System and method for synchronizing media presentation at multiple recipients

CROSS-REFERENCE TO RELATED APPLICATIONS

FIELD OF THE DISCLOSURE

0001The subject matter of the present disclosure relates to a system and method for synchronizing presentation of media at multiple recipients or devices on a network.

BACKGROUND OF THE DISCLOSURE

0002With the increasing capacity and capability of personal computers, as well as improved multimedia interfaces for these computers, it has become popular to use personal computers as a repository for multimedia content, such as songs, movies, etc. Particularly with music, the increased popularity of storing multimedia information on a personal computer has resulted in a variety of products and services to serve this industry. For example, a variety of stand-alone players of encoded multimedia information have been developed, including, for example, the iPod, produced by Apple Computer of Cupertino, California. Additionally, services have been developed around these devices, which allow consumers to purchase music and other multimedia information in digital form suitable for storage and playback using personal computers, including, for example, the iTunes music service, also run by Apple Computer.

0003These products and services have resulted in an environment where many consumers use their personal computer as a primary vehicle for obtaining, storing, and accessing multimedia information. One drawback to such a system is that although the quality of multimedia playback systems for computers, e.g., displays, speakers, etc. have improved dramatically in the last several years, these systems still lag behind typical entertainment devices, e.g., stereos, televisions, projection systems, etc. in terms of performance, fidelity, and usability for the typical consumer.

0004Thus, it would be beneficial to provide a mechanism whereby a consumer could easily obtain, store, and access multimedia content using a personal computer, while also being able to listen, view, or otherwise access this content using conventional entertainment devices, such as stereo equipment, televisions, home theatre systems, etc. Because of the increasing use of personal computers and related peripherals in the home, it would also be advantageous to integrate such a mechanism with a home networking to provide an integrated electronic environment for the consumer.

0005In addition to these needs, there is also increasing interest in the field of home networking, which involves allowing disparate devices in the home or workplace to recognize each other and exchange data, perhaps under the control of some central hub. To date a number of solutions in this area have involved closed systems that required the purchase of disparate components from the same vendor. For example, audio speaker systems that allow computer-controlled switching of music from one location to another may be purchased as a system from a single vendor, but they may be expensive and/or may limit the consumer's ability to mix and match components of a home network from different vendors according to her own preferences. Thus, it would be beneficial to provide a mechanism by which various home networking components from differing vendors can nonetheless interact in a home network environment.

0006The subject matter of the present disclosure is directed to overcoming, or at least reducing the effects of, one or more of the problems set forth above.

0007Further background to the present invention can be found in: <ul id="ul0001" list-style="none" compact="compact"><li><patcit id="pcit0001" dnum="EP1122931A"><text>EP 1122931</text></patcit> which discloses a method for media content synchronisation. Each multimedia receiver includes a buffer, "B", to store incoming media packets, and recalls a packet from the buffer to play the content of a multimedia packet at (TED-CND) after its reception.</li><li><patcit id="pcit0002" dnum="EP1523171A"><text>EP 1523171</text></patcit>, which describes an implementation facilitating fast start-up of a new media stream while avoiding temporal interruption of the presentation of that new media stream.</li><li><patcit id="pcit0003" dnum="US6630963B"><text>US6630963</text></patcit> which relates to synchronising two different media streams (an audio stream and a video stream) on the same client device, where the audio stream is slower than video stream. In order to synchronise the streams, the transmission delay is calculated and presentation of the audio stream is delayed based on the calculated delay.</li></ul>

SUMMARY OF THE DISCLOSURE

0008The invention is defined by the subject matter of the independent claims. Advantageous enhancements are subject to the dependent claims.

BRIEF DESCRIPTION OF THE DRAWINGS

0009The foregoing summary, preferred embodiments, and other aspects of subject matter of the present disclosure will be best understood with reference to a detailed description of specific embodiments, which follows, when read in conjunction with the accompanying drawings, in which: <ul id="ul0002" list-style="none" compact="compact"><li><figref idref="f0001">FIG. 1</figref> illustrates an embodiment of a network media delivery system according to certain teachings of the present disclosure.</li><li><figref idref="f0001">FIG. 2</figref> illustrates an embodiment of a networked media station or client device.</li><li><figref idref="f0002 f0003">FIG. 3</figref> illustrates a process of operating the disclosed system in flowchart form.</li><li><figref idref="f0004">FIG. 4</figref> illustrates an embodiment of an interface of a media application operating on a host device of the disclosed system.</li><li><figref idref="f0005">FIG. 5A</figref> illustrates portion of the disclosed system having a host device delivering packets to multiple client devices.</li><li><figref idref="f0006">FIG. 5B</figref> illustrates portion of the disclosed system having a host device and client devices performing retransmission of lost packet information.</li><li><figref idref="f0007">FIG. 6A</figref> illustrates an embodiment of a packet requesting retransmission of lost packets.</li><li><figref idref="f0007">FIG. 6B</figref> illustrates an embodiment of a response to retransmission request.</li><li><figref idref="f0007">FIG. 6C</figref> illustrates an embodiment of a response to a futile retransmission request.</li><li><figref idref="f0008">FIG. 7</figref> illustrates portion of the disclosed system having a host device and multiple client devices exchanging time information.</li><li><figref idref="f0009">FIG. 8A</figref> illustrates an embodiment of a packet for synchronizing time.</li><li><figref idref="f0010">FIG. 8B</figref> illustrates an embodiment of a packet for announcing time.</li><li><figref idref="f0011">FIG. 9</figref> illustrates portion of the disclosed system having a host device and a client device.</li><li><figref idref="f0012">FIG. 10</figref> illustrates an algorithm to limit stuttering in playback of audio.</li></ul>

0010While the subject matter of the present disclosure is susceptible to various modifications and alternative forms, specific embodiments thereof have been shown by way of example in the drawings and are herein described in detail. The figures and written description are not intended to limit the scope of the inventive concepts in any manner.

DETAILED DESCRIPTION

0011A network media delivery system having a host device and multiple client devices is described herein. The following embodiments disclosed herein are described in terms of devices and applications compatible with computer systems manufactured by Apple Computer, Inc. of Cupertino, California. The following embodiments are illustrative only and should not be considered limiting in any respect.

I. Components of the Network Media Delivery System

0012Referring to <figref idref="f0001">FIG. 1</figref>, an embodiment of a network media delivery system 10 according to certain teachings of the present disclosure is illustrated. The system 10 includes a host device or computer system 20 and one or more networked media stations or client devices 50, and various other devices. The system 10 in the present embodiment represents only one of several possible configurations and is meant to be illustrative only. Other possible configurations are discussed in <patcit id="pcit0004" dnum="US862115" dnum-type="L"><text>U.S. Patent Application Ser. No. 10/862,115</text></patcit>. For example, the host device 20 can have a wired or wireless connection to each of the client devices 50 without the use of a hub or base station 30, or the host device 20 can have a wireless connection to the hub or base station 30. The system 10 is used to distribute media (e.g., audio, video, multimedia, etc.) via network connections from the host device 20 to multiple client devices 50 located throughout an environment, such as a house, office, etc.

0013The host device 20 is a personal computer, such as an AirPort-equipped Mac or a Wi-Fi-compliant Windows-based PC. The client devices 50 are networked media stations, such as disclosed in <patcit id="pcit0005" dnum="US862115" dnum-type="L"><text>U.S. Patent Application Ser. No. 10/862,115</text></patcit>. The client devices 50 are plugged into wall sockets, which provide power to the client devices 50, and arc coupled to entertainment devices, such as amplifiers 80, powered speakers, televisions, stereo systems, videocassette recorders, DVD players, home theatre systems, or other devices capable of delivering media known in the art.

