NOTE: Some parts of this article are outdated. They will be updated once full-power tests have been completed. Please follow Samuli's user blog if you're interested in latest develoment in this project.

This reconstruction is based on analysis of the original Cheiroballistra text as well as available English editions of the cheiroballistra (Marsden 1971: 206-233; Wilkins 1995: 10-33). Ideas from a number of earlier scholars, especially Iriarte (2000; 2003), are utilized where they seem to make sense. Additionally I've used the manuscript diagrams available in Schneider's (1906) and Wescher's (1867) editions. Analysis of the archaeological finds is mostly based on the numerous publications of Baatz.

In this article and reconstruction I've made a few underlying assumptions and followed a few key principles:

  • Pseudo-Heron's (P.H.) cheiroballistra text is assumed to be more or less complete. No parts are assumed missing, unless it's certain that the reconstruction can't work without them. If a certain component is not described in much detail, it is assumed to have been well-known to the ancient artificer reading the text or so simple, that it required little explanation.
  • Archaeological finds are given preference to text in case of ambiguities. They are, however, only used to make design decisions where the text fails. Dimensions from archaeological finds are not used, even if they seem similar to those in the text.
  • The goal has been to make the reconstruction fit the text, not vice versa. This principle is followed as far as reasonably possible. For a good example of this principle, see the discussion below about the tenons of the rungs in the little ladder.
  • I do not try to hide problems in the source material or in my own theories. Therefore I've tried to make it clear what we really know and what is subjective. My goal is to make (constructive) critique as easy as possible, not to protect myself from critique by not mentioning the issues I've encountered.

The cheiroballistra text is a description of interrelated ballista components. Although the assembly instructions are very incomplete, the fact that parts must fit together helps a lot in making the reconstruction correct. If an incorrect change is made to dimensions of some component, it is likely that problems arise elsewhere. There are two possibilities to coping with this. The first option is what some scholars unfortunately seem to do: hang on to their assumptions and force the sources to fit them. For discussion of this issue see Iriarte's JRMES article (2000: 56-57). The second option is to question one's underlying assumptions and think "outside the box", trying to find the most logical explanation to the problem. I have tried to follow the second option according to my best ability. I've also tried to follow the principle of Occam's razor according to the best of my ability.

One big problem with the work of many previous scholars is that they have ignored the limitations of the metalworking techniques and tools used by the Greeks and the Romans. To arrive at a realistic reconstruction, these need to be taken into account. Again, there's a good example of this in the little ladder section.


The most useful cheiroballistra editions can be summarized as follows:

  • Marsden's (1971: 212-217) edition and English translation. Does not contain any manuscript diagrams.
  • Wilkins (1995: 5-59) edition and English translation. Contains most of the manuscript diagrams.
  • Wescher's (1867: 123-134) edition and Latin translation. Contains many manuscript diagrams.
  • Schneider's (1906: 142-168) edition and German translation. Contains photographs of many manuscript diagrams.
  • My own English translation. While certainly not the best translation/edition there is, it's the only modern translation that's freely available on the Internet. Moreover, it's released under a Creative Commons License allowing it to be used and improved with very few restrictions. For availability of the other editions take a look at the bibliography page.

In addition, several articles discussing the cheiroballistra have been written. Most noteworthy is the Iriarte's JRMES article (2000: 47-75). His follow-up article published in Gladius (2003: 111-140) also contains useful information regarding inswinging ballistas, including the cheiroballistra. All articles of Baatz are very useful, because they contain descriptions and pictures of archaeological finds belonging to late-Roman cheiroballistra-style ballistas. As these archaeological finds clear up lots of the confusion in the cheiroballistra text, Baatz' contributions have been extremely valuable to the research.

Wilkins' JRMES articles (1995; 2000) and his small book, "Roman Artillery" (2003) require special mention. They should be used with caution for two reasons: Wilkins' decided to reconstruct his cheiroballistra as a winched weapon and as an outswinger. There are very little evidence supporting either of those interpretations, so the authenticity of Wilkins actual reconstruction is questionable. That does not in any way diminish the valuable contribution he made by making another English edition of the cheiroballistra and by interpreting some of the cheiroballistra's components in a way that stands the test of time. Also, Wilkins' versions of the manuscript diagrams are of excellent clarity.

The availability of these sources varies greatly. A few of them can be ordered or bought from the Internet. Some are freely available. Some are nearly impossible to obtain without some creativity and help from fellow enthusiasts or friends in world's most highly rated universities. This is unfortunate, as it greatly limits the people who can - in practice - study this fascinating subject. Take a look at the bibliography page for more information about the availability of the various sources.

