Next Page »

Navigational note:  if you arrived here from the home page looking for day 1, keep going.  “Next page” will take you there.


Photo from, Greek and Roman Artillery, by E. W. Marsden.


For many years researchers had suggested that the square hole visible in the side of this ancient skull was the work of a square headed ballista bolt. Recent studies have fired bolts from a replica Skorpion ballista at sheep skulls and concluded that a neat four sided hole would not be the likely wound signature from a ballista bolt. Apparently the dried out sheep skulls shattered into multiple fragments, rather than perforating cleanly. The researchers concluded that this hole was more likely made by the thrust of a Roman pilum (a type of spear),  probably in some form of battlefield execution. This may well be the case, however the sheep skull experiments do not seem to cover all the variables that might influence how a human skull will react to this kind of intrusion.  Perhaps freshly killed pig skulls would be a more accurate model.



Photo by, Yale University Art Gallery; and copied from, Greek and Roman Artillery 399BC-AD 363, by Duncan Campbell.


This artifact is the only intact specimen of a Roman ballista bolt ever discovered.  It was excavated in Dura Europos, Syria, and is likely the type of projectile used by the Orsova ballista. Bolts of this type must have been made in the tens of thousands by the Romans. The wooden shaft is made from ash and the quadrobate iron tip is a four sided square, designed to punch through shields, armour, clothing, etc.  The thinner foreshaft of the Dura Europos bolt was designed to aid penetration, it also raised the ballistic coefficient to aid in a long flat trajectory. The swelling at the rear of the bolt not only helped absorb the powerful thrust of the bowstring, but also acted as a kind of aerodynamic counter weight to help keep it steady in flight. This inherent stability was further aided by the short stubby fins.  These fins were made from maple and glued into grooves cut in the ash body.   Overall length is 18″, thickness at the rear end is 1  3/16″ and it tapers down to a bit over 1/2″ in the front.  It was reported to be a good deal shorter than the type of bolt that came before it.  Late style ballistas like the Orsova model did not have any kind of narrow opening around the bolt groove that the projectile would have to pass through.  (See earlier postings of Gallwey ballista.)  The arched strut on these later iron framed machines meant they were better suited to firing the shorter style bolts.  The danger with a short bolt is that if something goes wrong, and the bolt turns sideways during the power stroke, it will likely smash into the supports in the middle of the box frame.  With an arched frame the whole mess will be cast out the front, avoiding unpleasant ricochets and perforated catapultiers.   

We see on the arched strut a pair of holes on either side of the arch. It has been suggested they were put there to allow a string to stretch across the arch. A simple bead on this string could have acted as a front sight. This was heartening news as it meant I was on right track trying to extract as much shooting accuracy out of the Orsova reconstruction as possible. In later centuries when the smoothbore cannon was in widespread use, artillery men would often just sight over their barrel without the aid of sights or plumb bob, pointing it at the target instinctively. Many researchers have supposed that the ancients used a similar instinctive technique for sighting their ballistas. Generally a lack of sights would seem to indicate a weapon not much prized for any kind of fluid, tactical accuracy. (Seige engines designed to batter down walls were another matter.  The trebuchet is quite accurate in its ability to hit to the same point.  But it was hardly designed to quickly acquire a new target.  Accurate shooting was a matter of shoot, observe the strike, and adjust.  The cycle was repeated until good hits were made.  No sights were needed.)  The two holes on the Orsova artifact gives us some evidence that perhaps this ballista was prized for its quick accuracy. Given their unique and perfect placement for this purpose, there doesn’t seem to be any logical explantion for these holes other than some type of front sight. Which, of course, begs the question, did they also have a backsight? The Romans were intimately familiar with surveying instruments, which no doubt required some form of sophisticated sighting. Surely they must have tried the simple peep sight?. The suggestion that these ancient crew served weapons had sights, further enhances the argument that they were used in the sniper role. Considering that the Romans were facing tribes of warriors often held together by the force of a single chieftan, the tactical utility of knocking out the opposing leadership with a well aimed shot seems strong motivation for them to use sights.

6 MOA at 100 yards seems like a reasonable goal for this project. Good sights will be vital to achieve that kind of accuracy.  I have some concerns that the very acute string angle of an inswinger at full draw, may somehow introduce instability into the bolt.  On the other hand, none of this would be any fun if it was all cut and dried. We’ll see.

