Important notice: The CSA web site was re-designed in August of 2010. Some documents then available were considered complete and static; so they were not included in the re-design and were not updated. This is one of those documents. Information about dates of posting and revision remains here, but there will be no revision of any kind after August, 2010.
Using CAD to document existing items is a difficult process. Documentation is the goal here; the point is to preserve information, preferably in three dimensions, with accuracy and with appropriate precision. Therefore, the process is dependent upon and must begin with determining the geometry of the item under consideration; that is, the documentation begins with the survey.
The first question that must be determined by the project director is that of survey methodology; all other plans for gathering data will depend on the survey methodology. In addition, if there are existing plans or other survey information that will be used in conjunction with the new survey data, their use should be included in the earliest planning stages. Specific issues regarding precision and accuracy have already been discussed in the previous chapter of this guide.
If existing plans are to be used, the process of importing the information from paper drawings may have implications for data precision and accuracy. The scale of the original drawings, their state of preservation, the experience of the CAD operator, and the size of the digitizer will impact the results if drawings are to be traced on a digitizer. Similar issues will impact the results if scanned images are to be traced on screen, automatically traced, or retained in the model as raster images.
In any of these cases it may be necessary to experiment with the original data sources to determine the levels of accuracy, precision, and reliability that can be obtained. Simple experiments can be undertaken to simulate the process of transferring paper drawings into a CAD model or real drawings can be used. In either case, comparisons of dimensions on the paper drawings with those in the CAD model will indicate the levels of accuracy and precision to be expected. (One such sobering experiment is to place a mark on a piece of paper taped to a digitizer surface, using a real-world scale. Then use the digitizer selecting device to make a point at the location of the mark ten times in a row -- looking away between each choice and its successor -- and compare the actual coordinates of the points. That will show the limit on precision when tracing from drawings and suggest the level of acceptable precision if a drawing is to be used for data input. A number of variations on that experiment may be used for other potential problems.)
Along with survey methodology, the issue of data precision must be established at the outset. As discussed previously, the precision required for survey data is related to the nature of the survey subject. Although high levels of precision may seem necessary, that need is often illusory. One should not impose artificially high standards simply to avoid the need to explain why lower levels of precision are being used. The level of data precision to be established should be one of the first matters for discussion and determination. Of course, available or intended survey methods may impose restraints on precision, but, within the limits imposed by the survey methods, precision should be matched to the realistic demands of the project.
Different aspects of the work may yield different levels of precision. For instance, survey data may be accurate to the nearest millimeter, while data from a drawing of missing parts of the same structure may be have only a 5 cm. tolerance. In such a case, different layers may be used to hold items with different levels of precision; one could also use cross-referenced drawings. In either case, however, the operator must be sure to avoid using points from the lower-tolerance data source as reference points for other material, and the project manager must be sure to document the use of different levels of precision and how users can determine the levels used for various parts of the model.
If the model will be based on drawings rather than survey data, precision will be determined primarily by drawing scale, though drawing quality may also affect model precision.
Another issue to be determined is that of data density, that is, how close together data points must be. Surveying the surface of a stucco or plaster wall might require no more than finding its boundaries, since irregularities in the surface itself could be considered accidental. Surveying the same size wall, but one made of cut marble blocks, might require four points for each block of marble. A similar wall made of mud brick and therefore having a more irregular surface might require a number of points along each edge as well as points spread throughout the surface. Similarly, a rubble masonry wall might require many points along the edges and some points spread along the surface. As with precision, appropriate data density depends on the nature of the project.
Damaged remains also present difficult choices for data density. If a wall has been partially destroyed, how important is the specific line of destruction? Must one take a data point every centimeter along the broken edge, every meter, or . . . ? Similarly, a mosaic floor may exist in an undulating surface today rather than an original, nearly planar shape. How many data points are required to define the current surface adequately? When considering data density, the project director must make practical choices; every added data point complicates the model, the model making, and the end product. As a result, every added point adds cost. By the same token, any necessary point that is not recorded represents a much greater cost if it must be obtained at a later time, a greater cost still if it is never taken and consequently compromises the value of the model.
