Calibration problems

[gegaus]

Hello,

When calibrating my lights some strange things happens

Im using Vicent's guide to contruct the Bias, Darks and flats.
The problem is that the Bias introduces a lot of noise to the lights and the flats have no effect at all?
Im using RAW(.CR2) files from unmodified Canon EOS 450D.

I`ve tried both with RAW and .fit converted files with the same result.
I think there must be something wrong with my filesettings/mode in PI?
When stacking the same files(RAW) in DeepSkyStacker the result is quite good with noise reduction and flats subtracted.

What am I doing wrong???

Thanks
Geir

[Juan Conejero]

Hi Geir,

Welcome to PixInsight Forum.

Do you know George's (astropixel's) DSLR raw tutorial? It is the reference for DSLR raw preprocessing. I strongly recommend you read it and follow all of the steps he describes:

http://pixinsight.com/forum/index.php?topic=2570.0

Quick tip: How are you loading your raw frames? looks like you are loading them debayered, which makes calibration impossible. In such case, in the ImageCalibration tool, write "raw" (without quotes) in the Input hints field (Format Hints section). This will force ImageCalibration to load pure raw data, irrespective of the current global settings in the DSLR_RAW module.

Let us know if this helps.

[gegaus]

Hello and thanks Juan.

I have read the tutorial and I also use pure raw flles but I think I have found whats causing the strange behavior.
The Bias files where old so I made some new ones and they did wounders

The flats issue is not that clear but it seems like the files are to "flat" and when they are subtracted they have no impact on the lights.
Im using an EL-panel of a rather cheap kind who is supposed to send out white light
I blend it down a little with paper sheets between the panel and the telescope so the exposure time is about 1/4sec.
Can it be that the color channels are to far spread out in the image?
Any tips on how to generate usable flats??

Regards
Geir

[astropixel]

Hello Geir.

A fresh set usually does the trick. If you compose a bias of 50 frames this should last for months, because of the higher sampling rate.

Your flats should have some effect. Perhaps, not optimal. Can you post a screenshot of a debayered flat and raw light, as follows; load raw file - convert to grayscale - debayer - automatic background extraction - STF. Let's see what you've got.

ap.

[gegaus]

Hello,

I´ve done what you told me and the flat seems to show the correct field curvature.
The light frame looks exactly like this RAW file after calibration so there is something I´m doing wrong for sure???
ImageCalibration  seems to have no effect on the lights what so ever.
I follow the tutorial step by step with the same settings and all but the result is like nothing..

PS. I subtracted the flat from the light in Photoshop and the light image became perfectly flat corrected 

Screendump:

[astropixel]

Hi Geir.

No issues with your frames - they all look normal.

Start here.

Create master bias, master dark and master flat, then calibrate light frames.

The master bias is processed differently. Convert raw bias frames to .fit and integrate.

If you use Format Hints, that is type 'raw' without the apostrophes in Output hints in the image calibration window, that will force proper raw conversion irrespective of any other settings.

I suspect that failure of the master flat to calibrate the lights may be related to generation of the master flat - can't be sure of that. Basically, load your raw flats into image calibration and calibrate with the master bias only, then integrate. No need to deBayer bias, flats and darks.

If you are still having problems, post your work flow.

ap.

[gegaus]

Hello and thank you,

It´s going in the right direction but I still scratch my head because of the results.
It still looks like the darks are filled with noise when i calibrate with the bias?
I tried to calibrate the lights directly with bias, darks(not calibrated) and flats(not calibrated) and the the result was better.
I also tried to calibrate with only flats and that to seem to generate noise into the image.
When stacking in DSS there seems to be less noise in the final result but more cromatic errors.

Some of the results is shown in the dump below(I´ve used Background Extraction on the last two images and STF on all):

[astropixel]

Hi Geir. How many bias, dark and flat sub frames are you using?

Increasing the bias Pedestal may work when bias calibrating your darks. The default is 0. Set a literal value. Try 5, 10, 15, until you get a good result. But I will be guided by others on this point.

ap.

[gegaus]

Hello,

The Master bias contains 50subs, the darks and flats are only five. The lights are six 3x600sec + 3x300sec.
The Lights are very light polluted so they are maybe a bad test example for the final image.
But this has nothing to do with the noise in the darks when bias calibrated.
I will try to set the literal value to see if that helps.

I have made a few dumps of the files with the noise (bayered and debayered) I dont know how these files are supposed to look like so I´m pleased if anyone can comment them :-)

Bayered:


DeBayered:

[astropixel]

Your frames look normal.

5 dark frames is not enough in my view. I would use 10 minimum. This is related to the way PI processes / scales data. I suspect a cause of your woes, particularly with 50 bias frames. Perhaps -ve numbers in your darks after calibration. If you can, take another 5 darks  or more at the same temperature as the existing set, and or increase the bias pedestal.

Generally, superior results are obtained with more frames. For this set maybe stick with DSS. I find however, that PI does a better and more predictable job than DSS with an appropriate data set.

[gegaus]

Hello,

I agree that this image contains to less data, but it was mostly a test because I´ve just started to use autoguiding.
Thank you all for your advise, I´ve got to the conclusion that converting all my files to Fits format prior to calibration gives the best end result 

In the future I will shoot a lot more subs to get as much data as possible into the image.

Regards
Geir