基于模型的DSP算法研究及高速FPGA实现

VIP免费

3.0 陈辉 2024-11-19 4 4 4.64MB 70 页 15积分

侵权投诉

摘要

信号的运算和处理几乎是所有的智能化仪器必不可少的组成部分。本文针对

“便携式智能光谱分析仪”的设计研究也不例外地选用了以现场可编程门阵列

（FPGA）为核心的完全嵌入式的 DSP 信号运算与处理平台。

FPGA 具有高度的并行性、重构的灵活性和最佳的性价等独特的优势，因此它

在信号处理方面与传统 DSP 处理方法相比有着巨大的改进。本文首先介绍了 FPGA

中的 DSP 硬件资源，并从内部结构上解释了 FPGA 是如何实现 DSP 的以及它的优

势所在，然后根据实际设计要求，在 Matlab-Simulink 环境下，结合 Xilinx 公司所

开发的插件“System Generator”建立了相应的 DSP 算法模型。具体的模型主要包

括LMS 自适应滤波器模块、归一化模块、存储器模块、比较器模块等。在建立每

一个独立 DSP 算法模型的时候，不但要保证功能上的完整性，也要力求在算法上

达到最优化；在硬件上的可实现性及资源消耗上达到最小化。

全部算法模型在通过软件仿真后将其打包，生成一个硬件可识别的“核”，

然后将这个“核”下载到“现场可编程门阵列器件”Xilinx-Sparten3E 中，进行硬

件协同仿真验证。再将验证结果和软件仿真的结果进行比较，从而证明先前所建

立的 DSP 算法模型是正确和可行的。

关键词：Xilinx-FPGA 嵌入式 DSP System Generator LMS 自适

应滤波器

ABSTRACT

The operating and processing of signal is a necessary part of an intellectualized

instrument. “The portable instrument which can analyze spectral” described in the paper

also select FPGA as a core where a digital signal processing (DSP) module is embedded

in. FPGA has three main advantages. Firstly, it can work in a parallel way. Secondly, it

is easy to be restructured. Thirdly, it supplies the best cost performance. So, FPGA

makes great progress in signal processing with the comparison of the traditional DSP.

The paper first introduces the hardware recourses of DSP in FPGA , and explains the

way of how to realize DSP from the structure of FPGA.

Then, some DSP modules are established based on the actual case with the help of

the “System Generator” offered by Xilinx in Matlab-Simulink. The concrete module

includes the LMS adaptive digital filter module, the normalization module, the memory

module, the comparison module and so on. In the process of establishing the DSP

module, not only the function of it should be correct, but some other factors should also

be considered such as whether the DSP arithmetic is the most optimal, whether the DSP

arithmetic consumes the least hardware recourses.

After all of the software simulation getting across, wrap the whole DSP module

and generate a “core” which the hardware could identify. Finally, download the “core”

into Xilinx-Sparten3E and do the hardware co-simulation. Compare the result of the

software simulation with the hardware co-simulation, and prove whether the DSP

module established above is correct and feasible.

Key words: Xilinx-FPGA, Embedded Digital Signal Process, System

Generator, LMS adaptive filter

中文摘要

ABSTRACT

第一章绪论·························································································· 1

§1.1 课题的来源及意义····································································· 1

§1.1.1 课题的来源···········································································1

§1.1.2 课题的意义···········································································1

§1.2 FPGA 功能简介以及目前的国内外应用情况······································ 2

§1.2.1 FPGA 的基本结构·································································· 2

§1.2.2 FPGA 的国内外应用情况························································· 3

§1.3 课题所要研究的内容及实施方案··················································· 5

§1.4 本章小结··················································································8

第二章蔬菜农药残留近红外光谱测量系统简介···········································9

§2.1 测量系统的原理与结构······························································· 9

§2.2 光路系统简介·········································································· 10

§2.2.1 光源·················································································· 10

§2.2.2 光纤·················································································· 12

§2.2.3 光谱仪··············································································· 12

§2.2.4 样品槽··············································································· 14

§2.2.5 光电探测器········································································· 15

§2.3 实验数据预处理与结果分析························································17

§2.3.1 采样数据的预处理································································ 17

§2.3.2 甲胺磷的光谱······································································ 17

§2.3.3 具体实验步骤及数据结果分析················································· 18

§2.4 本章小结················································································ 21

第三章基于 FPGA 的DSP 嵌入式系统··················································· 22

§3.1 DSP 的一般特性·······································································22

§3.2 FPGA 实现 DSP······································································· 23

§3.3 Xilinx-FPGA 的DSP 硬件资源·····················································27

