• 沒有找到結果。

OFDM Uplink Synchronization Problem and Tech- Tech-niques

OFDM TDD Uplink Synchronization

5.1 OFDM Uplink Synchronization Problem and Tech- Tech-niques

Accurate demodulation and detection of an OFDM signal requires carrier orthogonality.

Variations of the carrier oscillator, sample clock or the symbol time affect the orthogonality of the system. Then, before an OFDM receiver can demodulate the carriers, it has to perform three synchronization tasks. First, timing estimation is needed to detect the proper frame start time. Secondly, it has to estimate and correct the carrier frequency offset (CFO) of the received signal. Third, sampling frequency offset (SFO), or symbol clock offset) should be detected. Note that, in normal uplink transmission, frame synchronization is not needed because the base station knows roughly when the signal from each SS should arrive. Figure 5.1 shows the proposed synchronizer structure for the receiver.

SRRC filter and 4X downsample

preamble symbol symbol timing offset synchronization STO

CFO

synchronization CFO CFO

compensate FFT

Figure 5.1: The proposed synchronizer structure for the receiver.

5.1.1 Timing Offset and Fractional Carrier Frequency Offset

A popular algorithm to estimate timing offset and (fractional) CFO is proposed in [20]. By taking advantage of the cyclic prefix, the proposed technique can accurately estimate the symbol timing instant and frequency offset relatively accurately in additive white Gaussian noise (AWGN), blindly with no assistance from pilot symbols. However, it suffers consider-able performance degradation in multipath propagation or Rayleigh fading [5]. A modified technique proposed in [21] is shown to have better performance in fast Rayleigh fading. In addition, it can obtain a symbol timing estimate in parallel to the frequency offset estimate.

Figure 5.2 illustrates the algorithm structure proposed in [21]. Under the assumption that received samples are jointly Gaussian, symbol time offset ˆθ and fractional CFO ˆε is given by

To get a more accurate CFO estimation, we can average the estimated values over multiple OFDM symbols.

Figure 5.2: Structure of J.-C. Lin’s symbol timing and fractional carrier frequency synchro-nization method [21].

Our earlier study has considered the above approach. The uplink signal structure defined in the IEEE 802.16e standard, however, motivates another approach which could yield better performance in multipath fading. According to the standard, each UL burst contains a preamble, which consists 2 times 128 samples in the time domain. Although the contents of preamble is known, it seems good (we will see their performance later) to perform blind symbol timing detection based on the 128-sample periodical structure as

θ = arg maxˆ

i i+128X

k=i

|r(k + L)r(k + L + 128)− r(k + L + 128)r(k + L + 256)| (5.3) where i = 0, ..., 8 because all SSs should adjust their timing such that all OFDM symbols arrive time coincident at the BS to an accuracy of ±4 samples. Note that in the second term we subtract out the effect of CP, which interferes with the estimation of correct timing point.

The reason why we do not use the whole known preamble to correlate with the received preamble is because the received symbol is corrupted by channel response, so the correlation result of the corrupted preamble and the original preamble is not very well. On the other hand, if the mobile speed is not very high, the channel response is almost the same during

very similar and the performance of their correlation is better than using the whole known preamble.

5.1.2 Integer Carrier Frequency Offset

The technique discussed above can only estimate fractional CFO. Hence, theoretically, we still need to estimate any possible integer CFO between the transmitter and the receiver. But practically, this is unnecessary in the UL OFDM transmission because of the wide subscriber spacing. According to the specifications of the standard, an SS should synchronize their frequency to the BS to within a maximum tolerance of 0.02 times the subscriber spacing.

Now consider a mobile speed as high as 240 km/h. Then considering a 10 MHz signal bandwidth at a carrier frequency of 5 GHz, the maximum Doppler shift is on the order of 1 kHz, which is much smaller than the 44.531 kHz of subcarrier spacing even with the 2% maximum frequency error. (The conclusion is similar for the profile with the smallest bandwidth, namely, prof P 3 1.75, in Clause 12 of the standard.) Therefore, there is no need to estimate the integer CFO.

5.1.3 Sampling Frequency Offset

From [22], we know the frequency-domain symbol with phase rotation caused by SFO can be modeled as

zl,k = (ej2π((lNs+Ng)/N)ζk)α(φk)al,kHk+ nΩ;l,k+ nl,k. (5.4) where l is symbol number, Ns = N + Ng, k is subcarrier index, ζ = (T0 − T )/T with T0 being the sampling frequency of receiver and T the sampling frequency of transmitter, α(φk) = sinc(πφk) is very close to 1, al,k is transmitted data symbol, Hk is channel impulse response of subcarrier k, which is assumed to stay constant over two consecutive symbols, nΩ;l,k is interchannel interference (ICI), and nl,k is AWGN. In the IEEE 802.16e OFDM,

5 10 15 20

BER performance of imperfect frequency synchronization

perfect synchronization

imperfect frequency synchronization, BL=11 imperfect frequency synchronization, BL=22 imperfect frequency synchronization, BL=33

Figure 5.3: BER degradation at 5 ppm sampling clock error.

N = 256, and we let Ng = 32; hence Ns = 288. Therefore, in two consecutive symbols, the phase increment is given by

∆ϕk= 2π(Ns/N)ζk. (5.5)

Note that the IEEE 802.16e standard has specified the maximum tolerance for the sampling clock frequency at the SS as 5 ppm. This rule can simplify the synchronization work at BS, because the performance degradation caused by SFO is not serious if the transmission burst length is not too long. Figure 5.3 shows our simulation results. The modulation is 16-QAM, in AWGN channel. A burst with length 11 OFDM symbols transmits about 1kb data (200 used carriers × 4 bits/sample × 11 = 8800 bits )each time. We can see the BER degradation is very small when burst length is small.

Since the effects of SFO can be ignored, there is no need to do SFO synchronization.

相關文件