Time to Die

Designing and 3D printing (in metal!) some gaming dice! Mainly as a way to familiarise myself with the processes and pitfalls of metal 3D printing.

Since I need to get my head around 3D printing in various materials, and one of the classes I tech-support at work is a board-game-design class, I felt this would be useful on multiple fronts. Since there is no real point in custom 3D printing something that I can buy much cheaper mass-produced, I went for something genuinely unique too: My die are 8-sided and numbered zero (blank face) to seven, and in pairs, so you can roll 00-77 in octal (0-63 decimal equivalent). I rather like base-8 numerics, and my modified-octahedron came out both aestetically and dimensionally pleasing to me, and also 7 dots can be arranged in a way that is likewise aesthetically pleasing plus seems to bypass the normal human subitizing (sight-counting) limit.

I first generated an octahedron in Blender, chopped of the 'top' and 'bottom' points for aesthetic reasons, and generally smoothed out the edges with a bevel filter. Next I generated a bunch of cylinders to subtract out of that mesh, generating the below 3D model (coloured for clarity).

8-sided die.

Tetrahedrons are actually symetrical on all axis, but I found clipping back two of the points had a nice aesthetic, let the whole thing fit snugly inside an invisible cube (bounding box) with the truncated height the same as the width on the non-truncated side edges. I also found a further use for the flattened points: as I intended to use two die together to form a two-digit pair, I needed to mark one of the die to indicate it was the first digit of the pair. I cut out a circle on the top and bottom of one of the die, in the truncated bit, to be this indicator:

High and Low die.

So now you roll both die and take the marked one as the upper digit and the unmarked one as the lower digit for an octal result of 00 to 77.

I am also quite happy with how the hexagonal pattern of dots came out. I also tried some other ways of arranging the dots, such as a more traditional 2x4 grid, but found it was very difficult to tell at-a-glance what the number is. The hexagonal arrangement seems to bypass the numeric part of the brain and instead tap into the linguistic center, treating the various patterns as glyphs to be read rather than numbers to be counted (completely non-expert hypothesis). Also the hexagons fit quite nicely into the equilateral-triangle-based faces of the octahedron. I shy away from using numeric digit glyphs (and likewise specific languages and alphabets) as I like to keep things culturally neutral and universally applicable when I can.

And here is the STL file for these two die. I printed mine at 21.3mm high: 1/64th bigger than I wanted so I had a bit of room to sand-and-polish them back from the raw print output. You can print them at what ever size you like and your equimpent can handle. Tip: Use 100% infill as the mass (and center-of-mass) of die is important for them to roll well/fairly.

8-sided die with Hindu Arabic glyphs.

Above: What they might look like using familiar Hindu-Arabic numeric glyphs. Note that this level of detail is too fine to 3D-print at 21mm in many print processes, though a resin printer should handle it fine.


I did consider doing a test 3D-print in PLA plastic, but at the time only had easy access to a (rather clunky old) filament printer with a resolution too chunky to make this relatively-detailed item. Instead I went strait to a 3D printing house that can do metals. I chose stainless steel, which was a little more expensive than aluminium, mainly because I found a chemical 'blackening' process I was interested in trying out for that metal.

I had my first run of metal printing done by Zeal 3D Manufacturing Services in Melbourne, Australia (mainly because they were the first to get back to me with a quote). They were very helpful (and patient!) with assisting me get my model into a form that had a decent change of printing well - I am pushing hard on the resolution limits for that particular material! Print time from model-confirmation-and-payment to arrival-in-the-post was about 3-4 weeks as a standard-priority job.

TAEM Students note: If you are interested in going this route, I recomend bringing your 3D print files to me first so I can pre-check them before you submit them: I can likely save you, and your 3D print supplier, a bit of back-and-forth resolving issues - Metal printing is generally not super-high detail, plastics and other materials may vary in other ways: I am gradually trialling different materials and print processes to develop my knowledge in this area so I can advise students in this regard.