Expectation is the measure of what we hold as normal. I had hoped that my new digital dictaphone would be a convenient way to jot down ideas when I was ankle deep in metal chips. If I’d had the patience to wade through the 500 or so features that I didn’t need, to find the few that I did (ie. record, play, repeat etc.) it might have been a useful tool. As it is, I’m back to dry erase pens and a clumsy white board that is never there when I want it. When a new technology falls below its promise, we mourn something we never had in the first place. I am seriously considering a transcription program to turn spoken words into script on the computer. But that would probably be just another complicated nightmare. Perhaps if I wore the white board around my neck on a string. I don’t think that would be overly offensive to my sense of normalcy.

In static cultures where the simplest innovations may take generations, or happen not at all, the sense of what is normal overwhelms the need for change. Egyptian art rendered the human image in a highly stylized form that remained unchanged for over three thousand years. It was emblematic of a culture based on an agricultural technology utterly reliant on the regular rise and fall of the Nile. Their expectation of that cycle bred a dependency fearful of change. There was no need for them to develop the printing press, let alone digital dictaphones or other bothersome distractions. Their attitude is not surprising as any changes they would need to make were tied to cycles far longer than a human lifetime. If the patterns of the past are stable, then our expectations are not threatened, and change for its own sake becomes mere frivolity. If the patterns of the past become unstable, then our threatened expectations force change. The rub always lies in our ability to recognize a changing pattern when we are in the midst of it. Clearly my failure to embrace modern recording technology was going to make the white board necklace a reality.

In a sense we are victims of our own addiction to drama. It seems deeply ironic that it often takes a war, or the threat of losing a war, to spur innovation. Relabeling public policy issues with the appellation “war” is a feeble attempt to harness the power latent in our ability to recognize change. The War on Drugs, The War on Poverty, The War on Terror, they all seek to elevate our response out of the ordinary. If they work, we crowd around them like brave warriors gripped only by the fear that others will not also jump at the chance to bring about a better world. Our sense of business as usual becomes as fluid as the vanishing futures we had hoped for. Expectations change. In desperation we either plunge into reactionary versions of my mutating white board, or we learn how to use the dictaphone. Where did I put that instruction manual anyway?


A question has surfaced regarding the string movement during the power stroke of an inswinger style ballista. Let me set up the problem with the following givens. The sole factor governing bolt speed is how fast the bowstring can be made to move under load at the point it touches the rear of the bolt. In other words, the bolt will never move faster than the center of the string. A rather obvious point to be sure. However, if we forget about the bolt for a moment and concentrate just on the string movement over its entire stroke, an interesting phenomenon seems to happen. While playing with the full scale mock-up of the limbs and bowstring, the following was observed. When the nock point on the string has traveled through 50% of its stroke going from a draw length of around 62” down to 31”, the amount of distance traveled by the limb is 72% of its total rotation, going from 110 degrees at full draw down to 31 degrees at half draw. Does this mean that if the limbs only have 28% of their rotation left to go before they come to a stop, and the string has a full 50% further to go before it comes to a stop, that string speed will need to increase towards the end of the stroke because the string and limbs have to come to a stop at pretty much the same time? (From a functional standpoint there is zero stretch in the string)  My mathematical skills are not up to the task of answering whether or not this may give the bolt a boost in velocity just as it leaves the machine. It would be a welcome surprise if it did. Optimistic intuitions aside, with this did the Romans come up with the first form of compound leverage applied to a bowstring? My good friend Brian Kern suggests that stroboscopic photography would answer the question positively. Like so many things on this project, that is a bit down the road yet.

One of the more satisfying aspects of reconstructing a torsion catapult is to witness the level of amazement most people have when they first see it fire. The slow wind up process. The series of loud clicking sounds as various locking mechanisms snap into place. Bundles of rope creaking under enormous strain as the load starts to build. The exertion of the operator using the winch. It all adds to the relative drama . In public demonstrations* I have noticed a baited pause in the collective breathing of the audience a second before the shot is released. When the trigger is finally pulled, the resulting smack of the limbs into the frame is louder than most people expect. A split second later it is followed by an equally loud smashing sound as the bolt rips through several sheets of plywood.

