forked from darealshinji/dcaenc
-
Notifications
You must be signed in to change notification settings - Fork 0
dcaenc is an open-source implementation of the DTS Coherent Acoustics lossy audio codec (by Alexander E. Patrakov; personal mirror)
License
madmax2/dcaenc
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
Introduction ============ dcaenc is an open-source implementation of the DTS Coherent Acoustics lossy audio codec. Only the core part of the encoder is implemented, even though the full specification (ETSI TS 102 114 V1.3.1) is available for download. The project page is at http://aepatrakov.narod.ru/dcaenc/ The latest code can be obtained from git: git clone git://gitorious.org/dtsenc/dtsenc.git Windows-specific notes ====================== There is a fork of dcaenc specially targetted at Windows users. You can get it from http://gitorious.org/~mulder/dtsenc/mulders-dtsenc . As compared to this version, that fork adds a Visual Studio project file, a progress indicator and support for Unicode file names on Windows. It also breaks Linux support. Still, the original dcaenc can be compiled for Windows, here is how. The recommended compiler for dcaenc on Windows is MinGW. Get it from http://sourceforge.net/projects/mingw/files/Installer/ . You may want to use either the mingw-get.exe utility directly (just copy it to c:\mingw\bin, install nothing else), or use a graphical installer. In the latter case, install just the C compiler. After installation, add c:\mingw\bin and c:\mingw\msys\1.0\bin to the Path environment variable (right-click My Computer -> Properties -> Advanced -> Environment Variables) if it is not already there. Then install the required MinGW components from the command prompt: mingw-get install mingw32-base mingw32-autotools For autotools to work, you also have to copy or rename the c:\mingw\msys\1.0\etc\fstab.sample file to c:\mingw\msys\1.0\etc\fstab. Installation ============ The dcaenc library has no external dependencies. The optional ALSA plugin that comes with it, obviously, depends on the ALSA library compiled with support for external plugins. See the file INSTALL for the generic installation instructions. Quick summary for UNIX-like systems: autoreconf -f -i -v # only if building from git ./configure --prefix=/usr --libdir=??? # see below make su -c 'make install' # Or, on Ubuntu: sudo make install The --libdir parameter is distribution- and architecture-specific, thus, the default (/usr/lib) may be wrong for your system. This parameter influences the shared library and ALSA plugin location. If you specify this parameter incorrectly, ALSA will not find the plugin. You can verify that the value is correct by looking if there are already any ALSA plugins for the same architecture in $libdir/alsa-lib. Here are some values to consider: Most 32-bit x86 linux systems: /usr/lib Most x86-64 linux systems: /usr/lib or /usr/lib64 32-bit libraries on most x86-64 linux systems: /usr/lib or /usr/lib32 x86 Debian or Ubuntu (new versions): /usr/lib/i386-linux-gnu x86-64 Debian or Ubuntu (new versions): /usr/lib/x86_64-linux-gnu Quick summary for Windows: bash autoreconf -f -i -v # only if building from git ./configure MAKE=mingw32-make mingw32-make Package contents ================ On UNIX-like systems, the dcaenc package contains the libdcaenc.so shared library, the corresponding dcaenc.h header, the dcaenc command line tool, and (optionally, on Linux only) the ALSA plugin. On Windows, you only get the command-line tool, even if you use the forked version. The libdcaenc library ===================== The library may be useful if you want to create DTS streams in your own UNIX application. To do so, you have to call the following functions: dcaenc_context dcaenc_create( int sample_rate, int channel_config, int approx_bitrate, int flags); This function creates a library context according to its parameters. The resulting context is an opaque value that should be passed to the other library functions as the first parameter. The sample rate must be one of the following values: 32000, 44100, 48000 or those divided by 2 or 4. The channel configuration is one of the DCAENC_CHANNELS_* defines listed in the dcaenc.h header. Note that values greater than DCAENC_CHANNELS_3FRONT_2REAR always return an error now because their encoding requires the XCH/XXCH extensions that became documented only in