Also note that this isn't at all cheap: Even something as small as these two dice, you are looking at around $20 in high-res plastic or up to $100 in metal (so I won't be setting up an Etsy store any time soon!). That cost was one I was willing to bare in this case as it produced some really nice (and tax-deductable for me) results and was also a professional learning task in itself. YMMV, and a quick-and-nasty low-res print on our own in-house consumer-grade printers may be more appropriate, at least for prototyping! And keep the potential costs in mind before approaching a 3D print manufacturer, since I'm sure they already get plenty of glassy-eyed hopefuls that balk as soon as they get the production quote!

Undergrads shouldn't expect extra marks over a functional basic prototype in assignments, but may still want to go all-in if their project has personal value to them beyond passing the assessment... or they just want the extra learning experience of doing it! At HDR/postgrad level this may be a different matter, depending on the assessment goals agreed with tutors/supervisors.

After some back-and-forth with the printer to get my model within spec. for their printing process (my origional design had the holes a bit too close together and not deep enough to print well for this material), I got 4 dice pairs printed. At around $90 per pair, that wasn't cheap, but not a bank-breaker either.

They came back from the 3D printer already quite usable without further embelishment:

Rough-printed die as they arrived in the post.

Next, lets try some chemical processing, and plain old painting:

First pair, I lightly sanded their surfaces back (to make sure the surface was fully exposed to the chemicals I was about to apply) then dunked them in Caswell Stainless Steel Blackener for a few minutes, then washed the chemical off as per instructions. Then, one face per hour (so the paint wouldn't run as I turned the dice over), three times over, I hand-painted in the dots in a nice bright red modeling-paint:

Chemically-blackened die, hand-painted dots.

Conversely, I also tried using a fine brush to apply the stainless steel blackener inside the dots with the outside left alone, washing it back out after about 3 minutes chemical-reaction-time as per instructions. I used a rotary engraver to remove the oxide layer from the metal inside the holes for the blackener to take properly, and then I had to sand and polish the surface for better contrast since the 'black' is not as good as black paint would be. In the end, more work than painting for a not-as-good effect, but you have to try these things to find out, and while not as good, it isn't actually bad:

Sanded die with blackened-dots.

Since I already have a blackened-die with painted dots, I will keep the chemically-blackedned-dot die for my student-demo pieces and won't be doing a bright-face+black-paint-dots version for this set. Though maybe I should paint one and leave the other like this for comparison?

Conclusion: The stainless steel blackener doesn't 'stick' to the surface as well as I had hoped, so tends to rub off, particularly on the edges which get the most wear during use. I am going to try sand-blasting a dice and re-blackening as the rougher surface from the sand-blast may hold the colour better. Stay tuned! A student with some experience in the field has also suggested a different chemical process (which gives more shiny-blue than black) which I am interested in trying too.


Well! Sand-blasting had an odd effect: it took off most of the black very easily (I hit the die very breifly, not taking the paint out of the dots at all) but when I dunked the die back in the stainless-steel blackener, it didn't react at all! No idea why, but the surface of the die has obviously changed in some way. I could have sanded back the surface and re-blackened, but that would have just put me back where I started. In the end, I decided the (no-longer glossy, due to the sand-blasting) red dots looked quite good on the silver face, even the small amount of remaining black was nice, so I am keeping them like this, now:

Sand-blasted die with hand-painted dots.



Another day another die!

Using Amiga Engineering this time - they have a higher resolution - and proportionately more expensive - print process. I was planning to go with stainless steel again, however my job was too small to start up the machine for on its own and they didn't have any larger jobs pending in stainless at the time, so they offered to do them in titanium for a SS-price instead. My preference for stainless steel is that chemical-backening process I used above, but Amiga sent me some information on some interesting-looking heat-treating colour effects you can get with titanium, which convinced me to try it out.

