Merge pull request #2435 from hathach/enhance-uvc

Enhance UVC decriptors and example

5 files changed, 437 insertions, 92 deletions

diff --git a/examples/device/video_capture/src/CMakeLists.txt b/examples/device/video_capture/src/CMakeLists.txt
new file mode 100644
index 000000000..cef2b46ee
--- /dev/null
+++ b/examples/device/video_capture/src/CMakeLists.txt
@@ -0,0 +1,4 @@
+# This file is for ESP-IDF only
+idf_component_register(SRCS "main.c" "usb_descriptors.c"
+                    INCLUDE_DIRS "."
+                    REQUIRES boards tinyusb_src)
diff --git a/examples/device/video_capture/src/main.c b/examples/device/video_capture/src/main.c
index e63335e67..091eb8d0c 100644
--- a/examples/device/video_capture/src/main.c
+++ b/examples/device/video_capture/src/main.c
@@ -48,13 +48,25 @@ enum {
 
 static uint32_t blink_interval_ms = BLINK_NOT_MOUNTED;
 
-void led_blinking_task(void);
-void video_task(void);
+void led_blinking_task(void* param);
+void usb_device_task(void *param);
+void video_task(void* param);
 
-/*------------- MAIN -------------*/
+#if CFG_TUSB_OS == OPT_OS_FREERTOS
+void freertos_init_task(void);
+#endif
+
+
+//--------------------------------------------------------------------+
+// Main
+//--------------------------------------------------------------------+
 int main(void) {
   board_init();
 
+  // If using FreeRTOS: create blinky, tinyusb device, video task
+#if CFG_TUSB_OS == OPT_OS_FREERTOS
+  freertos_init_task();
+#else
   // init device stack on configured roothub port
   tud_init(BOARD_TUD_RHPORT);
 
@@ -64,10 +76,10 @@ int main(void) {
 
   while (1) {
     tud_task(); // tinyusb device task
-    led_blinking_task();
-
-    video_task();
+    led_blinking_task(NULL);
+    video_task(NULL);
   }
+#endif
 }
 
 //--------------------------------------------------------------------+
@@ -169,7 +181,7 @@ static void fill_color_bar(uint8_t* buffer, unsigned start_position) {
 
 #endif
 
-void video_task(void) {
+void video_send_frame(void) {
   static unsigned start_ms = 0;
   static unsigned already_sent = 0;
 
@@ -213,6 +225,21 @@ void video_task(void) {
 #endif
 }
 
+
+void video_task(void* param) {
+  (void) param;
+
+  while(1) {
+    video_send_frame();
+
+    #if CFG_TUSB_OS == OPT_OS_FREERTOS
+    vTaskDelay(interval_ms / portTICK_PERIOD_MS);
+    #else
+    return;
+    #endif
+  }
+}
+
 void tud_video_frame_xfer_complete_cb(uint_fast8_t ctl_idx, uint_fast8_t stm_idx) {
   (void) ctl_idx;
   (void) stm_idx;
@@ -231,16 +258,92 @@ int tud_video_commit_cb(uint_fast8_t ctl_idx, uint_fast8_t stm_idx,
 }
 
 //--------------------------------------------------------------------+
-// BLINKING TASK
+// Blinking Task
 //--------------------------------------------------------------------+
-void led_blinking_task(void) {
+void led_blinking_task(void* param) {
+  (void) param;
   static uint32_t start_ms = 0;
   static bool led_state = false;
 
-  // Blink every interval ms
-  if (board_millis() - start_ms < blink_interval_ms) return; // not enough time
-  start_ms += blink_interval_ms;
+  while (1) {
+    #if CFG_TUSB_OS == OPT_OS_FREERTOS
+    vTaskDelay(blink_interval_ms / portTICK_PERIOD_MS);
+    #else
+    if (board_millis() - start_ms < blink_interval_ms) return; // not enough time
+    #endif
+
+    start_ms += blink_interval_ms;
+    board_led_write(led_state);
+    led_state = 1 - led_state; // toggle
+  }
+}
+
+//--------------------------------------------------------------------+
+// FreeRTOS
+//--------------------------------------------------------------------+
+#if CFG_TUSB_OS == OPT_OS_FREERTOS
+
+#define BLINKY_STACK_SIZE   configMINIMAL_STACK_SIZE
+#define VIDEO_STACK_SIZE    (configMINIMAL_STACK_SIZE*4)
+
+#if TU_CHECK_MCU(OPT_MCU_ESP32S2, OPT_MCU_ESP32S3)
+  #define USBD_STACK_SIZE     4096
+  int main(void);
+  void app_main(void) {
+    main();
+  }
+#else
+  // Increase stack size when debug log is enabled
+  #define USBD_STACK_SIZE    (3*configMINIMAL_STACK_SIZE/2) * (CFG_TUSB_DEBUG ? 2 : 1)
+#endif
+
+// static task
+#if configSUPPORT_STATIC_ALLOCATION
+StackType_t blinky_stack[BLINKY_STACK_SIZE];
+StaticTask_t blinky_taskdef;
+
+StackType_t  usb_device_stack[USBD_STACK_SIZE];
+StaticTask_t usb_device_taskdef;
+
+StackType_t  video_stack[VIDEO_STACK_SIZE];
+StaticTask_t video_taskdef;
+#endif
+
+// USB Device Driver task
+// This top level thread process all usb events and invoke callbacks
+void usb_device_task(void *param) {
+  (void) param;
+
+  // init device stack on configured roothub port
+  // This should be called after scheduler/kernel is started.
+  // Otherwise, it could cause kernel issue since USB IRQ handler does use RTOS queue API.
+  tud_init(BOARD_TUD_RHPORT);
+
+  if (board_init_after_tusb) {
+    board_init_after_tusb();
+  }
 
-  board_led_write(led_state);
-  led_state = 1 - led_state; // toggle
+  // RTOS forever loop
+  while (1) {
+    // put this thread to waiting state until there is new events
+    tud_task();
+  }
 }
+
+void freertos_init_task(void) {
+  #if configSUPPORT_STATIC_ALLOCATION
+  xTaskCreateStatic(led_blinking_task, "blinky", BLINKY_STACK_SIZE, NULL, 1, blinky_stack, &blinky_taskdef);
+  xTaskCreateStatic(usb_device_task, "usbd", USBD_STACK_SIZE, NULL, configMAX_PRIORITIES-1, usb_device_stack, &usb_device_taskdef);
+  xTaskCreateStatic(video_task, "cdc", VIDEO_STACK_SIZE, NULL, configMAX_PRIORITIES - 2, video_stack, &video_taskdef);
+  #else
+  xTaskCreate(led_blinking_task, "blinky", BLINKY_STACK_SIZE, NULL, 1, NULL);
+  xTaskCreate(usb_device_task, "usbd", USBD_STACK_SIZE, NULL, configMAX_PRIORITIES - 1, NULL);
+  xTaskCreate(video_task, "video", VIDEO_STACK_SZIE, NULL, configMAX_PRIORITIES - 2, NULL);
+  #endif
+
+  // skip starting scheduler (and return) for ESP32-S2 or ESP32-S3
+  #if !TU_CHECK_MCU(OPT_MCU_ESP32S2, OPT_MCU_ESP32S3)
+  vTaskStartScheduler();
+  #endif
+}
+#endif
diff --git a/examples/device/video_capture/src/tusb_config.h b/examples/device/video_capture/src/tusb_config.h
index d563c6f5c..bdfc37d87 100644
--- a/examples/device/video_capture/src/tusb_config.h
+++ b/examples/device/video_capture/src/tusb_config.h
@@ -57,6 +57,11 @@
 #define CFG_TUSB_OS           OPT_OS_NONE
 #endif
 
+// Espressif IDF requires "freertos/" prefix in include path
+#if TU_CHECK_MCU(OPT_MCU_ESP32S2, OPT_MCU_ESP32S3)
+#define CFG_TUSB_OS_INC_PATH  freertos/
+#endif
+
 #ifndef CFG_TUSB_DEBUG
 #define CFG_TUSB_DEBUG        0
 #endif
@@ -103,6 +108,9 @@
 // use bulk endpoint for streaming interface
 #define CFG_TUD_VIDEO_STREAMING_BULK 1
 
+//#define CFG_EXAMPLE_VIDEO_READONLY
+//#define CFG_EXAMPLE_VIDEO_DISABLE_MJPEG
+
 #ifdef __cplusplus
  }
 #endif
diff --git a/examples/device/video_capture/src/usb_descriptors.c b/examples/device/video_capture/src/usb_descriptors.c
index d38f158c8..f308c75a7 100644
--- a/examples/device/video_capture/src/usb_descriptors.c
+++ b/examples/device/video_capture/src/usb_descriptors.c
@@ -40,6 +40,26 @@
 #define USB_VID   0xCafe
 #define USB_BCD   0x0200
 
+// String Descriptor Index
+enum {
+  STRID_LANGID = 0,
+  STRID_MANUFACTURER,
+  STRID_PRODUCT,
+  STRID_SERIAL,
+  STRID_UVC_CONTROL,
+  STRID_UVC_STREAMING,
+};
+
+// array of pointer to string descriptors
+char const* string_desc_arr[] = {
+    (const char[]) {0x09, 0x04}, // 0: is supported language is English (0x0409)
+    "TinyUSB",                     // 1: Manufacturer
+    "TinyUSB Device",              // 2: Product
+    NULL,                          // 3: Serials will use unique ID if possible
+    "TinyUSB UVC Control",         // 4: UVC Interface
+    "TinyUSB UVC Streaming",       // 5: UVC Interface
+};
+
 //--------------------------------------------------------------------+
 // Device Descriptors
 //--------------------------------------------------------------------+
@@ -60,9 +80,9 @@ tusb_desc_device_t const desc_device = {
     .idProduct          = USB_PID,
     .bcdDevice          = 0x0100,
 
-    .iManufacturer      = 0x01,
-    .iProduct           = 0x02,
-    .iSerialNumber      = 0x03,
+    .iManufacturer      = STRID_MANUFACTURER,
+    .iProduct           = STRID_PRODUCT,
+    .iSerialNumber      = STRID_SERIAL,
 
     .bNumConfigurations = 0x01
 };
@@ -77,6 +97,19 @@ uint8_t const* tud_descriptor_device_cb(void) {
 // Configuration Descriptor
 //--------------------------------------------------------------------+
 
+/* Time stamp base clock. It is a deprecated parameter. */
+#define UVC_CLOCK_FREQUENCY    27000000
+
+/* video capture path */
+#define UVC_ENTITY_CAP_INPUT_TERMINAL  0x01
+#define UVC_ENTITY_CAP_OUTPUT_TERMINAL 0x02
+
+enum {
+  ITF_NUM_VIDEO_CONTROL,
+  ITF_NUM_VIDEO_STREAMING,
+  ITF_NUM_TOTAL
+};
+
 // Select appropriate endpoint number
 #if TU_CHECK_MCU(OPT_MCU_LPC175X_6X, OPT_MCU_LPC177X_8X, OPT_MCU_LPC40XX)
   // LPC 17xx and 40xx endpoint type (bulk/interrupt/iso) are fixed by its number
@@ -89,61 +122,278 @@ uint8_t const* tud_descriptor_device_cb(void) {
   #define EPNUM_VIDEO_IN    0x81
 #endif
 
-// For mcus that does not have enough SRAM for frame buffer, we use fixed frame data.
-// To further reduce the size, we use MJPEG format instead of YUY2.
-// Select interface descriptor and length accordingly.
 #if defined(CFG_EXAMPLE_VIDEO_READONLY) && !defined(CFG_EXAMPLE_VIDEO_DISABLE_MJPEG)
-  #if CFG_TUD_VIDEO_STREAMING_BULK
-    #define ITF_VIDEO_DESC(epsize)  TUD_VIDEO_CAPTURE_DESCRIPTOR_MJPEG_BULK(4, EPNUM_VIDEO_IN, FRAME_WIDTH, FRAME_HEIGHT, FRAME_RATE, epsize)
-    #define ITF_VIDEO_LEN           TUD_VIDEO_CAPTURE_DESC_MJPEG_BULK_LEN
-  #else
-    #define ITF_VIDEO_DESC(epsize)  TUD_VIDEO_CAPTURE_DESCRIPTOR_MJPEG(4, EPNUM_VIDEO_IN, FRAME_WIDTH, FRAME_HEIGHT, FRAME_RATE, epsize)
-    #define ITF_VIDEO_LEN           TUD_VIDEO_CAPTURE_DESC_MJPEG_LEN
-  #endif
+  #define USE_MJPEG 1
+#else
+  #define USE_MJPEG 0
+#endif
+
+#define USE_ISO_STREAMING (!CFG_TUD_VIDEO_STREAMING_BULK)
+
+typedef struct TU_ATTR_PACKED {
+  tusb_desc_interface_t itf;
+  tusb_desc_video_control_header_1itf_t header;
+  tusb_desc_video_control_camera_terminal_t camera_terminal;
+  tusb_desc_video_control_output_terminal_t output_terminal;
+} uvc_control_desc_t;
+
+/* Windows support YUY2 and NV12
+ * https://docs.microsoft.com/en-us/windows-hardware/drivers/stream/usb-video-class-driver-overview */
+
+typedef struct TU_ATTR_PACKED {
+  tusb_desc_interface_t itf;
+  tusb_desc_video_streaming_input_header_1byte_t header;
+
+#if USE_MJPEG
+  tusb_desc_video_format_mjpeg_t format;
+  tusb_desc_video_frame_mjpeg_continuous_t frame;
+#else
+  tusb_desc_video_format_uncompressed_t format;
+  tusb_desc_video_frame_uncompressed_continuous_t frame;
+#endif
+
+  tusb_desc_video_streaming_color_matching_t color;
+
+#if USE_ISO_STREAMING
+  // For ISO streaming, USB spec requires to alternate interface
+  tusb_desc_interface_t itf_alt;
+#endif
+
+  tusb_desc_endpoint_t ep;
+} uvc_streaming_desc_t;
+
+typedef struct TU_ATTR_PACKED {
+  tusb_desc_configuration_t config;
+  tusb_desc_interface_assoc_t iad;
+  uvc_control_desc_t video_control;
+  uvc_streaming_desc_t video_streaming;
+} uvc_cfg_desc_t;
+
+const uvc_cfg_desc_t desc_fs_configuration = {
+    .config = {
+        .bLength = sizeof(tusb_desc_configuration_t),
+        .bDescriptorType = TUSB_DESC_CONFIGURATION,
+
+        .wTotalLength = sizeof(uvc_cfg_desc_t),
+        .bNumInterfaces = ITF_NUM_TOTAL,
+        .bConfigurationValue = 1,
+        .iConfiguration = 0,
+        .bmAttributes =  TU_BIT(7),
+        .bMaxPower = 100 / 2
+    },
+    .iad = {
+        .bLength = sizeof(tusb_desc_interface_assoc_t),
+        .bDescriptorType = TUSB_DESC_INTERFACE_ASSOCIATION,
+
+        .bFirstInterface = ITF_NUM_VIDEO_CONTROL,
+        .bInterfaceCount = 2,
+        .bFunctionClass = TUSB_CLASS_VIDEO,
+        .bFunctionSubClass = VIDEO_SUBCLASS_INTERFACE_COLLECTION,
+        .bFunctionProtocol = VIDEO_ITF_PROTOCOL_UNDEFINED,
+        .iFunction = 0
+    },
+
+    .video_control = {
+        .itf = {
+            .bLength = sizeof(tusb_desc_interface_t),
+            .bDescriptorType = TUSB_DESC_INTERFACE,
+
+            .bInterfaceNumber = ITF_NUM_VIDEO_CONTROL,
+            .bAlternateSetting = 0,
+            .bNumEndpoints = 0,
+            .bInterfaceClass = TUSB_CLASS_VIDEO,
+            .bInterfaceSubClass = VIDEO_SUBCLASS_CONTROL,
+            .bInterfaceProtocol = VIDEO_ITF_PROTOCOL_15,
+            .iInterface = STRID_UVC_CONTROL
+        },
+        .header = {
+            .bLength = sizeof(tusb_desc_video_control_header_1itf_t),
+            .bDescriptorType = TUSB_DESC_CS_INTERFACE,
+            .bDescriptorSubType = VIDEO_CS_ITF_VC_HEADER,
+
+            .bcdUVC = VIDEO_BCD_1_50,
+            .wTotalLength = sizeof(uvc_control_desc_t) - sizeof(tusb_desc_interface_t), // CS VC descriptors only
+            .dwClockFrequency = UVC_CLOCK_FREQUENCY,
+            .bInCollection = 1,
+            .baInterfaceNr = { ITF_NUM_VIDEO_STREAMING }
+        },
+        .camera_terminal = {
+            .bLength = sizeof(tusb_desc_video_control_camera_terminal_t),
+            .bDescriptorType = TUSB_DESC_CS_INTERFACE,
+            .bDescriptorSubType = VIDEO_CS_ITF_VC_INPUT_TERMINAL,
+
+            .bTerminalID = UVC_ENTITY_CAP_INPUT_TERMINAL,
+            .wTerminalType = VIDEO_ITT_CAMERA,
+            .bAssocTerminal = 0,
+            .iTerminal = 0,
+            .wObjectiveFocalLengthMin = 0,
+            .wObjectiveFocalLengthMax = 0,
+            .wOcularFocalLength = 0,
+            .bControlSize = 3,
+            .bmControls = { 0, 0, 0 }
+        },
+        .output_terminal = {
+            .bLength = sizeof(tusb_desc_video_control_output_terminal_t),
+            .bDescriptorType = TUSB_DESC_CS_INTERFACE,
+            .bDescriptorSubType = VIDEO_CS_ITF_VC_OUTPUT_TERMINAL,
+
+            .bTerminalID = UVC_ENTITY_CAP_OUTPUT_TERMINAL,
+            .wTerminalType = VIDEO_TT_STREAMING,
+            .bAssocTerminal = 0,
+            .bSourceID = UVC_ENTITY_CAP_INPUT_TERMINAL,
+            .iTerminal = 0
+        }
+    },
+
+    .video_streaming = {
+        .itf = {
+            .bLength = sizeof(tusb_desc_interface_t),
+            .bDescriptorType = TUSB_DESC_INTERFACE,
+
+            .bInterfaceNumber = ITF_NUM_VIDEO_STREAMING,
+            .bAlternateSetting = 0,
+            .bNumEndpoints = CFG_TUD_VIDEO_STREAMING_BULK, // bulk 1, iso 0
+            .bInterfaceClass = TUSB_CLASS_VIDEO,
+            .bInterfaceSubClass = VIDEO_SUBCLASS_STREAMING,
+            .bInterfaceProtocol = VIDEO_ITF_PROTOCOL_15,
+            .iInterface = STRID_UVC_STREAMING
+        },
+        .header = {
+            .bLength = sizeof(tusb_desc_video_streaming_input_header_1byte_t),
+            .bDescriptorType = TUSB_DESC_CS_INTERFACE,
+            .bDescriptorSubType = VIDEO_CS_ITF_VS_INPUT_HEADER,
+
+            .bNumFormats = 1,
+            .wTotalLength = sizeof(uvc_streaming_desc_t) - sizeof(tusb_desc_interface_t)
+                - sizeof(tusb_desc_endpoint_t) - (USE_ISO_STREAMING ? sizeof(tusb_desc_interface_t) : 0) , // CS VS descriptors only
+            .bEndpointAddress = EPNUM_VIDEO_IN,
+            .bmInfo = 0,
+            .bTerminalLink = UVC_ENTITY_CAP_OUTPUT_TERMINAL,
+            .bStillCaptureMethod = 0,
+            .bTriggerSupport = 0,
+            .bTriggerUsage = 0,
+            .bControlSize = 1,
+            .bmaControls = { 0 }
+        },
+        .format = {
+#if USE_MJPEG
+            .bLength = sizeof(tusb_desc_video_format_mjpeg_t),
+            .bDescriptorType = TUSB_DESC_CS_INTERFACE,
+            .bDescriptorSubType = VIDEO_CS_ITF_VS_FORMAT_MJPEG,
+            .bFormatIndex = 1, // 1-based index
+            .bNumFrameDescriptors = 1,
+            .bmFlags = 0,
 #else
-  #if CFG_TUD_VIDEO_STREAMING_BULK
-    #define ITF_VIDEO_DESC(epsize)  TUD_VIDEO_CAPTURE_DESCRIPTOR_UNCOMPR_BULK(4, EPNUM_VIDEO_IN, FRAME_WIDTH, FRAME_HEIGHT, FRAME_RATE, epsize)
-    #define ITF_VIDEO_LEN           TUD_VIDEO_CAPTURE_DESC_UNCOMPR_BULK_LEN
-  #else
-    #define ITF_VIDEO_DESC(epsize)  TUD_VIDEO_CAPTURE_DESCRIPTOR_UNCOMPR(4, EPNUM_VIDEO_IN, FRAME_WIDTH, FRAME_HEIGHT, FRAME_RATE, epsize)
-    #define ITF_VIDEO_LEN           TUD_VIDEO_CAPTURE_DESC_UNCOMPR_LEN
-  #endif
+            .bLength = sizeof(tusb_desc_video_format_uncompressed_t),
+            .bDescriptorType = TUSB_DESC_CS_INTERFACE,
+            .bDescriptorSubType = VIDEO_CS_ITF_VS_FORMAT_UNCOMPRESSED,
+            .bFormatIndex = 1, // 1-based index
+            .bNumFrameDescriptors = 1,
+            .guidFormat = { TUD_VIDEO_GUID_YUY2 },
+            .bBitsPerPixel = 16,
 #endif
+            .bDefaultFrameIndex = 1,
+            .bAspectRatioX = 0,
+            .bAspectRatioY = 0,
+            .bmInterlaceFlags = 0,
+            .bCopyProtect = 0
+        },
+        .frame = {
+#if USE_MJPEG
+            .bLength = sizeof(tusb_desc_video_frame_mjpeg_continuous_t),
+            .bDescriptorType = TUSB_DESC_CS_INTERFACE,
+            .bDescriptorSubType = VIDEO_CS_ITF_VS_FRAME_MJPEG,
+#else
+            .bLength = sizeof(tusb_desc_video_frame_uncompressed_continuous_t),
+            .bDescriptorType = TUSB_DESC_CS_INTERFACE,
+            .bDescriptorSubType = VIDEO_CS_ITF_VS_FRAME_UNCOMPRESSED,
+#endif
+            .bFrameIndex = 1, // 1-based index
+            .bmCapabilities = 0,
+            .wWidth = FRAME_WIDTH,
+            .wHeight = FRAME_HEIGHT,
+            .dwMinBitRate = FRAME_WIDTH * FRAME_HEIGHT * 16 * 1,
+            .dwMaxBitRate = FRAME_WIDTH * FRAME_HEIGHT * 16 * FRAME_RATE,
+            .dwMaxVideoFrameBufferSize = FRAME_WIDTH * FRAME_HEIGHT * 16 / 8,
+            .dwDefaultFrameInterval = 10000000 / FRAME_RATE,
+            .bFrameIntervalType = 0, // continuous
+            .dwFrameInterval = {
+                10000000 / FRAME_RATE, // min
+                10000000, // max
+                10000000 / FRAME_RATE // step
+            }
+        },
+        .color = {
+            .bLength = sizeof(tusb_desc_video_streaming_color_matching_t),
+            .bDescriptorType = TUSB_DESC_CS_INTERFACE,
+            .bDescriptorSubType = VIDEO_CS_ITF_VS_COLORFORMAT,
+
+            .bColorPrimaries = VIDEO_COLOR_PRIMARIES_BT709,
+            .bTransferCharacteristics = VIDEO_COLOR_XFER_CH_BT709,
+            .bMatrixCoefficients = VIDEO_COLOR_COEF_SMPTE170M
+        },
 
-#define CONFIG_TOTAL_LEN (TUD_CONFIG_DESC_LEN + ITF_VIDEO_LEN)
+#if USE_ISO_STREAMING
+        .itf_alt = {
+            .bLength = sizeof(tusb_desc_interface_t),
+            .bDescriptorType = TUSB_DESC_INTERFACE,
 
-// full speed descriptor
-uint8_t const desc_fs_configuration[] = {
-    // Config number, interface count, string index, total length, attribute, power in mA
-    TUD_CONFIG_DESCRIPTOR(1, ITF_NUM_TOTAL, 0, CONFIG_TOTAL_LEN, 0, 500),
+            .bInterfaceNumber = ITF_NUM_VIDEO_STREAMING,
+            .bAlternateSetting = 1,
+            .bNumEndpoints = 1,
+            .bInterfaceClass = TUSB_CLASS_VIDEO,
+            .bInterfaceSubClass = VIDEO_SUBCLASS_STREAMING,
+            .bInterfaceProtocol = VIDEO_ITF_PROTOCOL_15,
+            .iInterface = STRID_UVC_STREAMING
+        },
+#endif
 
-    // IAD for Video Control
-    ITF_VIDEO_DESC(CFG_TUD_VIDEO_STREAMING_BULK ? 64 : CFG_TUD_VIDEO_STREAMING_EP_BUFSIZE)
+        .ep = {
+            .bLength = sizeof(tusb_desc_endpoint_t),
+            .bDescriptorType = TUSB_DESC_ENDPOINT,
+
+            .bEndpointAddress = EPNUM_VIDEO_IN,
+            .bmAttributes = {
+                .xfer = CFG_TUD_VIDEO_STREAMING_BULK ? TUSB_XFER_BULK : TUSB_XFER_ISOCHRONOUS,
+                .sync = CFG_TUD_VIDEO_STREAMING_BULK ? 0 : 1 // asynchronous
+            },
+            .wMaxPacketSize = CFG_TUD_VIDEO_STREAMING_BULK ? 64 : CFG_TUD_VIDEO_STREAMING_EP_BUFSIZE,
+            .bInterval = 1
+        }
+    }
 };
 
