01 Press Fit

I started this project by evaluating the unique capabilities of a CNC router & the different ways of creating press-fit objects. This yielded two lists:

Unique features of a CNC router:

  1. Enhanced precision
  2. Hyper surface articulation (like Gramazio & Kohler’s Digital Materiality)
  3. Increased complexity – intricate shapes

Ways of press-fitting:

  1. Puzzle joint
  2. Finger joint (dovetail etc)
  3. Slotted connection
  4. Dadoed connection
  5. Double slotted (like the interior of a wine box)
  6. Pinned connection (dowels etc)
  7. Custom connectors (I found this nice example)

Next I looked at the different joint geometries that press-fitting permits:

  1. Puzzle joints enable organically shaped geometries
  2. Finger joints demands a 90 degree connection (if you were to taper one side, the other side would need to accommodate the taper and would require a sloped profile)
  3. Slotted or dadoed connections require a 90 degree connection (if you were to taper the peg the pocket would need to accommodate the taper and would require a sloped profile)
  4. Double slotted connections require that the connection detail mirrors itself, and would suggest a 90 degree connection (a taper would not fit snugly)
  5. Pinned connections require that the pocket or hole created accommodates the profile of the pin
  6. Custom connectors enable organically shaped geometries

After this analysis I ran through a few meta ideas (and a few thought tangents). Should I design an object that incorporates every kind of connection? Tectonically rich, but a bit busy. Would a projected-finger joint (like Williams & Tsien’s Cranbrook Natatorium) stay in place in a press-fit object? Is it possible to create translucency in MDF or plywood? How thin would one need to mill it to get there? Could it be interesting to try and get to transparency at moments, but fail a little and leave the material failures as part of the finished piece? A bit more material research than press-fit project. Is it possible to move away from something that looks so two-dimensional? Could you create a more complicated shape with press-fits like a hypar (or like the one I mentioned earlier)? Would work with a thinner or pliant material, but wouldn’t work with 1/2″ material.

After moving away from each of those I decided that it would be interesting to pursue a module that press-fit at both the joint scale and the module scale. I was also interested in a design that could accommodate different sized spaces and still seem intentional (like this piece from Peter Marigold). I developed this design with that in mind:

After designing this I milled a few pieces of the design in MDF  to practice using the Technomill and to test the joints & overall scale of the design. I learned a great deal from this mock-up. I set the plunge and pass speed on the Technomill to 100 which proved to be too rapid for the brittle MDF. As the machine milled through the material it delaminated considerably, leaving me with tattered edges and absent ridges. The geometry, because of its complexity, took approximately 30 minutes to cut. At that rate it would have taken 8 hours to mill the final prototype. Finally, the scale of the milled pieces was smaller than I had anticipated: at both module scale and at the scale of the joint.

First time on the mill

100 speed is a little too fast

This test prompted a redesign. I increased the size of each finger joint by a factor of 6, and with that the scale of the ridges on the sides of each module. Beyond that I increased the height and width of each module to snugly accommodate large hardcovers in the height dimension and small paperbacks in the width dimension. I also changed the number of ridges on the side of each unit. In the original design there were an odd number of ridges and channels. In the redesign I made that number even to allow the modules to stack vertically. I felt that this capability strengthened my idea about one module that could be deployed in different ways to create faux-custom millwork.

Assembled and exploded module

This redesign also enables the user to selectively remove modules from the wall to create porosity, differentiation and to reduce the amount of material used.

Tilt and stack to fill any space

One version of the units aggregated

Close up on an absent module

Before commencing with the final prototype I did one more joint mock-up, this time using Plywood. I tested several different offsets to ensure a snug fit at the joints. I found that no offset, or the actual depth of the material (.46″), was the best fit. It should be noted that the plywood ranged from .43″ to .49″. After taking measurements from several different sheets at several different locations using the calipers I decided to average my findings, and use that as the “actual depth” of the plywood.