September 2011 and thus were not implemented. Only non-LFE channels have to be specified here. The approximate bitrate is specified in bits per second and may be rounded slightly up or down by the library. The flags parameter should be a logical OR of zero or more of the DCAENC_FLAG_* defines. Their meanings: DCAENC_FLAG_28BIT: use only 28 out of 32 bits in each four bytes of output. This may be useful if the loudness of the hiss resulting from mis-interpretation of the encoded stream as PCM must be reduced. However, this also results in the reduction of the effective bitrate by 12.5%. DTS CDs are usually encoded with this option. DCAENC_FLAG_BIGENDIAN: produce a big-endian variant of the DTS bitstream. This may be useful for writing the stream as a DVD sound track. DCAENC_FLAG_LFE: indicates that a separate LFE channel has to be added to the layout indicated by the channel_config parameter. DCAENC_FLAG_PERFECT_QMF: selects a "perfect-reconstruction" version of the quadrature mirror filter. This reduces distortions inherent in the filterbank design, but makes the filter output more sensitive to quantization errors introduced later in the encoder. libdca version 0.0.5 or earlier will not be able to decode the resulting stream correctly due to a bug (mistyped table) in it. It makes no sense to use this option for streams with bitrate of 0.3 Mbps per channel or less. DCAENC_FLAG_IEC_WRAP: wraps DTS frames as defined by the IEC 61937-5 standard for transmission over SPDIF. Namely, the library adds a 8-byte header to each frame and pads the frame with zeroes to achieve the same bitrate as a stereo 16-bit PCM stream. On any error, dcaenc_create() returns NULL. There is no way to find out the reason for the error. int dcaenc_channel_config_to_count(int channel_config); Returns the number of channels used in a given channel configuration, or -1 if the channel configuration is invalid or unsupported. int dcaenc_bitrate(dcaenc_context c); Returns the actual bitrate that is used by the library. int dcaenc_input_size(dcaenc_context c); Returns the size of the input buffer that your application has to submit, in samples. Now this value is always equal to 512. int dcaenc_output_size(dcaenc_context c); Returns the size of the output buffer that the application should provide, in bytes. int dcaenc_convert_s32( dcaenc_context c, const int32_t *input, uint8_t *output); Performs the conversion of PCM samples stored in the input buffer to the DTS bitstream, stores one frame of the encoded bitstream in the output buffer. The input buffer should contain interleaved signed 32-bit samples. The channel order is as follows: DCAENC_CHANNELS_MONO: center DCAENC_CHANNELS_STEREO: left, right DCAENC_CHANNELS_3FRONT: center, left, right DCAENC_CHANNELS_2FRONT_1REAR: left, right, surround DCAENC_CHANNELS_3FRONT_1REAR: center, left, right, surround DCAENC_CHANNELS_2FRONT_2REAR: left, right, surround left, surround right DCAENC_CHANNELS_3FRONT_2REAR: center, left, right, surround left, surround right If the LFE channel is used, it should be added as the last one. The following layouts are also defined and do not return an error while encoding, but do not result in a high-quality bitstream due to incompatibility with the psychoacoustical model used by the library: DCAENC_CHANNELS_DUAL_MONO: A, B DCAENC_CHANNELS_STEREO_SUMDIFF: left+right, left-right DCAENC_CHANNELS_STEREO_TOTAL: left total, right total Unfortunately, there is currently no API to query the order of channels by the channel layout. This is because it is unclear what functionality related to channel mapping should be exposed. E.g., should speaker IDs compatible with ksmedia.h (as used in Windows and in WAVEFORMATEXTENSIBLE wav files) be used, even though a mapping between these speaker IDs and the positions used in the DTS specification is in some cases only approximate (e.g., there is no "Surround Left 2" speaker in ksmedia.h). Please mail your suggestions regarding this API to patrakov@gmail.com. dcaenc_convert_s32() returns the number of bytes written to the output buffer. Right now, it is always the same as returned by dcaenc_output_size(), but this will change if variable bitrate encoding is added to the library. int dcaenc_destroy(dcaenc_context c, uint8_t *output); Destroys the library context. If a non-NULL value is provided in the output parameter, the library encodes