0014An example of the client device 50 is discussed briefly with reference to <figref idref="f0001">FIG. 2</figref>. The client device 50 includes an AC power adapter portion 52 and a network electronics portion 54. The network electronics portion 54 includes a wired network interface 62, a peripheral interface 64, and a media interface 66. As illustrated, the wired network interface 62 is an Ethernet interface, although other types of wired network interface known in the art could be provided. Similarly, the peripheral interface 64 is illustrated as a USB interface, although other types of peripheral interfaces, such as IEEE 1394 ("Firewire"), RS-232 (serial interface), IEEE 1284 (parallel interface), could also be used. Likewise, the media interface 66 is illustrated as an audio interface including both an analog lineout and an optical digital audio functionality. However, other media interfaces known in the art, such as a multimedia interface or a video intcrfacc using composite video, S-vidco, component video, Digital Video Interface (DVI), High Definition Multimedia Interface (HTMI), etc., could also be provided.

0015The network electronics portion 54 also includes a wireless networking interface 68. The wireless network interface 68 preferably takes the form of a "Wi-Fi" interface according to the IEEE 802.11b or 802.11g standards know in the art. However, other wireless network standards could also be used, either in alternative to the identified standards or in addition to the identified standards. These other network standards can include the IEEE 802.11a standard or the Bluetooth standard, for example.

0016Returning to <figref idref="f0001">FIG. 1</figref>, the host device 20 runs a media application 22. In one exemplary embodiment, the media application 22 is iTunes software for media file management and playback produced by Apple Computer, Inc. In the present configuration, which is only one of several possibilities, the host device 20 is equipped with an Ethernet port that is connected via a cable 24 to a base station 30. The base station 30 can be any variety of access points known in the art. Preferably, the base station 30 includes wireless access, routing, switching and firewall functionality. The base station 30 is connected via a cable 42 to a modem 40, which receives an Internet connection through a connection 44. Using this arrangement, multimedia files stored on host device 20 can be played using stereo amplifiers 80, which are connected to client devices 50 using one of the audio interfaces on the client devices 50. The host device 20 and the client devices 50 preferably communicate via a wireless network segment (illustrated schematically by connections 32), but wired network segments formed by wired connections, such as Ethernet cables, could also provide communication between the host device and the client devices 50. The client devices 50 communicate with the entertainment devices via a wired network segment 82.

0017The client devices 50 act as wireless base stations for a wireless network and enable the host device 20 to deliver media (e.g., audio, video, and multimedia content) at multiple locations in an environment. For example, the client devices 50 are connected to stereo amplifiers 80 or other entertainment devices to playback media stored on the host device 20. In one embodiment, a line level audio or a digital fiber optic type of connector connects the client devices 50 to the stereo amplifiers 80. Either type of connector can plug into the multimedia port (66; <figref idref="f0001">FIG. 2</figref>), which is a dual-purpose analog/optical digital audio mini-jack. To interface with stereo amplifiers 80, a mini stereo to RCA adapter cable 82 is used, which connects to RCA-type right and left audio input ports on the stereo amplifier 80. Alternatively, a Toslink digital fiber optic cable can be used, which would connect to digital audio input port on the stereo amplifiers 80. These and other configurations are disclosed in <patcit id="pcit0006" dnum="US862115" dnum-type="L"><text>U.S. Patent Application Ser. No. 10/862,115</text></patcit>.

0018For the purposes of the present disclosure, the client devices 50 can also be connected to laptops 70 or personal computers that are capable of playing media (audio, video, etc.) so that the laptops and personal computers can also be considered entertainment devices. Moreover, the laptops 70 or personal computers can have the same functionality as both a client device 50 and an entertainment device so that the laptops 70 and personal computers can be considered both a client device and an entertainment device. Accordingly, the term "client device" as used herein is meant to encompass not only the networked media stations associated with reference numeral 50, but the term "client device" as used herein is also intended to encompass any device (e.g., laptop, personal computer, etc.) compatible with the network media delivery system 10 according to the present disclosure. In the present disclosure, however, reference is made to client devices 50 for ease in discussion. Furthermore, the term "entertainment device" as used herein is meant to encompass not only stereo amplifiers 80 as shown in <figref idref="f0001">FIG. 1</figref>, but the term "entertainment device" as used herein is also intended to encompass powered speakers, televisions, stereo systems, videocassette recorders, a DVD players, home theatre systems, laptops, personal computers, and other devices known in the art that capable of delivering media.

0019The client devices 50 receive media data from the host device 20 over network connections and output this media data to the entertainment devices. Although it is contemplated that audio, video, audio/video, and/or other forms of multimedia may be used, exemplary embodiments disclosed herein relate to sharing of audio with client devices 50 connected to entertainment devices, such as stereo amplifiers 80, or with laptops 70 or other computers having internal speakers or the like. The audio can be stored on the host device 20 or can be obtained from the Internet 46. However, it will be appreciated that the teachings of the present disclosure can be applied to video, audio/video, and/or other forms of multimedia in addition to the audio in the exemplary embodiments disclosed herein. Furthermore, in the discussion that follows, various details of the network media delivery system are implemented using hardware and software developed by Apple Computer, Inc. Although certain details are somewhat specific to such an implementation, various principles described are also generally applicable to other forms of hardware and/or software.

0020During operation, the system 10 delivers the same audio in separate locations of an environment (e.g., multiple rooms of a home). The system 10 addresses several issues related to playing the same audio in multiple, separate locations. One issue involves playing the audio in the separate locations in a synchronized manner with each other. Because the host device 20 and the client devices 50 have their own processors, memory, and transmission interfaces, sending or streaming audio from the host device 20 to the client devices 50 through a wireless or wired communication link will not likely result in synchronized playing of the audio at the separate locations. In addition, the client device 50 may be connected to different types of entertainment devices, which may have different latency and playback characteristics. It is undesirable to play the same audio in the separate locations out of sync because the listener will hear echoes and other undesirable audio effects. The system 10 addresses this issue by substantially synchronizing the playing of the audio in each location so that echo and other effects can be avoided. It should be noted that the level of precision required to substantially synchronize the playing of media at each location depends on the type of media being played, the perceptions of the user, spatial factors, and other details specific to an implementation.

0021Another issue related to playing of the same audio involves how to handle lost audio data at the separate locations. To address this issue, the disclosed system 10 preferably uses a retransmission scheme to recover lost audio. These and other issues and additional details of the disclosed network media delivery system arc discussed below.

II. Process of Operating the System

0022Referring to <figref idref="f0002">FIG. 3A</figref>, a process 100 of operating the network media delivery system of the present disclosure is illustrated in flowchart form. During discussion of the process 100, reference is concurrently made to components of <figref idref="f0001">FIG. 1</figref> to aid understanding. As an initial step in the process 100, network discovery is performed, and the networked client devices 50 and other configured devices (e.g., a configured laptop 70) publish or announce their presence on the network using a predefined service type of a transfer control protocol (Block 102). The host device 20 browses the local sub-net for the designated service type (Block 104).

0023The network discovery is used to initiate the interface between the host device 20 and client devices 50 and other compatible devices over the network of the system 10. One example of such a network discovery uses Bonjour, which is a technology that enables automatic discovery of computers, dcviccs, and scrviccs on IP networks. Bonjour uses standard IP protocols to allow devices to find each other automatically without the need for a user to enter IP addresses or configure DNS servers. Various aspects of Bonjour are generally known to those skilled in the art, and are disclosed in the technology brief entitled "<nplcit id="ncit0001" npl-type="b"><text>MAC OS X: Bonjour," dated April 2005, and published by Apple Computer</text></nplcit>. To provide the media sharing functionality between the host device 20 and the client devices 50, the client devices 50 advertise over the network that they support audio streaming and particular audio capabilities (e.g., 44.1 kHz sample rate, 16-bit sample size, and 2-channel/stereo samples). The client devices 50 may also advertise security, encryption, compression, and other capabilities and/or parameters that are necessary for communicating with the client devices 50.

0024When complaint client devices 50 are discovered, the addresses and port numbers of the discovered devices 50 are stored for use by the system 10. Then, the media application 22 displays information about the found client devices 50 in a user interface operating on the host device 20 (Block 106). In one embodiment, for example, the media application 22 discovers the client devices by obtaining information of the user's step up of computers and networks for their house, office, or the like from another application containing such information. In another embodiment, for example, the media application 22 discovers the client devices 50 and recognizes these client devices 50 as potential destinations for audio data. Then, the media application 22 automatically provides these recognized devices 50 as part of a selectable destination for audio playback in a user interface.