While I was very happy with the print-output of my first run of dice, I made some minor changes to my model - a slight adjustment to dot size and placement as well as increasing the bevel on the sharper edges a litle bit. Amiga also suggested I round out the sharp inside-bottom edges of the dot holes to make printing a bit easier, and I suspect this will also improve paint-flow on dice I hand-paint in the dots too (in the painted sample of my die-run above, you can see how the paint tended to pool more thickly around the bottom-edges of the dots due to the sharp angle's interaction with the wet paint's surface tension. With that sharp angle gone, I might even get away with only 2-coats instead of the 4 coats I had to do to (mostly) cover the inside-corner paint-pooling effect above!

Charm-Bead Model.

For this version, I also 'cored' the model with a 2.5mm tunnel through from top to bottom: I already had a bunch of nice playing dice, so wanted to do something different for this run. So I am printing them at 50% size (so 10.5mm bounding box) and with the mentioned hole vertically through the middle. I also didn't make two-digit variations this time, just one dice model. While these can still be used perfectly-well as dice like this, these are intended to be dice-themed jewelery beads for a charm bracelet, necklace or similar.

Charm-bracelet Die.

Next up, (when I find time - I presently have a whole new maker space to bring up at work!) I will try some of the heat-tempering colouration suggested by the print supplier. Will update here when I get some pics.



Mis-adventures with filament printing.

I managed to get the filament printer at work going again (I needed to make a replacement part by combining replacement parts from the MkII and MkIII versions - our MkI is 4 years old, so of course you can't get any parts, or a word out of the manufacturer - welcome to 21st Century Human Technology!)

Here are some prints:

Bad 3D prints.

Top is a fairly normal result when the print detaches from the print bed and gets dragged around melted onto the print nozzle. That is mainly down to my inexperience in setting up a sufficient bed-adhesion layer. Bottom is the normal result when web designers who think being able to copy-paste some Javascript qualifies them to call themselves 'programmers' - more specificaly, the spacing of the 'rifts' at 2.5mm, 5mm, 10mm (and I am sure 20mm, 40mm... if we went on) indicates a stupidly sub-novice screw-up in the handling of float-int conversion... we can also extropolate from this that the driver is working in centimeters as its base unit.... Anyway....

Below is the successful fresh-off-the-printer filament print. Done in black ABS plastic just because that was what happened to be spooled on the printer when I got it working again. It will actually look a bit better (though I doubt good) once I get the printing-supports off and give it a light sanding and polishing). Maybe it is just that ABS plastic is not the best material for this job... I will have to try a PLA print some time.

Good 3D print.

With ample application of supports and rafts, (and trying 3 different 3D print slicers, all of which appear to be inscestuiously copy-pasted from each other to differing degrees of failure-to-function, sharing a general UI that also makes me question the developers' wannabe-programmer-web-designers' design credentials!)... it is actually possible to get a passable 3D print out of a consumer-grade filament printer. And to be fair (though only on the hardware, not the reprehensible software!)... above, I am deliberately pushing the limits of what is really a rich-kids'-toy technology, mainly to show how far it can, and cannot, be pushed. There certainly are valid uses for this print technology, though:

Actually useful 3D print.

Above: A more appropriate job for this type of 3D printing technology: larger, lower-detail items are good. The insert trays were what I printed: the outside of the case was some sort of nifty-looking component box I bought several of online from a warehouse clearance sale some years ago and have been finding uses for ever since. I soaked it in near-boiling water to soften the glue, then gently pried out the component tray inserts that came with it, using them as a measurement guide for sizing the die-holder inserts I printed to replace them. Black ABS plastic again, as that was already loaded in the printer, and I thought it would look best anyway.

Actually useful 3D print.

Above:The label on the front was done in an office label-writer and carefully trimmed to fit an existing cavity on the top of the case.

....and for crude prototypes. But if I had a suitably large lithograpic resin printer right now, I would probably use that for all those too!

