September 2000 doc.: IEEE 802.11-00/282r2 Evaluating the Performance of HRb Proposals in the Presence of Multipath Steve Halford, Karen Halford, and Mark Webster Intersil Corporation September, 2000 ** **With assistance from Chris Heegard of Texas Instruments Note: This power point documents contains notes. Set view to Notes pages to see the notes. Submissio n Slide 1 S. Halford, K. Halford, and M. Webster September 2000 doc.: IEEE 802.11-00/282r2 Goals Multipath is recognized as major WLAN impairment To select best waveform, must include multipath performance Multipath Model was left as TBD by teleconference Want a model close to 802.11b model Want a model well-defined Compare proposal against the same measure Cross-validate multipath performance numbers Want a model that is fair to all proposals Want a model that reflects real radio conditions as much as possible Submissio n Slide 2 S. Halford, K. Halford, and M. Webster September 2000

Overview doc.: IEEE 802.11-00/282r2 Multipath Models for WLAN Motivation Exponential Channel Model (IEEE 802.11b model) Truncation to FIR model Sample Rate Normalization Rayleigh Fading Model AWGN with multipath Use of Channel Model : Block Diagrams Non-normalized Normalized Summary of Proposal Sample Code Submissio n Slide 3 S. Halford, K. Halford, and M. Webster September 2000 doc.: IEEE 802.11-00/282r2 Multipath in WLAN reflector receiver transmitter transmitted signal Channel Model reflector received signal Submissio

n Slide 4 time S. Halford, K. Halford, and M. Webster September 2000 doc.: IEEE 802.11-00/282r2 Exponential Model Used by Task Group b for 802.11 (see docs 97/96, 97/125, 97/157r1) Taps are independent complex gaussian variables with average power profile that decays exponentially 1 2 1 2 hk N (0, k ) j N (0, k ) 2 2 kmax 10 rms /Ts where Truncate to FIR e Ts / rms k2 o2 k o2 1 1 kmax 1 Submissio n for k 0,1,..., k max Slide 5 Normalize Average Power to 1 S. Halford, K. Halford, and M. Webster September 2000

doc.: IEEE 802.11-00/282r2 Exponential Channel Model Average Power Profile Sample Realization 0.7 0.4 0.6 0.35 0.5 0.3 0.4 0.25 0.2 0.3 0.15 0.2 0.1 0.1 0 0.05 Ts 2Ts 3Ts 4Ts 5Ts 6Ts 7Ts 8Ts 9Ts 10Ts 11Ts time Submissio n Slide 6 0 Ts 2Ts 3Ts 4Ts 5Ts 6Ts 7Ts 8Ts 9Ts 10Ts 11Ts time

S. Halford, K. Halford, and M. Webster September 2000 doc.: IEEE 802.11-00/282r2 Tap Truncation Truncate to represent with an FIR model. Value of last tap in truncated exponential channel: e kmax Ts / rms e 10 4.5 10 5 Exponential channel is monotonically decreasing Therefore, remaining unmodeled taps e 10 Unmodeled taps are insignificant. Submissio n Slide 7 S. Halford, K. Halford, and M. Webster September 2000 doc.: IEEE 802.11-00/282r2 Sample Rate and Exp Model Sample rate determines resolution of taps 0.7 1 11 MHz Example 0.9

44 MHz Example 0.6 0.8 0.5 0.7 Theoretical Delay Spread: 30 nsec Actual Delay Spread: 21 nsec 0.6 0.5 Theoretical Delay Spread: 30 nsec Actual Delay Spread: 29 nsec 0.4 0.3 0.4 0.3 0.2 0.2 0.1 0.1 0 0 50 100 150 200 nsec 250 300 350 400 0

0 50 100 150 200 nsec 250 300 350 50 45 Only a problem at low sample rate and low multipath delay. Actual Delay Spread 40 35 Theoretical Delay Spread 30 Sample Rate = 11 MHz 25 Sample Rate = 22 MHz 20 Sample Rate = 44 MHz 15 Sample Rate = 88 MHz 10 5 0 0

5 10 15 20 25 30 35 40 45 50 Theoretical Delay Spread Submissio n Slide 8 S. Halford, K. Halford, and M. Webster September 2000 doc.: IEEE 802.11-00/282r2 Normalization: Flat Fading & ISI Channel model is normalized in an expected value sense 1 1 Recall: hk N 0, k2 j N 0, k2 2 2 ( kmax 1)Ts / RMS kmax kmax 2 kmax 2 kTs / RMS 2 2 1 e E hk k 0 e 0 Ts / RMS

1 e k 0 k 0 k 0 1 e Ts / RMS 2 Select 0 to make average power = 1 ( kmax 1)Ts / RMS 1 e This normalization assumes an finite number of taps (unlike 802.11b model) Channel Model includes flat fading & Intersymbol Interference Power Gain of Channel (sum of taps squared) 2.4 Power varies on a per trial basis Average gain is one 2.2 2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 Submissio n Slide 9 0 500 1000

