Following the success of the 1989 experiments and analysis of the results, another experimental campaign was planned and carried out in September 1992. The 1992 on-site team consisted of Prof. G. Martelli, Mr. (now Dr.) Sam Coles, Dr. Mario Di Martino and Walter Ferreri from the Turin Astronomical Observatory, and myself.
The experiments differed from those in 1989 in two significant ways. Firstly, two NAC fast-framing cameras were used in order to provide a stereoscopic view of the disruption and allow a three-dimensional (3D) analysis. Although presenting a far more complicated problem in terms of analysis, the possibility of being able to reconstruct the fragment ejection field in 3D is appealing and offers, amongst other things, the possibility of identification of fragment jets while in flight - previously any jets were only detectable on the ground after the experiments.
Secondly, these experiments involved simulated impact (again using a contact charge) upon pre-fractured targets - i.e. targets fabricated in separate pieces and assembled prior to the experiments[1]. The motivation behind the use of these pre-fractured targets was to provide some insight into the possible differences between solid homogeneous bodies and most asteroids, particularly binary asteroids which might well consist of separate pieces. A possible mechanism which might allow gravitational re-accumulation of large fragments is highlighted by the confirmed existence of anisotropic fragment jets by our group (see Section 4.7). Recently discovered asteroid 4769 Castalia appears from radar observations (Ostro et al., 1990) to be a contact or near-contact binary - given the small (~1 km) diameter of this object it is highly unlikely that two bodies combined by chance under the influence of their mutual gravity. Similarly, 951 Toutatis exhibits an unusual lightcurve, with one possible explanation for this being that it is a contact binary consisting of two or more large fragments of comparable size. Such an object could be the result of shattering of a single body without significant dispersion, or the accretion of two large fragments from an earlier disruption.
The targets used in the 1992 campaign were of the same size (21 cm), shape (spherical) and material (strong cement mortar) as those used in 1989. The pre-fractured targets were constructed in three separate pieces and assembled on-site in an orientation which was hoped would maximise the chance of detecting any anisotropies in the fragment ejection field.
Figure 5.1: Pre-fragmented target arrangement used in 1992. The diagram represents a cross-section of the target in the vertical plane - the fractures were axially symmetric spherical caps.
Figure 5.1 shows a schematic of a pre-fragmented target and steel table - this can be directly compared to Fig. 2.2, which shows the 1989 target arrangement. Figure 5.2 is a digitised photograph of the shot 6 target, showing the two boundaries in the pre-fractured sphere.
Figure 5.2: Close-up of the 1992 shot 7 target.
Two homogeneous targets which remained from the 1989 experiments were disrupted as shots 7 and 8 in order to (i) provide a reference for direct comparison to the pre-fractured targets and (ii) allow the more advanced 3D analysis to be applied to these older targets as an 'addendum' to the 1989 experiments. The assumptions made during the analysis of the '89 experiments (specifically those required as a result of the 2D nature of the analysis) will hopefully be validated by the eventual analysis of shots 7 and 8. The type of target used in each shot is detailed in Table 5.1.
| Shot No. | Date | Description |
| 1 | 24th Sept '92 | 2 'normal' fracture planes (i.e. not glued) |
| 2 | 24th Sept '92 | 2 glued fracture planes |
| 3 | 25th Sept '92 | 1 'normal' plane, 1 glued fracture plane |
| 4 | 25th Sept '92 | 2 'normal' fracture planes |
| 5 | 26th Sept '92 | 2 glued fracture planes |
| 6 | 26th Sept '92 | 2 fracture planes, artificially spaced at 1mm |
| 7 | 28th Sept '92 | Homogeneous |
| 8 | 28th Sept '92 | Homogeneous |
Table 5.1: Target types used in 1992. All targets were constructed from strong cement mortar and were 21cm in diameter, as in 1989. Fractures described as normal are those held by (Earth) gravity alone, with contact but no direct acoustic coupling between the separate parts of the target.