0025<figref idref="f0004">FIG. 4</figref> shows an example of a user interface 200 associated with the media application, such as iTunes. Among other elements, the user interface 200 shows an icon 202 for selecting playback locations (<i>e.g</i>., networked client devices and other playback devices located in a house), which have detected on the network. A user may select the icon 202 to access a pop-up menu 204 in which the user can activate/deactivate (<i>i.e.,</i> check or uncheck) one or more of the playback locations as destinations for audio playback. Of course, the user interface 200 can display possible destinations for audio playback in a number of ways. For example, the display of possible destination can include a network schematic of the user's dwelling, office, or the like, that shows possible destination, or the display can be customized by the user.

0026Returning to <figref idref="f0002">FIG. 3A</figref>, the user selects one or more of the client devices to be used for playback in the user interface (Block 108). The host device 20 then uses Real-Time Streaming Protocol (RTSP) to set up and control the audio stream, and the host device 20 initiates an RTSP connection to each of the selected client devices 50 to determine which set of features the devices 50 support and to authenticate the user (if a password is required) (Block 110). On the host device 20, the user can then start playback using the user interface of the media application 22 (Block 112). The host device 20 makes an RTSP connection to each client device 50 to set it up for playback and to start sending the audio stream (Block 114). The host device 20 then sends a command to each client device 50 to initiate playback (Block 116). When each client device 50 receives the command, the device 50 negotiates timing information via User Datagram Protocol (UDP) packct exchanges with the host dcvicc 20 (Block 118). Each client device 50 then determines whether the timing negotiation either succeeds or fails (Block 119). The client devices 50 do not respond to the command to initiate playback until the timing negotiation either succeeds or fails. The timing negotiation occurs early to guarantee that the client devices 50 have the initial timing information needed to synchronize their clocks with the host device 20 before any audio packets are processed by the client devices 50. If the negotiation succeeds, the client device 50 can be used for playback (Block 120). If the negotiation fails, however, the associated client device 50 can perform a number of possible operations (Block 121). For example, the client device 50 can return an error to the host device 20 in response to the command, and the session on this device 50 can be terminated. In another possible operation, the associated client device 50 can retry to negotiate the timing information. Alternatively, the associated client device 50 can ignore the fact that negotiating timing information has failed. This may be suitable when the user is not interested in the audio playing in synchronized manner in the multiple locations associated with the client devices 50. For example, the client device may be located by the pool or out in the garage and does not necessarily need to deliver the audio in synch with the other devices.

0027During playback at Block 120, the host device 20 sends audio data to the client devices 50, which process the audio data and deliver processed audio to the connected entertainment devices. An example of the process of playing back audio is discussed below with reference to the flowchart of <figref idref="f0003">FIG. 3B</figref> with concurrent reference to element numerals of <figref idref="f0001">FIG. 1</figref>. Various buffering, error checking, and other data transfer steps have been omitted from the general description of <figref idref="f0003">FIG. 3B</figref>.

0028As discussed above, the host device 20 is connected to a wireless network established by the access point 30, which can also provide for a shared connection to the Internet or other network 46. The client devices 50 are also connected to the wireless network and have their multimedia ports connected to stereo amplifiers 80 or other entertainment device having output speakers or other multimedia output capability. A digital media file (<i>e.g.,</i> a song in ACC format) is stored on the host device 20. Once playback is started (Block 122), the host device 20 transcodes a portion of the media file from the format (<i>e.g</i>., AAC) in which it is stored to a format that is understood by client device 50 (Block 124). This transcoding step is not necessarily required if the file is stored on the host device 20 in a format that is understood by the client device 50. In any case, a block of audio data for transmission is created (Block 126). This audio data is preferably compressed and encrypted (Block 128). Encryption is not necessarily required, but it is advantageous for digital rights management purposes.

0029The host device 20 then transmits the audio data over the wireless network to the client devices 50 (Block 130). The client devices 50 decrypt and decompress the received audio data (Block 132), and the client devices 50 decode the audio data based on the encoding performed in Block 124 (Block 134). The decoding results in raw audio data, which may be, for example, in the form of PCM data. This data is converted to analog audio signals by digital-to-audio converters (DAC) (Block 136), and the audio signals are output to the stereo amplifiers 80 for playing with their loudspeakers (Block 138).

0030With the benefit of the description of the components of the disclosed network media delivery system and its process of operation provided in <figref idref="f0001 f0002 f0003 f0004">FIGS. 1 through 4</figref>, the discussion now turns to details related to how data is transferred between the host device and client devices, how lost data is handled, and how playback is synchronized, in addition to other details disclosed herein.

III. Network Transport Used for the System

0031To transfer audio data and other information, the network media delivery system 10 of the present disclosure preferably uses User Datagram Protocol (UDP) as its underlying transport for media data. UDP is beneficial for synchronized playback to the multiple client devices 50 because synchronized playback places time constraints on the network protocol. Because audio is extremely time sensitive and has a definite lifetime of usefulness, for example, a packet of media data, such as audio, can become useless if it is received after a point in time when it should have been presented. Accordingly, UDP is preferred because it provides more flexibility with respect to the time sensitive nature of audio data and other media data.

0032To use UDP or some similar protocol, the disclosed system is preferably configured to handle at least a small percentage of lost packets. The lost packets can be recovered using Forward Error Correction (FEC), can be hidden using loss concealment techniques (e.g. repetition, waveform substitution, etc.), or can be recovered via retransmission techniques, such as those disclosed herein. Although UDP is preferred for the reasons set forth herein, Transmission Control Protocol (TCP) can be used. Depending on the implementation, retransmission using TCP may need to address problems with blocking of transmissions. If a TCP segment is lost and a subsequent TCP segment arrives out of order, for example, it is possible that the subsequent segment is held off until the first segment is retransmitted and arrives at the receiver. This can result in a chain reaction and effective audio loss because data that has arrived successfully and in time for playback may not be delivered until it is too late. Due to some of the retransmission difficulties associated with TCP, the Partial Reliability extension of Stream Control Transmission Protocol (SCTP) can provide the retransmission functionality. Details related to the Partial Reliability of SCTP are disclosed in RFC 3758, which can be obtained from <u>http://www.ietf.org/rfc/rfc/3758.txt,</u>

0033UDP is preferred for time critical portions of the protocol because it can avoid some of the problems associated with blockage of transmission. For example, UDP allows the host's media application 22 to control retransmission of lost data because the media application 22 can track time constraints associated with pieces of audio data to be delivered. Based on the known time constraints, the media application 22 can then decide whether retransmission of lost packets of audio data would be beneficial or futile. All the same, in other embodiments, time critical portions of the disclosed system, such as time syncing, can be implemented using UDP, and audio data delivery can use TCP with a buffering system that addresses blocking problems associated with TCP.

IV. Audio Streaming and Playback with System

0034Before discussing how the client devices negotiate timing information in order to play audio in synchronization, the discussion first addresses how the disclosed system streams audio for playback. Referring to <figref idref="f0005">FIG. 5A</figref>, a portion of the disclosed system 300 is shown with a host device 320 and at least two client devices 350A-B. Each of the client devices 350 has a processor 352, a memory 354, a transmission interface 356, and an audio interface 358. The client devices 350 also include a UDP stack and can include a TCP stack depending on the implementation. As noted previously with reference to the client device of <figref idref="f0001">FIG. 2</figref>, the transmission interfaces 356 can be a Wi-Fi-compatible wireless network interface, and the audio interface 358 can provide an analog and/or an optical digital output. The processor 352 and memory 354 can be conventional hardware components known in the art. The memory 354 has two audio buffers 361 and 362. Although not shown in <figref idref="f0005">FIG. 5A</figref>, each of the client devices 350 has a local clock, a playback engine, and other features.

0035The host device 320 uses several commands to set up a connection with and to control operation of the client dcviccs 350. These commands include ANNOUNCE (used for identification of active client devices), SETUP (used to setup connection and operation), RECORD (used to initiate playback at client devices), PAUSE (used to pause playback), FLUSH (used to flush memory at the client devices), TEARDOWN (used to stop playback), OPTIONS (used to configure options), GET_PARAMETER (used to get parameters from the client devices), and SET_PARAMETER (used to set parameters at the client devices).

