DSP.rar_DSP编程_嵌入式/单片机/硬件编程下载-pudn.com

<html xmlns="http://www.w3.org/1999/xhtml"> <head> <meta charset="utf-8"> <meta name="generator" content="pdf2htmlEX"> <meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1"> <link rel="stylesheet" href="https://static.pudn.com/base/css/base.min.css"> <link rel="stylesheet" href="https://static.pudn.com/base/css/fancy.min.css"> <link rel="stylesheet" href="https://static.pudn.com/prod/directory_preview_static/62635d6a4f8811599e2396ea/raw.css"> <script src="https://static.pudn.com/base/js/compatibility.min.js"></script> <script src="https://static.pudn.com/base/js/pdf2htmlEX.min.js"></script> <script> try{ pdf2htmlEX.defaultViewer = new pdf2htmlEX.Viewer({}); }catch(e){} </script> <title></title> </head> <body> <div id="sidebar" style="display: none"> <div id="outline"> </div> </div> <div id="pf1" class="pf w0 h0" data-page-no="1"><div class="pc pc1 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/62635d6a4f8811599e2396ea/bg1.jpg"><div class="t m0 x1 h2 y1 ff1 fs0 fc0 sc0 ls0 ws0"> </div><div class="t m0 x2 h3 y2 ff2 fs1 fc0 sc0 ls0 ws0">·<span class="ff1">1</span>·</div><div class="t m0 x3 h4 y3 ff3 fs2 fc0 sc0 ls0 ws0">第<span class="_ _0"> </span><span class="ff4">1<span class="_ _0"> </span></span>章<span class="ff4 ls1"> </span>概<span class="ff4 ls1"> </span>述<span class="ff4"> </span></div><div class="t m0 x4 h5 y4 ff5 fs3 fc0 sc0 ls2 ws1">1.1 <span class="ff2 sc1 ls0 ws0">引</span><span class="ls3 ws0"> <span class="ff2 sc1 ls0">言</span><span class="ls0"> </span></span></div><div class="t m0 x5 h3 y5 ff2 fs1 fc0 sc0 ls0 ws0">数字信号处理（<span class="ff1 ls4 ws2">Digital Signal Proc<span class="_ _1"></span>essing<span class="ff2 ls0 ws0">，简称<span class="_ _2"> </span><span class="ff1 ls5">DSP</span>）是一门涉及许多学科而又广泛应</span></span></div><div class="t m0 x6 h3 y6 ff2 fs1 fc0 sc0 ls0 ws0">用于许多领域的新兴学科。<span class="ff1 ls6">20<span class="_ _3"> </span></span><span class="ls7">世纪<span class="_"> </span><span class="ff1 ls6">60<span class="_ _3"> </span></span></span>年代以来，随着计算机和信息技术的飞速发展，数</div><div class="t m0 x6 h6 y7 ff2 fs1 fc0 sc0 ls8 ws0">字信号处理技术应运而生并得到迅速的发展。在过去的二十多年时间里，数字信号处理已</div><div class="t m0 x6 h3 y8 ff2 fs1 fc0 sc0 ls0 ws0">经在通信等领域得到极为广泛的应用。<span class="ff1"> </span></div><div class="t m0 x5 h6 y9 ff2 fs1 fc0 sc0 ls8 ws0">数字信号处理是利用计算机或专用处理设备，以数字形式<span class="_ _1"></span>对信号进行采集、变换、滤</div><div class="t m0 x6 h3 ya ff2 fs1 fc0 sc0 ls0 ws0">波、估值、增强、压缩、识别等处理，以得到符合人们需要的信号形式。<span class="ff1"> </span></div><div class="t m0 x5 h6 yb ff2 fs1 fc0 sc0 ls8 ws0">数字信号处理是围绕着数字信号处理的理论、实现和应用<span class="_ _1"></span>等几个方面发展起来的。数</div><div class="t m0 x6 h6 yc ff2 fs1 fc0 sc0 ls8 ws0">字信号处理在理论上的发展推动了数字信号处理应用的发展。反过来，数字信号处理的应</div><div class="t m0 x6 h6 yd ff2 fs1 fc0 sc0 ls9 ws0">用又促进了数字信号处理理论的<span class="_ _1"></span>提高。而数字信号处理<span class="_ _1"></span>的实现则是理论和应用之间的桥</div><div class="t m0 x6 h3 ye ff2 fs1 fc0 sc0 ls0 ws0">梁。<span class="ff1"> </span></div><div class="t m0 x5 h6 yf ff2 fs1 fc0 sc0 ls8 ws0">数字信号处理是以众多学科为理论基础的，它所涉及的范<span class="_ _1"></span>围极其广泛。例如，在数学</div><div class="t m0 x6 h6 y10 ff2 fs1 fc0 sc0 ls8 ws0">领域，微积分、概率统计、随机过程、数值分析等都是数字信号处理的基本工具，与网络</div><div class="t m0 x6 h6 y11 ff2 fs1 fc0 sc0 ls8 ws0">理论、信号与系统、控制论、通信理论、故障诊断等也密切相关。近来新兴的一些学科，</div><div class="t m0 x6 h6 y12 ff2 fs1 fc0 sc0 ls8 ws0">如人工智能、模式识别、神经网络等，都与数字信号处理密不可分。可以说，数字信号处</div><div class="t m0 x6 h6 y13 ff2 fs1 fc0 sc0 ls8 ws0">理是把许多经典的理论体系作为自己的理论基础，同时又使自己成为一系列新兴学科的理</div><div class="t m0 x6 h3 y14 ff2 fs1 fc0 sc0 ls0 ws0">论基础。<span class="ff1"> </span></div><div class="t m0 x5 h3 y15 ff2 fs1 fc0 sc0 ls0 ws0">数字信号处理的实现方法一般有以下几种：<span class="ff1"> </span></div><div class="t m0 x5 h3 y16 ff1 fs1 fc0 sc0 lsa ws0">(1) <span class="_"> </span><span class="ff2 ls0">在通用的计算机（如<span class="_ _2"> </span></span><span class="lsb">PC<span class="_"> </span><span class="ff2 ls0">机）上用软件（如<span class="_ _2"> </span></span><span class="lsc">Fortran<span class="ff2 ls0">、<span class="ff1">C<span class="_ _4"> </span></span>语言）实现；<span class="ff1"> </span></span></span></span></div><div class="t m0 x5 h3 y17 ff1 fs1 fc0 sc0 lsa ws0">(2) <span class="_"> </span><span class="ff2 ls0">在通用计算机系统中加上专用的加速处理机实现；<span class="ff1"> </span></span></div><div class="t m0 x5 h3 y18 ff1 fs1 fc0 sc0 lsa ws0">(3) <span class="_"> </span><span class="ff2 ls8">用通用的单片机（如<span class="_ _3"> </span></span><span class="lsd">MCS-51<span class="_ _5"></span><span class="ff2 ls0">、</span><span class="lse">96<span class="_ _6"> </span><span class="ff2 ls8">系列等）实现，这种方法可<span class="_ _1"></span>用于一些不太复杂</span></span></span></div><div class="t m0 x6 h3 y19 ff2 fs1 fc0 sc0 ls0 ws0">的数字信号处理，如数字控制等；<span class="ff1"> </span></div><div class="t m0 x5 h3 y1a ff1 fs1 fc0 sc0 lsa ws0">(4) <span class="_"> </span><span class="ff2 ls0">用通用的可编程<span class="_ _7"> </span></span><span class="lsf">DSP<span class="_ _7"> </span><span class="ff2 ls0">芯片实现。