Three types of fracture were used in these experiments, the simplest of which ('normal') was arranged by resting the parts of the target on top of each other (these are therefore gravitationally bound, but not in the conventional sense!). A second type of fracture ('glued') used a fairly weak glue to join the inner target surfaces - sufficient to provide direct acoustic coupling while still placing a significant discontinuity at the fracture surface. The final type of fracture was made using a few 1mm spacers to artificially separate the parts of the target. It was hoped that these different types of fracture might exhibit various different properties under catastrophic fragmentation.
With the experience gained during the analysis of the 1989 experiments (using only one camera) we were well placed to expand our study to cover stereoscopic analysis. The two NAC fast-framing cameras used in these experiments were arranged at an angular separation of 60°, both 12 ± 0.1 m from the target table. This orientation was chosen in so that the cameras could provide sufficiently different views while still having some common components. Previous researchers have tried using two cameras at 90° separation (Nakamura and Fujiwara 1991) - while being important pioneering work these experiments have only resulted in the measurement of a few of the largest fragments.
The camera speed chosen for these experiments was 400 frames per second. This decision was based upon several factors:
Targets were placed upon the same steel table as used in 1989, but behind the table (at a distance of a few metres) were placed wooden panels, covered with a layer of foam polystyrene. These panels were intended to (i) provide a white background and thus reduce any errors in the analysis of the NAC film as well as (ii) hopefully 'capture' small fragment jets as they became embedded in the polystyrene. Figure 5.3 is a schematic of the arena, showing the target table, high-speed cameras and backing panels. Also shown is a CCD camera which was used to provide a 'quick look' analysis immediately after the experiments, though no measurements have been taken form the 8mm video films recorded by this unit. The most significant use of this film has been in confirming the identity of each of the shots, since (i) the cardboard labels (visible in Fig. 5.3) which we attached to the table are below the bottom of the frame in the NAC high-speed film, and (ii) the target identification stripes are in colour, whilst the NAC film is monochrome.
Figure 5.3: Schematic (not to scale) of the 1992 arena showing twin NAC cameras, CCD fast-shuttered camera and white backing panels. The numbering of the NAC E-10 cameras, namely with #1 on the right, is a result of Walter Ferreri's unilateral choice when labelling the film rolls!
Figure 5.4, below, shows shot 7 on the target table - the wooden panels can be seen behind the table, with large stones on their feet to hold them vertical in the strong wind.
Figure 5.3: Target table and backing screens as used in the 1992 experiments. The plywood circle around the base of the table (also visible at the bottom of Fig. 4.11) was marked with angular divisions to facilitate fragment position measurements. The dark line in the left foreground is the cable used to electrically detonate the plastic explosive (the contact charge).
In order to maximise the quality of the digitised images, the NAC film was professionally transferred in England from 16mm film to VHS video. This has considerable advantages over the in-house system described in Section 2.6, despite an increased cost. The result of this and the improvements to the experimental procedure has been an enormous increase in the data which can be collected from each shot. An unexpected drawback of the improved image quality, inevitable with hindsight, is a lower compression ratio for the video data once stored on disc. Coupled with the increased number of images overall, the data storage requirements begin to look rather daunting - while each shot recorded in 1989 constitutes approximately 10 Mb of data, the 1992 shot 6 data alone takes up 140 Mb of hard disc space. This is fortunately not a significant problem now we have access to a CD-ROM writer with each disc, costing only a few pounds, able to hold 700 Mb of data.
Figure 5.4: An example image pair from shot 6, immediately prior to detonation of the contact charge. Note the 60° separation between the left (top) and right (lower) images and the significant improvement in image quality over the 1989 data (see e.g. Fig. 3.9).
Figure 5.4 shows the initial images from each NAC camera in the shot 6 study (which is the only fully digitised shot so far) and gives some idea of the high quality of these newer digitised images. Camera 2, the left hand camera, has been placed at the top of the screen because this is more natural. As can be seen in Fig. 5.4 the stripes on the target simplify the visualisation of the target from the two different angles. Also apparent is the white panelled screen behind the target; this is not level but curves slightly down behind the target table (see Fig. 5.3) where the ground was not flat. Figure 5.5 shows an example of the disrupted shot 6 target, and can be compared to Fig. 3.6 which is a similar image from the 1989 films.