the final frame and puts it there. This may be useful because there is a 512-sample latency inherent in the DTS filterbank, so the output frame gets the last portion of the PCM input submitted earlier. The returned value indicates the number of bytes written to the output buffer. The command-line application ============================ The dcaenc command line utility converts multichannel wav files to DTS. The resulting DTS files can be written to an audio CD and played via a digital connection (SPDIF or HDMI) to a receiver, or, after changing the endianness, used as sound tracks for a DVD. Currently there are no options to select 28-bit encoding or change the endianness. This is a bug. Usage: dcaenc input.wav output.dts bitrate The input wav file should have the same channel order as defined by SMPTE, i.e.: left, right, center, lfe, surround left, surround right. Some destinations require a specific bitrate to be specified. To create a CD-compatible DTS file from a multichannel file (that needs to have the sample rate of 44100 Hz and either 16 or 32 bits per sample), run: dcaenc input.wav output.dts 1411200 To create a DVD-compatible track from a multichannel wav file that has the 48 kHz sample rate: dcaenc input.wav output.dts 1509000 or for a half-rate output: dcaenc input.wav output.dts 754500 and then byte-swap the resulting output.dts file. Mux it with your MPEG2 video track using the "mplex" tool from the mjpeg-tools package. Known bug: wav files with floating-point samples are misinterpreted as containing 32-bit integer samples. ALSA Plugin =========== The ALSA plugin may be useful for playing multichannel sound from arbitrary ALSA applications through an SPDIF link. This is needed because the SPDIF link cannot carry enough bits per second to transport the raw uncompressed 5.1 audio. The "alsa-plugins" package contains a similar plugin for on-the-fly AC3 encoding. The plugin should normally not be used with HDMI connections, because the HDMI standard defines enough bandwidth so that the uncompressed 5.1 PCM stream fits even at 192 kHz sample rate. So, attempting to encode that into DTS in the majority of cases is only a waste of CPU time and sound quality. There are, however, at least two valid use cases. 1. Radeon HD 6xxx video cards with open-source driver do not support more than two channels of PCM audio over HDMI. Encoding the output to DTS provides a useful workaround to the lack of multichannel capability in the driver. 2. Some soundbars are supposed to get the audio signal from the TV via the optical cable, while the TV itself is connected to the computer using HDMI. If you know another valid use case, please send an e-mail to patrakov@gmail.com. The ALSA plugin should work in real time on any modern CPU. Here on Intel Core i5 @ 1.20 GHz (i.e. in powersaving mode) it eats ~40% of a single core. To use the ALSA plugin, add the following line to your $HOME/.asoundrc file or to /etc/asound.conf: <confdir:pcm/dca.conf> It will create an additional ALSA device for each of your sound cards that have an SPDIF output. The name of the device will be similar to "dca:CARD=Intel,DEV=0", or, for the default card, simply "dca". There are also devices that encode to HDMI outputs, they have names like "dcahdmi:CARD=Intel,DEV=0". If you want to encode DTS and send it to something that is not SPDIF or HDMI, add a snippet similar to the following: pcm.dcacustom { type dca slave.pcm "custompcm" # if your receiver requires it: # iec61937 1 } Unlike the AC3 encoder, there is no bitrate configuration. This is because it does not make sense to have it. The hard-coded default (same bitrate as stereo PCM, all bits used) provides the best possible quality and should work for everyone. But it doesn't work with receivers that require IEC61937-5 wrapping of the DTS frames, that's why the "iec61937" option exists, which can be set to 1 for such receivers. To direct mplayer output to the default card via the encoder: mplayer -channels 6 -ao alsa:device=dca file.flac It is not possible to use dmix on top of the encoder. This is a limitation of dmix: it only works on direct hardware devices providing mmap, and the dca plugin is not a hardware device and provides (non-working) mmap only due to what seems to be at least partially an ALSA bug. Please use PulseAudio instead. Known bug: the ALSA plugin doesn't report the supported