与单片机相比，</span>DSP<span class="_ _7"> </span><span class="ff2 ls0">芯片具有更加适合于数字信</span></span></div><div class="t m0 x6 h3 y1b ff2 fs1 fc0 sc0 ls0 ws0">号处理的软件和硬件资源，可用于复杂的数字信号处理算法；<span class="ff1"> </span></div><div class="t m0 x5 h3 y1c ff1 fs1 fc0 sc0 lsa ws0">(5) <span class="_"> </span><span class="ff2 ls0">用专用的<span class="_ _0"> </span></span><span class="lsf">DSP<span class="_ _3"> </span><span class="ff2 ls0">芯片实现。在一些特殊的场合，要求的信号处理速度极高，用通用</span></span></div><div class="t m0 x6 h3 y1d ff1 fs1 fc0 sc0 lsf ws0">DSP<span class="_"> </span><span class="ff2 ls0">芯片很难实现，例如专用于<span class="_ _4"> </span></span><span class="lsb">FFT<span class="ff2 ls0">、数字滤波、卷积、相关等算法的<span class="_ _2"> </span></span></span>DSP<span class="_"> </span><span class="ff2 ls0">芯片，<span class="_ _5"></span>这种芯</span></div><div class="t m0 x6 h3 y1e ff2 fs1 fc0 sc0 ls0 ws0">片将相应的信号处理算法在芯片内部用硬件实现，无需进行编程。<span class="ff1"> </span></div><div class="t m0 x5 h3 y1f ff2 fs1 fc0 sc0 ls0 ws0">在上述几种方法中，第<span class="_ _2"> </span><span class="ff1">1<span class="_ _4"> </span></span>种方法的缺点是速度较慢，一般可用于<span class="_ _4"> </span><span class="ff1 lsf">DSP<span class="_"> </span></span>算法的模拟；第</div><div class="t m0 x6 h3 y20 ff1 fs1 fc0 sc0 ls0 ws0">2<span class="_ _6"> </span><span class="ff2 ls8">种和第<span class="_"> </span></span>5<span class="_ _3"> </span><span class="ff2 ls8">种方法专用性强，应用受到很大的限制，第<span class="_"> </span></span>2<span class="_ _6"> </span><span class="ff2 ls8">种方法也不便于系统的独立运</span></div><div class="t m0 x6 h3 y21 ff2 fs1 fc0 sc0 ls0 ws0">行；第<span class="_ _2"> </span><span class="ff1">3<span class="_ _4"> </span></span>种方法只适用于实现简单的<span class="_ _4"> </span><span class="ff1 lsf">DSP<span class="_"> </span></span>算法；只有第<span class="_ _4"> </span><span class="ff1">4<span class="_"> </span></span>种方法才使数字信号处理的应用</div><div class="t m0 x6 h3 y22 ff2 fs1 fc0 sc0 ls0 ws0">打开了新的局面。<span class="ff1"> </span></div><div class="t m0 x5 h3 y23 ff2 fs1 fc0 sc0 ls8 ws0">虽然数字信号处理的理论发展迅速，但在<span class="_"> </span><span class="ff1 lse">20<span class="_ _6"> </span></span>世纪<span class="_"> </span><span class="ff1 lse">80<span class="_ _6"> </span></span>年<span class="_ _1"></span>代以前，由于实现方法的限</div></div><div class="pi" data-data='{"ctm":[1.611639,0.000000,0.000000,1.611639,0.000000,0.000000]}'></div></div> </body> </html>

<div id="pf2" class="pf w0 h0" data-page-no="2"><div class="pc pc2 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/62635d6a4f8811599e2396ea/bg2.jpg"><div class="t m0 x5 h2 y24 ff1 fs0 fc0 sc0 ls0 ws0"> </div><div class="t m0 x6 h3 y2 ff2 fs1 fc0 sc0 ls0 ws0">·<span class="ff1">2</span>·<span class="ff1"> </span></div><div class="t m0 x6 h3 y25 ff2 fs1 fc0 sc0 ls0 ws0">制，数字信号处理的理论还得不到广泛的应用。直到<span class="_ _8"> </span><span class="ff1 ls6">20<span class="_ _8"> </span></span>世纪<span class="_ _8"> </span><span class="ff1 ls6">70<span class="_ _8"> </span></span>年代末<span class="_ _8"> </span><span class="ff1 ls6">80<span class="_ _8"> </span></span>年代初世界上</div><div class="t m0 x6 h3 y26 ff2 fs1 fc0 sc0 ls0 ws0">第一片单片可编程<span class="_ _4"> </span><span class="ff1 lsf">DSP<span class="_ _9"> </span></span>芯片的诞生，才将理论研究结果广泛应用到低成本的实际系统中，</div><div class="t m0 x6 h3 y27 ff2 fs1 fc0 sc0 ls0 ws0">并且推动了新的理论和应用领<span class="_ _5"></span>域的发展。可以毫不夸张地<span class="_ _5"></span>说，<span class="ff1 ls5">DSP<span class="_ _6"> </span></span>芯片的诞生及发展对近</div><div class="t m0 x6 h3 y28 ff1 fs1 fc0 sc0 ls6 ws0">20<span class="_"> </span><span class="ff2 ls0">年来通信、计算机、控制等领域的技术发展起到十分重要的作用。<span class="ff1"> </span></span></div><div class="t m0 x7 h5 y29 ff5 fs3 fc0 sc0 ls10 ws3">1.2 DSP<span class="ff2 sc1 ls11 ws0">系统</span><span class="ls0 ws0"> </span></div><div class="t m0 x5 h7 y2a ff4 fs4 fc0 sc0 ls12 ws4">1.2.1 DSP</div><div class="t m0 x8 h8 y2b ff3 fs4 fc0 sc0 ls0 ws0">系统构成</div><div class="t m0 x9 h7 y2a ff4 fs4 fc0 sc0 ls0 ws0"> </div><div class="t m0 x5 h3 y2c ff2 fs1 fc0 sc0 ls0 ws0">图<span class="_ _9"> </span><span class="ff1 ls13">1.1<span class="_ _7"> </span></span>所示为一个典型的<span class="_ _9"> </span><span class="ff1 lsf">DSP<span class="_ _9"> </span></span>系统。图中的输入信号可以有各种各样的形式。例如，</div><div class="t m0 x6 h6 y2d ff2 fs1 fc0 sc0 ls8 ws0">它可以是麦克风输出的语音信号或是电话线来的已调数据信号，可以是编码后在数字链路</div><div class="t m0 x6 h3 y2e ff2 fs1 fc0 sc0 ls0 ws0">上传输或存储在计算机里的摄像机图像信号等。<span class="ff1"> </span></div><div class="t m0 xa h9 y2f ff1 fs5 fc0 sc0 ls14 ws0"> </div><div class="t m0 xb ha y30 ff2 fs5 fc0 sc0 ls15 ws0">输入</div><div class="t m0 xc h9 y31 ff1 fs5 fc0 sc0 ls0 ws0"> </div><div class="t m0 xd ha y32 ff2 fs5 fc0 sc0 ls15 ws0">抗混叠</div><div class="t m0 xc h9 y33 ff1 fs5 fc0 sc0 ls14 ws0"> </div><div class="t m0 xe ha y34 ff2 fs5 fc0 sc0 ls16 ws0">滤波</div><div class="t m0 xf h9 y35 ff1 fs5 fc0 sc0 ls17 ws0">A/<span class="_ _5"></span>D</div><div class="t m0 x10 h9 y36 ff1 fs5 fc0 sc0 ls17 ws0">DSP</div><div class="t m0 x10 ha y37 ff2 fs5 fc0 sc0 ls15 ws0">芯片</div><div class="t m0 x11 h9 y38 ff1 fs5 fc0 sc0 ls17 ws0">D/<span class="_ _5"></span>A</div><div class="t m0 x12 hb y39 ff1 fs6 fc0 sc0 ls0 ws0"> </div><div class="t m0 x13 ha y3a ff2 fs5 fc0 sc0 ls15 ws0">平滑</div><div class="t m0 x12 hb y3b ff1 fs6 fc0 sc0 ls0 ws0"> </div><div class="t m0 x13 ha y3c ff2 fs5 fc0 sc0 ls15 ws0">滤波</div><div class="t m0 x14 ha y3d ff2 fs5 fc0 sc0 ls15 ws0">输出</div><div class="t m0 x15 hc y3e ff1 fs7 fc0 sc0 ls0 ws0"> </div><div class="t m0 x16 hd y3f ff2 fs0 fc0 sc0 ls0 ws0">图</div><div class="t m0 x17 h2 y40 ff1 fs0 fc0 sc0 ls18 ws5">1.1 </div><div class="t m0 x18 hd y3f ff2 fs0 fc0 sc0 ls0 ws0">典型的</div><div class="t m0 x19 h2 y40 ff1 fs0 fc0 sc0 ls19 ws0">DSP</div><div class="t m0 x1a hd y3f ff2 fs0 fc0 sc0 ls0 ws0">系统</div><div class="t m0 x1b h2 y40 ff1 fs0 fc0 sc0 ls0 ws0"> </div><div class="t m0 x5 h3 y41 ff2 fs1 fc0 sc0 ls0 ws0">输入信号首先进行带限滤波和抽样，然后进行<span class="_ _2"> </span><span class="ff1">A/D</span>（<span class="ff1 ls1a ws6">Analog to Digital</span>）变换将信号变</div><div class="t m0 x6 h6 y42 ff2 fs1 fc0 sc0 ls8 ws0">换成数字比特流。