fix: respect clk_phase_inverted in PARLIO backend (#114)

Bump version to 0.3.5 [skip ci]

feat: add SCAN_1_8_32PX_FULL wiring for P4-1921 and similar outdoor p…

…anels (#101)

Add intermediate clock speeds: 18, 23, 27 MHz (#79)

Bump version to 0.3.2 [skip ci]

Fix low brightness color issues across all platforms (#66)

* Fix BCM bitplane formula causing color distortion at low brightness

The original bitplane formula `(2 * bit_depth_ - bit) % bit_depth_` produced
incorrect mappings for 8-bit depth:
- bit 0 (LSB) → bitplane 0 (long OE time) - WRONG
- bit 1 → bitplane 7 (short OE time) - WRONG
- bit 7 (MSB) → bitplane 1 (long OE time) - correct

This caused bits 1-3 to have disproportionately short OE times at low
brightness, resulting in muddy/reddish colors because the middle bits
that differentiate hues were being suppressed.

The fix uses `bit_depth_ - 1 - bit` which correctly maps:
- bit 0 (LSB) → bitplane 7 → short OE time
- bit 7 (MSB) → bitplane 0 → long OE time

This ensures proper BCM weighting where MSB gets the longest display
time and LSB gets the shortest.

* Fix PARLIO brightness to use per-bit-plane BCM weighting

PARLIO's calculate_bcm_padding gives bits 0-4 (for lsbMsbTransitionBit=3)
the same padding size, but the brightness code was using uniform max_display
for all bits. This meant bits 0-4 all had equal display time, breaking BCM
ratios (should be 1:2:4:8:16, was 1:1:1:1:1).

Apply the same rightshift-based BCM weighting as GDMA/I2S to achieve proper
BCM ratios via OE duty cycle within the padding section.

* Improve low brightness color fidelity with hybrid minimum approach

Use a hybrid approach that gradually includes more bits for the minimum=1
floor as brightness increases:
- brightness 1-15:  only bit 7 (MSB) gets minimum=1
- brightness 16-31: bits 6-7 get minimum=1
- brightness 32-47: bits 5-7 get minimum=1
- ... and so on

This ensures:
1. Screen is never blank at brightness > 0 (MSB always gets at least 1)
2. At very low brightness, only MSB is visible (no ratio distortion)
3. As brightness increases, more bits naturally exceed 0 anyway
4. Better color fidelity than clamping all bits to 1

Formula: min_bit = bit_depth - 1 - (brightness / 16)

Applied to GDMA, I2S, and PARLIO platforms.

* Revert PARLIO brightness changes (keep GDMA/I2S fixes)

PARLIO uses padding-based BCM timing (not descriptor repetition like
GDMA/I2S), so applying the rightshift formula would cause double BCM
weighting and crush bits 1-2, causing color distortion.

The GDMA/I2S bitplane formula fix and hybrid minimum algorithm are
correct for those platforms.

* Fix PARLIO low brightness with hybrid BCM approach

PARLIO's BCM architecture requires different handling for LSB vs MSB bits:

- Bits > lsbMsbTransitionBit: Have exponential padding sizes that provide
  BCM timing. No rightshift needed (would double-weight).

- Bits <= lsbMsbTransitionBit: Have IDENTICAL padding sizes (all get
  base_padding + base_display). Need rightshift for BCM differentiation
  via OE duty cycle.

Also adds hybrid minimum algorithm (matching GDMA/I2S) to preserve color
ratios at low brightness instead of forcing all bits to display.

* Add minimum brightness floor to maintain BCM ratios

At low brightness (e.g., brightness=4), bits 2-7 all calculate to
display_pixels=1, destroying BCM ratios that should be 1:2:4:8:16:32.

Rather than accepting broken colors at low brightness, establish a
minimum brightness floor of 32 where proper BCM ratios can be maintained.
This gives ~4 bits of color resolution even at the lowest brightness.

Applied to all three platforms: GDMA, I2S, and PARLIO.

* Use smooth brightness remapping instead of clamp

Instead of clamping brightness 1-31 to 32 (losing 31 levels), linearly
remap the full 1-255 range to 32-255. This preserves all 255 brightness
levels while ensuring proper BCM ratios.

Formula: effective = 32 + (brightness * 223) / 255
- brightness=1   → effective=32
- brightness=128 → effective=144
- brightness=255 → effective=255

Intensity is still correctly applied (multiplied before this remapping).

* Calculate minimum brightness dynamically based on panel width

The minimum brightness needed for proper BCM ratios depends on dma_width:
- 64-wide panel: min = 65
- 128-wide panel: min = 32
- 256-wide panel: min = 16

Formula: min_brightness = (16 * 256) / (dma_width - latch_blanking)

This ensures 4-bit BCM ratio range (bits 4-7) regardless of panel size.
User brightness 1-255 is linearly remapped to min_brightness-255.

* Update comments

* simplify

* remove old brightness debugging code

* Lower minimum brightness threshold from 16 to 8

Reduces the minimum brightness floor to allow more usable low-brightness
range. New minimums by panel width:
- 64-wide: 33 (was 65)
- 128-wide: 16 (was 32)
- 256-wide: 8 (was 16)

Trade-off: 3-bit color depth at minimum (8:4:2 ratios) instead of 4-bit
(16:8:4:2). Lower bits contribute minimally due to CIE correction anyway.

* remove double bcm calc

* Fix validation code error

* clean up comments

---------

Co-authored-by: Claude <noreply@anthropic.com>

Make clock speed calculations more explicit (#59)

Add fill method (#45)

Add support for 4-bit and 5-bit color depth (#42)

Extends the minimum supported bit depth from 6-bit to 4-bit, enabling
higher refresh rates on large panel arrays at the cost of reduced
color gradation.

Changes:
- Update static_assert in color_lut.h to allow BitDepth >= 4
- Add DEPTH_4 and DEPTH_5 options to Kconfig with descriptive help text
- Update documentation (COLOR_GAMMA.md, MENUCONFIG.md, README.md) with
  new bit depth options and use case guidance

Fix clang tidy errors (#37)