<?xml version="1.0" encoding="utf-8" standalone="no"?>
<!DOCTYPE document PUBLIC "-//CNX//DTD CNXML 0.5 plus MathML//EN" "http://cnx.rice.edu/cnxml/0.5/DTD/cnxml_mathml.dtd">
<document xmlns="http://cnx.rice.edu/cnxml" xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="new1">
  <name xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">Design of Orthogonal PR-FIR Filterbanks via Halfband Spectral Factorization</name>
  <metadata xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">
  <md:version xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">2.2</md:version>
  <md:created xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">2003/01/13</md:created>
  <md:revised xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">2005/09/20 20:36:01.519 GMT-5</md:revised>
  <md:authorlist xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">
      <md:author xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="schniter">
      <md:firstname xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">Phil</md:firstname>
      
      <md:surname xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">Schniter</md:surname>
      <md:email xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">schniter@ee.eng.ohio-state.edu</md:email>
    </md:author>
  </md:authorlist>

  <md:maintainerlist xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">
    <md:maintainer xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="jago">
      <md:firstname xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">Adan</md:firstname>
      
      <md:surname xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">Galvan</md:surname>
      <md:email xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">jago@rice.edu</md:email>
    </md:maintainer>
    <md:maintainer xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="schniter">
      <md:firstname xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">Phil</md:firstname>
      
      <md:surname xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">Schniter</md:surname>
      <md:email xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">schniter@ee.eng.ohio-state.edu</md:email>
    </md:maintainer>
  </md:maintainerlist>
  
  <md:keywordlist xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">
    <md:keyword xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">halfband spectral Factorization</md:keyword>
    <md:keyword xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">orthogonal PR-Fir Filterbanks</md:keyword>
  </md:keywordlist>

  <md:abstract xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/"/>
</metadata>


  <content xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">

    <section xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="sec2">
      <name xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">Design of Orthogonal PR-FIR Filterbanks via Halfband
	Spectral Factorization</name>
      <para xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="p3">
	Recall that analysis-filter design for orthogonal PR-FIR
	filterbanks reduces to the design of a real-coefficient causal
	FIR prototype filter 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">
	      <m:msub>
		<m:mi>H</m:mi>
		<m:mn>0</m:mn>
	      </m:msub>
	    </m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> 
	that satisfies the power-symmetry condition 

	<equation xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="eq5a">
	  <m:math>
	    <m:apply>
	      <m:eq/>
	      <m:apply>
		<m:plus/>
		<m:apply>
		  <m:power/>
		  <m:apply>
		    <m:abs/>
		    <m:apply>
		      <m:ci type="fn">
			<m:msub>
			  <m:mi>H</m:mi>
			  <m:mn>0</m:mn>
			</m:msub>
		      </m:ci>
		      <m:apply>
			<m:exp/>
			<m:apply>
			  <m:times/>
			  <m:imaginaryi/>
			  <m:ci>ω</m:ci>
			</m:apply>
		      </m:apply>
		    </m:apply>
		  </m:apply>
		  <m:cn>2</m:cn>
		</m:apply>		
		<m:apply>
		  <m:power/>
		  <m:apply>
		    <m:abs/>
		    <m:apply>
		      <m:ci type="fn">
			<m:msub>
			  <m:mi>H</m:mi>
			  <m:mn>0</m:mn>
			</m:msub>
		      </m:ci>
		      <m:apply>
			<m:exp/>			
			<m:apply>
			  <m:times/>
			  <m:imaginaryi/>		    
			  <m:apply>
			    <m:minus/>
			    <m:pi/>
			    <m:ci>ω</m:ci>
			  </m:apply>
			</m:apply>
		      </m:apply>
		    </m:apply>
		  </m:apply>
		  <m:cn>2</m:cn>
		</m:apply>
	      </m:apply>
	      <m:cn>1</m:cn>
	    </m:apply>
	  </m:math>
	</equation>

	Power-symmetric filters are closely related to "halfband"
	filters.  A zero-phase halfband filter is a zero-phase filter 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">F</m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> with the property
	
	<equation xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="eq5">
	  <m:math>
	    <m:apply>
	      <m:eq/>
	      <m:apply>
		<m:plus/>
		<m:apply>
		  <m:ci type="fn">F</m:ci>
		  <m:ci>z</m:ci>
		</m:apply>
		<m:apply>
		  <m:ci type="fn">F</m:ci>
		  <m:ci>-z</m:ci>
		</m:apply>
	      </m:apply>
	      <m:cn>1</m:cn>
	    </m:apply>
	  </m:math>
	</equation>
	
	When, in addition, 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">F</m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> has real-valued coefficients, its DTFT is
	"amplitude-symmetric":

	<equation xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="eq6">
	  <m:math>
	    <m:apply>
	      <m:eq/>
	      <m:apply>
		<m:plus/>
		<m:apply>
		  <m:ci type="fn">F</m:ci>
		  <m:apply>
		    <m:exp/>
		    <m:apply>
		      <m:times/>
		      <m:imaginaryi/>
		      <m:ci>ω</m:ci>
		    </m:apply>
		  </m:apply>
		</m:apply>
		<m:apply>
		  <m:ci type="fn">F</m:ci>
		  <m:apply>
		    <m:exp/>
		    <m:apply>
		      <m:times/>
		      <m:imaginaryi/>			
		      <m:apply>
			<m:minus/>
			<m:pi/>
			<m:ci>ω</m:ci>
		      </m:apply>
		    </m:apply>
		  </m:apply>
		</m:apply>
	      </m:apply>
	      <m:cn>1</m:cn>
	    </m:apply>
	  </m:math>
	</equation>

