Data copyright © Harrison Eiteljorg, II unless otherwise stated
This work is licensed under the ADS Terms of Use and Access.
Harrison
Eiteljorg, II
Center for the Study of Architecture
Box 60
Bryn Mawr
PA 19010
USA
In order to understand the CAD model of the NW wing of the Propylaea; the drawings, survey data, and database from which it is descended; and the processes used to move data from the original sources to the model, it is crucial that a user understand fully the information sources, the survey methods used, the data storage and transfer processes, and the ways these enabled certain procedures to be used. These methods also determined the precision with which the model could be constructed.
This model has been constructed from two very different bases: the survey data obtained concerning the interior wall surfaces for the east and south walls, and the data published by Tasos Tanoulas (and Maria Ioannidou and A. Moraitou) in the form of drawings of the interior surfaces of the west and south walls, three drawings each, with the coordinates for each measured point, one value type per drawing.
The survey data obtained as part of the CSA Propylaea Project work provided points for many but not all blocks on the east and south walls. Not only were some blocks not surveyed, details of moldings had not been properly surveyed when the project was stopped.
There will be no attempt here to discuss or describe methods used in photogrammetric work, 3D scanning, or surveying without a hand-held target; the level of discussion and description required for understanding those methods may be found in the relevant CSA Newsletter articles listed on the "Final Report, CSA Propylaea Project Publications and Lectures" page, and those methods were not used here for the survey work generating our final results. This document does contain full and explicit descriptions of the methods used for total-station surveying with the assistance of hand-held survey targets, the methods that were used to generate the data for the CAD model of the east and south walls of the NW wing of the Propylaea. The methods were used because the survey team — primarily Mr. Eiteljorg, Mr. Kapokakis, and Mr. Kanellopoulos — determined that surveying without a target would not provide the necessary precision and that the other methods were impracticable for the reasons discussed in the CSA Newsletter articles and, more briefly, in the short history of the project.
To begin, this survey system assumed that every survey point would be taken with the aid of a hand-held survey target. A drawing of the target may be found in the CSA Newsletter article (Winter, 2004; XVI.3) entitled "A Final Survey Experiment on the NW Wing of the Propylaea." The target was printed on very thin, clear acetate so that the person holding the target could see through it to place it in the proper position and so that the thickness of the acetate would not impact the precision and accuracy of the survey data. A reflecting mini-prism could be positioned at the target center to provide the actual target for the total station's infrared beam. (The total station telescope was aimed at the target center before the mini-prism was put into position so that proper aim was based upon sighting of the acetate target, not the mini-prism.) Furthermore, the target had measuring scales on two edges so that, when required by problematic conditions, an offset from the point desired to the point actually surveyed could be measured without another tool. The center of the target was slightly indented with the aid of a sharp implement so that the reflecting mini-prism could more easily be positioned and held there.
The surveyors positioned the total station at the beginning of the process and, using fixed points as back-sights, located it very precisely. Then one of the archaeologists, generally Mr. Eiteljorg or Mr. Kanellopoulos and occasionally David Scahill, got into the right position for the point in question (on the man-lift when the points were high), held a target in place, determined the block number and the corner (eg., LR for lower right) for note-taking purposes, and then, after the operator had properly aimed the total station telescope at the target, added the mini prism and held it in place while the total station operator surveyed the point. The operator would then supply the point number, and the archaeologist not occupied with holding the prism and target recorded the point number and the information about its location. Blocks with bands,1 bosses, or other unusual surfaces required added information, of course. More detailed information about the note-taking and about various surfaces encountered may be found in this CSA Newsletter article (Winter, 2004; XVI.3), "Note-Taking with Total Station Survey Work." Note the point made in that article that notebook drawings were often more valuable than notes made at the same time. Additional information about the notebooks and PDF files with scanned images of the notebook pages may be found at the survey notebook page.
The use of offsets from points — always measured along the wall surface on a horizontal or vertical plane — was kept to a minimum, and the distances were always small, making it easier to assume that the measurements really did fall on one of the prescribed planes, or close enough that, at small distances, any error would be negligible. Using the offsets required that, when making adjustments from the surveyed point to the desired point, the CAD model-maker could apply the offset in the proper plane — one of the planes determined by real-world horizontal and vertical and the plane of the wall in question (two different walls, at one time the east wall and at another time the south wall). Thus, the person recording the data would simply note whether the offset was right, left, up, or down, and the distance (which was never more than 15 cm. and rarely more than 5 cm.). The notes did not specify that the plane underlying the offset measurement was the wall under study; that was assumed. The required adjustment to the data-point coordinates from the total-station reading was left to a later date.
As a result of this procedure, we had survey data based upon the local survey grid and adjustments based upon the planes of the two walls. Combining the two forms of data was accomplished later. (The survey team's north/south/east/west directions were determined by a local survey grid and might have been used without modification, but the 0,0,0 point was well away from the area of interest and the two walls needed different survey grids for uniform processing. Two temporary grids or coordinate systems were used as alternates for the sake of applying offsets, one for the east wall and one for the south wall, each taking the wall surface as defining the x-axis. For those who have long since forgotten their high-school geometrically, this may not be the best place to explain the process; so please see the page about coordinate systems for a fuller explanation.) Then the data were combined into a single-source database from which information for making the CAD model could be taken.
The process was as follows:
The database is also available as part of the data from the project. It is a FileMaker database consisting of multiple tables, and information about the database may be found at this page with database documentation.
Precision of measurements can be determined by reference to total station specifications and the like, but much depends upon the careful work of the survey team. In this work Manolis Kapokakis and his team were extremely careful in setting up with the back-sights, and I believe it is safe to say that the resulting data are as precise as reasonably possible with today's technology. That means, in my view, that the individual points are within a mm. or two of their true locations. (Five points were apparently surveyed twice in the course of this project, not intentionally. The differences between the two survey results (in mm.), expressed as difference in x,y,z were as follows:
The average discrepancies were .8 mm., 1.2 mm., and 1 mm. for the three axes. This suggests a reliable level of precision in the range of 2 mm., since two points separated by a single mm. in each axis are 1.7 mm. apart.) More important, it means that point-to-point measurements should be within 2 mm. of their true separation. It also means that point-to-point information may be used to make careful calculations about the construction of the building.
Comparing this method of surveying to manual methods involving fixed planes from which measurement may be taken with tapes, it seems safe to say that these measurements should be substantially more precise.
Accuracy is a different matter. Here the issue is simply whether or not the right points were surveyed. If the target had been improperly placed, the coordinates would be incorrect, not imprecise but incorrect. There is no way to know when such errors have been made unless they are so gross as to be visually apparent when the model has been finished. Substantial care was taken, there being multiple people involved in the target-locating process. As a result, we are aware of no need to assume inaccurate points arising from improperly-placed targets.
Inaccurate survey information might also be the result of improper note-keeping. While it was clear when the data were examined that some errors had been made in the recording processes (for instance, recording a point as on the right rather than left side of a block or vice versa), those errors were obvious and could be flagged for correction. Since there were never fewer than two people working to identify and target the points surveyed and the model does not show gross errors remaining, it seems reasonable to assert that there are no outright errors remaining. It must be acknowledged, however, that the combination of note-taking in Athens and database-building in Bryn Mawr is not ideal for error prevention.
That is not to say that all points are both accurate and precise. Some points were missed or impossible to survey with the equipment at hand. In some cases, as a result, points were interpolated. The database should make clear which points have been interpolated to anyone wishing to check. In general, when interpolation has been used, the operating assumption has been that the blocks were very regular. Otherwise, interpolation is effectively impossible.
The east and north walls had been surveyed earlier by Tasos Tanoulas, and the coordinates were published in drawings in Study for the Restoration of the Propylaea, Vol. I, by T. Tanoulas, M. Ioannidou, and A. Maraitou; Athens, 1994, figures 45 and following. Three drawings were published for each wall to provide coordinates, each drawing showing a coordinate in a single axis so that, taken together, all three coordinates were provided. State plans (actually elevations in this case) were also prepared and published so that true conditions on the walls could be ascertained. The drawings used to present the coordinate data were, by comparison, quite schematic.
Taking these measurements required hand measuring with tapes and fixed lines or planes as starting points and also required hand-recording the measurements on paper. (Note that multiple reference lines/planes were used in the course of measuring. Therefore, individual measurements in the model may appear to conflict with nearby measurements when, in fact, they are simply based on a different measuring grid.) While the precision of this method may be very high, there are reasons to argue that precision cannot reach the mm.-precision level. In addition, the recording process is one that invites error. The recording process is the most problematic part of the work because it is so easy to transpose numbers, misplace a decimal, or otherwise err. Since the dimensions must then be copied onto more finished drawings, the opportunity for errors in copying adds to the problems. As with the total station, the problems are more likely to be ones that impact accuracy than precision. That is, regardless of the precision with which measurements have been taken, errors in writing or copying will create inaccurate rather than imprecise numbers. (The CAD process results in a model directly derived from the numbers, and the result is something that can be seen and visually examined for obvious errors. The hand-drawing process for the wall blocks did not start with the numbers in the same fashion; so the errors would not be obvious in the drawings.)
