The challenge: Create a 3D printed case mold that incorporates a plaster section just for the finished surface.
Top right: The secret is M3 brass threaded inserts in pyramid-shaped 3D-printed mounts (I have just pressed them into the 4.4mm dia, 10mm deep, -3 degree tapered holes using a soldering iron). These brass/plaster pyramids embed into the plaster to provide a threaded hole that M3 bolts can screw into.
Upper left: We made a cross-section CAD drawing of a three-piece demonstration mold (upper left). The top plate has holes for the M3 bolts, air escape and natch clips and recesses for clamps to hold a 3D shell, with flanges, in place (not shown).
Lower left: The plaster pyramids have been screwed onto the upper plate.
Center left: The plaster was poured, and over-filled, then the top plate, with embeds, pressed down on top of it. After set, the plate was unscrewed and removed.
Bottom right: The plaster section has been reattached and natch inserts and embeds put in place. The plaster was not sanded or prepared, this is a demo.
What is this all about? A full master case mold, utilizing this technique, coming soon.

This is the most complex shape known that can fit together organically, without a pattern (dubbed the Einstein tile, it was discovered by mathematicians in 2023). It has six sides of 1 unit length, six of 1/1.73205080757 and one side measuring twice the latter. Placing the tiles is tricky because it is only logical to seek a pattern, but there is none. One method is to start with a center tile and move outward in a spiral, being ready to backtrack and place them upside down when a piece cannot be fit. The tile shape is a product of connecting four identical irregular pentagons - each made by cutting a regular hexagon into three pieces. To achieve the maximum precision, 3D print multiple cookie cutters and let them stiffen and shrink in the cutters. To round the upper corners use stretch wrap) when stamping (print cutters in both orientations if doing this). Tiling a floor or wall will present issues with the number of edge tiles that need to be cut. However, cutting at least some custom edge shapes is practical because only 90 and 120 degree angles are needed.

G2926S reduces the thermal expansion of the popular G2926B (a durable, crystal clear, easy-to-use general purpose cone 6 base glaze for stoneware and porcelain). However, some porcelains (e.g. these Plainsman P300 mugs) need the lower thermal expansion this offers (to avoid crazing). This recipe adjusts "B" chemistry by adding low-expansion MgO at the expense of high-expansion KNaO (while maintaining gloss). This is more expensive to make (because it calls for Frit 3249 or equivalent) - use it if G2926B (with 325 silica) fails an IWCT test for crazing. These mugs were fired using the PLC6DS firing schedule, the S glaze was opacified with 10% Zircopax and the outside glazes are G2934Y silky matte with added stains. Need to reduce COE even further? Try G2926J.

This inside glaze is G2926B (on Plainsman M340). It is capable of firing glassy smooth, crystal clear and un-crazed even on coarse stonewares. Watch the video 📹 to see the four unusual things we do to get reliable glazes like this. But the recipe is only part of getting success. Mixing it as a thixotropic slurry is another. And the firing schedule: Look closely at the two glazed tiles. The bottom one, although fired lower (cone 5.5) was slow cooled using the C5DHSC schedule - note how much smoother the glass is (the upper one was fired to cone 6 using the PLC6DS schedule).

The outside is a floating blue, GA6-C. These are a dime a dozen but a good transparent is priceless. Did you know that the outside glaze can be made from the inside one by simply adding 2:4:1 iron oxide:rutile:cobalt oxide? This glaze can be stained, opacified and variegated in an infinite number of ways. And it is adjustable (e.g. lower thermal expansion, lower or higher melting).

This is a "badlands" slope in the Frenchman river valley. The valley exposes the "Whitemud Formation" in many places (clearly visible here part way down on the left). Two surface mines of Plainsman Clays are nearby, in a place where lower-lying rolling hills leave much less over-burden to remove. These materials were laid down as marine sediments during the Cretaceous period. The skeleton of the world's largest T.Rex, dubbed "Scotty", was found 50km east of here (in the layers just above the Whitemuds). Where are the layers of Scotty's ancestors from the Jurassic period? Straight down 1 kilometer! And another kilometer to bedrock!

