FPGA implementation of timing synchronization in OFDM underwater acoustic communication system

0 Introduction

Orthogonal Frequency Division Multiplexing (OFDM) is a multi-carrier modulation technology that decomposes a broadband channel into a group of mutually orthogonal narrowband sub-channels and uses each sub-channel for parallel data transmission. Therefore, it has high spectrum utilization and strong resistance to multipath fading. It has been successfully applied in digital video broadcasting (DVB-T2), wireless local area network (802.11a/g]) and other systems, and has become one of the core technologies of the fourth generation of mobile communications. The underwater acoustic channel is a time-, space-, and frequency-varying multipath channel. It has the characteristics of strong multipath, narrow frequency band, and strong noise. Applying OFDM transmission technology to underwater acoustic communication has become one of the research hotspots of underwater acoustic communication.

The orthogonal multi-carrier modulation characteristics of the OFDM system itself determine that it is very sensitive to synchronization errors. Whether accurate symbol timing synchronization and carrier frequency synchronization can be achieved will directly affect the performance of the OFDM communication system. Since the linear frequency modulation (LFM) signal has good time-frequency aggregation, the LFM signal is suitable as a timing synchronization signal for the OFDM underwater acoustic communication system. At the receiving end, the position of the correlation peak of the LFM signal is detected using the autocorrelation characteristics of the LFM signal to achieve the timing synchronization of the OFDM underwater acoustic communication system.

1 Introduction to basic principles

1.1 Principle of OFDM underwater acoustic communication system

The principle block diagram of a typical OFDM underwater acoustic communication system is shown in Figure 1.

The input data symbols are mapped into a complex data sequence X[0], X[1], ..., X[N-1] through DQPSK. After serial-to-parallel conversion, the N parallel symbols are modulated onto N subcarriers and become time domain sample values ​​x[n] after IFFT:

After adding a cyclic prefix (CP), inserting an LFM synchronization signal, and D/A conversion, the signal is finally converted into an acoustic signal by an underwater acoustic transducer and transmitted in the underwater acoustic channel. At the receiving end, the signal is converted into an electrical signal by a receiving transducer, and after a series of inverse processes such as signal conditioning, A/D acquisition, and FFT, the data symbol demodulation can be completed.

In order to correctly restore the data symbols, this system uses the good autocorrelation characteristics of the LFM signal as the timing synchronization signal of the OFDM symbol. The frame structure of the signal sent by the OFDM underwater acoustic communication system is shown in Figure 2. The sliding correlation detection method is used at the receiving end to obtain the position of the correlation peak and achieve accurate synchronization of the timing symbols. Then, the OFDM signal can be demodulated through the inverse process of the transmitting end, and finally the original data symbols can be restored.

1.2 Characteristics of LFM Signal

LFM signal is a large time-width-bandwidth signal widely used in radar systems. The complex expression of LFM signal is:

Where: μ = B/r is the frequency change slope, B (= △f) is the frequency change range. The real signal is expressed as:

Its time domain waveform and autocorrelation output are shown in Figure 3. It can be clearly seen that the frequency of the LFM signal changes linearly within the pulse period, and the autocorrelation peak is very sharp.

The LFM signal has a parabolic nonlinear phase spectrum, and Bτ>1, τ is the signal time width, and B is the signal bandwidth. Therefore, the LFM signal has good pulse compression characteristics. Its fuzzy function (autocorrelation function) surface has a sharp main peak and a low skirt. It is insensitive to Doppler shift, and even if there is a large Doppler shift, it still has good pulse compression characteristics. The underwater acoustic channel has the characteristics of strong multipath, time, space, and frequency variation. Using the LFM signal as a synchronization signal can obtain better correlation detection performance, and will not cause obvious pseudo peaks due to multipath. After experiments, it is verified that the LFM signal can obtain better synchronization performance as the synchronization signal of the system. Therefore, this article focuses on the generation and synchronization detection of LFM signals on FPGA.

2 Generation and detection of LFM signals

2.1 Generation of LFM Signal

There are usually two methods for generating LFM signals: I, Q two-way digital generation method and intermediate frequency direct generation method. The former is more complicated to implement and is suitable for occasions with high frequency and large bandwidth. Underwater acoustic signals generally operate in a lower frequency band, which is suitable for generating LFM signals using the intermediate frequency direct generation method. According to the available bandwidth requirements of the OFDM underwater acoustic communication system in this laboratory, Direct Digital Synthesis (DDS) technology is used to directly generate LFM signals with a scanning frequency of 13 to 16 kHz.

