Digital::Analog—A Report from the Harvard Graduate School of Design
by Stephen M. Ervin, FASLA

The last twenty years have seen a huge effort at “digitization” – turning physical artifacts and sensor-readings into binary (digital) form (scanned photos, digital videos, satellite images, LIDAR data, 3D scans, etc.) This has occurred using various kinds of tools and techniques mostly to manipulate these digital models with software.

The process of going from digital to analog has been more limited. Digital drawings and renderings are plotted out at an increasingly prodigious rate, to be sure. The Harvard Graduate School of Design (GSD) consumes acres of bond and semi-gloss paper each semester, and every digital file is primarily displayed as light or sound energy (i.e., in analog form). Only in the last decade or so has fabrication really come into its own as a digitally controlled process. The “Fab Lab” phenomenon has increased the awareness and proliferation of digital fabrication tools. Today, many schools and offices have their own laser-cutters and 3D printers, which were considered quite experimental and prohibitively expensive just a decade ago.

At the GSD, we have been experimenting with four major kinds of digital-analog devices over the past decade: laser-cutters, 3-D printers, CNC machine tools (e.g. routers), and most recently, robotic arms in the CAD/CAM workshop fabrication laboratory.

Laser Cutters

At the GSD we deploy three Universal brand 24″x36″ laser cutters for student use, and they are heavily used, albeit more actively by architects making structures than by landscape architects or urban designers making spaces. The laser cutters have changed forever the speed of production, accuracy, and overall quality of topographic models, as well as reducing notably the number of X-acto blade injuries in the studio each semester! Laser cutters, which can cut up to ¼″ soft material (e.g.,cardboard, cork, plexiglass) can follow virtually any 2D pattern software can produce, such as Illustrator or VectorCAD. When set at lower power, they can burn or scribe the material surface, rather than cut all the way through, for a range of etching-like surface drawing and patterning/texturing effects. Laser cutters can create arbitrarily complex curves for topo lines, and every brick paving pattern burned in or window-opening cut out makes for wondrously precise and compelling chipboard models.

Because laser cutters burn their material, a ventilating system for smoke and vapors is required.  With our first several installations, we needed a substantial investment in ducts, blowers, filters, and vent stack, which often far out-weighed the installation cost of the laser cutter itself. Our most recent installation uses a self-contained filtration system, eliminating the need for a vent to the outside, and making it far more feasible to imagine a laser-cutter in every studio room, or at least one for every 12 students—on the current wish-list at the GSD!

Ervin Fig. 1
Chipboard model, laser-cut

Ervin Fig. 2
Chipboard model, laser-cut 

Ervin Fig. 3
Chipboard model, laser-cut

3D Printers

Laser cutters are the workhorses for all design students making topo models and structural assemblies. However, they are inherently limited to 2.5D stacked models, flat planar sheets (possibly cut, scored and folded to form 3D), and silhouette cutouts. A fully realized 3D object, conceptually composed of small voxels, analogous to 2D pixels, is possible only with the 3D printing (often now called “rapid prototyping”) technology. The first of these, introduced 20 years ago, was “stereo-lithography,” in which a laser beam was focused in a 3D volume filled with a polymer resin that was hardened in place to form the 3D solid model.
That heritage still lives on in the so-called “STL” files that are now the conventional standard for creating 3D water-tight models (in which water-tight is a criterion for “topologically consistent and possible to build,” a requirement for such models.) Software such as SolidWorks, DeskArtes "3DataExpert" or the open-source Mesh-Lab, are specially designed to create, analyze, modify and repair STL files.

Today’s more common technology uses a vat of powder and a modified ink-jet printhead delivering hardener (instead of ink) to form 3D solids in successive passes of  about .001 inch, building up hardened powder/polymer in each pass. It is even possible to embed color in these models now, and to use different plastic polymers and UV-hardening for even more robust finished pieces. These are still limited to about a cubic foot in volume, and are relatively expensive to create (the powders and polymer resins are approximately $5/cubic inch, which translates to tens to hundreds of dollars for a finished piece that you can hold, usually in one hand.)

At the GSD, we have three different 3D printers:  two “ZCorp” gypsum powder-based (the fastest); one “Dimension” ABS (polymer); and one “PolyJet” fused deposition (the latter two producing the strongest finished pieces).
Combining these plastic/resin 3D models with laser-cut cardboard site models is very much state-of-the-art for presentations these days. The combination of advanced CAD software that is capable of complex curved forms (e.g., Rhino, or CATIA) together with 3D printing, has made for the proliferation of “organic” forms in building and landscape that are hard to draw, and hard to model well with just a laser cutter.

