6.5 Implementation and Optimization
6.5.7 Data Channel Response Calculate
After we get the Wiener filter coefficients, we use them to calculate the data subcarrier channel responses. For optimization, Figure 6.19 shows the original code. We can employ the method used in Section 6.5.1 to replace the PilotPosition. Another optimization is
Figure 6.19: C code for data subcarrier response calculation before modification.
replacing the pointer. Because we save the Wiener filter coefficients in an array, when we call them for multiplications, we have to use an array pointer to obtain their values, which result in multilevel data loading and cost large amount of cycles. So we first place the Wiener filter coefficients into temporary registers as shown in Figure 6.20 and use the temporary registers for multiplication. Table 6.5 shows the cycles and efficiency evaluation of all the modifications for data subcarrier response calculation, the field names Rounding means we consider the rounding effect in our program, we can see it would increase additional cycles.
The average efficiency for fifteen data subcarrier can be calculated by : Symbols × SNR × PRU × multiplications = 28 × 11 × 48 × 6 = 88704. Efficiency is 88704/143896.8 = 61.64%.
We can see good performance from the result.
Figure 6.20: C code for data subcarrier response calculation after modification.
Table 6.5: Data Subcarrier Channel Response Calculation Clock Cycles Comparison
Method Carrier index (0~17)
0
1 425236 165841 166629 166265
2 426580 167244 165088 172172
3 426662 166937 164780 166529
4 426580 168685 163758 163720
5 428428 152192 112763 132963
6 425093 103918 106876 135828
7 423856 152460 107986 142604
8
9 423707 152460 113960 135828
10 426844 152460 113960 136064
11 424082 152460 113960 135828
12 425275 152460 106876 135828
13 426272 152460 105028 132992
14 424580 149819 104798 132913
15 423620 149720 104296 133090
16
17 424202 152460 113960 135828
Average 425401.133 152771.733 124314.533 143896.8
Efficiency 20.85% 58.06% 71.35% 61.64%
6.5.8 Summary
According to the above discussion, we reduce the total clock cycle time and code size suc-cessfully. Finally, DSP C6000 compiler compile our C code and packages to a out file, which size is 273 KB. Table 6.6 shows the clock cycles for all C code functions. Incl.Total cy-cle means total number of cycy-cles for all executions including function calls. Subfunction channel estimation f ixed includes all the functions described in Figure 6.8. And subfunc-tion interpolasubfunc-tion includes all the funcsubfunc-tion described in Figure 6.8 except pilot subcar-rier channel estimation. Subfunction main deals with SNR and symbol number setting.
P arameterT able record the pilot value. Because of we do not write all the C code as func-tions, so we just approximate the proportion for total cycles. Table 6.7 shows the proportion for total cycles without considering TI library.
The original C code without any optimization needs 13340267 clock cycles for calculation, after our optimized methods as describes in this chapter, which just need 9587041 cycles.
We reduced to approximate 72% cycles of the original program. We try to find the required time per OFDMA symbol, the required cycles are total cycles counts × clock cycle time ÷ (SNR × Symbols) = 9587041 × 10−9 ÷ (11 × 28) = 31.13(µs), and OFDMA symbol period are (FFT size + CP length ÷ sampling frequency = 1152 ÷ (11.2 × 106) = 102.86(µs). In the other word, we only use 30% symbol period for channel estimation, which remains 70%
symbol period for the other use, just like channel coding, synchronization, etc.