 #if TUD_OPT_HIGH_SPEED
-// Per USB specs: high speed capable device must report device_qualifier and other_speed_configuration
-uint8_t const desc_hs_configuration[] = {
-  // Config number, interface count, string index, total length, attribute, power in mA
-  TUD_CONFIG_DESCRIPTOR(1, ITF_NUM_TOTAL, 0, CONFIG_TOTAL_LEN, 0, 500),
+uvc_cfg_desc_t desc_hs_configuration;
 
-  // IAD for Video Control
-  ITF_VIDEO_DESC(CFG_TUD_VIDEO_STREAMING_BULK ? 512 : CFG_TUD_VIDEO_STREAMING_EP_BUFSIZE)
-};
+static uint8_t * get_hs_configuration_desc(void) {
+  static bool init = false;
+
+  if (!init) {
+    desc_hs_configuration = desc_fs_configuration;
+    // change endpoint bulk size to 512 if bulk streaming
+    if (CFG_TUD_VIDEO_STREAMING_BULK) {
+      desc_hs_configuration.video_streaming.ep.wMaxPacketSize = 512;
+    }
+  }
+  init = true;
+
+  return (uint8_t *) &desc_hs_configuration;
+}
 
 // device qualifier is mostly similar to device descriptor since we don't change configuration based on speed
 tusb_desc_device_qualifier_t const desc_device_qualifier = {
-  .bLength            = sizeof(tusb_desc_device_t),
-  .bDescriptorType    = TUSB_DESC_DEVICE,
-  .bcdUSB             = USB_BCD,
+    .bLength            = sizeof(tusb_desc_device_t),
+    .bDescriptorType    = TUSB_DESC_DEVICE,
+    .bcdUSB             = USB_BCD,
 
-  .bDeviceClass       = TUSB_CLASS_MISC,
-  .bDeviceSubClass    = MISC_SUBCLASS_COMMON,
-  .bDeviceProtocol    = MISC_PROTOCOL_IAD,
+    .bDeviceClass       = TUSB_CLASS_MISC,
+    .bDeviceSubClass    = MISC_SUBCLASS_COMMON,
+    .bDeviceProtocol    = MISC_PROTOCOL_IAD,
 
-  .bMaxPacketSize0    = CFG_TUD_ENDPOINT0_SIZE,
-  .bNumConfigurations = 0x01,
-  .bReserved          = 0x00
+    .bMaxPacketSize0    = CFG_TUD_ENDPOINT0_SIZE,
+    .bNumConfigurations = 0x01,
+    .bReserved          = 0x00
 };
 
 // Invoked when received GET DEVICE QUALIFIER DESCRIPTOR request
@@ -160,7 +410,11 @@ uint8_t const* tud_descriptor_device_qualifier_cb(void) {
 uint8_t const* tud_descriptor_other_speed_configuration_cb(uint8_t index) {
   (void) index; // for multiple configurations
   // if link speed is high return fullspeed config, and vice versa
-  return (tud_speed_get() == TUSB_SPEED_HIGH) ?  desc_fs_configuration : desc_hs_configuration;
+  if (tud_speed_get() == TUSB_SPEED_HIGH) {
+    return (uint8_t const*) &desc_fs_configuration;
+  } else {
+    return get_hs_configuration_desc();
+  }
 }
 #endif // highspeed
 
@@ -172,33 +426,19 @@ uint8_t const* tud_descriptor_configuration_cb(uint8_t index) {
 
 #if TUD_OPT_HIGH_SPEED
   // Although we are highspeed, host may be fullspeed.
-  return (tud_speed_get() == TUSB_SPEED_HIGH) ?  desc_hs_configuration : desc_fs_configuration;
-#else
-  return desc_fs_configuration;
+  if (tud_speed_get() == TUSB_SPEED_HIGH) {
+    return get_hs_configuration_desc();
+  } else
 #endif
+  {
+    return (uint8_t const*) &desc_fs_configuration;
+  }
 }
 
 //--------------------------------------------------------------------+
 // String Descriptors
 //--------------------------------------------------------------------+
 
-// String Descriptor Index
-enum {
-  STRID_LANGID = 0,
-  STRID_MANUFACTURER,
-  STRID_PRODUCT,
-  STRID_SERIAL,
-};
-
-// array of pointer to string descriptors
-char const* string_desc_arr[] = {
-    (const char[]) {0x09, 0x04}, // 0: is supported language is English (0x0409)
-    "TinyUSB",                     // 1: Manufacturer
-    "TinyUSB Device",              // 2: Product
-    NULL,                          // 3: Serials will use unique ID if possible
-    "TinyUSB UVC",                 // 4: UVC Interface
-};
-
 static uint16_t _desc_str[32 + 1];
 
