Layout Design

...

• I will use Figma to make these mock-ups, as I can easily create a component for the LED and the the auto-layout tool to create the desired grid for the matrix
  

• Assuming I select the devices and a pitch of , this produces a matrix that looks like the following:
  

  
  
  
  
  
  
  

• I learn that, although I can double up for an even maze, I would need many more pixels (and similarly matrix size) to have enough paths

• I do notice however that the odd maze is quite hard to read, perhaps a smaller pitch will make this better

• I will plan on moving forward with a single-pixel line width, and design for an odd-dimensioned matrix such that gaps work out as desired

• With a pitch of , I have a matrix that produces
  

  
  
  
  
  
  
  

• This definitely looks better, but I think the maze could/should still be more readable than what it is

• I chose to keep the row count to , as this allows me to to use a single 48-channel LED driver to drive a whole LED column at once, where I can then multiplex through

• As these are not addressable LEDs (ie incorporate a controller within the LED device), they do not require local decoupling—so I do not need to account for this in designing the pitch

• Reducing the pitch to , I have a matrix
  

  
  
  
  
  
  
  

• I am happy with this, I think this pitch is a good balance between being realistic (assembly, pixel count—price), and being readable with a decent resolution

• The row/column counts of 21 and 35 have also been deliberately chosen as I can use three 36-channel (ie 12 RGB LED) drivers to drive the full row, or two 36-channel drivers to drive a full column

• In terms of the driving mechanism, my first idea would be to use the below topology, with three 36-channel drivers to drive all the columns at once, and scan through the 21 rows with a set of three 8-bit shift registers chained together
  

  

• My next step will be to part out the LED Drivers and Shift Registers that I want to use

• A potential improvement (scan-wise) would be to add three more drivers such that I can split the rows in half, and display two whole rows at once (ie scanning through two lines at once), which would improve the refresh/scan rate of the display, at a cost of component count/cost/driver frequency

• As I am just wanting to scan through the rows one after another and do not need to load some arbitrary data into the shift register before enabling the output all at once, I will not need an output storage register

  • I can simply write the next bit into the shift register once I am ready to scan to the next row, as I only expect the shift register to contain a single 1 at any given instant

• I also do not need a tri-state output, as the shift register output will be the only thing connected to the common anode pins (probably through a buffer transistor)

• From the LED driver perspective—I may want to use more drivers simply such that each driver is controlling less than 32 channels, so that I can represent it using a uint32_t rather than needing to use a whole uint64_t and wasting 28 bits (for a total of 87 wasted bits)

#define DISPLAY_TOTAL_ROWS 35  
#define DISPLAY_TOTAL_COLUMNS 21  
#define DISPLAY_PIXEL_TOTAL_CHANNELS 3  
  
typedef uint128_t row_t;  
typedef uint32_t column_t;  
  
#define DISPLAY_DRIVER_TOTAL_ONE 12 * DISPLAY_PIXEL_TOTAL_CHANNELS  
#define DISPLAY_DRIVER_TOTAL_TWO 11 * DISPLAY_PIXEL_TOTAL_CHANNELS  
#define DISPLAY_DRIVER_TOTAL_THREE 12 * DISPLAY_PIXEL_TOTAL_CHANNELS  
  
typedef uint64_t driver_t; // !
  
row_t currentRow = 0b010/* ... */001;  
  
driver_t driverOne = (currentRow <<= DISPLAY_DRIVER_TOTAL_ONE);  
driver_t driverTwo = (currentRow <<= DISPLAY_DRIVER_TOTAL_TWO);  
driver_t driverThree = (currentRow <<= DISPLAY_DRIVER_TOTAL_THREE);
Copy

• Another option I have is to reduce the column count from 35 to 31, producing a more square matrix
  
  
  • I'm not sure how I feel about this, I think I will pursue the alternative of simply using more drivers for the time being
  • This may work out to be the better decision regardless, as I would expect 24-channel drivers to be easier to source than 48-channel drivers
  • Furthermore, a quick cursory search on DigiKey suggests that 24-channel drivers can be found that sink significantly more (vs ) current than 48-channel drivers
  • Also, although I can find 36-channel drivers, I can only find them in quad flat packages (and with less current capability), whilst I'd prefer a SOP package for routing/aesthetics

Having selected my Shift Registers and LED Drivers to produce the final layout shown below, I will now work through the below design tasks to determine my matrix Design Parameters.