	The amplitude-symmetry property is illustrated in <cnxn xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" target="fig3" strength="9"/>:
      </para>

      <figure xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="fig3">
	<media xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" type="image/png" src="ortho_f3.png"/>
      </figure>      

      <para xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="sec2_p2">
	If, in addition to being real-valued, 
	<note xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" type="footnote">
	  Recall that zero-phase filters have real-valued DTFTs.
	</note> 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">F</m:ci>
	    <m:apply>
	      <m:exp/>			
	      <m:apply>
		<m:times/>
		<m:imaginaryi/>		    
		<m:ci>ω</m:ci>
	      </m:apply>
	    </m:apply>
	  </m:apply>
	</m:math> is non-negative, then 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">F</m:ci>
	    <m:apply>
	      <m:exp/>			
	      <m:apply>
		<m:times/>
		<m:imaginaryi/>		    
		<m:ci>ω</m:ci>
	      </m:apply>
	    </m:apply>
	  </m:apply>
	</m:math> constitutes a valid power response.  If we can find 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">
	      <m:msub>
		<m:mi>H</m:mi>
		<m:mn>0</m:mn>
	      </m:msub>
	    </m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> 
	such that 

	<m:math>
	  <m:apply>
	    <m:eq/>
	    <m:apply>
	      <m:power/>
	      <m:apply>
		<m:abs/>
		<m:apply>
		  <m:ci type="fn">
		    <m:msub>
		      <m:mi>H</m:mi>
		      <m:mn>0</m:mn>
		    </m:msub>
		  </m:ci>
		  <m:apply>
		    <m:exp/>			
		    <m:apply>
		      <m:times/>
		      <m:imaginaryi/>		    
		      <m:ci>ω</m:ci>
		    </m:apply>
		  </m:apply>
		</m:apply>
	      </m:apply>
	      <m:cn>2</m:cn>
	    </m:apply>
	    <m:apply>
	      <m:ci type="fn">F</m:ci>
	      <m:apply>
		<m:exp/>			
		<m:apply>
		  <m:times/>
		  <m:imaginaryi/>		    
		  <m:ci>ω</m:ci>
		</m:apply>
	      </m:apply>
	    </m:apply>	    
	  </m:apply>
	</m:math>, 

	then this 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">
	      <m:msub>
		<m:mi>H</m:mi>
		<m:mn>0</m:mn>
	      </m:msub>
	    </m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> 
	will satisfy the desired power-symmetry property 	
	
	<m:math>
	  <m:apply>
	    <m:eq/>
	    <m:apply>
	      <m:plus/>
	      <m:apply>
		<m:power/>
		<m:apply>
		  <m:abs/>
		  <m:apply>
		    <m:ci type="fn">
		      <m:msub>
			<m:mi>H</m:mi>
			<m:mn>0</m:mn>
		      </m:msub>
		    </m:ci>
		    <m:apply>
		      <m:exp/>
		      <m:apply>
			<m:times/>
			<m:imaginaryi/>
			<m:ci>ω</m:ci>
		      </m:apply>
		    </m:apply>
		  </m:apply>
		</m:apply>
		<m:cn>2</m:cn>
	      </m:apply>		
	      <m:apply>
		<m:power/>
		<m:apply>
		  <m:abs/>
		  <m:apply>
		    <m:ci type="fn">
		      <m:msub>
			<m:mi>H</m:mi>
			<m:mn>0</m:mn>
		      </m:msub>
		    </m:ci>
		    <m:apply>
		      <m:exp/>			
		      <m:apply>
			<m:times/>
			<m:imaginaryi/>		    
			<m:apply>
			  <m:minus/>
			  <m:pi/>
			  <m:ci>ω</m:ci>
			</m:apply>
		      </m:apply>
		    </m:apply>
		  </m:apply>
		</m:apply>
		<m:cn>2</m:cn>
	      </m:apply>
	    </m:apply>
	    <m:cn>1</m:cn>
	  </m:apply>
	</m:math>.
      </para>

      <para xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="sec2_p3">
	First, realize 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">F</m:ci>
	    <m:apply>
	      <m:exp/>			
	      <m:apply>
		<m:times/>
		<m:imaginaryi/>		    
		<m:ci>ω</m:ci>
	      </m:apply>
	    </m:apply>
	  </m:apply>
	</m:math> 
	is easily modified to ensure non-negativity: construct 
	
	<m:math>
	  <m:apply>
	    <m:eq/>
	    <m:apply>
	      <m:ci type="fn" class="discrete">q</m:ci>
	      <m:ci>n</m:ci>
	    </m:apply>
	    <m:apply>
	      <m:plus/>
	      <m:apply>
		<m:ci type="fn" class="discrete">f</m:ci>
		<m:ci>n</m:ci>
	      </m:apply>
	      <m:apply>
		<m:times/>
		<m:ci>ε</m:ci>
		<m:apply>
		  <m:ci type="fn" class="discrete">δ</m:ci>
		  <m:ci>n</m:ci>
		</m:apply>
	      </m:apply>	     
	    </m:apply>
	  </m:apply>
	</m:math> for sufficiently large
	<m:math><m:ci>ɛ</m:ci></m:math>, which will raise 
	