Testing tolerances: .44", .46" & .48"

After determining the required tolerances at the joints, I began milling the final prototype. Unfortunately, the initial file as I had laid it out would have taken 3 and 1/2 hours to mill (it was set up for a 1/4″ bit).

Because of this I removed some of the geometry from my sheet and cut only the tops, bottoms and backs of the boxes (the pieces without channels).

This cut the milling time down to 40 minutes. Or at least it would have. Unfortunately, when I zeroed the router in the Z dimension I must have been off by a hundredth of an inch because after the job had run, all of the pieces were still attached by about a veneer’s thickness of wood. This necessitated running the machine for one final pass with a corrected Z dimension.

A veneer's worth of material left behind

The next day, after recalibrating the file to use a 1/2″  bit for the channels, I came back to cut the remainder of the pieces.

This change reduced the milling time by nearly half. Or at least it would have. Once again, there was a slight problem that nearly doubled my time on the mill. I set up the file to cut the channels first using a 1/2″ bit and then to change back to the 1/4″ bit and cut out the profile of each piece. After running the first pass and beginning to cut the pieces free of the material I realized that the channels were roughly .02″ too small. This meant that the ridges between each channel were too large, compounding the problem. Once I had realized this I stopped the machine with the intent of milling a hundredth of an inch off each side of the channels. In the end this is what I did. Unfortunately, by the time I realized this two pieces had already been cut from the material and could not be re-milled. I believe that the problem with the original channels has to do with the setting I selected in VisualMill 6.0. I selected “pocket” for the channels rather than “profile: inside”. While what I did was technically the more logical thing to do it resulted in a pocket that was slightly too small. Other people in the class have experienced similar problems when using “pocket” cuts. Another possibility is that the bit could have been slightly smaller in diameter than the 1/2″ I was expecting. When I changed the cut style for the second pass the dimension of the channel moved closer to the desired dimension but was still slightly too small. Once again I stopped the machine and this time I changed the bit and ran it using the “profile: inside” command. This combination achieved the desired result. From this point I proceeded with the profile cuts to extract each piece from the material.

The first pass at routing the channels

Cutting the side pieces

Getting the dimension of the channels right

The sheet at the end of the day

After removing the pieces from the mill I began assembling them. To assemble them I snapped the sides and back together and then slid them into the base piece as a unit. After using the rubber mallet to secure the connection, I inserted the top piece and hammered that in to place.

First module assembled, pre-sanding

The pieces had not been sanded yet, for fear of removing enough material to upset the tolerances. Once all three modules were assembled I used a power sander with a 250 grit sandpaper to remove the errant wood fibers.

All three modules, post-sanding

Here are some photos of the finished prototype:

4.196 Special Problems in Architectural Design Complete Fabrications Nick Gelpi Mon-Fri, Jan 5-7, 10-11, 13-14, 18, 20-21, 24-25, 27-28, 01-04:00pm, 3-402/7-432studio, 1st mtg Wed 1/5 Pre-register on WebSIS and attend first class. No listeners Prereq: Permission of instructor ; Yr-1 MArch students who have completed 4.123 only Level: H 9 units Standard A - F Grading Can be repeated for credit Lab Fee: 150 A comprehensive introduction to methods of “making” explored through a wide range of brief but focused exercises. Skills = developing complex geometries from flat components; fine-tuning press fit construction, molding and casting, and making repeatable molds for customization. A two-part workshop, the first half will contextualize contemporary tools and techniques within the trajectories of historical case studies of building, combined with hands on familiarization of tools. The second half will implement the tools of our workshop in the context of Design. Working on group design build process for three MIT 150 FAST installations, students will test and influencing designs through the instrumentality of production. These hands-on design build projects are intended to produce reciprocity between skills and design, making more complete the problems of fabrication. Subject limited to year-one MArch students who have completed core-1 studio. Contact: Nick Gelpi, 9-224, 253-9415, -


%d bloggers like this: