Hello again,
I tried hard to figure out, why my coadded images are looking so bad with the new version of weightwatcher.
I played around with the configs, checked and tested, but with every configuration I used, the old version (1.3) ran trough and produced perfect weight maps, while the new version (1.8.0) produced bad ones...
While looking at the weight maps produced, at first the new ones do not look that different, although some differences are visible.
Then I decided to blink the weights created by the new and old one, and I saw the cause for the problems: The weightmaps created by the new version are shifted in x-direction! I don't know why that is, but that causes my problems.
I will put my files and config onto [1], so you can see the problem yourself.
Is there anything I can do about that or am I forced to use the old version again?
Best regards Jan
[1]
http://www.aip.de/~jankoh/ww/
I got it!
The problem is in the "copyright" keywords (SOFTAUTH, SOFTVERS and so on) present in your input weight (correctly created by WeightWatcher) and not well managed during the re-processing.
Quote:Is there anything I can do about that or am I forced to use the old version again?
There is a test you can do for me.
Can you download the last version (1.8.3) from the SVN repository and see what happen?
If you have good results I will finalised the release.
Thanks.
Chiara.
Chiara Marmo Wrote:There is a test you can do for me.
Can you download the last version (1.8.3) from the SVN repository and see what happen?
My testing image ran through without any (visible) problems using the SVN version. I will have a deeper look at the WEIGHT-files on Monday, but I think everything is fine now.
Thanks and best regards Jan
jankoh Wrote:I will have a deeper look at the WEIGHT-files on Monday, but I think everything is fine now.
Confirmed from my side. The weights are looking fine now, so I think the bug is resolved.
Thanks again, best regards Jan
But another bug opened.
I some images the weight are still not created correctly. It's also a shift.
At this time, its the globalweight.fits, which is wrong, created out of skyflat.fits and dark.fits and fringe.fits image. The funny thing is, while creating the individual weights out of the globalweights, the individual weight is shifted once again.
I'll put the images onto
http://www.aip.de/~jankoh/ww/ once again.
Best regards Jan
I found something wrong yesterday evening in different images....
version 1.8.4 is available on
WeightWatcher page.
Hoping that this time the fix is good for all images in the universe...
Thanks for your patience.
Chiara.
Sorry to bother you again, but...
I had a quick look over the images my colleague processed while I was away, and they still are broken.
I'll put them on the usual place in a couple of minutes.
Best regards Jan
P.S.: Please note, that even the globalweight was created wrong.
Ok, I downloaded your images: I try a simple
~/software/weightwatcher/src/ww -WEIGHT_NAMES skyflat.fits -WEIGHT_MIN 14000 -WEIGHT_MAX 20000 -WEIGHT_OUTFLAGS 0 -FLAG_NAMES "" (skyflat.fits is an old skyflat you made available to me) and I have no problems...
I passed two times skyflat.fits in weightwatcher and I produce always right images: I think I do not catch how do you produce the "globalweight"...
Can you please send to me the exact command line and configuration file you used for the first WeightWatcher manipulation?
Thanks a lot.
Chiara.
OK, it was a bit difficult to get the code out, as the config is generated on the fly.
This is also *not* the same flat, which I gave you some time ago, but the problem seems to be more or less general, as I see the problem in many files.
Globalweight is generated the following:
ww -WEIGHT_NAMES SKYFLAT_norm_1.fits -WEIGHT_MIN 0.5 -WEIGHT_MAX 1.5 -WEIGHT_OUTFLAGS 2 -FLAG_NAMES "" -FLAG_MASKS "0x0" -FLAG_OUTFLAGS "0" -POLY_NAMES "" -POLY_OUTFLAGS "" -OUTWEIGHT_NAME globalweight_1.fits -OUTFLAG_NAME globalflag_1.fits
(There is a additional input flag image, I do not have available right now, but the error is viewable with an without the presence of that file.)
After the creation of the global weight, it is used with a sextractor generated cosmics image (mostly zeros and higher values for pixels with identified cosmics) and the image itself (for masking saturated pixels) to generate the individual *weight.fits for every single file.
I will provide my SKYFLAT file, at the usual location, I just tested and it gave me the wrong weight file.
Thanks in advance, Jan
The problem was (again!) in the length of the headers of the input flat.
Your (
) new version of WeightWatcher is available on the
svn repository.
Can you tell me if this is the good one?
Maybe, this time I will be patient and wait for an entire processing cycle of your pipeline, before to release the .rpm....
Thanks a lot!
Chiara.