Researching DVD Subtitle Format

I am attempting to stream webcam video from a BeagleBoneBlack to other computers over ethernet. I want to add an overlay with details about the video. I am capturing video from a Logitech C920 webcam, which is doing the hard work of creating the video on H.264 format, using FFMPEG to MUX the video into a network stream. The current video stream runs at 3Mb/s over ethernet, and seems to run at the same bitrate whether I’m sending video 30FPS at 1920×1080, 1280×720, or any other resolution I’ve tried. If I’m running the BBB at 1GHz FFMPEG uses only 3% load on the processor, while at 300MHz it uses 10% load. Either processor speed indicates that I should have plenty of CPU available for creating a subtitle frame a second.

If I transcode the H264 coming from the C920 to h.264 from FFMPEG the BBB CPU is 100% used and I’ve not been able to get over 5 FPS. This has led me to the idea of adding a second stream with much more compressible data and requiring the client computer to know how to enable subtitles.

My understanding of DVD Subtitles is that they are stored as image overlays. The images seem to be 3 color plus transparency, with the color indexed. They are RLE (Run Length Encoded) images but don’t seem to conform to any standard that would be created by an image library such as OpenCV.

The most useful links I’ve come across related to the DVD subtitles are these three:

Using FFMPEG to examine at a video that was ripped from a DVD into an MKV file with several subtitle layers shows the following:

Stream #0:0(eng): Video: mpeg2video (Main), yuv420p, 720x480 [SAR 32:27 DAR 16:9], SAR 186:157 DAR 279:157, 29.97 fps, 29.97 tbr, 1k tbn, 59.94 tbc
Stream #0:1(eng): Audio: ac3, 48000 Hz, 5.1(side), fltp, 448 kb/s (default)
Metadata:
  title           : 3/2+1
Stream #0:2(eng): Audio: ac3, 48000 Hz, 5.1(side), fltp, 384 kb/s
Metadata:
  title           : 3/2+1
Stream #0:3(spa): Audio: ac3, 48000 Hz, stereo, fltp, 192 kb/s
Metadata:
  title           : 2/0
Stream #0:4(fre): Audio: ac3, 48000 Hz, stereo, fltp, 192 kb/s
Metadata:
  title           : 2/0
Stream #0:5(eng): Subtitle: dvd_subtitle (default)
Stream #0:6(spa): Subtitle: dvd_subtitle
Stream #0:7(eng): Subtitle: dvd_subtitle
Stream #0:8(spa): Subtitle: dvd_subtitle
Stream #0:9(fre): Subtitle: dvd_subtitle

All of the descriptions of creating subtitle tracks are directly related to creating textual subtitles using tools that are wonderful for mainstream movie content but not what I want to do. e.g.

I’ve not figured out how to create my own subtitle stream and am still looking for information on that. I’ve not figured out what parameters may need to be passed to FFMPEG to indicate that I’m passing in a subtitle track. I’ve not figured out if there’s a way in FFMPEG to indicate that the subtitles should be on by default, or forced subtitles, while still keeping them as a separate stream.

It doesn’t help that the DVD subtitle files seem to use the STL extension and that same extension is used for the input files for many 3D Printers.

BeagleBoneBlack Webcam using a V4L2, FFMPEG and a FIFO

In my previous post I was starting FFMPEG with the command to read from standard input using a pipe. That’s a great method when you have a single input you want to supply to FFMPEG, but my ultimate goal is to supply multiple streams to FFMPEG and have it multiplex them into a single transport stream.

I believe that my end solution for this is going to be using several named pipes, which are First In First Out devices created by the command mkfifo() in linux.

The following code snippet functions nearly identically to the snippet in my previous post except that it creates a FIFO in the /tmp/ directory and tells FFMPEG to read from the FIFO instead of from stdin.

if (0 != mkfifo("/tmp/Wim_C920_FFMPEG", S_IRWXU))
	std::cerr << "[" << getTimeISO8601() << "] FiFo /tmp/Wim_C920_FFMPEG NOT created Successfully" << std::endl;
else
	std::cerr << "[" << getTimeISO8601() << "] FiFo /tmp/Wim_C920_FFMPEG created Successfully" << std::endl;
/* Attempt to fork */
pid_t pid = fork();
if(pid == -1)
{
	fprintf(stderr,"Fork error! Exiting.\n");  /* something went wrong */
	exit(1);        
}
else if (pid == 0)
{
	/* A zero PID indicates that this is the child process */
	/* Replace the child fork with a new process */
	if(execlp("ffmpeg", "ffmpeg", "-i", "/tmp/Wim_C920_FFMPEG", "-vcodec", "copy", "-f", "rtp", "rtp://239.8.8.8:8090/", "-metadata", "title=\"Wims BoneCam\"", "-vcodec", "copy", OutputFilename.c_str(), NULL) == -1)
	{
		std::cerr << "execlp Error! Exiting." << std::endl;
		exit(1);
	}
}
else
{
	/* A positive (non-negative) PID indicates the parent process */
	open_device();
	init_device();
	start_capturing();
	//mainloop();
	int pipe_fd = open("/tmp/Wim_C920_FFMPEG", O_WRONLY);
	if (pipe_fd <= 0)
		std::cerr << "[" << getTimeISO8601() << "] FiFo NOT opened Successfully" << std::endl;
	else
	{
		std::cerr << "[" << getTimeISO8601() << "] FiFo opened Successfully" << std::endl;
		time_t CurrentTime;
		time(&CurrentTime);
		time_t FileStartTime = CurrentTime;
		while ((difftime(CurrentTime, FileStartTime) < (60 * 60)) && (bRun))
		{
			fd_set fds;
			struct timeval tv;
			int r;

