Fix GC573 speckles and audio crackle on HDMI 2.0 sources

CHANGELOG.md

@@ -7,6 +7,18 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ---
 
+## [Unreleased]
+
+### Fixed
+- **GC573 green/white speckles and audio crackle on HDMI 2.0 sources (4K60, 1440p120):**
+  - The IT6664 input receiver was left at its fixed power-on equaliser (RS `0x1f`, forced), so >340 MHz TMDS links decoded a steady stream of wrong characters. The errors were re-encoded by the IT6664, so they showed up on the HDMI OUT passthrough too and were invisible to the IT6805. Windows was clean on the same hardware and cable.
+  - New `gc573_user_rx_eq.[ch]` drives both receivers (IT6664 input and IT6805 capture share the same ITE receiver core). On HDMI 2.0 lock it measures per-lane TMDS character errors (CED), leaves error-free links alone, and otherwise runs hardware auto-EQ with DFE taps, falling back to a per-lane RS sweep and skew retraining scored by measured errors.
+  - Hardware result (RetroTINK 4K, 4K60 RGB, ~593 MHz): input errors went from ~25,500 per 100 ms to 0 on every lane. Tuning adds ~0.75 s at signal lock.
+
+### Added
+- `[HDMI] TMDS errors B/G/R` line in `streamer.log` with per-lane error counts for the card input and the internal capture hop: every second while errors occur, otherwise every 10 s.
+- `EVERLIVE_NO_AUTO_EQ=1` disables link equalisation; `EVERLIVE_EQ_INPUT_DFE=0` keeps the input on RS-only EQ (for comparison and troubleshooting).
+
 ## [1.4.0] - 2026-11-06
 
 ### Added

tools/opengc573-userd/Makefile

@@ -17,10 +17,10 @@ PW_LIBS = $(shell pkg-config --libs libpipewire-0.3)
 # Frame generation is delegated to the system lsfg-vk Vulkan implicit layer,
 # which the SDL2 "vulkan" renderer feeds directly.  No GL, no Vulkan headers.
 
-COMMON_OBJS = gc573_vfio.o gc573_user_i2c.o gc573_user_hw.o gc573_user_edid.o gc573_user_it6805.o gc573_user_it6664.o third_party/jsmn.o
+COMMON_OBJS = gc573_vfio.o gc573_user_i2c.o gc573_user_hw.o gc573_user_edid.o gc573_user_it6805.o gc573_user_rx_eq.o gc573_user_it6664.o third_party/jsmn.o
 STREAM_OBJS = gc573_user_fpga.o gc573_user_video.o gc573_user_audio.o gc573_user_bitstream.o gc573_user_nosignal.o gc573_user_ipc.o gc573_user_audio_tap.o gc573_user_hdr.o gc573_user_framegen.o
 TARGETS = gc573_user_probe gc573_user_streamer everlive_userd gc573_user_diag
-HEADERS = gc573-user-compat.h gc573_vfio.h gc573_user_i2c.h gc573_user_hw.h gc573_user_edid.h gc573_user_it6805.h gc573_user_it6664.h gc573_user_fpga.h gc573_user_video.h gc573_user_audio.h gc573_user_bitstream.h gc573_user_nosignal.h gc573_user_ipc.h gc573_user_audio_tap.h gc573_user_hdr.h gc573_user_framegen.h
+HEADERS = gc573-user-compat.h gc573_vfio.h gc573_user_i2c.h gc573_user_hw.h gc573_user_edid.h gc573_user_it6805.h gc573_user_rx_eq.h gc573_user_it6664.h gc573_user_fpga.h gc573_user_video.h gc573_user_audio.h gc573_user_bitstream.h gc573_user_nosignal.h gc573_user_ipc.h gc573_user_audio_tap.h gc573_user_hdr.h gc573_user_framegen.h
 
 all: $(TARGETS)
 

tools/opengc573-userd/gc573_user_it6664.c

@@ -2597,6 +2597,21 @@ static int it6664_handle_rx_lock(void *bar0)
 	return 0;
 }
 
+int gc573_user_it6664_relink_outputs(void *bar0)
+{
+	if (!bar0)
+		return -EINVAL;
+	if (!g_rx_locked)
+		return 0;
+	/* Same state reset as an RX sync loss, then the normal lock path. */
+	g_tx2_powered = false;
+	g_tx2_output_enabled = false;
+	g_tx2_error_reported = false;
+	g_tx1_powered = false;
+	g_tx1_pclk_khz = 0;
+	return it6664_handle_rx_lock(bar0);
+}
+
 int gc573_user_it6664_poll(void *bar0, struct gc573_user_it6664_status *status)
 {
 	struct gc573_user_it6664_status local = { 0 };

tools/opengc573-userd/gc573_user_it6664.h

@@ -64,6 +64,14 @@ int gc573_user_it6664_enable_source(void *bar0);
 int gc573_user_it6664_poll(void *bar0,
 			   struct gc573_user_it6664_status *status);
 
+/**
+ * gc573_user_it6664_relink_outputs - Re-arm TX1 (capture) and TX2 (HDMI
+ * OUT) as after a fresh RX lock.  For when the RX input was disturbed (e.g.
+ * by equaliser changes) without the poll loop seeing sync drop, leaving the
+ * outputs on a stale clock.  No-op unless the RX is locked.
+ */
+int gc573_user_it6664_relink_outputs(void *bar0);
+
 /**
  * gc573_user_it6664_pulse_hpd - Pulse upstream HDMI RX HPD to trigger source re-sync
  * @bar0: Pointer to mapped BAR0 memory