Table 6.6: Downlink Channel Estimation Clock Cycle Table from the Function Aspect
Symbol Name Symbol Type Access Count cycle.Total: Incl. Total cycle.Total: Excl. Total
Levinson_Durbin function 8932 5056491 3592638 37.47%
interpolation function 308 9115806 3364697 35.10%
_divi <TI Library> function 56634 1206137 1206137 12.58%
memcpy <TI Library> function 11764 707276 707276 7.38%
Sin_Fix function 5202 315037 315037 3.29%
pilot_extraction function 308 104926 97842 1.02%
Cos_Fix function 2601 241925 84410 0.88%
_divu <TI Library> function 2396 53733 53733 0.56%
main function 1 26976923 41298 0.43%
Atan_Fix function 305 40096 31169 0.33%
channel_estimation_fixed function 341 9262628 29917 0.31%
DIV_fixed function 272 32249 23817 0.25%
_remi <TI Library> function 1232 20798 20798 0.22%
Sqrt_Fix function 350 13696 13696 0.14%
SqrtTable function 1 2556 2556 0.03%
AtanTable function 1 1159 1159 0.01%
SinTable function 1 861 861 0.01%
Total required cycles 9587041
Table 6.7: Downlink Channel Estimation Clock Cycle Proportional Distributed
Function name Percentage
Pilot subcarrier channel estimation 1.69%
Time domain interpolation of pilot channel response 4.66%
Calculation of R0and R1 3.22%
Rmstemp and angle calculation 1.11%
Correlation function calculate 7.19%
Wiener coefficient calculation 29.04%
Data channel response calculate 46.64%
Remains 6.45
Chapter 7
Fixed-Point Simulation Results
7.1 DL Simulation Results
7.1.1 Validation with AWGN Channel
We verify the correctness of the program code by simulation AWGN channel transmission.
Both SISO and SFBC MIMO transmission are consider. We run 105 symbols in each simula-tion to obtain the numerical results. Figs. 7.1 shows the DL channel estimasimula-tion performance in SISO transmission for AWGN channel. In our results, the MSE performance for fixed-point implementation is very close to floating-fixed-point result. Although we set a lower bound to noise variance, the error floor phenomenon is not obviously for MSE in AWGN channel.
7.1.2 SISO Transmission Results
We conduct simulation with the six SUI channels to examine the channel estimation perfor-mance. Figure 7.2 and Figure 7.3 shows the simulation results in SUI-2 channel and SUI-5 channel, respectively. In our result, because of the precise problem, the MSE performance of channel estimation different velocities become hardly distinguishably. When SNR small, the MSE curve for fixed-point is very close to floating-point, but if the SNR value is bigger than 12 dB, because we set a lower bound to avoid singular, the error floor would occurrence.
Another numerical results, please refer to Appendix E.
7.1.3 SFBC Transmission Results
SFBC is a technique that uses multiple antennas to achieve better diversity effect. Figure 7.5 and Figure 7.6 shows the simulation results in SUI-2 channel and SUI-5 channel, respectively.
In our simulation, we use two transmit antennas and one receive antenna. We estimate the two channel responses separately. So the MSE performance is not different by using SFBC as compared to SISO, but the SER performance is better. To check with our results, the fixed-point MSE curve is very close to floating point and which has similar SER performance to floating-point in AWGN channel. Another numerical results, please refer to Appendix F.
7.2 UL Simulation Results
7.2.1 Validation with AWGN Channel
We verify the correctness of the program code by simulation AWGN channel transmission.
Both SISO and SFBC MIMO transmission are consider. We run 105 symbols in each sim-ulation to obtain the numerical results. Figure 7.7 and Figure 7.8 shows the UL channel estimation performance for each frequency partition (FP). For downlink system, we use 48 PRUs for average to do the LMMSE channel estimation, but in uplink, we only consider the CRUs for each frequency partition. For FP0 which uses 6 PRUs for average to calculate LMMSE channel estimation. For FP1, FP2 and FP3 which use 8 PRUs for average. However the average numbers are different for each frequency partition, but we can not consider the performance of FP0 would worse than other frequency partitions, we can only confirm the standard deviation of mean delay for FP0 would higher than other frequency partitions. To check with our results, fixed-point performance is very close to floating-point.
7.2.2 SISO Transmission Result
Figures 7.9 to 7.12 shows the simulation results in SUI-2 and SUI-5 channel, respectively.
In our uplink results, however we set different noise variance lower bound (8.622 dB) with downlink (9.714 dB), the error floor would appear at 12 dB similarly. Another numerical results, please refer to Appendix G.
7.2.3 SFBC Transmission Result
Figures 7.15 to 7.18 shows the simulation results in SUI-2 and SUI-5 channel, respectively.
In our simulation, we use two transmit antennas and one receive antenna. We estimate the two channel responses separately. So the MSE performance is not different by using SFBC as compared to SISO, but the SER performance is better. Another numerical results, please refer to Appendix H.
0 5 10 15 20 25 30 10−4
10−3 10−2 10−1 100
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in AWGN
Es/N0(dB)
MSE
Floating point Fixed point
0 5 10 15 20 25 30
10−5 10−4 10−3 10−2 10−1 100
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in AWGN
Es/N0(dB)
SER
Floating point Fixed point Theory
Figure 7.1: Fixed-point channel estimation MSE and SER for QPSK in AWGN for IEEE 802.16m downlink.