Consumer 3D-printer software, like a lot of software in the 'maker' sector, is generally quite bad - even for 21st century cloud-ware junk. Messing around for the better part of the day to get a slicer set up to output something intelegable to the printer, and having to shuffle stuff about on SD cards because connecting to a printer with a USB cable isn't good enough for modern in-the-cloud consumer 3D-print software, is totally unacceptable, and is actually just a great way to make me really wish the windows in my office could be opened so I could toss the poorly-supported junk out of one!

And who's idiot idea is it to totally ignore the standard user-interface for the host OS and do something stupid and unintuitive - it doesn't make your software better, it just adds an extra level of learning between the user and their results. It definitely doesn't make the 'programmer(s)' appear in any way clever!: just shows them to be an unqualified and unprofessional hack without any actual training in the field.

And, no, I really don't want to subscribe to your pointless adver-zine and honestly resent having to sign up to your tediously under-resourced 'cloud' 'service' just for the privelege of downloading drivers/extensions. I will save the advertisers you sell my personal information to some CPU cycles on their profiling/analysis tools: I am in the RAGING MAD CONSUMER demographic!

New fillament printer: Ultimaker 2+ Connect, which is a fairly expensive one, and it shows: The software is noticably better and the printer itself is quite reliable (for this class of printer). I am generally very happy with it so far. My die are still pushing the limits of this type of 3D-print technology, so are still not brilliant, but are distinctly better, all the same. I also printed in PLA this time rather than ABS, though I suspect that wouldn't really make as much difference as the generally better printer itself.

A better 3D filament print.

Painted, they look... not great..., particularly the bottoms which suffered some distortion. I also had to clean up the bottom a bit with a hand-held rotary engraver to get it to work at all.

Tidied up and paninted filament-printed die.


The new printer takes a thicker gauge of filament to the old one (almost certainly a good thing considering how much filament-breakage that one suffers!) so I am still using the old printer for chunky, less-important jobs until I have used up all of its filament stocks. Though that printer's slicer software also got a major version upgrade recently, so while it does still have old-model-related issues I have to work around, it is working a good bit better now. Though my comments above still apply.



Resin to die!

Having recently added an AnyCubic MonoSE to our maker space, I had the opportunity to do a resin print of my die. The results at this much higher-resolution were beautiful!

I am mentioning the make and model specifically here as I have nothing but good things to say about his device - mechanical build quality was really good and the results were great. The only issues I have had were entirely due to my own inexperience with the process (learning the optimal settings for auto-generating rafts and supports) and a specific problem with my model (odd inverted normals - the bane of my 3D-printing life!). The only significant down-side for this class of printer, as far as I have found sofar, is there is a good bit of clean-up (of the resin tray specifically) needed after each print run.

I would recomend using Chitubox as your slicer rather than the included Photon Workshop slicer. The latter is (in a first for my experience-so-far with the maker sector) perfectly funtional and usable, but a bit basic. Chitubox is a much more featureful slicer (including detecting and warning me about those annoying invisible inverse normals in my model). The UI still more resembles a bad phone-app than an actual computer program, but 21st-century-tech-users can't be choosers, apparently! And I now have some 3D-scanner software for comparison that genuinely makes me yearn for 1991 GUIs all over again! Macintosh System 7 really was the pinacle of human-computer interface design, which is honestly just sad!

Anyway.... I painted the dots. Green isn't a colour I use much, but I felt it looked rather nice on the frosted-transparent of the die bodies.



Call me Al

I now have some aluminium prints, from Zeal 3D again. As with the staisless steel ones, even fresh from the factory, they are very nice:

Aluminium die as delivered from the printer.

I had four pairs printed again. Less expensive than stainless steel, but at around $60 a pair, still not cheap, but worth it, I consider: the weight of this material feels very suitable for dice.

Aluminium die on the skotch-brite pad I polished them with.

And here is a pair polished with a scotch-brite pad to remove the factory-oxide.

Polished aluminium die with dots hand-painted.