Data density will be a simpler matter when drawings are used as the core for a CAD model rather than survey data. Starting with drawings, the CAD operator will either use explicit dimensions, which will implicitly determine data density (making model points exist only where there are explicit dimensions), or follow lines to points of divergence from a given course. In any case, the decisions about data density will already have been made by the draftsperson.
Although both precision and data density should be determined in advance, that may not always be either practical or desirable. In some cases, it may be necessary to begin the modeling process in order to determine the appropriate level of precision and the appropriate data density. In those cases, a limited portion of the work should be done before the final decisions are made, and a definite cut-off date for making the decision should be established. The portions modeled during the testing phase should be chosen to help answer the outstanding questions about procedures.
Advance preparation should also include the decision as to whether the CAD model is to be a wire-frame model, surface model, or solid model. (As noted previously, models will rarely be exclusively of one type or another, but the intended dominant modeling style will guide many aspects of the work.) In some cases, it may even be prudent to make only plans and/or elevations, not a fully 3D model. Although a 2D approach is generally to be avoided, because it involves necessary simplifications, there are many times when basic 2D drawings, either plans or elevations, are appropriate. Many single-stratum archaeological sites may be appropriately documented with simple plans, and there are certainly other instances where 3D modeling serves no purpose. Even structures that should be fully modeled may have to be documented less fully, with plans and elevations only, simply for financial reasons.
If the project calls for 3D models, wire-frame models are the easiest to make but the most difficult to use. Hidden-line drawings cannot be made from them; so visualization is much more difficult. On the other hand, all the geometry is preserved in a wire-frame model; so all the data points are included -- though the differences between surfaces and voids, walls and windows, bounded openings and polygons may be very difficult to understand. In addition, wire-frame models cannot be effectively exported to rendering or virtual reality programs, but both surface models and solid models can.
Surface models are more difficult to make than wire-frame models, but they do not require more data. The draftsperson must work differently, since individual surfaces must be modeled, not lines in space. The results, though, are models that permit visualizations and hidden-line drawings.
Solid models are the most difficult to make and provide the most potential for structural analysis. More data are required, since a solid object must include all surfaces as boundaries of the solid, including surfaces not visible. (In addition complex features with multiple parts, a mud-brick wall with internal wood supports for instance, must have all internal parts separately modeled.) A surface model, on the other hand, may have isolated individual surfaces and need not include surfaces that cannot be seen. For example, a simple block in the midst of a wall and in the shape of a right rectangular prism can consist of only two surfaces (see Fig. 1). However, all six sides of that stone would be required for a solid model. A solid model permits the same visualizations and hidden-line drawings as a surface model; in addition, calculations about weight, center of gravity, and load-bearing capacity can be made (finite element analyses).
Generally speaking, CAD programs cannot automatically convert wire-frame models to surface models or surface models to solid ones. A draftsman should be able to use the data from a wire-frame model to make surfaces or to use surface data to make solids. However, that process is not quick or easy. The decision as to modeling mode should be made in the earliest stages of the project.
Projects considering the use of finite element analyses should consider use solid models. For the sake of economy, however, solids might be used for only a part of the project. If a proposed beam across a room must be modeled and examined for load-bearing capacity, for instance, the walls beneath the beam need not be modeled as solids.
Figure 1. Blocks in a wall drawn as solids (and completely surfaced) compared with using visible surfaces only.
Having prepared for the survey work, the project directors must also make choices regarding hardware and software. Hardware, of course, may depend on the hardware already at hand, but the most important factors for choosing hardware are actually the software and the need to provide data in digital formats that can be shared. That is, the first aim of any archaeological or architectural history project is the documentation, the CAD model in this case, in a form that can be shared with others; interpretation, of course, will follow as the culmination of the project. The suitability of the software to the task is crucial; so is the the format of the digital data files produced. The digital format must be one that other scholars can be expected to be able to use. As a result the CAD program, which determines the digital data format, is the critical choice; the hardware on which to run the appropriate program should be seen as a secondary choice that descends from the software choice.