Main controversies

Outswinger or inswinger

Archaeological finds strongly suggest that the cheiroballistra was an inswinger so I've reconstructed it as such. This issue has already been discussed in detail here.

Winched or not winched?

The cheiroballistra was almost certainly a personal weapon and as such did not have a winch. The reasons for this are discussed in detail on this page.


All measurements are in Greek dactyls (1,93cm). One Greek foot is 16 dactyls. The way dimensions are marked in the CAD drawings requires explanation:

  • Dimensions which are clearly stated in the cheiroballistra text are marked in green. Even these dimensions may be suspect as it is not always clear what P.H. means by width, thickness, length and breadth. That said, the vast majority of these dimensions can't really be questioned.
  • Dimensions which roughly know are marked in orange. These are the ones given in the text as "about x dactyls". This applies mostly to the slider width.
  • Dimensions which are derivable from other dimensions are marked in magenta. These dimensions are not stated in the text, but can be calculated from dimensions of other parts. This applies especially to thickness of the field frame bars, which are referred to throughout the text.
  • Dimensions which are entirely subjective and not given in text are marked in red. These dimensions are the ones which have allowed scholars to reconstruct the cheiroballistra as a winched weapon without amending the text too heavily.

In the few cases where the clearly stated or roughly known dimensions have been amended, the following notation has been used:

  • X d (Y d)
  • X d (Y d)

Where X is the amended measurement and Y the original measurement stated by P.H.

Overview of the cheiroballistra

The contents of this section have been taken from my article "The cheiroballistra - Producing a viable weapon based on historical manuscripts, archaeological finds and experimentation" (Seppänen 2014). An outline drawing of the weapon is shown below:

320hz ballista overview 3 - small

Main components

The wooden case (1) forms the core of the weapon. The case has a female dovetail matching the male dovetail of the wooden slider (2). The rear-end of the slider has the triggering mechanism composed of several steel parts: the trigger (3), the claw (4), the fork (5), the pitarion (6), the handle (7) and a steel rod (8) The steel field-frames (9) and washers (10) made from bronze or steel house the sinew torsion spring bundles (11), through which the arms composed of wooden cones (12), steel bars (13), soft iron hoops (14) and wrappings (15) are inserted. The little ladder beams (16) are made from steel and held at a proper distance by wooden rungs and crosspieces (17). The little ladder is braced against the wooden projecting block (18) under the case and attached to the case with T-clamps (19) made from steel. The notches in tenons (20) in the little ladder beams are locked into the the field-frame bars inside the lower pi-brackets (21) and tightened using wooden shims and wedges (22). The field-frames are further stabilized by the little arch (23), the ends of which are inserted into the upper pi-brackets (24) and held in place by pairs of pins (25) and wooden wedges (26). The crescent-shaped piece (27) is attached to the end of the slider to serve as a stomach-rest during cocking. The bowstring (28) is inserted into the hooks (29) at the end of the bars.

Preparations for use

Each cord in the torsion spring is stretched using the winch in the stretcher. The power output of the weapon is directly proportional to the amount of pretension applied to the cords during this phase. Once the torsion spring bundles are full of cord, the arms are inserted between the two halves of the springs. The washers at the top and bottom of the field-frames are rotated against the direction of the arm rotation to increase tension further and to ensure that arms are rotated synchronously during pullback. Finally the washers are locked in place using pins (30) going through holes in the washer rim (31) and in the field-frame rings (32).

Using the cheiroballistra

The cheiroballistra is fairly simple weapon to operate. The trigger is first pulled from under the claw. The slider is then pushed forward, the claw locked to the bowstring and trigger pushed under the claw. This way the bowstring is locked to the slider. The slider is then braced against a sufficiently hard surface, and the operator pushes the weapon with his belly while simultaneously pulling the handle with both hands. This rotates the arms in the torsion spring bundles from their forward-pointing position, first towards the case, and then towards the operator, for an arc of 90-120 degrees. Once the slider has been fully drawn back, the handle is pushed through the steel rod attached to the case, so that the slider is locked into place. Finally a bolt is inserted into the groove (33) in the slider and pushed between the fingers of the claw against the bowstring. The cheiroballistra can be aimed accurately by bracing the left elbow against the hip and by placing the crescent-shaped piece behind the neck from the right side. The right hand is thus free to operate the trigger. Using this technique the weight of the weapon is actually an asset in that it stabilizes the weapon a great deal. The point of balance of the cheiroballistra, which is near the projecting block, also helps stabilize the weapon.

Cheiroballistra parts


Content moved to the Cheiroballistra case, slider and crescent-shaped piece article.


Content moved to the Cheiroballistra case, slider and crescent-shaped piece article.

Crescent-shaped piece

Content moved to the Cheiroballistra case, slider and crescent-shaped piece article.

