论文部分内容阅读
研究和试验需要处理越来越多的数据,普遍认为:实时提供压缩了的工程单位的和派生的参数数据能减少试验费用,因为试验者能对试验点(方式)的有效性作实时判断。传统的遥测预处理器不能作这么多的实时处理,因为在设计这些处理器时已假定了原始数据是在单个PCM字内并仅仅做一个简单的mx+b的工理单位转换,或者只要求简单的数据压缩。目前的预处理器已用专门的位片技术来加速实时处理。它们只是一个或二个总线系统。它们处理参数的能力常小于每秒300k到400k。此外,许多预处理器都不能处理字连结(相邻的PCM字除外),没有一个能处理象推力、升力、总重量、重力中心、失速速度,谐波分析等复杂的派生参数。为了寻找突破这些限制的方法:开发了一个整体并行的计算机系统,该系统拥有60个通用并行工作的IMFLOP浮点计算机,用以支持吞吐量达1.5Mw/s的任何类型的实时处理。这个系统能够合并多达8个不同异步源来的实时数据,每个源具有高过2.0Mw/s的字速率,多达32768个不同参数与附加的32768个标识符一起能被作为输入接收。标识符用作级联和派生参数的标记。一个功能很强的实时软件包允许计算机系统用户将任一个或多个算法用于任一个或全部被处理的参数。
Research and experimentation require more and more data to be processed. It is generally agreed that providing contracted engineering units and derived parametric data in real-time reduces the cost of testing because the tester can make real-time judgments about the effectiveness of the test site. Conventional telemetry preprocessors can not do so much real-time processing because in designing these processors it has been assumed that the raw data is within a single PCM word and only makes a simple mx + b conversion of the unit of work or only requires Simple data compression. Current preprocessors have used specialized bit-slicing techniques to speed up real-time processing. They are just one or two bus systems. Their ability to process parameters is often less than 300k to 400k per second. In addition, many preprocessors can not handle word links (except adjacent PCM words), none of which can handle complex derived parameters such as thrust, lift, total weight, gravity center, stall speed, harmonic analysis and more. In an attempt to find a way out of these limitations, a fully parallel computer system has been developed that has 60 generic, parallel-running IMFLOP floating-point computers to support any type of real-time processing with up to 1.5 Mbps throughput. This system is capable of combining real-time data from up to eight different asynchronous sources, each source having a word rate higher than 2.0Mw / s and up to 32,768 different parameters, together with an additional 32,768 identifiers, can be received as input. Identifiers are used as flags for concatenated and derived parameters. A powerful real-time package allows computer system users to use one or more algorithms for any or all of the parameters being processed.