The digital data format of the CAD model is a key element in choosing a program. Unfortunately, data format in the CAD world is a very difficult issue. There is no standard format for sharing CAD data, though there are three formats that have some claim to that distinction. A rather old international format (International Graphics Exchange Standard or IGES) was to have been a proper format for exchanging CAD data, but it has not been able to keep up with the improvements in CAD software. The drawing exchange format (DXF) was created by Autodesk to provide an output format from AutoCAD® that would allow users to exchange AutoCAD data; it has not been stable enough to permit confidence for complex models, though it is very widely used. Nor has DXF become a standard controlled by a non-proprietary group or organization; Autodesk can -- and does -- alter the format at will. Standards for the Exchange of Product Model Data (ISO 10303 standard, or STEP), a new standard still under development, is so broad as to be of little utility, at least at this point.
In the absence of a standard for data portability, users are left with only proprietary formats, despite the disadvantages of using them. The most widely used of the proprietary formats is the DWG format used by AutoCAD and developed by its parent company, Autodesk. (It should be acknowledged that Autodesk helped to sponsor the creation of this guide.) Because of its market dominance, AutoCAD's DWG file format has become dominant as well, and other manufacturers have purchased from Autodesk software that permits users to read and/or write DWG files in other programs. That is not always completely effective, though, since incompatibilities between programs -- either different programs or different versions of the same program -- can create data incompatibilities, even when there seem to be none.
Another common data format is the DGN format used by Microstation® (from Bentley Systems, Inc.). Although Microstation can also read and write the DWG format now, its native format, DGN, is used very widely.
A recent development is intended to make the DWG format into a public standard. The OpenDWG Alliance was formed in the late 1990s to parse the DWG format, maintain information on it, and to supply to its members software to read and write DWG files. Software incompatibility may frustrate this work as well, but it is conceivable that the OpenDWG Alliance (http://opendwg.org) will make the DWG format into a public standard.
Perhaps the best general advice on this issue was given by Peter Lyman and Howard Besser (in their article "Defining the Problem of Our Vanishing Memory: Background, Current Status, Models for Resolution," in the Getty Conservation Institute Publication, Time and Bits: Managing Digital Continuity, Margaret MacLean and Ben H. Davis ed., 1998, p. 15): "Save in the most common file formats. The more files that exist in a given format, the more likely that file converters or emulators will be written for that format (because of economies of scale)."
Since the choice of software is complicated by the file format issue, what is one to do? While the safe choice is AutoCAD, since its file format is the most common, one should consult personnel at the intended digital archival repository for the model to take advantage of their expertise and perspective. Not only will they have good advice, they may be unwilling to archive files in some CAD formats. Other issues to keep in mind when choosing CAD software:
Competence. Will it do everything you need it to do?
It is not always obvious that a specific program will or will not meet all necessary requirements for a given job. Generally speaking, the more irregular the material to be modeled, the more competent the software must be. Almost all programs can model very complicated geometric shapes if they are regular. The irregular shapes create the problems. Therefore, when examining software, modeling of irregular items should be very carefully investigated. The best test is to try to model typical material from the project or from a similar project with any program under consideration. (It is important to test CAD programs with the kinds of tasks actually expected. A large proportion of the commands available in a modern CAD program will never be needed by scholars. Therefore, it is important to test with real examples that reflect scholarly needs, not the needs of designers and architects.)
A particularly important issue is that of layer names -- or the names of the equivalent data segments used by the particular CAD program. If the layer names cannot be used effectively in the manner discussed below and briefly in the first chapter of this work (and explained more fully in the CSA Layer Naming Convention at http://csanet.org/inftech/csalnc.html), the model will be less easily used and much more difficult to analyze.
Ease of use.
Data input systems.
Data entry is often restricted by those programs that try hardest to use a graphical user interface. It is easy to draw with a mouse if one is designing. However, if one is entering data points from dimensions taken in the field, the keyboard is the essential tool. It should not be difficult to enter absolute data points or to define points in relation to others (either by relative coordinates or by direction and distance indicators). Here again, users should experiment with typical data from their own work to be sure programs meet their needs. It may also be desirable to consult with a colleague who is familiar both with the program under consideration and the kind of work to be undertaken.
Data entry with a digitizer, of course, requires that the CAD program include an interface system for accepting digitizer input. The system must not only permit the digitizer to be used in place of a mouse but to be used in the more sophisticated ways required to trace scaled drawings.
Data entry via a cut-and-paste process from a spreadsheet or database may also be desirable, as described previously. Therefore, the program chosen should permit that.
Stability of and support from the parent company.
Hardware requirements, if any. This is rarely a problem, but less popular software may not support a wide array of peripheral equipment.
