path: root/thesis/thesis.tex

\documentclass[12pt,a4paper,oneside,titlepage]{paper}
\usepackage[english]{babel}
\usepackage[utf8]{inputenc}
\usepackage[T1]{fontenc}
\usepackage{textcomp}
\usepackage{listings}
\lstdefinelanguage{Ini}{basicstyle=\ttfamily\tiny,
  columns=fullflexible,
  tag=[s]{[]},
  tagstyle=\color{blue}\bfseries,
  usekeywordsintag=true
}[html]
\lstdefinelanguage{bash}{basicstyle=\ttfamily\tiny}
\usepackage{ulem}
\usepackage{lmodern}
\usepackage{multirow}
\usepackage{url}
\usepackage{graphicx}
\usepackage{pdfpages}
\usepackage{float}
\floatstyle{boxed}
\restylefloat{figure}
\usepackage[usenames,dvipsnames,svgnames,table]{xcolor}
\definecolor{osc-out}{RGB}{150,0,255}
\definecolor{osc-in}{RGB}{0,0,255}
\definecolor{audio-in}{RGB}{255,0,0}
\definecolor{audio-out}{RGB}{0,206,0}

%\usepackage{color}
\usepackage{hyperref}
\hypersetup{hidelinks, colorlinks = false}
\usepackage[font=scriptsize]{caption}
\usepackage[authoryear]{natbib}

% glossary
\usepackage[acronym,nonumberlist,toc]{glossaries}
\newacronym{bs}{BS}{Binaural Synthesis}
\newacronym{hoa}{HOA}{Higher Order Ambisonics}
\newacronym{ip}{IP}{Internet Protocol}
\newacronym{jack}{JACK}{JACK Audio Connection Kit}
\newacronym{oop}{OOP}{Object-oriented Programming}
\newacronym{osc}{OSC}{Open Sound Control}
\newacronym{pubsub}{PubSub}{Publish-subscribe message pattern}
\newacronym{pd}{Pd}{PureData}
\newacronym{ssr}{SSR}{SoundScape Renderer}
\newacronym{tcp}{TCP}{Transmission Control Protocol}
\newacronym{vbap}{VBAP}{Vector Based Amplitude Panning}
\newacronym{wfs}{WFS}{Wave Field Synthesis}
\newacronym{xml}{XML}{Extensible Markup Language}
\makeindex
\makeglossaries



\graphicspath{{../images//}}

\begin{document}
  \begin{titlepage}
    \centering
    \includegraphics[width=0.3\textwidth]{tu-berlin-logo.pdf}\par\vspace{1cm}
    {\scshape\LARGE Technische Universität Berlin\par}
    \vspace{1cm}
    {\scshape\Large Master Thesis\par}
    \vspace{1.5cm}
    {\huge\bfseries A Networking Extension for the SoundScape Renderer\par}
    \vspace{2cm}
    {\Large\itshape David Runge\par}
    \href{dave@sleepmap.de}{dave@sleepmap.de}
    \vfill
    supervised by\par
    Henrik von Coler and Stefan Weinzierl
    \vfill
    {\large \today\par}
  \end{titlepage}

  \pagestyle{empty}
  \section*{Eidesstattliche Erklärung}
  \vspace{1cm}
  Hiermit erkläre ich, dass ich die vorliegende Arbeit selbstständig und
  eigenhändig sowie ohne unerlaubte fremde Hilfe und ausschließlich unter
  Verwendung der aufgeführten Quellen und Hilfsmittel angefertigt habe.\\
  Berlin, den \today\par\\
  \vspace{2cm}
  \noindent\ldots\ldots\ldots\ldots\ldots\ldots\ldots\ldots\ldots\ldots\ldots\\
  David Runge

  \begin{abstract}
    \gls{wfs} as a technological concept has been around for
    many years now and all over the world several institutions run small and
    some even large scale setups ranging from single speaker lines to those
    facilitating a couple of hundred loudspeakers respectively.\\ 
    The still evolving implementations are driven by several rendering
    engines, of which two free and open-source ones, namely sWONDER and
    SoundScape Renderer, have (partially) been developed at TU Berlin.\\
    The latter due to its current design is not yet able to render for large
    scale setups, ie.\ those using several computers to render audio on a
    loudspeaker setup, due to the high amount of channels.\\
    Its solid codebase however, which additionally offers a framework for many
    more renderering types, and the ongoing development, deems further work on
    this application a good future investment.\\ 
    This work is about the extension of the SoundScape Renderer's functionality
    to turn it into a networking application for large scale \gls{wfs} setups.
  \end{abstract}

  \tableofcontents
  \cleardoublepage
  \pagestyle{headings}
  \setcounter{page}{1}

