Difference between revisions of "BeagleBoard/GSoC/2020 Projects/Media IP Streaming"

From eLinux.org
Jump to: navigation, search
(Timeline update)
Line 120: Line 120:
| August 3 || Milestone #9 Joining AVB ALSA drivers with ctag face audio card drivers
| August 3 || Milestone #9 Joining AVB ALSA drivers with ctag face audio card drivers
* Finish implementation of AVDECC in kernel space
* Finish implementation of AVDECC (device enumeration part) in kernel space
* Join Face drivers with AVB drivers and try to get SPI Bitbang to work
* Join Face drivers with AVB drivers and try to get SPI Bitbang to work
* Document AVB driver stack
* Document AVB driver stack

Revision as of 02:25, 13 July 2020

Proposol of equipping the Beaglebone AI with Media IP Streaming capabilities


This project will equip the Beagleboard AI with Media IP Streaming capabilities, by porting the sound card drivers for CTAG face2|4 Audio Card and the AVB protocol stack from BeagleBone AVB to the BeagleBone AI.

Student: nwan
Mentors: rma
Code: https://github.com/NiklasWan/linux
Progress and Documentation/Research Results: https://niklaswan.github.io/GSoC-Overview
Wiki: http://elinux.org/BeagleBoard/GSoC/MediaIpStreaming
GSoC: [N/A]


This project is currently just a proposal.


All requirements have been fullfilled, the Pull Request can be found here #139

About you

IRC: nwan
Github: NiklasWan
School: Kiel University of Applied Sciences
Country: Germany
Primary languages: German, English
Typical work hours: 8AM-5PM CET
Previous GSoC participation: I want to participate at GSoC because I want to gather experience in working within an open source community and try to apply theoretical knowledge into the practical domain. Also I hope to learn new awesome things. This would be my first time participating in GSoC.

About your project

Project name: Media Ip Streaming


The BeagleBone AI is equipped with a high amount of processing power due to the Dual Core ARM Cortex-A15 chip as a main computing unit and its accompanying co-processors. This makes the AI a perfect fit for highly demanding applications regarding CPU consumption, like audio applications which have extremely strong realtime constraints. Professional audio/video studios have to guarantee for small latencies when transmitting media signals between different devices and different media channels in a transmitted stream need to be synchronized. Latency and snychronicity are both extremely important when transmitting e.g. a video channel together with the accompanying audio channel. Those two channels have to be transmitted in a manner, that lip synchronicity can be guaranteed because humans are extremely sensitive to voice offset to accompanying video signals.

To bring media ip streaming capabilities to the BeagleBone AI, the following steps are planned: A previous GSoC project ported a sound card driver from the BeagleBone Green/Black to the BeagleBoard-X15 (https://summerofcode.withgoogle.com/archive/2016/projects/5351212496977920/). This port will now be ported to the BegleBone AI. With the sound card driver successfully ported, the next step would be to port the AVB protocol driver stack from BeagleBone AVB enabling media streaming over the network. This would allow to use the BeagleBone AI as a media streaming device in professional audio/media applications and bring audio stream synchronization features to the BeagleBone AI. Thus allowing for tight synchronization between different audio and video streams which are transmitted over the network. Additionally for people who don't own the CTAG Face 2|4 cape HDMI audio output should be realized.

Practical Use: AVB is primarily used in large scale media productions, like sports venues, broadcasting studios or concert halls. Basically AVB can be used everywhere where media data has to be transmitted over larger distances in a local network. Implementing this on a BeagleBone AI would allow for a low cost alternative for proprietary hardware and further allow for customization by the Beagleboard.org community.


Provide a development timeline with a milestone each of the 11 weeks and any pre-work. (A realistic timeline is critical to our selection process.)

Mar 30 Proposal complete, Submitted to https://summerofcode.withgoogle.com
May 4

Proposal accepted or rejected Community Bonding Period starts.