If programs have relatively small markets, users may be restricted to specific graphics cards, plotters, or input devices. This can obviously cause availability and pricing problems. This is less likely with Windows and Macintosh systems, since the hardware interaction normally does not depend on the CAD programs, but there are still possible problems.
Software for data transfer from total stations.
If a total station is to be used for survey work, it is obviously necessary for the data collected in the total station to be transferred to the computer and the CAD model. Anyone planning to use a total station should be certain that there are programs that will accept data from a total station and format the data appropriately for the CAD system under consideration. (Data transfer from a total station is often accomplished with DXF files. They are quite acceptable for the kind of data collected with a total station, though, as noted above, they may not be for complex CAD models.)
Potential for connecting to external data tables. This feature is common enough that it may be used by vendors to recommend one program over another. However, the proprietary links provided by CAD software to external data tables should be avoided. Connections to external files should not be used except as outlined in Harrison Eiteljorg, II, "Linking Text and Data to CAD Models, "CSA Newsletter XIV, 3, Winter 2002 (http://csanet.org/newsletter/winter02/nlw0201.html). That system will work with any CAD program and does not require any data-linkage capabilities in the CAD program.
Training -- either through academic institutions or through CAD vendors -- is not readily available for all CAD programs, but good training programs are available for many of the more widely used programs.
Cost. Academic discounts can be very large.
Drawing output formats.
Publication will often require individual drawings from a CAD model, and, generally speaking, those drawings will be produced in illustration programs like Illustrator®, Freehand®, or CorelDraw®, starting with a CAD output file. That is, the CAD program will create a basic view and then produce a drawing file that can be manipulated with added text and other features for publication. Any program should be able to produce drawing files in a format appropriate for such publication plans.
Though some, including the author, may deplore the fact, many of the requirements noted above will drive the user to IBM-compatible PCs rather than Macintosh computers or workstations using UNIX, LINUX, or another operating system. In the end, the software must be considered more important than the hardware.
Other items of hardware for consideration: the computer itself, monitors, printers and plotters, digitizers. Changes in computers come too often; a recommendation for a computer for CAD work would not be good for more than a few months, if that. In general, though, the fastest computers are not needed for data entry. Speed bottlenecks with data entry come from the "wetware" (human brain) not the hardware. Once the basic data have been entered, considerable manipulation may still be required, but, with proper use of layers, speed or operation is not likely to be greatly affected by the speed of the computer. On the other hand, working with large models -- changing points of view, making hidden line drawings, rendering the model, and the like -- does require a fast computer, and it seems that each improvement in CAD software adds to the computer requirements. Regardless of the speed of the computer, a large hard drive should be included. Equally important there should be ample memory (RAM) as well as additional specialized memory on the graphics card.
If a laptop is needed for work on site, it should have a good color screen. A desktop machine should have the largest and sharpest monitor affordable. It should be as free from distortion as possible. An LCD monitor should be considered; though expensive, good LCD monitors are sharper and have more contrast.
Plotters are used to make paper drawings with either pens, an electrostatic charge, or ink jets. Large plotters are desirable, since they make it possible to produce very large drawings -- often up to thirty-six inches in width -- but it will be more economical to use a service bureau for large drawings if such large drawings are not needed often. A good ink jet or laser printer that will make drawings up to eleven by seventeen inches is much more economical, may be used for other jobs, and will provide excellent results. Similarly, color ink jets can produce excellent drawings in small sizes at modest prices.
One last factor should be decided prior to the beginning of work. If a database management system is to be used in conjunction with the CAD model, the specific database management program should be chosen. This is not the place to make a recommendation for a database management program, but assistance is available in the Arts and Humanities Data Service Guide to Good Practice called Digitizing History: A Guide to Creating Digital Resources from Historical Documents, by Sean Townsend, Cressida Chappell, and Oscar Struivjé (http://hds.essex.ac.uk/g2gp/digitising_history/index.asp).
The CAD program should not determine the database manager. The system described in Harrison Eiteljorg, II, "Linking Text and Data to CAD Models," CSA Newsletter XIV, 3, Winter 2002 (http://csanet.org/newsletter/winter02/nlw0201.html) separates the CAD model from the database in such a way that any CAD program can be used with any database program. In addition, notes may be connected to the model without a database program at all, using simple text files instead.
