Creating Master Flats

[sreilly]

I had stopped making my calibration frames in PI some time back and was calibrating my images first in MaxImDl. I used the tutorials for creating my master frames posted on the PI website but didn't seem to have good luck. Now I'm back trying again and I have a few questions. I created my master bias and master dark frames as per instructions here http://www.pixinsight.com/tutorials/master-frames/index.html . Now I'm trying again to to make my master flats. I've calibrated the flats with the same temperature dark master and used the master bias, same temperature here as I understand with the STL-11002M camera this makes a difference, and have several messages in the Process Console I don't understand.

They are:

1) ImageCalibration: Global context

Loading master calibration frames:
Loading calibration frame image:
G:/Images/STL-Bias/-30-Bias/PI-MasterBias-30Bias-1x1.fit
Reading FITS: 32-bit floating point, 1 channel(s), 4008x2672 pixels: done
Loading calibration frame image:
G:/Images/STL-Darks/-30-900/Dark Master bin1x1 -30C 900 seconds.FIT
Reading FITS: 32-bit floating point, 1 channel(s), 4008x2672 pixels: done
Rescaling sample values: [890.689636230469,65535] -> [0,1]: done
* Subtracting pedestal keyword 'PEDESTAL': 100.0000 DN

The question here is why is the pedestal  keyword being subtracted. Actually I guess the real question is what is this. If I've loaded a master bias frame isn't it being used?

2) Writing output file: G:/Images/20140217/AutoFlat/pi/AutoFlat-PA075-Ha-Bin1-014-East_c.fit
Dark scaling factors:
k0 = 0.003
** Warning: No correlation between the master dark and target frames (channel 0).
Gaussian noise estimates:
s0 = 3.770e-003, n0 = 0.963 (MRS)
Writing FITS image: 32-bit floating point, 1 channel(s), 4008x2672 pixels: done

Out of 72 flat frames, 6 give the above message. All the flats were taking using a flat panel with ACP running an automatic flats script. The background ADUs are all about 34,000 and the same bias and dark masters are used on all 72 frames. 36 are east side and 36 are west side (rotator).

Any suggestions would be great.

-

[sreilly]

I also just noticed that the sequence seems to calibrate and then save randomly. The flats are calibrated in order but may not be written until later. I've attached the long Process Console readout from the procedure to show the entire run for the flats. It's a bit long but shows what I'm referring to. I should mention that this is using a Windows 7 Pro 64 bit system with 16GBs memory, Intel i7-2600K processor running 3.4GHz, with a SSD and plenty of hard drive space. PI Version 01.08.01.1087 Ripley (x64)

[jerryyyyy]

If you are using Maxim, there is a no-brainer way to make good flats.    Find the Maxim plugin call Sky Flats Assistant.  Read a bit and run this at sunset pointing east for me.  You get 10 flats for each filter at the desired ADU (25000) then you can follow the manual instructions in PI to create a master or just use the Preprocessing script and load in the raw files... plus darks and biases.  Saved my life. 

[sreilly]

Thanks but taking them isn't the problem. Using ACP AutoFlats.vbs script and the FlatMan XL panel make the taking part easy. Making the master in PI is where I have questions. I take 36 flats for each side of mount. I also run these when it's cloudy, raining, or just a bad night altogether. That's time spent making calibration frames and I use an updated library that usually has no more than a few months old frames. Make and keep dated masters for reprocessing images at a later time as I get more proficient with PI.

Steve

[pfile]

if the flats are short, forget about the master dark and just calibrate the flat subs with the bias frame. that works great for me.

if the pedestal subtraction is a problem i think you can change what keyword it's looking for so that it does find "PEDESTAL" and thus won't subtract it.

rob

[sreilly]

Thanks Rob, I'll give that a try.

The Ha flats are 8.37 seconds @-30. The SII and OIII usually go 18.02s and 3.98s respectively.