  \section{Introduction}
    \label{sec:introduction}


  \cleardoublepage
  \section{Free and open-source spatial audio renderers}
    \label{sec:freespatialaudiorenderers}
    To date there exist three (known of) free and open-source spatial audio
    renderers, which are all \href{http://jackaudio.org/}{\gls{jack}}
    \citep{website:jackaudio2016} clients:
    \begin{itemize}
      \item \href{https://sourceforge.net/projects/swonder/}{sWONDER}
        \citep{website:swonder2016}, developed by Technische Universität
        Berlin, Germany 
      \item \href{https://github.com/GameOfLife/WFSCollider}{WFSCollider}
        \citep{website:wfscollider2016}, developed by
        \href{http://gameoflife.nl/en}{Game Of Life Foundation}
        \citep{website:gameoflife2016}, The Hague, Netherlands
      \item \href{http://spatialaudio.net/ssr/}{\gls{ssr}}
        \citep{website:ssr2016}, developed by Quality \& Usability Lab,
        Deutsche Telekom Laboratories and TU Berlin and Institut für
        Nachrichtentechnik, Universität Rostock 
    \end{itemize}
    Currently only WFSCollider and the \gls{ssr} are actively maintained and
    developed, thus sWONDER, although used in some setups, loses significance.
    Generally it can be said, that different concepts apply to the three
    renderers, which are about to be explained briefly in the following
    sections.

    \subsection{Spatial audio renderers and their appliance}
      \label{subsec:spatialaudiorenderersandtheirappliance}
      \subsubsection{Wave Field Synthesis}
        \label{subsubsec:wavefieldsynthesis}
      \gls{wfs} describes a spatial technique for rendering
      audio. As such it aims at synthesizing a sound field of desired acoustic
      preference in a given listening area, assuming a planar reproduction to be
      most suitable for most applications.\\
      \gls{wfs} is typically implemented using a curved or linear loudspeaker array
      surrounding the listening area.\\
      Several free and open-source renderer applications exist for \gls{wfs}
      environments, with varying stages of feature richness.\\
      The proposed work will focus on one of them and its extension towards \gls{wfs}
      on large scale systems.
      \subsubsection{\gls{hoa} and \gls{vbap}}
        \label{subsubsec:hoaandvbap}
      \subsubsection{\gls{bs}}
        \label{subsubsec:binaural}

    \subsection{WONDER}
      \label{subsec:WONDER}
      sWONDER \citep{baalman2007} consists of a set of C++ applications that
      provide \gls{bs} and \gls{wfs} rendering. In 2007 it was specifically
      redesigned \citep{baalmanetal2007} to cope with large scale \gls{wfs} setups in
      which several (computer) nodes, providing several speakers each, drive a
      system together.\\
      In these setups each node receives all available audio streams (which
      represent one virtual audio source respectively) redundantly and a master
      application signals which node is responsible for rendering what source
      on which speaker.\\
      It uses \gls{osc} for messaging between its parts and for
      setting its controls. Apart from that, it can be controlled through a
      Graphical User Interface (GUI), that was specifically designed for it.
      Unfortunately sWONDER has not been actively maintained for several years,
      has a complex setup chain and many bugs, that are not likely to get fixed
      any time soon.

    \subsection{HOA-Pd}
      \label{subsec:hoapd}
    \subsection{WFSCollider}
      \label{subsec:wfscollider}
      WFSCollider was built on top of
      \href{https://supercollider.github.io}{SuperCollider} 3.5
      \citep{website:supercollider2016} and is also capable of driving large
      scale systems. It uses a different approach in
      doing so, though: Whereas with sWONDER all audio streams are distributed
      to each node, WFSCollider usually uses the audio files to be played on
      all machines simultaneously and synchronizes between them.\\
      It has a feature-rich GUI in the ``many window'' style, making available
      time lines and movement of sources through facilitating what the sclang
      (SuperCollider programming language) has to offer.\\
      As WFSCollider basically is SuperCollider plus extra features, it is also
      an \gls{osc} enabled application and can thus also be used for mere
      multi-channel playback of audio.\\
      Although it has many useful features, it requires MacOSX (Linux version
      still untested) to run, is built upon a quite old version of
      \href{https://supercollider.github.io}{SuperCollider} and is likely never
      to be merged into it, due to many core changes to it.