  • Learn about embedded linux structure ✓
  • Learn about Linux kernel driver development ✓
  • Set up general development environment for embedded Linux systems and required periphery ✓
  • Work through current code base on CTAG drivers ✓
  • Work through current code base on AVB drivers ✓
  • Learn about ALSA SoC driver development ✓
  • Learn about Beaglebone AI hardware structure ✓
June 1 Pre-work complete, Coding officially begins!
June 8 Milestone #1, Introductory YouTube video, review of existing drivers for ctag face audio interface, identifying challenges for porting drivers to Beagle AI and selection of appropriate kernel
June 15 Milestone #2 Implementation / porting of ALSA audio drivers for ctag face to Beagle AI --> toolchain setup, driver adoptions, coding
  • Setup Beaglebone AI specific Toolchain ✓
  • Evaluate if existing kernel driver needs changes for AI ✓
  • Implement a base DTS for CTAG Face BBAI in arch/arm/boot/dts ✓
  • Change existing DTS Pin Configurations to match BBAI pin muxing ✓
June 22 Milestone #3 Port of sound card drivers, testing, performance check
  • Testing of all implemented functionality ✓
  • Bug fixing of implemented functionality ✓
  • Implement Overlay for BBB ✓
  • Documentation of CTAG Face installation process ✓
  • First evaluation of sound card performance using https://github.com/henrix/ALSA-sound-card-evaluation
July 3 18:00 UTC Milestone #4 (Phase 1 evaluations), finalzing port of ctag face audio card driver to Beagle AI and getting pull request to Beagleboard Mainline
  • Refactoring of device tree configuration, to just include base configuration in DTS ✓
  • Fork https://github.com/beagleboard/bb.org-overlays
  • Implement Face Specific configuration as overlay for BBAI ✓
  • Second evaluation of sound card performance ✓
  • Submit pull request to mainline kernel ✓
July 10 Milestone #5 Review of existing AVB network driver architecture for real-time audio streaming, basis is https://elinux.org/BeagleBoard/GSoC/2017_Projects#Project:_BeagleBone_AVB_Stack , identifying challenges for porting to Beagle AI
  • Review of existing AVB driver ✓
  • Review of gPTP daemon ✓
  • Review new version of gPTP from 2020 ✓
  • Start porting gPTP to kernel 5.4-rt ✓
July 17 Milestone #6 Implementation / porting of ALSA AVB network drivers to Beagle AI --> toolchain setup, driver adoptions
  • Finish implementation of gPTP in kernel space
  • Review MSRP protocol
  • Implement ALSA dummy driver
July 24 Milestone #7 AVB ALSA drivers implementation for Beagle AI
  • Finish implementation of MSRP in kernel space
  • Review AVTP protocol
  • Start implementing AVTP
July 31 18:00 UTC Milestone #8 (Phase 2 evaluations), Getting ALSA AVB network drivers finished and document everything till now
  • Finish implementation of AVTP in kernel space
  • Review AVDECC protocol
  • Start implementing AVDECC (device enumeration part)
  • Document AVB driver stack
August 3 Milestone #9 Joining AVB ALSA drivers with ctag face audio card drivers
  • Finish implementation of AVDECC (device enumeration part) in kernel space
  • Join Face drivers with AVB drivers and try to get SPI Bitbang to work
  • Document AVB driver stack
August 10 Milestone #10 Performance and integration testing of driver ports
  • Test AVB drivers and Face drivers in conjunction
  • Make completion video
  • Finish documentation
August 17 Milestone #11, Completion YouTube video, pull request of driver architecture for mainline
  • Finish youtube video
  • Last Bug fixes
  • Submit pul request of AVB driver stack
August 24 - 31 18:00 UTC Final week: Students submit their final work product and their final mentor evaluation
August 31 - September 7 18:00 UTC Mentors submit final student evaluations

Experience and approach

During my bachelor's degree in information technology I had several courses like programming in C, programming in C++, operating systems and embedded system programming which layed down the basis for developing embedded software. Due to my additional bachelor's degree in audio production I have additional experience in audio applications and audio and media codecs, which will help me to understand the theory behind the different needed algorithms. With my previous development work for the Strämpler project I already have experience in working on complex embedded C projects and which potential pitfalls could occur.


If I get stuck and my mentor is not around I will follow the following steps in displayed order:

  1. Search the internet for the problem.
  2. Serach through literature acquired during milestone #1.
  3. Ask in the GSoC IRC, if fellow students know a solution to the specific problem.
  4. If the problem is still not solved, postpone the problem until mentor is available again and work on another part of the project.


Equipping the BeagleBone AI with media ip streaming capabilities would allow the Beagleboard.org community to use those capabilities to implement the system in professional media applications. The community could also implement further media protocols like AES/Ravenna to allow the usage of the AI for even more media streaming tasks.


Link to pull request #139.


  1. [1] „4.3. PTP — Processor SDK Linux Documentation“. https://software-dl.ti.com/processor-sdk-linux/esd/docs/06_02_00_81/linux/Industrial_Protocols_PTP.html (accessed March 30, 2020).
  2. [2] 1733-2011 IEEE Standard for Layer 3 Transport Protocol for Time-Sensitive Applications in Local Area Networks. ///.
  3. [3] M. A. Yoder und J. Kridner, BeagleBone cookbook, First edition. Sebastopol, CA: O’Reilly Media, Inc, 2015.
  4. [4] C. Hallinan, Embedded Linux primer: a practical real-world approach, 2nd ed. Upper Saddle River, NJ: Prentice Hall, 2011.
  5. [5] A. Liberal de los Ríos, Linux driver development for embedded processors: Learn to develop Linux embedded drivers with kernel 4.9 LTS, Second edition. .
  6. [6] R. Love, Linux kernel development, 3rd ed. Upper Saddle River, NJ: Addison-Wesley, 2010.
  7. [7] E. White, Making embedded systems: design patterns for great software, 1. ed. Beijing: O’Reilly, 2012.
  8. [8] D. Molloy, Molloy_exploring BeagleBone 2e. Indianapolis, NY: John Wiley and Sons, 2018.
  9. [9] „The Linux Kernel documentation — The Linux Kernel documentation“. https://www.kernel.org/doc/html/latest/index.html (accessed March 26, 2020).

PTP Overview

  • is used by the AVB protocol to achieve synchronization between devices
  • based on IEEE 1588v2
  • up to nanoseconds accuracy
  • sharing timestamps over the network for synchronization of devices
  • uses master/slave hierarchy
  • slave retrieves time from master ==> network dely has to be taken into account

PTP Clocks:

Ordinary Clock

  • normally endpoint of the network
  • single port
  • BMCA (best master clock algorithm) determines which clock is used as master (the one with the highest accuracy)

Grandmaster Clock

  • is used as an endpoint master and has extremely high accuracy (normally timed by GPS or NTP)
  • there can be more than one in a network to achieve redundancy

Boundary Clock

  • mutli port
  • a network switch with master/slave ports

Transparent Clock

  • accounts for queuing delays when a standrad switch is used and thus improves accuracy

Delay mechanisms:

  • calculate network delay End-To-End
  • no need of PTP equipment but this results in added cost in accuracy


  • calculate network delay Peer-To-Peer
  • results in high accuracy, but all devices in the network need to be PTP enabled