			FD_ZERO(&fds);
			FD_SET(fd, &fds);

			/* Timeout. */
			tv.tv_sec = 2;
			tv.tv_usec = 0;

			r = select(fd + 1, &fds, NULL, NULL, &tv);
			if (-1 == r) 
			{
				if (EINTR == errno)
					continue;
				errno_exit("select");
			}
			if (0 == r) 
			{
				fprintf(stderr, "select timeout\n");
				exit(EXIT_FAILURE);
			}
			struct v4l2_buffer buf;
			WIMLABEL:
			CLEAR(buf);
			buf.type = V4L2_BUF_TYPE_VIDEO_CAPTURE;
			buf.memory = V4L2_MEMORY_MMAP;

			if (-1 == xioctl(fd, VIDIOC_DQBUF, &buf)) 
			{
				switch (errno) 
				{
				case EAGAIN:
					goto WIMLABEL;
				case EIO:
					/* Could ignore EIO, see spec. */
					/* fall through */
				default:
					errno_exit("VIDIOC_DQBUF");
				}
			}
			else
			{
				assert(buf.index < n_buffers);
				write(pipe_fd, buffers[buf.index].start, buf.bytesused);
			}
			if (-1 == xioctl(fd, VIDIOC_QBUF, &buf))
					errno_exit("VIDIOC_QBUF");
			time(&CurrentTime);
		}
		stop_capturing();
		uninit_device();
		close_device();
		close(pipe_fd);		/* Close side of pipe I'm writing to, to get the child to recognize it's gone away */
		std::cerr << "\n[" << getTimeISO8601() << "] Pipe Closed, Waiting for FFMPEG to exit" << std::endl;
	}
	int ffmpeg_exit_status;
	wait(&ffmpeg_exit_status);				/* Wait for child process to end */
	std::cerr << "[" << getTimeISO8601() << "] FFMPEG exited with a  " << ffmpeg_exit_status << " value" << std::endl;
	if (0 == remove("/tmp/Wim_C920_FFMPEG"))
		std::cerr << "[" << getTimeISO8601() << "] FiFo /tmp/Wim_C920_FFMPEG removed Successfully" << std::endl;
	else
		std::cerr << "[" << getTimeISO8601() << "] FiFo /tmp/Wim_C920_FFMPEG NOT removed Successfully" << std::endl;
}

An interesting side effect is that ffmpeg now can read from stdin for arbitrary commands, as if it had been run from the linux command line, and so the "q" key will cause it to quit running. I've not fully investigated this behavior to see what happens to the parent program when the fifo consumer suddenly quits. Initially, it seems that the fifo supplier quits as well.

A bad thing about going with named pipes is that there should be a way of ensuring that the names used are unique and don't conflict with anything else on the system. They should also be cleaned up at the end of the process. My current method has very little error correction and should not be used as an example beyond to get started.

BeagleBoneBlack Webcam using a V4L2, FFMPEG and a Pipe

I’ve been working on streaming video from my BeagleBoneBlack (BBB) over WiFi and also keeping a copy of the video on the local flash. My starting point for capturing the video using Video For Linux 2 (V4L2) came from demonstration code from Derek Molloy’s site. He has some very well done instructional videos that are much more accessible and complete than most of what I’ve put together.

I am doing most of my development in C/C++ simply because it’s my strongest language. I’m doing the development on the BBB under linux which means that some of the skills I’m using are fairly rusty.

Because I wanted to have the program pulling the video from the camera in charge of the entire process, including managing which local output files are created, and making sure that there is disk storage space, I decided to use pipe(), fork(), and exec(), to manage ffmpeg as a child process of my program.

I started with the same demonstration program from V4L2 and then modified it so that instead of writing out to stdout, it’s writing to a file descriptor I supply. This all happens inside of the primary loop that generates the outputfilename for the ffmpeg command line and makes sure that there is enough free storage for an hours worth of video. Here’s the main code snippet.

int	pipefd[2];		/* This holds the fd for the input & output of the pipe */
/* Setup communication pipeline first */
if(pipe(pipefd))
{
	std::cerr << "[" << getTimeISO8601() << "] Pipe error! Exiting." << std::endl;
	exit(1);
}
/* Attempt to fork */
pid_t pid = fork();
if(pid == -1)
{
	std::cerr << "[" << getTimeISO8601() << "] Fork error! Exiting." << std::endl; /* something went wrong */
	exit(1);        
}
else if (pid == 0)
{
	/* A zero PID indicates that this is the child process */
	dup2(pipefd[0], 0);	/* Replace stdin with the in side of the pipe */
	close(pipefd[1]);	/* Close unused side of pipe (out side) */
	/* Replace the child fork with a new process */
	if(execlp("ffmpeg", "ffmpeg", "-i", "-", "-vcodec", "copy", "-f", "rtp", "rtp://239.8.8.8:8090/", "-metadata", "title=\"Wims BoneCam\"", "-vcodec", "copy", OutputFilename.c_str(), NULL) == -1)
	{
		std::cerr << "[" << getTimeISO8601() << "] execlp Error! Exiting." << std::endl;
		exit(1);
	}
}
else
{
	/* A positive (non-negative) PID indicates the parent process */
	close(pipefd[0]);		/* Close unused side of pipe (in side) */
	open_device();
	init_device();
	start_capturing();
	//mainloop();

	time_t CurrentTime;
	time(&CurrentTime);
	time_t FileStartTime = CurrentTime;
	while ((difftime(CurrentTime, FileStartTime) < (60 * 60)) && (bRun))
	//unsigned int count = frame_count;
	//while ((count-- > 0) && (bRun))
	{
		fd_set fds;
		struct timeval tv;
		int r;

