Recent changes to WSnapshotAlgorithm

WSnapshotAlgorithm modified by André Offringa

André Offringa — Sun, 23 Aug 2015 09:27:42 -0000

--- v9
+++ v10
@@ -1,8 +1,10 @@
-I've implemented an algorithm that has approximately the same performance as the W-snapshot algorithm suggested by Cornwell et al. (2012). The algorithm in WSClean is slightly different though. In WSClean, it consists of phase rotating the visibilities to zenith, and then recentring the image during w-stacking. Mathematical details are explained in [the WSClean paper](http://arxiv.org/abs/1407.1943). It seems to be worthwhile for MWA snapshots of a few minutes with images of 3072 x 3072 pixels at zenith angles > 20 degree, causing a speed-up of about a factor of 3 at zenith angles > 45 degrees. For larger images the speed-up will be greater.
+I've implemented an algorithm that has approximately the same performance as the W-snapshot algorithm suggested by Cornwell et al. (2012). The algorithm in WSClean is slightly different though. In WSClean, it consists of phase rotating the visibilities to zenith, and then recentring the image during w-stacking. Mathematical details are explained in [the WSClean paper](http://arxiv.org/abs/1407.1943). It seems to be worthwhile for MWA snapshots of a few minutes with images of 3072 x 3072 pixels at zenith angles > 20 degree, and provides a speed-up of about a factor of 3 at zenith angles > 45 degrees. For larger images the speed-up will be greater.

-To use this method, start by making a copy of your measurement set. This measurement set should be phased up to the phase centre that you want to image (as it normally would be). Then, run the chgcentre tool with the following parameters:
+To use this method, start by making a copy of your measurement set. This measurement set should be phased up to the phase centre that you want to image (as it normally would be). Then, run [the chgcentre tool](chgcentre) with the following parameters:

     :::text
     chgcentre -minw -shiftback copy.ms

 This will calculate the optimal projection direction, perform required phase rotations and shifts, and add some keywords to the measurement set. WSClean will recognize the presence of those keywords and perform the recentring, so WSClean can be run in the normal way. Since the visibilities are rewritten by chgcentre in a non-standard projection, you cannot use that measurement set with most other tools anymore (that's why it should be copied).
+
+[The chgcentre page](chgcentre) provides further information about the chgcentre tool.

WSnapshotAlgorithm modified by André Offringa

André Offringa — Sat, 13 Sep 2014 02:12:01 -0000

--- v8
+++ v9
@@ -1,4 +1,4 @@
-I've implemented an algorithm that is similar to the W-snapshot algorithm suggested by Cornwell et al. (2012). The algorithm in WSClean is slightly different though. In WSClean, it consists of phase rotating the visibilities to zenith, and then recentring the image during w-stacking. Mathematical details are explained in [the WSClean paper](http://arxiv.org/abs/1407.1943). It seems to be worthwhile for MWA snapshots of a few minutes with images of 3072 x 3072 pixels at zenith angles > 20 degree, causing a speed-up of about a factor of 3 at zenith angles > 45 degrees. For larger images the speed-up will be greater.
+I've implemented an algorithm that has approximately the same performance as the W-snapshot algorithm suggested by Cornwell et al. (2012). The algorithm in WSClean is slightly different though. In WSClean, it consists of phase rotating the visibilities to zenith, and then recentring the image during w-stacking. Mathematical details are explained in [the WSClean paper](http://arxiv.org/abs/1407.1943). It seems to be worthwhile for MWA snapshots of a few minutes with images of 3072 x 3072 pixels at zenith angles > 20 degree, causing a speed-up of about a factor of 3 at zenith angles > 45 degrees. For larger images the speed-up will be greater.

 To use this method, start by making a copy of your measurement set. This measurement set should be phased up to the phase centre that you want to image (as it normally would be). Then, run the chgcentre tool with the following parameters:

WSnapshotAlgorithm modified by André Offringa

André Offringa — Sat, 13 Sep 2014 01:17:04 -0000

--- v7
+++ v8
@@ -1,4 +1,4 @@
-I've implemented an algorithm that is similar to the W-snapshot algorithm suggested by Cornwell et al. (2012). The algorithm in WSClean is slightly different though. In WSClean, it consists of phase rotating the visibilities to zenith, and then recentring the image during w-stacking. Mathematical details are explained in the WSClean paper. It seems to be worthwhile for MWA snapshots of a few minutes with images of 3072 x 3072 pixels at zenith angles > 20 degree, causing a speed-up of about a factor of 3 at zenith angles > 45 degrees. For larger images the speed-up will be greater.
+I've implemented an algorithm that is similar to the W-snapshot algorithm suggested by Cornwell et al. (2012). The algorithm in WSClean is slightly different though. In WSClean, it consists of phase rotating the visibilities to zenith, and then recentring the image during w-stacking. Mathematical details are explained in [the WSClean paper](http://arxiv.org/abs/1407.1943). It seems to be worthwhile for MWA snapshots of a few minutes with images of 3072 x 3072 pixels at zenith angles > 20 degree, causing a speed-up of about a factor of 3 at zenith angles > 45 degrees. For larger images the speed-up will be greater.

 To use this method, start by making a copy of your measurement set. This measurement set should be phased up to the phase centre that you want to image (as it normally would be). Then, run the chgcentre tool with the following parameters:

WSnapshotAlgorithm modified by André Offringa

André Offringa — Tue, 09 Sep 2014 15:21:22 -0000

--- v6
+++ v7
@@ -1,4 +1,4 @@
-I've implemented an algorithm similar to the W-snapshot algorithm suggested by Cornwell et al. (2012). In WSClean, it consists of phase rotating the visibilities to zenith, and then recentring the image during w-stacking. Mathematical details are explained in the WSClean paper. It seems to be worthwhile for MWA images with 3072 x 3072 pixels at zenith angles > 20 degree., causing a speed-up of about a factor of 3 at zenith angles > 45 degrees. For larger images the speed-up will be greater.
+I've implemented an algorithm that is similar to the W-snapshot algorithm suggested by Cornwell et al. (2012). The algorithm in WSClean is slightly different though. In WSClean, it consists of phase rotating the visibilities to zenith, and then recentring the image during w-stacking. Mathematical details are explained in the WSClean paper. It seems to be worthwhile for MWA snapshots of a few minutes with images of 3072 x 3072 pixels at zenith angles > 20 degree, causing a speed-up of about a factor of 3 at zenith angles > 45 degrees. For larger images the speed-up will be greater.

 To use this method, start by making a copy of your measurement set. This measurement set should be phased up to the phase centre that you want to image (as it normally would be). Then, run the chgcentre tool with the following parameters:

WSnapshotAlgorithm modified by André Offringa

André Offringa — Wed, 20 Aug 2014 11:48:20 -0000

--- v5
+++ v6
@@ -1,6 +1,6 @@
-I've implemented an algorithm similar to the W-snapshot algorithm suggested by Cornwell et al. (2012). In WSClean, it consists of phase rotating the visibilities to zenith, and then recentring the image during w-stacking. Mathematical details are explained in the WSClean paper. It seems to be worthwhile for MWA at zenith angles > 20 degree., causing a speed-up of about a factor of 3 at zenith angles > 45 degrees.
+I've implemented an algorithm similar to the W-snapshot algorithm suggested by Cornwell et al. (2012). In WSClean, it consists of phase rotating the visibilities to zenith, and then recentring the image during w-stacking. Mathematical details are explained in the WSClean paper. It seems to be worthwhile for MWA images with 3072 x 3072 pixels at zenith angles > 20 degree., causing a speed-up of about a factor of 3 at zenith angles > 45 degrees. For larger images the speed-up will be greater.

-To do this, start by making a copy of your measurement set. This measurement set should be phased up to the phase centre that you want to image (as it normally would be). Then, run the chgcentre tool with the following parameters:
+To use this method, start by making a copy of your measurement set. This measurement set should be phased up to the phase centre that you want to image (as it normally would be). Then, run the chgcentre tool with the following parameters:

     :::text
     chgcentre -minw -shiftback copy.ms

WSnapshotAlgorithm modified by André Offringa

André Offringa — Fri, 27 Jun 2014 00:42:24 -0000

--- v4
+++ v5
@@ -1,6 +1,6 @@
 I've implemented an algorithm similar to the W-snapshot algorithm suggested by Cornwell et al. (2012). In WSClean, it consists of phase rotating the visibilities to zenith, and then recentring the image during w-stacking. Mathematical details are explained in the WSClean paper. It seems to be worthwhile for MWA at zenith angles > 20 degree., causing a speed-up of about a factor of 3 at zenith angles > 45 degrees.