Click here for a grainy peak at the Gallwey in action,   visible-tech-1

These are the effects of a visible technology. A machine whose functioning and performance people can understand on an instinctive level. We live in a world ruled by invisible powers. How many of us really understand how our plasma and LCD t.v.’s operate? Computers and cell phones are generally a mystery. Fuel injected, oxygen sensing automobiles may get us around, but who wants to try and fix one. All these things exist in a world made opaque with the magic of invisible technology. I have had enthusiastic five year olds start explaining to me how they thought a ballista worked. Many times they got it exactly right. A catapult represents a source of power that is utterly uncomplicated. It doesn’t rely on chemical detonation or puffing up huge quantities of steam or the arcane movement of electrons running through copper wires. It comes from a time when the water wheel and the oxen were the forces that moved civilization forward. When twisted skeins and the craft of cordage were rendered into instruments of lethal propulsion. Before gunpowder, the world was ruled by people that knew how to harness the power of visible technology. They were gifted artisans and engineers, no less intelligent than ourselves, just without the knowledge of how to harness unseen physical forces we now take for granted.

Discovery and invention have made the modern world possible and the catapult has morphed into the nuclear tipped missile. By todays standards, a torsion engine is an archaic and whimsical piece of equipment. Hollywood notwithstanding, it is as far from a weapon of mass destruction as the horse and buggy are from a Ferrari. As impressive as a ballista is to see function, its actual performance compared to a modern firearm is a pathetic joke. My reconstruction of the Gallwey ballista weighs close to 800 lbs, takes a couple of minutes to wind up with a draw force measured in tons, and for all that, still only generates the same power as a single round out of six shot .44 magnum revolver you can hide under your coat. The limits of visible technology spelled the end of the ancient world. If the Romans could have developed gunpowder, or steam power or electricity,  perhaps their empire would have dominated the world for another thousand years.   As it was, they crumbled under the weight of their own ambition.

*  Public demonstrations are always performed at a fraction of full power for safety reasons.  It is one of the unique features of a ballista that it can be fired at anywhere along its draw length.  Unlike a crossbow that always has to be fired at full power, a ballista operator can choose the power setting on the machine.  One wonders if the ancients used this feature for playing war games.  A classical version of paintball perhaps.

A few notes about fit-up work seem to be in order. The loops on the field frames of the original artifacts are described as being somewhat crude in terms of how close they all are to one another dimensionally. This is not necessarily an impediment to getting the whole assembly to lock up tightly if we are using wedges to hold the joints together. My reconstruction uses lugs that wrap around the tangs of the arched strut to trap the loops of the field frames in place, driving in wedges behind the tangs causes everything to lock up solidly.

This particular approach is the one place where my reconstruction diverges from the artifacts. It is a temporary concession needed to conduct some full power accuracy tests I have in mind. A more authentic joint utilizing the elongated holes seen in the original tangs will be substituted later. (The idea being that we will work backwards from the best group size possible, and see if the Roman approach using some kind of cross wedge system with those holes they put in their tangs, will degrade the group size.) It has been suggested that the Romans probably used hardwood wedges to lock everything together. I used some made of oak when I was trial fitting the parts together and they did seem to work okay. However, in keeping with the maximum performance goal of this project, I soon replaced them with steel and bronze wedges built around a three degree taper. The matching surfaces on the struts that fit into both the top and bottom loops, were also given a three degree taper. It would certainly not have been beyond the capabilities of any competent Roman armorer to have likewise used matching tapers to insure maximum rigidity in the final assembly. One thing is for sure, the loose and rattling nature of the eight juncture points where the loops and struts went together, instantly disappeared when well fitted wedges were driven into place. The whole assembly had the feel of being locked together with remarkable strength. In short, if the fit-up work on the wedges is done properly, the varying dimensions evident in the loops of the original field frames would not have affected the integrity of the final assembly.


In his treatise on catapult design, Belopoeica, the ancient Greek author Philon tells us this about successful designs, “The man who wants to shoot far must try to put on as much spring-cord as possible: of course, not only do we believe that the secret of power lies particularly in this, but all others believe it, too. But, since the spring cord passes through the holes of the hole-carrier, he who intends to put in more spring-cord must make the hole in the hole-carrier larger (otherwise they will not take more), so that the surrounding edges are left extremely thin and are naturally weak. It is impossible to make the hole-carrier broader, for it will then exceed its dimensional scale”. In other words, the impulse to make the spring bundles larger in diameter is limited by the size of the holes in the frame through which they must pass. There comes a point when that desire to make larger springs requires the whole frame to become bigger, and then of course we are talking about making a different machine than the one originally envisaged.


With the Orsova reconstruction the original artifacts presented a bit of a dilemma.   The following model should help illustrate the problem.