		FD_ZERO(&fds);
		FD_SET(fd, &fds);

		/* Timeout. */
		tv.tv_sec = 2;
		tv.tv_usec = 0;

		r = select(fd + 1, &fds, NULL, NULL, &tv);
		if (-1 == r) 
		{
			if (EINTR == errno)
				continue;
			errno_exit("select");
		}
		if (0 == r) 
		{
			fprintf(stderr, "select timeout\n");
			exit(EXIT_FAILURE);
		}
		struct v4l2_buffer buf;
		WIMLABEL:
		CLEAR(buf);
		buf.type = V4L2_BUF_TYPE_VIDEO_CAPTURE;
		buf.memory = V4L2_MEMORY_MMAP;

		if (-1 == xioctl(fd, VIDIOC_DQBUF, &buf)) 
		{
			switch (errno) 
			{
			case EAGAIN:
				goto WIMLABEL;
			case EIO:
				/* Could ignore EIO, see spec. */
				/* fall through */
			default:
				errno_exit("VIDIOC_DQBUF");
			}
		}
		else
		{
			assert(buf.index < n_buffers);
			write(pipefd[1], buffers[buf.index].start, buf.bytesused);
		}
		if (-1 == xioctl(fd, VIDIOC_QBUF, &buf))
				errno_exit("VIDIOC_QBUF");
		time(&CurrentTime);
	}

	stop_capturing();
	uninit_device();
	close_device();
	close(pipefd[1]);		/* Close side of pipe I'm writing to, to get the child to recognize it's gone away */
	std::cerr << "\n[" << getTimeISO8601() << "] Pipe Closed, Waiting for FFMPEG to exit" << std::endl;
	int ffmpeg_exit_status;
	wait(&ffmpeg_exit_status);				/* Wait for child process to end */
	std::cerr << "[" << getTimeISO8601() << "] FFMPEG exited with a  " << ffmpeg_exit_status << " value" << std::endl;
}

You can see the calls to open_device(), init_device(), and start_capturing() that I didn’t change at all. My loop runs for a particular amount of time instead of number of frames. I flattened the call to mainloop() into this routine in a very quick and dirty fashion to make sure I had access to the file descriptor to write the data from the camera to the input for ffmpeg.

ffmpeg is started with the execlp() command. The parameters -i – tell ffmpeg to read it’s input from standard input, which in this case I’ve pointed to be the output end of the pipe I created.

Webcam on BeagleBoardBlack using OpenCV

I’ve been working with my BBB and Logitech C920 webcam trying to stream video at low latency for some time and have not yet managed to get the latency under 2 seconds.

As a side project I wanted to use the BBB to create a time lapse video, capturing a picture a second, and then later stitching all of the pictures into a video using ffmpeg.

I’m using OpenCV for the first time. I’m really only using it for the capture/save and to draw some text and lines onto the image, which probably makes OpenCV significant overkill.

My C++ code for the process is:

#include <iostream> // for standard I/O
#include <string>   // for strings
#include <iomanip>  // for controlling float print precision
#include <sstream>  // string to number conversion
#include <unistd.h> // for sleep
#include<opencv2/opencv.hpp>
using namespace std;
using namespace cv;

std::string timeToISO8601(const time_t & TheTime)
{
	std::ostringstream ISOTime;
	struct tm * UTC = gmtime(&TheTime);
	ISOTime.fill('0');
	ISOTime << UTC->tm_year+1900 << "-";
	ISOTime.width(2);
	ISOTime << UTC->tm_mon+1 << "-";
	ISOTime.width(2);
	ISOTime << UTC->tm_mday << "T";
	ISOTime.width(2);
	ISOTime << UTC->tm_hour << ":";
	ISOTime.width(2);
	ISOTime << UTC->tm_min << ":";
	ISOTime.width(2);
	ISOTime << UTC->tm_sec;
	ISOTime << "Z";
	return(ISOTime.str());
}
std::string getTimeISO8601(void)
{
	time_t timer;
	time(&timer);
	return(timeToISO8601(timer));
}

int main()
{
    VideoCapture capture(-1);	// Using -1 tells OpenCV to grab whatever camera is available.
    if(!capture.isOpened()){
	    std::cout << "Failed to connect to the camera." << std::endl;
		return(1);
    }
    capture.set(CAP_PROP_FRAME_WIDTH,1920);
    capture.set(CAP_PROP_FRAME_HEIGHT,1080);
    //capture.set(CAP_PROP_FRAME_WIDTH,2304);	// This should be possible for still images, but not for 30fps video.
    //capture.set(CAP_PROP_FRAME_HEIGHT,1536);

	for (int OutputFolderNum = 100;	OutputFolderNum < 1000; OutputFolderNum++)
		for (int OutputImageNum = 1; OutputImageNum < 10000; OutputImageNum++)
		{
			Mat C920Image;
		    capture >> C920Image;
			if(!C920Image.empty())
			{
				std::ostringstream OutputFilename;
				OutputFilename.fill('0');
				OutputFilename << "/media/BONEBOOT/DCIM/";
				OutputFilename.width(3);
				OutputFilename << OutputFolderNum;
				OutputFilename << "WIMBO/img_";
				OutputFilename.width(4);
				OutputFilename << OutputImageNum;
				OutputFilename << ".jpg";

