Author Topic: Colors not correct after Flatdivision  (Read 12481 times)

Offline sctall

  • PixInsight Enthusiast
  • **
  • Posts: 87
    • View Profile
Re: Colors not correct after Flatdivision
« Reply #30 on: 2013 December 05 11:58:21 »
OK Thanks
That did it
So now I would like to know, reading all the posts, just what is a good reading for a flat, and how to read it.
One says 4000ADU ( how is that determined), or 4000 on the statistics 14bit scale.
I avg between 500-1500 on the statistics 14bit scale. That sounds low.
But I usually get flats at ~ 1sec, not much longer. Seems to get to 4000, I would be exposing too long.
Ever since I got the IDAS LP2 filter, I have been trying to dial in the flats. It seriously cut back on my red channel. So I have added in a pinkish shirt to get the channel back up.
I have attached an image of my latest values. If I read the previous posts, this is very low.
But I could have sworn that the camera histogram was at least 33%. Is it safe to expose for longer in LP skies?
What would be real helpful is if someone could let me download a good RAW flat to compare to, or as a goal.
I do have an image named flat_L_001, ( forget where I got it, but it was in the forum post), and it has a value of 7000+. Almost perfectly balanced though.
Man that seems high

Scott

Offline pfile

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 4472
    • View Profile
Re: Colors not correct after Flatdivision
« Reply #31 on: 2013 December 05 12:17:24 »
okay here's the thing - the back of the camera shows you what the histogram looks like after a stretch. CCD data is linear (human eye has a log response) so all terrestrial image processing packages debayer the image, then stretch it, then boost the saturation. this includes the camera firmware.

the 33% histogram rule is a rule of thumb for DSLRs which is intended to help you figure out when your sub exposure has become sky-noise limited rather than CCD read-noise limited. in other words, the act of reading out the CCD destroys the fainter data recorded, so you want to make sure you've pushed all the signal well past the point where the DSO signal would be destroyed.

but that does not apply to flats. with flat sub exposures you want the brightest area of the image (usually the center) to be as bright as possible without departing what's known as the 'linear response' area of the CCD. for most cameras this is near 50% well depth, but it does differ from chip to chip.

so you want to expose your flats so that the histogram is near the center of the display, but when the data is represented in linear mode. this is what PI is showing you and that's why it looks (and is) really underexposed. in addition because the 14-bit file is opened as a 16-bit file, a completely overexposed image would show all the data at x=0.25. so you're really looking for a histogram that's peaked somewhere in the 0.125 area, but probably a little less since half the pixels are brighter than the peak of course.

the pink shirt is a good idea to get the red channel strength comparable to the others.

i guess another shortcut here is with a DSLR you can tell it to meter +2EV over the Av-computed exposure. that usually works fine. but you can also open your linear RGB images in PI (as you did) and check what's going on and adjust your exposure accordingly.

note that it's not necessarily fatal to have underexposed flat subs since you normally stack them to boost the SNR, but it's better to start with 'clean' data of course.

 rob

Offline NGC7789

  • PixInsight Old Hand
  • ****
  • Posts: 391
    • View Profile
Re: Colors not correct after Flatdivision
« Reply #32 on: 2013 December 05 12:20:33 »
Ever since I got the IDAS LP2 filter, I have been trying to dial in the flats. It seriously cut back on my red channel. So I have added in a pinkish shirt to get the channel back up.

If I am not mistaken a pink shirt will reduce the red. In fact any filter can only subtract. The only way to add is more exposure or a brighter target. I would recommend either making or buying an EL panel for taking your flats. I made one for under $20 and it is bright enough that my flat exposures are about 1/30 sec.

