一种VLBI数字基带转换器的阈值计算方法及其FPGA实现

A Method of Calculating the Quantization Threshold for a VLBI DBBC and Its FPGA Implementation

随着电子技术的进步,数字化的设备成为发展的趋势。在VLBI2010中定义了新型VLBI数据终端,新的系统使用数字逻辑电路完成VLBI的数据采集。数字基带转换器(Digital Base Band Converter,DBBC)是VLBI数据终端的核心部件,其功能是将射频接收机输出的宽带模拟中频信号数字化处理后,选取若干频道转换为基频信号。与模拟设备相比,其在带通特性、长基线条纹信噪比性能等方面有很大优势。其中经过基带转换后的信号需要通过自动增益控制(Auto Gain Control,AGC)模块进行阈值比较,得到2 bits量化信号作为输出。目前2 bits数字自动增益控制模块设置阈值的方法是基于传统的自动增益控制方法,即信号通过平方、累加和开方,计算出信号的平均功率,再通过平均功率得到阈值。提出了一种量化阈值计算的新方法,该方法通过统计数字信号各比特位的状态分布,并与原阈值进行比较,从而确定新阈值的相应比特位的值,使经阈值比较后的量化输出符合预定的比例。阈值随每N个输入信号的统计情况进行更新,从而实现2 bits动态量化。这种方法可以避免多位数据的平方、累加和开方的复杂计算,从而减少数字自动增益控制模块的资源占用。通过对其FPGA设计的仿真结果分析,验证了该方法的可行性。对2 bits量化原理及其量化误差分析的关键部分作了论述,并通过MATLAB计算出最佳量化门限和量化状态。

With the development of digital signal processing circuits, especially the using of programmable components, digital equipments are replacing analog ones. The VLBI2010 defines a modem digital backend system using digital logic circuits to perform VLBI data acquisition. A Digital Base-Band Converter （DBBC）, which digitizes the broadband Intermediate Frequency （IF） analog signals output by the radio receiver and transforms them into base-band signals through different channels, is the core component of a VLBI data acquisition system. It has many advantages over the traditional Analog Base-Band Converter （ABBC） such as better bandpass properties and higher signal-to-noise ratios in long-baseline measurements. In order to obtain the 2-bit output of a DBBC, the converted base-band signals are compared with a quantization threshold by a digital AGC and transformed into 2-bit quantized signals. A currently used 2-bit digital AGC is based on the traditional AGC method, which calculates the threshold value from the average power as obtained by taking squares, additions, and square roots of the base-band signals. In this paper, we present a new method to calculate the quantization threshold. Through counting and analyzing the states of all bits of the digital signals, the new method determines the best-fit threshold value according to the statistics of the distribution of the digital-signal bits, so as to conform the statistics to prescribed number proportions of bits. The threshold is updated every N input signals with the best-fit approach to achieve dynamic quantization. With this method, which avoids complex calculations in the traditional method, the number of resources occupied in the digital AGC module is reduced. The simulation results of an FPGA design of the method have verified its feasibility. Key parts of the 2-bit quantization and the related theory of quantization noise are also briefly introduced in the paper. The optimal quantization thresholds and the quantitative weights based on the theory are calculated by the MATLAB in the simulations.