropes (fig. 3, n. M) passing over pulleys (fig. 3, n. N) allowed the frames to lower as the work commenced. To control the weight of the frame, counterweights (fig. 3, n. O) were needed. In order to ensure a light pressure, the four counterweights together would have had a total weight approaching that of the frame and blades. Depending upon hardness of the stone the pressure could be changed by varying the size of the counterweights. With optimal adjustment, no manual lowering was necessary, and the saw would cut into the stone step by step automatically under its own weight.

19The construction of the saw does not allow the misalignment of connecting rod and cut axis. The stone blocks had to be positioned by crowbars horizontally and in such a position that the first cut was central under the frame and parallel to the intended line. Before each new cut, the stone block would be positioned parallel to the earlier saw cuts by crowbars.

Conclusion

20It is believed that the interpretation of the evidence in conjunction with our present day knowledge of stone sawing, gives us considerable confidence in the proposed reconstruction (fig. 5). This machine was a semi-automatic, double two-blade gate, powered by a water-wheel. A water supply branched off the headrace to provide lubrication and the flushing of the abrasives from the saw cuts. The lowering of the saw cuts took place automatically by the counterbalanced weight of the sawing frames. This confident reconstruction suggests that water-powered mechanical stone sawing was well advanced by the Byzantine and maybe already the Roman period, and was essentially the same as that which existed in the 18^th and 19^th centuries. The use of decorative marble slabs was very common in the architecture of the Roman and the Byzantine period. Such a large quantity of stone slabs can only have been produced with the help of this water-powered type of machine. For a better understanding of this machine and to get data for technical and economical quantification, it is proposed to reconstruct a replica of the machine in the scale of 1/1.