根据奈奎斯特抽样定理，为保证信息不丢失，抽样频率至少必须是输入</div><div class="t m0 x6 h3 y43 ff2 fs1 fc0 sc0 ls0 ws0">带限信号最高频率的<span class="_ _2"> </span><span class="ff1">2<span class="_"> </span></span>倍。<span class="ff1"> </span></div><div class="t m0 x5 h3 y44 ff1 fs1 fc0 sc0 lsf ws0">DSP<span class="_"> </span><span class="ff2 ls0">芯片的输入是<span class="_ _2"> </span><span class="ff1">A/D<span class="_"> </span></span>变换后得到的以抽样形式表示的数字信号，</span><span class="ls5">DSP<span class="_ _4"> </span><span class="ff2 ls0">芯片对输入的</span></span></div><div class="t m0 x6 h3 y45 ff2 fs1 fc0 sc0 ls1b ws0">数字信号进行某种形式的处理，如进行一系列的乘累加操作（<span class="ff1 lsd">MAC</span>）。数字处理是<span class="_"> </span><span class="ff1 lsf">DSP</span></div><div class="t m0 x6 h6 y46 ff2 fs1 fc0 sc0 ls8 ws0">的关键，这与其他系统（如电话交换系统）有很大的不同，在交换系统中，处理器的作用</div><div class="t m0 x6 h6 y47 ff2 fs1 fc0 sc0 ls8 ws0">是进行路由选择，它并不对输入数据进行修改。因此虽然两者都是实时系统，但两者的实</div><div class="t m0 x6 h3 y48 ff2 fs1 fc0 sc0 ls0 ws0">时约束条件却有很大的不同。最后，经过处理后的数字样值再经<span class="_ _2"> </span><span class="ff1">D/A</span>（<span class="ff1 ls1c ws7">Digital to Analog</span>）</div><div class="t m0 x6 h3 y49 ff2 fs1 fc0 sc0 ls0 ws0">变换转换为模拟样值，之后再进行内插和平滑滤波就可得到连续的模拟波形。<span class="ff1"> </span></div><div class="t m0 x5 h3 y4a ff2 fs1 fc0 sc0 ls0 ws0">必须指出的是，上面给出的<span class="_ _0"> </span><span class="ff1 ls5">DSP<span class="_ _0"> </span></span>系统模型是一个典型模型，但并不是所有的<span class="_ _0"> </span><span class="ff1 ls5">DSP<span class="_ _3"> </span></span>系</div><div class="t m0 x6 h6 y4b ff2 fs1 fc0 sc0 ls8 ws0">统都必须具有模型中的所有部件。如语音识别系统在输出端并不是连续的波形，而是识别</div><div class="t m0 x6 h3 y4c ff2 fs1 fc0 sc0 ls0 ws0">结果，如数字、文字等；有些输入信号本身就是数字信号（如<span class="_ _2"> </span><span class="ff1 ls1c">CD</span>：<span class="ff1 ls1d ws8">Compact Disk</span>），因此</div><div class="t m0 x6 h3 y4d ff2 fs1 fc0 sc0 ls0 ws0">就不必进行模数变换了。<span class="ff1"> </span></div><div class="t m0 x5 h7 y4e ff4 fs4 fc0 sc0 ls12 ws4">1.2.2 DSP</div><div class="t m0 x8 h8 y4f ff3 fs4 fc0 sc0 ls0 ws0">系统的特点</div><div class="t m0 x1c h7 y4e ff4 fs4 fc0 sc0 ls0 ws0"> </div><div class="t m0 x5 h3 y50 ff2 fs1 fc0 sc0 ls0 ws0">数字信号处理系统是以数字信号处理为基础，因此具有数字处理的全部优点：<span class="ff1"> </span></div><div class="t m0 x5 h3 y51 ff1 fs1 fc0 sc0 lsa ws0">(1) <span class="_ _4"> </span><span class="ff2 ls9">接口方<span class="_ _5"></span>便。</span><span class="ls5">DSP<span class="_ _a"> </span><span class="ff2 ls9">系统与其他以现代数字技<span class="_ _5"></span>术为基础的系统或设备都是相互<span class="_ _5"></span>兼容</span></span></div><div class="t m0 x6 h3 y52 ff2 fs1 fc0 sc0 ls0 ws0">的，与这样的系统接口以实现某种功能要比模拟系统与这些系统接口要容易得多；<span class="ff1"> </span></div><div class="t m0 x5 h3 y53 ff1 fs1 fc0 sc0 lsa ws0">(2) <span class="_"> </span><span class="ff2 ls0">编程方便。</span><span class="lsf">DSP<span class="_ _7"> </span><span class="ff2 ls0">系统中的可编程<span class="_ _7"> </span></span>DSP<span class="_ _7"> </span><span class="ff2 ls0">芯片可使设计人员在开发过程中灵活方便地</span></span></div><div class="t m0 x6 h3 y54 ff2 fs1 fc0 sc0 ls0 ws0">对软件进行修改和升级；<span class="ff1"> </span></div></div><div class="pi" data-data='{"ctm":[1.611639,0.000000,0.000000,1.611639,0.000000,0.000000]}'></div></div>

<div id="pf3" class="pf w0 h0" data-page-no="3"><div class="pc pc3 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/62635d6a4f8811599e2396ea/bg3.jpg"><div class="t m0 x1 h2 y1 ff1 fs0 fc0 sc0 ls0 ws0"> </div><div class="t m0 x2 h3 y2 ff2 fs1 fc0 sc0 ls0 ws0">·<span class="ff1">3</span>·</div><div class="t m0 x5 h3 y25 ff1 fs1 fc0 sc0 lsa ws0">(3) <span class="_"> </span><span class="ff2 ls8">稳定性好。</span><span class="lsf">DSP<span class="_ _6"> </span><span class="ff2 ls8">系统以数字处理为基础，受环境温度以及噪声的影响较小，可靠</span></span></div><div class="t m0 x6 h3 y26 ff2 fs1 fc0 sc0 ls0 ws0">性高；<span class="ff1"> </span></div><div class="t m0 x5 h3 y55 ff1 fs1 fc0 sc0 lsa ws0">(4) <span class="_"> </span><span class="ff2 ls0">精度高。</span><span class="ls6">16<span class="_"> </span><span class="ff2 ls0">位数字系统可以达到</span></span></div><div class="t m0 x10 he y56 ff1 fs8 fc0 sc0 ls0 ws0">5</div><div class="t m0 x1d hf y57 ff1 fs9 fc0 sc0 ls0 ws0">10</div><div class="t m0 x1e h10 y56 ff6 fs8 fc0 sc0 ls0 ws0">−</div><div class="t m0 x1f hf y58 ff2 fs1 fc0 sc0 ls0 ws0">的精度；<span class="ff1 fs9"> </span></div><div class="t m0 x5 hf y59 ff1 fs9 fc0 sc0 ls1e ws0">(5) </div><div class="t m0 x20 h6 y5a ff2 fs1 fc0 sc0 ls1b ws0">可重复性好。