8x8x8 LED Cube
8x8x8 LED Cube
Embedded Files
Going into Junior year of High School, I saw a video of an 8x8x8 LED cube made by Harry Le. I knew right away that I had to make one for myself. Over to the right is a picture of the completed cube and driver.
Going into Junior year of High School, I saw a video of an 8x8x8 LED cube made by Harry Le. I knew right away that I had to make one for myself. Over to the right is a picture of the completed cube and driver.
To start off, I ordered 600 blue LEDs (not all pictured) from Digikey, my go-to for electronic parts. I went through and made sure each LED worked, an extremely tedious yet necessary task. If one does not work in the final cube, it's nearly impossible to replace.
To start off, I ordered 600 blue LEDs (not all pictured) from Digikey, my go-to for electronic parts. I went through and made sure each LED worked, an extremely tedious yet necessary task. If one does not work in the final cube, it's nearly impossible to replace.
Making Layers.MP4The plan was to make 8 layers of 64 LEDs with their cathodes connected in horizontal layers, and anodes connected vertically. This way, the number of outputs needed to control the cube drops dramatically with some clever coding.
The plan was to make 8 layers of 64 LEDs with their cathodes connected in horizontal layers, and anodes connected vertically. This way, the number of outputs needed to control the cube drops dramatically with some clever coding.
I used a piece of wood which I drilled 5mm holes in at every inch as a stencil where I could put the LEDs before soldering them.
I used a piece of wood which I drilled 5mm holes in at every inch as a stencil where I could put the LEDs before soldering them.
Each layer took about 45 minutes and change. I finally finished the 8 layers, which I then stacked on top of each other.
Each layer took about 45 minutes and change. I finally finished the 8 layers, which I then stacked on top of each other.
After countless hours of soldering and trying to get 64 wires to behave at one time, I finished the cube. It was really rewarding to see the all of the LEDs together for the first time.
After countless hours of soldering and trying to get 64 wires to behave at one time, I finished the cube. It was really rewarding to see the all of the LEDs together for the first time.
On the underside of the block of wood where I mounted the LEDs, I connected the 64 anode columns and 8 cathode layers to some homemade 8 pin female connectors, which could be easily plugged and unplugged from a driving circuit.
On the underside of the block of wood where I mounted the LEDs, I connected the 64 anode columns and 8 cathode layers to some homemade 8 pin female connectors, which could be easily plugged and unplugged from a driving circuit.
The next step was to build a circuit to control the cube. I started off with the one that Harry Le, the creator of the original video I saw, had designed. This is the original board I made
The next step was to build a circuit to control the cube. I started off with the one that Harry Le, the creator of the original video I saw, had designed. This is the original board I made
The bottom side showcases all of the connections I made with 30 AWG wire.
The bottom side showcases all of the connections I made with 30 AWG wire.
Although the cube worked with Le's design, I wasn't satisfied. I wanted to be able to say that I made it, rather than followed a tutorial. I set out and designed my own circuit which I then was able to use to make a PCB, taking inspiration from Le which I then was able to use to make a PCB.
Although the cube worked with Le's design, I wasn't satisfied. I wanted to be able to say that I made it, rather than followed a tutorial. I set out and designed my own circuit which I then was able to use to make a PCB, taking inspiration from Le which I then was able to use to make a PCB.
schem v3.pdfboard v3.pdfI ordered boards from a PCB manufacturer and assembled them right away. The left board is a prototype of the final version on the right.
I ordered boards from a PCB manufacturer and assembled them right away. The left board is a prototype of the final version on the right.
Note: some designers use MOSFETs rather than BJTs to connect each layer to ground. This was not needed in this design, as each LED is limited to the amount of current the shift register can supply. Coming from the HC595 datasheet, this current measures at around 6 milliamps. That times 64 gives 384mA, so for a rule of thumb, around 400. The transistors I chose have a continuous collector current rated at 600mA. Given that a single layer will only be on 1/8th of the time and for most animations not all 64 LEDs will be lit, the BJTs are well within their limits.
Note: some designers use MOSFETs rather than BJTs to connect each layer to ground. This was not needed in this design, as each LED is limited to the amount of current the shift register can supply. Coming from the HC595 datasheet, this current measures at around 6 milliamps. That times 64 gives 384mA, so for a rule of thumb, around 400. The transistors I chose have a continuous collector current rated at 600mA. Given that a single layer will only be on 1/8th of the time and for most animations not all 64 LEDs will be lit, the BJTs are well within their limits.
So How Does It Actually Work?
So How Does It Actually Work?
Recall that to control each LED individually, 512 I/O ports would be needed on a microcontroller. The cube connects each anode in 64 vertical columns and all cathodes for each layer. This reduces the number of ports needed to 64+8 = 72. 72 connections is better than 512, but it is still way too many for a cost effective microcontroller. The solution is to use nine 8-bit serial-in-parallel-out shift registers. They do what they say; they take an 8 bit (1 byte) number in serial and, when told to, express that number on 8 output pins all at once. They can be chained together to act as one 72 output shift register, which is what is done for the cube.
