欢迎您访问程序员文章站本站旨在为大家提供分享程序员计算机编程知识!
您现在的位置是: 首页  >  数据库

lcd debug notes 2(qcom and orise)

程序员文章站 2022-06-10 15:29:26
...

1 polarity inversion 面板极性变换方式 可使用的common电极驱动方式 Frame inversion 固定与变动 Row inversion 固定与变动 Column inversion 只能使用固定的common电极电压 Dot inversion 只能使用固定的common电极电压 面板极性变换方式 Flicker的现象 Cr

1 polarity inversion

面板极性变换方式 可使用的common电极驱动方式
Frame inversion 固定与变动
Row inversion 固定与变动
Column inversion 只能使用固定的common电极电压
Dot inversion 只能使用固定的common电极电压


面板极性变换方式 Flicker的现象 Crosstalk(串扰; 串音)的现象
Frame inversion 明显 垂直与水平方向都易发生
Row inversion 不明显 水平方向容易发生
Columninversion 不明显 垂直方向容易发生
Dot inversion 几乎没有 不易发生


Current Orise IC is supported for Column Inversion and Dot Inversion (0xC0B4h Panel Driving Mode)


2 The method of showing layer (qcom)

2.1 Insert the following patch

diff --git a/libhwcomposer/hwc.cpp b/libhwcomposer/hwc.cpp
index ae7bb95..3f1a657 100644
--- a/libhwcomposer/hwc.cpp
+++ b/libhwcomposer/hwc.cpp
@@ -331,11 +331,121 @@ static int hwc_query(struct hwc_composer_device_1* dev,
 
 }
 
+inline void getHalPixelFormatStr(int format, char pixelformatstr[])
+{
+    if (!pixelformatstr)
+        return;
+
+    switch(format) {
+        case HAL_PIXEL_FORMAT_RGBA_8888:
+            strcpy(pixelformatstr, "RGBA_8888");
+            break;
+        case HAL_PIXEL_FORMAT_RGBX_8888:
+            strcpy(pixelformatstr, "RGBX_8888");
+            break;
+        case HAL_PIXEL_FORMAT_RGB_888:
+            strcpy(pixelformatstr, "RGB_888");
+            break;
+        case HAL_PIXEL_FORMAT_RGB_565:
+            strcpy(pixelformatstr, "RGB_565");
+            break;
+        case HAL_PIXEL_FORMAT_BGRA_8888:
+            strcpy(pixelformatstr, "BGRA_8888");
+            break;
+        case HAL_PIXEL_FORMAT_RGBA_5551:
+            strcpy(pixelformatstr, "RGBA_5551");
+            break;
+        case HAL_PIXEL_FORMAT_RGBA_4444:
+            strcpy(pixelformatstr, "RGBA_4444");
+            break;
+        case HAL_PIXEL_FORMAT_YV12:
+            strcpy(pixelformatstr, "YV12");
+            break;
+        case HAL_PIXEL_FORMAT_YCbCr_422_SP:
+            strcpy(pixelformatstr, "YCbCr_422_SP_NV16");
+            break;
+        case HAL_PIXEL_FORMAT_YCrCb_420_SP:
+            strcpy(pixelformatstr, "YCrCb_420_SP_NV21");
+            break;
+        case HAL_PIXEL_FORMAT_YCbCr_422_I:
+            strcpy(pixelformatstr, "YCbCr_422_I_YUY2");
+            break;
+        case HAL_PIXEL_FORMAT_NV12_ENCODEABLE:
+            strcpy(pixelformatstr, "NV12_ENCODEABLE");
+            break;
+        case HAL_PIXEL_FORMAT_YCbCr_420_SP_TILED:
+            strcpy(pixelformatstr, "YCbCr_420_SP_TILED_TILE_4x2");
+            break;
+        case HAL_PIXEL_FORMAT_YCbCr_420_SP:
+            strcpy(pixelformatstr, "YCbCr_420_SP");
+            break;
+        case HAL_PIXEL_FORMAT_YCrCb_420_SP_ADRENO:
+            strcpy(pixelformatstr, "YCrCb_420_SP_ADRENO");
+            break;
+        case HAL_PIXEL_FORMAT_YCrCb_422_SP:
+            strcpy(pixelformatstr, "YCrCb_422_SP");
+            break;
+        case HAL_PIXEL_FORMAT_R_8:
+            strcpy(pixelformatstr, "R_8");
+            break;
+        case HAL_PIXEL_FORMAT_RG_88:
+            strcpy(pixelformatstr, "RG_88");
+            break;
+        case HAL_PIXEL_FORMAT_INTERLACE:
+            strcpy(pixelformatstr, "INTERLACE");
+            break;
+        default:
+            sprintf(pixelformatstr, "Unknown0x%X", format);
+            break;
+    }
+}
+
+void dump_layers(int layercount, size_t layerIndex, hwc_layer_1_t hwLayers[])
+{
+    hwc_layer_1_t *layer = &hwLayers[layerIndex];
+    hwc_rect_t sourceCrop = layer->sourceCrop;
+    hwc_rect_t displayFrame = layer->displayFrame;
+    private_handle_t *hnd = (private_handle_t *)layer->handle;
+    char *layer_format=new char[50];
+    getHalPixelFormatStr(hnd->format,layer_format);
+    int composition_type =layer->compositionType;
+    char sfdumpfile_name[256];
+
+    sprintf(sfdumpfile_name, "/data/layerdump/dump%03d_layer%d_%dx%d_%s_comp%d.raw",
+               layercount, layerIndex, hnd->width, hnd->height,
+               layer_format,composition_type);
+    FILE* fp = fopen(sfdumpfile_name, "w+");
+    if (fp != NULL) {
+        fwrite((void*)hnd->base, hnd->size, 1, fp);
+        fclose(fp);
+    }
+    
+}
+
 static int hwc_set_primary(hwc_context_t *ctx, hwc_display_contents_1_t* list) {
     ATRACE_CALL();
     int ret = 0;
     const int dpy = HWC_DISPLAY_PRIMARY;
     if (LIKELY(list) && ctx->dpyAttr[dpy].isActive) {
+
+    char property_fb[PROPERTY_VALUE_MAX];
+    static int layer_count =0;
+    int count_limit=0;
+    if (property_get("debug.sf.dump", property_fb, NULL) > 0) {
+       if(atoi(property_fb) > 0) {
+               count_limit=atoi(property_fb) ;
+               if (0 == mkdir("/data/layerdump", 0777))
+                       layer_count=0;
+               if(layer_countnumHwLayers;i++)
+                               dump_layers(layer_count, i, list->hwLayers);
+                       layer_count ++ ;
+               }
+       }
+    }
+
+
+
         uint32_t last = list->numHwLayers - 1;
         hwc_layer_1_t *fbLayer = &list->hwLayers[last];
         int fd = -1; //FenceFD from the Copybit(valid in async mode)