模拟系统的性能受<span class="_ _1"></span>元器件参数性能变化比<span class="_ _1"></span>较大，而数字系统基本<span class="_ _1"></span>不</div><div class="t m0 x6 h6 y5b ff2 fs1 fc0 sc0 ls0 ws0">受影响，因此数字系统便于测试、调试和大规模生产；</div><div class="t m0 x1b hf y5c ff1 fs9 fc0 sc0 ls0 ws0"> </div><div class="t m0 x5 hf y5d ff1 fs9 fc0 sc0 ls1e ws0">(6) </div><div class="t m0 x20 hf y5e ff2 fs1 fc0 sc0 ls0 ws0">集成方便。<span class="ff1 fs9 ls1f">DSP<span class="_"> </span></span>系统中的数字部件有高度的规范性，便于大规模集成。<span class="ff1 fs9"> </span></div><div class="t m0 x5 h6 y5f ff2 fs1 fc0 sc0 ls8 ws0">当然，数字信号处理也存在一定的缺点。例如，对于简单<span class="_ _1"></span>的信号处理任务，如与模拟</div><div class="t m0 x6 h6 y60 ff2 fs1 fc0 sc0 ls20 ws0">交换线的电话接口，若采用</div><div class="t m0 x9 hf y61 ff1 fs9 fc0 sc0 ls1f ws0">DSP<span class="_ _6"> </span><span class="ff2 fs1 ls20">则使成本增加。</span>DSP<span class="_ _6"> </span><span class="ff2 fs1 ls20">系统中的高速时钟可能带来高频干</span></div><div class="t m0 x6 h6 y62 ff2 fs1 fc0 sc0 ls21 ws0">扰和电磁泄漏等问题，而且</div><div class="t m0 x21 hf y63 ff1 fs9 fc0 sc0 ls1f ws0">DSP<span class="_ _a"> </span><span class="ff2 fs1 ls21">系统消耗的功率也较大。此外<span class="_ _5"></span>，</span><span class="ls22">DSP<span class="_ _a"> </span><span class="ff2 fs1 ls21">技术更新的速度</span></span></div><div class="t m0 x6 h6 y64 ff2 fs1 fc0 sc0 ls0 ws0">快，数学知识要求多，开发和调试工具还不尽完善。</div><div class="t m0 x22 hf y65 ff1 fs9 fc0 sc0 ls0 ws0"> </div><div class="t m0 x5 hf y66 ff2 fs1 fc0 sc0 ls23 ws0">虽然<span class="_"> </span><span class="ff1 fs9 ls24">DSP<span class="_ _6"> </span></span>系统存在着一些缺点，但其突出的<span class="_ _1"></span>优点已经使之在通信、语音、图像、雷</div><div class="t m0 x6 h6 y67 ff2 fs1 fc0 sc0 ls0 ws0">达、生物医学、工业控制、仪器仪表等许多领域得到越来越广泛的应用。</div><div class="t m0 x23 hf y68 ff1 fs9 fc0 sc0 ls0 ws0"> </div><div class="t m0 x5 h7 y69 ff4 fs4 fc0 sc0 ls12 ws4">1.2.3 DSP</div><div class="t m0 x8 h8 y6a ff3 fs4 fc0 sc0 ls0 ws0">系统的设计过程</div><div class="t m0 x24 h7 y69 ff4 fs4 fc0 sc0 ls0 ws0"> </div><div class="t m0 x5 hf y6b ff2 fs1 fc0 sc0 ls0 ws0">总的来说，<span class="ff1 fs9 ls1f">DSP<span class="_ _4"> </span></span>系统的设计还没有非常好的正规设计方法。图<span class="_ _9"> </span><span class="ff1 fs9 ls25">1.2<span class="_ _9"> </span></span>所示是<span class="_ _4"> </span><span class="ff1 fs9 ls1f">DSP<span class="_ _9"> </span></span>系统设</div><div class="t m0 x6 h6 y6c ff2 fs1 fc0 sc0 ls0 ws0">计的一般过程。</div><div class="t m0 x25 hf y6d ff1 fs9 fc0 sc0 ls0 ws0"> </div><div class="t m0 x26 h11 y6e ff1 fsa fc0 sc0 ls26 ws9"> <span class="fc1 sc0"> </span></div><div class="t m0 x27 h12 y6f ff2 fsb fc0 sc0 ls27 ws0">软件编程</div><div class="t m0 x28 h12 y70 ff2 fsb fc0 sc0 ls27 ws0">硬件设计</div><div class="t m0 x29 h12 y71 ff2 fsb fc0 sc0 ls28 ws0">软件调试</div><div class="t m0 x2a h12 y72 ff2 fsb fc0 sc0 ls27 ws0">硬件调试</div><div class="t m0 x2b h12 y73 ff2 fsb fc0 sc0 ls28 ws0">选择</div><div class="t m0 x2c h13 y74 ff1 fsb fc0 sc0 ls29 ws0">DS<span class="_ _1"></span>P</div><div class="t m0 x2d h12 y73 ff2 fsb fc0 sc0 ls27 ws0">芯片</div><div class="t m0 x2e h12 y75 ff2 fsb fc0 sc0 ls27 ws0">定义系统性能指标</div><div class="t m0 x2f h13 y76 ff1 fsb fc0 sc0 ls2a ws0"> </div><div class="t m0 x30 h12 y77 ff2 fsb fc0 sc0 ls27 ws0">系统集成</div><div class="t m0 x31 h12 y78 ff2 fsb fc0 sc0 ls27 ws0">系统测试和调试</div><div class="t m0 x18 h13 y79 ff1 fsb fc0 sc0 ls29 ws0">DS<span class="_ _1"></span>P</div><div class="t m0 x32 h12 y7a ff2 fsb fc0 sc0 ls27 ws0">应用</div><div class="t m0 x33 hf y7b ff1 fs9 fc0 sc0 ls0 ws0"> </div><div class="t m0 x34 hd y7c ff2 fs0 fc0 sc0 ls0 ws0">图</div><div class="t m0 x16 h2 y7d ff1 fs0 fc0 sc0 ls2b wsa">1.2 DSP</div><div class="t m0 x35 hd y7c ff2 fs0 fc0 sc0 ls0 ws0">系统的设计流程</div><div class="t m0 x36 h2 y7d ff1 fs0 fc0 sc0 ls0 ws0"> </div><div class="t m0 x5 hf y7e ff2 fs1 fc0 sc0 ls23 ws0">在设计<span class="ff1 fs9 ls2c wsb"> DSP </span>系统之前，首先必须根据应用系统的目标确定<span class="_ _1"></span>系统的性能指标、信号处</div><div class="t m0 x6 h6 y7f ff2 fs1 fc0 sc0 ls0 ws0">理的要求，通常可用数据流程图、数学运算序列、正式的符号或自然语言来描述。</div><div class="t m0 x37 hf y80 ff1 fs9 fc0 sc0 ls0 ws0"> </div><div class="t m0 x5 h6 y81 ff2 fs1 fc0 sc0 ls9 ws0">第二步是根据系统的要求进行高级语言的模拟。一般来说，为了实现系统的最终目</div><div class="t m0 x6 h6 y82 ff2 fs1 fc0 sc0 ls8 ws0">标，需要对输入的信号进行适当的处理，而处理方法的不同会导致不同的系统性能，要得</div><div class="t m0 x6 h6 y83 ff2 fs1 fc0 sc0 ls2d ws0">到最佳的系统性能，就必须在这一步确定最佳的处理方法，即数字信号处理的算法</div></div><div class="pi" data-data='{"ctm":[1.611639,0.000000,0.000000,1.611639,0.000000,0.000000]}'></div></div>

<div id="pf4" class="pf w0 h0" data-page-no="4"><div class="pc pc4 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/62635d6a4f8811599e2396ea/bg4.jpg"><div class="t m0 x5 h2 y24 ff1 fs0 fc0 sc0 ls0 ws0"> </div><div class="t m0 x6 hf y2 ff2 fs1 fc0 sc0 ls0 ws0">·<span class="ff1 fs9">4</span>·<span class="ff1 fs9"> </span></div><div class="t m0 x6 hf y84 ff2 fs1 fc0 sc0 ls0 ws0">（<span class="ff1 fs9 ls2e">Algorithm</span><span class="ls20">），因此这一步也称算法模拟阶段。例如，语音压缩编码<span class="_ _5"></span>算法就是要在确定的</span></div><div class="t m0 x6 h6 y85 ff2 fs1 fc0 sc0 ls8 ws0">压缩比条件下，获得最佳的合成语音。算法模拟所用的输入数据是实际信号经采集而获得</div><div class="t m0 x6 h6 y86 ff2 fs1 fc0 sc0 ls8 ws0">的，通常以计算机文件的形式存储为数据文件。如语音压缩编码算法模拟时所用的语音信</div><div class="t m0 x6 h6 y87 ff2 fs1 fc0 sc0 ls8 ws0">号就是实际采集而获得并存储为计算机文件形式的语音数据文件。有些算法模拟时所用的</div><div class="t m0 x6 h6 y88 ff2 fs1 fc0 sc0 ls8 ws0">输入数据并不一定要是实际采集的信号数据，只要能够验证算法的可行性，输入假设的数</div><div class="t m0 x6 h6 y89 ff2 fs1 fc0 sc0 ls0 ws0">据也是可以的。</div><div class="t m0 x25 hf y8a ff1 fs9 fc0 sc0 ls0 ws0"> </div><div class="t m0 x5 hf y8b ff2 fs1 fc0 sc0 ls0 ws0">在完成第二步之后，接下来就可以设计实时<span class="_ _0"> </span><span class="ff1 fs9 ls24">DSP<span class="_ _0"> </span></span>系统，实时<span class="_ _0"> </span><span class="ff1 fs9 ls24">DSP<span class="_ _0"> </span></span>系统<span class="_ _5"></span>的设计包括硬</div><div class="t m0 x6 h6 y8c ff2 fs1 fc0 sc0 ls9 ws0">件设计和软件设计两个方面。硬<span class="_ _1"></span>件设计首先要根据系统<span class="_ _1"></span>运算量的大小、对运算精度的要</div><div class="t m0 x6 h6 y8d ff2 fs1 fc0 sc0 ls0 ws0">求、系统成本限制以及体积、功耗等要求选择合适的</div><div class="t m0 x38 hf y8e ff1 fs9 fc0 sc0 ls24 ws0">DSP<span class="_ _0"> </span><span class="ff2 fs1 ls0">芯片。然后设计<span class="_ _3"> </span></span>DSP<span class="_ _0"> </span><span class="ff2 fs1 ls0">芯片的外</span></div><div class="t m0 x6 h6 y8f ff2 fs1 fc0 sc0 ls2f ws0">围电路及其他电路。软件设计和编程主要根据系统要求和<span class="_ _1"></span>所选的</div><div class="t m0 x39 hf y90 ff1 fs9 fc0 sc0 ls22 ws0">DSP<span class="_ _a"> </span><span class="ff2 fs1 ls2f">芯片编写相应的</span></div><div class="t m0 x6 hf y91 ff1 fs9 fc0 sc0 ls1f ws0">DSP<span class="_ _6"> </span><span class="ff2 fs1 ls8">汇编程序，若系统运算量不大且有高级语言编译器支持，也可用高级语言（如<span class="_"> </span></span><span class="ls0">C<span class="_ _6"> </span><span class="ff2 fs1">语</span></span></div><div class="t m0 x6 h6 y92 ff2 fs1 fc0 sc0 ls8 ws0">言）编程。由于现有的高级语言编译器的效率还比不上手工编写汇编语言的效率，因此在</div><div class="t m0 x6 h6 y93 ff2 fs1 fc0 sc0 ls9 ws0">实际应用系统中常常采用高级语<span class="_ _1"></span>言和汇编语言的混合编<span class="_ _1"></span>程方法，即在算法运算量大的地</div><div class="t m0 x6 h6 y94 ff2 fs1 fc0 sc0 ls8 ws0">方，用手工编写的方法编写汇编语言，而运算量不大的地方则采用高级语言。采用这种方</div><div class="t m0 x6 h6 y95 ff2 fs1 fc0 sc0 ls8 ws0">法，既可缩短软件开发的周期，提高程序的可读性和可移植性，又能满足系统实时运算的</div><div class="t m0 x6 h6 y96 ff2 fs1 fc0 sc0 ls0 ws0">要求。</div><div class="t m0 x3a hf y97 ff1 fs9 fc0 sc0 ls0 ws0"> </div><div class="t m0 x5 hf y98 ff1 fs9 fc0 sc0 ls1f ws0">DSP</div><div class="t m0 x3b h6 y99 ff2 fs1 fc0 sc0 ls0 ws0">硬件和软件设计完成后，就需要进行硬件<span class="_ _5"></span>和软件的调试。软件的调试一般借助<span class="_ _5"></span>于</div><div class="t m0 x6 hf y9a ff1 fs9 fc0 sc0 ls1f ws0">DSP<span class="_ _3"> </span><span class="ff2 fs1 ls0">开发工具，如软件模拟器、</span><span class="ls22">DS<span class="_ _5"></span>P<span class="_ _3"> </span><span class="ff2 fs1 ls0">开发系统或仿真器等。调试<span class="_"> </span></span></span>DSP<span class="_ _3"> </span><span class="ff2 fs1 ls0">算法时一般采用比</span></div><div class="t m0 x6 h6 y9b ff2 fs1 fc0 sc0 ls8 ws0">较实时结果与模拟结果的方法，如果实时程序和模拟程序的输入相同，则两者的输出应该</div><div class="t m0 x6 h6 y9c ff2 fs1 fc0 sc0 ls8 ws0">一致。应用系统的其他软件可以根据实际情况进行调试。硬件调试一般采用硬件仿真器进</div><div class="t m0 x6 h6 y9d ff2 fs1 fc0 sc0 ls8 ws0">行调试，如果没有相应的硬件仿真器，且硬件系统不是十分复杂，也可以借助于一般的工</div><div class="t m0 x6 h6 y9e ff2 fs1 fc0 sc0 ls0 ws0">具进行调试。</div><div class="t m0 x3c hf y9f ff1 fs9 fc0 sc0 ls0 ws0"> </div><div class="t m0 x5 h6 ya0 ff2 fs1 fc0 sc0 ls8 ws0">系统的软件和硬件分别调试完成后，就可以将软件脱离开<span class="_ _1"></span>发系统而直接在应用系统上</div><div class="t m0 x6 h6 ya1 ff2 fs1 fc0 sc0 ls0 ws0">运行。当然，</div><div class="t m0 x3c hf ya2 ff1 fs9 fc0 sc0 ls1f ws0">DSP<span class="_ _6"> </span><span class="ff2 fs1 ls0">系统的开发，特别是软件开发是一个需<span class="_ _5"></span>要反复进行的过程，虽<span class="_ _5"></span>然通过算</span></div><div class="t m0 x6 h6 ya3 ff2 fs1 fc0 sc0 ls8 ws0">法模拟基本上可以知道实时系统的性能，但实际上模拟环境不可能做到与实时系统环境完</div><div class="t m0 x6 h6 ya4 ff2 fs1 fc0 sc0 ls8 ws0">全一致，而且将模拟算法移植到实时系统时必须考虑算法是否能够实时运行的问题。如果</div><div class="t m0 x6 h6 ya5 ff2 fs1 fc0 sc0 ls0 ws0">算法运算量太大不能在硬件上实时运行，则必须重新修改或简化算法。</div><div class="t m0 x3d hf ya6 ff1 fs9 fc0 sc0 ls0 ws0"> </div><div class="t m0 x3e h5 ya7 ff5 fs3 fc0 sc0 ls2 ws1">1.3 <span class="ff2 sc1 ls11 ws0">可编程</span><span class="ls30 ws0">DSP<span class="ff2 sc1 ls11">芯片</span><span class="ls0"> </span></span></div><div class="t m0 x5 h7 ya8 ff4 fs4 fc0 sc0 ls31 wsc">1.3.1 </div><div class="t m0 xa h8 ya9 ff3 fs4 fc0 sc0 ls0 ws0">什么是</div><div class="t m0 x3f h7 ya8 ff4 fs4 fc0 sc0 ls32 ws0">DSP</div><div class="t m0 x40 h8 ya9 ff3 fs4 fc0 sc0 ls33 ws0">芯片</div><div class="t m0 x1c h7 ya8 ff4 fs4 fc0 sc0 ls0 ws0"> </div><div class="t m0 x5 hf yaa ff1 fs9 fc0 sc0 ls1f ws0">DSP<span class="_ _6"> </span><span class="ff2 fs1 ls0">芯片，也称数字信号处理器，是一种特别适合于进行数字信号处理运算的微<span class="_ _5"></span>处理</span></div><div class="t m0 x6 h6 yab ff2 fs1 fc0 sc0 ls9 ws0">器，其主要应用是实时快速地实<span class="_ _1"></span>现各种数字信号处理算<span class="_ _1"></span>法。根据数字信号处理的要求，</div><div class="t m0 x6 hf yac ff1 fs9 fc0 sc0 ls1f ws0">DSP<span class="_"> </span><span class="ff2 fs1 ls0">芯片一般具有如下主要特点：</span><span class="ls0"> </span></div><div class="t m0 x5 hf yad ff1 fs9 fc0 sc0 ls1e ws0">(1) </div><div class="t m0 x20 hf yae ff2 fs1 fc0 sc0 ls0 ws0">在一个指令周期内可完成一次乘法和一次加法；<span class="ff1 fs9"> </span></div><div class="t m0 x5 hf yaf ff1 fs9 fc0 sc0 ls1e ws0">(2) </div><div class="t m0 x20 hf yb0 ff2 fs1 fc0 sc0 ls0 ws0">程序和数据空间分开，可以同时访问指令和数据；<span class="ff1 fs9"> </span></div><div class="t m0 x5 hf yb1 ff1 fs9 fc0 sc0 ls1e ws0">(3) </div><div class="t m0 x20 hf yb2 ff2 fs1 fc0 sc0 ls0 ws0">片内具有快速<span class="ff1 fs9 ls34 wsd"> RAM</span>，通常可通过独立的数据总线在两块中同时访问；<span class="ff1 fs9"> </span></div><div class="t m0 x5 hf yb3 ff1 fs9 fc0 sc0 ls1e ws0">(4) </div><div class="t m0 x20 hf yb4 ff2 fs1 fc0 sc0 ls0 ws0">具有低开销或无开销循环及跳转的硬件支持；<span class="ff1 fs9"> </span></div></div><div class="pi" data-data='{"ctm":[1.611639,0.000000,0.000000,1.611639,0.000000,0.000000]}'></div></div>

