<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Recent changes to ImageWeighting</title><link>https://sourceforge.net/p/wsclean/wiki/ImageWeighting/</link><description>Recent changes to ImageWeighting</description><atom:link href="https://sourceforge.net/p/wsclean/wiki/ImageWeighting/feed" rel="self"/><language>en</language><lastBuildDate>Fri, 23 Nov 2018 18:53:21 -0000</lastBuildDate><atom:link href="https://sourceforge.net/p/wsclean/wiki/ImageWeighting/feed" rel="self" type="application/rss+xml"/><item><title>ImageWeighting modified by André Offringa</title><link>https://sourceforge.net/p/wsclean/wiki/ImageWeighting/</link><description>&lt;div class="markdown_content"&gt;&lt;pre&gt;--- v11
+++ v12
@@ -17,6 +17,11 @@
 ####Weighting rank filter
 The weighting rank filter works in combination with one of the other weighting modes, and suppresses visibilities that would receive a much larger weight than its uv-neighbouring visibilities. This is basically done by smoothing the weights -- see the [weight rank filter page](WeightRankFilter) for a detailed description. In uniform weighting, some visibilities can receive order of magnitudes more weight, causing these visibilities to dominate the image noise. By smoothing the weights, this issue is mitigated. The weighting rank filter does the smoothing in such a way that the PSF is not visibly affected.

+####Relation to visibility weights
+In WSClean, image weighting normally works 'on top of' the visibility weights stored in the measurement set. This in particular means that uniform weighting will weight all scales (baselines) the same independent of their visibility weight. One therefore cannot increase the weight on long baselines by using uniform weighting and increasing the visibility weights of those baselines. The visibility weights nevertheless still have a function: they determine the weight in their particular uv-cell with respect to other visibilities *in that same uv-cell*. All in all, this means that WSClean assumes that the visibility weights specify the inverse variance weight of the visibility. Up or downweighting of scales has to be done by the image weighting (which can include [tapering](Tapering)). I've had reports that this behaviour is not consistent with how CASA treats the combination of visibility weights and imaging weights. 
+
+This behaviour can be altered by using the `-use-weights-as-taper` option. In that case, imaging weights will be determined without taking the visibility weights into account. Uniform weighting, for example, will make the visibilities have uniform weights before the visibility weights have been applied, and the visibility weights are applied afterwards. In this case, the visibility weights can be used to increase/decrease the weights of certain scales.
+
 ####Other weighting-related settings
 * [MFS weighting](MFSWeighting) -- If the bandwidth is split in multiple subbands for imaging (e.g. with `-channels-out`), it is good to know about MFS weighting.
 * [Tapers](Tapering) -- can be used to further shape the synthesized beam.
&lt;/pre&gt;
&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">André Offringa</dc:creator><pubDate>Fri, 23 Nov 2018 18:53:21 -0000</pubDate><guid>https://sourceforge.net3fb26d217bcf71867debb7502b57d1ecdd5e728b</guid></item><item><title>ImageWeighting modified by André Offringa</title><link>https://sourceforge.net/p/wsclean/wiki/ImageWeighting/</link><description>&lt;div class="markdown_content"&gt;&lt;pre&gt;--- v10
+++ v11
@@ -10,7 +10,7 @@
 Briggs' weighting is a compromise between uniform and natural weighting. A parameter, called the robustness parameter, selects the desired level between uniform and natural. Typical values are between -1 and 1. Lower values give more uniform-like weights, higher values give more natural-like weights.  Note that values of -1 / 1 do not *equal* uniform / natural weighting. Briggs weighting mode is selected with `-weight briggs &amp;lt;robustness&amp;gt;`, for example `-weight briggs 0.5`.

