An approach to hardware-oriented software optimization during the implementation of real-time digital radar signal processing is described. A typical moving target indication algorithm is considered. The main principles of the optimization approach are given. The speed of the optimized algorithm in comparison to the implementation performed by standard software means is shown. The possibility of potential increase in signal processing speed with hardware-oriented optimization is indicated.

Введение. Бистатическая просветная радиолокация -одно из перспективных и нетрадиционных направлений современной радиолокации. Это связанно с тем, что в данном методе основную роль играет теневая компонента рассеянного поля [1, 2], которая, как правило, по интенсивности на несколько порядков больше, чем компонента обратного рассеяния в случае моностатической радио-локации. К тому же теневая эффективная поверхность рассеяния (ЭПР) не за-висит от наличия на объекте радиопоглощающих покрытий [3] и применения технологии «Стелс» [4]. Эти особенности позволяют эффективно использовать бистатические просветные радиолокационные системы (БП РЛС) для обнару-жения малозаметных маловысотных воздушных объектов [5]. Традиционное построение БП РЛС предполагает использование узкопо-лосного (немодулированного) сигнала и несканирующих антенн на приемной и передающей позициях. При этом определение координат и траекторий целей имеет невысокую точность и осуществляется путем длительных измерений ча-стоты Доплера и азимута цели на приемной позиции за время пролета целью барьерной зоны обнаружения, сосредоточенной вблизи линии базы БП РЛС [6].

In advanced portable radars using the principles of MIMO, obtaining a high resolution and a sufficient S/N ratio is important due to the use of a reduced number of widely directed transmitting and receiving antenna elements, for example, sub lattices with low directional coefficients. In such radars, probing signals in the form of burst of broadband LFM pulses are applicable. When processing these signals reflected from moving targets, two methods are fundamentally possible: classical one in the time domain with matched filtering of the burst of LFM-pulses and subsequent inter-period accumulation, and the spectral one based on preliminary calculation of the full spectrum of the pulses burst with subsequent accumulation over the frequency intervals of Doppler ambiguity. In this case, there are two options for the accumulation of spectra: 1) incoherent and 2) coherent, and their efficiency for a burst of LFM-pulses has not been studied before. In this work on the simulation model of the processing system in the spectral domain the output S/N ratios for these variants of the spectral processing method for a burst of LFM-pulses are determined and compared. In addition, the issues of presenting the results of spectral accumulation in the coordinates «Doppler frequency – delay» plane were worked out. The following results are obtained for a specific variant of the burst signal: the efficiency of coherent spectrum accumulation under the influence of noise exceeds the efficiency of incoherent by a value of about 12 dB; in the case of incoherent accumulation of spectrum modules in the case of a group targets, the accumulated amplitude spectrum, with a sufficient duration of the pulses burst, contains all the Doppler-resoluted target marks; in the case of coherent accumulation of complex spectra of group targets with different Doppler frequencies and ranges, the targets can be resolved based on the analysis of the module of the accumulated complex spectrum in the «Doppler frequency – range» plane; 4) with incoherent and coherent of spectral accumulation over ambiguous Doppler intervals, a single FFT of a LFM-pulses burst is required. The remaining operations are the algebraic summation of the spectral samples over the ambiguity intervals and taking the modulus from the accumulated complex spectrum.