And now the dots are painted on one polished pair. I did one red and one black simply because for demo pieces it gives me more variety to demo.


I am currently in the process of building a small bench-top anodizing setup to try dying the aluminium surface in a reasonably permonant way before (as usual) hand-painting in the dots. This will hopefully be the fate of the remaining two aluminum pairs. Stay tuned!



Blankedy Blanks

3D printing, particularly at professional quality, is fiddly and expensive. As a much cheaper and simpler approach, particularly good at the game-design prototyping stage, doing something by hand may be more effective.

Here I used a rotary hand engraver to crudely make my die set out of octahedral blanks I bought for a few dollars online (other common die shapes such as D4, D6, D10, D12, D20 also available). You could even just write-on the faces with a permonant marker, but by engraving in the holes, and then colouring in those holes with a marker pen, I get something a bit more wear-resistant. This one was done at home during COVID lockdown, so is a bit rough: with a small drill press and an X-Y-adjustable bed-clamp, I could do a much neater job with the holes properly aligned, but even this crude job is a good example for if you just want to quickly make up some custom die for play-testing an idea.

Hand-engraved die.

More to come: We recently got a CO2 laser cutter/engraver in our maker space, and once I get the CnC machine assembled and working, I intend also making a set of trays that will let me burn nice elaborate die-face graphics onto such plastic blanks off a template that students can use to design their own faces into. (One face per run, manually turning the die between, so no D20's please!)



So... what to do with a pair of 8-sided die that count 0-7 each or 00-77 together? Well, a tabletop-role-playing-game is an obvious use! You could simply modify one of the open-source "D100" systems, which use two 10-sided die to roll 00-99 in decimal, to instead use 00-77 in octal (base-8), for example.

Of course I made one .... But it is a very rough early draft with many bits still missing and completely un-play-tested, so I haven't released it just yet! - I didn't follow my own advice and modify an existing D100 system and instead worked it up from scratch, so it is really badly play-balanced right now! Though I have some interesting game-mechanics that I am quite proud of (assuming they survive an actual play-test!) Stay tuned, I will slowly pick at it and improve it as my interest and free time allows.

Note that while the game uses octal die, this is largely hidden from players: as long as they know their (decimal) 8-times tables up to 7x8, players may not even notice as the (default) rules are still decimal. There is also an appendix for playing in true octal, as well as another for adapting the rules to play with a pair of D10 die. Even playing in a numeric base of 6 or 12 is likely possible, in either the native base 6/12 or using decimal, with a pair of 0-5 or 0-11 die: I haven't covered this in my rules-conversion appendixes, but one can use either the true-octal-play or base-10-play conversion appendix as a guide to what needs to be altered if you really wanted to.



Notes on Dice Design.

Formally, The singular is dice and the plural is die. This gets really confusing and I have undoubtedly messed it up multiple times in this document too! I don't think it really matters - I take the view that any use of language that remains clear in meaning is correct usage. When I talk about resin-casting dice, I deliberately use dice as both singuular and plural, to differentiate it from the casting-die, which is the mould into which the resin is poured!

There is a convention to the way numbers are put on faces of a dice. That convention is that all opposing faces should add up to the same number. So for a standard 6-sided dice, the opposing faces are 1:6 2:5 3:4, all adding to 7. On my 8-sided die, the opposing faces are 0:7 1:6 2:5 3:4, also all adding up to 7 (because I am counting from zero... both happening to add to 7 is coincidence: nothing special about the number 7 here!). On a 20-sided dice, 1:20 2:19 3:18 etc. all adding to 21. For dice without opposing faces, such as tetrahedrons or some of the wacky odd-number-faced die from my sample bag, this system is out the window and you probably should just do your best to disperse the numbers around as far from their numeric-sequence neighbors as possible.

