hd108.h

Go to the documentation of this file.

#pragma once
 
#include "eorder.h"
// IWYU pragma: begin_keep
#include "platforms/shared/spi_pixel_writer.h"  // ok platform headers
// IWYU pragma: end_keep
#include "fastspi.h"
#include "cpixel_ledcontroller.h"
#include "fl/math/ease.h"
#include "fl/stl/noexcept.h"

//
// HD108/NS108 16-bit SPI chipset
// - SPI mode: MODE3 recommended (CPOL=1, CPHA=1)
//
// Key features:
// 1. 16-bit color depth (65536 levels per channel)
// 2. 5-bit global current control (0-31) for brightness management
// 3. Higher color accuracy than APA102, especially at low brightness
// 4. Dual-byte header encoding for brightness/current control
//
// Protocol:
// - Start frame: 64 bits (8 bytes) of zeros
// - Per LED: 2 header bytes + 6 color bytes (RGB @ 16-bit each)
// - End frame: (num_leds / 2) + 4 bytes of 0xFF for latching
//
// References:
// - GitHub Issue #1045: Community protocol discussion
// - Pull Request #2119: Initial implementation by arfoll
// - Manufacturer: www.hd108-led.com protocol documentation
//
 
// HD108 Controller class
template <int DATA_PIN, fl::u8 CLOCK_PIN, EOrder RGB_ORDER = GRB, fl::u32 SPI_SPEED = DATA_RATE_MHZ(25)>
class HD108Controller : public CPixelLEDController<RGB_ORDER> {
    typedef fl::SPIOutput<DATA_PIN, CLOCK_PIN, SPI_SPEED> SPI;
    SPI mSPI;
 
public:
    HD108Controller() FL_NOEXCEPT {}
 
    void init() override { mSPI.init(); }
 
protected:
void showPixels(PixelController<RGB_ORDER> &pixels) override {
    mSPI.select();
 
    // ---- Start frame: 64 bits of 0 ----
    // HD108 requires 8 bytes (64 bits) of zeros to initialize the strip
    // Note: Some sources mention 128 bits (16 bytes), but 64 bits works reliably
    for (int i = 0; i < 8; i++) mSPI.writeByte(0x00);
 
    while (pixels.has(1)) {
        // Load and scale pixel data in OUTPUT order (respects RGB_ORDER template parameter)
        fl::u8 c0_8, c1_8, c2_8;
        pixels.loadAndScaleRGB(&c0_8, &c1_8, &c2_8);
 
        // Apply gamma correction (2.8) to convert 8-bit to 16-bit for HD108
        // This provides smooth perceptual brightness transitions across the full 65K range
        // Note: Brightness scaling is applied by loadAndScaleRGB() before gamma correction
        fl::u16 c0_16 = fl::gamma_2_8(c0_8);
        fl::u16 c1_16 = fl::gamma_2_8(c1_8);
        fl::u16 c2_16 = fl::gamma_2_8(c2_8);
 
        // Header bytes: HD108 per-channel gain control encoding
        // f0: [1][RRRRR][GG] - marker bit, 5-bit R gain, 2 MSBs of G gain
        // f1: [GGG][BBBBB]   - 3 LSBs of G gain, 5-bit B gain
        // Use maximum gain (31) for all channels to maximize precision
        // Brightness control via 16-bit PWM values (already applied via loadAndScaleRGB)
        constexpr fl::u8 r_gain = 31, g_gain = 31, b_gain = 31;
        constexpr fl::u8 f0 = fl::u8(0x80 | ((r_gain & 0x1F) << 2) | ((g_gain >> 3) & 0x03));
        constexpr fl::u8 f1 = fl::u8(((g_gain & 0x07) << 5) | (b_gain & 0x1F));
 
        // Transmit LED frame: 2 header bytes + 6 color bytes (16-bit, big-endian, in RGB_ORDER)
        mSPI.writeByte(f0);
        mSPI.writeByte(f1);
        mSPI.writeByte(fl::u8(c0_16 >> 8)); mSPI.writeByte(fl::u8(c0_16 & 0xFF));  // Channel 0 MSB, LSB
        mSPI.writeByte(fl::u8(c1_16 >> 8)); mSPI.writeByte(fl::u8(c1_16 & 0xFF));  // Channel 1 MSB, LSB
        mSPI.writeByte(fl::u8(c2_16 >> 8)); mSPI.writeByte(fl::u8(c2_16 & 0xFF));  // Channel 2 MSB, LSB
 
        pixels.stepDithering();
        pixels.advanceData();
    }
 
    // ---- End frame: 0xFF bytes to latch data into LEDs ----
    // Formula: (num_leds / 2) + 4 provides sufficient clock pulses for 40 MHz operation
    // This is more conservative than APA102's (num_leds + 15) / 16 formula
    const int latch = pixels.size() / 2 + 4;
    for (int i = 0; i < latch; i++) mSPI.writeByte(0xFF);
    mSPI.endTransaction();
}
};