0036Preferably, the client devices 350 are authenticated when initially establishing a connection to the media application 322 running on the host device 320. Upon successful authentication, the media application 322 opens network connections to the transmission interface 356 of the client devices 350. Preferably, network connections between the host device 320 and the client devices 350 are separated into an audio channel for sending audio data and a control channel used to set up connection and operation between the devices 320 and 350. However, a single channel could be used for data and control information. Once the connections are established, the host device 320 begins sending data to the client devices 350. In turn, the clicnt dcviccs 350 rcccivc the audio data, buffer some portion of the data, and begin playing back the audio data once the buffer has reached a predetermined capacity.

0037Communication between the host device 320 and the client devices 350 preferably uses the Real Time Streaming Protocol (RTSP) standard. The media application 322 at the host device 320 preferably uses Real-Time Transport Protocol (RTP) encapsulated in User Datagram Protocol (UDP) packets 330 to deliver audio data from the host device 320 to the client devices 350. RTSP, RTP, and UDP are standards known to those skilled in the art. Therefore, some implementation details are not discussed here. Details of RTSP can be found in "<nplcit id="ncit0002" npl-type="s" url="http://www.ietf.ors/rfc/rfc2326.txt"><text>Real-Time Streaming Protocol," RFC 2326, which is available from http://www.ietf.ors/rfc/rfc2326.txt</text></nplcit><u>.</u> Details of RTP can be found in "<nplcit id="ncit0003" npl-type="s" url="http)://www.ietf.org./rfc/rfc3550.txt"><text>Real-Time Transport Protocol," RFC 3550, which is available from http)://www.ietf.org./rfc/rfc3550.txt</text></nplcit>.

0038The packets 330 have RTP headers and include both sequence numbers and timestamps. The data payload of the RTP packets 330 contains the audio data to be played back by the client devices 350. The media files, from which the packets 330 are derived, can be stored on host device 320 in one or more formats, including, for example, MP3 (Motion Picture Expert's Group Layer 3), AAC (Advanced Audio Coding a/k/a MPEG-4 audio), WMA (Windows Media Audio), etc. Preferably, the media application 322 running on the host device 320 decodes these various audio formats to construct the packets 330 so that the client devices 350 do not need decoders for multiple formats. This also reduces the hardware performance requirements of the client devices 350. Another advantage of performing decoding on the host device 320 is that various effects may be applied to the audio stream, for example, cross fading between tracks, volume control, equalization, and/or other audio effects. Many of these effects would be difficult or impossible to apply if the client device 350 were to apply them, for example, because of the computational resources required. Although not preferred in the present embodiment, other embodiments of the present disclosure can allow for decoding at the client devices 350 for audio and other forms of media.

0039The host device 320 preferably uses a separate unicast stream 310A-B of RTP packets 330 for each of the client devices 350A-B. In the present embodiment, the separate unicast streams 310A-B are intended to deliver the same media information (<i>e.g</i>., audio) to each of the client devices 350A-B so that the same media can be presented at the same time from multiple client devices 350A-B. In another embodiment, each of the separate unicast streams 310A-B can be used to deliver separate media information (<i>e.g</i>., audio) to each of the client devices 350A-B. The user may wish to unicast separate media information in some situations, for example, if a first destination of a first unicast stream of audio is a client device in a game room of a house and a second destination of a second unicast stream of different audio is a client device in the garage of the house. Therefore, it may be preferred in some situations to enable to the user to not only select sending the same media information by unicast streams to multiple client devices by to also allow the user to send different media information by separate unicast streams to multiple client devices. The user interface 200 of <figref idref="f0004">FIG. 4</figref> can include a drop down menu or other way for the user to make such a related selection.

0040Separate unicast streams 310 are preferred because multicasting over wireless networks can produce high loss rates and can be generally unreliable. All the same, the disclosed system 300 can use multicasting over the wireless network. In general, though, bandwidth limitations (i.e. fixed multicast rate), negative effects on unicast performance (low-rate multicast slows down other unicast traffic due to multicast packets taking longer), and loss characteristics associated with multicasting over wireless (multicast packets are not acknowledged at the wireless layer) make multicasting less desirable than using multiple, unicast streams 310A-B as preferred. Use of multiple, unicast streams 310A-B does correspond to an increase in bandwidth as additional client devices 350 are added to a group of designated locations for playback. If the average compression rate for audio data is about 75%, the increase in bandwidth associated with multiple, unicast streams 310A-B may correspond to about 1 Mbit/sec bandwidth required for each client device 350 so that the host device 320 can send compressed audio data to the access point (e.g., 30; <figref idref="f0001">FIG. 1</figref>) and another 1 Mbit/sec so that the access point can forward the compressed audio data to the client device 350.

0041Once an RTSP session has been started and the RECORD command has been sent from the host device 320 to the client devices 350, the host device 320 begins sending normal RTP packets 330 containing the audio data for playback. These RTP packets 330 are sent at regular intervals, based on the number of samples per second, which can be about 44,100 Hz for audio. The RTP packets 330 are sent at the regular intervals in a throttled and evenly spaced manner in order to approximate the audio playback rate of the remote client devices 350 because the UDP-based connection does not automatically control the sending of data in relation to the rate at which that data is consumed on the remote client devices 350.

0042Because each of the multiple client devices 350 has their own audio buffers 361, 362, network conditions, etc., it may not be desirable to use a feedback scheme when sending the packets 330. Accordingly, the host device 320 sends audio data at a rate that preferably does not significantly under-run or overrun a playback engine 353 of any of the remote client devices 350. To accomplish this, the host device 320 estimates a fixed delay 340 to insert between packets 330 to maintain the desired audio playback rate. In one embodiment, the packets 330 of audio data are sent with a delay of about 7.982-ms between packets 330 (<i>i.e.,</i> 352 samples per packet / 44,100 Hz = ∼7.982-ms per packet), which corresponds to a rate of about 125 packets/sec. Because the delay 340 is fixed, each of the client devices 350 can also detect any skew between its clock and the clock of the sending host device 320. Then, based on the detected skew, each client device 350 can insert simulated audio samples or remove audio samples in the audio it plays back in order to compensate for that skew.

0043As alluded to above, the RTP packets 330 have timestamps and sequence numbers. When an RTP packet 330 is received by a client device 350, the client device 350 decrypts and decompresses the payload (see Encryption and Compression section below), then inserts the packet 320, sorted by its timestamp, into a packet queue. The two audio buffers 361 and 362 are alternatingly cycled as audio is played back. Each audio buffer 361 and 362 can store a 250-ms interval of audio. The received RTP packets in the packet queue are processed when one of the two, cycling audio buffers 361 and 362 completes playback. In one embodiment, the audio is USB-based so this is a USB buffer completion process.

0044To process the queued packets, the engine 353 assembles the queued RTP packets in one of the audio buffers 361 or 362. During the assembly, the engine 353 calculates when each of queued RTP packets should be inserted into the audio stream. The RTP timestamp in the packets combined with time sync information (see the Time Synchronization section below) is used to determine when to insert the packets. The engine 353 performs this assembly process and runs through the queued packets to fill the inactive audio buffer 361 or 362 before the currently playing audio buffer 361 or 362 has completed. Because each of the audio buffers 361 and 362 can store 250-ms of audio, the client device 350 has a little less than 250-ms to assemble all the RTP packets, conceal any losses, and compensate for any clock skew. If there are any gaps in the audio (<i>e.g</i>., the device's audio clock is skewed from the host's audio clock, a packet was lost and not recovered, etc.), then those gaps can be concealed by inserting simulated audio samples or removing existing audio samples.

