hub75-framebuffer
DMA-friendly framebuffer implementations for driving HUB75 RGB LED matrix
panels with Rust. The crate focuses on performance, correct timing,
and ergonomic drawing by integrating tightly with the embedded-graphics
ecosystem.
How HUB75 LED panels work (very short recap)
A HUB75 panel behaves like a long daisy-chained shift-register:
- Color data for one pair of rows is shifted in serially on every cycle of
CLK. - After the last pixel of the row pair has been clocked, the controller blanks
the LEDs (
OEHIGH), sets the address lines A–E, and produces a short pulse onLATto latch the freshly-shifted data into the LED drivers. OEgoes LOW again and the row pair lights up while the next one is already being shifted.
Color depth is achieved with Binary/Bit-Angle Code Modulation (BCM): lower bit-planes are shown for shorter times, higher ones for longer, yielding 2^n intensity levels per channel while keeping peak currents low.
If you want a deeper explanation, have a look inside src/lib.rs — the crate
documentation contains an extensive primer.
Two framebuffer flavors
| Module | Extra hardware | Word size | Memory use | Pros / Cons |
|---|---|---|---|---|
plain |
none | 16 bit (14 used) | high | Simplest, wires exactly like a standard HUB75 matrix. |
latched |
external latch gate (see below) | 8 bit | ×½ of plain |
Lower memory footprint, but needs a tiny glue-logic board. |
Multiple Panels
- Use
tiling::TiledFrameBufferto drive several HUB75 panels as one large display. - Combine it with a pixel-remapping policy like
ChainTopRightDownand any of the framebuffers above (plain or latched). - The wrapper exposes a single
embedded-graphicscanvas, so a 3 × 3 stack of 64 × 32 panels simply looks like a 192 × 96 screen while all coordinate translation happens transparently.
The latch circuit
The latched implementation assumes a small external circuit that holds the row address while gating the pixel clock. A typical solution uses a 74xx373 latch along with a few NAND gates:

The latch IC stores the address bits whilst one NAND gate blocks the CLK
signal during the latch interval. The remaining spare gate can be employed
to combine a global PWM signal with OE for fine-grained brightness control
as shown.
Getting started
Add the dependency to your Cargo.toml:
[]
= "0.6.0"
Choose your parameters
use ;
use DmaFrameBuffer;
// or ::plain::DmaFrameBuffer
const ROWS: usize = 32; // panel height
const COLS: usize = 64; // panel width
const BITS: u8 = 3; // colour depth ⇒ 7 BCM frames
const NROWS: usize = compute_rows; // 16
const FRAME_COUNT:usize = compute_frame_count; // (1<<BITS)-1 = 7
// Create a framebuffer (already initialized/cleared)
let mut framebuffer =
new;
You can now draw using any embedded-graphics primitive:
use *;
use ;
use Color;
new
.into_styled
.draw
.unwrap;
new
.into_styled
.draw
.unwrap;
Finally hand the raw DMA buffer off to your MCU's parallel peripheral.
Crate features
esp-hal-dma (required when using esp-hal)
Required when using the esp-hal crate for ESP32 development. This
feature switches the ReadBuffer trait implementation from embedded-dma
to esp-hal::dma. If you're targeting ESP32 devices with esp-hal, you
must enable this feature for DMA compatibility.
[]
= { = "0.6.0", = ["esp-hal-dma"] }
esp32-ordering (required for original ESP32 only)
Required when targeting the original ESP32 chip (not ESP32-S3 or other variants). This feature adjusts byte ordering to accommodate the quirky requirements of the ESP32's I²S peripheral in 8-bit and 16-bit modes. Other ESP32 variants (S2, S3, C3, etc.) do not need this feature.
[]
= { = "0.6.0", = ["esp32-ordering"] }
skip-black-pixels
Skip drawing black pixels for performance boost in UI applications. When
enabled, calls to set_pixel() with Color::BLACK return early without
writing to the framebuffer, assuming the framebuffer was already cleared.
defmt
Implement the defmt::Format trait so framebuffer types can be logged with
the defmt ecosystem.
doc-images
Embed documentation images when building docs on docs.rs. Not needed for normal usage.
Enable features in your Cargo.toml:
[]
= { = "0.6.0",
features = ["esp-hal-dma", "esp32-ordering"] }
Running tests
cargo test
All logic including bitfields, address mapping, brightness modulation and
the embedded-graphics integration is covered by a comprehensive test-suite
(≈ 300 tests).
License
Licensed under either of
- Apache License, Version 2.0, (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
- MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)
at your option.