When advance preparations have been completed, equipment and software assembled, and survey procedures determined, the actual survey begins. As with any survey work, the most important part of the process is the note taking, even if electronic survey equipment is used. It is easy to assume that on-site sketches and notes are not needed if the process is "automatic," but that is not the case. Only if one is using an on-site survey-and-modeling system, one that builds the model as the surveying is done, may the note-taking process be reduced in importance -- and then only if the modeling is completed on site.
Sketches of the survey subject with indicators for data points must be made as the process goes forward. The easiest error to make is to assume that a data point taken today will be clearly understood without notes tomorrow. (Taking digital photographs can be very beneficial, providing photos to which notes can easily be added.)
If the survey process involves more traditional hand-measuring processes, the note-taking and the data gathering require equally careful attention. In addition, the actual survey process, assuming a 3D model is to be the result, requires added care. As noted elsewhere, more data points are required, and it is more difficult to make certain that the points are fully specified when three coordinates are required. Line levels and plumb bobs will be needed, since proper orientation on horizontal or vertical planes will need to be checked. With inclined surfaces, measurements of inclination will be needed. In general, users should be certain that the survey data are sufficient to permit the geometry of the subject to be fully specified. That may seem obvious, but recording for 3D modeling is much more demanding than recording for plans and elevations. It is remarkably easy to have a large number of data points with one or two crucial ones still missing.
Once the survey process has begun, the modeling should proceed in tandem with the survey work. This is especially true for those who have not previously constructed 3D models. It is almost inevitable that essential data points will be skipped or insufficiently specified until some experience has been gained. Even with experience, it is far better to be able to make the model as the survey process moves forward. (That makes the power of portable computers taken into the field more important, though it is possible to use less powerful computers by working with smaller files and combining them later, after returning to an office setting and stronger computers.)
As the model is built, it is the responsibility of the survey/CAD personnel to make notes to indicate the processes. Field drawings should, for instance, be annotated with the date(s) on which the items drawn were modeled as well as the name of the CAD operator. Data transfer files for electronic survey data should include dates in their names, and notations of work done and personnel involved should be kept in a log.
Data capture from existing paper drawings should be noted with dates of work and processes employed, the name of the CAD operator, and indications of precision achieved. If dimensions from the drawings are used to correct scanned or digitized drawings, the manner in which that is done should be carefully described.
Other work processes should be recorded in similar fashions, with dates, personnel, and processes noted.
When surveying, each day's data should be modeled before more data are gathered. The process of data modeling requires that the operator visualize the spaces under study, and that is not easy if too much time has passed since the survey work. It is more difficult if 3D models are being created, but even relatively simple plans will involve many points, not all of which are as obviously related to one another as they seem to be when the notes are taken in the field. (At a more mundane and practical level, it is also problematic to leave the field and assume that the necessary time will be found for model making before returning for more survey work. The backlog can then continue to grow uncontrollably.)
When using electronic data from a total station, only data points will normally be brought into the CAD model from the total station. (In some cases, 2D drawing information or wire-frame lines can be imported as well.) Building the model from the data points involves the following steps. First, the points themselves are brought into the model (usually on their own layer) with numbers or other indicators that permit them to be individually identified. Second, the points are used to define points in model entities (lines, surfaces, etc.). At the conclusion, both points and model entities exist in the model, but the points are no longer useful and may be removed. Nonetheless, the points are important to the process and should not be discarded completely.
There are at least two ways to keep the points without cluttering the model. One, the points can be brought into a related model (one using the same grid system and explicitly related to the principal model) and kept there. That would result in having two parallel models, one with the survey information (the data points) and the other with the actual objects. The other possibility for retaining the data points without cluttering the CAD model is to save the transfer files that are created to take data from the total station to the CAD system. In either case, though, once the data points have been used, they need not be left to clutter the model (and to make the file much larger). Their preservation is only for documentation purposes and to provide the potential for later accuracy checking.
If, as is likely, the total station coordinates are rounded during the process of moving the data from the data collector to the CAD model, the raw data, rather than the rounded-off numbers or points in a related CAD file, should be kept.