	<m:math>
	  <m:apply>
	    <m:ci type="fn">F</m:ci>
	    <m:apply>
	      <m:exp/>			
	      <m:apply>
		<m:times/>
		<m:imaginaryi/>		    
		<m:ci>ω</m:ci>
	      </m:apply>
	    </m:apply>
	  </m:apply>
	</m:math>
	by <m:math><m:ci>ε</m:ci></m:math> uniformly over
	<m:math><m:ci>ω</m:ci></m:math> (see <cnxn xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" target="fig4" strength="9"/>).
      </para>

      <figure xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="fig4">
	<media xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" type="image/png" src="ortho_f4.png"/>
      </figure>
      
      <para xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="sec2_p3a">
	The resulting 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">Q</m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> 
	is non-negative and satisfies the amplitude-symmetry condition 

	<m:math>
	  <m:apply>
	    <m:eq/>
	    <m:apply>
	      <m:plus/>
	      <m:apply>
		<m:ci type="fn">Q</m:ci>
		<m:apply>
		  <m:exp/>			
		  <m:apply>
		    <m:times/>
		    <m:imaginaryi/>		    
		    <m:ci>ω</m:ci>
		  </m:apply>
		</m:apply>
	      </m:apply>
	      <m:apply>
		<m:ci type="fn">Q</m:ci>
		<m:apply>
		  <m:exp/>			
		  <m:apply>
		    <m:times/>
		    <m:imaginaryi/>	
		    <m:apply>
		      <m:minus/>
		      <m:pi/>
		      <m:ci>ω</m:ci>
		    </m:apply>
		  </m:apply>
		</m:apply>
	      </m:apply>
	    </m:apply>
	    <m:apply>
	      <m:plus/>
	      <m:cn>1</m:cn>
	      <m:apply>
		<m:times/>
		<m:cn>2</m:cn>
		<m:ci>ε</m:ci>
	      </m:apply>
	    </m:apply>
	  </m:apply>
	</m:math>.  

	We will make up for the additional gain later.  The procedure
	by which 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">
	      <m:msub>
		<m:mi>H</m:mi>
		<m:mn>0</m:mn>
	      </m:msub>
	    </m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> 
	can be calculated from the raised halfband 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">Q</m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math>, known as <term xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">spectral factorization</term>, is
	described next.
      </para>

      <para xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="sec2_p5">
	Since 
	<m:math>
	  <m:apply>
	    <m:ci type="fn" class="discrete">q</m:ci>
	    <m:ci>n</m:ci>
	  </m:apply>
	</m:math> 
	is conjugate-symmetric around the origin, the roots of 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">Q</m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> come in pairs 

	<m:math>
	  <m:apply>
	    <m:set>
	      <m:ci><m:msub>
		<m:mi>a</m:mi>
		<m:mi>i</m:mi>
	      </m:msub></m:ci>
	      <m:apply>
		<m:divide/>
		<m:cn>1</m:cn>
		<m:apply>
		  <m:conjugate/>
		  <m:ci><m:msub>
		    <m:mi>a</m:mi>
		    <m:mi>i</m:mi>
		  </m:msub></m:ci>
		</m:apply>
	      </m:apply>
	    </m:set>
	  </m:apply>
	</m:math>.  This can be seen by writing 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">Q</m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> in the factored form below, which clearly
	corresponds to a polynomial with coefficients
	conjugate-symmetric around the 
	<m:math>
	  <m:msup>
	    <m:mn>0</m:mn>
	    <m:ci>th</m:ci>
	  </m:msup>
	</m:math>-order coefficient.
	
	<equation xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="eq21">
	  <m:math>
	    <m:apply>
	      <m:eq/>
	      <m:apply>
		<m:ci type="fn">Q</m:ci>
		<m:ci>z</m:ci>
	      </m:apply>

	      <m:apply>
		<m:sum/>
		<m:bvar>
		  <m:ci>n</m:ci>
		</m:bvar>
		<m:lowlimit>
		  <m:apply>
		    <m:minus/>
		    <m:apply>
		      <m:minus/>
		      <m:cn>N</m:cn>
		      <m:cn>1</m:cn>
		    </m:apply>
		  </m:apply>
		</m:lowlimit>
		<m:uplimit>
		  <m:apply>
		    <m:minus/>
		    <m:ci>N</m:ci>
		    <m:cn>1</m:cn>
		  </m:apply>
		</m:uplimit>
		<m:apply>
		  <m:times/>
		  <m:apply>
		    <m:ci type="fn" class="discrete">q</m:ci>
		    <m:ci>n</m:ci>
		  </m:apply>
		  <m:apply>
		    <m:power/>
		    <m:ci>z</m:ci>
		    <m:apply>
		      <m:minus/>
		      <m:ci>n</m:ci>
		    </m:apply>
		  </m:apply>
		</m:apply>
	      </m:apply>

	      <m:apply>
		<m:times/>
		<m:ci>A</m:ci>
		<m:apply>
		  <m:product/>
		  <m:bvar>
		    <m:ci>i</m:ci>
		  </m:bvar>
		  <m:lowlimit>
		    <m:cn>1</m:cn>
		  </m:lowlimit>
		  <m:uplimit>
		    <m:apply>
		      <m:minus/>
		      <m:ci>N</m:ci>
		      <m:cn>1</m:cn>
		    </m:apply>
		  </m:uplimit>
		  <m:apply>
		    <m:times/>
		    <m:apply>
		      <m:minus/>
		      <m:cn>1</m:cn>
		      <m:apply>
			<m:times/>
			<m:ci>
			  <m:msub>
			    <m:mi>a</m:mi>
			    <m:mi>i</m:mi>
			  </m:msub>
			</m:ci>
			<m:apply>
			  <m:inverse/>
			  <m:ci>z</m:ci>
			</m:apply>
		      </m:apply>
		    </m:apply>
		    <m:apply>
		      <m:minus/>
		      <m:cn>1</m:cn>
		      <m:apply>
			<m:times/>
			<m:apply>
			  <m:conjugate/>
			  <m:ci>
			    <m:msub>
			      <m:mi>a</m:mi>
			      <m:mi>i</m:mi>
			    </m:msub>
			  </m:ci>
			</m:apply>
			<m:ci>z</m:ci>
		      </m:apply>
		    </m:apply>
		  </m:apply>
		</m:apply>
	      </m:apply>
	    </m:apply>
	  </m:math>
	</equation>

	where 
	<m:math>
	  <m:apply>
	    <m:in/>
	    <m:ci>A</m:ci>
	    <m:ci><m:msup>
	      <m:mi>ℝ</m:mi>
	      <m:mi>+</m:mi>
	    </m:msup></m:ci>
	  </m:apply>
	</m:math>.  Note that the complex numbers 

	<m:math>
	  <m:apply>
	    <m:set>
	      <m:ci><m:msub>
		<m:mi>a</m:mi>
		<m:mi>i</m:mi>
	      </m:msub></m:ci>
	      <m:apply>
		<m:divide/>
		<m:cn>1</m:cn>
		<m:apply>
		  <m:conjugate/>
		  <m:ci><m:msub>
		    <m:mi>a</m:mi>
		    <m:mi>i</m:mi>
		  </m:msub></m:ci>
		</m:apply>
	      </m:apply>
	    </m:set>
	  </m:apply>
	</m:math>
	
	are symmetric across the unit circle in the <cnxn xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" document="m10556" target="zplane" strength="8">z-plane</cnxn>.
	Thus, for ever root of
	<m:math>
	  <m:apply>
	    <m:ci type="fn">Q</m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> 
	inside the unit-circle, there exists a root outside of the
	unit circle (see <cnxn xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" target="fig5" strength="9"/>).
      </para>

      <figure xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="fig5">
	<media xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" type="image/png" src="ortho_f5.png"/>
      </figure>
      
      <para xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="sec2_p6">
	Let us assume, without loss of generality, that 
	<m:math>
	  <m:apply>
	    <m:lt/>
	    <m:apply>
	      <m:abs/>
	      <m:ci><m:msub>
		<m:mi>a</m:mi>
		<m:mi>i</m:mi>
	      </m:msub></m:ci>
	    </m:apply>
	    <m:cn>1</m:cn>
	  </m:apply>
	</m:math>.  If we form 
	
	<m:math>
	  <m:apply>
	    <m:ci type="fn">
	      <m:msub>
		<m:mi>H</m:mi>
		<m:mn>0</m:mn>
	      </m:msub>
	    </m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> 
	
	from the roots of 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">Q</m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> 
	with magnitude less than one:  

	<equation xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="eq19">
	  <m:math>
	    <m:apply>
	      <m:eq/>	    
	      <m:apply>
		<m:ci type="fn">
		  <m:msub>
		    <m:mi>H</m:mi>
		    <m:mn>0</m:mn>
		  </m:msub>
		</m:ci>
		<m:ci>z</m:ci>
	      </m:apply>
	      <m:apply>
		<m:times/>
		<m:apply>
		  <m:root/>
		  <m:ci>A</m:ci>
		</m:apply>
		<m:apply>
		  <m:product/>
		  <m:bvar>
		    <m:ci>i</m:ci>
		  </m:bvar>
		  <m:lowlimit>
		    <m:cn>1</m:cn>
		  </m:lowlimit>
		  <m:uplimit>
		    <m:apply>
		      <m:minus/>
		      <m:ci>N</m:ci>
		      <m:cn>1</m:cn>
		    </m:apply>
		  </m:uplimit>
		  <m:apply>
		    <m:minus/>
		    <m:cn>1</m:cn>
		    <m:apply>
		      <m:times/>
		      <m:ci>
			<m:msub>
			  <m:mi>a</m:mi>
			  <m:mi>i</m:mi>
			</m:msub>
		      </m:ci>
			<m:apply>
			<m:inverse/>
			<m:ci>z</m:ci>
		      </m:apply>
		    </m:apply>
		  </m:apply>
		</m:apply>
	      </m:apply>
	    </m:apply>
	  </m:math>
	</equation>

	then it is apparent that 
	<m:math>
	  <m:apply>
	    <m:eq/>
	    <m:apply>
	      <m:power/>
	      <m:apply>
		<m:abs/>
		<m:apply>
		  <m:ci type="fn">
		    <m:msub>
		      <m:mi>H</m:mi>
		      <m:mn>0</m:mn>
		    </m:msub>
		  </m:ci>
		  <m:apply>
		    <m:exp/>
		    <m:apply>
		      <m:times/>
		      <m:imaginaryi/>
		      <m:ci>ω</m:ci>
		    </m:apply>
		  </m:apply>
		</m:apply>
	      </m:apply>
	      <m:cn>2</m:cn>
	    </m:apply>
	    <m:apply>
	      <m:ci type="fn">Q</m:ci>
	      <m:apply>
		<m:exp/>
		<m:apply>
		  <m:times/>
		  <m:imaginaryi/>
		  <m:ci>ω</m:ci>
		</m:apply>
	      </m:apply>
	    </m:apply>
	  </m:apply>
	</m:math>.  This 
		
