A project log for Pixels

A set of smart dice, the same size as traditional ones, full of LEDs and Bluetooth-enabled.

Jean SimonetJean Simonet 03/18/2020 at 16:585 Comments

Hi there friends,

The electronic dice know have a new name: Pixels. I always knew "the electronic dice" wasn't a great product name, but also didn't want to rush into finding a new name for them. I'm happy with Pixels, it hits a bunch of notes I really like: It sounds like Pixies and evokes video games. Pixels have a new website and twitter account; feel free to check them out ;)

I posted a few videos already, and the dice have really picked up a lot of attention, it's been amazing! Of course it also means that the pressure is on to get this darn things manufactured!
So were are things? In short: more revisions, more updates and more research.

New boards

I think I've lost track of the number of board revisions, but new D20 boards just arrived as I write this. I ordered them in all the colors that PCBWay offered, as I want to see how the color of the board affects the final look of transparent dice.

---------- more ----------

The BOM is pretty much the same as previously, the main changes are minor functionality-wise, but important for manufacturing and end use. Let me go through them!

Proper outline

The previous version of the fold-able board didn't perfectly line up once folded. It was pretty close, but not exactly right. So this time I took the time to really model it in Fusion 360. The "Sheet Metal" tool is really great for this. I created a custom material with a 0.2mm thickness (the PCB thickness) and 0.0 K factor (K factor is a representation of how much a given material stretches when you try to bend it) and designed the folding pattern one face at a time, making sure that everything fit together every step of the way.

Once I had the outline properly figured out, I exported it to Kicad to use as a board outline and started re-routing the board. Once again, I added small areas of metal and solder paste to stiffen the board on each side of a bend. This has worked really well in the past to make the board naturally bend into shape without needing to be preformed.

Of course the new outline features the interlocking notches I mentioned in the previous update. Fingers crossed that they function exactly as intended and reduce the folding/assembly time drastically.

Lastly, the new outline features specific cut-outs on certain faces. This is again targeted at manufacturing. Those cutouts will match registration pins inside the molds, so that a folded circuit board can only be inserted in the molds in a specific way. This is important because otherwise the numbers on the faces may not match what the microcontroller thinks they are. The presence of the coil was already providing some registration, but it still allowed a certain rotational symmetry, meaning that it was possible for the 1 and 20 to be correct but not the other faces; a huge pain in the rear for 'calibration'.

More voltage monitoring

I updated the schematic a tiny bit to use a couple more A2D pins to monitor the charging and LED voltages. This won't be super useful during use but will make testing boards during manufacturing that much better. The board will be able to quickly check if there is something wrong with the LEDs and/or charging circuitry, hopefully leading to a faster test on the assembly line.

Dual monopolar Hall-effect sensor

You read that right, it means the sensor has two elements, each only sensitive to a single pole north for one, south for the other. This is in contrast to an omnipolar sensor, which would have a single element, sensitive to either pole. Previously I was only using the north pole output, and had it connected to the voltage regulator enable pin, so that I could easily power-cycle the microcontroller (a crititcal feature during development). I was also planning on using that magnet to power down the board during transport - each charging case would have a small magnet in it - so that the dice doesn't drain its battery for no reason while being shaken, and that is working great.

However, I have started designing the 'To Go' case. This is a case that can charge 8 dice at a time, and has a battery of its own so that it can recharge the dice without itself needing to be plugged in. That's great, but being on battery itself means that this To Go case needs to be smart and only charge the dice when necessary, stopping when the dice are full. Otherwise it would keep driving the charging coils forever and drain its own battery. This meant that some form of communication was necessary between the case and the dice. The Qi standard has a fairly complicated way of doing this by modulating the load capacitance, and I didn't want to go down that route. In fact in the previous log I explain a bit how I started using Qi but then moved away from it.

