ns-3 and wifi - An overview of physical layer models

Transcription

1 ns-3 and wifi - An overview of physical layer models Workshop on ns-3 in conjunction with SIMUTools 2009 March 2nd, 2009 Decentralized Systems and Network Services Research Group and Junior Research Group for Traffic Telematics Institute of Telematics University of Karlsruhe Jens Mittag

2 I. Background» Research background: Vehicular Ad-hoc NETworks: Protocol development, evaluation and optimization» Characteristics of VANETs: Very high mobility of network nodes Diverse environment urban scenarios rural scenarios highway scenarios Radio signal propagation conditions are changing rapidly over time different w.r.t. environmental effects Fully distributed communication system» Our ns-2 / ns-3 experience ns-2 PHY/MAC improvements, e.g. cumulative interference, capture capabilities or Nakagami-m distribution (2006) Port of improvements to ns-3 finished merge into main branch pending 2

3 I. Motivation» How should we model the quality of the wireless communication channel? Radio Propagation Modeling» Based on which set of rules should we decide whether a packet can be successfully decoded? Transceiver Reception Modeling 3

4 II. Wifi Architecture of ns-3 MAC MacHigh PHY Queue DcaTxop FOCUS OF THIS TALK WIRELESS CHANNEL DcfManager StationManager MacLow WifiPhy WifiChannel MacRxMiddle InterferenceHelper ErrorRateModel PropagationLossModel 4

5 III. Radio Signal Propagation» 3 different scales of signal strength variation» PathLoss: Friis Two-Ray Ground LogDistance ThreeLogDistance» Shadowing: LogNormal Shadowing» Fast fading: Nakagami-m Rician Fading Rayleigh Fading 5

6 III. Radio Signal Propagation» 3 different scales of signal strength variation» ns-3 calculates one signal strength for each packet» Principle: chaining of several propagation loss models TxPwr Friis Shadowing Nakagami-m RxPwr 6

7 III. Radio Signal Propagation» Issues with model usage Currently, (most) models are applied in a probabilistic way no correlation for receivers in a close proximity no possible correlation of successive packet receptions No consideration of scenario semantics e.g. no radio obstacles such as buildings, trucks, No consideration of signal strength variations during packet reception e.g. due to a time- and frequency-selective channel Choosing the right model and parametrization is a tough job and requires a thorough understanding of the communication system and of influencing environmental effects! 7

8 IV. Transceiver Reception Modeling» How to model the reception behavior of a transceiver? How to decide whether a packet can be successfully decoded? 1. Detection of the preamble 2. 1st decision: could the header be successfully decoded? 3. 2nd decision: could the payload be successfully decoded?» How are interfering packets and background noise modeled? Additive White Gaussian Noise Channel model 8

9 IV. Additive White Gaussian Noise Channel 9

10 IV. Additive White Gaussian Noise Channel» Reception quality of packet Ratio of Signal Strength to Noise & Interference SINR = Signal Noise + Interference 10

11 V. Reception Criterion» Bit-Error Rate based decision For each packet segment with a constant SINR compute corresponding BER Mapping Φ: SINR BER can be derived analytically or empirically for each modulation scheme (coded/uncoded) by Krishna Pillai ( Combine the BERs into a Packet Error Rate (PER) P = 1 (1 BER ) err i i L i Assumption: BitErrors are uniformly distributed and independent! 11

12 V. Reception Criterion» SINR based decision Determine the minimum experienced SINR level of a packet Compare this SINR with a threshold Thresholds are measured experimentally using real hardware e.g. 5dB for BPSK with Atheros chipsets e.g. 8dB for QPSK with Atheros chipsets 12

13 V. Reception Criterions» Capture Effect So far, synchronization to a packet is only possible when receiver is in idle state, i.e., Phy is searching for a preamble Modern chipsets support a feature called packet capturing even if receiver is already synchronized to a packet, it is able switch over to a new arriving packet SINR of new packet has to be sufficiently high capture threshold Value for capture threshold is a trade-off capture threshold too low aggressive capture policy capture threshold too high conservative capture policy 13

14 VI. Conclusion» We have different models to account for radio propagation characteristics Pathloss Shadowing Fast Fading» We have different models to reflect transceiver technology Additive White Gaussian Noise channel BER-based reception criterion SINR-based reception criterion Capture model Again, choosing the right model and the right parametrization is difficult. A wrong configuration of the wifi might lead to invalid protocol results! 14