	<m:math>
	  <m:apply>
	    <m:ci type="fn">
	      <m:msub>
		<m:mi>H</m:mi>
		<m:mn>0</m:mn>
	      </m:msub>
	    </m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> 
	is the so-called <term xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">minimum-phase spectral factor</term> of 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">Q</m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math>.
      </para>


      <para xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="sec2_p7">
	Actually, in order to make 
	<m:math>
	  <m:apply>
	    <m:eq/>
	    <m:apply>
	      <m:power/>
	      <m:apply>
		<m:abs/>
		<m:apply>
		  <m:ci type="fn">
		    <m:msub>
		      <m:mi>H</m:mi>
		      <m:mn>0</m:mn>
		    </m:msub>
		  </m:ci>
		  <m:apply>
		    <m:exp/>
		    <m:apply>
		      <m:times/>
		      <m:imaginaryi/>
		      <m:ci>ω</m:ci>
		    </m:apply>
		  </m:apply>
		</m:apply>
	      </m:apply>
	      <m:cn>2</m:cn>
	    </m:apply>
	    <m:apply>
	      <m:ci type="fn">Q</m:ci>
	      <m:apply>
		<m:exp/>
		<m:apply>
		  <m:times/>
		  <m:imaginaryi/>
		  <m:ci>ω</m:ci>
		</m:apply>
	      </m:apply>
	    </m:apply>
	  </m:apply>
	</m:math>, we are not required to choose all roots inside the
	unit circle; it is enough to choose one root from every
	unit-circle-symmetric pair.  However, we do want to ensure
	that
	<m:math>
	  <m:apply>
	    <m:ci type="fn">
	      <m:msub>
		<m:mi>H</m:mi>
		<m:mn>0</m:mn>
	      </m:msub>
	    </m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math>
	has real-valued coefficients.  For this, we must ensure that
	roots come in conjugate-symmetric pairs,
	<foreign xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">i.e.</foreign>, pairs having symmetry with respect to
	the real axis in the complex plane (<cnxn xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" target="fig6" strength="9"/>).
      </para>
      
      <figure xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="fig6">
	<media xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" type="image/png" src="ortho_f6.png"/>
      </figure>

      <para xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="sec2_p8">
	Because 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">Q</m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> 
	has real-valued coefficients, we know that its roots satisfy
	this conjugate-symmetry property.  Then forming 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">
	      <m:msub>
		<m:mi>H</m:mi>
		<m:mn>0</m:mn>
	      </m:msub>
	    </m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> 
	from the roots of
	<m:math>
	  <m:apply>
	    <m:ci type="fn">Q</m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> 
	that are strictly inside (or strictly outside) the unit
	circle, we ensure that 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">
	      <m:msub>
		<m:mi>H</m:mi>
		<m:mn>0</m:mn>
	      </m:msub>
	    </m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> has real-valued coefficients.
      </para>

      <para xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="sec2_p9">
	Finally, we say a few words about the design of the halfband
	filter 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">F</m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math>.  The <term xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">window design</term> method is one technique that
	could be used in this application.  The window design method
	starts with an ideal lowpass filter, and windows its
	doubly-infinite impulse response using a window function with
	finite time-support.  The ideal real-valued zero-phase halfband
	filter has impulse response (where 
	<m:math>
	  <m:apply>
	    <m:in/>
	    <m:ci>n</m:ci>
	    <m:integers/>
	  </m:apply>
	</m:math>): 

	<equation xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="eq20">
	  <m:math>
	    <m:apply>
	      <m:eq/>
	      <m:apply>
		<m:ci type="fn" class="discrete">
		  <m:mover accent="true">
		    <m:mi>f</m:mi>
		    <m:mo>¯</m:mo>
		  </m:mover>
		</m:ci>
		<m:ci>n</m:ci>
	      </m:apply>
	      <m:apply>
		<m:divide/>
		<m:apply>
		  <m:sin/>
		  <m:apply>
		    <m:times/>
		    <m:apply>
		      <m:divide/>
		      <m:pi/>
		      <m:cn>2</m:cn>
		    </m:apply>
		    <m:ci>n</m:ci>
		  </m:apply>
		</m:apply>
		<m:apply>
		  <m:times/>
		  <m:pi/>
		  <m:ci>n</m:ci>
		</m:apply>
	      </m:apply>
	    </m:apply>
	  </m:math>
	</equation>
	
	which has the important property that all even-indexed
	coefficients except 
	<m:math>
	  <m:apply>
	    <m:ci type="fn" class="discrete">
	      <m:mover accent="true">
		<m:mi>f</m:mi>
		<m:mo>¯</m:mo>
	      </m:mover>
	    </m:ci>
	    <m:cn>0</m:cn>
	  </m:apply>
	</m:math>
	equal zero.  It can be seen that this latter property is
	implied by the halfband definition 
	<m:math>
	  <m:apply>
	    <m:eq/>
	    <m:apply>
	      <m:plus/>
	      <m:apply>
		<m:ci type="fn"> 
		  <m:mover accent="true">
		    <m:mi>F</m:mi>
		    <m:mo>¯</m:mo>
		  </m:mover>
		</m:ci>
		<m:ci>z</m:ci>
	      </m:apply>
	      <m:apply>
		<m:ci type="fn"> 
		  <m:mover accent="true">
		    <m:mi>F</m:mi>
		    <m:mo>¯</m:mo>
		  </m:mover>
		</m:ci>
		<m:apply>
		  <m:inverse/>
		  <m:ci>z</m:ci>
		</m:apply>
	      </m:apply>
	    </m:apply>
	    <m:cn>1</m:cn>
	  </m:apply>
	</m:math> since, due to odd-coefficient cancellation, we find 