Making a CAD model from the kinds of numbers provided in Tasos Tanoulas' drawings — x-, y-, and z-coordinate value for each point — is not as easy or as error-free as one might hope. Needless to say, everyone makes errors in measurement; so there are some of those. In fact, however, they are both few and usually fairly obvious because, when present, they yield clearly incorrect results, as when the lower corners of a block have been measured as above the upper corners. In addition to the measuring errors, though, there are other possibilities. For instance, handwriting can be obscure, as we have all learned at one time or another by failing to read our own accurately — and we do not know how many generations of copies were made between making a measurement and publishing the results. On occasion the actual location of the measurement may also be less than crystal-clear.
Of course, there are also missing numbers when no measurement could be taken.
These problems can lead to confusion.
Missing numbers on a drawing are quite different from missing numbers in a CAD model. In the former case, the drawing can and should be seen as schematic, with the absent number making clear the absence of hard data. In a CAD model, on the other hand, the model (and any drawing made from it) IS the numbers. A missing number makes modeling (drawing) an object impossible. Therefore, it was necessary to fill in a missing number with . . . .
Often the choices are obvious. If there is no value for a given location, the comparable value from a nearby location that ought, assuming geometric regularity, to mimic the missing one may be used. In the case of two candidates for a value, they may be averaged. If no easy substitute can be found, the only remaining choice is to omit the object in question. Insufficient data yield a poor model. ("For want of a nail . . . the war was lost.")
Mr. Kanellopoulos made the CAD models of the north and west as two individual models, one of the north wall and another of the west wall. He had access to Mr. Tanoulas, other drawings, and the monument at the time. Susan C. Jones and Mr. Eiteljorg brought those CAD models into the overall model of the NW wing, in the process checking the coordinates as the transfer was made. It was difficult to know precisely how to react to the discrepancies found. Since we did not have Mr. Kanellopoulos' access to other drawings, not to mention Mr. Tanoulas and the actual monument, we were obliged to make choices when there were discrepancies. If we read Mr. Tanoulas' handwriting differently, for instance, we might change Mr. Kanellopoulos' coordinate to our reading. But that is not always a simple choice. If we could see that Mr. Kanellopoulos' number was an alternate reading and assure ourselves that we were reading the number correctly (and if the difference was very small, as was the case generally), we could change his number to match the one from the drawing. On the other hand, if Mr. Kanellopoulos' number was different from the drawing number and the drawing number was clearly wrong, we would keep Mr. Kanellopoulos' number even though we could not be sure of the source. This we did on the assumption that he had a reason for his choice, though we might not know that reason. On other occasions we saw the number differently but did not believe we could be sure that our version was better than Mr. Kanellopoulos' reading. In such a case, we kept his number.
There are also numerous instances (all the peritenia bands on the west and south walls, for instance) when Mr. Kanellopoulos modeled portions of the wall for which we did not have the necessary data.
Finally, there are instances of apparent errors for which there seem to be no logical corrections. In such cases we must assume that there is an explanation that is simply beyond our reach.
All these explanations are provided to emphasize the point that the data for the south and west walls are the best we can provide but not perfect. And to point out that we can claim no uniform method by which all judgments were made. Individual choices were made to reflect individual concerns. These issues are very important here due to the unique problem with CAD models and other forms of digital data: their seeming infallibility as a positivist record of the monument coming from an unbiased source. They are data, after all. So who will look at the data retrieved from such a model and ask the necessary questions, exhibit the necessary skepticism? Such a skeptical approach is required, though, particularly since corrections, when made, have been very much one-off choices, choices made because of the individual circumstances of specific issues.
So we will do our best to explain various concerns and choices here for each block where some choice was required on the south wall and the west wall, not including simple choices of different readings of numbers. Please note that this discussion assumes the coordinate/measuring systems used by Mr. Tanoulas for the original data. (We did not make the same kinds of choices on the east and north walls. Instead, we either did or did not have data and therefore did or did not model a given block. The exceptions were missing corners defined by related corners, and the explanations for those choices are provided in the database.)
1. The peritenia band is the smooth band along an edge of a block that has already been finished to its final surface, unlike the remainder of the block that awaits final finishing to trim the surface back (to the level defined by the surface of the peritenia). The peritenia is carved in advance or when the block is set in place only at corners where carving it during final finishing would be problematic and where the complex meeting surfaces make it preferable to carve the finished surfaces at the time the wall is built rather than during the period of final finishing. Return to text.
2. "Lifting boss" is in quotation marks because the term, though widely used, has long been deemed questionable by many scholars. It has now been thoroughly discredited by A. Trevor Hodge in "Bosses Reappraised," in Stephan T.A.M. Mols and Eric M. Moorman, eds., Omni pede stare. Saggi architettonici e circumvesuviani in memoriam Jos de Waele. Studi della Soprintendenza archeologica di Pompei 9, Electa Napoli and Ministero per i Beni e le Attività Culturali, 2005. Return to text.