V. Encryption and Compression

0045For digital rights management purposes, it is desirable to determine whether the client devices 350 are authorized to receive an audio data stream and/or whether the communications links between the host device 320 and the client devices 350 are secure (encrypted). This requires some form of authentication, which is preferably based on a public key/private key system. In one embodiment, each client station 350 is provided with a plurality of private keys embedded in read only memory (ROM). The media application at the host device 320 is then provided with a corresponding plurality of public keys. This allows identification data transmitted from the networked client devices 350 to the media application to be digitally signed by the client device 350 using its private key, by which it can be authenticated by the media application at the host device 320 using the appropriate public key. Similarly, data sent from the media application at the host device 320 to the networked client stations 350 is encrypted using a public key so that only a client device 350 using the corresponding private key can decrypt the data. The media software and networked media station can determine which of their respective pluralities of keys to use based on the exchange of a key index, telling them which of their respective keys to use without the necessity of transmitting entire keys.

0046In addition to encryption, the decoded audio data is preferably compressed by host device 320 before transmission to the client devices 350. This compression is most preferably accomplished using a lossless compression algorithm to provide maximum audio fidelity. One suitable compressor is the Apple Lossless Encoder, which is available in conjunction with Apple's iTunes software. The client devices 350 require a decoder for the compression codec used.

0047The RTP packets 330 are preferably compressed using the Apple Lossless algorithm and arc preferably encrypted using the Advanced Encryption Standard (AES) with a 128-bit key size. Loss is still inevitable even though the system 300 uses a UDP-based protocol that attempts to recover from packet loss via retransmission and/or Forward Error Correction (FEC). For this reason, encryption and compression preferably operate on a per-packet basis. In this way, each packet 330 can be completely decoded entirely on its own, without the need for any surrounding packets 330. The Apple Lossless algorithm is used to compress each individual packet 330 rather than compressing a larger stream of audio and packetizing the compressed stream. Although compressing each individual packet 330 may reduce the effectiveness of the compression algorithm, the methodology simplifies operation for the client devices 350 and allows them to be more tolerant to packet loss. Although compression rates are highly dependent on the content, music audio can have an average compression rate of about 75% of the original size when used by the disclosed system 300.

0048The AES-128 algorithm is used in frame-based cipher block chaining (CBC) mode to encrypt payloads of the RTP packets 330 and the RTP payload portion of RTCP retransmission packets (380; <figref idref="f0006">FIG. 5B</figref>) discussed below. Because each packet 330 represents a single audio frame, no other packets are required to decrypt each packet correctly. The system preferably supports any combination of encryption and compression, such as both encryption and compression, encryption only, compression only, or neither encryption nor compression. Encryption and compression are configured during the RTSP ANNOUNCE command. The format used to configure encryption and compression is based on the Session Description Protocol (SDP) and embedded as RTSP header fields. Compression uses an SDP "m" (media description) combined with an "rtpmap" and "fmtp" to specify the media formats being used numerically and how those numbers map to actual compression formats and algorithms.

VI. Retransmission of Lost Packets of Audio Data

0049As noted above, the RTP packets 330 received from the host device 320 have RTP sequence numbers. Based on those RTP sequence numbers, the client device 350 can determine whether packets 330 that have been lost during transmission or for other reasons. The lost RTP packets 330 cannot be queued for playback in the audio buffers 361 and 362 of the client devices 350 so that gaps will result in the audio. To address this issue, the client devices 350 requests that the lost packet(s) be retransmitted. Referring to <figref idref="f0006">FIG. 5B</figref>, portion of the disclosed system 300 is shown again to discuss how the system 300 attempts to retransmit packets lost during original transmission.

0050To handle retransmissions, the system 300 preferably uses Real-Time Transport Control Protocol (RTCP) when packet loss is detected. As note above, the sequence numbers associated with the received RTP packets (330; <figref idref="f0005">FIG. 5A</figref>) are used to determine if any packets have been lost in the transmission. If there is a gap in the sequence numbers, the client device 350 sends a retransmission request 370 to the sender (<i>e.g</i>., host device 320 or other linked client device 350) requesting all the missing packets. In one embodiment, the retransmission request 370 can request up to a maximum of 128 lost packets per detected gap.

0051In response to the retransmission request 370, the host device 320 sends one or more retransmission responses 380 for lost packets. Due to limitations of the maximum transmission unit (MTU) on RTCP packet sizes, only one response can be sent per retransmission response packct 380. This means that a single retransmission request packet 370 from a device 350 may generate up to 128 retransmission response packets 380 from the host device 320 if all of the lost packets are found in the host's recently sent packets.

0052Because RTP does not currently define a standard packet to be used for retransmissions, an RTP extension for an RTCP Retransmission Request packet is preferably defined. <figref idref="f0007">FIG. 6A</figref> shows an example of an RTCP Retransmit Request Packet 370 for use with the disclosed system. The Sequence Number Base refers to the sequence number of the first (lost) packet requested by this RTCP Retransmit Request Packet 370. The Sequence Number Count refers to the number of (lost) packets to retransmit, starting at the base indicated.

0053In <figref idref="f0005">FIG. 5A</figref>, the client device 350 sending the RTCP Retransmission Request packet 370 tracks the retransmission requests that it sends in a queue to facilitate sending additional requests if a response to the retransmission request 370 is not received in a timely manner. When a retransmission request 370 has not been responded to in a timely manner, another retransmission request 370 is sent from the client device 350. The process of retrying can be continued until a maximum time has elapsed since the first retransmission request 370 was sent. After that maximum time, it is likely too late to deal with the lost packet anyway because the lost packets time for insertion in one of the audio buffers 361 or 362 has passed.

0054When multiple, contiguous packets have been lost, the initial retransmit request 370 includes all the missing packets. However, if a response 380 is not received in a timely manner, the missing packets are spread out among multiple requests 370 over time when reattempts are made. Spreading out among multiple requests can maintain a uniform delivery of request and response packets. This also prioritizes packets by time and defers delivery of packets whose presentation time is later.

0055When the host device 320 receives a retransmission request 370, the host device 320 searches a list of recently sent packets stored at the device 320. If the requested packet in the request 370 is found, the host device 320 sends a retransmission response 380 to the clicnt device 350. An example of an RTP extension for an RTCP Retransmit Response Packet 380 is shown in <figref idref="f0007">FIG. 6B</figref>. The RTCP Retransmit Response Packet 380 includes the complete RTP packet (<i>e.g</i>., header and payload) being retransmitted. The retransmission packet 380, however, is only sent to the sender of the retransmission request 370, unlike the normal RTP packets (330; <figref idref="f0005">FIG. 5A</figref>) that are sent to all devices participating in the session.

0056If the requested packet is not found by the host device 320, however, a negative response 390 is sent so the corresponding client device 350 knows that any further attempt to request that particular packet is futile. An example of an RTP extension for an RTCP Futile Retransmit Response Packet 390 is shown in <figref idref="f0007">FIG. 6C</figref>. The RTCP Futile Retransmit Response Packet 390 includes the 16-bit sequence number of the failed packet followed by a 16-bit pad containing zero.

0057In <figref idref="f0006">FIG. 5B</figref>, the client device 350 receiving a retransmission response packet 380 inserts the packet 380 into the packet queue in the same way used for inserting packets received as part of the normal RTP packet stream discussed above with reference to <figref idref="f0005">FIG. 5A</figref>. By definition, however, the retransmission response packet 380 is already out-of-sequence and, therefore, does not trigger new retransmission requests based on its sequence number. If an existing packet already exists at the same timestamp as the incoming packet, either via the normal RTP stream or via retransmission, the packet is dropped as a duplicate.

0058Scheduling retransmission is based on regular reception of RTP packets (330; <figref idref="f0005">FIG. 5A</figref>) rather than explicit timers. This simplifies the code required and reduces retransmission overhead, but it also throttles retransmission during burst outages (e.g. wireless interference resulting in packet loss during a period). Since retransmissions only occur when RTP packets 330 are received, retransmissions are deferred beyond a possible window when packets 330 may have been lost anyway.

VII. Controlling Relative Volume at Multiple Client Devices During Playback

0059Because the disclosed system 330 plays music at multiple locations at the same time, it may be desirable to be able to adjust the volume at each location individually. The disclosed system 300 supports individual volume control by using a relative volume setting specified using a header field as part of an RTSP SET_PARAMETER request. The volume is expressed as a floating-point decibel level (e.g. 0 dB for full volume). In addition to volume, the disclosed system 330 can set other parameters related to the delivery of media at multiple locations using similar techniques. For example, the disclosed system 300 can be used to set equalization levels at each location individually.