    \subsection{SoundScape Renderer}
      \label{subsec:soundscaperenderer}
      \gls{ssr}, also a C++ application, running on Linux and
      MacOSX, is a multi-purpose spatial audio renderer, as it is not only
      capable of \gls{bs} and \gls{wfs}, but also \gls{hoa}
      and \gls{vbap}.\\
      It can be used with a GUI or headless (without one), depicting the
      virtual sources, their volumes and positions, alongside which speakers
      are currently used for rendering a selected source.
      \gls{ssr} uses TCP/IP sockets for communication and is therefore not directly
      \gls{osc} enabled. This functionality can be achieved using the capapilities of
      other applications such as \gls{pd} \citep{website:puredata2016} in
      combination with it though.\\
      Unlike the two renderers above, the \gls{ssr} is not able to run large-scale
      \gls{wfs} setups, as it lacks the features to communicate between instances of
      itself on several computers, while these instances serve a subset of the
      available loudspeakers.

  \cleardoublepage
  \section{Methods}
    \label{sec:methods}
    The \gls{ssr}, due to its diverse set of rendering engines, which are made
    available through an extensible framework, and its relatively clean
    codebase, is a good candidate for future large scale \gls{wfs} setups. These type
    of features are not yet implemented though and will need testing.\\
    Therefore I propose the implementation and testing of said feature, making
    the \gls{ssr} capable of rendering on large scale \gls{wfs} setups with many nodes,
    controlled by a master instance.\\
    The sought implementation is inspired by the architecture of sWONDER, but
    instead of creating many single purpose applications, the master/node
    feature will be made available through flags to the \gls{ssr} executable, when
    starting it. This behavior is already actively harnessed eg.\ for selecting
    one of the several rendering engines.
    \begin{figure}[!htb]
      \centering
      \includegraphics[scale=0.9, trim = 31mm 190mm 24mm 8mm, clip]
      {ssr-networking.pdf}
      \caption{A diagram displaying the \gls{ssr} master/node setup with TCP/IP
        socket connections over network (green lines), audio channels (red
        dashed lines) and \gls{osc} connection (blue dashed line). Machines are
          indicated as red dashed rectangles and connections to audio hardware
        as outputs of \gls{ssr} nodes as black lines below them.}
      \label{fig:ssr-networking}
    \end{figure}
    While the \gls{ssr} already has an internal logic to know which loudspeaker will
    be used for what virtual audio source, this will have to be extended to be
    able to know which renderer node has to render what source on which
    loudspeaker (see Figure~\ref{fig:ssr-networking}).
    To achieve the above features, the \gls{ssr}'s messaging (and thus also settings)
    capabilities have to be extended alongside its internal logic concerning
    the selection of output channels (and the master to node notification
    thereof). To introduce as little redundant code as possible, most likely a
    ``the client knows all'' setup is desirable, in which each node knows about
    the whole setup, but is also set to only serve its own subset of
    loudspeakers in it. This will make sure that the rendering engine remains
    functional also in a small scale \gls{wfs} setup.\\
    The lack of a direct \gls{osc} functionality, as provided by the two other
    renderers, will not be problematic, as master and nodes can communicate
    through their builtin TCP/IP sockets directly and the master can, if
    needed, be controlled via \gls{osc}.
    \subsection{Prelimenaries}
      \label{subsec:preliminaries}
      In preparation to the work an implement a side-by-side
      installation, using Arch Linux on a medium scale setup, facilitating the
      \gls{wfs} system of the Electronic Studio at TU Berlin. Unfortunately the
      proprietary Dante driver, that is used in that system is very complex to be
      built, as well as underdeveloped and thus keeps the system from being
      easily updated, which is needed for testing purposes (finding a suitable
      real-time, low-latency Linux kernel), trying out new software features,
      building new software and keeping a system safe. The driver will most
      likely require changes to the hardware due to implemention of hardware
      branding by the vendor and dire testing before usage.\\
      Although eventually using a proper \gls{wfs} setup for testing will be necessary,
      it is luckily not needed for implementing the features, as they can already
      be worked out using two machines running Linux, \gls{jack} and the development
      version of \gls{ssr}.\\
      The hardware of the large scale setup at TU Berlin in H0104 is currently
      about to be updated and therefore a valuable candidate for testing of the
      sought after \gls{ssr} features.
    \subsection{Outline}
      \label{subsec:outline}
      Initially extending the \gls{ssr}'s features was aimed at 
      \subsubsection{Remote controlling a server}
        \label{subsubsec:remote_controlling_a_server}
      \subsubsection{Remote controlling clients}
        \label{subsubsec:remote_controlling_a_client}
      \subsubsection{Rendering on dedicated speakers}
        \label{subsubsec:rendering_on_dedicated_speakers}
    \subsection{Publisher/Subscriber interface}
      \label{subsec:publisher_subscriber_interface}
      The \gls{ssr} internally uses a \gls{pubsub}, which is a design pattern,
      to implement control through and over several parts of its components.\\
      In \gls{oop} \gls{pubsub} - also called observer, listener messaging - is
      usually comprised of a publisher class, handling the messages, without
      explicitely implementing how they will be used and a subscriber class,
      that allows for its implementations to subscribe to the messages
      provided. Filtering takes place to enable subscribers to only receive a
      certain subset of the messages.\\
      The \gls{ssr} implements a content-based filtering system, in which each
      subscriber evaluates the messages received and acts depending on its own
      constraints to implement further actions upon it.\\
      The abstract class Publisher defines the messages possible to send and
      provides means to subscribe to them. The global Controller class is its
      only implementation within the \gls{ssr}.\\
      The abstract class Subscriber in turn defines the messages understood,
      while its implementations in RenderSubscriber, Scene, OscSender and
      NetworkSubscriber take care of how they are used.\\
      This system enables a versatile messaging layout, in which components can
      call the publisher functionality in Controller, which in turn will send
      out messages to all of its subscribers.

    \subsection{\gls{ip} interface}
      \label{subsec:ip-interface}
      The \gls{ssr} from early on incorporated a network interface, that
      accepts specially terminated \gls{xml}-formatted strings over a \gls{tcp}
      port, called “\gls{ip} interface”. This has the benefit of reusing the
      same \gls{xml} parser code in use for scene and reproduction
      description.\\
      A downside is however, that - from the perspective of other software - it
      is complicated to use, as a conversion to \gls{xml} has to be attempted
      before sending a message to the \gls{ssr}. Additionally the message has
      to be linted (error checked) before sending and again parsed, after
      receiving an answer from the application.\\

      \paragraph{OSC through PureData}
        \label{par:osc_through_puredata}
        To allow \gls{osc} communication, the \gls{ssr} incorporates a Lua
        based \gls{pd} external. It uses two externals (iemnet and pdlua)
        alongside a Lua library for parsing and creating \gls{xml} (SLAXML).

      \paragraph{Sending and receiving}
        \label{par:sending_and_receiving}
        As mentioned in
        section~\nameref{subsec:publisher_subscriber_interface}, the
        NetworkSubscriber class (part of the \gls{ip} interface) implements the
        subscriber interface. This means: The network interface subscribes to
        the messages the publisher (the Controller instance) has to offer.
        Every time a function of the \gls{ssr}'s Controller instance, that was
        inherited from Publisher, is called, it will issue the call on all of
        its subscribers, too.\\

  \cleardoublepage
  \section{Results}
    \label{sec:results}
    \subsection{\gls{osc} interface}
      \label{subsec:osc-interface}
      \subsubsection{liblo}
        \label{subsubsec:liblo}
      \subsubsection{Client-Server setup}
        \label{subsubsec:client_server_setup}
        \begin{figure}[!htb]
          \centering
          \includegraphics[scale=1.0, trim = 20mm 204mm 10mm 10mm, clip]
          {ssr-client-server-shared-output.pdf}
          \caption{A diagram displaying a \gls{ssr} client/server setup, in
            which the server and the clients render audio collectively (e.g.
            \gls{wfs}). The server instance is not controlled via \gls{osc},
            but controls its clients through it.\\
            {\color{osc-in}\textbf{--}} \gls{osc} input
            {\color{osc-out}\textbf{--}} \gls{osc} output
            {\color{audio-in}\textbf{--}} Audio input
            {\color{audio-out}\textbf{--}} Audio output
          }
          \label{fig:ssr-client-server-shared-output}
        \end{figure}

      \subsubsection{Layered clients}
        \label{subsubsec:layered_clients}
      \subsubsection{Message interface}
        \label{subsubsec:message_interface}
  \cleardoublepage
  \section{Discussion}
    \label{sec:discussion}
    \paragraph{Stress testing the \gls{osc} interface}
      \label{par:stress_testing_the_osc_interface}
    \paragraph{Implementing a NullRenderer}
      \label{par:implementing_a_nullrenderer}
    \paragraph{Implementing AlienLoudspeaker}
      \label{par:implementing_alienloudspeaker}
    \paragraph{Interpolation of moving sources}
      \label{par:interpolation_of_moving_sources}

  \pagestyle{empty}
  \cleardoublepage
  \addcontentsline{toc}{section}{\listfigurename}
  \listoffigures
  \cleardoublepage
  \addcontentsline{toc}{section}{\listtablename}
  \listoftables
  \cleardoublepage
  \printindex
  \glsaddall
  \printglossaries
  \cleardoublepage
  \bibliographystyle{plainnat}
  \bibliography{../bib/ssr-networking}
\end{document}