The theory behind Flats?

[Spoon]

evening,

2 things

1. what is the theory behind flat frames (aka how do they work)? 

and 2. what is the recommended length of a flat to be?

short and simple

Cam

[D4N]

1. Remove dust bunnies and vignetting by taking an image that should be plain the software can fix the bits that aren't.

2. Nothing set but the histogram should be roughly half way exposed without over exposing any colours. Mine are 2-4 ms depending on filters.

TSED70Q, iOptron Smart EQ pro, ASI-120MM, Finepix S5 pro.

[jambouk]

Take a flat frame and you'll see all the vignetting and dust in the optical train. I believe in part they help remove these artefacts from your final image.

I think it depends on what sort of camera / sensor you are using. With my dslr i get the histogram up and try and get the peak to be roughly at about 50% (half way along the x axis).

They do help massively with my images.

James

[Spoon]

1. Remove dust bunnies and vignetting by taking an image that should be plain the software can fix the bits that aren't.

2. Nothing set but the histogram should be roughly half way exposed without over exposing any colours. Mine are 2-4 ms depending on filters.

TSED70Q, iOptron Smart EQ pro, ASI-120MM, Finepix S5 pro.

 i know that part but i meant how does it actually work? sorry if i was unclear

[jambouk]

You are still being unclear

The software must make a master flat file and them by some fancy footwork subtract that from the stacked image... Is that was you mean?

James

[Spoon]

That's what I mean yeah

Sent from my iPhone using Tapatalk

[Earl]

The lenght of a flat is dependant on the intensity of the light source, the speed of the optics and what filter if any are been used.

Each filter will behave differently Lum letting the most light through and a Ha Filter the least, leading to a longer exposure.

I aim for a mean adu of 25000 some aim a bit higher its a bit of trial and error to find what works best for your setup.

For example I have to use a minimum of 2 seconds as I have a mechanical shutter on my CCD any shorter and the shutter can cause a smear. My Ha filter with my Flip flat on full needs about 30 seconds to get to 25000.

I use the Flat calibration wizard in Sequence generator pro to work out the needed sub lenght.

[steppenwolf]

A flat frame captures a facsimile of the light cone falling on the sensor. However, there is no detail in this light cone as it isn't focused on anything but it is the exact 'shape' of the light from the night sky that falls when a normal image is taken so it captures any vignetting. As this 'facsimile' light cone travels down through the telescope optics/tube/filters it is interrupted by unwanted items like dust - this dims the light cone in specific areas and, therefore, mimics what that dust is doing to the light from a real image.

The software 'division' of this facsimile data into the real sky data removes the unwanted artefacts caused by the dust and compensates for variations in the light intensity across the sensor caused by vignetting.

[Cath]

Yes Steppenwolf is right

A flat is not subtracted but used as a division coefficient, although when coding the algorithm it's often quicker to do a 1/ on the flat frame pixel data and then do a quicker multiply on the data using the flat frame data. Multplies are often quicker than division, although not always the case these days with the lovely FPU's etc you find on todays CPU chips.

[ollypenrice]

When teaching imaging I usually take a master flat, which looks like a bright inner circle with darker coners and a smattering of dust bunnies, and flatfield-calibrate it by applying a copy of itself to this master flat. The result comes out dead flat. The same correction is made to your images. It's a simple and effective demo. As a spin on this I also take a luminance flat, say, and apply it to a red flat. The result is exactly the same, proving that there is no need, on the rig in question, to shoot separate red flats. There is no way of knowing whether this will be the same for another rig but on the three that I use this turns out to be the case.

Olly

[D4N]

Flattened flat, nice

Mine don't seem to have any vignetting on them, wish I could say the same for the dust bunnies but unfortunately I have loads of those

Sent from my iPad using Tapatalk

[ollypenrice]

Flattened flat, nice

Mine don't seem to have any vignetting on them, wish I could say the same for the dust bunnies but unfortunately I have loads of those

Sent from my iPad using Tapatalk

The vignetting here (greatly exaggerated by the stretch) comes from using a huge full frame sensor with 2 inch mounted filters. Unmounted would have been better but the flats sort it anyway.

Olly

[Spoon]

When teaching imaging I usually take a master flat, which looks like a bright inner circle with darker coners and a smattering of dust bunnies, and flatfield-calibrate it by applying a copy of itself to this master flat. The result comes out dead flat. The same correction is made to your images. It's a simple and effective demo. As a spin on this I also take a luminance flat, say, and apply it to a red flat. The result is exactly the same, proving that there is no need, on the rig in question, to shoot separate red flats. There is no way of knowing whether this will be the same for another rig but on the three that I use this turns out to be the case.

Olly

I do flats and they don't look anything like this. And also they still don't remove vignetting?

