Progress Report in
Progress Report in
DVB-T’s Research
DVB-T’s Research
Speaker: Keng-Hsien Lin
Speaker: Keng-Hsien Lin
Advisor: Prof. Tzi-Dar Chiueh
Advisor: Prof. Tzi-Dar Chiueh
Date: December 27, 2004
Outline
Outline
• Soft Demapping for Soft Viterbi DecodingSoft Demapping for Soft Viterbi Decoding • Multipath Channel ModelingMultipath Channel Modeling
• Propose Two-stage Mode Detection AlgorithmPropose Two-stage Mode Detection Algorithm • Conclusions Conclusions && Future Work Future Work && References References
Outline Overview
Outline Overview
• Soft Demapping for Soft Viterbi DecodingSoft Demapping for Soft Viterbi Decoding
– Bit Interleaved Coded ModulationBit Interleaved Coded Modulation – Bit Log-likelihood RatioBit Log-likelihood Ratio
– Soft Demapping Approximation in ML SenseSoft Demapping Approximation in ML Sense
• Multipath Channel ModelingMultipath Channel Modeling
• Propose Two-stage Mode Detection AlgorithmPropose Two-stage Mode Detection Algorithm • Conclusions Conclusions && Future Work Future Work && References References
Bit Interleaved Coded Modulation
Bit Interleaved Coded Modulation
• Interleaver
– The performance is better over Rayleigh fading – Needs appropriate soft-decision metrics
• Soft bit metrics for soft-decision decoding
[1] Encoder Interleaver Modulation
Channel
Bit Log-likelihood Ratio
Bit Log-likelihood Ratio
• Bit log-likelihood ratio
, Pr( | ) ln 0 Pr( | ) j dj i d j j ij r c r c
Pr( | 0) ln Pr( |1) j j j r r , (( 1)cdj ( 1)cij ) 0 j j i d j
(( 1)cdj ) (( 1)cij ) j j j j
• ML criterionSoft Demapping Approximation
Soft Demapping Approximation
in ML Sense (1/5)
in ML Sense (1/5)
• Relation between received and transmitted signals yk, xk 2 : 0 2 : 1 max exp( | ( ) | ) ln max exp( | ( ) | ) j j k k b b j k k b b y h f b y h f b ( ) k k k y h f b w : : 2 0 0 2 : 1 : 1 Pr( | ) exp( | ( ) | ) Pr( | 0) ln ln ln Pr( | 1) Pr( | ) exp( | ( ) | ) j j j j b b b b k k k k j j k j k k k b b b b y b y h f b y b y b y b y h f b
k k y s h ( ) k x f b • Calculate bit log-likelihood ratio2 2 2
: 1 : 0
Soft Demapping Approximation
Soft Demapping Approximation
in ML Sense (2/5)
in ML Sense (2/5)
2 2 2 2
0 | k | (| Re( ) 1|k | Re( ) 1| ) |k k | (4 Re( ))k
b metric h s s h s
2 2 2 2 2
0 | k | (min(| Re( ) 1| ,| Re( ) 3 | ) min(| Re( ) 1| ,| Re( ) 3 | ))k k k k
b metric h s s s s
2 2 2 2 2
2 | k | (min(| Re( ) 1| ,| Re( ) 1| ) min(| Re( ) 3 | ,| Re( ) 3 | ))k k k k
b metric h s s s s
• For QPSK using Gray mapping
• For 16-QAM using Gray mapping
2 2 2 2
1 | k | (| Im( ) 1|k | Im( ) 1| ) |k k | (4 Im( ))k
b metric h s s h s
2 2 2 2 2
1 | k | (min(| Im( ) 1| ,| Im( ) 3 | ) min(| Im( ) 1| ,| Im( ) 3 | ))k k k k
Soft Demapping Approximation
Soft Demapping Approximation
in ML Sense (3/5)
in ML Sense (3/5)
Soft Demapping Approximation
Soft Demapping Approximation
in ML Sense (4/5)
in ML Sense (4/5)
Soft Demapping Approximation
Soft Demapping Approximation
in ML Sense (5/5)
in ML Sense (5/5)
Outline Overview
Outline Overview
• Soft Demapping for Soft Viterbi DecodingSoft Demapping for Soft Viterbi Decoding
• Multipath Channel ModelingMultipath Channel Modeling
– Simple Case AnalysisSimple Case Analysis
– Frequency Responses of Several Type InterpolatorsFrequency Responses of Several Type Interpolators – Circular Concepts in OFDM SystemCircular Concepts in OFDM System
• Circular ConvolutionCircular Convolution
