Overview on the Parallel Iterative Multi-layered Detection Method

4.3.2 Overview on the Parallel Iterative Multi-layered

Detection Method

The parallel multi-layered detection method exploits the soft information instead of hard decision results to implement adaptive MMSE equalization, MPIC and LAIC, thus the information of layered detection and decoding can be passed to each other and exchanged between layers adequately. For the reason of saving process time, we adopt parallel interference cancellation architecture in our proposed method.

The scheme diagram for the first iteration is shown in Fig.4.6. After OFDM demodulation, the baseband signal at jth receiver, r , passes through MMSE equalizer to suppress ^j layered-antenna interference and noise. After being despread, scrambled and summed up for all L receivers to combine receiver diversity, the signal is delivered to the BICM decoder block.

In the decoder block, the signal is demapped to the information of three coded bits firstly,

Fig.4.6: The first iteration of the multilayered detection method.

and the bit deinterleavers are used to return the information to the encoding order. After iteratively decoding and demapping, the a-posteriori information of coded bits will be obtained from the output of the decoder and be exploited to generate the soft symbol.

For the following iteration, the soft symbols of the previous iterations are utilized to reconstruct and cancel LAI. The soft symbol can also be exploited to adjust the factors of adaptive MMSE equalizer and calculate the statistics of signals to help demap. The block diagram is illustrated in Fig. 4.7.

Through several iterations, the soft symbols become more and more reliable, we can replace the MMSE equalizer by the MPIC scheme shown Fig. 3.3. After LAIC, MPIC exploits soft symbols to reconstruct and delete MPI, and matches each path individually to suppress intercode interference (ICI) and combine multipath diversity. The block diagram of the iteration with MPIC is shown in Fig. 4.8. The details of adaptive MMSE equalizer, MPIC, symbol demapper and soft symbol block are described as follows:

A. Adaptive MMSE equalizer

For the first iteration, because the soft information of the transmit symbols is not available, we use the factor derived in Eq.(3.1) to equalize r . Thus the signal before demapping in ^j Fig.4.6 can be expressed as

Fig. 4.7: The second iteration of the multilayered detection method.

Fig. 4.8: The MPIC iteration of the multilayered detection method.

( )

Layered Antenna Interference Enhanced Noise

M L K M N

Assuming transmitted symbols are independent to each other, z can be model as a _u^Q complex Gaussian distribution from the central limit theorem. It consists of four components:

the desired signal (DS), the inter-code interference (ICI), the layered antenna interference (LAI) and the enhanced noise (EN).The mean of the distribution comes from the DS term, and the variance comes from the other terms. Here we use h_eq to represent the equivalent channel gain for the desired signal. The means of the real part Re{ }z_u^Q and the imaginary

, ,

induced by the frequency selective channel effect, thus the variance of ICI is

, 2

= −

∑

, the equivalent channel gain of the ith subcarrier which deletes the common part for all N subcarriers, E denotes the power of 8-PSK symbol. _av

The variance of LAI and EN are, respectively, calculated as

, , 2

Afterward the mean and variance of z are applied in the demapper block. _u^Q

For the second and following iteration, the soft symbols ˆd^l are supplied, and LAIC can be applied. At the left of Fig. 4.7, the receive signal which has deleted LAI can be presented as

, , term is produced by LAI reconstruction block in the top of Fig. 4.7. Then the adaptive equalizer parameter σ² in Eq.(4.22) is adjusted to

where [ ]E • is the average operator for the transmitted symbol. Because the soft symbol d^ˆ_k^l is the mean of the transmitted symbol d and the mean of the 8-PSK symbol power is 1, it _k^l can be rewritten as

2 2 , 2 2 2

where the only difference with Eq.(4.21) is the residual LAI term which is induced by the

Then the mean and variance of ˆz are computed and delivered to the demapper block. _u^Q

B. MPIC

The MPIC scheme is applied in the condition that the soft symbol is accurate enough to cancel the multipath interference. As the LAI reduces gradually, the soft information becomes more and more reliable and the accuracy of soft symbol is improved, and we start to apply

( )

MPIC scheme which is illustrated in Fig. 3.3. The iteration of multi-layered detection method with MPIC is shown in Fig. 4.8.

In the same way mentioned in Section 3.2, the MPI-cancelled signal for the qth path can be derived as

Desired Signal Residual Multipath Interference

( ˆ )

where the diagonal matrix of multipath channel gain ^{, ,}

1

a complex amplitude multiplied by a phase which shifts with the subcarrier frequency. When

, j Q

rq is matched by the complex conjugate of H_q^j,Q, chips on all subcarriers will suffer the same equivalent channel gain |a_p^{j Q,} |² , and the ICI that induced by the loss of code orthogonality will be suppressed. After maximum ratio combining for all paths, the signal is despread and descrambled. The signal after combining for all receivers can be expressed as

( )

Desired Signal Residual Multipath Interference

( ) ( )

where the residual multipath interference (RMPI) term which induced by incorrect soft symbols replace the ICI term in Eq.(4.32). We can obtain the means of z _u^Q

The variance of RMPI, RLAI and EN are, respectively,

These parameters are exploited to help demap.

C. Demapper and soft symbol

The damapper block which translates symbol to bit information is applied after all receive signals are combined. The statistic parameters of the demapping signal, mean and variance, are sent to the demapper block and substituted into Eq.(4.10). We can compute the extrinsic LLR of the ithbit of the symbol for the lth layer as

where i= ∼0 2, w is the symbol included in Ω or ₀ⁱ Ω , ₁ⁱ h and _eq σ_z² change with different processes ( MMSE or MPIC ). Notice that the a-priori information Λ^l_a(V )_t^i' is equal to zero for the first demapping.

After demapping for all layers, Λ^l_ex(V )_tⁱ is dinterleaved to return to the coding order, and the soft-in soft-out decoder described in Section 4.1.2.B is used to output the extrinsic information of coded bit Λ_ex(C )ⁱ_t . Λ_ex(C )ⁱ_t is interleaved and taken as Λ^l_a(V )_tⁱ of the demapper. After several iteration of demapping and decoding, the final a-posteriori LLR

i

(C )t

Λ is passed to the soft symbol block.

2 z

=Var[ { }] Var[ { }]= (Var[RMPI]+Var[RLAI]+Var[EN])1 2

The soft symbol is the expectation of the transmitted symbol. The probability of the transmitted symbol is computed from the a-posteriori information of coded bits Λ(C )ⁱ_t by

{ }

and we can obtain the soft symbol by

3 i i soft symbols are passed to the next iteration for LAIC or MPIC.

在文檔中 應用於多輸入/輸出多載波分碼多重進接系統的遞迴式多層級偵測方法之研究 (頁 54-67)