 // Invoked when received GET STRING DESCRIPTOR request
diff --git a/examples/device/video_capture/src/usb_descriptors.h b/examples/device/video_capture/src/usb_descriptors.h
index fa3c1fbd2..12d41b2f3 100644
--- a/examples/device/video_capture/src/usb_descriptors.h
+++ b/examples/device/video_capture/src/usb_descriptors.h
@@ -27,21 +27,11 @@
 #ifndef _USB_DESCRIPTORS_H_
 #define _USB_DESCRIPTORS_H_
 
-/* Time stamp base clock. It is a deprecated parameter. */
-#define UVC_CLOCK_FREQUENCY    27000000
-/* video capture path */
-#define UVC_ENTITY_CAP_INPUT_TERMINAL  0x01
-#define UVC_ENTITY_CAP_OUTPUT_TERMINAL 0x02
-
 #define FRAME_WIDTH   128
 #define FRAME_HEIGHT  96
 #define FRAME_RATE    10
 
-enum {
-  ITF_NUM_VIDEO_CONTROL,
-  ITF_NUM_VIDEO_STREAMING,
-  ITF_NUM_TOTAL
-};
+// NOTE: descriptor template is not used but leave here as reference
 
 #define TUD_VIDEO_CAPTURE_DESC_UNCOMPR_LEN (\
     TUD_VIDEO_DESC_IAD_LEN\
@@ -146,7 +136,7 @@ enum {
         /* Video stream frame format */ \
         TUD_VIDEO_DESC_CS_VS_FRM_UNCOMPR_CONT(/*bFrameIndex */1, 0, _width, _height, \
             _width * _height * 16, _width * _height * 16 * _fps, \
-            _width * _height * 16, \
+            _width * _height * 16 / 8, \
             (10000000/_fps), (10000000/_fps), (10000000/_fps)*_fps, (10000000/_fps)), \
         TUD_VIDEO_DESC_CS_VS_COLOR_MATCHING(VIDEO_COLOR_PRIMARIES_BT709, VIDEO_COLOR_XFER_CH_BT709, VIDEO_COLOR_COEF_SMPTE170M), \
   /* VS alt 1 */\
@@ -210,7 +200,7 @@ enum {
         /* Video stream frame format */ \
         TUD_VIDEO_DESC_CS_VS_FRM_UNCOMPR_CONT(/*bFrameIndex */1, 0, _width, _height, \
             _width * _height * 16, _width * _height * 16 * _fps, \
-            _width * _height * 16, \
+            _width * _height * 16 / 8, \
             (10000000/_fps), (10000000/_fps), (10000000/_fps)*_fps, (10000000/_fps)), \
         TUD_VIDEO_DESC_CS_VS_COLOR_MATCHING(VIDEO_COLOR_PRIMARIES_BT709, VIDEO_COLOR_XFER_CH_BT709, VIDEO_COLOR_COEF_SMPTE170M), \
         TUD_VIDEO_DESC_EP_BULK(_epin, _epsize, 1)
@@ -223,7 +213,7 @@ enum {
     TUD_VIDEO_DESC_CS_VC(0x0150, TUD_VIDEO_DESC_CAMERA_TERM_LEN + TUD_VIDEO_DESC_OUTPUT_TERM_LEN, UVC_CLOCK_FREQUENCY, ITF_NUM_VIDEO_STREAMING), \
       /* Camera Terminal: ID, bAssocTerminal, iTerminal, focal min, max, length, bmControl */ \
       TUD_VIDEO_DESC_CAMERA_TERM(UVC_ENTITY_CAP_INPUT_TERMINAL, 0, 0, 0, 0, 0, 0), \
-      TUD_VIDEO_DESC_OUTPUT_TERM(UVC_ENTITY_CAP_OUTPUT_TERMINAL, VIDEO_TT_STREAMING, 0, 1, 0), \
+      TUD_VIDEO_DESC_OUTPUT_TERM(UVC_ENTITY_CAP_OUTPUT_TERMINAL, VIDEO_TT_STREAMING, 0, UVC_ENTITY_CAP_INPUT_TERMINAL, 0), \
   /* Video stream alt. 0 */ \
   TUD_VIDEO_DESC_STD_VS(ITF_NUM_VIDEO_STREAMING, 0, 1, _stridx), \
     /* Video stream header for without still image capture */ \