	<equation xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="eq22">
	  <m:math>
	    <m:apply>
	      <m:eq/>
	      <m:cn>1</m:cn>
	      <m:apply>
		<m:plus/>
		<m:apply>
		  <m:ci type="fn"> 
		    <m:mover accent="true">
		      <m:mi>F</m:mi>
		      <m:mo>¯</m:mo>
		    </m:mover>
		  </m:ci>
		  <m:ci>z</m:ci>
		</m:apply>
		<m:apply>
		  <m:ci type="fn"> 
		    <m:mover accent="true">
		      <m:mi>F</m:mi>
		      <m:mo>¯</m:mo>
		    </m:mover>
		  </m:ci>
		  <m:apply>
		    <m:inverse/>
		    <m:ci>z</m:ci>
		  </m:apply>
		</m:apply>
	      </m:apply>	      
	      <m:apply>
                <m:ci><m:mo>⇔</m:mo></m:ci>
		<m:apply>
		  <m:times/>
		  <m:cn>2</m:cn>
		  <m:apply>
		    <m:sum/>
		    <m:bvar>
		      <m:ci>m</m:ci>
		    </m:bvar>
		    <m:lowlimit>
		      <m:apply>
			<m:minus/>
			<m:infinity/>
		      </m:apply>
		    </m:lowlimit>
		    <m:uplimit>
		      <m:infinity/>
		    </m:uplimit>
		    <m:apply>
		      <m:times/>
		      <m:apply>
			<m:ci type="fn" class="discrete"> 
			  <m:mover accent="true">
			    <m:mi>f</m:mi>
			    <m:mo>¯</m:mo>
			  </m:mover>
			</m:ci>
			<m:apply>
			  <m:times/>
			  <m:cn>2</m:cn>
			  <m:ci>m</m:ci>
			</m:apply>
		      </m:apply>
		      <m:apply>
			<m:power/>
			<m:ci>z</m:ci>
			<m:apply>
			  <m:minus/>
			  <m:apply>
			    <m:times/>
			    <m:cn>2</m:cn>
			    <m:ci>m</m:ci>
			  </m:apply>
			</m:apply>
		      </m:apply>
		    </m:apply>
		  </m:apply>
		</m:apply>		
		<m:apply>
		  <m:eq/>
		  <m:apply>
		    <m:ci type="fn"> 
		      <m:mover accent="true" class="discrete">
			<m:mi>f</m:mi>
			<m:mo>¯</m:mo>
		      </m:mover>
		    </m:ci>
		    <m:apply>
		      <m:times/>
		      <m:cn>2</m:cn>
		      <m:ci>m</m:ci>
		    </m:apply>
		  </m:apply>
		  <m:apply>
		    <m:times/>
		    <m:apply>
		      <m:divide/>
		      <m:cn>1</m:cn>
		      <m:cn>2</m:cn>
		    </m:apply>
		    <m:apply>
		      <m:ci type="fn" class="discrete">δ</m:ci>
		      <m:ci>m</m:ci>
		    </m:apply>
		  </m:apply>
		</m:apply>	
	      </m:apply>
	    </m:apply>
	  </m:math>
	</equation>

	Note that windowing the ideal halfband does not alter the
	property 
	<m:math>
	  <m:apply>
	    <m:eq/>
	    <m:apply>
	      <m:ci type="fn"> 
		<m:mover accent="true">
		  <m:mi>f</m:mi>
		  <m:mo>¯</m:mo>
		</m:mover>
	      </m:ci>
	      <m:apply>
		<m:times/>
		<m:cn>2</m:cn>
		<m:ci>m</m:ci>
	      </m:apply>
	    </m:apply>
	    <m:apply>
	      <m:times/>
	      <m:apply>
		<m:divide/>
		<m:cn>1</m:cn>
		<m:cn>2</m:cn>
	      </m:apply>
	      <m:apply>
		<m:ci type="fn" class="discrete">δ</m:ci>
		<m:ci>m</m:ci>
	      </m:apply>
	    </m:apply>
	  </m:apply>
	</m:math>, thus the window design 
	<m:math>
	  <m:apply>
	    <m:ci type="fn">F</m:ci>
	    <m:ci>z</m:ci>
	  </m:apply>
	</m:math> is guaranteed to be halfband feature.  Furthermore,
	a real-valued window with origin-symmetry preserves the
	real-valued zero-phase property of 
	<m:math>
	  <m:apply>
	    <m:set>
	      <m:apply>
		<m:ci type="fn" class="discrete">
		  <m:mover>
		    <m:mi>f</m:mi>
		    <m:mo>¯</m:mo>
		  </m:mover>
		</m:ci>
		<m:ci>n</m:ci>
	      </m:apply>
	    </m:set>
	  </m:apply>
	</m:math> above.  It turns out that many of the other popular
	design methods (<foreign xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">e.g.</foreign>, LS and equiripple)
	also produce halfband filters when the cutoff is specified at
	<m:math>
	  <m:apply>
	    <m:divide/>
	    <m:pi/>
	    <m:cn>2</m:cn>
	  </m:apply>
	</m:math> radians and all passband/stopband specifications
	are symmetric with respect to 
	<m:math>
	  <m:apply>
	    <m:eq/>
	    <m:ci>ω</m:ci>
	    <m:apply>
	      <m:divide/>
	      <m:pi/>
	      <m:cn>2</m:cn>
	    </m:apply>
	  </m:apply>
	</m:math>.
      </para>