Offline sctall

  • PixInsight Enthusiast
  • **
  • Posts: 87
    • View Profile
Re: Colors not correct after Flatdivision
« Reply #33 on: 2013 December 05 14:43:41 »
Thanks great info
Thanks for the verbosity Rob.
I am still a bit of a lamen when it comes to all this image detail.
But I learn something new every day.
I usually take skyflats, but this time of year on weekdays I don't get home until dark, so I cant.
One issue I have with the skyflats is the blue channel is higher than the red or green.
I have been contemplating an el panel.
But I see several people complaining about the brightness.
I don't have the funds for a gerdman.
It would make thing more consistent, and I would like to be able to take flats whenever.
Is there a preference on the EL-panel color?
I see aqua, white, and pink mostly

Scott

Offline pfile

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 4472
    • View Profile
Re: Colors not correct after Flatdivision
« Reply #34 on: 2013 December 05 16:13:20 »
Ever since I got the IDAS LP2 filter, I have been trying to dial in the flats. It seriously cut back on my red channel. So I have added in a pinkish shirt to get the channel back up.

If I am not mistaken a pink shirt will reduce the red. In fact any filter can only subtract. The only way to add is more exposure or a brighter target. I would recommend either making or buying an EL panel for taking your flats. I made one for under $20 and it is bright enough that my flat exposures are about 1/30 sec.

if you take a picture of a pink wall with an OSC, the red channel will be 'hotter' than the green and blue… try it.

the point is that with a light pollution filter, especially the astronomik CLS, your images come out with a strong blue cast. which means the blue histogram hump is already to the right. and then when you increase the exposure in an attempt to fix the red channel, the blue channel becomes overexposed.

hence the use of a colored light source, basically to undo the cast that the filter has introduced. not to fix the color mind you, but simply to make sure all 3 channels can be increased together with one another rather than one channel becoming overexposed.

rob

Offline pfile

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 4472
    • View Profile
Re: Colors not correct after Flatdivision
« Reply #35 on: 2013 December 05 16:26:39 »
Thanks great info
Thanks for the verbosity Rob.
I am still a bit of a lamen when it comes to all this image detail.
But I learn something new every day.
I usually take skyflats, but this time of year on weekdays I don't get home until dark, so I cant.
One issue I have with the skyflats is the blue channel is higher than the red or green.
I have been contemplating an el panel.
But I see several people complaining about the brightness.
I don't have the funds for a gerdman.
It would make thing more consistent, and I would like to be able to take flats whenever.
Is there a preference on the EL-panel color?
I see aqua, white, and pink mostly

Scott

oh - one thing i forgot to mention above is that if you look at the back of camera histogram for a properly exposed flat, it's going to look overexposed… but don't panic, that's fine.

if an EL panel is too bright you can always put sheets of binder paper between the telescope and the panel to try to dim it out.

i don't understand why, but i get better results from sky flats when using my CCD camera. the EL flats look like the light source is shining in from an angle. it is a cheap one i bought on ebay though, so maybe it's just bad.

rob

Offline NGC7789

  • PixInsight Old Hand
  • ****
  • Posts: 391
    • View Profile
Re: Colors not correct after Flatdivision
« Reply #36 on: 2013 December 05 18:43:41 »
the EL flats look like the light source is shining in from an angle.

Are you aware of measuring the flatness of your EL flat? Take two sets of flats, one with the EL rotated 90 degrees. Create a master flat out of one set and then use it to calibrate the other set. An ideal flat light source should produce no variation across the image on the "flattened flat" . I think in practice a 2% (or is it .2%) variation is flat enough. I think when I measured the one I made (which also came from ebay) it was 3% (or was it .3%) which is good enough for me  :)

Offline pfile

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 4472
    • View Profile
Re: Colors not correct after Flatdivision
« Reply #37 on: 2013 December 05 20:23:21 »
i have not tried that but the difference between a sky flat and an EL flat is actually obvious to the eye, and dividing one by the other really does show that light seems to be coming in at an angle. of course it's impossible to know which is which but my lights are not flattened properly by the EL flats, so i'm assuming the EL panel is bad.

i'll try the EL rotation and see what i get.

rob

Offline sctall

  • PixInsight Enthusiast
  • **
  • Posts: 87
    • View Profile
Re: Colors not correct after Flatdivision
« Reply #38 on: 2013 December 06 09:37:59 »
I think sky flats are the best possible way to make flats. They are very even.
It is just that if you have to tear down after imaging for whatever reason, you must take them prior to the lights.
This limits the times you can perform this.
I see the panels are not very expensive, so I will try one to see.
As far as my current flats:  sky flats
I would expect my higher blue channel to be coming from the blue sky. Is that correct?
Maybe it is a byproduct of the IDAS LPS filter as well.