				line(C920Image, Point(0, C920Image.rows/2), Point(C920Image.cols, C920Image.rows/2), Scalar(255, 255, 255, 32)); // Horizontal line at center
				line(C920Image, Point(C920Image.cols/2, 0), Point(C920Image.cols/2, C920Image.rows), Scalar(255, 255, 255, 32)); // Vertical line at center

				circle(C920Image, Point(C920Image.cols/2, C920Image.rows/2), 240, Scalar(255, 255, 255, 32)); // Circles based at center
				putText(C920Image, "10", Point((C920Image.cols/2 + 240), (C920Image.rows/2)), FONT_HERSHEY_SIMPLEX, 1.0, Scalar(0, 0, 255));
				circle(C920Image, Point(C920Image.cols/2, C920Image.rows/2), 495, Scalar(255, 255, 255, 32)); // Circles based at center
				putText(C920Image, "20", Point((C920Image.cols/2 + 495), (C920Image.rows/2)), FONT_HERSHEY_SIMPLEX, 1.0, Scalar(0, 0, 255));
				circle(C920Image, Point(C920Image.cols/2, C920Image.rows/2), 785, Scalar(255, 255, 255, 32)); // Circles based at center
				putText(C920Image, "30", Point((C920Image.cols/2 + 785), (C920Image.rows/2)), FONT_HERSHEY_SIMPLEX, 1.0, Scalar(0, 0, 255));
				circle(C920Image, Point(C920Image.cols/2, C920Image.rows/2), 1141, Scalar(255, 255, 255, 32)); // Circles based at center
				putText(C920Image, "40", Point((C920Image.cols/2 + 1141), (C920Image.rows/2)), FONT_HERSHEY_SIMPLEX, 1.0, Scalar(0, 0, 255));

				string DateTimeText = "WimsWorld.com " + getTimeISO8601();
				int baseline=0;
				Size textSize = getTextSize(DateTimeText, FONT_HERSHEY_SIMPLEX, 1, 1, &baseline);
				putText(C920Image, DateTimeText, Point((C920Image.cols - textSize.width), (C920Image.rows - baseline)), FONT_HERSHEY_SIMPLEX, 1.0, Scalar(0, 0, 255));
				imwrite(OutputFilename.str(), C920Image);
				std::cout << DateTimeText << " Wrote File : " << OutputFilename.str() << std::endl;
			}
			std::cout << getTimeISO8601() << "\r" << std::flush;
			sleep(1);
		}
    return 0;
}

I compile it on the BBB with the command:

g++ -O2 `pkg-config --cflags --libs opencv` TimeLapse.cpp -o TimeLapse

I’ve got a bug in that I don’t automatically create the directory structure that I’m saving files into. That’s in the to-do list.

I had been interested in the angle of view on the C920 and found it defined on the Logitech support site that the “Diagonal Field of View (FOV) for the Logitech C920 is 78°”. Unfortunately I was not able to understand if that varied based on the resolution being used. I’m currently using the resolution of 1920×1080, but for stills the camera can capture up to 2304×1536.

I did the geometry math to figure out that 10° off center would be a radius of 240, 20° off center would be a radius of 495, and 30° off center would be a radius of 785. Remembering SOHCAHTOA as Some Old Hags Can’t Always Hide Their Old Age from 9th grade math class came in useful. Using 1920×1080 and 78°angle, my diagonal radius (opposite) works out at 1101 and angle of 39° for tangent, allowing me to calculate my eye height of 1360 = (1101/Tan(39°)). Once I had my eye height I could calculate the radius of circles at any angle by Radius = Tan(Angle) * EyeHeight.

I wanted the circles and angles of vision for my streaming video application and decided that seeing them drawn on the images created here would be helpful, along with both the horizontal and vertical center lines.

The thing I’m not happy with is that the application seems to be running between 30% and 60% of the CPU load on the BBB. When I stream video from the C920 using the native H.264 output the C920 can produce, I was only using about 3% of the BBB CPU. I’ve commented out my drawing code, and verified that the CPU load is primarily related to acquiring the image from the capture device and saving it out to a jpeg file. The lines and text drawing produce minimal incremental CPU. I want to keep the CPU load as low as possible because I’m powering this device from a battery and want it to have as long a runtime as possible.

I believe that the OpenCV library is opening the capture device in a movie streaming mode, and it’s using more CPU interpreting the stream as it’s coming in than the method I was using for streaming to a file. I’ve not yet figured out if there’s a way to define what mode OpenCV acquires the image from the camera.

I was trying to draw the lines and circles with some alpha transparency, but it seems that my underlying image is not the right number of channels and so the lines are being drawn fully opaque.

When the capture opens, it outputs several instances of the same error “VIDIOC_QUERYMENU: Invalid argument” that I’ve not figured out what they mean, or stopped procucing.

I am working on a 32GB flash card, partitioned into two 16GB filesystems. The first is Fat32, has a simple uEnv.txt file in the root allowing the BBB onboard flash to be used, and following the Design rules for Camera File systems standard for the image naming. It allows me to take out the card put it in a PC and it’s recognized just like a normal camera memory card.