For non-numeric dice, there are no real rules, but you probably should again think about what may constitute an 'opposite' in the context of your game and go with that. One example (or counter-example, since it is sort-of the opposite of what I just said!) is 'Fateā„¢' or 'fudge' dice which are 6-sided dice configured effectively as 3-sided dice, as there are two faces with a + symbol, two with a - symbol and two with no symbol. These dice just have the same symbol on opposing faces, so +:+ -:- blank:blank.

Fudge Die

Roll big handfulls of these (fun in itself) and add one for each + and subtract one for each to get your result - it has a distinctly different probability-distribution of the final number compared to more tratitional dice techniques.

Here is a bunch of other non-numeric die from my demo/sample collection:

Random collection of non-numeric die

Top Row: constellations, planets, colours, shapes, rock-paper-scissors, I-Ching, seafood (actually, from a form of Mahjong, I believe).
Bottom Row: emotes, weather, direction, terrain, dungeon-interor, furniature, treasure. (Some of these bottom row might be used for on-the-fly random map generation in a tabletop roleplplaying game - you can use standard numeric die and a look-up table, of course, but special die like this can be fun in themselves. A numeric dice and a look-up table can do the same thing, but specialist die can be fun in their own right!)

Shape is also important: Below are two different designs of 16-sided dice. The left one has sides of slightly different shape and angle between faces, which means it is not going to be truly 'fair' (though it will be close enough for most purposes). The right dice uses an octagonal trapezohedron design (as per the pentagonal trapezohedron used for TTRPG 10-sided dice) and is far better for ensuring reliable randomness, at the cost of sharp angles that make the physical roll a little awkward. Two types of 16-sided dice.

100-sided die have the same issue. I did get one amongst the two multi-die sets I purchased to get my big variety of numeric-die samples, but I removed it from my collection because it was big, stupid and I don't care much for numeric base-10 anyway! I wouldn't make myself a 64-sided dice for the (first two) same reasons, and my dual-8 die set performs exactly the same function anyway.

The platonic solids are obvious good choices for dice shape as they are very regular, providing 4,6,8,12 and 20 sides identical in size, shape and angles between. There are a number of other semi-regular shapes with identicaly-shaped sides though the faces themselves are not regular, and the angles between them differ, though in consistent ways, so the die can probably still be considered 'fair'. These are usually constructed by further triange-dividing the faces of the platonic solids, providing face counts such as 12, 24, 24, 30 and 60 (repeats there are deliberate - there are multiple ways to get some of those side counts).

A bunch of various-sided die.

The 4-sided dice is an interesting one to consider separately too, as the tetrahedron (below, left) is often used but rolls really poorly due to its sharp angles and low center-of-mass:

Three styles of 4-sided dice.

Above center is a tetrahedron with the corners rounded off for better roll, which also gives you an up-facing surface to put the numbers on. And above right is an entirely different approach to a 4-sided dice. ... you could also make a D4 by doubling up numbers on a D8 (or trippling on a D12, quintupling on a D20!)

Two slightly different D8s.

Similarly, above are two D8s. Left is the more conventional sharp octahedron, and right is one with rounded corners (I got this one after I had shaved the tops off my own D8 design. After getting this die, I played around with taking off my edge corners, but because my dot pattern only has two-axis symetry on some faces, and shaving all the corners lends itself to three-axis symetry, I decided to stick with dice with a definite 'top/bottom'. A purely personal-aesthetic decision).

Collection of odd die.

Above are some assorted 'weird' die, (yes, I consider D10s 'weird'). The D10's, being pentagonal trapezohedron in shape, are 'fair' (if a little pointy) as is the D3 made from a D6 with numbers I II and III doubled-up on opposite sides. The D7, D5 and D16 are not strictly 'fair' for reasons discussed above, but are close enough for just having fun with! A truly 'fair' D5 couid be made by quadrupling-up numbers on a D20. You could also make a 'rounder' D10 by doubling up numbers on a D20. The only way to get a 'fair' D7 would be to double up numbers on a 14-sided heptagonal trapezohedron, though.


upBack to Glenn's Project page.