VIII. Time Synchronization Between Host Device and Multiple Client Devices

0060Referring to <figref idref="f0008">FIG. 7</figref>, portion 300 of the disclosed system is shown having a host device 320 and multiple client devices 350 exchanging timing information. To play the same audio on the multiple client devices 350 in synchronization with each other, the timebase on the multiple client devices 350 is synchronized with a reference clock 324 on the host device 320. As noted previously, the host device 320 can be a Mac or Windows-based system running the media application 322. The host device 320 does not need to run any special server software, and only the media application 322 according to the present disclosure is required. The reference clock 324 at the host device 320 does not need to be synchronized with an external clock, such provided by an NTP server. Rather, the client devices 350 only need to be synchronized to the same reference clock 324 even if that clock 324 is wrong with respect to an external clock.

0061The reference clock 324 is maintained within the media application 322 running on the host device 320. If the host device 320 is a Macintosh computer, then the reference clock 324 can use the PowerPC timebase registers. If the host device 320 is a Windows-based computer, the reference clock 324 can use the Pentium performance counter registers. The reference clock 324 of the host's media application 322 is separate from the normal wall-clock time of the host device 320, which is maintained by an NTP agent and synchronized to an external clock. The reference clock 324 of the host's media application 322 does not need to be synchronized to an external clock and in some cases this would actually be undesirable. For example, a time difference between the reference clock 324 and the local clock of a client device 350 can be explicitly skewed or adjusted to account for spatial effects or differences, such at the client device 350 being located farther away than another. In addition, there may be situations where a user may want to intentionally skew the clocks to produce effects. Accordingly, the user interface associated with the disclosed system 300, such as interface 200 of <figref idref="f0004">FIG. 4</figref>, may include a drop-down menu or other control for intentionally manipulating skew.

0062To synchronize the timebase between the client devices 350 and the host device 320, the media application 322 uses time sync information based on the principals of the Network Time Protocol (NTP) encapsulated in Real-Time Transport Control Protocol (RTCP) packets. Preferably, NTP is not used directly to avoid collisions with existing NTP services (e.g., date/time synchronization with an external clock) and to avoid permission issues due to NTP's use of a privileged port number. Even though the time sync information of the media application 322 is encapsulated in RTCP packets, the time synchronization works substantially the same as NTP and will be referred to as NTP henceforth. NTP is known in the art and provides the basis for inter-media synchronization support in the Real-Time Transport Protocol (RTP). Details of NTP can be found in "<nplcit id="ncit0004" npl-type="s" url="http://www.ietf.org/rfc/rfc1305.txt"><text>Network Time Protocol," RFC 1305, which is available from http://www.ietf.org/rfc/rfc1305.txt</text></nplcit>.

0063Techniques of NTP, however, are preferably not used to provide moment-to-moment time directly to each client device 350 due to issues related to network latency, bandwidth consumption, and CPU resources. Accordingly, techniques of NTP are used for periodic synchronization of time. In addition, each client device 350 is provided with a high-resolution clock 364 based on the local clock hardware of each client device 350 (sec Local Clock Implementation section below), the high-resolution clocks 364 are synchronized with the reference clock 324 of the host device 320 using the NTP techniques.

0064Synchronizing the local clocks 364 of the client devices 350 with the reference clock 324 preferably does not jump to a new time with every correction (referred to as stepping) because stepping can introduce discontinuities in time and can cause time to appear to go backward, which can create havoc on processing code that relies on time. Instead, the time synchronization techniques of the present disclosure preferably correct time smoothly using clock slewing so that time advances in a linear and monotonically increasing manner. In the clock slewing techniques of the present disclosure, frequent micro-corrections, below a tolerance threshold, are performed to the running clocks 364 at the client devices 350 to bring their timebase gradually in sync with the timebase of the reference clock 324 of the host's media application 322. The clock slewing techniques also predict the relative clock skew between the local clocks 364 and the host's reference clock 324 by analyzing past history of clock offsets and disciplining the local clocks 364 to run at the same rate as the host's reference clock 324.

0065Because a centralized reference clock 324 is used for several client devices 350 on a local network, one way to disseminate time information is to send broadcast/multicast NTP packets periodically from the host device 320 to the client devices 350. Sending NTP packets by multicasting must account for losses and performance degradation that may result from the wireless 802.11b and 802.11g communication links between the host device 320 and the client devices 350. Due to issues of performance degradation, loss rates, and lack of propagation delay information associated with broadcasting or multicasting, unicast NTP transactions 400 are preferably used.

0066As part of the unicast NTP transactions 400, the client devices 350 periodically send unicast requests 410 to the host device 320 so that the client devices 350 can synchronize their clocks 364 with the reference clock 324. Then, the client devices 350 use responses 420 from the host device 320 corresponding to their requests 410 to continually track the clock offset and propagation delay between the client device 350 and host device 320 so the client devices 350 can update their local clocks 364. Thus, synchronization of the audio playback at the client devices 350 is achieved by maintaining local clocks 364 that arc synchronized to the host device's clock 324. Since all client devices participating in a particular session are synchronized to the reference clock 324. When the clocks 324 and 364 are synchronized, the client devices 350 can play audio in-sync without ever communicating with each other.

0067With the timebase at the client devices 350 synchronized with the reference clock 324 at the host device 320, the client devices 350 can use the synchronized timebase to determine when to playback packets of audio data. As noted previously, audio data is delivered to the client devices 350 using RTP packets (330; <figref idref="f0005">FIG. 5A</figref>) that contain an RTP timestamp describing the time of a packet's audio relative to other packets in the audio stream. The client device 350 uses this timestamp information to reconstruct audio at the correct presentation time for playback. Accordingly, each client device 350 correlates the NTP timebase of its local clock 364 with the RTP timestamps provided in the RTP packets of the audio stream.

0068With respect to the unicast requests and responses 410 and 420 noted above, RTP does not define a standard packet format for synchronizing time. There is an RTCP sender report, which contains some timing information, but not everything that is needed to synchronize time (e.g., there is no originate time for receivers to determine the round trip time). There are also rules preventing sender reports from being sent before any RTP data has been sent, which is critical for playing the initial audio samples in sync.

0069Therefore, the host's media application 322 preferably defines an RTP extension for an RTCP TimeSync packet for the requests and responses 410 and 420. An embodiment of an RTCP TimeSync packet 430 is shown in <figref idref="f0009">FIG. 8A</figref>. The RTCP TimeSync Packet 430 includes a header, the RTP timestamp at NTP Transmit (T3) time, NTP Originate (T1) timestamp, most significant word; NTP Originate (T1) timestamp, least significant word; NTP Receive (T2) timestamp, most significant word; NTP Receive (T2) timestamp, least significant word; NTP Transmit (T3) timestamp, most significant word; NTP Transmit (T3) timestamp, least significant word. The Marker bit (M) is not used for these TimeSync packets 430. The packet types (PT) include '210' for a clicnt dcvicc request to synchronize time in a manner similar to an NTP client device request and include '211' for a host device response to a client device request. The 'RTP Timestamp' is the RTP timestamp at the same instant as the transmit time (T3). This should be 0. The times T1-T3 come from NTP and are used in the same manner as NTP.

0070In <figref idref="f0008">FIG. 7</figref>, the RTCP TimeSync request packets 410 from the client devices 350 are sent once the RTSP RECORD command is received so that the client devices 350 can initially synchronize time. Then, the client devices 350 periodically send RTCP TimeSync request packets 410. In one embodiment, the periodic intervals for synchronizing time can be at random intervals between two and three seconds apart. The RTCP TimeSync response packets 420 are sent by the host device 320 in response to receiving a valid RTCP TimeSync request packet 410.