Contents of uEnv.txt:

mmcdev=1
bootpart=1:2
mmcroot=/dev/mmcblk1p2
optargs=quiet

The camera seems to be focusing on the building across the street instead of West Seattle.

View from 1200 Western Ave, 13th Floor Elevator Room

1200 Western Ave, 13th Floor Elevator Room

I replaced Google Reader with InoReader

Google reader has been my primary method of keeping up with technological changes over the past few years. I’ve liked the cloud based system that I can read from my laptop, desktop, or tablet, and it will keep track of what I’e read so I don’t have to read the same thing multiple times.

In the early days of the internet, I used usenet news feeds and an nntp reader to keep up with what was going on around the world and with various programming projects. A decade after Usenet died under its own weight Google Reader made RSS a usable replacement for me.

Google reader announced they were shutting down after July 1st, and they followed through with it.

I’ve tried out Feedly, Digg, and looked at several more. I decided to go with InoReader because while it only has a cloud based interface, it is most similar to the quick method I’ve been used to for reading my news.

Feedly has a really nice iPad app that syncs with their system. Digg has an iPad app, but it requires iOS6, which I’ve delayed installing, and I’ve not seen it in use yet.

I’m certain that Google exiting the RSS reader market has increased the innovation in RSS Readers, and may create a revitalization of the technology in general.

New FFMPEG install on BeagleBone Black

There was a 1.0 release of FFMPEG that came out in December of 2012. It moved the bar forward significantly in working with video formats. The Angstrom distribution is still including a 0.8 version which doesn’t even have direct support for h264.

My solution was to use “git” to download the latest code image for both ffmpeg and libx264, and build the libraries. The interesting complications I had were removing the old libraries from the machine, because they caused ffmpeg not to build correctly. Because I do not plan on using thwe machine interactively, I don’t care that I had to remove gimp. I don’t know what the other dependencies I removed may be related to.

I repeatedly issued the command “opkg remove –force-removal-of-dependent-packages libav” until no packages were removed.

I setup a usb drive partition, and a second user. As the second user on that drive partition I issued the following commands.

git clone git://source.ffmpeg.org/ffmpeg.git
git clone git://git.videolan.org/x264.git
cd x264/
./configure --enable-static --enable-shared
date ; make ; date

I switched to the root user, changed to the x264/ directory and

make install
ldconfig

back to the second user

cd ../ffmpeg/
./configure --enable-gpl --enable-libx26
date ; make ; date

and as root

make install
ldconfig

I had a problem because the default installation for libx264 puts the libraries in /usr/local/lib. I had to add that line to the file /etc/ld.so.conf before running ffmpeg.

root@beaglebone:/usr/local/lib# ffmpeg -?
ffmpeg: error while loading shared libraries: libx264.so.133: cannot open shared object file: No such file or directory

I ran the make commands surrounded by the date commands so that I’d be able to get an estimate on time required. With the BeagleBone running at 1GHz, libx264 gets built in slightly less than 10 minutes and FFMPEG takes 1 hour and 40 minutes.

I was left with an ffmpeg that I wanted to use for my video streaming project.

ffmpeg version N-54212-g034b31d Copyright (c) 2000-2013 the FFmpeg developers
  built on Jun 27 2013 04:05:20 with gcc 4.7.3 (Linaro GCC 4.7-2013.02-01) 20130205 (prerelease)
  configuration: --enable-gpl --enable-libx264
  libavutil      52. 37.101 / 52. 37.101
  libavcodec     55. 17.100 / 55. 17.100
  libavformat    55. 10.100 / 55. 10.100
  libavdevice    55.  2.100 / 55.  2.100
  libavfilter     3. 77.101 /  3. 77.101
  libswscale      2.  3.100 /  2.  3.100
  libswresample   0. 17.102 /  0. 17.102
  libpostproc    52.  3.100 / 52.  3.100
Hyper fast Audio and Video encoder
usage: ffmpeg [options] [[infile options] -i infile]... {[outfile options] outfile}...

Use -h to get full help or, even better, run 'man ffmpeg'

BeagleBoneBlack WiFi configuration problems

I got a BeagleBone Black last week. It seems to be a nice system, with a 1GHz ARM7 processor, ethernet, USB, HDMI and plenty of expansion possibilities, all for $45.

It’s shipping configuration runs the Angstrom Linux distribution. I’ve worked with Angstrom in the past on a Beagleboard used for an embedded application. It seems that the newer Angstrom is using ConnMan 1.4 as the network manager. The version that’s available at the ConnMan site, 1.7, has a command line configuration tool, while the version included in the distribution does not.

I’ve not been able to figure out how to enable wireless networking. Plugging in an ethernet cable just works. Plugging in the USB WiFi adapter gives me a wlan0 device, and the MAC matches that printed on the device.