[1] It is important to make a distinction between pre-fragmented (as used by our group) and pre-shattered targets (as used by Ryan et al., 1991). The former are fabricated in separate pieces and assembled before impacting once, while the latter are prepared by impacting and re-building solid targets.
Fig 5.5: Two images of the highly disrupted shot 6 target. This is the twentieth image pair, i.e. 
of a second, or 0.05 s (± 0.0025 s) after the detonation of the contact charge. As in Fig. 5.3 (but less obviously) the top half shows the left hand view and the lower half the right hand view.
Several problems arose during the digitisation of the 1992 film. The first was that the increased clarity of the video images highlighted any inconsistency in the video grabbing - this was usually manifested as a slipping from left to right, of a few picture elements (pixels), from one frame to the next. The joins in the backing screen clearly moved between images, making this very obvious. The problem was solved by 'brute force' - since it was clear that some images were being sampled perfectly, without any horizontal positioning errors, I just repeatedly re-sampled the images until a complete run had been grabbed without errors. This increased the time required to digitise the images, but not by much since the re-sampling is a fairly fast process. Any software patch to tackle this problem - perhaps by sliding the image left or right a little under user control - would have required too much development and 'user' time. To date only shot 6 has been fully digitised; the process of digitising and sorting these images took approximately 3 days including these corrections.
The second problem is somewhat more fundamental and cannot be solved directly, though it can be worked around.. It arises from the fact that the cameras were either running steadily at different speeds or (more likely) that they took different lengths of time to reach the required speed and had not quite done so by the time of the first useful frame. The discrepancy was found by chance when testing the HV-3 program (see next chapter) by running it through all digitised frames - a process which takes quite some time even from hard disc. A well-defined event, when a large fragment hits the steel table, was seen to occur at slightly different times in each view. Figure 5.6 shows the four frames which are critical to the calculation of the error, i.e. the start and end frames, in each view, of this particular timing test. The clock shown in each image is the Burnt-In Time Code (BITC) added to one set of tapes during transcription - the display shows hour, minute, second and frame. Since the films were digitised with both odd and even fields the same, the VHS video frame count corresponds to the NAC frame count with an offset.
Figure 5.6: Four frames from the shot 6 films which reveal an inconsistency in the timing of the high-speed film sequences. Note that these images have been rendered 'unscaled', i.e. 512 pixels wide, and have been trimmed to half height. The vertical difference can easily be corrected within HV-3.
It is straightforward to translate the BITC into a simple integer frame number. The time of flight for the large fragment (just hitting the table in each view) turns out to be 353 frames for the left view (which is also the left hand NAC camera) and 340 frames for the right view, suggesting an error of approximately 4% in the timing over this period. A correction for this known error can, in principle, be applied to the timing calculations within any analysis software. However, we do not know which film is correctly timed, and indeed it is unlikely that either film sequence is precisely correct. The NAC E-10 camera is nominally accurate to X% and it is unlikely that the film sequences suffer from any large systematic timing errors, so this 4% error figure should be borne in mind and applied where necessary.
As mentioned in section 5.3, the polystyrene panelling was expected to catch some of the smaller ejected fragments, possibly highlighting some small scale jetting or similar phenomena. After each shot, the panels were photographed with the intention of digitising the images and possibly carrying out some kind of analysis of the impact patterns. Fig. 5.7 shows a typical example - the 35 mm negative has been put onto Kodak Photo CD, this format producing a very high quality scan. The resulting digital image can easily be copied from the CD for analysis.
Figure 5.7: A typical fragment impact pattern. The original images are 24-bit colour scans at a resolution of 3072 ´ 2048 pixels, i.e. quite adequate for high resolution analysis. This image has been slightly processed in order to make it suitable for inclusion here (see Appendix 1.1).
No analysis has yet been carried out on this set of data; to perform such a study will probably require a special computer program and considerable development. However, a visual inspection of the panels after each shot suggested that some considerable anisotropies were indeed present in the ejection field of small fragments, and that a formal analysis of these impact patterns would be worthwhile.