 The end caps in the field frames have 3 1/8” holes for the spring bundles and do not have any kind of recess cut in them for the bronze tightening washers to ride in.  Absent some other component, that would mean that the flange on the bottom of the washer would need to fit down inside the holes of the end cap to allow the washer to run concentrically in the frame as the pre-load tightening procedure is being applied. The wall thickness of such a flange would need to be at least 1/4”, therefore the effective hole diameter for the spring would naturally reduce the room available for the spring bundle down to 2 5/8”.  This represents a 16% reduction in spring size. An unthinkable amount for an obsessive catapultier lusting after more foot pounds of energy. Either we accept that the Romans devised a way of torquing up the springs in their ballistas without having the washer’s axial movement controlled mechanically, or they were unconcerned about the washer drifting around on the end cap of the field frame until it was pinned in place. Possibly two strong men with a tightening lever that projected from either side of the washer could control the axial drift by sheer muscle power, while a third popped the locking pins in at the appropriate place. Would they still be able to pre-load the springs with as much torque as a system that had some kind of flange projecting from the underside of the washer. Probably not. Especially considering that any frictional braking that might impede maximum torque with a flange based system would be caused by bronze rubbing against iron, a pairing of materials well known to prevent galling and reduce friction. Moreover, from the earliest of times, torsion engines had relied on a flange riding in a groove to allow maximum pre-loading of the springs. Probably this part of their past designs they would have little incentive to change.


How then to use a flange based system for tightening the washers, when the original field frames unearthed in Orsova had no recess in the end caps for the flange to run in?  Enter my proposal for the “vernier” plate.  A simple 3/8” thick plate that had four projecting pins to lock it into the end cap and a vernier based hole pattern for the washer’s locking pins. With a 3 5/8” through hole it would naturally providing a recess for the flange on the underside of the washer to ride in. Now we can have the spring bundles at the maximum diameter allowed by the artifacts (around 3”), and all we need to accept to get there is that,  like so many of the other parts of the original machine, with the ravages of time, they just turned up missing. I will favor this interpretation as it gets us to a point of maximum performance, with a minimum of incredulity.


As a side note, we might speculate that the forging of a very simple plate containing the hole patterns described above,  would have great utility in the ancient workshops as it would allow an easy way to mix and match the various field frames and washers they might have on hand.  Such an approach would obviate the need for much of the close tolerancing of the hole patterns between the field frames and washers (esp. radially), as final assembly is now controlled by the fit-up work on a single separate component.   In my experience, this is a fairly common strategy when large amounts of handwork are involved.


The arguments set forth by Mr.Lewis and Mr. Iriarte that the Orsova ballista was an inswinger, struck me as the most logical explanation for the unique geometry inherent in the original artifacts.  If we examine all of the ways these components can be fitted together with the following mock ups, the validity of their hypothesis becomes readily apparent.  Because the tangs on the arched strut are of different lengths, inserting them into the loops on the field frames limits the number of arrangements possible to the following three cases.  (I don’t include the fourth possible case of “curved stanchions facing out on the aft side of frame”  as even a remote possibility because the curved stanchion would interfere with any kind of sensible limb movement,  i.e. at rest limbs held nearly parallel to stock)

Case One. Curved stanchion facing inwards and on the aft side of the frame

In the first photo we might suspect that the curved stanchion allows the limbs to be retracted to an extreme degree to increase the length of the power stroke. However, the more the limb and string are brought into a straight line with one another the less leverage is available to move the limb tip back further. This is the commonly known phenomenon of “stacking” that the archer finds when shooting poorly designed short bows. Things go fine through mid draw and then at some point the increasingly unfavorable string angle puts the draw force into an exponential curve.   A classic case of diminishing returns.  Moreover, as seen in the second photo, the straight stanchion is not in the correct position to act as a limb arrestor, nor would it be substantial enough to act as one even if it was. We can forget about the commonly held view that the bowstring would act to stop the limb moving forward at the end of its stroke much like it does on the simple hand bow.  The 12,000 lb. test dacron yacht braid used on the Gallwey bowstring was easily able to handle its 4000 lb maximum draw weight, but it (or the eye loops spliced into its ends) soon stretched out when subjected to the full snap of the limbs going forward.  The sudden jerk on the bowstring coming from a powerful ballista is going to quickly stretch any string to the point that the limbs will keep rotating forward until they hit something substantial.   Clearly, some type of shock absorbing pad is needed to prevent damage to the limb and to the stanchion it smashes into.   For these reasons, the Case One arrangement seems pretty dubious.

Case Two. Curved stanchion facing inwards and on the forward side of the frame.