 ####Super-uniform and super-Briggs weighting
-In some cases, slightly better results can be obtained by using super weighting. Super weighting refers to performing the counting of visibilities on larger grid cells, such that neighbouring uv-samples share their weight. This can be selected with `superweight &amp;lt;value&amp;gt;`. The value represents how much larger the weight-counting cell should be. CASA's super uniform weighting mode is equivalent to `-superweight 3 -weight uniform`. Additionally, it is also possible to perform sub-uniform or sub-Briggs weighting, by giving a value less than one. The super-weighting can also be used in combination with Briggs' weighting. Super-weighting was implemented because many astronomers are familiar with the "super uniform" weighting option in CASA. In many cases, the weighting rank filter described below is a better option.
+In some cases, slightly better results can be obtained by using super weighting. Super weighting refers to performing the counting of visibilities on larger grid cells, such that neighbouring uv-samples share their weight. This can be selected with `-superweight &amp;lt;value&amp;gt;`. The value represents how much larger the weight-counting cell should be. CASA's super uniform weighting mode is equivalent to `-superweight 3 -weight uniform`. Additionally, it is also possible to perform sub-uniform or sub-Briggs weighting, by giving a value less than one. The super-weighting can also be used in combination with Briggs' weighting. Super-weighting was implemented because many astronomers are familiar with the "super uniform" weighting option in CASA. In many cases, the weighting rank filter described below is a better option.

 Super-pixel weighting can be nice on very large (full-sky) images, since the immense uv resolution might otherwise start to make your beam become natural-ish weighted. [Dan Briggs' thesis](http://www.aoc.nrao.edu/dissertations/dbriggs/) describes sub/super-pixel weighting very well.

&amp;lt;/value&amp;gt;&amp;lt;/value&amp;gt;&amp;lt;/robustness&amp;gt;&lt;/pre&gt;
&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">André Offringa</dc:creator><pubDate>Wed, 19 Sep 2018 12:34:00 -0000</pubDate><guid>https://sourceforge.nete8f5ea4b6efaa79bad0b610e13d989440ab79d5f</guid></item><item><title>ImageWeighting modified by André Offringa</title><link>https://sourceforge.net/p/wsclean/wiki/ImageWeighting/</link><description>&lt;div class="markdown_content"&gt;&lt;pre&gt;--- v9
+++ v10
@@ -15,11 +15,7 @@
 Super-pixel weighting can be nice on very large (full-sky) images, since the immense uv resolution might otherwise start to make your beam become natural-ish weighted. [Dan Briggs' thesis](http://www.aoc.nrao.edu/dissertations/dbriggs/) describes sub/super-pixel weighting very well.

 ####Weighting rank filter
-The weighting rank filter works in combination with one of the other weighting modes, and suppresses visibilities that would receive a much larger weight than its uv-neighbouring visibilities. This is basically done by smoothing the weights. In uniform weighting, some visibilities can receive order of magnitudes more weight, causing these visibilities to dominate the image noise. By smoothing the weights, this issue is mitigated. The weighting rank filter does the smoothing in such a way that the PSF is not visibly affected.
-
-The syntax for the weighting rank filter is `-weighting-rank-filter &amp;lt;level&amp;gt;`. The 'level' parameter specifies how much deviation is allowed. A value of '3' would trim any value larger than three times the local medium weight value, and set it to that value.  Additionally, the size over which the medium is calculated can be set with `-weighting-rank-filter-size &amp;lt;size&amp;gt;`.
-
-For uniform weighting, it is almost always desirable to have a weak rank filter set, e.g. -weighting-rank-filter 3.
+The weighting rank filter works in combination with one of the other weighting modes, and suppresses visibilities that would receive a much larger weight than its uv-neighbouring visibilities. This is basically done by smoothing the weights -- see the [weight rank filter page](WeightRankFilter) for a detailed description. In uniform weighting, some visibilities can receive order of magnitudes more weight, causing these visibilities to dominate the image noise. By smoothing the weights, this issue is mitigated. The weighting rank filter does the smoothing in such a way that the PSF is not visibly affected.