0 5 10 15 20 25 30
DSP Implementation for IEEE 802.16M Downlink Channel Estimation in SUI−2
Es/N0(dB)
MSE
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
0 5 10 15 20 25 30
10−3 10−2 10−1 100
DSP Implementation for IEEE 802.16M Downlink Channel Estimation in SUI−2
Es/N0(dB)
SER
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
Figure 7.2: Fixed-point channel estimation MSE and SER for QPSK at different velocities in SUI-2 channel for IEEE 802.16m downlink, where the speed v is km/h.
0 5 10 15 20 25 30 10−2
10−1 100
DSP Implementation for IEEE 802.16M Downlink Channel Estimation in SUI−5
Es/N0(dB)
MSE
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
0 5 10 15 20 25 30
10−1 100
DSP Implementation for IEEE 802.16M Downlink Channel Estimation in SUI−5
Es/N0(dB)
SER
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
Figure 7.3: Fixed-point channel estimation MSE and SER for QPSK at different velocities in SUI-5 channel for IEEE 802.16m downlink, where the speed v is km/h.
0 5 10 15 20 25 30 10−4
10−3 10−2 10−1 100
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in AWGN with SFBC
Es/N0(dB)
MSE
Antenna 1 (Fixed−point) Antenna 2 (Fixed−point) Antenna 1 (Floating−point) Antenna 2 (Floating−point)
0 5 10 15 20 25 30
10−3 10−2 10−1 100
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in AWGN with SFBC
Es/N0(dB)
SER
Floating−point Fixed−point Theory
Figure 7.4: Fixed-point channel estimation MSE and SER for QPSK with SFBC in AWGN channel for IEEE 802.16m downlink.
0 5 10 15 20 25 30
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in SUI−2 with SFBC
Es/N0(dB) MSE v=30 with Antenna 1 (Fixed−point)
v=30 with Antenna 2 (Fixed−point) v=30 with Antenna 1 (Floating−point) v=30 with Antenna 2 (Floating−point) v=60 with Antenna 1 (Fixed−point) v=60 with Antenna 2 (Fixed−point) v=60 with Antenna 1 (Floating−point) v=60 with Antenna 2 (Floating−point) v=90 with Antenna 1 (Fixed−point) v=90 with Antenna 2 (Fixed−point) v=90 with Antenna 1 (Floating−point) v=90 with Antenna 2 (Floating−point)
0 5 10 15 20 25 30
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in SUI−2 with SFBC
Es/N0(dB)
Figure 7.5: Fixed-point channel estimation MSE and SER for QPSK with SFBC at different velocities in SUI-2 channel for IEEE 802.16m downlink, where the speed v is km/h.
0 5 10 15 20 25 30
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in SUI−5 with SFBC
Es/N0(dB)
MSE
v=30 with Antenna 1 (Fixed−point) v=30 with Antenna 2 (Fixed−point) v=30 with Antenna 1 (Floating−point) v=30 with Antenna 2 (Floating−point) v=60 with Antenna 1 (Fixed−point) v=60 with Antenna 2 (Fixed−point) v=60 with Antenna 1 (Floating−point) v=60 with Antenna 2 (Floating−point) v=90 with Antenna 1 (Fixed−point) v=90 with Antenna 2 (Fixed−point) v=90 with Antenna 1 (Floating−point) v=90 with Antenna 2 (Floating−point)
0 5 10 15 20 25 30
10−2 10−1 100
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in SUI−5 with SFBC
Es/N0(dB)
Figure 7.6: Fixed-point channel estimation MSE and SER for QPSK with SFBC at different velocities in SUI-5 channel for IEEE 802.16m downlink, where the speed v is km/h.
0 5 10 15 20 25 30
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in AWGN for FP0
Es/N0(dB)
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in AWGN for FP1
Es/N0(dB)
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in AWGN for FP2
Es/N0(dB)
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in AWGN for FP3
Es/N0(dB)
MSE
Floating point Fixed point
(c) (d)
Figure 7.7: Fixed-point channel estimation MSE use QPSK modulation for different fre-quency partitions at different velocities in AWGN channel for IEEE 802.16m uplink.
