660 likes | 672 Views
This overview discusses the implementation of adaptive playout for real-time audio applications, focusing on buffering and addressing issues such as packet ordering, packet loss, and jitter. Various methods for estimating network delay and adjusting audio playout delay are explored.
E N D
You are Here Encoder Decoder Middlebox Receiver Sender Network
How to recv and play? open socket while not done if socket is readable read packet from socket remove RTP header decode play back
What’s Wrong? • packet ordering • packet loss • next packet arrive in-time? Especially bad for audio applications
Overview Decode RTP Classifier Network
Implementation • Single Thread • using select() • Multi-Threads
Packet Buffer • Sorted by sequence number • When ADU is complete, send to decoder RTP Classifier
Playout Buffer • Stored decoded data in playout order • Post-processing/Mixing may happens Decode
Sending Packets Packet Time
Receiving Packets Packet Time
With Jitter Packet Time
With Jitter Packet Time
What causes Jitter? • Network delay = Transmission Delay (fixed) + Propagation Delay (fixed) + Queuing Delay (variable) • Delay jitter is caused by variable queuing delay
Delay Jitter Transit Time small jitter large jitter Time
Spike Transit Time Time
Today’s Question • How big is the playout buffer? • When to playback?
Types of Applications • Non-interactive • Buffer can be large • Interactive • As small as possible
Types of Applications • Video • Frames are discrete (easier problem) • Audio • Samples are “continuous”
Naive Answer • How big is a buffer? • Fixed at a small value, to reduce latency • When to playback? • Playback as soon as possible, to reduce latency
Audio Conferencing • Live, interactive application • Latency is important • Each packet 20-30ms of audio
Silence Suppression • Silence Detection • if no sound, no need to send • Talk spurt • consecutive audio packets (between silence) • hundreds of ms
Recall: RTP Header • marker bit: • depends on payload • e.g. beginning of frame
RTP and Talkspurt • First packet of a talkspurt will have marker bit set to 1
RTP and Talkspurt • Deduce talkspurt from sequence number and timestamp SeqNo TimeStamp 1 20 2 40 3 60 5 190
Consequences of Talkspurt • Opportunity to adjust playout delay • if jitter is large, increase delay • if jitter is small, decrease delay
Fixed Playout Delay SEND RECV PLAY
Adaptive Playout Delay SEND RECV PLAY
Adaptive Playout Delay SEND RECV PLAY
Trade-Off Latency vs. Packet Loss
Latency vs Loss-Rate Loss Rate Latency
Adaptive Playout Mechanisms for Packetized Audio Applications in WAN R. Ramjee, J. Kurose, D. Towsley, H. Schulzrinne INFOCOM 1995
Variables and Notations Tnet(i) Tbuf(i) Tarrive(i) Tplay(i) Tdelay(i) Tsend(i)
1st Packet in Talkspurt We can estimate as
How to estimate Enet(i) • Method 1: Jacobson’s Method
Spike Tnet Time
Problems • Does not react to spike fast enough
How to estimate Enet(i) • Ramjee’s Method SPIKE NORMAL
Ramjee’s Idea if Tnet(i) suddenly increase SPIKE NORMAL if “slope” is small enough
In Spike Mode SPIKE Tnet i
In Normal Mode NORMAL Tnet i
Evaluations Loss Rate Delay
Problems with Ramjee’s Method Transit Time Time
Packet Audio Playout Delay Adjustment: Performance Bounds and Algorithms S. Moon, J Kurose, D. Towsley Multimedia Systems 1998
How to Set Tdelay(i) • Moon’s Method • Collect statistics on packets that have arrived. • Find t such that q% of last w packets have Tnet(i) < t.