§3.3.1 逻辑资源的 DSP 特性···························································· 28

§3.3.2 乘法器专用模块··································································· 30

§3.3.3 DSP 专用模块······································································ 31

§3.4 本章小结················································································ 34

第四章模型的建立及基于模型的 DSP 算法研究··········································35

§4.1 模型的初步建立········································································ 35

§4.1.1 算法模型的组成部分······························································35

§4.1.2 Matlab-Simulink 环境下所建立的 DSP 算法模型··························· 36

§4.2 自适应滤波器的设计思路····························································· 37

§4.2.1 数字滤波器的理论算法介绍·····················································37

§4.2.2 自适应滤波器······································································· 39

§4.2.3 基于 FPGA 的自适应数字滤波器的模型建立······························· 42

§4.2.4 基于模型的自适应数字滤波器的仿真与验证······························· 43

§4.3 归一化算法·············································································· 48

§4.4 信号存储的方法比较··································································· 53

§4.5 比较算法的实现········································································· 55

§4.6 本章小结·················································································· 56

第五章复杂 DSP 算法在高速 FPGA 上的实现··········································· 57

§5.1 Xilinx-SPARTEN-3E硬件平台························································57

§5.2 硬件平台协同仿真····································································· 59

§5.3 结论························································································63

第六章总结与展望··············································································64

§6.1 总结························································································64

§6.2 展望························································································65

参考文献····························································································· 66

在读期间公开发表的论文和承担科研项目及取得成果··································· 67

致谢··································································································· 68