If total station data points are put into an intermediate file, a spreadsheet for instance, to permit model building in more efficient ways, then the file transferred to the intermediate program should be retained. So should the intermediate file.
In almost all survey work there will be some use of standard measurements with tapes and line levels, even if the major part of the work is done with more advanced equipment. When hand measurements are to be used in an area inaccessible to the advanced equipment, some fixed points within the area to be measured should be surveyed with the advanced equipment so that the relationships between the two areas can be properly specified.
Modeling processes in a CAD program depend on the specific program. However, if the model being constructed is a solid model, there are two general methods that may be used to specify the shapes of objects. One is called boundary-representation (b-rep) and the other is Boolean (also called constructive solid geometry or csg). As the names imply, the former determines the shape of the object by reference to the shapes of its bounding edges. Boolean construction involves adding and subtracting simple shapes to create more complex ones. Programs may permit some cross-over, but users should become familiar with the basic process permitted by their program long before beginning to build an important model. (It should be noted, however, that learning the intricacies of CAD modeling, as opposed to designing, is much better done with one's own objects and survey data than through any tutorial or sample process. Trying to match real geometry requires that the user understand that geometry at the outset and have some sense of what a good finished model should look like -- and what inadequacies are acceptable. This applies to learning to make both solid and surface models.)
Regular solids -- boxes, cones, spheres, cylinders, and the like -- can often be created directly with the tools of a CAD program. Unfortunately, most real-world objects are not regular -- even if they were intended to be -- so the easy-to-use creation tools will rarely be as useful as one might hope.
There are also two ways to build surfaces. Both involve the use of combined small surfaces to create larger, more complex objects. One process involves the actual construction of small, discrete surfaces that are then combined to make more complex objects. The other relies upon the CAD program's ability to construct the individual surfaces automatically when constructing larger geometric shapes -- boxes, cones, spheres, etc. As with solids, virtually any regular geometric surface(s) can be modeled automatically by most CAD programs, but the key word, once again, is regular. The real world is not full of regular geometric shapes. Thus, documenting real-world objects usually requires building large, complex surfaces from small, simple ones.
The small surfaces that will eventually make up larger ones must be planar; they must lie in a single plane. The simplest surfaces are triangular (three points may always be placed in a single plane, but more points may or may not fall in a single plane.) Even if a program permits the construction of more complex irregular surfaces without using individual planar facets, it can be crucial to have the ability to manipulate small surfaces in order to model very complex geometry.
Figure 2. Small surfaces used together to make a single, larger surface.
When complex objects must be modeled as related surfaces, this can be more problematic than may be apparent. For instance, the block shown here in Figure 3 (from the Mycenaean wall adjacent to the older propylon) was surveyed -- using photogrammetry -- to obtain the outer boundaries of its face, but its face is far from a simple plane. It is very irregular. The polygon (A) represents the boundary of the block. B and C are two versions of the next step in making the surface model -- adding surfaces to fill the boundary, using only the points that define the boundary. In each case triangular surfaces have been used to fill the boundary. A hidden line drawing, using either B or C would show only the boundary as shown in A. However, a rendering will show the facets made by the triangular surfaces, as in D and E. D is derived from the surfaces indicated in B; E from the surfaces shown in C. The ways the surfaces were drawn changed dramatically the appearance of the block in the rendering. Neither version is "right" or "wrong," but the CAD modeler must, at the least, understand that the way such a block is modeled will have an impact on the way it appears in a rendering.
Figure 3. Model of a single block from the Mycenaean wall near the older propylon. A shows the outline of the block. B and C are two different models of the block with triangular surfaces filling the outline. D and E are renderings based, respectively, on models B and C.
As model subjects become more and more complex and irregular, the ability of CAD systems to create fully accurate and precise models is more and more strained. At some point, one that cannot be defined in the abstract, the scholar in charge must decide that a limit has been reached, that a given level of detail is sufficient. This is not different from the process of making a paper drawing; in either case there is a limit to the detail expected or desired. As discussed previously, this is an inescapable problem with any model.
