Chapter 4 Architecture Design and Circuit Implementation
4.4 Interference Shaping Filter Block
The interference shaping filter contains constant complex coefficients and a set of accumulating register. From the simulation results presented in Section 3.6.1 we realize that the improvement achieved by ICI cancellation saturates when W 2 , that is we should only consider the ICI from nearest four subcarriers and set W 2. We approximate Doppler spread and CCI noise by multiplying the precomputed coefficients Sb w , which are described in (3.29), with the variation of noise-free received signal reconstructed by STBC re-encoder. We apply CSD coding on the coefficients S to minimize the number of addition needed. As a b result, the complex multiplications involved inside the interference shaping filter are implemented using only several adders, as illustrated in Fig.4.6 (a) and (b). The shift registers accumulate the ICI noises from 2W neighboring subcarriers and the CCI noise from the same subcarrier. When the input is the reconstructed interference-free received signal of subcarrier f , the output of interference shaping filter is the total interference noise at received subcarrier
2 f .
64
Fig.4.6 Design of shaping filter circuit design of (a) the first transmit antenna branch and (b) the second transmit antenna branch
65
4.5 Simulation Results
The performances of the proposed STBC interference canceller are demonstrated through the simulation of an STBC-OFDM system with the two stage DFT-based channel estimator introduced in Section 3.2. The STBC decoding flow is defined in Section 2.4. The multipath channel adopt the ITU Veh-A channel model with relative path power profiles of 0, -1, -9, -10, -15, and -20 (dB), and the path excess delays are uniformly distributed from 0 to 50 sampling periods. Jakes model is also used to generate Raleigh fading environment. Fig.3.12 – Fig.3.15 shows the BER versus Eb/No performances of different decoding scheme under QPSK modulation or 16QAM modulation at vehicle speed of 360 km/hr or 240 km/hr. The results of perfect channel estimation and perfect ICI model (W=2) are included as benchmarks, denoted as “P2”. The maximum Doppler f is 555.6 Hz and the normalized Doppler is 0.05 for vehicle d speed 240 km/hr, similarly fd 833.4 Hzand fd f 0.075 for vehicle speed 360 km/hr.
“T2” denotes the floating point performance of proposed decoding scheme. The hardware version performance of “T2” scheme is simulated with fixed-point word length and denoted as
“F2”. The curves of proposed scheme and the hardware version are very close. The detail of each “T2” and “ P2” scheme are described in Table 3.3.
As shown in Fig.4.7, in 16QAM modulation and vehicle speed 240 km/hr, the curve of the hardware version has about 0.5 dB gap in Eb/No as compared with proposed scheme and 1 dB gap as compared with perfect ICI model case at BER=5 10 3. As shown in Fig.4.8, in 16QAM modulation and vehicle speed 360 km/hr, the curve of the hardware version has about 0.7 dB gap in Eb/No as compared with proposed scheme and 2.5 dB gap as compared with perfect ICI model case at BER=3 10 2. As shown in Fig.4.7, in QPSK modulation and vehicle speed 240 km/hr, the curve of the hardware version has about 0.5 dB gap in Eb/No as compared with proposed scheme and 1 dB gap as compared with perfect ICI model case at BER=103. As shown in Fig.4.8, in QPSK modulation and vehicle speed 360 km/hr, the curve of the hardware version has about 0.7 dB gap in Eb/No as compared with proposed scheme and 1 dB gap as compared with perfect ICI model case at BER=5 10 3.
66
Fig.4.7 BER performance versus Eb/No at vehicle speed 240 km/hr
Fig.4.8 BER performance versus Eb/No at vehicle speed 360 km/hr
67
We use the improved performance ratios to demonstrated the improvement of proposed STBC interference canceller, which are given by
BER of 0 Improved Performance Ratio of 2
BER of 2 BER of 0 Improved Performance Ratio of 2
BER of 2 BER of 0 Improved Performance Ratio of 2
BER of 2
Fig.4.9 shows the improved performance of 16QAM at vehicle speed 240 km/hr. Nearly 5 time improvement can be achieved if perfect channel estimation is applied to our proposed decoding scheme. Our proposed interference canceller and two-stage channel estimator implementation provides 3.9 and 2.2 time improvement in Eb/No of 15 and 30 dB, respectively. Fig.4.10 shows the improved performance of 16QAM at vehicle speed 360 km/hr. 2.8 time improvement can be achieved if perfect channel estimation is applied to our proposed decoding scheme. Our proposed interference canceller and two-stage channel estimator implementation provides 2.4 and 1.9 time improvement in Eb/No of 15 and 30 dB, respectively.
Fig.4.11 and Fig.4.12 shows the improved performance of QPSK modulation at vehicle speed of 240 km/hr and 360 km/hr, respectively. The improvements of our proposed implementation do not exceed 2 times. This is because QPSK modulation has better resolution and is more robust compared to 16QAM modulation, the effect of STBC interference noses are not significant. As a result, the modeling error of our proposed algorithm counteracts the cancelled interference noise and results in limited performance improvement.
68
Fig.4.9 Improved performance ratio of 16QAM at vehicle speed 240 km/hr
Fig.4.10 Improved performance ratio of 16QAM at vehicle speed 360 km/hr
0 3 6 9 12 15 18 21 24 27 30
0 2 4 6 8 10 12 14
Eb/No (dB)
Improved Performnce Ratio
I2 P2 F2
2 times improvement
0 3 6 9 12 15 18 21 24 27 30
1 2 3 4
Eb/No (dB)
Improved Performnce Ratio
I2 P2 F2
2 times improvement
69
Fig.4.11 Improved performance ratio of QPSK at vehicle speed 240 km/hr
Fig.4.12 Improved performance ratio of QPSK at vehicle speed 360 km/hr
0 3 6 9 12 15 18 21 24 27 30
1 2
Eb/No (dB)
Improved Performnce Ratio
I2 P2 F2
2 times improvement
0 3 6 9 12 15 18 21 24 27 30
1 2
Eb/No (dB)
Improved Performnce Ratio
I2 P2 F2
2 times improvement
70