Questions of the real value to landscape architecture of such expensive, somewhat fragile, and complicated modeling processes, and the role(s) of 3D printing in professional LA practice, are far from settled. That said, our graduates will have an increasingly informed opinion and experience on which to base a judgment, or an innovation. The current merging of built and natural forms, often complex and curved in shape, is especially amenable to visualization and design using these techniques.

Ervin Fig. 4
3D printer

Ervin Fig. 5
3D “prints”

Ervin Fig. 6
3D “prints”

Ervin Fig. 7
3D “prints”

CNC Milling/Router 

Whereas the laser-cutters are limited to 2′x3′ and cardboard and the 3D printers can only make fragile models, we also have some industrial-duty Computer-Numerically-Controlled (CNC) machines that work at larger sizes and with stronger materials. A typical milling machine is basically a fancy drill-press, that can drill holes and carve metal at any angle by hydraulic and mechanical control of carbide bits, and can cut shapes and forms out of wood, metal, ceramic, and glass.

Similarly, a CNC router (we have one with a 4′ x 8′ bed) can drive an industrial-duty router with a variety of cutting and shaping bits over a rectangular area, much like a laser-cutter. The router has the added dimension of some amount of  “Z” (depth, or “plunge”) over which the tool can be operated, giving rise to 3D (or, technically, 2.5D) models out of plywood, or most commonly, dense foam. Such routers have been used for some time by boat-builders and professional topo-model building services. In this case, the terrain is not built up out of layers of chipboard, but carved whole, in relief, from the foam material, constrained only by the amount of Z-axis, usually less than 20″. Matching these cutting technologies with “LIDAR” or other grid-based data sets (DEMs) is a natural for landscape and urban models. Last year at the GSD, Professor Paula Meijerink had all MLA students in second semester cutting 3D topo models using the router as a component of their studio project.

Ervin Fig. 8
4′x8′ CNC Router

Robotic Arm

Finally, in a category that includes chronology, cost, complexity, and sheer power, we have been using several robotic arms in the CAD/CAM workshop/fabrication laboratory.

These industrial robots (technically, six-axis robotic manipulators, from ABB Corporation) are superficially similar to a human arm and hand: a multi-segmented “arm” with a controllable set of bends and rotations, terminating in a “hand” that can be any of many tool ends: grabber, welder, drill, paint-gun, etc. The motion of the arm and tools is numerically controlled and can be accurate to within extremely small tolerances. We have two—one with a reach of only about half a meter, that is capable of moving an object or tool, up to about 6Kg., and the other with a reach of several meters, that is capable of a load up to 60 Kg. Used extensively by auto manufacturers and other heavy industries, applications of these devices in the field and in construction and design are still being explored.  One professor at the GSD, Inge Rocker, has led a group of students in programming the arm to pick up and stack blocks—like bricks—in mathematically derived patterns. (See the "serpentine wall.")

Other applications include cutting with a “water-jet,” a tool using grit delivered by a very-high-pressure water stream, to cut steel, glass, and stone, etc. We have produced complex topographic forms cut in marble and granite by this technique.

Ervin Fig. 9
ABB Robotic arm, (small in class)

Ervin Fig. 10
ABB Robotic arm (large in class) at work

When I first came to the GSD in 1989, I had an Apple Macintosh II, the school had an installation of Sun Unix servers, a ComputerVision CAD system, and a legacy of GIS software development and use in the studio, although almost no CAD there. In twenty years, we have gone from that limited and cumbersome state to being awash in digital devices from USB keys and PDAs to 60″ plasma screens, and all manner of PCs (including an ever-increasing number of Macs). We have gone almost entirely from celluloid to silicon slide collections (although we still keep a few pairs of slide projectors running), and “pin-ups” are increasingly “screenings.” Amidst this proliferation of bits, bytes, operating systems and file formats, the continued primacy of physical artifacts, now re-energized by the kinds of digital-to-analog converters described above, is telling—and no doubt reassuring—to some.

Just as landscape architects have come to grips with software and digital data,  workflows, and interoperability, now landscape architects are confronted with the opportunity (and constraints) of CNC fabrication, rapid prototyping, and 3D scanning. This has occurred while advantages, disadvantages, affordances, costs, and benefits have been roundly debated and discussed and never settled (for no one answer applies to all cases—although almost all projects now have some Photoshop component). What role(s) these technologies will play in practice and education, going forward, besides just making fabulous chipboard topo models, is yet to be seen.

Stephen M. Ervin, MLA, PhD, FASLA, is Assistant Dean for Information Technology and Lecturer in the Landscape Department, at the Harvard Graduate School of Design, where he teaches digital landscape modeling. He can be reached at
CommentsComments(0)  |  Print PDFPrint PDF  |  Send to a FriendEmail Newsletter
PPN Home




David Leonard, ASLA, Chair
(619) 294-4477, ext.130