DDS technology can be further divided into direct digital waveform synthesis (DDWS) and direct digital frequency synthesis (DDFS), which are slightly different in implementation structure. DDWS is also called digital waveform storage direct reading waveform generation system, which pre-stores the ideal sampled digital waveform, and obtains the required analog signal by D/A conversion through table lookup when in use. The LFM signal generated by this method is basically not limited by the frequency modulation slope, and can be used to generate arbitrary waveforms (including complex waveforms and large data volume combined waveforms). It can also perform pre-distortion processing on the pre-stored data waveform to improve the performance of the system. This design uses DDWS to generate LFM signals, and the basic principle block diagram of generating LFM is shown in Figure 4.

[page]

Under the control of the 50 MHz master clock, the internal logic of the FPGA controls the output of the LFM signal at a frequency of 120 kHz. After the digital signal is D/A converted, a step-shaped time domain signal is output, and then a bipolar LFM signal is obtained after the out-of-band noise is filtered out by a bandpass filter.

2.2 Detection of LFM Signal

The detection of LFM synchronization signal at the receiving end is essentially the process of obtaining the compressed narrow pulse of LFM signal, so as to achieve the purpose of synchronization signal extraction. The methods generally used are matched filtering method and correlation extraction method. The implementation of matched filtering requires processing using FFT and IFFT transformations in the frequency domain, which consumes a large amount of FPGA resources and has high complexity. Considering the hardware resources and computational complexity, this design uses the sliding correlation method in the time domain to realize the detection of LFM signal. This method takes advantage of the sharp autocorrelation characteristics of LFM signal. According to the formula of correlation operation:

When the received LFM signal is the same as the locally stored LFM signal (j=0 in the above formula), the correlation value is the largest and a sharp correlation peak appears. Figure 5 is a block diagram of the principle of implementing the LFM signal correlation algorithm using FPGA.

At the transmitting end, the number of points of a periodic LFM signal is 256. After A/D sampling at the receiving end, an 8-bit digital quantity is obtained and stored in a receiving buffer with a length of 256 B. The buffer is designed as a first-in first-out (FIFO) and is used as a sliding window to perform correlation operations with the local correlation sequence. The local correlation sequence (stored in ROM) is the same as the LFM sequence sent by the transmitting end, and the capacity of the ROM is also 256×8 bits.

Each time an A/D sampling is completed, the obtained 8-bit data is stored in the FIFO, and then a correlation operation is performed to obtain 256 16-bit data. These 256 data are then added together to obtain the corresponding correlation value at this moment (stored in 24 bits). After processing the sequence of the obtained 256 consecutive correlation values, the maximum value is obtained to determine the position where the LFM signal is received.

3 Experimental Results

In order to verify the performance of LFM signal as synchronization signal in underwater acoustic communication, relevant experiments were carried out in a laboratory pool. The FPGA used in the experiment is CycloneⅡEP2C20Q240C8. Considering the situation of half-duplex communication, the generation and detection of LFM signal are realized in the same FPGA, and a total of 3693 logic units (Logic: Elements, LE) are used, accounting for 20% of the total LE of EP2C20 chip. The basic block diagram of the experimental system is shown in Figure 6.

The oscilloscope model in Figure 7 is TDS2024, and the signals observed in each channel are as follows:

CH1 is the LFM signal sent by the transmitter. Since the signal output by the D/A is filtered by a bandpass filter, the high-frequency and low-frequency parts of the signal are attenuated.

CH2 is the receiving signal (the signal output by the transducer is amplified 5,000 times and band-pass filtered).

CH3 is the synchronous pulse output after the receiving FPGA detects the LFM signal.

As can be seen from Figure 7: this scheme realizes the generation of LFM signals. In the laboratory water pool with severe multipath, the synchronous detection of LFM signals is correctly completed at the receiving end, and the correlation peak position of the LFM signal can be extracted more accurately, proving that this method is feasible as a timing synchronization scheme for OFDM underwater acoustic communication systems.