0071The host's media application 322 also defines an RTP extension for an RTCP TimeAnnounce packet 450. The RTCP TimeAnnounce packets 450 are sent periodically (e.g., once a second) by the host device 320 to update the client devices 350 with the current timing relationship between NTP and RTP. The RTCP TimeAnnounce packets 450 can be sent sooner if the host device 320 changes the NTP to RTP timing relationship. For example, when a new song starts, the host's media application 322 can send a new RTCP TimeAnnounce packet 450 with the marker bit (M) set to indicate that the NTP to RTP timing relationship has changed.

0072As shown in the embodiment of <figref idref="f0010">FIG. 8B</figref>, the RTCPTimeAnnounce Packet 450 includes an RTP timestamp; an NTP timestamp, high 32 bits; an NTP timestamp, low 32 bits; and an RTP timestamp when the new timeline should be applied. The Marker bit (M) is used to indicate an explicit change in the NTP to RTP timing relationship. The packet type (PT) is defined as '212' to indicate that the host device is announcing a new NTP to RTP relationship. The "RTP Timestamp" is the RTP timestamp at the same instant as the NTP timestamp. The "NTP Timestamp" is the NTP timestamp at the same instant as the RTP timestamp. The field "RTP Apply Timestamp" refers to the RTP timestamp when the new timeline should be applied.

IX. Local Clock Implementation at Host Device

0073Returning to <figref idref="f0008">FIG. 7</figref>, the local clock 364 of the client device 350 is discussed in more detail. The local clock 364 maintains a 64-bit nanoseconds counter that starts at zero on system boot and uses the 60-Hz clock interrupt to increment the nanoseconds counter. When an interrupt occurs, the 32-bit timer counter is used to determine how much time has passed since the last clock interrupt. This determined amount of time since the last clock interrupt is referred to as the tick delta and is in units of 1/100 of a microsecond. The tick delta is then converted to nanoseconds and is added to the nanoseconds counter to maintain the current time. The tick delta is used in this manner to avoid drift due to interrupt latency.

0074To maintain more accurate time, it may be preferable to allow time to be adjusted gradually. Accordingly, the nanoseconds counter is adjusted in very small increments during each clock interrupt to "slew" to the target time. These small increments are chosen based on a fraction of the amount of adjustment needed and based on the tick delta. This prevents time from appearing to go backward so that time always increases in a linear and monotonic manner.

0075Additionally, the client device 350 can predict what the next NTP clock offset will be in the future to further adjust the local clock 364. To make the prediction, the client device 350 uses a moving average of NTP clock offsets to estimate the slope of the clock skew between each of client device 350 and host device 320. This slope is then extrapolated to estimate the amount of adjustment necessary to keep the local clock 364 at the client device 350 in sync with the reference clock 324. The client device 350 then makes very small adjustments to the per-clock interrupt increment, in addition to the adjustments made for clock slewing, to simulate the faster or slower clock frequency of the host's reference clock 324. This allows the local clock 364 to remain synchronized between NTP update intervals and may even allow the reference clock 324 to remain synchronized in the absence of future NTP clock updates).

X. Simulated Timelines for Audio Playback

0076Referring to <figref idref="f0011">FIG. 9</figref>, additional details related to synchronized delivery of media with multiple client devices are discussed. In <figref idref="f0011">FIG. 9</figref>, portion of the network media delivery system 300 is again illustrated. The host device 320 is schematically shown having the media application 322 and reference clock 324, as described previously. In addition, the host device 320 is schematically shown having an engine 323, a processor 325, a transmission interface 326, and an audio interface 327. As disclosed herein, the host device 320 can be a computer. Therefore, the processor 325 can be a conventional computer processor, the transmission interface 326 can be a Wi-Fi compatible wireless network interface, and the audio interface 327 can be a sound card or the like for playing audio. In addition, the media application 322 can be a software program stored in memory on the computer and operating on the computer processor 325. Furthermore, the media application 322 can include the engine 324 for processing media (e.g., audio) data and can include the reference clock 324 for synchronizing time.

0077To play audio in a synchronized manner on multiple client devices 350 (only one of which is shown in <figref idref="f0011">FIG. 9</figref>), audio data needs to be scheduled for playback at a constant or consistent rate. One way to achieve this is for the media application 322 on the host device 320 to send packets 330 of audio data at a constant rate and to have the timeline for presenting that audio data with the client device 350 tied to the send rate of the packets 330. For example, packets of audio data can be sent about every 7.982-ms (<i>i.e</i>., 352 samples per packet / 44,100 Hz = ~7.982-ms per packet, which corresponds to a rate of about 125 packets/sec), and the timeline for presenting that audio can correspond directly to this rate. While this works, the send rate of the packets 330 and the presentation timeline at the client device 350 must have a one-to-one correspondence, which can restrict the ability to buffer the audio data at the client device 350. As discussed herein, buffering of the audio data at the client devices 350 is desirable for handling lost packets, clock skew, etc. If five seconds of buffering is desired at the client device 350, there will be a five-second delay between the time when the audio data arrives at the client device and the time when it is actually played. Unfortunately, users can readily perceive such a high level of latency when buffering is used with such a one-to-one correspondence between the packet send rate and the presentation time of the audio.

0078To provide buffering without this high level of latency, the sending of packets 330 is preferably decoupled or separated from the timeline for presenting the audio data of those packets 330. To achieve this, the media application 322 maintains two simulated timelines 328 and 329. A first packet timeline 328 corresponds to when packets 330 should be sent, and a second playback timeline 329 corresponds to when the audio data in those packets 330 should be presented or delivered (<i>i.e.,</i> played for the user). The separate timelines 328 and 329 allow the send rate of the packets 330 to vary as needed so that the system 300 can provide buffering without introducing latency. If more buffering is needed, for example, the packet send rate of the first packet timeline 328 can be temporarily increased to front-load the buffers in memory 354 on the client devices 350 and can be later reduced back to the real-time send rate of the packets 330. The separate timelines 328 and 329 also avoid problems associated with fluctuations in the presentation time of audio caused by scheduled latency of the operating systems on the devices.

0079The second playback timeline 329, which corresponds to when the audio data in the packets 330 should be presented or delivered, is constructed by the host device 320. Using the reference clock 324 and a desired playback rate of the audio, the host device 320 estimates the number of audio samples that would have played at a given point in time at the client device 350 to construct the playback timeline 329. This second playback timeline 329 is then published from the host device 320 to the client devices 350 as part of the time announcements 450 sent periodically from the host device 320 to the client devices 350. As discussed in greater detail previously, the client device 350 uses the periodic time announcements 450 to establish and maintain the relationship between the RTP timestamps in the audio packets 330 and the corresponding NTP presentation time for the audio packets 330 so that the client device 350 can deliver the audio in synch with other dcviccs.

0080By having the send rate of the packets 330 (represented by the packet timeline 328) separate from the presentation time (represented by the playback timeline 329), the periodic time announcements 450 are not designed to take effect immediately when received by the client devices 350 since the announcements 450 may come in advance of when they are effective. As noted previously, however, the time announcement packets 450 contain an additional RTP timestamp that indicates when the announced time should take effect at the client device 350. Therefore, a time announcement packet 450 is saved at a client device 350 once it is received. When audio playback reaches the RTP timestamp of that saved time announcement packet 450, the client device 350 applies the time change contained in that saved time announcement package 450.

0081To play audio in a synchronized manner on multiple client devices 350 (only one of which is shown in <figref idref="f0011">FIG. 9</figref>), it is also preferred to consider the amount of latency or delay between the time when the audio data is scheduled to be delivered at the device 350 and the time when the audio is actually delivered by the device 350 (and associated entertainment devices). Different types of client devices 350 (and associated entertainment devices) will typically have different latency characteristics. Accordingly, the disclosed system 300 preferably provides a way for each client device 350 to report its latency characteristics (and that of its associated entertainment device) to the host device 320 so that these latency characteristics can be taken into consideration when determining how to synchronize the playback of media at the client devices 350.