I originally spent a fair amount of time thinking that an argument could be made for having the curved stanchions in this position as a way to pre-load the spring bundles by using the limbs themselves. If the limbs were allowed to rotate forward into the pockets of the curved stanchions and then the washers were locked into place to the field frames, retracting the limbs to slip on the bowstring would naturally twist up the spring bundles to pre-load them. What a lot of nonsense that turned out to be.  Again there is no limb arrestor with this configuration,  which pretty well kills the idea right there.  Also, if we measure the length of draw available before the limbs rearward movement is stopped by the straight stanchions, it is an unimpressive 34”.  For a machine of this size, that is way wimpy.

Case Three. Curved stanchions are facing outwards and on the forward side of the frame.

In this position,  the heavy duty curved stanchions are in the perfect position to act as limb arrestors. Length of draw goes to a whopping 63”.  Also, the great distance apart between the spring bundles (compared to a conventional outswinger) now makes sense as it provides the room necessary for the limbs to travel through their arc of movement without hitting the stock.  Comparing this configuration to the other two cases, there doesn’t seem to be much of an argument in their favor.  The Orsova ballista was clearly intended to be an inswinger.  To eyes that are accustomed to seeing the limbs and bowstring in the conventional position of an outswinger, the above inswinger looks awkward and can violate more traditional sensibilities.  At least that is how I felt when I first looked at it.  However, I am also a firm believer in the notion that form must follow function.  After actually handling the reproductions of these artifacts and fitting them together in every possible configuration, this radical interpretation does seem to be inescapable.

Was the inswinger, like so many of the innovations used by the Romans, the invention of some unknown Greek engineer?  Or was it something truly new and radical that the Romans had come up with themselves?  Probably that is something we will never know.  I can recall reading that towards the end of the Roman Empire there were certain politicians who spoke of a whole new class of secret weapons that would be able to save the day.  Was the Orsova ballista one of the designs they were talking about?  Very likely I should think.  However, like so many promises of that kind,  larger events overtook any tactical advantage the new design may have rendered.


The key to developing power in a torsion engine is to pre-load the springs with as much torque as is consistent with the capacity of the spring material to utilize without failure. In the case of the Orsova reconstruction, we are looking at a 3” diameter bundle of nylon line. The amount of tensile force required to break a bundle that thick is around 40 tons,  however the important number for a ballista would be the point at which the spring fails to develop full recovery.   No doubt this would be considerably less than the final breaking strength.  However,  nylon, like sinew, is renown for its elasticity and so it seems reasonable to suggest that even a puny 3” spring can be quite powerful if we are not too timid in how it is pre-loaded.   Locking in maximum torque while the machine is at rest is largely dependent on the number of locking stations available to the tightening washer.   The Gallwey reconstruction has 32 notches , 11 ¼ degrees of rotation per locking station.  Using a 5 foot spanner I often find myself straining to get that one extra notch of tightening.  It follows that the closer together the notches are, the more pre-load can be put into the springs without having to let the washer slip back to the preceding locking station.  Field trials with the Gallwey showed how important it was to balance the torque between the springs to get good accuracy.  I found with that machine, if I used good body mechanics each time, my maximum physical exertion with a 5 foot spanner was a pretty accurate way of judging the torque applied to the springs. 





The value of introducing high levels of pre-load into the springs was well understood by the ancients. The first record of a “vernier” system can be traced to the hole pattern evident in the tightening washer of a second century ballista excavated in Ampurias, Spain just before World War 1.  It allowed the ancient engineers to lock in the amount of twist introduced to the springs to a fine 7 ½ degrees of rotation.   This level of rotational control also allows the two spring bundles to be more closely matched in power than was possible with my Gallwey ballista.   Clearly this was the way to go with the Orsova reconstruction.  With a spanner maybe eight or ten feet long, and the field frames locked into some kind of fixture mounted on the ground, it should be possible to take that 3″ bundle of nylon and torque it up to some ungodly level.  Taking it to the extreme I’m contemplating,  the rest of the machine will need to be as stout as possible.  I guess I’ll know I’ve gone too far when something breaks.



It takes about one minute for an old man to cock this contraption to the yellow mark on the side plate.  It seems like some of these notches are better than others in the accuracy department.  Need more work to test this premise for sure.  Sand bags are used to bed the rear end of the machine when actual accuracy testing is in progress.  Full power testing requires more elaborate safety protocols (i.e. no civilians in the area and a thick plywood pavis) than is apparent with this relatively tepid series of shots. 

Next Page »