Little ladder

Contents moved here.

Little arch

Content moved here.

Field frames

Content moved here.



Triggering mechanism

Content moved here.


Content moved here.

Assembling the components


Correctly reconstructing the cheiroballistra involves assembling the components so that they work perfectly together. If some of the individual parts are misinterpreted, problems almost certainly arise when assembling the components. Marsden (1971) and Wilkins (1995) encountered a number of these problems because they had arbitrarily changed various dimensions of the cheiroballistra. I've used the relationship and interaction of the components as a guide: if the components don't seem to fit together, there more likely an issue with the interpretation rather than the sources themself.

Alignment of the field-frame bars

Content moved to Cheiroballistra field-frames article.

Attaching field-frames to arch and ladder

Attaching the little arch

All archaeological field-frames have four Pi-brackets attached to the field-frame bars. We can say without a doubt that the cheiroballistra was an inswinger. We can also say with reasonable certainty how their field-frame bars were aligned. With some help from archaeological finds - especially the Orsova one - it's relatively easy to see the most obvious way to attach field-frames to the little arch (viewed from the top):

Attaching the little arch - top

The forked ends of the little arch pass through the Pi-brackets attached to the field-frame bars.

Attaching the little ladder tenons to Pi-brackets from outside of the bars

Unfortunately no remains of little ladders remain. In archaeological field-frames the lower pair of Pi-brackets is usually larger. One possible explanation for this is that their little ladders were made of wood. In cheiroballistra, however, the Pi-brackets are all the same size, so the little ladder was almost certainly made from metal similarly to the little arch. If we follow the example of the archaeological finds, the Pi-brackets have to be placed outside the field-frames. This means the tenons of the little ladder have to be spread out like this (top view):

Attaching the little ladder tenons to Pi-brackets outside the bars - top

There's probably no issue with this approach from structural perspective, but it means we have to make changes to the measurements P.H. gave us. This is the best point to discuss the critique Wilkins had against passing the little ladder tenons and the forked ends of the little arch through the Pi-brackets.

Wilkins (1995: 34) is correct in that the field-frame bars can't be made to fit neatly between the little ladder beams or the ends of the little arch. This is because dimensions given by P.H. are somewhat too small (see Marsden 1971: 215; Wilkins 1995: 24, 28). It is also true that both the arch and the ladder could have been simply made wider from the beginning, as Wilkins (1995: 34) says. That said, this problem becomes worse the thicker the field-frame bars are, and Wilkins' (1995: 20) bars were 9mm thick. In my reconstruction which has 4mm bars the gap in the little arch is only 0,34 d too small, which is easy to correct with a small bend. Also, as Iriarte (2000: 62) points out, it was impossible for P.H. to give an exact figure for the distance between forked ends of the arch, as it depended on the thickness of the field-frame bars which was left for the blacksmith (or engineer) to decide. In any case the little ladder is definitely too narrow: it should be 1d wider to fit over even my thin field-frames. However, as Iriarte (2000: 57-58) points out, the exact form of the little ladder beam tenons is unknown: this may be enough to explain the confusing dimensions P.H. gave us.

All this said, we can't ignore the most logical solution to the dimensions P.H. gave us: attaching the little ladder tenons to Pi-brackets inside the bars.

Attaching the little ladder tenons to Pi-brackets inside the bars

If we want to follow P.H.'s description closely, we have to make them slightly differently from the archaeological field-frames. This may be a lot to stomach for most scholars, but we should not forget that none of the existing field-frames belong to P.H.'s cheiroballistra. Therefore we can't take it for granted that it's field-frames were of the same kind. The easiest and most logical solution is to pass the tenons of the little ladder through a pair of Pi-brackets placed inside the field-frames. This option is shown in below diagrams, first from the top:

Attaching the little ladder tenons to Pi-brackets inside the bars - top

The green circle represents the diameter of the cord bundle. The same setup from the front:

Attaching the little ladder tenons to Pi-brackets inside the bars - front

If we place the lower Pi-brackets inside the field-frame all measurements fit perfectly, unless the spring diameter is arbitrarily increased (e.g. Wilkins 1995: 24). In fact, the width of the end of the little ladder beams is given as 1,25 d, and the width of the inside of the Pi-brackets as 2/3 d. This means that if we simply fold the ends of the beams along the long axis, they fit neatly inside the tenons (as shown above).

Even if the Pi-brackets are placed inside the field-frames, there's still a ~3mm gap between the the Pi-brackets and the torsion bundle before the arm is inserted. Experiments will show whether this is enough.

