moodlite - Coming soon

V3 idea with better electronics and multiple shape support

tg44 · 386


  • Newbie
  • *
    • Posts: 3
    • Likes: +0/-0
    • View Profile
Hy there!

I'm thinking about making my own nanoleaf. I read a lot of posts, guides, teardown images etc. I think building one has several challenges. The biggest one, in my opinion, is diffusion.
This group seems to have a lot of good ideas, but the actual electronics and software are really "cheap". I mean, you can't even create a decent tree :D You can't create randomly shaped squares, because they need to be in a snake pattern. Because of the electronics, it's hardly plug'n'play.
BTW the original nanoleaf has some really bad design too. I mean you need to address all tiles one-by-one, and if you have a big enough shape, you can't refresh it more than 2 times/sec. (And you can't mix the squares with the triangles... And how hard would it be to create a 5, 6, 8 edged shape?)

So my proposal is to create a small and cheap circuit, which can handle the communication, and can help to detect the shape of the graph.

If we want an 8-edged tile too, we need 8 data pins for the edges, 1 data pin for the inner led strip, 1 software reset just in case, counted with the needed other pins, we need a 12-16 legged MCU. The MCU can be a lowcost one, we basically just need to transfer bits, no expensive computation (would be nice to stream bits paralell to the strip and to other nodes, but its not a musthave).

The lightning protocol could be something like the ledstrip one. (The ledstrip protocol works like: first node gets the first pack, after that it just repeats every bit on the wire, so the second node gets the first (globally the second) pack, and repeats every remaining bits. With that you makes a big bitstream, and every led gets the information meant to it.) We need only minor modifications, somehow we need to count how much nodes have right and left to the actual node, if we have that, and we count the running through packs we can switch from right to left output after all the right node gets its pockets.
The initial graph detection is an interesting problem, but if we have fixed pocket size, we can do direct addressing (like the original nanoleaf, but on one wire), we can do full-rewrite like the ledstrip, so our protocol can be much faster/better reaction than the original.
Also if we have a concrete board, and we can flash them with a really generic software we can probably do the burnin as a part of the protocol (like tell the pcb that it will be go to a square, and it have 3 leds/edges, and you can tell an other pcb that it is a triangle with 3 leds/edges as a burn in instruction as part of the protocol).

If we have proper graph detection, we can visualize the shape on mobile/web interface! We can create a generic api to switch patterns, switch tiles one-by-one (for example it can be used as temp/humidity metrics, weather forecast, you can assign them to server load metrics, build statuses, motion sensors, who is home statuses, etc.)
Bcs we will have a proper graph detection, and multiple shapes, probably we can do much better looking sound/music visualisation than the original one. (Even bcs we will have an own protocol we have a lot of room to play.)

If we massproduce the electronics, I think one tile pcb can be created about 1$. I think printing a tile is about 1$ too, one led is about 0.2$, so a dualled triangle/square would be ~4$, the 5/6 edges version would be around ~5$. The only question is the diffusion, which is based on the perfection can be another 5-10$/tile.

If anybody is interested creating a PCB or creating a new protocol, or just have questions about this idea just comment :D (OFC it is much bigger than a weekend project (or at least with my current electronics knowledge), but it can be a much better product than the original one.)


SMF spam blocked by CleanTalk