0082Determination of the latency characteristics of the client devices 350 preferably occurs at initial set up of the system 300. For example, the media application 322 at the host device 320 sends RTSP SETUP requests 312 to the client devices 350 at initial set up. In responses 314 to the RTSP SETUP requests 312, the client devices 350 use a header field to report the latency characteristics associated with the client devices 350. The values of the field are preferably given as the number of RTP timestamp units of latency. For example, a client device 350 having 250-ms of latency at a 44,100-Hz sample rate would report its audio-latency as 11025 RTP timestamp units. Based on the reported latency characteristics from the client devices 350, the host's media application 322 determines a maximum latency of all client devices 350 in the group being used for playback. This maximum latency is then added to the playback timeline 329.

XI. Synchronized Local Playback at Host Device

0083In addition to synchronized playback at multiple client dcviccs 350, the disclosed system 300 allows for synchronized local playback at the host device 320 running the media application 322. For example, the host device 320 can play the same audio to its local speakers (not shown) that is being played by the client devices 350, and the host device 350 can have that same audio play in sync with the all the other devices 350. To achieve this, the host device 320 uses many of the same principles as applied to the client devices 350. Rather than receiving packets of audio data over a wireless network, however, audio data is delivered directly to a local playback engine 323 of the media application 322. In addition, because local playback on the host device 320 is handled by the media application 322, there is no need for the host device 320 to synchronize time with its own reference clock 324.

0084The packets of audio data delivered to the synchronized local playback engine 323 within the media application 322 are generated before being compressed and encrypted. Since these packets do not leave media application 322, no compression or encryption is necessary. In one embodiment, the host device 320 uses CoreAudio to playback audio. CoreAudio can be used for both Mac-based or Windows-based computers because QuickTime 7 provides support for CoreAudio on Windows-based computers. During operation, an output AudioUnit is opened, and a callback is installed. The callback is called when CoreAudio needs audio data to play. When the callback is called, the media application 322 constructs the relevant audio data from the raw packets delivered to it along with the RTP->NTP timing information. Since CoreAudio has different latency characteristics than the latency characteristics associated with the client devices 350, information is also gathered about the presentation latency associated with the audio stream of CoreAudio. This information is used to delay the CoreAudio audio stream so that it plays in sync with the known latency of the audio streams associated with the client devices 350.

XII. Stutter Avoidance During Audio Playback

0085In addition to the techniques discussed previously for handling lost RTP packets of audio data and for synchronizing clocks between the host device 320 and the client devices 350, the disclosed system 300 preferably limits stuttering in the playback of media. Referring to <figref idref="f0012">FIG. 10</figref>, an algorithm 500 for limiting stutter in the playback of media is shown in flowchart form. This algorithm 500 can be performed by the host device of the disclosed system for each of the client devices. Using the algorithm 500, the disclosed system detects audible "glitches" caused by gaps in the media (e.g., audio). These gaps can be caused by loss of packets, packets arriving too late, changes to the synchronized timeline, large amounts of clock skew, or other reasons. First, the system determines the number of such "glitches" occurring in a period of time for each of the client devices (Block 502). Then, a determination is made whether the number of glitches is greater than a predetermined limit (Block 504). For example, the audio is analyzed over a period of 250-ms to determine whether the 250-ms period is either "glitching" (bad) or "glitch-free" (good). A credit system is used to make this determination. Each time a glitching period is detected, the system takes away a number of credits from a credit score of the client device. The credit score is capped at a minimum value to prevent a long sequence of glitching periods from requiring protracted period of time for the client device to recover, because the intention is to allow the client device to recover quickly as soon as its audio situation clears up.

0086If the number of credits goes below a predefined threshold at Block 504, the client device is put on probation (Block 506). When on probation, audio is disabled and silenced, but the client device can still send retransmit requests to the host device as needed to recover lost packets of audio data. The audio is silenced during probation so that the client device will not produce an annoying stutter sound when a significant number of glitching periods are successively delivered in an interval of time. Even though the audio is silenced, retransmits remain enabled so that operation of the client device can improve to a point suitable to resume playback.

0087If the number of glitches is not greater than the limit at Block 504, then the client device is set as "glitch free" (Block 505). Each time a "glitch-free" period is detected, for example, a number of credits is added to the credit score for the client device. The number of credits is capped at a maximum value to prevent a long sequence of glitch-free periods from extending the number of glitches required before going into stutter avoidance mode because the intention is to be able to go into stutter avoidance mode quickly so that there is not any significant stutter produced.

0088For the client device on probation with audio silenced and retransmits enabled, the number of glitches occurring in a predetermined unit of time (e.g., X seconds) is determined (Block 508). The number of glitches is compared to a predetermined limit or threshold (Block 510). If the client device is on probation for the predetermined unit of time (X seconds) and the number of credits reaches an upper threshold at Block 510, the client devices is placed back into normal playback mode at Block 505.

0089If the client device remains on probation for the predetermined unit of time (X seconds) and the number of credits has not reached an upper threshold at Block 510, then the client device is put in jail (Block 512). When in jail, the audio remains disabled and silenced. However, retransmits are now disabled. In this situation, the client device has not recovered for a significant period of time, and any retransmits may actually be making the situation worse. By disabling retransmits, the recovery time may be improved by reducing congestion on the network. In addition, disabling retransmits may at least reduce the amount of traffic on the network and may allow other client devices to receive packets of audio data more reliably.

0090If the client device remains in jail for a predetermined unit of time (<i>e.g.,</i> Y seconds) at Block 514, the client device goes on parole to see if its situation has improved (Block 516). When on parole, audio is still disabled and silenced. However, retransmits are re-enabled. The number of glitches occurring in a predetermined unit of time (e.g., Z seconds) is determined (Block 518) and compared to a predetermined limit (Block 520). If the client device is on parole for the predetermined unit of time and the number of credits reaches an upper threshold at Block 520, then client device returns to normal playback mode at Block 505 where audio and retransmits are both enabled. If the client device stays on parole for the predetermined unit of time and the number of credits does not reach the upper threshold at Block 520, however, the client device goes back to jail at Block 512.

XIII. Handling Address Resolution Protocol

0091With reference again to <figref idref="f0005">FIG. 5A</figref>, for example, the high volume of data being exchanged by the disclosed system 300 can cause Address Resolution Protocol (ARP) requests, which are broadcast, to become lost. This may be the case especially when the ARP requests are wirelessly broadcast. Address Resolution Protocol (ARP) is a network protocol used to map a network layer protocol address to a data link layer hardware address. For example, ARP can be used to resolve an IP address to a corresponding Ethernet address. When ARP requests are lost, ARP entries at the host device 320 can expire and can fail to be renewed during operation of the disclosed system 300 so that connections between the host device 320 and client devices 350 may appear to go down. Because steady, unicast streams 310 of packets 330 are being exchanged during operation of the disclosed system 300, one solution to this problem is to extend the expiration times of the ARP entries at the host device 320 as long as packets 330 from the host device 320 are being received by the client devices 350. By extending the expiration time, the ARP entry for a given client device 350 does not time out (as long as packets 330 are being received by that client device 350), and the client device 350 does not need to explicitly exchange ARP packets, which may tend to get lost as noted previously, with the host device 320.

0092In another solution, the client devices 350 periodically (e.g., once a minute) send unsolicited, unicast ARP request packets (not shown) to the host device 320. These unicast ARP request packets contain source addresses (Internet Protocol (IP) address and the hardware address of the client device 350) and target addresses (IP address and hardware address of the host device 320). The unicast ARP request packets are more reliable than broadcast packets because the unicast packets are acknowledged and retried at a wireless layer. To keep the ARP entries on the host device 320 for the client devices 350 from expiring, the host device 320 updates its ARP cache when it receives these unicast ARP request packets by refreshing the timeout for the corresponding ARP entries. This prevents the host device 320 from needing to issue a broadcast ARP request when the ARP entry for a client device 350 expires because the ARP entries effectively never expire as long as the client devices 350 unicast ARP request packets to the host device 320.

0093The foregoing description of preferred and other embodiments is not intended to limit or restrict the scope or applicability of the inventive concepts conceived of by the Applicants. In exchange for disclosing the inventive concepts contained herein, the Applicants desire all patent rights afforded by the appended claims. Therefore, it is intended that the appended claims include all modifications and alterations to the full extent that they come within the scope of the following claims.