<div id="pf5" class="pf w0 h0" data-page-no="5"><div class="pc pc5 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/62635d6a4f8811599e2396ea/bg5.jpg"><div class="t m0 x1 h2 y1 ff1 fs0 fc0 sc0 ls0 ws0"> </div><div class="t m0 x2 hf y2 ff2 fs1 fc0 sc0 ls0 ws0">·<span class="ff1 fs9">5</span>·</div><div class="t m0 x5 hf y84 ff1 fs9 fc0 sc0 ls1e ws0">(5) <span class="_"> </span><span class="ff2 fs1 ls0">快速的中断处理和硬件<span class="_ _2"> </span></span><span class="ls35">I/O<span class="_"> </span><span class="ff2 fs1 ls0">支持；</span><span class="ls0"> </span></span></div><div class="t m0 x5 hf yb5 ff1 fs9 fc0 sc0 ls1e ws0">(6) </div><div class="t m0 x20 hf yb6 ff2 fs1 fc0 sc0 ls0 ws0">具有在单周期内操作的多个硬件地址产生器；<span class="ff1 fs9"> </span></div><div class="t m0 x5 hf yb7 ff1 fs9 fc0 sc0 ls1e ws0">(7) </div><div class="t m0 x20 hf yb8 ff2 fs1 fc0 sc0 ls0 ws0">可以并行执行多个操作；<span class="ff1 fs9"> </span></div><div class="t m0 x5 hf yb9 ff1 fs9 fc0 sc0 ls1e ws0">(8) </div><div class="t m0 x20 hf yba ff2 fs1 fc0 sc0 ls0 ws0">支持流水线操作，使取指、译码和执行等操作可以重叠执行。<span class="ff1 fs9"> </span></div><div class="t m0 x5 hf ybb ff2 fs1 fc0 sc0 ls0 ws0">当然，与通用微处理器相比，<span class="ff1 fs9 ls1f">DSP<span class="_"> </span></span>芯片的其他通用功能相对较弱些。<span class="ff1 fs9"> </span></div><div class="t m0 x5 h7 ybc ff4 fs4 fc0 sc0 ls12 ws4">1.3.2 DSP</div><div class="t m0 x8 h8 ybd ff3 fs4 fc0 sc0 ls0 ws0">芯片的发展</div><div class="t m0 x1c h7 ybc ff4 fs4 fc0 sc0 ls0 ws0"> </div><div class="t m0 x5 hf ybe ff2 fs1 fc0 sc0 ls0 ws0">世界上第一个单片<span class="ff1 fs9 ls1f wse"> DSP <span class="_ _5"></span></span>芯片应当是<span class="_"> </span><span class="ff1 fs9 ls25">1978<span class="_ _6"> </span></span>年<span class="ff1 fs9 ls36 wsf"> AMI<span class="_"> </span></span>公司发布的<span class="ff1 fs9 ls37 ws10"> S2811</span>，<span class="ff1 fs9 ls38">1979<span class="_ _6"> </span></span>年美国</div><div class="t m0 x6 hf ybf ff1 fs9 fc0 sc0 ls39 ws0">Intel<span class="_ _7"> </span><span class="ff2 fs1 ls0">公司发布的商用可编程器件<span class="_ _b"> </span></span><span class="ls25">2920<span class="_ _b"> </span><span class="ff2 fs1 ls0">是<span class="_ _7"> </span></span><span class="ls1f">DSP<span class="_ _b"> </span><span class="ff2 fs1 ls0">芯片的一个主要里程碑。这两种芯片内部都</span></span></span></div><div class="t m0 x6 h6 yc0 ff2 fs1 fc0 sc0 ls20 ws0">没有现代</div><div class="t m0 x41 hf yc1 ff1 fs9 fc0 sc0 ls1f ws0">DSP<span class="_ _6"> </span><span class="ff2 fs1 ls20">芯片所必须有的单周期乘法器。</span><span class="ls38">1980 <span class="_"> </span><span class="ff2 fs1 ls0">年，日本</span><span class="ls3a ws11"> NEC </span><span class="ff2 fs1 ls20">公司推出的μ<span class="_ _c"></span><span class="ff1 fs9 ls0">P<span class="_ _d"></span><span class="ls25 ws12"> D7720</span></span></span></span></div><div class="t m0 x6 hf yc2 ff2 fs1 fc0 sc0 ls0 ws0">是第一个具有乘法器的商用<span class="ff1 fs9 ls2c ws13"> DSP </span>芯片。<span class="ff1 fs9"> </span></div><div class="t m0 x5 hf yc3 ff2 fs1 fc0 sc0 ls0 ws0">在这之后，最成功的<span class="_ _2"> </span><span class="ff1 fs9 ls1f">DSP <span class="_ _4"> </span></span>芯片当数美国德州仪器公司（<span class="ff1 fs9 ls3b ws14">Texas Instruments</span>，简称<span class="_ _2"> </span><span class="ff1 fs9 ls3c">TI</span>）</div><div class="t m0 x6 h6 yc4 ff2 fs1 fc0 sc0 ls0 ws0">的一系列产品。</div><div class="t m0 x25 hf yc5 ff1 fs9 fc0 sc0 ls3a ws0">TI <span class="_"> </span><span class="ff2 fs1 ls0">公司在<span class="_ _7"> </span></span><span class="ls25">1982<span class="_ _7"> </span><span class="ff2 fs1 ls0">年成功推出其第一代</span><span class="ls2c ws13"> DSP </span><span class="ff2 fs1 ls0">芯片</span><span class="ls3d ws15"> TMS32010<span class="_ _7"> </span></span><span class="ff2 fs1 ls0">及其系列产品</span></span></div><div class="t m0 x6 hf yc6 ff1 fs9 fc0 sc0 ls3e ws0">TMS32011<span class="_ _e"> </span><span class="ff2 fs1 ls0">、<span class="_ _e"> </span></span><span class="ls39">TMS320C10/C14/C15/C<span class="_ _1"></span>16/C17<span class="_ _f"> </span><span class="ff2 fs1 ls3f">等，之后相继推出了第二代<span class="_ _6"> </span></span><span class="ls24">DSP<span class="_ _f"> </span><span class="ff2 fs1 ls3f">芯片</span></span></span></div><div class="t m0 x6 hf yc7 ff1 fs9 fc0 sc0 ls40 ws0">TMS32020<span class="ff2 fs1 ls0">、</span><span class="ls41">TMS320C2<span class="_ _5"></span>5/C26/C28<span class="ff2 fs1 ls0">，第三代<span class="_"> </span></span><span class="ls1f">DSP<span class="_ _3"> </span><span class="ff2 fs1 ls20">芯片<span class="_"> </span></span><span class="ls37">TMS320C30/C31/C32<span class="_ _1"></span><span class="ff2 fs1 ls0">，第四代<span class="_"> </span><span class="ff1 fs9 ls22">DSP</span></span></span></span></span></div><div