Impulse Burst Position Detection and Channel Estimation Scheme for OFDM Systems 高永安 老師...

Impulse Burst Position Detection and Channel Estimation Scheme for OFDM Systems

高永安老師

學生：吳林達報告日期： 2006/5/18

Jukka Rinne, Ali Hazmi, Student Member, IEEE, and Markku Renfors, Senior Member, IEEE,” Impulse Burst Position Detection and Channel

Estimation Schemes for OFDM Systems,”IEEE Transactions on Consumer Electronics, Vo

l. 49, No. 3, AUGUST 2003

Outline• Introduction

• Channel estimation step

• Pre-estimation method

• MSE analysis

• Simulation environment(1)

• Channel model

• Implementation method

• Simulation environment(2)

• Simulation result

Introduction(1)• Impulse noise is a common impairment in a c

ommunications systems arising, e.g., from motors or lightning.

• Impulse noise： This model can be physically thought of as each transmitted data symbol being hit independently by an impulse with probability p and with a random amplitude g(gauss).

• OFDM is inherently more immune to this form of impairment when compared to single carrier modulation.

Introduction(2)

• The idea is to use pilots of the other (previous and future) symbols to construct reliable channel estimates

• We study the limitations and performances of those approaches in terms of MSE and BER in the case of single frequency network (SFN) hilly terrain static and mobile cases.

Channel estimation step1

• 1. Carry out the channel estimation in time domain.

• 2. Carry out channel estimation in frequency domain.

• 3. Compensate channel on data-carriers.

Channel estimation step2

• pre-estimation:

• 1. Carry out pre-estimation of pilot of burst

contaminated symbol in time or frequency domain using impulse noise free symbols.

• 2. Use conventional estimation.

• Method #1: Four Away OFDM Symbols

• Method #2: Six Pilots Spacing

• Method #3: Twelve Pilots Spacing

MSE analysis environment

• The symbol contains impulse burst

• Other symbols are assumed to be impulse free

• Mean Square Error (MSE) is calculated analytically after pre-estimation

thm

• In general MSE can be written as:

2

, ,J E H k m H k m

:

: -

:

:

:

,

E the average operation in time domain

k sub carrier index

m OFDM symbol number

H channel response

H pre estimated channel response

Here k takes pilot position values defined

by pilot pattern

• In the case of method #1,

• For the method #2, the estimate at pilots (of both

neighboring symbols) is calculated first by

• and then pre-estimation interpolation is used. In the above

equations, AWGN and ICI exist in the pilots, i.e., before pre-estimation.

, , 4 awgn ICIH k m H k m n n

, , 1 awgn ICIH k m H k m n n

• For method #3, it can be found that

• where AWGN and ICI are defined after pre-estimation

' '

1, , 1

21

, 12

awgn ICI

awgn ICI

H k m H k m n n

H k m n n

• 條件： Assume all the pilots contain the same amount of estimation errors.

• MSE for method #1：

2

1

2 2 2

, , 4

2 2 4

awgn ICI

H awgn ICI

J E H k m H k m n n

R

2 2 2

: - var

:

0 , , H awgn ICI

R autocorrelation function of time ient channel

refer to time in OFDM symbol lengths

R AWGN powers ICI powers

• for method #2, the MSE at the pilots is given by:

• The total MSE in the pilot estimates after pre-estimation can be found from:

• The optimum pre-estimation filter coefficients are given by:

' 2 2 22 2 2 1H awgn ICIJ R

H H H2 Topt opt opt opt2 -c c +c cHJ

: - cov

: - cov

cross ariance vector

auto ariance vector

-1optc =

• For method #3, the MSE at the pilots can be found from:

• The total MSE for method #3 after combining the pilots is given by:

H H H2 Topt opt opt opt-c c +c cHJ

2 '2 '23 awgn ICI

3 1 1-2 1 + R 2 + +

2 2 4HJ R

• The ICI power used in the above equations can be expressed in the case of wide sense stationary uniformly distributed uncorrelated scatters (WSSUS)

-1

2ICI 02

1

11- 2 - 2

Nu

di

TN N i J f i

N N

0

:

:

:

: d

u

N the number of carriers

J zeroth order bessel function of first kind

f Doppler frequency

T useful OFDM symbol duration

Simulation environment(1)

int : 4 :1

(30 )

uTGuard erval

Power of the channel

channel PDP is known to the receiver

Uniform PDP of given MAX length is used

Doppler spectrum and SNR dB is known to the reseiver

Doppler spectrum is defined b mod

Im inf

#2 #3: 91

: 16u

y WSSUS el

pulse are assumed to occur quite requently

Interpolator length for method and

TMax delay in channel

MSE vs. Doppler frequency for different pre-estimation methods

• Interpolator length :91 • Max delay of the channel: .16

uT 2 1H

Channel model

• Impulse noise model:• A standard Bernoulli-Gaussian process is used to

model impulsive noise• Burst length L=500 samples corresponding in 8k

mode DVB-T system to the duration of 54.5• For easier implementation, we suppose that the

impulse burst occurs exactly once in every 8th OFDM symbol.

s

• Time-Frequency Selective Channel:• SFN configuration, based on DVB-T channel model

for hilly terrain reception .• Two transmitters, both paths to the receiver are

modeled as a hilly terrain, static and mobile reception. • Received powers from transmitter were defined

relatively to the most powerful signal (and first to arrive).

• A value of 0 dB for both paths has been chosen. • The channel model is mainly parameterized by the

delay and the Doppler frequency when we consider an SFN mobile channel.

s df

Implementation method(1)

• Four Away OFDM Symbols (FOA)

• 適用條件： impulse bursts occur quite rarely.

• In the case of static single frequency network (SFN), this method is very effective and simple to implement.

Implementation method(2)

• Six Pilot Spacing (SPS)

• To tolerate larger Doppler frequencies

• An interpolator designed for a delay spread of 1024 samples is used to interpolate the channel coefficients.

Implementation method(3)

• Twelve Pilot Spacing (TPS)

• An interpolator designed for a delay spread of 512 samples is used based on the twelve pilot spacing pattern.

Simulation environment(2)

• 8k DVB-T system model• Sub-modulation:64QAM• Code rate:1/2 or 2/3• Impulse noise burst length :• L=500samples(54.5 )• Interpolator length:125• Max delay spread = respectively for me

thod2 and method3• SNR=30dB

s

u uT T or 8 16s

BER performance for three methodin SFN static channel(1)

512 s samples

BER performance for three methodin SFN static channel(2)

1024 s samples

BER performance for three methodin SFN static channel(3)

2048 s samples

BER performance for three methodin SFN mobile channel(1)

code rate:1/2 64QAM-OFDM

BER performance for three methodin SFN mobile channel(2)

code rate:2/3 64QAM-OFDM

Tolerability of the three methods in Doppler frequency and delay spread

cases