• Propose Circular InterpolationPropose Circular Interpolation
• Propose Two-stage Mode Detection AlgorithmPropose Two-stage Mode Detection Algorithm • Conclusions Conclusions && Future Work Future Work && References References
Simple Case Analysis
Simple Case Analysis
[ ] [0] [ ] [0.8] [ 0.8] y n h x n h x n ( 0.8) ( jw) [0] ( jw) [0.8] ( jw) j w ( jw) ( jw) Y e h X e h X e e H e X e ( 0.8) ( jw) [0] [0.8] j w [ ] jwn n H e h h e
h n e 2 1 [ ] [0] [ ] [0.8]( [ ] [ ]) k y n h x n h i k x n k
( jw) [0] ( jw) [0.8] ( jw) ( jw) ( jw) ( jw) Y e h X e h I e X e H e X e ( ) ( jw) [0] [0.8] | ( jw) | j I ejw H e h h I e e | (I ejw) | 1 e j I e ( jw) e j( 0.8) w • Good interpolatorFrequency Responses of
Frequency Responses of
Several Type Interpolators (1/5)
Several Type Interpolators (1/5)
Frequency Responses of
Frequency Responses of
Several Type Interpolators (2/5)
Several Type Interpolators (2/5)
Frequency Responses of
Frequency Responses of
Several Type Interpolators (3/5)
Several Type Interpolators (3/5)
Frequency Responses of
Frequency Responses of
Several Type Interpolators (4/5)
Several Type Interpolators (4/5)
Frequency Responses of
Frequency Responses of
Several Type Interpolators (5/5)
Several Type Interpolators (5/5)
Circular Convolution
Circular Convolution
1 mod 0 [ ] [ ] [ ] N N m y n h m x n m
( )
( ) ( )
Y k
H k X k
• Circular convolution • OFDM system– Uses cyclic prefix (CP)
– Only needs one-tap equalizer
2 1 0 ( ) ( ) nk N j N n Y k y n e
Propose Circular Interpolation
Propose Circular Interpolation
2 1 [ ] [0] [ ] [0.8]( [ ] [ ]) k y n h x n h i k x n k Y e( jw) h[0] (X ejw) h[0.8] (I ejw) (X ejw) mod mod 2 1 [ ] [0] [ ] [0.8]( [ ] [ ] ) N N k y n h x n h i k x n k
• Circular interpolation in OFDM system
( )
[0] ( )
[0.8] ( ) ( )
Y k
h
X k
h
I k X k
Outline Overview
Outline Overview
• Soft Demapping for Soft Viterbi DecodingSoft Demapping for Soft Viterbi Decoding
• Multipath Channel ModelingMultipath Channel Modeling
• Propose Two-stage Mode Detection AlgorithmPropose Two-stage Mode Detection Algorithm
– Why Needs Mode DetectionWhy Needs Mode Detection
– Critical Issues in Mode Detection of DVB-T SystemCritical Issues in Mode Detection of DVB-T System – Two-stage Mode Detection AlgorithmTwo-stage Mode Detection Algorithm
• FFT Mode DetectionFFT Mode Detection
• Guard Interval Mode DetectionGuard Interval Mode Detection
Why Needs Mode Detection
Why Needs Mode Detection
• DVB-T system
– Two FFT size to select: 2K-mode, 8K-mode
– Four Guard Interval to select: 1/4, 1/8, 1/16, 1/32
• Severe frequency-selective fading channel
– Select 8K-mode and 1/4 guard interval ratio
• Severe fast fading channel
– Select 2K-mode and 1/32 guard interval ratio
Critical Issues in Mode Detection
Critical Issues in Mode Detection
of DVB-T System
of DVB-T System
• Mode detection should be finished in time domain
– No preamble
– Known signals are put in frequency domain – Only CP characteristic can be used
• Total eight kinds of mode need to detect one
– ML or MMSE criterion may not work
FFT Mode Detection (1/5)
FFT Mode Detection (1/5)
1 * 0 ( , , ) GI ( ) ( ) k d N GI r d k r d k N 63 * 2 0 1 | ( , 2 , 2 ) | | ( ) ( 2 ) | 32 K k d K K r d k r d k K 63 * 8 0 1 | ( ,8 , 2 ) | | ( ) ( 8 ) | 32 K k d K K r d k r d k K • Delay correlation • FFT mode detection • Criterion CP CPEither one will have peaks!