1500 2000 2500 3000 Trial Number 3500 4000 4500 5000 S. Halford, K. Halford, and M. Webster September 2000 doc.: IEEE 802.11-00/282r2 Flat Fading & ISI: Block Diagram Multipath results should include Eb/N0 Setting Desired Eb Channel Sample rate Delay spread Packet Length (1000 bytes) Packet Error Rate N0 Calculate Noise Power (N0) Generate Noise Measure energy per bit Measure Packet Error Rate Transmitter Model Exponential

Channel Model Packet Length Data Rate Sample Rate Delay Spread Submissio n Receiver Model Slide 10 Packet Error Rate S. Halford, K. Halford, and M. Webster September 2000 doc.: IEEE 802.11-00/282r2 Normalization: ISI Only Want to evaluate the proposals for ISI robustness w/o flat fading Need to normalize the relative gain (loss) of exp channel Care must be taken to ensure fair application of model Potential Approach K max h If we force each channel realization to unit gain k 2 1 k 0 Problem: Inconsistent results across channel sample rates Normalization applies to entire bandwidthnot signal bandwidth 3

Example 2.5 Sample Rate = 88 MHz Normalized power of each realization Power Gain for 22 MHz signal Shows the power variation Penalizes samples rates >> bandwidth Submissio n 2 1.5 1 0.5 0 Slide 11 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 S. Halford, K. Halford, and M. Webster September 2000

doc.: IEEE 802.11-00/282r2 ISI Only: Suggest Approach For consistency -- Need to normalize at output of the channel Normalizes the signal bandwidth rather than entire channel bandwidth Desired Eb N0 Calculate Noise Power (N0) Generate Noise Measure energy per bit Measure Packet Error Rate Transmitter Model Packet Length Data Rate Submissio n Exponential Channel Model Receiver Model Packet Error Rate Sample Rate Delay Spread Slide 12 S. Halford, K. Halford, and M. Webster September 2000 doc.: IEEE 802.11-00/282r2

Rayleigh Fading Classic model for multipath components with delay much less than sample rate Amplitude has a Rayleigh Distribution with uniform random phase Memoryless -- affects all signal frequencies the same (flat fade) For convenience, can consider to be a limiting case of exponential channel Single tap channel with 0 RMS delay spread fix kmax equal to one Single tap will scale and rotate the received signal affect all frequencies in the same way since it is a multiplication not a convolution hRayleigh 1 2 1 2 h0 N (0, 0 ) j N (0, 0 ) 2 2 2 0 1 Submissio n Slide 13 S. Halford, K. Halford, and M. Webster September 2000 doc.: IEEE 802.11-00/282r2 Multipath with AWGN Q: Does HRb include additive noise in multipath comparisons? Additive noise can have a major impact on multipath performance Example: Zero-forcing & MMSE equalizer have same performance w/o noise Performance can be vastly different in presence of noise Realistic to include both impairments simultaneously Sweep Packet Error Rates across a range of noise & multipath values A: Yes. Submissio n Slide 14 S. Halford, K. Halford, and M. Webster

September 2000 doc.: IEEE 802.11-00/282r2 Summary of Proposal Propose using the current exponential channel model Identical to IEEE802.11b Truncate using kmax 10 rms /Ts Sample rate used to generate channel should always be given Normalization: Include both normalized and un-normalized results Normalization must be done at output of the channel Rayleigh fading included as special case of exp model Showed suggested block diagrams Recommend using PER with 1000 byte packets Include noise with multipath Vary levels of both noise and multipath Cross-Verification: Include description of equalizer type (not required to give design details) Submissio n Slide 15 S. Halford, K. Halford, and M. Webster September 2000 doc.: IEEE 802.11-00/282r2 Matlab Code for Exponential Channel %**************************************************************** % function [taps] = ExpChanTaps(sampRateMHz, delaySprdNsec) % % Return the FIR channel taps for the exponential channel model % for indoor multipath. % % INPUT PARAMETERS: % sampRateMHz = sampling rate in MHz % (reciprocal of tap spacing in usec) % delaySprdNsec = delay spread in nsec (0 generates Rayleigh

% % OUTPUT VALUE: % taps = complex channel taps for the exponential % channel model % *************************************************************** % % % % % % % % % % % % % % % function [taps] = ExpChanTaps(sampRateMHz, delaySprdNsec) sampTimeNsec = 1000 / sampRateMHz; if delaySprdNsec == 0 Kmax = 0; vark = 1; else Kmax = ceil( 10 * delaySprdNsec/sampTimeNsec ); var0 = (1 - exp( -1*(sampTimeNsec)/delaySprdNsec ))/ ... (1 - exp( -1*((Kmax+1)* sampTimeNsec)/delaySprdNsec )); k = (0:Kmax)'; vark = var0 * exp( - k*sampTimeNsec/delaySprdNsec ); end stdDevReOrIm = sqrt(vark/2); taps = stdDevReOrIm .* (randn(Kmax+1,1) + j*randn(Kmax+1,1)); Submissio n Slide 16 S. Halford, K. Halford, and M. Webster