Time domain synchronisation estimation algorithm for FBMC vector signal analysis in 5G system

: For the challenges of filter bank multi-carrier (FBMC) test applications in the fifth-generation mobile communication network (5G) system, presented is a time domain synchronisation estimation algorithm based on power synchronisation in frequency domain. Through MATLAB simulation, the proposed algorithm is verified to achieve good synchronisation at low signal-to-noise ratio (SNR) and large frequency offset without signal quality changes. As a result, it can be applied in FBMC signal vector analysers, promoting FBMC signal vector analysis functions used in 5G test applications.


Introduction
Rapidly growing mobile communications are bringing people infinite convenience. With fourth-generation mobile communication network (4G) coming into large-scale commercial applications phase, R&D globally has been focused on the fifthgeneration mobile communication network (5G) towards 2020 year and future beyond. Filter bank multi-carrier (FBMC), as one of the alternative waveforms for the 5G system, has advantages such as good out-band suppression, very high utilisation rate of the spectrum, suitability for scattered fragment frequencies utilisation and so on, compared with orthogonal frequency division multiplexing technology (OFDM) used in the 4G system [1].
FBMC can use scattered frequencies in any combination to effectively improve spectrum usage and solve the problem with high bandwidth applications at the lower band, therefore satisfying expected flexibility needed for various services in the 5G landscape. As one of the hotspot technologies, FBMC has been extensively studied on their core techniques, such as prototype filter design, channel equalisation, and multiple input multiple output (MIMO) transmission as well.
The prototype filters involve PHYDYAS filter [2], Rossi filter [3] and frequency selection algorithm [4]. Channel equalisation involves frequency sampling technique-based linear multitap equaliser [5], minimum mean square error (MMSE)-based enhanced linear multitap equaliser [6] etc. MIMO transmission technology involves Alamouti-based coding scheme [7], maximum likelihood test (MLD)-based innovative mitigation technique [8], and others. However, the investigations are still much insufficient on time domain synchronisation as the first step for FBMC reception. Related literature on domain synchronisation is highly limited [9]. Furthermore, these investigations are mainly on modifications for synchronisation technique based on OFDM. So this cannot meet requirements of FBMC signal vector analysis for domain synchronisation. We presented a time domain synchronisation estimation algorithm with low complexity, easy implementation, and high performance, which can achieve favourable synchronisation performance at low SNR and large frequency offset without signal quality changes. It can be used in FBMC signal vector analysers used in 5G test applications.

FBMC principle
PHYDYAS began to conduct FBMC project since 2009 and provided a fast FBMC implementation solution and detailed principle [10-12] that will not be described any more herein. The packet filtering coefficients in FBMC modulation can be expressed as where M is the symbol length; B(Z) is the Z domain matrix for the filtering coefficients; and w = e j2π/M ; H(Z) is the Z domain matrix for the prototype filtering coefficients. The filter matrix for packet filtering can be transferred into two matrices in the product form where the first is the matrix for inverse Fourier transform (IFFT), and the second is in the expanded multiphase form to perform the filtering function. So the FBMC modulation can be implemented by firstly converting serial data into parallel through the IFFT module, then processing IFFT output through the multiphase network. The system block diagram is shown in Fig. 1.
For the FBMC signals, sub-carriers are not orthogonal to each other. It is common to use real and imaginary cross mapping as shown in Fig. 2 to ensure orthogonality of a sub-carrier to its adjacent two sub-carriers. However, this obviously will lead to a decrease in transmission speed and spectrum usage by half. It is not expected for 5G. Also, it will cause huge waste because spectrum resources are at a premium in wireless communication systems. Consequently, offset quadrature amplitude modulation (OQAM) modulation is proposed in the FBMC system, where its mapping is shown as in Fig. 3. The basic idea remains that QAM signals are modulated by real and imaginary cross-mapping to avoid interferences between adjacent sub-carriers. Based on this, additionally, OQAM is modulated by splitting data into two channels, which doubles the amount of data to solve decreasing data traffic problem due to real and imaginary cross-mappings. Interferences between two channels are prevented using

Time domain synchronous estimation algorithm
Similar to OFDM modulation, FBMC modulation also uses channel equalisation to eliminate influences of time, phase, frequency offsets, and noise, but this will cause damage to signal quality. What is more, its analysis results, associated with used channel equalisation, cannot intuitively reflect signal quality as well. On the other hand, it is estimated that the standard has not been established for the FBMC channel, so it is unpersuasive for using channel equalisation for FBMC analysis; besides, vector analysis, as a traditional means for signal modulation analysis, is a mature technique. FBMC vector analysis, as a critical test means used for FBMC applications, performs signal analysis through time domain synchronisation, frequency offset synchronisation, phase offset correction, and other methods without channel equalisation used. Time domain synchronisation is the first step in vector analysis to determine the optimal moment for signal time domain sampling, and the synchronisation performance will affect the accuracy of the whole vector analysis. For vector signal modulation, interpolation and filtering are conducted in the time domain, but for FBMC modulation, interpolation and filtering in the time domain. We proposed a time domain synchronisation estimation algorithm with power synchronisation in the frequency domain, as shown in Fig. 4.
The symbol synchronisation module performs the judgment of initial position for symbol time domain. For bursting FBMC signal, synchronisation is implemented using power, but for continuous FBMC signal, synchronisation is implemented using the OFDM frame synchronisation method to interpolate ZC sequences.
The time domain interpolation module performs recovery of time domain data to equivalently improve sampling rate and is used to subdivide time domain scale for reducing time domain synchronisation errors. Detailed steps are as the following:
Second, according to the time domain synchronisation estimation algorithm, select O as 16 to perform simulations, with obtained power accumulation module as shown in Fig. 8. Using   power to judge output position, output synchronisation position as shown in Fig. 9. To verify the time domain synchronisation estimation algorithm, first, simulations are proceeded at different SNR values: 0, 10, 20, and 30 dB, respectively. Results from 1000 simulations are shown in Table 2.
Second, simulations are proceeded at different DFs for frequency offset, 100 Hz, 1, 10, and 20 kHz, respectively, with 60 dB SNR. Results from 1000 simulations are shown in Table 3.
Through these simulations, it can be found that good performances are obtained at low SNR, e.g. 30 dB, using the time domain synchronisation estimation algorithm, and to increase synchronisation probability, more symbols can be added to conduct synchronisation; however, good performances are also obtained at large frequency offset using the time domain synchronisation estimation algorithm.

Conclusion
The paper proposed a time domain synchronisation estimation algorithm for FBMC vector signal analysis. Through simulations, the algorithm is verified to have resistance to frequency offset and anti-jamming advantages, satisfying requirements of vector analysers for time domain synchronisation. As time domain synchronisation estimations are conducted in the frequency domain, available bandwidth has a direct effect on the performance of the algorithm. For further efforts, we will improve the algorithm to mitigate effluence on available bandwidth; in the meanwhile, we will investigate the frequency domain estimation algorithm for vector analyses to promote FBMC vector analyses used in 5G test applications.

Acknowledgments
This work was supported by the National Science and Technology Major Project of the Ministry of Science and Technology of China (no. 2018ZX030117002).