0 5 10 15 20 25 30
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in AWGN for FP0
Es/N0(dB)
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in AWGN for FP1
Es/N0(dB)
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in AWGN for FP2
Es/N0(dB)
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in AWGN for FP3
Es/N0(dB)
Figure 7.8: Fixed-point channel estimation SER use QPSK modulation for different fre-quency partitions at different velocities in AWGN channel for IEEE 802.16m uplink.
0 5 10 15 20 25 30
DSP Implementation for IEEE 802.16M Uplink Channel Estimation in SUI−2 for FP0
Es/N0(dB)
MSE
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
0 5 10 15 20 25 30
DSP Implementation for IEEE 802.16M Uplink Channel Estimation in SUI−2 for FP1
Es/N0(dB)
MSE
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
(a) (b)
DSP Implementation for IEEE 802.16M Uplink Channel Estimation in SUI−2 for FP2
Es/N0(dB)
MSE
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
0 5 10 15 20 25 30
DSP Implementation for IEEE 802.16M Uplink Channel Estimation in SUI−2 for FP3
Es/N0(dB)
MSE
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
(c) (d)
Figure 7.9: Fixed-point channel estimation MSE use QPSK modulation for different fre-quency partitions at different velocities in SUI-2 channel for IEEE 802.16m uplink, where the speed v is km/h.
0 5 10 15 20 25 30 10−3
10−2 10−1 100
DSP Implementation for IEEE 802.16M Uplink Channel Estimation in SUI−2 for FP0
Es/N0(dB)
SER
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
0 5 10 15 20 25 30
10−3 10−2 10−1 100
DSP Implementation for IEEE 802.16M Uplink Channel Estimation in SUI−2 for FP1
Es/N0(dB)
SER
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
(a) (b)
DSP Implementation for IEEE 802.16M Uplink Channel Estimation in SUI−2 for FP2
Es/N0(dB)
SER
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
0 5 10 15 20 25 30
10−3 10−2 10−1 100
DSP Implementation for IEEE 802.16M Uplink Channel Estimation in SUI−2 for FP3
Es/N0(dB)
SER
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
(c) (d)
Figure 7.10: Fixed-point channel estimation SER use QPSK modulation for different fre-quency partitions at different velocities in SUI-2 channel for IEEE 802.16m uplink, where the speed v is km/h.
0 5 10 15 20 25 30 10−2
10−1 100
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in SUI−5 for FP0
Es/N0(dB)
MSE
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
0 5 10 15 20 25 30
10−2 10−1 100
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in SUI−5 for FP1
Es/N0(dB)
MSE
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
(a) (b)
0 5 10 15 20 25 30
10−2 10−1 100
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in SUI−5 for FP2
Es/N0(dB)
MSE
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
0 5 10 15 20 25 30
10−2 10−1 100
DSP Implementation for IEEE 802.16M Uplink Channel Estimation in SUI−5 for FP3
Es/N0(dB)
MSE
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
(c) (d)
Figure 7.11: Fixed-point channel estimation MSE use QPSK modulation for different fre-quency partitions at different velocities in SUI-5 channel for IEEE 802.16m uplink, where the speed v is km/h.
0 5 10 15 20 25 30 10−1
100
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in SUI−5 for FP0
Es/N0(dB)
SER
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
0 5 10 15 20 25 30
10−1 100
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in SUI−5 for FP1
Es/N0(dB)
SER
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
(a) (b)
0 5 10 15 20 25 30
10−1 100
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in SUI−5 for FP2
Es/N0(dB)
SER
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
0 5 10 15 20 25 30
10−1 100
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in SUI−5 for FP3
Es/N0(dB)
SER
v=30 for floating point v=60 for floating point v=90 for floating point v=30 for fixed point v=60 for fixed point v=90 for fixed point
(c) (d)
Figure 7.12: Fixed-point channel estimation SER use QPSK modulation for different fre-quency partitions at different velocities in SUI-5 channel for IEEE 802.16m uplink, where the speed v is km/h.
0 5 10 15 20 25 30
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in AWGN for FP0 with SFBC
Es/N0(dB)
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in AWGN for FP1 with SFBC
Es/N0(dB)
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in AWGN for FP2 with SFBC
Es/N0(dB)
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in AWGN for FP3 with SFBC
Es/N0(dB)
Figure 7.13: Fixed-point channel estimation MSE use QPSK modulation with SFBC for different frequency partitions at different velocities in AWGN Fixed-point channel for IEEE 802.16m uplink.