 ####Other weighting-related settings
 * [MFS weighting](MFSWeighting) -- If the bandwidth is split in multiple subbands for imaging (e.g. with `-channels-out`), it is good to know about MFS weighting.
&amp;lt;/size&amp;gt;&amp;lt;/level&amp;gt;&lt;/pre&gt;
&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">André Offringa</dc:creator><pubDate>Wed, 19 Sep 2018 12:12:55 -0000</pubDate><guid>https://sourceforge.net149f54f0abd3c90889d58a953036c0d16a79d515</guid></item><item><title>ImageWeighting modified by André Offringa</title><link>https://sourceforge.net/p/wsclean/wiki/ImageWeighting/</link><description>&lt;div class="markdown_content"&gt;&lt;pre&gt;--- v8
+++ v9
@@ -25,5 +25,6 @@
 * [MFS weighting](MFSWeighting) -- If the bandwidth is split in multiple subbands for imaging (e.g. with `-channels-out`), it is good to know about MFS weighting.
 * [Tapers](Tapering) -- can be used to further shape the synthesized beam.
 * [The weight rank filter](WeightRankFilter) -- used to decrease noise by avoiding giving too much weight to a single visibility.
+* [Continuing deconvolution](ContinueDeconvolution) -- can be used to deconvolve a model with one weighting setting, and image the final result with different weighting settings.

 **Next chapter: [Tapering](Tapering)**
&lt;/pre&gt;
&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">André Offringa</dc:creator><pubDate>Wed, 11 Apr 2018 17:37:10 -0000</pubDate><guid>https://sourceforge.net2532a6da5022ee7f7a987bc581c99798dc31c410</guid></item><item><title>ImageWeighting modified by André Offringa</title><link>https://sourceforge.net/p/wsclean/wiki/ImageWeighting/</link><description>&lt;div class="markdown_content"&gt;&lt;pre&gt;--- v7
+++ v8
@@ -22,7 +22,7 @@
 For uniform weighting, it is almost always desirable to have a weak rank filter set, e.g. -weighting-rank-filter 3.

 ####Other weighting-related settings
-* [MFS weighting](MFSWeighting) -- If the bandwidth is split in multiple subbands for imaging (e.g. with `-channelsout`), it is good to know about MFS weighting.
+* [MFS weighting](MFSWeighting) -- If the bandwidth is split in multiple subbands for imaging (e.g. with `-channels-out`), it is good to know about MFS weighting.
 * [Tapers](Tapering) -- can be used to further shape the synthesized beam.
 * [The weight rank filter](WeightRankFilter) -- used to decrease noise by avoiding giving too much weight to a single visibility.

&lt;/pre&gt;
&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">André Offringa</dc:creator><pubDate>Fri, 01 Dec 2017 11:20:24 -0000</pubDate><guid>https://sourceforge.netb11f9e2b68788f95f4dabdf8ec10541f71f7b03c</guid></item><item><title>ImageWeighting modified by André Offringa</title><link>https://sourceforge.net/p/wsclean/wiki/ImageWeighting/</link><description>&lt;div class="markdown_content"&gt;&lt;pre&gt;--- v6
+++ v7
@@ -10,7 +10,7 @@
 Briggs' weighting is a compromise between uniform and natural weighting. A parameter, called the robustness parameter, selects the desired level between uniform and natural. Typical values are between -1 and 1. Lower values give more uniform-like weights, higher values give more natural-like weights.  Note that values of -1 / 1 do not *equal* uniform / natural weighting. Briggs weighting mode is selected with `-weight briggs &amp;lt;robustness&amp;gt;`, for example `-weight briggs 0.5`.

