I have banding in my images. I suspect its introduced by the bias frames - but it might be the darks.
Have a look at the attached screenshot. This is an integration of 60 x 5 minute subs. Note the horizontal bands.
I can eliminate the bands by ditching the bias frames - but then this plays havoc with my flat division. As can be seen from the second attached screenshot.
Any thoughts? The Superbias module in PI gives me a similar result to the second attachment.
I'm certainly not in the PI expert camp, but that looks like (rotated) column defects in the sensor to me - my guess is that will probably come from the flats. Maybe you could check the process for making the masterflat - could there be bias, dark or pedestal in there where it should not be - or missing where it should be included?
FWIW, I found that there are quite few settings that have to be right or the master cal images can get screwed up - I ended up using the following to the letter and all went well: http://www.pixinsight.com/tutorials/...mes/index.html good luck.
I'm certainly not in the PI expert camp, but that looks like (rotated) column defects in the sensor to me - my guess is that will probably come from the flats. Maybe you could check the process for making the masterflat - could there be bias, dark or pedestal in there where it should not be - or missing where it should be included?
FWIW, I found that there are quite few settings that have to be right or the master cal images can get screwed up - I ended up using the following to the letter and all went well: http://www.pixinsight.com/tutorials/...mes/index.html good luck.
Thanks Ray
my diagnosis was the same.
I spent several hours playing with the Cosmetic Correction module in PI yesterday - with some success but its not as good as I'd like.
I am now taking new flats and will start from scratch with new bias and dark frames too.
That link is great - I have followed it to the letter to assemble the masters but I may have a setting wrong.
Will see how that goes
Its a pain - I have about 4.5 hours of each of R, G and B and a bit more Lum - I find that its my processing skills that are letting me down rather than my equipment - which is probably better than the other way around
Its a pain - I have about 4.5 hours of each of R, G and B and a bit more Lum - I find that its my processing skills that are letting me down rather than my equipment - which is probably better than the other way around
Pete
Definitely better this way! Bad data is always going to be bad data, but processing, no matter how bad, can always be turned into great processing.
Definitely better this way! Bad data is always going to be bad data, but processing, no matter how bad, can always be turned into great processing.
You are not wrong there Geoff
The odd feature of this is I am now processing some data for Centaurus A - less RGB but more Lum - probably 3 hours each RGB and 5 Lum. I used the cosmetic correction module in PI - absolutely no lines to be seen.
I might go back and check my data to see if it is a feature of some images but not others
Hi Pete,
Take a short series of bias frames every night. See if the artifacts come and go (electronics flakyness or interference?) or are related to ambient temp. If the pattern never changes it's in the chip and you can calibrate it out.
Does the STX have the readout register on the side? I know the ST4000 did. Those single pixel dark streaks look like they are from hot pixel bleed (quite common in KAF chips), they are only found in columns. So either the image has been rotated 90 deg by processing, or the readout register is on the side.
I've attached a sample master bias frame from my camera (Apogee U16 with KAF16803, readout register at bottom), scaled down 25% and screen stretched to 1 ADU range (i.e. black = 1406.4 ADU and white = 1407.4 ADU) this is from 48 bias frames at -30C. The black lines are from non-uniformity in the readout register itself.
As a zero second exposure is not possible with SBIG cameras, I am puzzled as to why anyone would want to use them, as they are already contained in the dark frame data.
A bias frame is simply a zero second dark, hence if you take any dark frame it consists of: a zero-second dark + x-second dark.
Also the dark map in SBIG cameras, from what I understand, while superbly consistent, is not truly linear.
Hence best practice to to use a library of darks taken at temperature x, and exposure time T, that obviously need to match your lights.
I've found much more value in simply getting excellent quality flats (I prefer sky flats....in fact they were the only flats that consistently worked with a back-illuminated chip I was running some years back )
Playing with Bias frames can often induce a bunch of artefacts that simply aren't in the data. I personally never use them.
Peter may have identified the problem. I understand that Pixinsight automatically compensates for variations in chip temperature and exposure time by scaling the master dark, so bias subtraction is essential. If the SBIG cameras cannot provide bias data or if they have non-linear temperature/time profiles, the Pixinsight calibration scaling algorithm will not work properly. Not sure what you can do about it though, short of using some other software to calibrate your data, using a set of darks that exactly match the times and temperature of your light subs, as Peter suggests.
