Hints and Experiences in the Use of Quickslice

Some Experimements Concerning the Fabrication Phase

To resolve to my own satisfaction a few puzzling questions, I constructed a part "cubestack" whith which I then tested several things:

1.) What happens when there are missing slices ?
==> The machine does not deal with absolute Z, and deposits the higher slices correspondingly lower, right on top of the last slice of the lower set.

2.) Does program start/pause cause problems during the building process ?
==> No -- it continues without any mishaps.

3.) What happens when you manually adjust X,Y, after the programm has been started (during PAUSE) ?
==> The machine does not get confused and returns to the original coordinate system.

4.) How does a fine internal cross-hatch align from layer to layer ?
==> If the slice shape is the same in the same position, it does align. If the shape is moved, the cross-hatch also moves and thus no longer aligns. However, the emerging (soft) bead can span rather large gaps without much sagging.

5.) What happens if you build a part without a support platform, partially over a hole ?
==> It first messes up and piles up loose curled "spaghetti" in the hole area. After a while it recovers and seems to continue the building without too much distortion.

Implicit Boolean Operations

The STL boundary representation does not support Boolean operations.

However, the QuickSlice software is robustly enough designed that one can do implicit Boolean operations if one carefully watches the surface normal orientation of the different components in the Boolean expression. QuickSlice seems to do a nice job of determining the occupancy of material in each layer based on a positive winding number test. However it has the added feature to invert the sense of direction in the contour if it is entirely of winding number -1 (Which normally does just the right thing if some part descriptions have consistently inward pointing surface normals). In the context of Boolean operations between
two correctly specified shells, it may leave a few unwanted features that then have to be pruned away manually.

With a little care and finesse, the individual Boolean STL components can be loaded individually into the QuickSlice viewer. One has to be aware though of the automatic coordinate adjustements taking place at loading time: When QuickSlice loads a new STl component, it will do a translation so that the minimal coordinates in x, y, and z will be at x=0.1, y=0.1, and z=0.0 inches, respectively; this transformation is appliaed to all the geometrical contents already inside the viewer ! Thus parts that had been previously aligned to this preferred position, will now lie at a quite different location, and the separately loaded components my no longer have the registration that seem to be implied by the raw coordinates in the original STL files. One typically has to move the current content in the viewer to an x, y, z- location that corresponds to the coordinates of the part that will be brought newly into the viewer. Carefully check the resulting composite.

The stack of slices generated may need some editing and clean-up. If the Boolean operation performed is A-B, and if there are layers with only the B component present, these layers will show positive B features. It will be necessary to remove those layers manually. In this context, remember, that layers in the middle of the build structure that get removed in this manner will not lead to an empty slab of space; instead, that slab will be collapsed to thickness zero, and the upper occupied layers will be built directly on top of the lower one.

Do you really need a support base ?

By default, QuickSlice builds a "base" consiting of four wide ly spaced meandering support layers and one dense platform layer on which the actual part is then built.  One can choose to omit the construction of this "base" and build test parts directly on the foam mats. This is easier on the old pink mats (summer 99) than on the newer yellow mats (fall99). The model bead does not stick well to the yellow pads, and when it builds a dense layer, the back and forth motion will rip the loose the ends of the dense zig-zag runs, which makes the edges of the layer curl up. It tends to even out after 2-3 layers.

On the pink pads the bead sticks well enough to get a good first layer -- just make sure that the nozzle is brought down to the mat deep enough -- about 20/1000 of an inch INTO the mat. On the first pass, the nozzle should be carving the mat, and you should not see the first layer of beads "lie on top of the mat."

If one wants clean parts one should then take active measures not to build the part directly onto the red foam pad, but to start it a little higher up, so that there is some support material under ALL features of the part. Unfortunately, simply raising the STL part in +z will not create extended supports underneath. By adding a small 30 mil cube underneath the desired part, one can trick the system into building an extra few support layers. This is normally enough to isolate the part from the special situation near the beginning of the build. However one should remember that these supports start from z = 0.0, whereas the default base is constructed to start at about z = -0.02 inch.
Thus, when one omits the standard base, one should remember to set the nozzle at the start of the build process slightly into the foam pad by about 10 or 20 mils to guarantee good adhesion of the first couple of support layers. It is particularly important that all the bead roads are firmly fused to the foam pad, since there is no platform layer to form a rigid compound; badly attached support membranes, can cause trouble later in the build process.

A Clean Start

Often the first couple of layers of material are discolored and seem to be of a more brittle consitency. First I suspected that the
trouble might arise from this platform (hypothesizing that too much heat might be stored in the platform layer). However, the platform has nothing to do with it.  If the FDM machine has been sitting idle for a while, the filament close to the nozzle may have been heated for too long and changed its consistency and color.  Thus at the start of a new build, dispense some filament, perhaps as much as 1 minute worth, until it floows clean and smooth and shows the proper color.

After changing the color of the model material, you will see for a while a mixture of the old and the new color in which the old color will gradually fade away. The effect is particularly pronounced when white material follows blue; the blue color may still be visible over dozens of new layers ! -- I is actually neat to make such color graded parts.

Entering at the SSL Level

We have developed the capability to enter information directly at the SSL level, rather than just at the STL level. This makes it possible to resolve Boolean operations more cleanly than by the intrinsic use of oriented STL shells described above. We also hope to develop the capability to produce thin dense shells with just a loose, spidery fill on the inside in a more or less automatic way from the description of the outer geometry alone.

One idiosyncrasy that we discovered when working with the SSL level, is that the QuickSlice software is rather sensitive to coinciding points. In particular, -- and this is rather annoying -- when the first and the last point of a contour fall in the same place, the program acts up. This may happen even if the original data set has different coordinate values for these two points, since the QuickSlice program may reduce the accuracy of these coordinate values upon read-in.

There are other idiosyncrasies ! One would think that the proper description for a thin solid shell with a spiderweb fill would consits of three contours: the outher shell contour (red), the inner shell contour (red), and -- in the same place as the latter -- an outer contour for the filler zone (ligth-red). However, the coinciding red and light-red contours are not properly processed by QuickSlice. One can work around this by making the two contours slightly different. An easier way -- but less obvious -- is to make both of the coinciding contours in the light-red layer. It does not seem logical, but it produces the desired results when the roads are generated.

Page Editor: Carlo H. Séquin