Orthogonal Frequency Division Multiplexing Part 2 • Multipath channel • Multipath time variant channel • Multipath channel with frequency offset Algoritmo per la stima di canale in multipath channel S. Jain, P. Gupta and D. K. Mehra, “EM-MMSE based Channel Estimation for OFDM Systems,” in Proc. IEEE, I.C.I.T., 2006, Mumbai, pp.2598-2602 • Modello di riferimento • Uso dei toni pilota • Algoritmo EM-MMSE: – The (Expectation Maximization) EM è una forma di maximum likelihood (ML) detection con parametri non noti. – E-STEP: Calcola il valore atteso dei dati inviati usando le stime correnti dei parametri e dei dati osservati. – M-STEP: usa le stime dei dati inviati per determinare il parametro che massimizza la probabilità di aver osservato quei dati inviati • E step: dati trasmessi con MMSE usando – simboli ricevuti – stima del canale alla p-th iterazione • M step: canale • 4-tap multipath channel, QPSK, N=128 Algoritmi per la stima di canale in multipath time variant channel Y. Mostofi and D. C. Cox, “ICI Mitigation for Pilot-Aided OFDM Mobile Systems,” IEEE Transaction on Wireless Comminication, vol.4, no. 2, March 2005 • Modello di riferimento • Uso dei toni pilota ICI • Approssimazione lineare a tratti • Estrazione toni pilota valori medi • Metodo iterativo No – Stima di X – Calcolo pendenze con prefisso ciclico Inizializzazione Estrazione toni pilota Calcolo X Calcolo pendenze Converge? Sì • Approssimazione lineare a tratti (2 pendenze) • Equazione risolutiva • Estrazione toni pilota valori medi • Calcolo pendenze con simboli precedente e successivo • Canale – Profilo di canale (dB): [-3.01 -8.24 -13.01 -5.23] – Velocità utente: 100 Km/h • Sistema WiMax Metodo I Metodo II Algoritmi per la stima di canale in multipath channel with frequency offset L. Favalli, P. Savazzi, A. Vizziello, “Frequency Domain estimation and Compensation of Intercarrier Interference in OFDM Systems,” proc. Of the 10th International Symposium on Spread Spectrum Techniques and Application, Bologna, August 2008. L. Favalli, P. Savazzi, A. Vizziello, “Estimation and Mitigation of Intercarrier Interference for OFDM Systems in Multipath Fading Channels,” proc. Of the 4th IEEE International Conference on Wireless and Mobile Computing, Networking and Communications, Avignon, October 2008. • Proposed methods: –Pilot tones –Algorithm I: AWGN channel with frequency offset • Estimate ICI coefficients: CORRELATION • Recover data: MMSE –Algorithm II: extension to multipath channel Radio access architecture – Modulation/demoduation: IFFT/FFT – Time variant multipath propagation: channel h(t) – AWGN noise – Frequency offset at local oscillator • Relationship in the frequency domain • Channel matrix includes ICI effects l,m accounts for the interference of the m-th subcarrier onto the l-th one • Several factors simultaneosly • Large size of matrix some simplifications: – Frequency offset or one time variant path is circular – Mutipath and frequency offset is split in two parts: one circular and another diagonal – AWGN channel and frequency offset: • can be represented as follows: the first row the other rows (circular shift) • Compute only the coefficients of the first row • Further simplification: The loss of orthogonality interests a limited number of adiacent subcarriers Consider only the first I coefficients of each row, with I<<N – Multipath channel and frequency offset: • can be represented as follows: • C takes into account frequency offset effect: circular • H is the channel frequency response: diagonal • AWGN channel and frequency offset • Iterative Method • Equations in the frequency domain to – Channel Estimation – Symbol Estimation R. R. Lopes and J. R. Barry, “Blind Iterative Channel Identification and Equalization,” IEEE International Conference on Comminication, vol.7, pp. 2256-2260, June 2001 • First row • d is the position of interfering subcarrier with respect to the considered k • I is the maximum number of interfering subcarriers – Equation: • Frequency domain • Multicarrier system • Recover ICI channel matrix – Proposed Method: • Time domain • Single carrier system • Recover channel impulse response to reduce ISI – Derived from Lopes and Barry1: Channel Estimation through CORRELATION 1 Symbol Estimation via MMSE – Complex coefficients obtained by MMSE equalization • is supposed known at the receiver is the transpost conjugate – Coefficients multiply received vector • Proposed Scheme • After FFT • Channel Estimation: Received Data and Estimated Transmitted Data are used • Symbol Estimation: Received Data and Estimated Channel are used • Multipath channel and frequency offset • Matrix Approximation: • Compute matrix H • Iterative Method – Calculate matrix C: CORRELATION – Symbol Estimation: MMSE Pilot tones equally spaced on subcarrier2 Compute matrix H: • through an IFFT of length • Diagonal elements of H on these positions • Taps coefficients Y. Mostofi and D. C. Cox, “ICI Mitigation for Pilot-Aided OFDM Mobile Systems,” IEEE Transaction on Wireless Comminication, vol.4, no. 2, March 2005 • H in frequency domain 2 • Compute matrix C by CORRELATION • Calculate matrix • Data Recovery by MMSE The algorithm is summarized in the following table: 5 4 3 2 1 Estimate Data by MMSE Calculate Estimate C by CORRELATION Estimate H Extact pilot tones Initialization: set to 0 estimate of Operation 6 If Method does not converge go to step 4 Step 7 Expectation Maximization framework: – E step – M step General EM Scheme Application to the Proposed Algorithm – Compute matrix H – E step: Recover Data by MMSE – M step: Calculate ICI matrix Re-compute MMSE equalizer coefficients Two sets of simulation for Method I: – AWGN channel with carrier frequency offset • Normalized frequency offset – One tap time-variant channel • Channel condition of the compared method • Time varying channel with Jakes model • Speed = 80 km/h fD = 371 Hz fD Norm = 3.8% • Increasing the number of pilot tones the performance improve BER varing pilot tones AWGN channel with offset 0.2 • Pilot tones to initialize the algorithm and improve convergence speed ICI coefficients convergence • After a dozen iterations there is a perfect match with the target value BER for different offset values • • 2 or 3 estimated ICI coefficients Larger coefficients are better recovered BER in one-tap time variant channel • • 2 estimated ICI coefficients Compared with pre-DFT algorithm to recover ICI Set of simulation for Method II: – Multipath channel: two taps with power -0.4576 dB, -5.2288 dB – Normalized frequency offset: 0.1 and 0.2 Constant multipath Slow time variant multipath