Sent from my iPhone using Tapatalk

[steppenwolf]

Have to say that my flats look very similar to Olly's and they most certainly remove vignetting.

[Spoon]

Yea mine don't. How do you measure ADU?

Sent from my iPhone using Tapatalk

[ollypenrice]

You can measure the ADU in Artemis Capture at a glance by looking at the White Point value. Or you can scroll over the image in many programmes, like AstroArt. Nebulosity gives you a max (white point) value as well.

Most accurate flats do resemeble those posted by myself and Steve. I see far less vignetting in the TEC140 flats because the more elongated F7 light cone is less vignetted by the filters than the steeper F5 one from the Taks.

I often take flats which don't, for some reason, look like this and I scrap them and try again!

Flats need to be calibrated with darks-for-flats. This is vital. A master bias makes a perfectly good dark-for-flats, though. I wouldn't bother taking dedicated ones.

Another issue arises in shutter cameras where the wipe of the shutter will put a gradient into flats if the exposures are too short. 8300 chip users beware.

Edit; Oh, another flat bugbear arises in Newts if light gets into the bottom of the tube. Even with refractors I shoot flats in the dark, though many get twilight flats to work.

Olly

[russellhq]

Maybe a daft question, but after you divide the light frames by the flat, is the light frame re-scaled? Otherwise wouldn't you end up with a lot of quantization errors?

[Cath]

Maybe a daft question, but after you divide the light frames by the flat, is the light frame re-scaled? Otherwise wouldn't you end up with a lot of quantization errors?

Yes it has to be rescaled back up to cover the number of bits per pixel. You do that either before you do the divide (keeping calculations as integers) or afterwards (divide calculations would probably be done using floating point).

[dph1nm]

Maybe a daft question, but after you divide the light frames by the flat, is the light frame re-scaled?

You will probably find that most software normalises the flat before division (i.e. makes the average flux per pixel in the flat 1.0000 by dividing throughout by the sum of the flux in the whole flat). This all has to be done in floating point, not integers, of course.

NIgelM

[Dave_D]

i've yet to produce a single useable flat... even though  my images have the vignette central on the frame, whenever i try to do a flat i just get a gradient fading off to the bottom of the camera.

this isn't a recent attempt but they come out pretty much the same. been using a 12w 12" square ceiling panel downlight so i don't know if they're any good for flats, but also tried the white shirt over the tube and got the same. then again, my primary is off centre and my spider arms are a bit wonky and off 90 degrees so who knows...

[ollypenrice]

I get this sometimes, usually when doing flats other than in the dark. That's why I do them in the dark! Where is the histogram peak in this example? Try a third and also a quarter of the way over to the right. Sometimes half way works for me but sometimes it doesn't, depending on the rig. I can't account for any of this but pass on experiences.

Olly

[jambouk]

Dave_d, what scope are you using and how close are you putting the light source? Is it possible any light is by-passing the primary and secondary (if using a newt) and shining straight into the focus tube and onto the sensor?

James

[Dave_D]

it's a 10" orion optics GX250 'agricultural engineering edition' (plus, hacked to mars and back). the last time i tried was straight after imaging so the scope was in the (as close to) dark. the tube was vertical with the panel sat on the end of the tube on a spacer i made to stop any light leaks. the bottom of the tube was covered so i can't think where there would be any light leaks. i also tried the next day with the white shirt and got exactly the same.

i did a bit of research on the GX250 and the cell and spider are definitely not the standard GX250 parts but the scope was very second hand so i don't know it's history. the cell, although is 'seems' like a good fit, doesn't put the primary in the centre of the tube. on one side i can slide my fingers between the mirror and tube up to the knuckle, but the other side i can barely get my fingertips in the gap. gonna test this further as the first thing i did when i treated myself to a router was to make a circle cutting template and made a wooden 'cell' that's a push-fit into the tube out of 25mm ply with another 25mm ring that will be glued on top. i'm drilling this to make a 'test cell' that i can centralize the mirror in the tube with.

the three arms of the cell also are not true 120 degrees relative to eachother either... methinks that this is the scope that jack built... optics are fine, mechanicals are a bag of doo-doo.

@olly, the histogram on the last set i tried was at about 30-35%. i think i stall have them on my camera so if they are still there, i'll post one

[Alien 13]

If you are using the Canon camera you can get a shadow of the moving mirror if the shutter speed is too high which is quite likely if shooting a bright light in the daytime. It can be fixed by using mirror lockup if available or shooting in live view mode.

Alan

[Dave_D]

If you are using the Canon camera you can get a shadow of the moving mirror if the shutter speed is too high which is quite likely if shooting a bright light in the daytime. It can be fixed by using mirror lockup if available or shooting in live view mode.

Alan

I've modded the camera a few months ago and completely removed the flip mirror and viewfinder