arg max ( )
FFT Mode Detection (2/5)
FFT Mode Detection (2/5)
FFT Mode Detection (3/5)
FFT Mode Detection (3/5)
FFT Mode Detection (4/5)
FFT Mode Detection (4/5)
FFT Mode Detection (5/5)
FFT Mode Detection (5/5)
Guard Interval Mode Detection (1/4)
Guard Interval Mode Detection (1/4)
1 32 * 0 1 | ( , , ) | | ( ) ( ) | 32 N k d N N r d k r d k N
• Using the lease guard interval length to do delay correlation
• Theoretical analysis
| |
Different platform width
• Detection algorithms
– Platform width detection
Guard Interval Mode Detection (2/4)
Guard Interval Mode Detection (2/4)
• Platform width detection in 2K mode
– The best case: Through Tx to Rx
Guard Interval Mode Detection (3/4)
Guard Interval Mode Detection (3/4)
• Platform width detection in 2K mode
– The worst case: P1, SNR=3dB, 20ppm CFO&TO > 1/2 Peak > 3/4 Peak
Guard Interval Mode Detection (4/4)
Guard Interval Mode Detection (4/4)
1 32 * 0 1 | ( , , ) | | ( ) ( ) | 32 N k d N N r d k r d k N 3 1/ 4 0 5 1 max( | ( ( ), , ) |) 4 32 GR l d l N N N 1/8 3 0 9 1 max( | ( ( ), , ) |) 8 32 GR l d l N N N
• Different accumulated window detection
3 1/16 0 17 1 max( | ( ( ), , ) |) 16 32 GR l d l N N N 1/32 3 0 33 1 max( | ( ( ), , ) |) 32 32 GR l d l N N N 1/ 4 1/8 1/16 1/ 32 arg max GR i i or or or Guard Ratio
Outline Overview
Outline Overview
• Soft Demapping for Soft Viterbi DecodingSoft Demapping for Soft Viterbi Decoding • Multipath Channel ModelingMultipath Channel Modeling
• Propose Two-stage Mode Detection AlgorithmPropose Two-stage Mode Detection Algorithm
• Conclusions Conclusions && Future Work Future Work && References References
– ConclusionsConclusions – Future WorkFuture Work – ReferencesReferences
Conclusions
Conclusions
• Both analytical and simulation results are shown
– Soft demapping method
• Use piecewise linear approximation in ML sense
– Multipath channel modeling
• Use circular interpolation with good interpolator
– Mode detection algorithm
Future Work
Future Work
• Finish the simulation on mode detection error rate
• Transform the DVB-T code to silicon IP
– Synthesizable RTL code – Testbench
References
References
• [1] G. Caire, G. TariccoE. Biglieri, “Bit-Interleaved Coded Modulation”, IEEE Trans. Inform. Theory, vol.44, May. 1998