I’m not able to issue the command “ifup wlan0” without an error. It looks like the connman settings file should enable WiFi by default.

root@beaglebone:~# ifup wlan0
ifup: can't open '/etc/network/interfaces': No such file or directory
root@beaglebone:~# ifconfig wlan0 up
ifconfig: SIOCSIFFLAGS: No such file or directory
root@beaglebone:~# lsusb
Bus 001 Device 002: ID 04b4:6560 Cypress Semiconductor Corp. CY7C65640 USB-2.0 "TetraHub"
Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
Bus 002 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
Bus 001 Device 003: ID 13b1:002f Linksys AE1000 v1 802.11n [Ralink RT3572]
root@beaglebone:~# ls -alFR /var/lib/connman/
/var/lib/connman/:
total 16
drwxr-xr-x  3 root root 4096 Jan  1  2000 ./
drwxr-xr-x 17 root root 4096 Jan  1  2000 ../
drwx------  2 root root 4096 Jan  1  2000 ethernet_c8a030a62b80_cable/
-rw-------  1 root root   68 Jan  1  2000 settings

/var/lib/connman/ethernet_c8a030a62b80_cable:
total 16
drwx------ 2 root root 4096 Jan  1  2000 ./
drwxr-xr-x 3 root root 4096 Jan  1  2000 ../
-rw------- 1 root root 4096 Jan  1  2000 data
-rw------- 1 root root  186 Jan  1  2000 settings
root@beaglebone:~# cat /var/lib/connman/settings 
[global]
OfflineMode=false

[Wired]
Enable=true

[WiFi]
Enable=true

I went so far as to connect to an HDMI monitor with keyboard and mouse and was able to see the graphical connection manager. I tried both with and without having the ethernet cable plugged in, but was not able to click on the “Enable” button on the wireless networks dialog.

BeagleBoneWiFi

I believe that I need to do something with wpa_supplicant to get the password properly accepted on my machine, but I’m more interested in getting the wireless up and running than worrying about getting the security set.

I included the lsusb command in my code listing before because it lists the WiFi device that I’m using. I’m powering the device by a standalone power supply, not via USB, so I believe that I should have enough power to run the wifi. This connection has it connected via an external hub, but I’ve had the same results when connected directly to the board.

WimTiVoServer Processes

I have a TiVoHD, and so have written my program specifically to support the features of the TiVoSeries3 and TiVoHD. The newer TiVoPremier units support more codecs and faster transfer speeds, and the earlier Series2 units do not support the 720p or 1080i resolutions that I allow to be transferred to the TiVo. My initial TiVo server serves files from specific shares on the server, \\server\TiVo\* and \\server\Videos\*, which are specified in the registry key and does not have a user interface to modify them. I recursively parse the shares for files FFMPEG reports as video files and present them as a flattened container sorted by date.
[HKEY_LOCAL_MACHINE\SOFTWARE\WimsWorld\WimTiVoServer]
"Container"="//server/TiVo/*;//server/Videos/*"

The TiVo Server operates using the HTTP protocol. It can also operate using SSL/HTTPS. The TiVo itself uses HTTPS for all of the XML transfers, but transfers its videos using HTTP. The videos the TiVo transfers are encrypted using the Media Access Key (MAK) specific to each TiVo. Videos transferred to the TiVo do not need to be encrypted.

There are multiple ways that your server can announce itself to the TiVo. I’m using the simplest to develop, which is broadcasting a specially formatted UDP packet every minute on the local network to port number 2190. An example packet looks like:
tivoconnect=1
method=broadcast
platform=pc/win-nt
machine=FREDS-PC
identity={D936E980-79E3-11D6-A84A-00045A43EEE7}
services=FooService:1234,BarService:4321

Here is a current broadcast from my server indicating that it will serve TiVo compatible media via the HTTP protocol on port 56667. My program follows the tivo recommendation of using a TCP port supplied by the OS, and attempting to use the same port on subsequent restarts of the server. It also generates the GUID and attempts to save and reuse it.
tivoconnect=1
method=broadcast
platform=pc/WinNT:6.2.9200
machine=Acid
identity={FF121976-2D7B-4682-A1DA-464510DCACEB}
services=TiVoMediaServer:56667/http
swversion=20130604103659

After the TiVo has received a broadcast with the service it is interested in, it will make a request of the server to get the top level containers. http://{machine}/TiVoConnect?Command=QueryContainer&Container=/ The server responds with the XML describing the top level containers.

Here’s an example of the code produced by my server, which is named Acid and has a single top level container with the same name. URLs embedded in the XML can be either relative or absolute. I’ve chosen to always use relative URLs because that way I don’t have to embed the TCP port I’m serving data from in the XML.

<?xml version="1.0" encoding="UTF-8"?>
<TiVoContainer xmlns="http://www.tivo.com/developer/calypso-protocol-1.6/">
  <Details>
    <Title>Acid</Title>
    <ContentType>x-tivo-container/tivo-server</ContentType>
    <SourceFormat>x-tivo-container/folder</SourceFormat>
    <TotalItems>1</TotalItems>
  </Details>
  <Item>
    <Details>
      <Title>Acid</Title>
      <ContentType>x-tivo-container/tivo-videos</ContentType>
      <SourceFormat>x-tivo-container/folder</SourceFormat>
    </Details>
    <Links>
      <Content>
        <Url>/TiVoConnect?Command=QueryContainer&amp;Container=%2FTiVoNowPlaying</Url>
        <ContentType>x-tivo-container/tivo-videos</ContentType>
      </Content>
    </Links>
  </Item>
  <ItemStart>0</ItemStart>
  <ItemCount>1</ItemCount>
</TiVoContainer>

When a user goes to the Now Playing screen on the TiVo, the top level containers should show up in the list at the bottom. Selecting the container will cause the TiVo to request the particular container from the server via its content URL. http://{machine}/TiVoConnect?Command=QueryContainer&Container=/Foo

That container may contain TiVoItems which have URLS for both details and content. The TiVo will only request the content URL after it’s successfully requested the URL for the details, which was something that it took a while for me to understand.

