The proposed HPSU AC is used for PD and CFO estimation. For accurate coarse and fine CFO estimation, in the below simulations the parameter of (3-3): LS is set as 2 in short symbols and 1 in long symbols. Actually the LS of short symbols is lower than [27] which sets LS as 10 in short symbols therefore the LS setting as 2 also achieves low auto-correlation complexity. To understand the PD performance based on the proposed AC, the FER of PD with different ωAC values and LS=2 in (3-3) is shown in Figure 3-21. The FER is to calculate the error rate to detect both packet and FFT-window. When the packet is successful detected however maybe the success of PD is too late to make the FWD error. And then the frame is still error. So we can find not only the PD accuracy but also the influence of PD to the overall synchronization
802.11a system, we focus to make FER below 10%. As shown in Figure 3-21, when the reduction factor ωAC is increased from 1 to 2, the SNR for 10% FER is only increased by 0.1dB. Although the used signal amount is reduced by half, the used high-power signal can assist to keep the design performance with only 0.1dB additional SNR loss. And the 0.1dB SNR loss could be acceptable for the trade-off of high performance and low complexity. And when ωAC is increased to 4 or 8, the SNR loss for 10% FER will be increased by 2.0dB ~ 3.8dB.
Channel Condition:
RMS=50ns, CFO=40ppm+phase noise, SCO=40ppm, Doppler=50Hz ωAC
ωAC ωAC ωAC
Figure 3-21: FER of the proposed low-complexity PD for OFDM-based WLAN system
To understand the CFO estimation performance based on the proposed HPSU AC. The RMSE of CFO estimation with different ωAC values is shown in Figure 3-22.
And the phase noise is also added in the channel model [33, 34]. To achieve low residual CFO to reduce the frequency-domain ICI, the CFO estimation error is suggested to be less than about ±0.8ppm [24]. As shown in Figure 3-22, when the ωAC
value becomes higher, the CFO estimation error will be increased. That is because
that the amount of the signal used for CFO estimation becomes lower. For 0.8ppm CFO error, the design with ωAC = 2 has 2.4dB SNR loss than that with ωAC = 1. And the design with ωAC ≥ 4 will result ≥ 5.6dB SNR loss for 0.8ppm CFO estimation error. The degraded CFO estimation performance may also cause the PER performance loss.
ωAC ωAC ωAC ωAC
Channel Condition:
RMS=50ns, CFO=40ppm+phase noise, SCO=40ppm, Doppler=50Hz
Figure 3-22: RMSE of the proposed low-complexity CFO estimation for OFDM-based WLAN system
To understand the PER performance based on the proposed low-complexity AC, the PER curves of 6Mb/s (lowest data rate) and 54Mb/s (highest data rate) in AWGN channel and the multipath channel with 50ns RMS delay spread and frequency selective fading ≥ -15dB are shown in Figure 3-23 and Figure 3-24. As shown in Figure 3-23, in 6Mb/s mode and AWGN channel, when ωAC is 1 and 2, the SNR loss for 10% PER is only 0.1dB and 0.15dB compared with the perfect PD and CFO estimation design. That means the increased FER and CFO estimation error of ωAC = 2 increase the SNR loss by only 0.05dB. However when the ω is increased to ≥ 4,
the SNR loss for 10% PER will be arrived at 0.75dB and 2.1dB.
ωAC ωAC ωAC ωAC ωAC ωAC ωAC ωAC
Channel Condition:
AWGN or RMS=50ns, CFO=40ppm+phase noise, SCO=40ppm, Doppler=50Hz
Figure 3-23: 6Mb/s PER of the proposed low-complexity auto-correlation
ωAC ωAC ωAC
ωAC ωAC ωAC ωAC ωAC
Channel Condition:
AWGN or RMS=50ns, CFO=40ppm+phase noise, SCO=40ppm, Doppler=50Hz
Figure 3-24: 54Mb/s PER of the proposed low-complexity auto-correlation
Also as shown in Figure 3-23, when the multipath channel with 50ns RMS delay spread is simulated, the difference of SNR loss between different ωAC values looks smaller than that in AWGN channel. When the multipath channel is simulated, the SNR for 10% PER will be increased from about 2dB to about 6dB in 6Mb/s data rate because of the frequency-selective fading. As shown in Figure 3-21, when the SNR is moved from 2dB to 6dB, the FER will be decreased from about 1% to less than 10-4. And as shown in Figure 3-22, the CFO estimation error in SNR = 6dB is only about 20% of that in SNR = 2dB. In the multipath channel with 50ns RMS delay spread, the SNR loss for 10% PER of ωAC = 1, 2, 4, and 8 is 0.4dB, 0.6dB, 0.8dB, and 1.4dB.
The FER with ω = 8 are shown in Figure 3-24-2 (a). We can find the SNR for 10% FER in multipath channel is 3.1dB. But in Figure 3-23, the SNR for 10% PER of perfect synchronization in multipath channel is 6dB. Since the SNR for 10% PER in multipath channel is higher than that for 10% FER, the PER degradation caused by FER in multipath channel can be less than that in AWGN channel. The CFO estimation error with ω = 4 are shown in Figure 3-24-2 (b). We can find the CFO estimation RMSE curves of AWGN and multipath channel are similar to each other.
That is because the auto-correlation used for CFO estimation can be robust to multipath channel. We can find the CFO estimation error is lower when SNR condition is higher. And the SNR for 10% PER in multipath channel is higher than that in AWGN channel. Hence the CFO estimation error which degrades the SNR for 10% PER can be less when the channel condition becomes from AWGN to multipath.
And the PER degradation caused by FER and CFO estimation error in multipath channel can be less than that in AWGN channel.
Channel Condition:
AWGN or RMS=50ns, CFO=40ppm+phase noise, SCO=40ppm, Doppler=50Hz
(a) (b)
Figure 3-24-2: FER and CFO estimation error with ω = 8
To understand the performance difference between high-power-signal-used and random-signal-used design, the PER curves with ωAC=1~4 are drawn in Figure 3-24-3.
We can find the PER with ωAC=2 and random-signal-used design is higher than that with high-power-signal-used design and lower that with ωAC = 4. By using high-power-signal-used scheme the SNR for 10% PER can be saved.
Channel Condition:
AWGN or RMS=50ns, CFO=40ppm+phase noise, SCO=40ppm, Doppler=50Hz
Figure 3-24-3: CFO estimation RMSE and range for OFDM-based WLAN system
The RMSE and range of the proposed CFO estimation is shown in Figure 3-25.
With the estimation in both short and long symbols, the estimation range can be -120
~ 120ppm CFO (TX+RX).
CFO Between Tx and Rx [ppm]
CFO Estimation RMSE [ppm]
Figure 3-25: CFO estimation RMSE and range for OFDM-based WLAN system