The only evidence against this interpretation comes from the manuscript diagrams, which definitely show both upper and lower Pi-brackets facing the same direction - away from the field-frames (see Schneider 1906: 154-115; Wilkins 1995: 18; Iriarte 2000: 54). The text itself does not state where the Pi-brackets are attached to (***reference***). This means that at this point we are forced to adjust the evidence to meet our expectations.

Whatever the truth is, Wilkins' use of little ladder's width as evidence against using the Pi-brackets for their most intuitive purpose does not sound very convincing (Wilkins 1995: 34).

Use of wedges

Most scholars seem to agree that wedges were used to bind the little arch, the little ladder, pi-brackets and field-frame bars together. Practical tests have shown that a pair of small wooden wedges per pi-bracket is enough to keep the little ladder in place during shooting. When aiming for maximum power levels pin or such may have to be passed through the wedges to prevent them from becoming loose in repeated shooting.

For the little ladder these two wedges are not enough by themselves. The problem is that the forces exerted upon the ladder are much greater than those affecting the little arch. The simplest way to fix this is to cut small notches to the ends of the little ladder tenons:

Little ladder tenon in place

The little ladder beams are bent outward during assembly, so that the notches snap into the field-frame bars tightly. After this a pair of wedges is hammered between the tenon and the pi-bracket to rigidify the construction. A small piece of hardwood could be beaten between the side of the tenon and pi-bracket would make this construction even stronger.

An elaborate wedge system such as that described by Iriarte (2000: 62) does not seem necessary.

Attaching little ladder to the case

The projecting block under the case supports the little ladder, preventing it from moving towards the operator during use. The role of the fairly light T-clamps is simply to keep the little ladder from falling of the case and possibly to keep it aligned sideways. This method of attaching the little ladder is not only very simple and intuitive, but it has also proven itself in practice. This does not mean that differing opinions have not been expressed in the past.

Wilkins (1995: 11) interpreted the block as an attachment point for the base whereas Marsden (1971: plates 7-8) ultimately ignored it. Both placed the little ladder close to the forward end of the case where it was not supported by the projecting block. This meant that the dimensions of the T-clamps had to be increased, because they had to bear the brunt of cocking the weapon. In addition, this soon lead both into issues with draw length: even when the arms were drawn to the maximum, the slider was not entirely pulled back. The only way to fix this issue was to lengthen the arms, which allowed longer draw length (Marsden 1971: 226; Wilkins 1995: 33). As the length of the "cones" or wooden portions was known, only the metal hooks could be lengthened without contradicting the text. However, lengthening the hooks beyond the cones is a bad idea for several reasons.

Below the little ladder, case and T-clamps from the side:

Attaching little ladder to the case - side

And from the front:

Attaching little ladder to the case - front

A few interesting things can be seen from these diagrams:

  • There's no reason to change P.H.'s measurements (3d long, 1d wide) like Marsden (1971: 225) and Wilkins (1995: 35) did. Of course, both of these scholars assumed that the cheiroballistra was a winched weapon, for which small T-clamps were not adequate, especially when the T-clamps are not supported by the projecting block.
  • The distance between T-clamps is given as 2.5d, probably across the case as Wilkins (1995: 29) assumed. This means that the edges of the T-clamps press against little ladder crosspiece, which thus provides additional support for them.
  • If we assume that T-clamps don't project above the case, their width must be 0,5d.

Assembling the triggering mechanism

Content moved here.

Defining the initial arm angle

NOTE: This section is outdated. To see why, look at the Cheiroballistra arms article.

There are several factors that affect the draw length of the cheiroballistra. We know with relative certainty the location of cord bundles (see this and this) and the claw. It is also almost certain that the slider was fully draw back before the shot, and likely that the handle was pushed through a nail(?) in the case to keep it in place. This allow us to check if our assumptions about the cheiroballistra are even remotely possible. For example, if the bowstring can't be draw far enough to catch the claw, we have probably made a mistake somewhere.

The location of the bowstring at full draw depends on a number of factors:

  • Arm angle at rest: determines the bowstring length
  • Arm angle at full draw: determines when the bowstring movement stops
  • Length of the arms: lengthening the arms increases the bowstring movement

Below is a set of diagrams illustrating this issue. First, inwinger with relatively limited arm movement:

Inswinger with small arm arc

Then one with a lot more arm movement:

Inswinger with large arm arc

And finally the maximum amount the cheiroballistra design can hope to achieve:

Inswinger with maximum arm arc

As can be seen, fully drawing back the bowstring is only possible if the arms rotate the absolute maximum amount, roughly 170 degrees. There's only one potential issue with this much rotation: the arm hits the curved field-frame bar at a ~60 degree angle when 90 degree angle would be safest for the bars.

Samuli.seppanen 17:05, June 19, 2011 (UTC)

Community content is available under CC-BY-SA unless otherwise noted.