I attached an image of what the histogram looks like with canon_DPP. You can see the channels are not balanced and the blue is high. The pink shirt brought out the red a bit more , but the G,B remained the same.

But when debayering in PI, the red is lower( but not terribly)  but the G,B are fairly balanced.

Does this look like a good flat, or does it need more?

scott

Offline NGC7789

  • PixInsight Old Hand
  • ****
  • Posts: 391
    • View Profile
Re: Colors not correct after Flatdivision
« Reply #39 on: 2013 December 06 09:56:09 »
really does show that light seems to be coming in at an angle

Not to belabor the point, but it could be that the diffusion in front of the panel is not sufficient (I use several sheets of diffusion paper about 1/4" in front of the panel) or that the panel holder is not square to the scope tube. I've found EL panels more convenient as I can take my flats whenever I want, but one should use whatever technique works.

Offline pfile

  • PTeam Member
  • PixInsight Jedi Grand Master
  • ********
  • Posts: 4472
    • View Profile
Re: Colors not correct after Flatdivision
« Reply #40 on: 2013 December 06 11:14:26 »
no, it's not belaboring the point. using the EL panel would be a heck of a lot more more convenient for me.

it could be simply that i'm putting the EL panel right on top of the scope rather than at some distance... but it's not really big enough to put in on a wall. at any rate it should be square with the tube like that. i point the OTA straight up and put a book bigger than the aperture over the panel.

regardless i think it should work right up against the OTA as that's how the flip-flat works. unless there's some special sauce there that i don't know about.

for the lum flats i recently took i had to put in about 6 or 7 pieces of printer paper to get the shutter times up (with this camera, shutter speeds <= 1 sec definitely cause artifacts...) maybe the printer paper does not diffuse evenly.

rob

Offline NGC7789

  • PixInsight Old Hand
  • ****
  • Posts: 391
    • View Profile
Re: Colors not correct after Flatdivision
« Reply #41 on: 2013 December 06 11:30:06 »
I think one of the things you pay for with flip flat is the quality of their panel. When using a cheap panel you definitely need diffusion. I mounted my panel to some 1/4" foam core. In another piece of foam core I cut a hole the size of my dew shield. Onto this I attached several layers diffusion paper. I'll look for the link.

You can also find fancier stuff in the lighting department of camera stores.

Then I attached another piece of foam core also with a hole the size of the dew shield.

For me this was pretty simple as my scope is only 71mm aperture  :)
« Last Edit: 2013 December 06 11:35:23 by NGC7789 »

Offline NGC7789

  • PixInsight Old Hand
  • ****
  • Posts: 391
    • View Profile
Re: Colors not correct after Flatdivision
« Reply #42 on: 2013 December 06 12:03:38 »
Turns out it was just vellum aka tracing paper.

Offline Phil Leigh

  • PixInsight Addict
  • ***
  • Posts: 220
    • View Profile
Re: Colors not correct after Flatdivision
« Reply #43 on: 2013 December 08 04:33:46 »
Unfortunately not all EL panels are equal. Some of them have spectrum  gaps which play havoc with narrowband and interference filters. Some of them do not illuminate evenly even though to the eye they look fine... some of them have a strong colour imbalance although to the eye they appear "white"...

I've found the Gerd Neumann panels to be the best (albeit at a price!)

Offline StefanG

  • Newcomer
  • Posts: 19
    • View Profile
Re: Colors not correct after Flatdivision
« Reply #44 on: 2013 December 08 06:33:46 »
The Gerd Neumann panels are supposed to be even suitable for spectroscopic applications.
GSO RC10
QSI583wsg, SX Lodestar Guider
AP Mach 1 Mount, Eagle Pier
PixInsight, MaximDL, CCD Commander