 ####Super-uniform and super-Briggs weighting
-In some cases, slightly better results can be obtained by using super weighting. Super weighting refers to performing the counting of visibilities on a larger grid cells, such that neighbouring uv-samples share their weight. This can be selected with `superweight &amp;lt;value&amp;gt;`. The value represents how much larger the weight-counting cell should be. CASA's super uniform weighting mode is equivalent to `-superweight 3 -weight uniform`. Additionally, it is also possible to perform sub-uniform or sub-Briggs weighting, by giving a value less than one. The super-weighting can also be used in combination with Briggs' weighting. Super-weighting was implemented because many astronomers are familiar with the "super uniform" weighting option in CASA. In many cases, the weighting rank filter described below is a better option.
+In some cases, slightly better results can be obtained by using super weighting. Super weighting refers to performing the counting of visibilities on larger grid cells, such that neighbouring uv-samples share their weight. This can be selected with `superweight &amp;lt;value&amp;gt;`. The value represents how much larger the weight-counting cell should be. CASA's super uniform weighting mode is equivalent to `-superweight 3 -weight uniform`. Additionally, it is also possible to perform sub-uniform or sub-Briggs weighting, by giving a value less than one. The super-weighting can also be used in combination with Briggs' weighting. Super-weighting was implemented because many astronomers are familiar with the "super uniform" weighting option in CASA. In many cases, the weighting rank filter described below is a better option.

 Super-pixel weighting can be nice on very large (full-sky) images, since the immense uv resolution might otherwise start to make your beam become natural-ish weighted. [Dan Briggs' thesis](http://www.aoc.nrao.edu/dissertations/dbriggs/) describes sub/super-pixel weighting very well.

&lt;/pre&gt;
&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">André Offringa</dc:creator><pubDate>Fri, 01 Dec 2017 11:18:58 -0000</pubDate><guid>https://sourceforge.net57cb6dbc8cba77a7707c6b64ea9896de9281fac4</guid></item><item><title>ImageWeighting modified by André Offringa</title><link>https://sourceforge.net/p/wsclean/wiki/ImageWeighting/</link><description>&lt;div class="markdown_content"&gt;&lt;pre&gt;--- v5
+++ v6
@@ -7,7 +7,7 @@
 In natural weighting, the visibilities are not weighted before gridding, causing scales that have more baselines to dominate the image. Arrays have more small baselines, which will mean that large scales will be more dominantly present. This mode will make sure that the system noise has the least effect in the image, but resolution and sidelobe noise are worse. This mode is activated with option `-weight natural`.

 ####Briggs' weighting
-Briggs' weighting is a compromise between uniform and natural weighting. A parameter, called the robustness parameter, selects the desired level between uniform and natural and is typically between -1 and 1. Lower values give more uniform-like weights, higher values give more natural-like weights. This mode is selected with `-weight briggs &amp;lt;robustness&amp;gt;`, for example `-weight briggs 0.5`.
+Briggs' weighting is a compromise between uniform and natural weighting. A parameter, called the robustness parameter, selects the desired level between uniform and natural. Typical values are between -1 and 1. Lower values give more uniform-like weights, higher values give more natural-like weights.  Note that values of -1 / 1 do not *equal* uniform / natural weighting. Briggs weighting mode is selected with `-weight briggs &amp;lt;robustness&amp;gt;`, for example `-weight briggs 0.5`.