      <!-- ************************************* -->
      <!-- ************************************* -->
      <!-- ************************************* -->


      <section xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="sub2">
	<name xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">Design Procedure Summary</name>
	<para xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="sec2_p10">
	  We now summarize the design procedure for a
	  length-<m:math><m:ci>N</m:ci></m:math> analysis lowpass
	  filter for an orthogonal perfect-reconstruction FIR
	  filterbank:

	  <list xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" id="l_design" type="enumerated">
	    <item xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">
	      Design a zero-phase real-coefficient halfband lowpass
	      filter 
	      <m:math>
		<m:apply>
		  <m:eq/>
		  <m:apply>
		    <m:ci type="fn">F</m:ci>
		    <m:ci>z</m:ci>
		  </m:apply>
		  <m:apply>
		    <m:sum/>
		    <m:bvar>
		      <m:ci>n</m:ci>
		    </m:bvar>
		    <m:lowlimit>
		      <m:apply>
			<m:minus/>
			<m:apply>
			  <m:minus/>
			  <m:ci>N</m:ci>
			  <m:cn>1</m:cn>
			</m:apply>
		      </m:apply>
		    </m:lowlimit>
		    <m:uplimit>
		      <m:apply>
			<m:minus/>
			<m:ci>N</m:ci>
			<m:cn>1</m:cn>
		      </m:apply>
		    </m:uplimit>
		    <m:apply>
		      <m:times/>
		      <m:apply>
			<m:ci type="fn" class="discrete">f</m:ci>
			<m:ci>n</m:ci>
		      </m:apply>
		      <m:apply>
			<m:power/>
			<m:ci>z</m:ci>
			<m:apply>
			  <m:minus/>
			  <m:ci>n</m:ci>
			</m:apply>
		      </m:apply>
		    </m:apply> 
		  </m:apply>
		</m:apply>
	      </m:math> where <m:math><m:ci>N</m:ci></m:math> is a
	      positive even integer (via, <foreign xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">e.g.</foreign>,
	      window designs, LS, or equiripple).
	    </item>

	    <item xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">
	      Calculate 
	      <m:math>	      
		<m:ci>ε</m:ci>
	      </m:math>, 
	      the maximum negative value of 
	      <m:math>
		<m:apply>
		  <m:ci type="fn">F</m:ci>
		  <m:apply>
		    <m:exp/>
		    <m:apply>
		      <m:times/>
		      <m:imaginaryi/>
		      <m:ci>ω</m:ci>
		    </m:apply>
		  </m:apply>
		</m:apply>
	      </m:math>.  

	      (Recall that 
	      <m:math>
		<m:apply>
		  <m:ci type="fn">F</m:ci>
		  <m:apply>
		    <m:exp/>
		    <m:apply>
		      <m:times/>
		      <m:imaginaryi/>
		      <m:ci>ω</m:ci>
		    </m:apply>
		  </m:apply>
		</m:apply>
	      </m:math> is real-valued for all 
	      <m:math><m:ci>ω</m:ci></m:math> because it has a
	      zero-phase response.)  Then create "raised halfband" 
	      <m:math>
		<m:apply>
		  <m:ci type="fn">Q</m:ci>
		  <m:ci>z</m:ci>
		</m:apply>
	      </m:math> 
	      via 
	      <m:math>
		<m:apply>
		  <m:eq/>
		  <m:apply>
		    <m:ci type="fn" class="discrete">q</m:ci>
		    <m:ci>n</m:ci>
		  </m:apply>
		  <m:apply>
		    <m:plus/>
		    <m:apply>
		      <m:ci type="fn" class="discrete">f</m:ci>
		      <m:ci>n</m:ci>
		    </m:apply>
		    <m:apply>
		      <m:times/>
		      <m:ci>ε</m:ci>
		      <m:apply>
			<m:ci type="fn" class="discrete">δ</m:ci>
			<m:ci>n</m:ci>
		      </m:apply>
		    </m:apply>
		  </m:apply>
		</m:apply>
	      </m:math>, ensuring that 
	      
	      <m:math>
		<m:apply>
		  <m:geq/>
		  <m:apply>
		    <m:ci type="fn">Q</m:ci>
		    <m:apply>
		      <m:exp/>			
		      <m:apply>
			<m:times/>
			<m:imaginaryi/>		    
			<m:ci>ω</m:ci>
		      </m:apply>
		    </m:apply>
		  </m:apply>
		  <m:cn>0</m:cn>
		</m:apply>
	      </m:math>, forall <m:math><m:ci>ω</m:ci></m:math>. 
	    </item>
	    

