动态双目标VLBI软件相关处理技术研究

Dynamic dual-target VLBI correlator technology

本文介绍了应用于中国VLBI网软件相关处理机(CVNScorr)的动态双目标VLBI相关处理技术.嫦娥五号(CE-5)实现了我国首次对地外天体的采样返回,测控目标多、技术难度大.在交会对接过程中,需采用实时同波束VLBI技术完成对轨道器、上升器两个动态目标的高精度测轨, CVNScorr需具备2个相位中心同时处理能力.由于在交会对接、近月制动、动力下降、月面起飞等特殊动力飞行段,预报轨道精度不高,无法获得VLBI高精度时延模型,若采用常规VLBI数据处理方法,无法获取干涉条纹.因此, VLBI相关处理机增加了实时双目标条纹搜索与时延模型重构功能. CVNScorr采用FX型软件相关处理机原理,具备双目标相关处理模块和多信标条纹搜索及时延模型重构功能,可利用测控信号特点,自动搜索多个信标,同时获取两个目标的时延值,重构高精度的时延模型. CVNScorr运行于CPU+GPU混合并行集群平台,利用MPI和GPU等技术实现并行计算,在CE-5任务中实现了中国VLBI网4站双目标实时数据相关处理,并具有6站双目标数据事后处理能力.实时数据率达到128 Mbps/站,VLBI时延精度达到0.4 ns,实时性优于25 s.

The dynamic dual-target very long baseline interferometry(VLBI) processing technology applied to the CVNScorr software processor is introduced. Chang’e 5(CE-5) has achieved China’s first sample return mission from an extraterrestrial body. The characteristics of VLBI are that there are several target probes and the telemetry tracking and command(TT&C) technical requirements are high in CE-5. In the rendezvous and docking process, a real-time in-beam VLBI technology is required to perform the high-precision orbit measurement of two dynamic targets, the orbiter and the ascender. In addition, the VLBI correlator(CVNScorr) should tackle two-phase centers from two probes simultaneously.In the special power flight phases such as rendezvous and docking, near-moon braking, powered descending, and lunar takeoff, the VLBI high-precision delay model cannot be derived owing to the low accuracy of the predicted orbit. If the conventional data processing method is adopted, the VLBI fringe cannot be obtained. Therefore, the VLBI correlator should have the function of real-time dual-target fringe search and delay model reconstruction. CVNscorr adopts the principle of FX-type software correlation and has the function of dual-target correlation module, multibeacon fringe search, and delay model reconstruction. By taking advantage of the characteristics of TT&C signals from different probes, multiple beacons fringes are automatically searched and the delay values of the two targets are obtained simultaneously to reconstruct the high-precision delay models. CVNScorr uses MPI, GPU, and other technologies to realize parallel computing, and it is run on a CPU+GPU parallel cluster platform. In the CE-5 mission, the real-time processing capability is up to four stations of the Chinese VLBI Network and the postprocessing capability is up to six stations. The real-time data rate reaches 128 Mbps per station, the VLBI delay accuracy reaches 0.4 ns, and the real-time performance is better than 25 s.