 ####Super-uniform and super-Briggs weighting
 In some cases, slightly better results can be obtained by using super weighting. Super weighting refers to performing the counting of visibilities on a larger grid cells, such that neighbouring uv-samples share their weight. This can be selected with `superweight &amp;lt;value&amp;gt;`. The value represents how much larger the weight-counting cell should be. CASA's super uniform weighting mode is equivalent to `-superweight 3 -weight uniform`. Additionally, it is also possible to perform sub-uniform or sub-Briggs weighting, by giving a value less than one. The super-weighting can also be used in combination with Briggs' weighting. Super-weighting was implemented because many astronomers are familiar with the "super uniform" weighting option in CASA. In many cases, the weighting rank filter described below is a better option.
&lt;/pre&gt;
&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">André Offringa</dc:creator><pubDate>Mon, 08 May 2017 10:09:20 -0000</pubDate><guid>https://sourceforge.net90cf34332b2648425a80009172b46e8bc3e7f602</guid></item><item><title>ImageWeighting modified by André Offringa</title><link>https://sourceforge.net/p/wsclean/wiki/ImageWeighting/</link><description>&lt;div class="markdown_content"&gt;&lt;pre&gt;--- v4
+++ v5
@@ -25,3 +25,5 @@
 * [MFS weighting](MFSWeighting) -- If the bandwidth is split in multiple subbands for imaging (e.g. with `-channelsout`), it is good to know about MFS weighting.
 * [Tapers](Tapering) -- can be used to further shape the synthesized beam.
 * [The weight rank filter](WeightRankFilter) -- used to decrease noise by avoiding giving too much weight to a single visibility.
+
+**Next chapter: [Tapering](Tapering)**
&lt;/pre&gt;
&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">André Offringa</dc:creator><pubDate>Sat, 21 Jan 2017 10:04:54 -0000</pubDate><guid>https://sourceforge.net9f23cc090305a35ef85e7e741716ac07b35bef61</guid></item><item><title>ImageWeighting modified by André Offringa</title><link>https://sourceforge.net/p/wsclean/wiki/ImageWeighting/</link><description>&lt;div class="markdown_content"&gt;&lt;pre&gt;--- v3
+++ v4
@@ -24,3 +24,4 @@
 ####Other weighting-related settings
 * [MFS weighting](MFSWeighting) -- If the bandwidth is split in multiple subbands for imaging (e.g. with `-channelsout`), it is good to know about MFS weighting.
 * [Tapers](Tapering) -- can be used to further shape the synthesized beam.
+* [The weight rank filter](WeightRankFilter) -- used to decrease noise by avoiding giving too much weight to a single visibility.
&lt;/pre&gt;
&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">André Offringa</dc:creator><pubDate>Tue, 23 Feb 2016 15:12:37 -0000</pubDate><guid>https://sourceforge.net70999a47e69172a6a38e55fa82637c3e3aeb16bf</guid></item><item><title>ImageWeighting modified by André Offringa</title><link>https://sourceforge.net/p/wsclean/wiki/ImageWeighting/</link><description>&lt;div class="markdown_content"&gt;&lt;pre&gt;--- v2
+++ v3
@@ -1,4 +1,4 @@
-WSClean supports uniform, natural and Briggs' weighting. These work exactly like in Casa and other imagers:
+WSClean supports uniform, natural and Briggs' weighting. These work exactly like in Casa and other imagers, and are further described below.

 ####Uniform weighting
 Uniform weighting will give each scale the same weight. If multiple visibilities fall in one uv cell, they are downweighted by the number of visibilities in that cell. In this weighting mode, the image will have the highest resolution, but the system noise will not be optimal. This gridmode is the default, but can be explicitly selected with `-weight uniform`.
@@ -20,3 +20,7 @@
 The syntax for the weighting rank filter is `-weighting-rank-filter &amp;lt;level&amp;gt;`. The 'level' parameter specifies how much deviation is allowed. A value of '3' would trim any value larger than three times the local medium weight value, and set it to that value.  Additionally, the size over which the medium is calculated can be set with `-weighting-rank-filter-size &amp;lt;size&amp;gt;`.

 For uniform weighting, it is almost always desirable to have a weak rank filter set, e.g. -weighting-rank-filter 3.
+
+####Other weighting-related settings
+* [MFS weighting](MFSWeighting) -- If the bandwidth is split in multiple subbands for imaging (e.g. with `-channelsout`), it is good to know about MFS weighting.
+* [Tapers](Tapering) -- can be used to further shape the synthesized beam.
&lt;/pre&gt;
&lt;/div&gt;</description><dc:creator xmlns:dc="http://purl.org/dc/elements/1.1/">André Offringa</dc:creator><pubDate>Sat, 30 Jan 2016 23:18:43 -0000</pubDate><guid>https://sourceforge.net9457858bebe688499cc396e0316c3e200f5d2be5</guid></item></channel></rss>