BCWS oct19th06 partc


Presentation Description

No description available.


Presentation Transcript

WINNER2 WG1 presentation Usage of correlation properties and model evolution in WINNER related MIMO radio-channel models. : 

WINNER2 WG1 presentation Usage of correlation properties and model evolution in WINNER related MIMO radio-channel models.


Content Introduction Large-Scale Control Parameters Correlations Link-level System-level Drop concept Model evolution in time Conclusion Discussion


Introduction MIMO (Multiple Input Multiple Output) antenna systems are offering high throughputs necessary for future communication services (e.g. multimedia …) Reliable models are necessary for evaluation of a system performance. WINNER MIMO targeted model was supposed to provide reliable tool for estimation of system performance, covering frequencies up to 5 GHz and bandwidths of 100 MHz in different types of environment. Generic spatial-channel-model; large-scale (e.g. pathloss) as well as small-scale (e.g fast fading) effects of MIMO radio-channel Parameterization is based upon the real channel-sounding measurements; Measurements are conducted in real environments, at targeted frequency ranges (5GHz) by using the same system bandwidth that will be modelled (100 MHz).

Introduction (cont.): 

Introduction (cont.) Modeling phases: initial: 3GPP SCM (Spatial Channel Model) Spatial dimension of models is necessary to support MIMO concepts Extension of SCM (SCME) WINNER model Methodology Sum-of-sinusoids used to generate CIR between all Tx-Rx antenna element pairs. Stochastically controlled spatial channel model parameters of each multipath-component (MPC) are related to virtual spatial propagation, but they are genetated randomly from appropriate probability distributions

Introduction (cont.): 

Introduction (cont.) MPC parameters: departure (from Tx) and arriving (to Rx) angles, propagation delay and power Evolution of these parameters can not be based on ray-tracing since positions of scattering clusters are not known. MPC are clustered according to observations from measured data

Large Scale Parameters (cont.): 

Large Scale Parameters (cont.) WINNER LSP: Delay spread (DS), Angular spread (AS) of departure and arrival angles in azimuth or elevation, Shadow fading. Deviation of measured data from expected pathloss Spreads are roots of statistical second order central moments, where pdf is estimated from Power-Delay-Profile (PDP) or Power-Angular-Spectrum (PAS).

Large Scale Parameters: 

Large Scale Parameters Due to non-stationary behavior of radio-channel distributions of MPC (low-level) parameters are changing in time. probability distributions are parameterized and their changes are modelled through the change of control parameters. Since these parameters are controlling probability distributions of other (low-level) parameters in WINNER terminology they are called Large-Scale-Parameters (LSPs). Large-scale control parameters by them self also represent random numbers, described by probability distributions. Distributions of LSP are determined from measurements.

Correlations at Link-Level (single spatial position) : 

Correlations at Link-Level (single spatial position) Link describes radio connection between two radio-stations being at certain positions. Each link is governed by its own set of Large Scale Control Parameters Since LSPs are estimated from marginal power distributions (independently for angles and delays) their probability distributions are also independent. Marginal distributions are used since for joint distributions more data would be required. However, chosen Large Scale Control Parameters (although being quite intuitive) are NOT INDEPENDENT in reality e.g. increased DS measure is usually accompanied by increased AS Correlation coefficient (CorrC) describe inter-dependence of Large-Scale control Parameters (LSP) over distance. Cxy if cross-covariance of stochastic processes x and y. for d=0: correlation of LSP are related to single spatial position.

Correlation coefficient at System Level: 

Correlation coefficient at System Level Figures: correlation coefficient for DS as a function of distance, for B3 NLOS (indoor hotspot) and C1 LOS (suburban macro cell). Decorrelation distance (CorrC=1/e) is scenario dependant.

Cross-Correlations at System-Level (cont.): 

Cross-Correlations at System-Level (cont.) Correlation coefficients of LSP are defining changes with distance (similarity of environment in space-time dimension): Movement of single MS or MSs at different positions. Multidimensional problem: correlation of multiple LSPs at different space-time positions. WINNER approximation: Calculate how auto-correlation coefficients are changing with distance. Use distance dependant auto-correlation coefficients to approximate changes of cross-correlation coefficients with distance For this purpose scaling of auto-correlation functions with cross-correlation function at zero distance is used Square root of matrix (cross-correlations values at zero distance) is calculated trough eigen-decomposition MODELLING: relative distance between MSs being connected to the same BS is used to introduce scenario-specific correlation of LSPs  for MSs. Full correlation for links between one MS and multiple sectors of the same BS (since we have same MS: d=0). Correlations of link LSP of single MS toward multiple base-stations is open issue. Complex measurement set-up is necessary. Not many results in literature.

Polarization dependence : 

Polarization dependence Correlations between co- and cross-polarized antenna elements (especially of interest for small portable terminals supporting MIMO capability.) Example (W. Kotterman at al.) for OFDM system dependance of correlation parameters over frequency may be more important then for other systems

Drop concept: 

Drop concept Channel segment (drop) represent period of quasi-stationarity in witch probability distributions of low-level parameters are not changed. Corresponds to area where radio-environment (scatterers) of MS is not changed significantly. During this period all large-scale control parameters, as well as velocity and direction-of-travel for mobile station (MS), are held constant in SCM/WINNER models. This reduces necessity to update delays and angles of MPCs in each simulation step. Simulated channel impulse response is not constant: Fast fading during one drop is modeled through shifts in Doppler phases (caused by motion of MS)

Model evolution in time: 

Model evolution in time In SCM/WINNER models simulation are based on consecutive INDEPENDENT drops low-level (MPC) parameters (delays and departure/arrival angles) are generated independently between consecutive drops. Insight to channels realizations not being properly ordered in time. abrupt changes could influence performance of time-dependant algorithms. SCM extension (SCME) has allowed simultaneous drifting of arriving angles and delays for every MPC at each simulation step inside drop. Drifting of angles/delays was based on randomly generated distance from Tx/Rx antennas to the closest scatters. In respect to this property SCME can be classified as model with continuous evolution (in discrete steps, being smaller than drop). Consequence was substantial increase of complexity and simulation time length in comparison to SCM/WINNER


Conclusion Main issue in the system-level modelling is to find simple enough mechanisms to mimic real properties of channel. Drop concept reduces simulation complexity. Sacrificed smooth time-behaviour property – only statistically equivalent to real channel Future WINNER models would probably keep drop concept in order to keep simulation complexity low. Chosen control parameter set (LSP) influences further modelling solutions. LSP cross-correlations link-level are providing simpler analysis tool if compared to joint distribution estimation. Correlations of LSP over distance are introduced to reestablish space-time dependance being present in measured channels. Different system may depend on different properties of channel

authorStream Live Help