	    <item xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">
	      Compute the roots of 
	      <m:math>
		<m:apply>
		  <m:ci type="fn">Q</m:ci>
		  <m:ci>z</m:ci>
		</m:apply>
	      </m:math>, which should come in unit-circle-symmetric
	      pairs 	    
	      <m:math>
		<m:apply>
		  <m:set>
		    <m:ci><m:msub>
		      <m:mi>a</m:mi>
		      <m:mi>i</m:mi>
		    </m:msub></m:ci>
		    <m:apply>
		      <m:divide/>
		      <m:cn>1</m:cn>
		      <m:apply>
			<m:conjugate/>
			<m:ci><m:msub>
			  <m:mi>a</m:mi>
			  <m:mi>i</m:mi>
			</m:msub></m:ci>
		      </m:apply>
		    </m:apply>
		  </m:set>
		</m:apply>
	      </m:math>. Then collect the roots with magnitude less than
	      one into filter 
	      <m:math>
		<m:apply>
		  <m:ci type="fn">
		    <m:msub>
		      <m:mover accent="true">
			<m:mi>H</m:mi>
			<m:mo>^</m:mo>
		      </m:mover>
		      <m:mn>0</m:mn>
		    </m:msub>
		  </m:ci>
		  <m:ci>z</m:ci>
		</m:apply>
	      </m:math>.
	    </item>


	    <item xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/">
	      <m:math>
		<m:apply>
		  <m:ci type="fn">
		    <m:msub>
		      <m:mover accent="true">
			<m:mi>H</m:mi>
			<m:mo>^</m:mo>
		      </m:mover>
		      <m:mn>0</m:mn>
		    </m:msub>
		  </m:ci>
		  <m:ci>z</m:ci>
		</m:apply>
	      </m:math> is the desired prototype filter except for a
	      scale factor.  Recall that we desire 
	      <m:math display="block">
		<m:apply>
		  <m:eq/>
		  <m:apply>
		    <m:plus/>
		    <m:apply>
		      <m:power/>
		      <m:apply>
			<m:abs/>
			<m:apply>
			  <m:ci type="fn">
			    <m:msub>
			      <m:mi>H</m:mi>
			      <m:mn>0</m:mn>
			    </m:msub>
			  </m:ci>
			  <m:apply>
			    <m:exp/>
			    <m:apply>
			      <m:times/>
			      <m:imaginaryi/>
			      <m:ci>ω</m:ci>
			    </m:apply>
			  </m:apply>
			</m:apply>
		      </m:apply>
		      <m:cn>2</m:cn>
		    </m:apply>		
		    <m:apply>
		      <m:power/>
		      <m:apply>
			<m:abs/>
			<m:apply>
			  <m:ci type="fn">
			    <m:msub>
			      <m:mi>H</m:mi>
			      <m:mn>0</m:mn>
			    </m:msub>
			  </m:ci>
			  <m:apply>
			    <m:exp/>			
			    <m:apply>
			      <m:times/>
			      <m:imaginaryi/>		    
			      <m:apply>
				<m:minus/>
				<m:pi/>
				<m:ci>ω</m:ci>
			      </m:apply>
			    </m:apply>
			  </m:apply>
			</m:apply>
		      </m:apply>
		      <m:cn>2</m:cn>
		    </m:apply>
		  </m:apply>
		  <m:cn>1</m:cn>
		</m:apply>
	      </m:math>

	      Using <cnxn xmlns:md="http://cnx.rice.edu/mdml/0.4" xmlns:m="http://www.w3.org/1998/Math/MathML" xmlns:bib="http://bibtexml.sf.net/" document="m10769" strength="8" target="parsec">Parseval's Theorem</cnxn>, we see that
	      <m:math>
		<m:apply>
		  <m:set>
		    <m:apply>
		      <m:ci type="fn" class="discrete">
			<m:msub>
			  <m:mover accent="true">
			    <m:mi>h</m:mi>
			    <m:mo>^</m:mo>
			  </m:mover>
			  <m:mn>0</m:mn>
			</m:msub>
		      </m:ci>
		      <m:ci>n</m:ci>
		    </m:apply>
		  </m:set>
		</m:apply>
	      </m:math> should be scaled to give 
	      <m:math>
		<m:apply>
		  <m:set>
		    <m:apply>
		      <m:ci type="fn" class="discrete">
			<m:msub>
			  <m:mi>h</m:mi>
			  <m:mn>0</m:mn>
			</m:msub>
		      </m:ci>
		      <m:ci>n</m:ci>
		    </m:apply>
		  </m:set>
		</m:apply>
	      </m:math> 
	      for which 
	      
	      <m:math>
		<m:apply>
		  <m:eq/>
		  <m:apply>
		    <m:sum/>
		    <m:bvar>
		      <m:ci>n</m:ci>
		    </m:bvar>
		    <m:lowlimit>
		      <m:cn>0</m:cn>
		    </m:lowlimit>
		    <m:uplimit>
		      <m:apply>
			<m:minus/>
			<m:ci>N</m:ci>
			<m:cn>1</m:cn>
		      </m:apply>
		    </m:uplimit>
		    <m:apply>
		      <m:power/>
		      <m:apply> 
			<m:ci type="fn" class="discrete">
			  <m:msub>
			    <m:mi>h</m:mi>
			    <m:mn>0</m:mn>
			  </m:msub>
			</m:ci>
			<m:ci>n</m:ci>
		      </m:apply>
		      <m:cn>2</m:cn>
		    </m:apply>
		  </m:apply>
		  <m:apply>
		    <m:divide/>
		    <m:cn>1</m:cn>
		    <m:cn>2</m:cn>
		  </m:apply>
		</m:apply>
	      </m:math>.
	    </item>
	  </list>
	</para>
	
      </section>  
    </section>
  </content>
  
</document>
