l09 Phy Simulations

download l09 Phy Simulations

of 12

Transcript of l09 Phy Simulations

  • 8/3/2019 l09 Phy Simulations

    1/12

  • 8/3/2019 l09 Phy Simulations

    2/12

    Simulation hierarchySimulation hierarchy

    Networks

    Links

    DSP Circuits RF

    Event driven simulations:

    ns2, Opnet

    Time driven simulations:

    SPW, Cossap, Simulink/Matlab

    Algorithm simulations:

    TI CodeComposer

    Packets, messages, flows

    Waveforms

    Circuit simulations:

    NC-{VHDL/Verilog}, Scirroco,

    RF simulations:

    PSpice, ADS,XFDTD

    Technology

  • 8/3/2019 l09 Phy Simulations

    3/12

    Waveform Level SimulationsWaveform Level Simulations

    Usually used when analytical evaluationUsually used when analytical evaluation

    of performance is difficult (of performance is difficult (nonlinearities, ISInonlinearities, ISI

    caused by bandlimiting filterscaused by bandlimiting filters))

    Typically:Typically:1.1. Generate sampled values of the inputGenerate sampled values of the input

    waveforms (process)waveforms (process)

    2.2. Process them through system models andProcess them through system models and

    generate outputgenerate output

    3.3. Estimate the performance by comparing inputsEstimate the performance by comparing inputs

    and outputsand outputs

  • 8/3/2019 l09 Phy Simulations

    4/12

    MethodologyMethodology

    Ideally model is a perfect replica ofIdeally model is a perfect replica of

    the real system hard to dothe real system hard to do

    Instead we introduce approximationsInstead we introduce approximations

    to reduce complexity or run-time:to reduce complexity or run-time:

    Modeling level simplification of theModeling level simplification of the

    specific functionsspecific functions

    Performance evaluation level Performance evaluation level

    estimation of performance measuresestimation of performance measures

  • 8/3/2019 l09 Phy Simulations

    5/12

    Methodology (cont.)Methodology (cont.)

    Modeling:Modeling: System ModelingSystem Modeling - highest level of- highest level of

    description; complexity reductiondescription; complexity reduction

    Device ModelingDevice Modeling block or subsystem (e.g. block or subsystem (e.g.transfer function on every clock cycle: "input-transfer function on every clock cycle: "input-transfer-output)transfer-output)

    Random Process Modeling:Random Process Modeling:

    Source random process (imitated withSource random process (imitated withpseudo random number generator RNG)pseudo random number generator RNG) Time-variant random channelTime-variant random channel Equivalent random process (ERP)Equivalent random process (ERP)

  • 8/3/2019 l09 Phy Simulations

    6/12

    Methodology (cont.)Methodology (cont.)

    Monte Carlo simulation as the nameMonte Carlo simulation as the nameimplies relates to game of chanceimplies relates to game of chance

    Input signals are assumed to be randomInput signals are assumed to be random

    processesprocesses

    Objective is to find statistical properties ofObjective is to find statistical properties of

    )(tV Model ofCommunication System

    )(tU

    )(tY

    )(tW

    )(tY

    If we do time evolution of all the waveforms -pure Monte Carlo simulation Generating sampled values of all the input processes

  • 8/3/2019 l09 Phy Simulations

    7/12

    Methodology (cont.)Methodology (cont.)

    =

    =

    N

    i

    iYgN

    tYgE1

    ^

    ))((1

    ))((

    Procedure: Generate sampled values of the inputs

    (e.g. bit sequence {U(k)}, k=1,2,,N and noise {V(j)}, j=1,2,,mN)

    Process samples through the model andgenarate Y(k) (received bits):

    Estimate the performance by counting errors

    In general find expected value of E{g(Y(t)}

    from the simulation according to:

  • 8/3/2019 l09 Phy Simulations

    8/12

    Methodology (cont.)Methodology (cont.)

    For our example: whereFor our example: where

    If only some input processes areIf only some input processes are

    simulated explicitly partial MCsimulated explicitly partial MC(quasianalytical simulation)(quasianalytical simulation)

    Random number generation is essentialRandom number generation is essentialfor MC simulationsfor MC simulations

    Requires RNG generation methods fromRequires RNG generation methods froma wide variety of distributions and witha wide variety of distributions and witharbitrary autocorrelation (PSD).arbitrary autocorrelation (PSD).

    =

    =N

    k

    kYgN

    Pe1

    ^))((1

    ==

    )()(0)()(1))((kUkYkUkYkYg

  • 8/3/2019 l09 Phy Simulations

    9/12

    RNGRNG

    Important properties:Important properties:

    AlgebraicAlgebraicStructure (uncorrelated samples)Structure (uncorrelated samples)

    PeriodPeriod

    StatisticalStatisticalDistributionDistribution

    Uniform RNGUniform RNG Congruent or the power residue methodCongruent or the power residue method

    MckaXkX mod])1([)( +=

  • 8/3/2019 l09 Phy Simulations

    10/12

    RNG (cont)RNG (cont)

    wherewhere M>0 large (prime) integer - modulusM>0 large (prime) integer - modulus

    0

  • 8/3/2019 l09 Phy Simulations

    11/12

    RNG (cont)RNG (cont)

    Few good menFew good men(for 32 bit machines)(for 32 bit machines)

    For longer periodsFor longer periods Wichman-Hill Algorithm combines 3 RNGs:Wichman-Hill Algorithm combines 3 RNGs:

    periodperiod

    )2mod(]1)1(069.69[)(

    )12mod()]1(807.16[)(

    32

    31

    +=

    =

    kXkX

    kXkX

    1mod30269

    )(

    30307

    )(

    30269

    )()(

    30323mod)]1(170[)(

    30307mod)]1(172[)(

    30269mod)]1(171[)(

    ++=

    =

    =

    =

    nZnYnXnU

    nZnZ

    nYnY

    nXnX

    12107x

  • 8/3/2019 l09 Phy Simulations

    12/12

    RNG (cont)RNG (cont)