Here is an example of the output when the TiVo requests the container with the URL http://acid:56667/TiVoConnect?Command=QueryContainer&Container=/TiVoNowPlaying&ItemCount=1 :

<?xml version="1.0" encoding="UTF-8"?>
<TiVoContainer xmlns="http://www.tivo.com/developer/calypso-protocol-1.6/">
  <ItemStart>0</ItemStart>
  <ItemCount>1</ItemCount>
  <Details>
    <Title>Acid</Title>
    <ContentType>x-tivo-container/folder</ContentType>
    <SourceFormat>x-tivo-container/folder</SourceFormat>
    <TotalItems>2116</TotalItems>
  </Details>
  <Item>
    <Details>
      <Title>ShowTitle</Title>
      <ContentType>video/x-tivo-mpeg</ContentType>
      <SourceFormat>video/h264</SourceFormat>
      <SourceSize>1805804000</SourceSize>
      <Duration>1289860</Duration>
      <CaptureDate>0x51b0ff42</CaptureDate>
    </Details>
    <Links>
      <Content>
        <ContentType>video/x-tivo-mpeg</ContentType>
        <Url>/TiVoConnect/TivoNowPlaying///Acid/TiVo/ShowTitle.mp4</Url>
      </Content>
      <CustomIcon>
        <ContentType>image/*</ContentType>
        <AcceptsParams>No</AcceptsParams>
        <Url>urn:tivo:image:save-until-i-delete-recording</Url>
      </CustomIcon>
      <TiVoVideoDetails>
        <ContentType>text/xml</ContentType>
        <AcceptsParams>No</AcceptsParams>
        <Url>/TiVoConnect?Command=TVBusQuery&amp;Url=/TiVoConnect/TivoNowPlaying///Acid/TiVo/ShowTitle.mp4</Url>
      </TiVoVideoDetails>
    </Links>
  </Item>
</TiVoContainer>

When I select “ShowTitle” in the tivo onscreen interface, it will request this same URL, with the anchoritem set to the content URL. It will then display the details about the show on the TV screen. After I select that I want the show to be transferred it first requests an undocumented URL, my server responds that the URL was not found, then the TiVo requests the TiVoVideoDetails URL, and I respond with an XML chunk that I couldn’t find documentation for anywhere and had to fake together from looking at the output of TiVoDecode and pyTiVo. The really ugly bit of the XML is in the initial element declaration. I wasn’t able to make the XML be accepted by the TiVo if it didn’t have the full namespace declarations, even though the hosts listed are not accessable.

<?xml version="1.0" encoding="UTF-8"?>
<TvBusMarshalledStruct:TvBusEnvelope xmlns:xs="http://www.w3.org/2001/XMLSchema-instance" xmlns:TvBusMarshalledStruct="http://tivo.com/developer/xml/idl/TvBusMarshalledStruct" xmlns:TvPgdRecording="http://tivo.com/developer/xml/idl/TvPgdRecording" xmlns:TvBusDuration="http://tivo.com/developer/xml/idl/TvBusDuration" xmlns:TvPgdShowing="http://tivo.com/developer/xml/idl/TvPgdShowing" xmlns:TvDbShowingBit="http://tivo.com/developer/xml/idl/TvDbShowingBit" xmlns:TvBusDateTime="http://tivo.com/developer/xml/idl/TvBusDateTime" xmlns:TvPgdProgram="http://tivo.com/developer/xml/idl/TvPgdProgram" xmlns:TvDbColorCode="http://tivo.com/developer/xml/idl/TvDbColorCode" xmlns:TvPgdSeries="http://tivo.com/developer/xml/idl/TvPgdSeries" xmlns:TvDbShowType="http://tivo.com/developer/xml/idl/TvDbShowType" xmlns:TvPgdBookmark="http://tivo.com/developer/xml/idl/TvPgdBookmark" xmlns:TvPgdChannel="http://tivo.com/developer/xml/idl/TvPgdChannel" xmlns:TvDbBitstreamFormat="http://tivo.com/developer/xml/idl/TvDbBitstreamFormat" xs:schemaLocation="http://tivo.com/developer/xml/idl/TvBusMarshalledStruct TvBusMarshalledStruct.xsd http://tivo.com/developer/xml/idl/TvPgdRecording TvPgdRecording.xsd http://tivo.com/developer/xml/idl/TvBusDuration TvBusDuration.xsd http://tivo.com/developer/xml/idl/TvPgdShowing TvPgdShowing.xsd http://tivo.com/developer/xml/idl/TvDbShowingBit TvDbShowingBit.xsd http://tivo.com/developer/xml/idl/TvBusDateTime TvBusDateTime.xsd http://tivo.com/developer/xml/idl/TvPgdProgram TvPgdProgram.xsd http://tivo.com/developer/xml/idl/TvDbColorCode TvDbColorCode.xsd http://tivo.com/developer/xml/idl/TvPgdSeries TvPgdSeries.xsd http://tivo.com/developer/xml/idl/TvDbShowType TvDbShowType.xsd http://tivo.com/developer/xml/idl/TvPgdBookmark TvPgdBookmark.xsd http://tivo.com/developer/xml/idl/TvPgdChannel TvPgdChannel.xsd http://tivo.com/developer/xml/idl/TvDbBitstreamFormat TvDbBitstreamFormat.xsd" xs:type="TvPgdRecording:TvPgdRecording">
  <recordedDuration>PT21M29S</recordedDuration>
  <vActualShowing />
  <vBookmark />
  <recordingQuality value="75">HIGH</recordingQuality>
  <showing>
    <showingBits value="0" />
    <time>2013-06-06:21:29:38Z</time>
    <duration>PT21M29S</duration>
    <program>
      <vActor />
      <vAdvisory />
      <vChoreographer />
      <colorCode value="4">COLOR</colorCode>
      <description></description>
      <vDirector />
      <episodeTitle></episodeTitle>
      <vExecProducer />
      <vProgramGenre />
      <vGuestStar />
      <vHost />
      <isEpisode>false</isEpisode>
      <originalAirDate>2013-06-06:21:29:38Z</originalAirDate>
      <vProducer />
      <series>
        <isEpisodic>false</isEpisodic>
        <vSeriesGenre />
        <seriesTitle>ShowTitle</seriesTitle>
      </series>
      <showType value="5">SERIES</showType>
      <title>ShowTitle</title>
      <vWriter />
    </program>
    <channel>
      <displayMajorNumber />
      <displayMinorNumber />
      <callsign />
    </channel>
  </showing>
  <startTime>2013-06-06:21:29:38Z</startTime>
  <stopTime>2013-06-06:21:51:07Z</stopTime>
</TvBusMarshalledStruct:TvBusEnvelope>