0 5 10 15 20 25 30
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in AWGN for FP0 with SFBC
Es/N0(dB)
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in AWGN for FP1 with SFBC
Es/N0(dB)
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in AWGN for FP2 with SFBC
Es/N0(dB)
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in AWGN for FP3 with SFBC
Es/N0(dB)
Figure 7.14: Fixed-point channel estimation SER use QPSK modulation with SFBC for different frequency partitions at different velocities in AWGN Fixed-point channel for IEEE 802.16m uplink.
0 5 10 15 20 25 30
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in SUI−2 for FP0 with SFBC
Es/N0(dB) MSE v=30 with Antenna 1 (Fixed−point)
v=30 with Antenna 2 (Fixed−point) v=30 with Antenna 1 (Floating−point) v=30 with Antenna 2 (Floating−point) v=60 with Antenna 1 (Fixed−point) v=60 with Antenna 2 (Fixed−point) v=60 with Antenna 1 (Floating−point) v=60 with Antenna 2 (Floating−point) v=90 with Antenna 1 (Fixed−point) v=90 with Antenna 2 (Fixed−point) v=90 with Antenna 1 (Floating−point) v=90 with Antenna 2 (Floating−point)
0 5 10 15 20 25 30
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in SUI−2 for FP1 with SFBC
Es/N0(dB)
MSE
v=30 with Antenna 1 (Fixed−point) v=30 with Antenna 2 (Fixed−point) v=30 with Antenna 1 (Floating−point) v=30 with Antenna 2 (Floating−point) v=60 with Antenna 1 (Fixed−point) v=60 with Antenna 2 (Fixed−point) v=60 with Antenna 1 (Floating−point) v=60 with Antenna 2 (Floating−point) v=90 with Antenna 1 (Fixed−point) v=90 with Antenna 2 (Fixed−point) v=90 with Antenna 1 (Floating−point) v=90 with Antenna 2 (Floating−point)
(a) (b)
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in SUI−2 for FP2 with SFBC
Es/N0(dB)
MSE
v=30 with Antenna 1 (Fixed−point) v=30 with Antenna 2 (Fixed−point) v=30 with Antenna 1 (Floating−point) v=30 with Antenna 2 (Floating−point) v=60 with Antenna 1 (Fixed−point) v=60 with Antenna 2 (Fixed−point) v=60 with Antenna 1 (Floating−point) v=60 with Antenna 2 (Floating−point) v=90 with Antenna 1 (Fixed−point) v=90 with Antenna 2 (Fixed−point) v=90 with Antenna 1 (Floating−point) v=90 with Antenna 2 (Floating−point)
0 5 10 15 20 25 30
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in SUI−2 for FP3 with SFBC
Es/N0(dB) MSE v=30 with Antenna 1 (Fixed−point)
v=30 with Antenna 2 (Fixed−point) v=30 with Antenna 1 (Floating−point) v=30 with Antenna 2 (Floating−point) v=60 with Antenna 1 (Fixed−point) v=60 with Antenna 2 (Fixed−point) v=60 with Antenna 1 (Floating−point) v=60 with Antenna 2 (Floating−point) v=90 with Antenna 1 (Fixed−point) v=90 with Antenna 2 (Fixed−point) v=90 with Antenna 1 (Floating−point) v=90 with Antenna 2 (Floating−point)
(c) (d)
Figure 7.15: Fixed-point channel estimation MSE use QPSK modulation with SFBC for different frequency partitions at different velocities in SUI-2 Fixed-point channel for IEEE 802.16m uplink, where the speed v is km/h.
0 5 10 15 20 25 30
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in SUI−2 for FP0 with SFBC
Es/N0(dB)
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in SUI−2 for FP1 with SFBC
Es/N0(dB)
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in SUI−2 for FP2 with SFBC
Es/N0(dB)
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in SUI−2 for FP3 with SFBC
Es/N0(dB)
Figure 7.16: Fixed-point channel estimation SER use QPSK modulation with SFBC for different frequency partitions at different velocities in SUI-2 Fixed-point channel for IEEE 802.16m uplink, where the speed v is km/h.
