Consideration of Dual Mode Architecture Design

Two kinds of methods can be used for the symbol boundary detection and they are atuo-correlation based and cross-correlation based method. The atuo-correlation based method uses time shifted and repeated signals and calculates the correlation among them. A maximum correlation method [6][43] with a moving sum architecture [25]

can reduce the required multiplier in atuo-correlation. The cross-correlation based method correlates the received signal and the known transmitting signal and a match filter is an example. In the 60 GHz standards, the known preamble uses Golay sequence; hence, a Golay correlater [63] can be adopted. The Golay correlater is also a kind of match filter, but it has lower complexity. The comparison of these method are listed in TABLE 5-3. In summary, the Colay correlator has the lowest complexity;

however, in the view of the whole system, the maximum correlation operation can also estimate the CFO when searching the boundary. Therefore, the proposed receiver will use maximum correlation algorithm.

TABLE 5-3 Comparison of Boundary Detection

Multiplication pre Sample Storage Unit (Samples) Maximum Correlation

(Moving Sum)

1 2N

Match Filter

N (0)

(if the known pattern is +1 or -1; then, multiplications can be replaced with additions )

N

Golay Correlator

log₂(N) (0)

(if Golay sequence is +1 or -1; then, multiplications can be replaced with additions )

N

( N is the length of preamble or repeatd signal)

Moreover, three methods can be used for the carrier synchronization, the maximum correlation (a delay correlation method) [43], the pilot-based method [6], and the decision direction method [64]. The discussions of three methods are followed:

 Maximum correlation method reuses the hardware of the symbol boundary detection. When the maximum value of the correlation output is found, the phase is an estimation of the CFO. The required information can use preamble or guard interval (cyclic prefix or postfix). Both OFDM and SC system can adopt this method by using preamble. Furthermore, a SC system with GI can also adopt this method.

 The pilot-based method uses the known pilots inserted in frequency domain and calculates the difference of phase between the successive symbols. However, the SC system does not have plots in the frequency domain; hence, it cannot adopt this method.

 Decision direction method uses the sliced data and calculates the difference of the phase with the un-sliced data. This method will have large performance degradation when decision errors happen. This method does not require any known data; hence, both a SC receiver and an OFDM receiver can use this

method.

TABLE 5-4 Comparison of CFO Estimations

Required Information System

Maximum Correlation Repeated signals

SC OFDM Pilot-based Plots in frequency domain OFDM

Decision Direction ^{None SC}

OFDM

For the SCO estimation, the traditional methods in wire line systems are the edge detection method, such as the early-late gate method, Gardner‟s method [65] and Mueller-Muller‟s method [66]. However, these methods are not suitable for wireless system because multi-path channels distort the edge of the transmitted signal. The most usable method in wireless channel is based on the phase rotation due to SCO in the frequency domain. The effect of SCO in the time domain is not obvious in short time period. For a wireless SC system, it is difficult to estimate SCO in the time domain. Fortunately, a typical transmitter and receiver usually use the same reference clock source so it is reasonable to assume that CFO and SCO have the same ratio [29].

Hence, the estimation of CFO can also be an estimation of SCO.

For the channel estimation, due to the cyclic prefix which forms a circular convolution, an OFDM receiver can use a one tap equalizer in the frequency domain.

The channel estimation usually uses interpolation based algorithms which use pilots inserted in frequency domain to estimate channel frequency response (CFR). In 802.15.3c standard, the interval between known plots is 22 sub-carrier spaces. Thus, it is hard to do interpolation in an environment with long channel dispersion.

802.15.3c standard provides a preamble in time domain called Pilot Channel Estimation Sequence (PCES). Hence, PCES can be transferred into the frequency domain and estimation CFR. Moreover, a LMS tracking the CFR can be adopted to improve the performance of the estimation [67] [68].

A data block of SC mode in 802.15.3c standard is composed of 448 data and 64 known GI. The data structure is also called single carrier block transmission (STBC) and a SC-FDE [57] can be adopted. The known GI is a Golay Sequence; hence, a time domain estimation method [69] can be adopted by using the property of Golay Sequence. After getting the channel impulse response (CIR), the inverse of CIR is required for a time equalizer or a FFT which transfers CIR into CFR is required for the frequency equalizer.

In summary, the proposed architecture of the OFDM/SC dual mode receiver for 802.15.3c standard is as described below:

 The symbol/Preamble detection uses the auto-correlation and CFO is also estimated by using the auto-correlation [70].

 Assume the ADC and Mixer use the same reference clock source [29]; hence, the estimation of CFO is also an estimation of SCO.

 The equalizer of SC mode uses SC-FDE and shares the FFT unit with the OFDM mode. Besides, the additional tracking loop (LMS estimation) is adopted to improve the system performance [68].

Finally, the block diagram of the proposed dual mode receiver is shown in Fig. 5-1.

Fig. 5-1 OFDM/SC dual modes receiver for 802.15.3c