<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	>
<channel>
	<title>Comments on: Antialiasing filters and multirate systems</title>
	<atom:link href="http://eli.thegreenplace.net/2006/05/10/antialiasing-filteres-and-multirate-systems/feed/" rel="self" type="application/rss+xml" />
	<link>http://eli.thegreenplace.net/2006/05/10/antialiasing-filteres-and-multirate-systems/</link>
	<description>Eli Bendersky's personal website</description>
	<pubDate>Thu, 04 Dec 2008 01:36:56 +0000</pubDate>
	<generator>http://wordpress.org/?v=2.6.2</generator>
		<item>
		<title>By: John Handwork</title>
		<link>http://eli.thegreenplace.net/2006/05/10/antialiasing-filteres-and-multirate-systems/#comment-82</link>
		<dc:creator>John Handwork</dc:creator>
		<pubDate>Wed, 10 May 2006 17:34:13 +0000</pubDate>
		<guid isPermaLink="false">http://eli.thegreenplace.net/2006/05/10/antialiasing-filteres-and-multirate-systems/#comment-82</guid>
		<description>Rather than thinking of the filter for the oversampled system as providing cutoff at 1/2 the higher sampling rate (Fsh), think of the filter requirements delivering a passband up to the end of your frequency of interest - 10 kHz in your examples - and a stopband beyond Fsh-10kHz (or repeating stopband locations of n*Fsh+/-10kHz).

I wouldn't think it's a "common" misconception that antialiasing can be accommodated entirely in the digital domain.  *Anyone* who has been introduced to sampling systems *should* know about the frequency domain "folding" that provides aliazing.

Otherwize, it's a decent synopsis.

One additional note:  halfband filters are used heavily in digital systems such that sampling at your frequency of interest multiplied by 2^n may make the implementation easier.  It is often more compact to use 3 halfband filters to decimate by 8 rather than decimating by 5 in one step.  What I didn't know about halfband filters until a few days ago is that they aren't all the same; the first halfband filter only needs to pass the small bandwidth for the frequency of interest and reject the small stopband around Fsh.

- John_H</description>
		<content:encoded><![CDATA[<p>Rather than thinking of the filter for the oversampled system as providing cutoff at 1/2 the higher sampling rate (Fsh), think of the filter requirements delivering a passband up to the end of your frequency of interest - 10 kHz in your examples - and a stopband beyond Fsh-10kHz (or repeating stopband locations of n*Fsh+/-10kHz).</p>
<p>I wouldn&#8217;t think it&#8217;s a &#8220;common&#8221; misconception that antialiasing can be accommodated entirely in the digital domain.  *Anyone* who has been introduced to sampling systems *should* know about the frequency domain &#8220;folding&#8221; that provides aliazing.</p>
<p>Otherwize, it&#8217;s a decent synopsis.</p>
<p>One additional note:  halfband filters are used heavily in digital systems such that sampling at your frequency of interest multiplied by 2^n may make the implementation easier.  It is often more compact to use 3 halfband filters to decimate by 8 rather than decimating by 5 in one step.  What I didn&#8217;t know about halfband filters until a few days ago is that they aren&#8217;t all the same; the first halfband filter only needs to pass the small bandwidth for the frequency of interest and reject the small stopband around Fsh.</p>
<p>- John_H</p>
]]></content:encoded>
	</item>
</channel>
</rss>