Anyway, I eventually decided to just have the case talk to the dice using the existing Bluetooth :) This meant that I couldn't have the dice simply turn off while they were charging in the case. So I hooked up the south-pole output of the hall-effect sensor to an interrupt pin on the micro. This way, the charging case can have magnets embedded in it such that the south pole triggers the hall-effect sensor and essentially "tells" the dice that it is inside the case and that it can go into an extremely-low power mode, waking up as necessary to communicate, while still being able to use the north pole of a magnet to reset the micro-controller!

And finally, the last change was to move the hall-effect sensor to the 20-face of of the dice. Since the charging coil is on the 1 face, it means that when a die is placed in its charger (with face 1 down) the sensor is in the right spot to be triggered by a magnet inside the top cover of the charger, regardless of how rotated the die is inside the charger. This is one of those completely-obivous things in hindsight, but that you don't really appreciate until you've started actually handling and using your product.

Material testing

I think I've mentioned previously how I wanted to try different combinations of material for the dice, and this is also going on as I write. I had many many questions:

I needed to run a lot of experiments. The thing is: I didn't really want to sacrifice a pcb for every single try, I don't have that many...

So instead I made a little "Core"; essentially I poured resin around a folded pcb so that it would be protected, but kept it as thin as I could. Then I created molds for hollow half dice, like little caps I could snap on top of the core to see how the light would shine through! These half-dice caps are still cast in two steps (see previous updates), giving me a chance to try clear/opaque resin combinations, as well as different designs, without wasting pcbs.

I have only made a few castings so far, but I have already learned a number of very interesting things:

The first is that liquid additives (for instance the Pinata alcohol-based ones) tend to mix more uniformly than powder-based. Take this with an asterisk though, because it may entirely be my fault, not so much the additive's.

The second is that making a really thin layer (1mm) of transparent resin opaque takes a lot more additive than you think. In fact I got much better results starting with a different kind of resin altogether. Using additives caused the light to 'bleed' through thinner parts of the casting in unpleasant ways.

Then glitter / mica powders don't really interact much with the LED light. I bet there are really interesting reasons why (involving the single wavelength spectrum of LEDs and the microstructure of glitter kind of thing...) but the bottom line is that those really are only useful for improving the look of the dice when NOT lit up. Also, another one of those obvious-in-hindsight things: if you mix glitter in your opaque resin to try and have a glittery surface finish, you need to mix A LOT of glitter in, because most of it will be wasted under the surface. What I have seen dice makers do instead is brush the glitter on the inside of the mold before casting in resin. I haven't tried yet...

But the most important thing I learned what the difference between Urethane-based resin and Epoxy-based resin. It's not always clear which one you're using, but it matters. The end product looks and feel the same, in clear or opaque. Urethane resins are often thinner than Epoxy resins, but that it not by nature: it's just that a lot of Epoxy resins are "Art" resins or counter-top resin and are therefore formulated to be thicker and have longer pot-life (to give you more time to work with it). But you can actually make thick Urethane resin or thin Epoxy resin. Smooth-on, for instance, sells all variations of both. No the main difference, besides the smell, is that Urethane resin gets softer with heat. Epoxy resins doesn't.

I realized this when I tried to recharge my little led core with a Urethane resin cap on: the cap got all soft on me! The core had warmed up to maybe 50 degrees C from the charging and the cap was completely squishable. This is not something I really want for the final dice. Note though that it hardened right back after cooling down, and it didn't seem deformed at all, but still...

So far I have had some success with material combinations. For opaque dice, I've found that an inner layer of just slightly tinted black transparent epoxy and an outer layer of neutral opaque epoxy looks really good! Surprisingly, a totally black outer layer doesn't look as good, and I can't figure out why... It's fascinating and I can't wait to try more stuff!

I have a few more types of resin on order for even more experiments. I am really excited to try the aluminium-filled one; first to see if it is significantly denser than the regular stuff, and then to see if it's possible to get a polished finish on it... would love to be able to create burning-metal dice :)


I have also spent some time designing chargers for the dice, and that in itself is another interesting story!