第一章绪论

§1.1 课题的来源及意义

§1.1.1 课题的来源

本课题是上海市科委国际合作项目（编号 051407092）“用于检测食品中化学

成分的便携式智能光谱分析仪研究”的一部分。随着人口的快速增长和工业的迅

猛发展，食品健康问题再次被提出。尤其当今社会已进入绿色环保时代，人们对

菜田受工业三废和过量化学农药的污染越来越重视，它直接关系到人们的身体健

康。主要表现在蔬菜，瓜果等食品普遍存在有害农药残留超标现象。而传统的食

品有害物质检测以化学和生化检测手段为主，这些检测手段较耗时，操作繁琐，

价格高昂，难以普及应用。为此上海理工大学联合德国 MUT 公司并得到上海市科

委的资助，共同开发一种快速先进的便携式的有害成分的光谱分析仪。主要用于

检测食品中所含化学成分或有毒含量。如何使用最先进的全光谱分析技术，并结

合光电检测方式、采用以目前世界上已广泛使用的可编程芯片 FPGA 为核心的完

全嵌入式的 DSP 信号运算与处理平台，附加同个人 PC 相交互控制的方式使人机

交互与成分分析完全可视化，免去传统对食品检测的繁复的化学处理，这是本课

题主要的研究内容和目标。

§1.1.2 课题的意义

信号运算和处理模块几乎是所有的智能化仪器和设备所必不可少的组成部

分。现场可编程门阵列（FPGA）作为当前的研究热点，已经被越来越多的应用于

信号处理中。它与传统的 DSP 处理方法相比，有着明显的优势：(1) FPGA 是高度

并行处理的引擎，对于多通道的 DSP 设计是理想的器件；(2) FPGA 的硬件可再配

置特性使其实现的高性能 DSP 具有极大的灵活性，对于所设想的算法可以定制化

结构实现；(3) FPGA 随着半导体工艺的飞速发展，价格不断降低，可以以低成本

来实现高集成度设计[1]。因此，它也理所当然地成为了本课题研究的“便携式智能

光谱分析仪”中首选的信号运算处理平台。如何为“便携式智能光谱分析仪”建

立最合理的 DSP 算法模型，以及如何把这些算法嵌入到 FPGA 中去则是本文所需

要研究的重点。

基于模型的 DSP 算法研究及高速 FPGA 实现

§1.2 FPGA 功能简介以及目前的国内外应用情况

§1.2.1 FPGA 的基本结构

图1-1 FPGA 结构简图

FPGA 全称为“现场可编程门阵列器件（Field Programmable Gate Array）”，是

实现电子设计自动化(EDA)技术的核心器件。运用 FPGA 器件可方便的配置电子系

统的硬件结构，

FPGA 器件又可嵌入高速微处理器，从而可灵活的设计出高性能的

电子软硬件系统，与 DSP 和ARM 等流行的电子设计手段相比，FPGA 在高速 DSP

算法和并行系统处理上具有不可相比的优势。运用 FPGA 设计电子系统，可方便

的进行代码自动生成，系统方案修改，还可直接使用各种 IP 核，将极大的提高产

品开发速度。

现场可编程门阵列（FPGA）是基于通过可编程互联连接的可配置逻辑块

（CLB）矩阵的可编程半导体器件。与为特殊设计而定制的专用集成电路（ASIC）

相对，FPGA 可以针对所需的应用或功能要求进行编程。当今的 FPGA 已经远远超

出了先前版本的基本性能，并且整合了常用功能（如 RAM、时钟管理和 DSP）的

硬（ASIC 型）块[2]。图 1-1 则是 FPGA 的基本结构图，它所包含的基本元件如下：

(1) 可配置逻辑块（CLB）

CLB 是FPGA 内的基本逻辑单元。实际数量和特性会依器件的不同而不同，

但是每个 CLB 都包含一个由 4或6个输入、一些选型电路（多路复用器等）和触

第一章绪论

发器组成的可配置开关矩阵。开关矩阵是高度灵活的，可以进行配置以便处理组

合逻辑、移位寄存器或 RAM。

(2) 互连

CLB 提供了逻辑性能，灵活的互联布线在 CLB 和I/O 之间发送信号。有几

种布线方法，从专门实现 CLB 互联的到快速水平和垂直长线，再到实现时钟与其

它全局信号的低歪斜发送的器件。除非特别规定，设计软件使得互联布线任务从

用户眼前消失，这样就极大地降低了设计复杂度。

(3) SelectIO（IOB）

当今的 FPGA 可支持多达 24 类I/O 标准， FPGA 内的 I/O 按组分类，每组都

能够独立的支持不同的 I/O 标准，这样就实现了 I/O 支持的灵活性。

(4) 存储器

大多数 FPGA 均提供嵌入式 Block RAM 存储器，这可以在设计中实现片上存

储器。这可以为您的设计实现片上存储器。 Xilinx FPGA 在36 kbit 块中提供高达

10 Mbits 的片上存储器，可以支持真正的双端口操作。

(5) 完整的时钟管理

大多数 FPGA 均提供数字时钟管理（Xilinx 的全部 FPGA 均具有这种特性）。

Xilinx 推出的最先进的 FPGA 提供数字时钟管理和相位环路锁定。相位环路锁定能

够提供精确的时钟综合，且能够降低抖动，并能够实现过滤功能。

§1.2.2 FPGA 的国内外应用情况

1. 国内 FPGA 研究情况

空间太阳望远镜项目是我国太阳物理学家为了实现对太阳的高分辨率观测而

提出的科学计划。它可以得到空间分辨率为 0.1"的向量磁图和 0.5"的X射线图像，

实现这样高的观测精度的前提就是采用高精度的姿态控制系统和高精度的相关跟

踪系统。从整个系统来看，相关运算所需的时间成为限制系统性能能否提高的一

个重要环节。目前，国际国内相关计算比较通用的实现方法有两种：用高速 DSP

或者专用（FFT）处理芯片。用 DSP 完成相关计算（关键是 FFT）受到航天级 DSP

性能的限制，现有的航天级 DSP（如 ADSP21020）计算一个 32×32 点8bit 的二

维FFT 所用时间需要 1.5ms 以上，远远不能满足系统设计要求；而现有的 FFT 处

理芯片在处理速度、系统兼容性、抗辐射能力等方面不能满足空间太阳望远镜所

提出的要求。为克服这一矛盾，现代科学技术人员利用 FPGA 资源丰富、易于实

现并行流水的特点设计专用的 FFT 处理芯片来完成复杂的、大量的数据处理；并

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

15 积分 4人已下载

立即下载 VIP免费下载

摘要：

摘要信号的运算和处理几乎是所有的智能化仪器必不可少的组成部分。本文针对“便携式智能光谱分析仪”的设计研究也不例外地选用了以现场可编程门阵列（FPGA）为核心的完全嵌入式的DSP信号运算与处理平台。FPGA具有高度的并行性、重构的灵活性和最佳的性价等独特的优势，因此它在信号处理方面与传统DSP处理方法相比有着巨大的改进。本文首先介绍了FPGA中的DSP硬件资源，并从内部结构上解释了FPGA是如何实现DSP的以及它的优势所在，然后根据实际设计要求，在Matlab-Simulink环境下，结合Xilinx公司所开发的插件“SystemGenerator”建立了相应的DSP算法模型。具体的模型主要包括...

展开>> 收起<<

基于模型的DSP算法研究及高速FPGA实现.pdf

共70页,预览7页

还剩页未读，继续阅读

基于模型的DSP算法研究及高速FPGA实现

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

推荐作者

热门标签

举报选择: