\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}