sample rates correctly and pretends to support mmap. Fixing this requires rewriting the plugin from the extplug infrastructure to ioplug, or talking to ALSA developers. So you may need to add one of the following flags to mplayer command line: -af resample=44100 -af resample=48000 Use with PulseAudio =================== The ALSA plugin can be used with PulseAudio. To do so, add the following lines to the end of the /usr/share/pulseaudio/alsa-mixer/profile-sets/default.conf file: [Mapping iec958-dts-surround-51] device-strings = dca:%f channel-map = front-left,front-right,rear-left,rear-right,front-center,lfe paths-output = iec958-stereo-output priority = 3 direction = output [Mapping hdmi-dts-surround-51] device-strings = dcahdmi:%f channel-map = front-left,front-right,rear-left,rear-right,front-center,lfe paths-output = hdmi-output-0 priority = 1 direction = output Newer and/or patched versions of PulseAudio use the /usr/share/pulseaudio/alsa-mixer/profile-sets/extra-hdmi.conf file instead for Intel and NVidia sound cards. If that file exists, add the above lines there, too, adjusting if needed for multiple HDMI outputs. Note that the iec958-dts-surround-51 part of the above is already provided by PulseAudio 3.0 or later. After restarting PulseAudio, it will see additional output profiles having "DTS" in their names and will allow you to select them in the volume control application such as pavucontrol or gnome-volume-control. Compatibility ============= The author has an LG 47LM640T TV that can decode DTS on HDMI inputs. It works. He also tests the encoder by decoding its output with ffmpeg, libdca or ArcSoft DTS decoder (the same engine as used in WinDVD). The ALSA plugin has been tested and found to work with the following receivers by other people: Logitech Z5500 JVC TH-A25 Samsung HT-Z310 Sony STR-DB780 Some receivers (including JVC TH-A25 and Sony STR-DB780) mute their outputs when receiving full 32-bit DTS stream (as generated by the ALSA plugin) over SPDIF with AES0=6 (the default for the "dca" and "dcahdmi" families of ALSA devices). To overcome this problem, add the following line to the end of .asoundrc: defaults.pcm.dca.aes0 0x04 However, it can cause your receiver to unmute its output even if it does not support DTS streams or does not detect them reliably. This will result in very loud hiss that can damage the loudspeakers. So try this setting with the lowest possible volume, and this is why it is not the default. Similar settings exist for AES1, AES2 and AES3 SPDIF parameters. Some other receivers (including the LG 47LM640T TV) need not raw DTS frames, but DTS frames wrapped according to IEC61937-5, and either mute the output or produce loud hiss otherwise. For such receivers, add the following line to the end of .asoundrc: defaults.pcm.dca.iec61937 1 This could not be made the default, because other receivers (such as Sony STR-DB780) reject DTS frames wrapped into IEC61937-5. Quality ======= There are debates on the Internet about the relative quality of AC3 vs DTS. AC3 uses more advanced compression algorithms, DTS allows for higher bit rates. There were no blind tests comparing the output of the encoder with anything else. However, this encoder uses only the most basic compression techniques defined in the DTS specification, and thus cannot win any comparison with commercial DTS encoders. Still, at 754 kbps, the internal psychoacoustical model considers the distortions to be just below the threshold of detection by human ears. How to report bugs ================== Bugs should be reported by email to patrakov@gmail.com, preferrably with a short (< 10 seconds) flac sample that demonstrates the problem, and a patch that fixes it. Thanks ====== The following people helped me to test the encoder (including the versions that did not work): Arun Raghavan Colin Guthrie Mikhail Elovskikh cryptonymous from the linux.org.ru forum rulet from the linux.org.ru forum Steven Newbury The following people contributed useful information: Adam Thomas-Murphy The following people reported bugs: LoRd_MuldeR from https://gitorious.org/~mulder
About
dcaenc is an open-source implementation of the DTS Coherent Acoustics lossy audio codec (by Alexander E. Patrakov; personal mirror)
Resources
License
Stars
Watchers
Forks
Packages 0
No packages published
Languages
- C 96.4%
- Shell 1.2%
- M4 1.2%
- Makefile 1.2%