After the TvBusMarshalledStruct XML has been transferred to the TiVo the tivo will finally request the content url from the original container listing. When the content is sent, it should be sent using HTTP Chunked Encoding.

My current ffmpeg command line is fairly simple, and mostly copied from my reading of the pyTiVo source code. I make a couple of simple decisions based on the source file codecs. If the video is mpeg2video or the audio is ac3 I use the “-vcodec copy” or “-acodec copy” command,
otherwise I specify those as codecs. I also include the options “-b:v 16384k -maxrate 30000k -bufsize 4096k -ab 448k -ar 48000 -f vob” which I need to spend more time learning about.

One of the gotchas that I spent a significant amount of time debugging was that the SourceSize element in the Item listing is significant. It appears that the tivo pre-allocates space for the file as it starts the transfer and if the transfer is proceeding significantly larger than the reported source size the tivo will close its end of the http connection. I have not verified importance in any internal part of the TvBusMarshalledStruct. I got it working, and have been adding support from metadata as available but that’s it currently.

It’s interesting to me that the transfer speed seems completely governed by the speed of the TiVo processor and what the TiVo is doing. The maximum speed I’ve been able to transfer a file to the TiVo is about 16Mb/s. That speed was achieved while the TiVo was not recording any shows and had a 480p program displayed on the screen in a paused state. If I’m playing a 1080i or 720p program the transfer rate is usually much closer to 7Mb/s. This means that if I’m transferring a program that was recorded in HD on my windows media center from my HDHomeRunPrime, I generally cannot watch in real time because it transfers slightly slower than the data rate of the video and audio.

WimTiVoServer reasons for existence.

I had multiple reasons for writing this program.

I wanted to work with the XMLLite library for XML processing which Microsoft has included as part of their operating system for several years. http://msdn.microsoft.com/en-us/library/windows/desktop/ms752838(v=vs.85).aspx I’ve used XML as my data storage and transfer medium for years but most of the time have manually processed the XML as opposed to using a prebuilt library. My reasoning behind my own parsing and generation was that I was building cross platform solutions that ran on embedded systems working with windows programs. It was easier to build the same code to run on both platforms and assure compatibility than to deal with the overhead and maintenance of third party libraries. I’ve still not formed a full opinion on using the library for reading XML, but it certainly makes creating XML easier by making sure tags are properly closed and nested, and the option of auto indentation in the formatting is nice.

I wanted experience using FFMPEG and its associated video processing libraries. With the release of FFMPEG v1 late last year it became much more capable in dealing with all of the container formats and encoding types that I was interested in, including the WTV containers that Windows Media Center uses and the MKV containers that are common on the internet for high definition files. In my functioning version of my server, I’m using the libraries directly linked into the program to parse the media files for metadata, but spawning a full copy of FFMPEG to do the required transcoding to send the proper format to the TiVo. I’m considering migrating entirely to the spawned FFMPEG process to simplify licensing if I want to make my program publicly available. It would also simplify future support for codecs and containers that FFMPEG may support if my service itself didn’t need to be relinked with updated libraries.

I’ve been frustrated with the state of the TiVo Desktop software provided by the TiVo company. It was designed so that it plugs into the windows video codec stack for video transcoding, as well as the apple protocol stack for Bonjour protocol support. Both of those lead to complications when upgrading programs. Apple regularly releases updates to iTunes. Whenever an update to iTunes needed to be installed, it caused problems because the TiVo service was running and keeping some of the apple services locked in use. It essentially required a full uninstall and reinstall of iTunes every time I needed to update it, with several machine reboots in that process.  Somehow I’d managed to get a codec installed on my machine that allowed the TiVo desktop to support the MKV container. Duplicating that installation on a new machine or server was not something I wanted to attempt. Since the modern FFMPEG supports MKV, I get that support without manipulating a video codec stack in windows.

The TiVo Desktop software only runs as a user process, when a user is logged in. Files are not served from a background service. This is an issue when I’d like to run the process on a headless server. There are ways around the issue, but my program solves it simply by being a pure service. The TiVo Desktop both announces itself via the UDP beacon process on port 2190 and also listens for other servers on the same port. Because my program is purely a server I saw no reason to listen for incoming beacons, and so I do not tie up the UDP port for receiving. This allows secondary programs to be started at the users discretion to listen for TiVo processes on the network.