2.2 Issue the following command

# setprop debug.sf.dump 100

The dump data is in /data/layerdump folder.


2.3 Issue the following command

# adb pull 

2.4 Download the tool: Irfanview and PlugIns

The newest version of Irfanview is 4.36


2.5 Open the corresponding files, such as *.raw

Please notes that the pixels need to be matched to oringinal size. Then the pictures can be shown, as follows:


dump035_layer0_544x960_RGBX_8888_comp1.raw // layer0 means layer 0; RGBX X means no alpha

dump034_layer1_544x960_RGBA_8888_comp1.raw

dump035_layer2_416x608_RGBA_8888_comp1.raw // layer2 means layer 2; comp1 means mdpcomp

dump033_layer3_544x960_RGBA_8888_comp1.raw

dump035_layer4_544x960_RGBA_8888_comp3.raw // layer4 means layer 4; comp3 means fb_target


3 popup menu jitter(抖动; 颤动)

According to the methods in div 2, we can capture all layers from hwc.cpp. The root cause is related to the progress of animation. Because of the animation, the actual size of popup menu is variable and the alpha value isn't 0xff until the animation is shown completely. But mdp RGB pipe can't handle this scenario smoothly which causes the edge of popup menu jitter. So if this scenario appears, mdp composition should be disabled.


The animation can be disabled via the following steps:

Settings -> Developer options ->

Window animation scale (off, .5x, 1x, 2x, 5x, 10x)

Transition animation scale

Animator duration scale


So the animation progress can be adjusted if you need to do this. Meanwhile we can disable hw overlay via the option of Disable HW overlays.


4 overlay process

4.1 mdp4_overlay_set (key point)
a. Get pipe from request (struct mdp_overlay ), choose corresponding pipe type;
Where the function checks if the request is valid or not, such as width, height, downscale ratio(1/8), upscale ratio(20(mdp version > 4.1, otherwise 8)) and so on.
b. Return pipe index;
c. Calc mdp clock for current frame and bandwidth;
d. Done.

5 Ghost Shadow

This issue is caused by VCOM. For orise 9605, we can adjust 0xD900h(VCOMDC: VCOM voltage setting, 0x39:-1)

影像残留主要是PANEL Pixel上的正负电压值不对称,解决方案是调整VCOM(寄存器D900)。如果改Panel的翻转方式,也可以改善。如果LCM上有GVDD、NGVDD的测试点,可以量一下这两个点的电压,理论上这两个电压值是一致的!

If we adjust the mode of inversion to dot inversion, the power will increase about 10mA, althrough this way can fix this issue too. I don't think it is a good method.


6 DTYPE_DCS_WRITE (long, short)

#define DTYPE_DCS_LWRITE	0x39	/* long write */
#define DTYPE_DCS_WRITE		0x05	/* short write, 0 parameter */
#define DTYPE_DCS_WRITE1	0x15	/* short write, 1 parameter */
#define DTYPE_DCS_READ		0x06	/* read */

With regard to opening lcd or closing lcd(sleep out/in), it is proper to use DTYPE_DCS_WRITE. If the commands have one parameter, it is proper to use DTYPE_DCS_WRITE1. If the commands have more than one parameter, it is proper to use DTYPE_DCS_LWRITE.