Maybe you could try CCDStack to dark scale your images and see if that gives a different result to PixInsight. I Use CCDStack for this process with great success. I use a master dark from 16 x 30min dark exposures with an SBIG STL11000.
As a zero second exposure is not possible with SBIG cameras, I am puzzled as to why anyone would want to use them, as they are already contained in the dark frame data.
A bias frame is simply a zero second dark, hence if you take any dark frame it consists of: a zero-second dark + x-second dark.
Also the dark map in SBIG cameras, from what I understand, while superbly consistent, is not truly linear.
Hence best practice to to use a library of darks taken at temperature x, and exposure time T, that obviously need to match your lights.
I've found much more value in simply getting excellent quality flats (I prefer sky flats....in fact they were the only flats that consistently worked with a back-illuminated chip I was running some years back )
Playing with Bias frames can often induce a bunch of artefacts that simply aren't in the data. I personally never use them.
Cheers
Peter
Understood Peter. I took the opportunity of lousy weather up at Coona yesterday to take a huge number of flats with an EL panel - 80 - 100 of each. The download pipe has been running hot. I've also taken plenty more darks.
Unfortunately, I have yet to find a way to calibrate files in PI without using a bias frame. Without it, the flats overcorrect. Time for some more research.
Quote:
Originally Posted by Shiraz
Pete
Peter may have identified the problem. I understand that Pixinsight automatically compensates for variations in chip temperature and exposure time by scaling the master dark, so bias subtraction is essential. If the SBIG cameras cannot provide bias data or if they have non-linear temperature/time profiles, the Pixinsight calibration scaling algorithm will not work properly. Not sure what you can do about it though, short of using some other software to calibrate your data, using a set of darks that exactly match the times and temperature of your light subs, as Peter suggests.
Ray, you are right. There may be a way to do this in PI but I have yet to find it. I am trying to use PI for all processing and have overcome the UI hurdle. Now its down to getting it to play nicely.
At the weekend I compiled a new library of dark, bias and flat frames. A heap of them.
I then compiled master dark and master bias frames in PI last night. They were integrated separately ie the bias frames didn't come near the darks.
When severely stretched, I see readout artefacts in the same columns for both the master dark and master bias frames. They also line up with the lights.
So the bias frames aren't adding to the problem - its inherent in the camera readout.
So I'm back to 2 alternatives - a hot pixel that gets shuffled along a column till its read - assuming that the KAF16803 chip is read along the bottom rather than at the side - or I have a few column defects.
Could it be anything else?
Playing with Cosmetic Correction in PI last night, I also found that the most troublesome defect - which only extends about 1/3 of the way up a frame - wasn't being corrected as I had set the column number incorrectly. Moving it by 1 pixel seemed to deal with it. I need to play with it a bit more this evening but I'm hopeful that I can simply remove these in calibration.
I've posted on the other thread dealing with overscan. I'd like to play with that feature in PI but I don't think that it will solve this readout issue.
Hi Pete,
Take a short series of bias frames every night. See if the artifacts come and go (electronics flakyness or interference?) or are related to ambient temp. If the pattern never changes it's in the chip and you can calibrate it out.
Does the STX have the readout register on the side? I know the ST4000 did. Those single pixel dark streaks look like they are from hot pixel bleed (quite common in KAF chips), they are only found in columns. So either the image has been rotated 90 deg by processing, or the readout register is on the side.
I've attached a sample master bias frame from my camera (Apogee U16 with KAF16803, readout register at bottom), scaled down 25% and screen stretched to 1 ADU range (i.e. black = 1406.4 ADU and white = 1407.4 ADU) this is from 48 bias frames at -30C. The black lines are from non-uniformity in the readout register itself.
Good luck,
EB
Eric
I must have missed your post yesterday.
What you present is precisely what I see in my bias frames. They don't vary from night to night.
If its the readout register itself, then presumably you calibrate them out. Do flats and darks alone deal with this or do you have some form of defect map applied as well?
The flats need to have their bias subtracted, by either a master bias or a master short dark (sometimes called flat-darks), I never bother with flat-darks since the sensor is at -30C and there isn't much dark accumulation in 10 seconds! Ideally flats are calibrated before being combined into a master when using Sigma or SD combine.
All your images and your cal files have that built-in bias. This needs to be removed in order to scale a dark frame or divide/multiply two frames (flatten an image) otherwise the bias offset will contaminate the division. Darks only need to have the bias removed if they are to be scale to a different exposure length.
Bias frames are the easiest to acquire, there is no reason not to use them, unless of course the camera electronics drift!