class="t m0 x6 hf yc8 ff2 fs1 fc0 sc0 ls0 ws0">芯片<span class="_ _7"> </span><span class="ff1 fs9 ls42">TMS320C40/C44</span>，第五代<span class="ff1 fs9 ls2c ws13"> DSP </span>芯片<span class="_ _7"> </span><span class="ff1 fs9 ls1e">TMS320C5X/C54X</span>，第二代<span class="_ _7"> </span><span class="ff1 fs9 ls24">DSP<span class="_ _8"> </span></span>芯片的改进型</div><div class="t m0 x6 hf yc9 ff1 fs9 fc0 sc0 ls43 ws0">TMS320C2XX<span class="ff2 fs1 ls0">，集多片<span class="_ _b"> </span></span><span class="ls44">DSP<span class="_ _7"> </span><span class="ff2 fs1 ls0">芯片于一体的高性能<span class="_ _7"> </span></span><span class="ls1f">DSP<span class="_ _7"> </span><span class="ff2 fs1 ls0">芯片<span class="_ _7"> </span></span><span class="ls45">TMS320C8X<span class="_ _9"> </span><span class="ff2 fs1 ls0">以及目前速度最</span></span></span></span></div><div class="t m0 x6 h6 yca ff2 fs1 fc0 sc0 ls0 ws0">快的第六代</div><div class="t m0 x42 hf ycb ff1 fs9 fc0 sc0 ls1f ws0">DSP<span class="_ _3"> </span><span class="ff2 fs1 ls0">芯片<span class="_ _0"> </span></span><span class="ls37">TMS320C62X/C67X<span class="_ _0"> </span><span class="ff2 fs1 ls0">等。</span><span class="ls46">TI<span class="_ _3"> </span><span class="ff2 fs1 ls0">将常用的<span class="_ _3"> </span></span></span></span>DSP<span class="_ _3"> </span><span class="ff2 fs1 ls0">芯片归纳为三大系列，</span></div><div class="t m0 x6 h6 ycc ff2 fs1 fc0 sc0 ls47 ws0">即：</div><div class="t m0 x43 hf ycd ff1 fs9 fc0 sc0 ls48 ws0">TMS320C2000<span class="_ _f"> </span><span class="ff2 fs1 ls47">系列（包括<span class="_ _6"> </span></span><span class="ls49">TMS320C2X/C2XX<span class="_ _e"> </span><span class="ff2 fs1 ls47">）、</span><span class="ls4a">TMS320C5000<span class="_ _f"> </span><span class="ff2 fs1 ls47">系列（包括</span></span></span></div><div class="t m0 x6 hf yce ff1 fs9 fc0 sc0 ls49 ws0">TMS320C5X/C54<span class="_ _5"></span>X/C55X<span class="ff2 fs1 ls0">）、</span><span class="ls37">TMS320C6000<span class="_ _8"> </span><span class="ff2 fs1 ls0">系列（</span><span class="ls4b">TMS320C62<span class="_ _5"></span>X/C67X<span class="ff2 fs1 ls0">）。如今，</span><span class="ls3c">TI<span class="_ _0"> </span><span class="ff2 fs1 ls0">公司</span></span></span></span></div><div class="t m0 x6 h6 ycf ff2 fs1 fc0 sc0 ls0 ws0">的一系列</div><div class="t m0 x3b hf yd0 ff1 fs9 fc0 sc0 ls22 ws0">DSP<span class="_ _7"> </span><span class="ff2 fs1 ls0">产品已经成为当今世界上最有影响的<span class="_ _7"> </span></span><span class="ls1f">DSP<span class="_ _7"> </span><span class="ff2 fs1 ls20">芯片。</span><span class="ls3c">TI<span class="_ _7"> </span><span class="ff2 fs1 ls0">公司也成为世界上最大</span></span></span></div><div class="t m0 x6 h6 yd1 ff2 fs1 fc0 sc0 ls0 ws0">的</div><div class="t m0 x44 hf yd2 ff1 fs9 fc0 sc0 ls2c ws13"> DSP <span class="ff2 fs1 ls0 ws0">芯片供应商，其<span class="_ _2"> </span></span><span class="ls1f ws0">DSP<span class="_"> </span><span class="ff2 fs1 ls0">市场份额占全世界份额近</span><span class="ls38 ws16"> 50<span class="_ _5"></span></span><span class="ff2 fs1 ls7">％。</span><span class="ls0"> </span></span></div><div class="t m0 x5 hf yd3 ff2 fs1 fc0 sc0 ls0 ws0">第一个采用<span class="_ _4"> </span><span class="ff1 fs9 ls4c">CMOS<span class="_ _9"> </span></span>工艺生产浮点<span class="_ _9"> </span><span class="ff1 fs9 ls1f">DSP<span class="_ _9"> </span></span>芯片的是日本的<span class="ff1 fs9 ls4a ws17"> Hitachi </span>公司，它<span class="_ _1"></span>于<span class="_ _4"> </span><span class="ff1 fs9 ls25">1982<span class="_ _9"> </span></span>年推</div><div class="t m0 x6 h6 yd4 ff2 fs1 fc0 sc0 ls0 ws0">出了浮点</div><div class="t m0 x41 hf yd5 ff1 fs9 fc0 sc0 ls22 ws0">DSP<span class="_ _3"> </span><span class="ff2 fs1 ls0">芯片。</span><span class="ls25">1983 <span class="_"> </span><span class="ff2 fs1 ls0">年</span><span class="ls0"> <span class="_"> </span><span class="ff2 fs1">日本<span class="_ _3"> </span></span><span class="ls4d">Fuji<span class="_ _1"></span>tsu <span class="_"> </span><span class="ff2 fs1 ls0">公司推出的<span class="_ _0"> </span></span><span class="ls4e">MB87<span class="_ _5"></span>64<span class="ff2 fs1 ls0">，其指令周期为</span><span class="ls38 ws16"> 120ns</span><span class="ff2 fs1 ls0">，</span></span></span></span></span></div><div class="t m0 x6 h6 yd6 ff2 fs1 fc0 sc0 ls0 ws0">且具有双内部总线，从而使处理吞吐量发生了一个大的飞跃。而第一个高性能浮点</div><div class="t m0 x45 hf yd7 ff1 fs9 fc0 sc0 ls1f ws0">DSP<span class="_ _4"> </span><span class="ff2 fs1 ls0">芯</span></div><div class="t m0 x6 h6 yd8 ff2 fs1 fc0 sc0 ls0 ws0">片应是</div><div class="t m0 x3a hf yd9 ff1 fs9 fc0 sc0 ls4c ws18"> AT&T <span class="ff2 fs1 ls0 ws0">公司于<span class="_ _2"> </span></span><span class="ls38 ws0">1984 <span class="_"> </span><span class="ff2 fs1 ls0">年推出的<span class="_ _4"> </span></span><span class="ls34">DSP32<span class="ff2 fs1 ls0">。</span><span class="ls0"> </span></span></span></div><div class="t m0 x5 hf yda ff2 fs1 fc0 sc0 ls0 ws0">与其他公司相比，<span class="ff1 fs9 ls4f">Motorola <span class="_"> </span></span>公司在推出<span class="ff1 fs9 ls2c ws13"> DSP </span>芯片方面相对较晚。