Thanks Again,

Steve

[sreilly]

if the pedestal subtraction is a problem i think you can change what keyword it's looking for so that it does find "PEDESTAL" and thus won't subtract it.

rob

Are you referring to editing the FITS Header for the master Bias? I've attached two FITS Header screen shots. The BP image is a master bias created using Batch Processing and The PI is made using the instructions for making master Bias and Darks by the aforementioned tutorial.

[pfile]

well, no what i meant is that if you use ImageCalibration to calibrate your images i think you can tell it a custom keyword for the pedestal value. i need to double-check this though.

rob

[jerryyyyy]

Can I ask a flats related question.  I am using my Nikon D800 and it is tricky using Sky Flats Assistant to make flats.  Isn't there a way you use an uncorrected image, "defocus" or use atrous wavelets to wipe out the detail, and then use that image as a master flat?

[jkmorse]

Jerry,

While it may help with vignetting, wouldn't the defocus destroy any ability of the Flats to correct for dust donuts?  For the C8, which I think is a f8 system, dounts seem like they would be a problem you would need to address.  The best thing I have found is just to use a flat panel.  Altinak makes a great set of different sized panels and they work perfectly.

For what it's worth.

Jim

[jkmorse]

Steve,

I have had issues with calibration/integration in PI with MaximDL raw images for some time (in fact I just abandoned MaximDL as my capture software because of it and went to Sequence Generator Pro based on recommendations here in the forum). 

But one thing I found when working with MaximDL raw images and PI calibration/Integration is that the biggest problems involved weird things happening with Bias frames.  What I got to work was to simply shoot my darks at the same time iteration as my lights and therefore you don't need the bias frame at all, just darks and flats.  That worked for me and may help in your case as well.

For what it's worth.

Jim

[sreilly]

Thanks for the reply Jim.

With ACP I have too much invested to abandon MaxIm not to mention that it is highly supported. Automated imaging sessions make gathering the needed data a bit more easy these days. I have almost always used matched darks as I have built a dark library that I usually rebuild each season. The season usually last 4-5 months and I only use 10, 15, and 30 minute exposures. If I'm working on a high dynamic image and use shorter exposures, let's say M24 or M45, I'll take as needed on nonimaging days.

The most use my bias usually gets is for calibrating the flats but as Rob suggested I have been experimenting with creating them not using a scaled dark and that seems to be working. Not being a heavy math fan, I can do the usual math but those expressions that take a line or two are a bit beyond my grasp these days, I need to find an easier way of qualifying my calibration images. By that I mean there must be a relatively easy way to determine the quality of any given image. I assume, yes I know, that PI does this somewhat already by assigning a weight to each image when creating master frames. So I need a way to order the calibration frames in order of weight. This may not come across as I hope. As an example, I have been taking darks frames since mid-night last night and now the morning is brighter. The inside of my roll off roof observatory will be getting lighter than it was at 5am. Does this have any affect on my later darks? Can I continue taking them still?

What I'm looking at now is a way to determine the difference, if any, between the dark taken at a totally dark time versus one at a lighter time. As far as I can see visually, there aren't any light leaks with my setup but then that's as far as I can see.

[pfile]

you can use the statistics process to see if the mean/median values of the nighttime darks and daytime darks differ significantly. if so you should not use the daytime darks; any light leaking in is really going to screw them up.

rob

[sreilly]

Excellent advice Rob, thank you. I just checked darks from different nights and daytime as well and they are very close, within 4-5 counts of each other. Happy to learn something new. I had never used that process before.

[jerryyyyy]

I have used Sky Flats Assistant with Maxim and I think it gives excellent results for flats... you can dial in the ADU and drop the original files into BatchProcessing to make a Master.  The only problem with the Master is you need to copy the FITS headers from an old Master. 

Unfortunately this script does not seem to work with the DSLR D800.  I have been working with my Nikon D800 because my SBIG STT 8300M is in the shop.  I am very happy with the combination of the fast Tak scope and ISO 6400 Nikon, but can't take flats with the Sky Flats Assistant the way I can with the SBIG filterwheel.  Pfiles is right I could not create flats via AtrousWavelets... the flats of the D800 are weird, sorta a rectangle required. 

Scratching my head on this one.