doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 1 Project: IEEE P Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [NICT Indoor 60 GHz Channel Measurements and Analysis update] Date Submitted: [7 March 2006] Source: [Z. Krusevac, S. Krusevac, A. Gupta, T. Pollock, D. Miniutti, C. Liu, E. Skafidas] Company [National ICT Australia Limited] Address [Level 2, 216 Northbourne Ave, Canberra, ACT 2602, Australia] Voice:[ ], FAX: [ ], Source: [Hirokazu Sawada, Chang-Soon Choi, Yozo Shoji, Hiroyo Ogawa] Company [National Institute of Information and Communications Technology] Address [3-4, Hikarino-Oka, Yokosuka, Kanagawa, , Japan] Voice:[ ], FAX: [ ], Re: [Response to the TG3c channel model subgroup call for channel models] Abstract:[Analysis of the NICT data for proposed 60 GHz Channel Model ref ] Purpose:[Contribution to TG3c at March 2006 meeting in Denver, USA] Notice:This document has been prepared to assist the IEEE P It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein. Release:The contributor acknowledges and accepts that this contribution becomes the property of IEEE and may be made publicly available by P

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 2 Purpose Channel model sub-committee requires a common channel model for evaluation of PHY layer proposals Bulk of proposed applications require directional antennas Committee needs a model with AoA Require a measurement campaign to: Determine an appropriate channel model with AoA Determine the set of parameters that accurately describes the 60 GHz channel for indoor situations

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 3 Recall: Proposed Channel Model ref: S-V Model with AoA extension L = number of clusters; K l = number of multipath components (number of rays) in the l-th cluster; = multipath gain coefficient of the k-th ray in the l-th cluster; T l = arrival time of the first ray of the l-th cluster;  k,l = delay of the k-th ray within the l-th cluster relative to the first path arrival time, T l ; = mean angle of arrival of l-th cluster; = angle of arrival of the k-th ray within the l-th cluster.

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 4 Recall: Model Parameters l l cluster arrival rate l ray arrival rate cluster decay factor ray decay factor,, …multipath gain distribution factor(s) AoA distribution standard deviation

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 5 NICT AoA Channel Measurements ref: LOS Residential and NLOS Office environments Polarization: Vertical Antenna height: 1.1m Antenna separation: 3m(home), 10m(office) Tx antenna: always fixed Rx antenna: rotated from 0 to 360 degree in 5 degree steps

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 6 Equipment 3.75 MHz7.5 MHzFrequency step 128Times of average ns133.3 nsMax. excess delay No. frequency points 19.1cmDistance resolution 0.25 nsTime resolution 3 GHzBandwidth 62.5 GHzCenter frequency OfficeResidential Environment Using Vector Network Analyzer HP8510C

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 7 Residential Environment (LOS) Living room without fixtures (LOS) Surface of a wall and ceiling are covered with wallpaper Window is plane glass Wooden door and floor

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 8 Office Environment (NLOS) The office room is made of steel wall, steel ceiling and steel floor The floor and the ceiling are covered with carpet and plaster board, respectively Plate glass windows are attached on wall

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 9 Single Directional Channel Model deconvolve Rx antenna pattern from measured data –Solution to a large set of linear simultaneous equations

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 10 Cluster Identification Summary 1.PDPs are converted to a grayscale image 2.Matlab tool imtool.m is used for the picture analysis Individual pixel analysis - contrast adjustment 3.Blind deconvolution tool deconvblind.m for cleaning the picture CLEAN algorithm to uncorrelate pixels 4.Observed clusters (PDPs) then obtained from original data for more precise details and parameter extraction

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 11 Cluster Identification

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 12 Cluster Identification Identify cluster shape Pixel level indicates strength of MPC Used to determine rough cluster size/shape Indicates where in data to extract more precise cluster MPCs

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 13 MPC Selection Many possible methods: –Visual –CLEAN algorithm –Lucy algorithm –Maximum Entropy Methods –Maximum likelihood algorithms based on blind deconvolution techniques Currently using CLEAN and visual ray extraction Looking to improve using other algorithms

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 14 Residential – Parameter Extraction I Cluster decay time constant Ray decay time constant Method: Linear Regression

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 15 Residential – Cluster Decay Rate v360

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 16 Residential – Ray Decay Rate v360

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 17 Residential – Parameter Extraction II Cluster arrival rate 1/  Ray arrival rate 1/ Method: distribution fitting tool (dfittool.m) Note: arrival times from Poisson distribution (95% confidence – see Goodness of Fit testing)

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 18 Residential – Cluster Arrival Rate v360

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 19 Residential – Ray Arrival Rate v360

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 20 Distribution Extraction Multipath Gains Angle of Arrival Method: Goodness of Fit and distribution fitting tool

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 21 Goodness of Fit Summary 1.Probability Plotting –selection of the distribution hypothesis based on preliminary fitting analysis –use an inverse distribution scale so that a cumulative distribution function (CDF) plots as a straight line 2.Regression testing –to fit a straight line on the probability plot 3.Error plotting –gives the confidence level of the goodness of fit –Mean Squared error gives the error value between the data and the hypothesis distribution –Steps 1-3 done for every hypothesis distribution 4.Parameter extraction –dfittool used to extract distribution parameters

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 22 Residential – AoA Distribution

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 23 Residential – AoA Distribution

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 24 Residential – AoA Distribution

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 25 Residential – Multipath Gain Distribution No conclusive result at this time Similar in process to AoA and arrival rate distributions Further work required

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 26 Residential Summary of Parameters Extracted Mean excess delay spread [ns] RMS delay spread [ns] v v v Number of clusters [deg] [ns] [ns] v15 [ns] Tx ant. 1/  1/

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 27 Office Summary of Parameters Extracted Mean excess delay spread [ns] RMS delay spread [ns] v v30 Number of clusters [deg] [ns] [ns] [ns] Tx ant. 1/  1/ Identical analysis as for Residential environment

doc.: IEEE c Submission March, 2006 Tony Pollock, NICTASlide 28 Remaining Work Improve MPC selection Determine multipath gain distribution and parameters Determine time and angle correlation Generate model realization technique Timeline: completed before Jacksonville meeting (May)