As the model is constructed, various portions of the model must be placed on different drawing layers. (See introductory chapter for basic description of layers.) The layering system used should be designed to distinguish material in the model according to all important criteria -- building part, building or site phase/stratum, material, chronological standing, scholar responsible for specific hypothetical additions, etc. Each layer should hold only a portion of the model, and putting too much of the model on a single layer will usually cause problems when the time comes to use the model for analytic purposes. Objects can be moved from layer to layer, but many objects on a single layer make it hard to deal with individual objects. More important, the way the model is segmented will have an enormous impact on the usefulness of the model -- how easily it can be manipulated, what questions can be asked of it, what portions isolated, and so on.
Using different layers for a model requires some system for assigning portions of the model to particular layers -- and a system for naming the layers. For example, an archaeological site model might have a unique layer for each locus, but there should also be ways to find all loci of certain types or dates. A model of an historic structure also needs many layers. Not only must each phase of the structure be placed on a different layer, it may also be necessary to segregate parts of a model on the basis of material, function, designer or builder, and so on.
In each of the foregoing cases, the model might be segmented in any number of ways. The segmentation should make it possible to find material according to multiple criteria -- e.g., parts of an historic structure made of marble and from the first phase and designed by the initial architect or loci from an excavation from a given stratum and containing hearths and belonging to a common time period. In other words, the layering scheme should permit users to access the layers very much as they might access parts of a database -- by asking for parts of the model that meet specific criteria.
The normal approach to layer naming is very straight-forward, misses the database approach described above, and is almost inevitably unsuccessful. Naive users will begin with layer names like wall and door, then graduate to wall1, wall2 and door1, door2, and then wonder how to deal with further complexity. As the names grow longer and more complicated, they also become harder to remember, and the end result is often chaotic and confusing. Furthermore, there is, at best, no way to find layers according to their characteristics, rather than their names. That is, one cannot select all the layers that contain cut-stone exclusively, or rubble, or any other particular material type -- or layers that belong to one or another building phase -- or layers associated with one or another scholarly reconstruction. The manner of naming layers must permit complex selections and must be carefully designed to do so. Thus, the system of naming the layers must be carefully designed to make the analytic possibilities into realities. Without it, a user must know all the layer names employed for a particular project and laboriously type in some subset of the complete list when trying to select specific portions of a model. Having done that, the user will still be unsure whether or not he/she has accessed -- or rejected -- all relevant layers.
The layer-naming scheme to be used should be carefully designed and specified as early in the project as possible. As the model grows, the use of the scheme will become more and more important. The scheme should not, however, be considered inviolable. If it does not work well, it should be changed. Before the system is changed, the model should be backed up and saved separately so that, if necessary, the original model and layer names can be checked. Once a new layer naming system has been established, all layer names should be changed so that the model is not cluttered with layers following different naming schemes.
The layer-naming scheme is critical to a CAD project. In a well-designed system the names will not appear to reflect the actual material on the layers, since they will not be mnemonics or abbreviations. The names will follow a strict system but may appear to consist of random characters. Therefore, documenting the layer-naming system is crucial, as is maintaining that documentation for other users. In addition, it may be desirable to use available features of the particular CAD program to create access mechanisms for specific groups of layers, all from a specific building phase, for instance. CAD programs generally provide one or more ways to accomplish this.
Some CAD systems make it quite easy to systematize the naming of layers, since they permit searches with typical "wild-card" search processes. Using the search potential, it is possible to create a layer-naming system that is designed to enhance searching possibilities and to provide superior access to the layers -- for ultimate analytic work as well as working with the model as surveyor/draftsman/scholar. CSA has developed such a system, utilizing search features and meeting scholars' needs; the CSA Layer-Naming Convention (http://csanet.org/inftech/csalnc.html) is available on line at the Center for the Study of Architecture Web site. The document provides both a general explanation of the way naming layers can enhance any project and a more specific set of suggestions for archaeological work.
Some disciplines -- architecture, for example -- have layer naming conventions intended for use by professionals, usually defined by their professional organizations. Those systems tend to be designed for practitioners and to meet specific, practical needs of the profession. The system described in the CSA Web site, on the other hand, is a conceptual scheme that permits the layers of any CAD model to be accessed according to logical analytic categories that are meaningful and useful for the specific project; it is even possible to use more than one conceptual scheme at a time. It is both more general and more adaptable than discipline-specific schemes, but it works well only with programs that permit "wild-card" searches for named layers.