0 5 10 15 20 25 30 10−2
10−1 100
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in SUI−5 for FP0 with SFBC
Es/N0(dB)
MSE
v=30 with Antenna 1 (Fixed−point) v=30 with Antenna 2 (Fixed−point) v=30 with Antenna 1 (Floating−point) v=30 with Antenna 2 (Floating−point) v=60 with Antenna 1 (Fixed−point) v=60 with Antenna 2 (Fixed−point) v=60 with Antenna 1 (Floating−point) v=60 with Antenna 2 (Floating−point) v=90 with Antenna 1 (Fixed−point) v=90 with Antenna 2 (Fixed−point) v=90 with Antenna 1 (Floating−point) v=90 with Antenna 2 (Floating−point)
0 5 10 15 20 25 30
10−2 10−1 100
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in SUI−5 for FP1 with SFBC
Es/N0(dB)
MSE
v=30 with Antenna 1 (Fixed−point) v=30 with Antenna 2 (Fixed−point) v=30 with Antenna 1 (Floating−point) v=30 with Antenna 2 (Floating−point) v=60 with Antenna 1 (Fixed−point) v=60 with Antenna 2 (Fixed−point) v=60 with Antenna 1 (Floating−point) v=60 with Antenna 2 (Floating−point) v=90 with Antenna 1 (Fixed−point) v=90 with Antenna 2 (Fixed−point) v=90 with Antenna 1 (Floating−point) v=90 with Antenna 2 (Floating−point)
(a) (b)
0 5 10 15 20 25 30
10−2 10−1 100
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in SUI−5 for FP2 with SFBC
Es/N0(dB)
MSE
v=30 with Antenna 1 (Fixed−point) v=30 with Antenna 2 (Fixed−point) v=30 with Antenna 1 (Floating−point) v=30 with Antenna 2 (Floating−point) v=60 with Antenna 1 (Fixed−point) v=60 with Antenna 2 (Fixed−point) v=60 with Antenna 1 (Floating−point) v=60 with Antenna 2 (Floating−point) v=90 with Antenna 1 (Fixed−point) v=90 with Antenna 2 (Fixed−point) v=90 with Antenna 1 (Floating−point) v=90 with Antenna 2 (Floating−point)
0 5 10 15 20 25 30
10−2 10−1 100
DSP Implementation for IEEE 802.16m Downlink Channel Estimation in SUI−5 for FP3 with SFBC
Es/N0(dB)
MSE
v=30 with Antenna 1 (Fixed−point) v=30 with Antenna 2 (Fixed−point) v=30 with Antenna 1 (Floating−point) v=30 with Antenna 2 (Floating−point) v=60 with Antenna 1 (Fixed−point) v=60 with Antenna 2 (Fixed−point) v=60 with Antenna 1 (Floating−point) v=60 with Antenna 2 (Floating−point) v=90 with Antenna 1 (Fixed−point) v=90 with Antenna 2 (Fixed−point) v=90 with Antenna 1 (Floating−point) v=90 with Antenna 2 (Floating−point)
(c) (d)
Figure 7.17: Fixed-point channel estimation MSE use QPSK modulation with SFBC for different frequency partitions at different velocities in SUI-5 Fixed-point channel for IEEE 802.16m uplink, where the speed v is km/h.
0 5 10 15 20 25 30 10−2
10−1 100
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in SUI−5 for FP0 with SFBC
Es/N0(dB)
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in SUI−5 for FP1 with SFBC
Es/N0(dB)
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in SUI−5 for FP2 with SFBC
Es/N0(dB)
DSP Implementation for IEEE 802.16m Uplink Channel Estimation in SUI−5 for FP3 with SFBC
Es/N0(dB)
Figure 7.18: Fixed-point channel estimation SER use QPSK modulation with SFBC for different frequency partitions at different velocities in SUI-5 Fixed-point channel for IEEE 802.16m uplink, where the speed v is km/h.
Chapter 8
Conclusion and Future Work
8.1 Conclusion
In this thesis, first half of that we discussed about the LMMSE channel estimation method for OFDMA downlink and uplink for IEEE 802.16m. We summarize the process to following steps:
• First, use least-square method on pilot position to get each pilot response.
• Second, do linear interpolation in time to get all the pilot response and related
• Second, do linear interpolation in time to get all the pilot response and related