Recall that to control each LED individually, 512 I/O ports would be needed on a microcontroller. The cube connects each anode in 64 vertical columns and all cathodes for each layer. This reduces the number of ports needed to 64+8 = 72. 72 connections is better than 512, but it is still way too many for a cost effective microcontroller. The solution is to use nine 8-bit serial-in-parallel-out shift registers. They do what they say; they take an 8 bit (1 byte) number in serial and, when told to, express that number on 8 output pins all at once. They can be chained together to act as one 72 output shift register, which is what is done for the cube.
You may have noticed that connecting the LEDs like I have makes it impossible to individually control all LEDs at the same time. For instance, turning on LEDs at position 1,1,1 and 8,8,8 also turns on 1,8,1 and 8,1,8. To get around this, the cube relies on a phenomenon called persistence of vision. When a light is flashed on and off very quickly, it appears to be on all the time. By lighting one layer of the cube at a time in succession at high speeds, the entire cube appears to be on at once.
You may have noticed that connecting the LEDs like I have makes it impossible to individually control all LEDs at the same time. For instance, turning on LEDs at position 1,1,1 and 8,8,8 also turns on 1,8,1 and 8,1,8. To get around this, the cube relies on a phenomenon called persistence of vision. When a light is flashed on and off very quickly, it appears to be on all the time. By lighting one layer of the cube at a time in succession at high speeds, the entire cube appears to be on at once.
The code theory is pretty straightforward. There is a 2 dimensional array with 64 slots containing values 0-255 which correspond to which LEDs will be turned on. In other words, it has the values in binary to send to each shift register for a given layer. There are functions throughout the program which then change the values in the array, producing animations. When rendering the cube, there is a for loop which sends out a '1' bit-shifted to the left with each increment (or 128 shifted to the right to invert the cube vertically), and a nested for loop which sends one row of the array at a time using the SPI protocol. One loop cycle may look like so:
The code theory is pretty straightforward. There is a 2 dimensional array with 64 slots containing values 0-255 which correspond to which LEDs will be turned on. In other words, it has the values in binary to send to each shift register for a given layer. There are functions throughout the program which then change the values in the array, producing animations. When rendering the cube, there is a for loop which sends out a '1' bit-shifted to the left with each increment (or 128 shifted to the right to invert the cube vertically), and a nested for loop which sends one row of the array at a time using the SPI protocol. One loop cycle may look like so:
00000001 11111111 11111111 11111111 11111111 11111111 11111111 11111111 11111111
In this example, the first byte sent out enables the first layer. The next 8 bytes fill each register corresponding to a column of the array, so each LED on the first layer is turned on. The loop then increments, this time starting with '00000010' followed by the next row of columns in the array and so on.
In this example, the first byte sent out enables the first layer. The next 8 bytes fill each register corresponding to a column of the array, so each LED on the first layer is turned on. The loop then increments, this time starting with '00000010' followed by the next row of columns in the array and so on.
Below is a PDF of the code that I have been working on, and here is the github . For a while, I used Harry Le's code and didn't play around too much, as writing functions to produce different animations is a time intensive task. One day, I opened a blank file and decided to write my own code. I borrowed his render and shift functions, and the rest is mine. What I wrote is in no way supposed to look beautiful; it is a work in progress and I have not yet tried to optimize it.
Below is a PDF of the code that I have been working on, and here is the github . For a while, I used Harry Le's code and didn't play around too much, as writing functions to produce different animations is a time intensive task. One day, I opened a blank file and decided to write my own code. I borrowed his render and shift functions, and the rest is mine. What I wrote is in no way supposed to look beautiful; it is a work in progress and I have not yet tried to optimize it.
Cube Code.pdfA Select Few Animations
A Select Few Animations
IMG_4248.MOVFuture Improvements
Future Improvements
The connectors I made work, but they are fragile and somewhat tedious to use. In the future, I may go for one 2x40 header instead.
The connectors I made work, but they are fragile and somewhat tedious to use. In the future, I may go for one 2x40 header instead.
I used 16 AWG wire to connect the layers on the outside of the cube for extra support and 24 AWG wire on the inside. The 16 AWG wire is hard to solder to, so some of the connections have come loose from time to time and require a very scientific procedure known as "poking it until it works again". I may possibly replace these columns with wire that is easier to work with.
I used 16 AWG wire to connect the layers on the outside of the cube for extra support and 24 AWG wire on the inside. The 16 AWG wire is hard to solder to, so some of the connections have come loose from time to time and require a very scientific procedure known as "poking it until it works again". I may possibly replace these columns with wire that is easier to work with.
Page updated
Google Sites
Report abuse