English Heritage has presented a layer-naming system in the appendix of the publication The Presentation of Historic Building Survey in CAD. (Undated. Contributors: Dave Andrews, Bill Blake, Nigel Fradgley, Sarah Lunnon, and Paul Roberts. Available over the Web as two PDF files through the English Heritage site at http://www.english-heritage.org.uk -- apparently accessible only with Internet Explorer. Follow publication links to free publications to reach the download page.) The system described provides limited search potential, but it is a prescriptive list of layer names that happens to permit searches for some characteristics (but not others) rather than a systematic approach to using layer names for analytic purposes.
It seems that colors should also be used, like layers, to specify portions of a model. For instance, it seems logical to assign a specific color to a given period in a complex structure or to a given stratum in an archaeological site. This can be done, but one should not simply assign different colors to objects on the same layer of a drawing. Rather, the drawing entities should be placed onto appropriate layers first; then a specific color may be assigned to each layer. All entities on a given layer will then be the same color. The visual result is the same, but the process is different, because it is important to use layers, not colors, to hold analytic distinctions.
There are two reasons to resist the temptations to use colors, rather than layers, to hold meaning. First, it is generally much easier to change colors than to change layers, and inadvertently changing colors would be most unfortunate if important differences were denoted by color distinctions and could be lost in the blink of an eye. Second, colors are often used to direct the making of paper drawings in ways that require changing colors for different drawing effects, making the possibility of inadvertently changing color even stronger. (Print processes usually use color to determine the printed line weight or color. That, in turn, requires that colors in the model be changed nearly every time a paper drawing is produced, since each drawing is intended to serve a particular purpose and to emphasize particular points. Changing colors in the model is therefore a common practice.) Since changing colors in the model should not involve any risk of losing important distinctions, distinctions between/among portions of the model should be made with layers, not colors.
There are two levels of required documentation for a CAD project. First, there is the documentation done for internal use, the record keeping required for project personnel and for coping with the materials generated in the course of the work. Second, there is the documentation required by anyone external to the project who may ultimately need to use the model and associated data.
Some parts of this have already been addressed, but a list of documentation requirements, with some added notes, should be of help. (Asterisked items are required for external documentation as well.)
Any potential user of the project data (either the CAD model or the associated database information) must have documentation available in order to use the information. Archival repositories will require such documentation, and any user will need it.
Some would argue that all documentation required for the project is required for any user, and, once created, all the documentation should accompany the model. However, some portions seem to be unnecessary, specifically the back-up files, most of the personnel information, and the equipment information. The crucial pieces in the foregoing list have been asterisked. In addition to that documentation, however, a user should have a general statement of purpose and scope for the project. Such a statement may seem unnecessary, but it will often help to clarify methods and choices regarding precision.
Finally, there should also be bibliographic references that specify publications regarding the project and related to the project. Such citations should include data sets of potential interest. In no case should such information be available only through a Web site as a substitute for included documentation. Rather, it should be transmitted to a user along with the actual data files so that, should the Web site cease operation, the documentation will already be in the possession of anyone who had the data files. (Advancing technology may change this point.) This information is probably most easily contained in a text document -- in a non-proprietary format -- that can be automatically passed to a user along with the CAD and/or data files.
Some necessary information about a project falls into a different category; that is the information needed for indexing or searching for relevant material -- information for what is often called resource discovery. This is the more basic data needed by a potential user so that he or she can know whether to investigate the data further. The information is comparable to that which is placed on a library catalog form; it alerts the potential user to the potential utility of a resource for the purposes at hand.
This information is much more general and includes the following at a minimum:
For a fuller discussion of the information required for resource discovery, see H. Eiteljorg, II, "Where Is the Information?" CSA Newsletter, X, 2 (http://csanet.org/newsletter/fall97/nlf9707.html). Since that article was written, I have come to believe that the term metadata has too often been used in ways that confuse. Since its meaning is data about data, it should apply to both documentation required to make a resource useful and indexing information required for searching and resource discovery. People have used the term, however, to refer to one or the other category rather than both.
When digital materials are deposited in an archival respository (to be discussed in Chapter V), the repository will have a list of data items that are considered crucial for resource discovery, perhaps one from international efforts to define such requirements, the most well-known of which is called the Dublin Core.
|Continue to next chapter.||Go to glossary.||Go to Table of Contents.||Go to CSA home page.|