Flat field problem

[pfile]

the correct way to measure the ADU in a flat is to leave it linear. stretching the flat makes no sense... how much do you stretch it? you can stretch it until the values are 65535 but that does not change the underlying fact that the flat is very, very underexposed. the imagecalibration tool is only going to normalize the flat; it's not going to stretch it, so stretching it tells you nothing.

rob

[Stu]

You measure the ADU linearly.  I measure it directly off the camera when I am taking my flats using SGP and the statistics module. 

[gianpri]

Surely I'm wrong. But I do not understand why the picture looks properly exposed in BYEOS while it seems grossly underexposed in PI:

[Stu]

Oh!  Because ByeEOS autostretches and presents a .jpg image.  The histogram is of that image.  Best to ignore that histogram completely when doing flats. 

[magnusl]

Hi!

Ok, so it seems that there is something with the statistics values here. Two suggestions, the HT and the debayering.

First, the HT: I'm not sure what you mean here. Sure, I can do a HT.  But how more precisely do you mean? I move the midtone setting, and it gets brighter and the statistics values increase - but what do I gain? Part of the problem here, is that these images are so saturated so there is no vignetting left in them - they produce an even light grey or white when I raise the values, but no vignetting....so what do I gain?

As for debayering: yes, I tried that. Using Superpixel, all I get i the same values in three channels. But no increase in values to the levels I would need. And again, these are actually saturated images. To make the vignetting visible, I need far LESS exposure, producing even lower values....

Magnus

[magnusl]

Hi!

Here is a link to a dropbox folder with a bunch of files, taken at 0.1 secs, 1 sec, and far longer, and both in fits and in cr2 format. All of them display the same type of behavior in my PI. Please play with them as much as you want! From just looking at them, the shortest exposures in both fits and cr2 seems "best" in terms of actually representing the vignetting in the camera.

https://www.dropbox.com/sh/fs744ujp8oed79d/AACfnztnt-E98N8km8Tv2dUBa?dl=0

Unfortunately the file names do not fully display exposure times, but it is in the files, so the program should show.

Magnus

[Stu]

Oh, can you post a bias frame also?

[gianpri]

I am getting more confused. Ultimately my flat field is underexposed as it appears in the PI or is exposed correctly as it appears with BYEOS? To evaluate the flat I have to use statisc of PI and not the histogram of BYEOS?

[pfile]

the results from a program like BYEOS can be confusing. they use a software library that auto-stretches the image before it computes the histogram. this is because when a DSLR creates a jpeg image (or the image that you see on the back-of-camera display), it also auto-stretches the data. there is a rule of thumb about lights taken with a DSLR where you try to get the back-of-camera histogram hump detached from the left side of the display. BYEOS emulates this so that there is no confusion about how well a light is exposed. but it also does the same stretch when showing you a flat. note that it does not actually stretch the image, it just saves the raw CR2, which is un-stretched.

no matter how the flat looks in BYEOS, you should expose the flats so that the histogram peak is about 1/2 of a completely overexposed flat. because the CR2 is 14 bits, PI will show a completely overexposed flat as 0.25. so you should aim for 0.125 on PI's histogram display.

also watch out that all 3 channels are well-exposed. when using a CLS filter you end up with a very strong blue cast and so the red channel can be underexposed while the green and blue are properly exposed. i used to solve this problem with a slightly pink-colored t-shirt stretched over the telescope's aperture.

rob

[Stu]

And just to give a different perspective on Rob's post, if you don't have a built in ADU measurement (ByeEOS does not), then what I would do is expose a flat, open that image in PI, open the statistics process, and then use the drop down menu to pick 14 bit.  Then go for an ADU exposure of about 7500 or so (somewhere in the mid range).  Then go back to your program and take your flats at that exposure.   I don't use histograms--I prefer numbers. 

Rob also makes a good point about the color channels channels.  You want to make sure that your red, green, and blue channels are all saturated about 50% or so.  They don't have to be identical, but somewhere in the middle of your max ADU. 

[gianpri]

Thanks Rob. So, as I understand it, a flat properly complaint should be so

[gianpri]

But even so, if imposed 14-bit, the average is about 2000 and not 7500. Why? I am more and more confused.

[Stu]

Probably because it just needs more exposure time. 

[pfile]

But even so, if imposed 14-bit, the average is about 2000 and not 7500. Why? I am more and more confused.

can you see if there is a difference if you debayer this flat? it has been a long time since i worked with an OSC.

rob

[gianpri]

Here are the flat debayer statistics. I do not think that changes anything, apart from the red channel underexposed. Is it possible that in an almost saturated BYEOS file is underexposed in PI? So even when I take the light frames can not trust the histogram of BYEOS? Now I give up to understand.