I need to distribute a charger for each die that I sell, obviously. This means that the cost of manufacturing the charger is, for all intents and purposes, part of the cost of a die. That cost must be as low as possible. The coils are so small that in order to couple properly, they need to be really well aligned between the charger and the die. This means, in essence, that each die requires a custom charger, or more specifically, a custom cup that it will sit in to get recharged.

On the other hand, most people will buy more than one die (hopefully many more) and having to juggle 6 or 8 individual chargers does not sound like fun. So I want to offer a multi-die charger, the To-Go charger I mentioned earlier. But of course, players may not necessarily all want to purchase the exact same set. Maybe one player wants to get eight D10, or six D6, etc... In short I can't make a one-size fits-all multi-dice charger. And of course, if you buy the To-Go charger, then you've overpaid for the dice and now have a bunch of individual chargers you don't need anymore. A huge waste!

It would also be a SKU nightmare! "Do you want to get the 'midnight black set with 2 D6s' or the 'clear set with 8 D10'?", "Don't forget to add a single-die charger if you didn't get the To-Go case!" etc...

So I came up with something that I *think* will work; it's a bit of a gamble though.

First off, each die you buy will come with a charging case. There is no way around it. You can't buy a die without a charger. The case, off course is important anyway because it allow you to "turn the die off" while you're transporting it. The charging circuitry is still based on the XKT-412 and XKT-335, super dumb but super cheap. When you need to charge the die you can plug it in to a usb plug and you're good to go. I'll only include a micro USB cable, no wall-wart. A lot of cheap electronic devices do this and nobody seems to mind.

However, this simple charger will be built in a way that it can easily be taken appart, and that's the gamble! There are only 4 parts to it:

Only the insert is specific to the die that the charger is sold with, the rest of the components are common to all chargers. This is already a benefit in terms of manufacturing costs. The insert will have to be vacuformed for the simple reason that it needs to be really thin (0.5mm) and that seems too thin for injection molding.

To-Go chargers will come without inserts and receptacles to plug in the coil PCBs directly. So in order to make your To-Go case functional, you'll have to plug in the Coil PCBs and add the inserts yourself. This will give you the greatest flexibility in how you want to organize your dice, while reducing both the cost and combinatorial complexity of offering a million different versions of the chargers.

I could make it that you only need to add the inserts yourself (i.e. the To-Go charger already has Coils for each die) but the Coils ARE the most expensive component of the charger, so... it may be worth it. I don't know, I may still change my mind. :)

In any case that the jest of it. I am hopeful that rather than put people off ("Oh no I have to assemble this thing myself...") it will make it feel more personal, customized. We'll have to wait and see...


rcombs wrote 04/13/2020 at 02:20 point

The interchangeable-inserts case concept seems really neat!

My only issue is how you mentioned using micro-USB. Is this essential from a BOM perspective? I'd personally much prefer USB-C in essentially any new device I buy nowadays.
Additionally, would it be cost-prohibitive to have the "on-the-go" case include a Qi _receiver_? I find that capability extremely convenient on my AirPods charging case (which this seems pretty similar to in concept), and would happily pay a bit extra for the feature.

  Are you sure? yes | no

Finn Lynch wrote 03/19/2020 at 22:32 point

I'm incredibly excited about this project. If it launches you can count on me day one at least one set purchase. I just hope I can grab an early bird ;)

  Are you sure? yes | no

hornswroggle wrote 03/18/2020 at 18:55 point

Every time I look at these dice I get giddy like a child on Christmas Day.

I cannot wait to show up to my playgroup with these.

  Are you sure? yes | no

Richard Hogben wrote 03/18/2020 at 17:25 point

Nice update! Those chargers look really cool.

  Are you sure? yes | no

Jean Simonet wrote 03/18/2020 at 18:22 point

Thanks! I have to say, being able to get a nicely ray-traced image out of Fusion 360 really helps make even simple shapes look fancy ;)

  Are you sure? yes | no