-To do this, make a copy of your measurement set(!). This measurement set should be phased up to the phase centre that you want to image (as it normally would be). Then, run the chgcentre with the following parameters:
+To do this, start by making a copy of your measurement set. This measurement set should be phased up to the phase centre that you want to image (as it normally would be). Then, run the chgcentre tool with the following parameters:

     :::text
     chgcentre -minw -shiftback copy.ms

WSnapshotAlgorithm modified by André Offringa

André Offringa — Wed, 25 Jun 2014 10:18:58 -0000

--- v3
+++ v4
@@ -5,4 +5,4 @@
     :::text
     chgcentre -minw -shiftback copy.ms

-This will calculate the optimal projection direction, perform required phase rotations and shifts, and add some keywords to the measurement set. WSClean will recognize the presence of those keywords and perform the recentring, so WSClean can be run in the normal way. Since the visbilities are rewritten by chgcentre in a non-standard projection, you cannot use that measurement set with any other tool anymore (that's why it should be copied).
+This will calculate the optimal projection direction, perform required phase rotations and shifts, and add some keywords to the measurement set. WSClean will recognize the presence of those keywords and perform the recentring, so WSClean can be run in the normal way. Since the visibilities are rewritten by chgcentre in a non-standard projection, you cannot use that measurement set with most other tools anymore (that's why it should be copied).

WSnapshotAlgorithm modified by André Offringa

André Offringa — Tue, 10 Jun 2014 11:03:38 -0000

--- v2
+++ v3
@@ -5,4 +5,4 @@
     :::text
     chgcentre -minw -shiftback copy.ms

-This will calculate the optimal projection direction, perform required phase rotations and shifts, and add some keywords to the measurement set. WSClean will recognize the presence of those keywords and perform the recentring, so WSClean can be run in the normal way. Since the visbilities are rewritten by chgcentre in a non-standard projection, you cannot use that measurement set with any other tool anymore (that's why it should be copied). These timings speed-ups include the time that chgcentre takes. 
+This will calculate the optimal projection direction, perform required phase rotations and shifts, and add some keywords to the measurement set. WSClean will recognize the presence of those keywords and perform the recentring, so WSClean can be run in the normal way. Since the visbilities are rewritten by chgcentre in a non-standard projection, you cannot use that measurement set with any other tool anymore (that's why it should be copied).

WSnapshotAlgorithm modified by André Offringa

André Offringa — Sat, 15 Mar 2014 23:19:59 -0000

--- v1
+++ v2
@@ -2,6 +2,7 @@

 To do this, make a copy of your measurement set(!). This measurement set should be phased up to the phase centre that you want to image (as it normally would be). Then, run the chgcentre with the following parameters:

-chgcentre -minw -shiftback copy.ms
+    :::text
+    chgcentre -minw -shiftback copy.ms

 This will calculate the optimal projection direction, perform required phase rotations and shifts, and add some keywords to the measurement set. WSClean will recognize the presence of those keywords and perform the recentring, so WSClean can be run in the normal way. Since the visbilities are rewritten by chgcentre in a non-standard projection, you cannot use that measurement set with any other tool anymore (that's why it should be copied). These timings speed-ups include the time that chgcentre takes.

WSnapshotAlgorithm modified by André Offringa

André Offringa — Sat, 15 Mar 2014 07:03:15 -0000

I've implemented an algorithm similar to the W-snapshot algorithm suggested by Cornwell et al. (2012). In WSClean, it consists of phase rotating the visibilities to zenith, and then recentring the image during w-stacking. Mathematical details are explained in the WSClean paper. It seems to be worthwhile for MWA at zenith angles > 20 degree., causing a speed-up of about a factor of 3 at zenith angles > 45 degrees.

To do this, make a copy of your measurement set(!). This measurement set should be phased up to the phase centre that you want to image (as it normally would be). Then, run the chgcentre with the following parameters:

chgcentre -minw -shiftback copy.ms

This will calculate the optimal projection direction, perform required phase rotations and shifts, and add some keywords to the measurement set. WSClean will recognize the presence of those keywords and perform the recentring, so WSClean can be run in the normal way. Since the visbilities are rewritten by chgcentre in a non-standard projection, you cannot use that measurement set with any other tool anymore (that's why it should be copied). These timings speed-ups include the time that chgcentre takes.