These are the original cone 6 Perkins Studio Clear (left) beside our fritted version (right). You cannot just substitute a frit for Gerstley Borate (GB), they have very different chemistries. But, using the calculation tools in my account at insight-live.com, I compensated for the differences by juggling other materials in the recipe. I even upped the Al2O3 and SiO2 a little on the belief they would dissolve in the more active melt the frit would create. I was right - a melt-flow GLFL test comparison (inset left) shows that the GB version flows less. Using this on ware exhibited another issue (after doing a IWCT test): Crazing. The very good melt flow on my G2926A fritted version is thus good news: It can accept more silica - the more silica, the more durable and craze resistant it will be. How much did it take? 10% more! That ultimately became the recipe for our standard G2926B cone 6 transparent.

This is G2934Y (a version of the G2934 cone 6 matte base recipe that supplies much of the MgO from a frit instead of dolomite). Like the original, it has a beautiful fine silky matte surface and feels like it would not cutlery mark. But, as you can see on the left, it does! The marks can be cleaned off easily. But still, this is not ideal. The degree of matteness that a glaze has is a product of its chemistry. But can we fix this without doing any chemistry? Yes. By blending in some G2926B clear glossy (90:10 proportions). The result: The marks are gone and the surface is only slightly less matte. This underscores the need to compromise the degree of matteness, on food surfaces, enough to avoid staining and cutlery marking.

Plastic natches are cast into plaster molds to provide a durable and good-fitting interlock between pieces. The traditional self-interlocking 3/8" or 9.5 mm (nipple diameter) one has not proven suitable for mold making based on 3D printing. Our solution is a four-part system. To use it, your 3D printed mold shells only need matched 13.5mm holes.
-13.5mm holes in 3D printed case molds are all that is needed to adapt to these.
-3D printing case and block molds necessitates pouring plaster and rubber into shells with planar mating surfaces downward (they must sit flat on the table). The thin flanges on the clips cause minimal issues.
-Casting an embed into a mold enables gluing (or friction fitting) a natch or a spacer inside.
-The use of embeds permits flat mating surfaces - these can be sanded (for better flatness and fit). They also allow replacing natches if they get broken (assuming friction fit).
-A set of four interlocks (4 embeds, 4 clips, 2 spacers, 2 natches) weighs 8.7g.
Our drawing shows the measurements we use. 3D printing is precise enough that the inside dimension of the embed is the same as the outside of the natch shoulder, yet the natch fits. The same good fit happens with the clip and embed and the natch nipple and spacer (although it is necessary to chamfer the bottom corners and bevel the top corners of the spacer for better insert).
Some dimension changes may be needed to fine-tune for printing in your circumstances.

-Size of mixing container needed: 1L per 1000g of plaster.
-To determine the amount of plaster and water needed our normal process is to counterbalance the mold and fill it with water to get the cc volume, in this case 2000. Then use the glazy.org plaster calculator for the amounts of plaster and water needed. But this time I just used the gram/weight as the amount of plaster, 2000g. I derived the water, 1400g, according to the 70/100 ratio recommended by USG (it was 200g less plaster and was sufficient).
-Using a 4-minute soak and a good propeller mixer there is no need to sprinkle the plaster in.
-Make sure no plaster is hanging on sides of the mixing container after soak.
-Mix with a whirlpool shallow enough that it does not suck air but deep enough that it is pulling bubbles up to the surface.
-Make sure the mold is strong enough not to split at the bottom and is a strong shape that will not bow out. 3D print it so the artifacts are at the top. For PLA mold soap is not normally needed.
-Make sure the mold is sitting on a very flat surface.
-We had to apply tape to the bottom flat surface to make sure it is watertight.
-Pour the plaster into the mold in a steady stream that does not pull in bubbles.
-Remove using a heat gun when the plaster is set.

Stains can work surprisingly well in matte base glazes like the DIY G2934 recipe. The glass is less transparent and so varying thicknesses do not produce as much variation in tint as glossy bases do. Notice how low many of the stain percentages are here, yet most of the colors are bright. We tested 6600, 6350, 6300, 6021 and 6404 overnight in lemon juice, they all passed leach-free. The 6385 is an error, it should be purple (that being said, do not use it, it is ugly in this base). And chrome-tin pink and maroon stains do not develop the color (e.g. 6006). But our G1214Z1 CaO-matte comes to the rescue, it both works better with some stains and has a more crystal matte surface. The degree-of-matteness of both can be tuned by cooling speed and blending in some G2926B glossy base. You can mix any of these into brushing or dipping glazes.