<span class="ff1 fs9 ls25">1986<span class="_ _9"> </span></span>年，该公司推</div><div class="t m0 x6 h6 ydb ff2 fs1 fc0 sc0 ls2f ws0">出了定点处理器</div><div class="t m0 x46 hf ydc ff1 fs9 fc0 sc0 ls4e ws0">MC56001<span class="_ _10"></span><span class="ff2 fs1 ls0">。<span class="_ _5"></span></span><span class="ls38">1990<span class="_ _11"> </span><span class="ff2 fs1 ls2f">年，推出了与<span class="_ _6"> </span></span><span class="ls50">IEEE <span class="_ _4"> </span><span class="ff2 fs1 ls2f">浮点格式兼容的浮点</span><span class="ls2c ws19"> DSP </span><span class="ff2 fs1 ls2f">芯片<span class="_ _1"></span><span class="ff1 fs9 ls0"> </span></span></span></span></div><div class="t m0 x6 hf ydd ff1 fs9 fc0 sc0 ls4e ws0">MC96002</div><div class="t m0 x47 hf yde ff2 fs1 fc0 sc0 ls0 ws0">。<span class="ff1 fs9"> </span></div><div class="t m0 x5 hf ydf ff2 fs1 fc0 sc0 ls0 ws0">美国模拟器件公司（<span class="ff1 fs9 ls51 ws1a">Analog<span class="_ _5"></span> Devices</span>，简称<span class="_"> </span><span class="ff1 fs9 ls22">AD</span>）在<span class="_"> </span><span class="ff1 fs9 ls24">DSP<span class="_ _3"> </span></span>芯片市场上也占有一定的份</div><div class="t m0 x6 h6 ye0 ff2 fs1 fc0 sc0 ls52 ws0">额，相继推出了一系列具有自己特点的</div><div class="t m0 x48 hf ye1 ff1 fs9 fc0 sc0 ls1f ws0">DSP<span class="_ _12"> </span><span class="ff2 fs1 ls52">芯片，其定点<span class="_ _6"> </span></span>DSP<span class="_ _12"> </span><span class="ff2 fs1 ls53">芯片有</span></div><div class="t m0 x6 hf ye2 ff1 fs9 fc0 sc0 ls2e ws0">ADSP2101/2103/<span class="_ _1"></span>2105<span class="ff2 fs1 ls0">、</span><span class="ls54">AS<span class="_ _1"></span>DP2111/211<span class="_ _1"></span>5<span class="ff2 fs1 ls0">、</span><span class="ls55">ADSP2161/<span class="_ _1"></span>2162/2164<span class="_"> </span><span class="ff2 fs1 ls7">以及<span class="_ _2"> </span></span><span class="ls56">ADSP2171/2181<span class="ff2 fs1 ls0">，浮点</span></span></span></span></div><div class="t m0 x6 hf ye3 ff1 fs9 fc0 sc0 ls1f ws0">DSP<span class="_"> </span><span class="ff2 fs1 ls0">芯片有<span class="_ _2"> </span></span><span class="ls57">ADSP21000/21020<span class="ff2 fs1 ls0">、</span></span>ADSP21060/21062<span class="_"> </span><span class="ff2 fs1 ls0">等。</span><span class="ls0"> </span></div><div class="t m0 x5 hf ye4 ff2 fs1 fc0 sc0 ls0 ws0">自<span class="_ _0"> </span><span class="ff1 fs9 ls25">1980<span class="_ _0"> </span></span>年以来，<span class="ff1 fs9 ls1f">DSP<span class="_ _0"> </span></span>芯片得到了突飞猛进的发展，<span class="ff1 fs9 ls1f">DSP<span class="_ _0"> </span></span>芯片的应用越来越广泛。从</div><div class="t m0 x6 h6 ye5 ff2 fs1 fc0 sc0 ls0 ws0">运算速度来看，</div><div class="t m0 x25 hf ye6 ff1 fs9 fc0 sc0 ls58 ws0">MAC<span class="ff2 fs1 ls0">（一次乘法和一次加法）时间已经从<span class="_ _b"> </span></span><span class="ls25">20<span class="_ _7"> </span><span class="ff2 fs1 ls0">世纪<span class="_ _b"> </span></span><span class="ls38">80<span class="_ _7"> </span><span class="ff2 fs1 ls0">年代初的<span class="_ _7"> </span></span></span>400ns<span class="ff2 fs1 ls0">（如</span></span></div><div class="t m0 x6 hf ye7 ff1 fs9 fc0 sc0 ls40 ws0">TMS32010<span class="ff2 fs1 ls0">）降低到<span class="_ _4"> </span></span><span class="ls25">10ns<span class="_"> </span><span class="ff2 fs1 ls0">以下（如<span class="_ _4"> </span></span><span class="ls3e">TMS320C54X<span class="ff2 fs1 ls0">、</span><span class="ls59">TMS320C62X/67X<span class="_ _4"> </span><span class="ff2 fs1 ls0">等），处理能力提高</span></span></span></span></div><div class="t m0 x6 h6 ye8 ff2 fs1 fc0 sc0 ls0 ws0">了几十倍。</div><div class="t m0 x49 hf ye9 ff1 fs9 fc0 sc0 ls5a ws0">DSP<span class="_ _0"> </span><span class="ff2 fs1 ls23">芯片内部关<span class="_ _5"></span>键的乘法器部件从<span class="_ _0"> </span></span><span class="ls5b">1980<span class="_ _3"> </span><span class="ff2 fs1 ls23">年的占模片区（</span><span class="ls5c ws1b">die area</span><span class="ff2 fs1 ls23">）的<span class="_ _8"> </span></span>40<span class="_ _5"></span>%<span class="ff2 fs1 ls0">左</span></span></div><div class="t m0 x6 h6 yea ff2 fs1 fc0 sc0 ls1b ws0">右下降<span class="_ _1"></span>到</div><div class="t m0 x3b hf yeb ff1 fs9 fc0 sc0 ls5d ws0">5%<span class="ff2 fs1 ls23">以下，片内<span class="_ _4"> </span></span><span class="ls5e">RAM<span class="_ _4"> </span><span class="ff2 fs1 ls23">数量增加一个数<span class="_ _5"></span>量级以上。从制造工艺来看<span class="_ _5"></span>，</span><span class="ls5b">1980<span class="_ _4"> </span><span class="ff2 fs1 ls23">年采用</span></span></span></div><div class="t m0 x6 hf yec ff1 fs9 fc0 sc0 ls0 ws0">4<span class="_ _13"></span><span class="ff2 fs1">μ<span class="_ _14"></span><span class="ff1 fs9">m<span class="_ _d"></span> <span class="_"> </span><span class="ff2 fs1">的</span><span class="ls5f"> N<span class="_ _e"> </span></span><span class="ff2 fs1">沟道<span class="_"> </span></span><span class="ls58">MOS</span><span class="ff2 fs1">（</span><span class="ls2c">NMOS</span><span class="ff2 fs1">）工艺，而现在则普遍采用亚微米（</span><span class="ls48">Micron</span><span class="ff2 fs1">）</span><span class="ls60">CMOS<span class="_ _3"> </span><span class="ff2 fs1 ls7">工艺。</span></span></span></span></div></div><div class="pi" data-data='{"ctm":[1.611639,0.000000,0.000000,1.611639,0.000000,0.000000]}'></div></div>