Do I need darks and flats for planetary work?

[toml42]

and if not, will they hurt?

I have noticed in some of my planetary images that one half of the ccd in my spc900nc always seems a little brighter, and it annoys me somewhat.

As i understand it a few dark frames should sort that out, but apart from that, can i expect any improvement in image quality?

[riklaunim]

You don't need darks for standard planetary work. If the camera would get very hot you could see brighter top left corner.

[toml42]

thanks i'll bear that in mind!

more to the point now, i actually stacked up a long dark frame to find the banding issue that had been irritating me:

does anyone know if this looks normal for an spc880?

[riklaunim]

That looks bad... or it's not debayered image. What software do you use?

[toml42]

its a raw, dark, 1 minute video stacked into a single frame in Lynkeos (I'm on a mac) but i've seen that band crop up in images before using registax, too. since it's raw data straight from the webcam, i'm assuming it's not debayered.

I'm also pretty sure i've noticed it in some other spc900nc images.

I wondered if it was that annoying little led light on the camera, but it seems far too neat...

oh, and i did tweak the contrast levels a little to make it more detail, but other than that nothing.

[toml42]

I've done some more testing, that pattern of banding is very consistent when the gain value is set to default, i've tested it multiple times for different exposure lengths and shutter speeds (oh what the cloudy nights do to me...)

for all other gain values, the bands look like this, regardless of exposure length or shutter speed, again tested multiple times.

it's always blue, too, i was imaging in monochrome the first time around. I had also noticed that my images often came out with unusually blue backgrounds, so next time the sky is clear i'm going to try using this as a darkframe and see if it helps!

[toml42]

update: The banded pattern and blue colour in the images disappeared when i reduced the Gamma setting to nearly zero, anyone else having this problem, give it a try!

[teh_orph]

I get the same nasty pattern with my camera too..it's really annoying! I'll try changing the gamma. But surely that'll make the image much darker too, forcing longer exposures (or more gain etc) and bringing back that dark pattern?

[riklaunim]

Maybe it's some electronic interference, not so good USB port in the computer. The pattern is very odd.

[jgs001]

I've only seen that sort of pattern appear on my SPC900 images when using the RAW mod. You have to set the RAW options in registax to debayer. I don't know how you would do that in other software.

[toml42]

I didn't know how to set to debayer in lynkeos, so i googled "lynkeos debayer". this thread is the top, and only relevant result. 'sigh'

[jgs001]

Did you apply the raw mod to the spc900 ? If so, you could try setting it back to normal again.

[riklaunim]

use another software then FireCapture 1.2 beta has some debayer, and Registax should also be able to do something about it.

[narrowbandpaul]

do you need flats?

well quite surprisingly perhaps...yes! Its not obvious as to why until you delve in to the underlying theory, but flats are essential.

Once you stack several hundred frames together the random noise will be significantly reduced, but what will remain is the spatially invariant noise....which is removed using flats. In fact you need lots of flats to do this well.

But yes, use flats.......

[toml42]

I'm working on a mac whilst at university, lynkeos is the only free stacking program i could find that was compatible. it does the job ok, but using Registax whilst at home is certainly more pleasurable...

do you need flats?

well quite surprisingly perhaps...yes! Its not obvious as to why until you delve in to the underlying theory, but flats are essential.

Once you stack several hundred frames together the random noise will be significantly reduced, but what will remain is the spatially invariant noise....which is removed using flats. In fact you need lots of flats to do this well.

But yes, use flats.......

see that's what i would have thought, but i never hear people talking about them in planetary imaging, i'll certainly give them a go.

would you say there is little point to dark frames though, considering the number of frames being stacked?

[riklaunim]

Note that Paul likes narrowband DS long exposure imaging I didn't saw that Anthony Wesley, Christhoper Go or other planetary guru would take flats and darks for planetary imaging. For methane-band yes, darks are usually there as the exposures are in seconds and some pixels can glow etc. But for RGB? no, don't think so.

[teh_orph]

I remember seeing a pic of Uranus on here recently and due to the zoom factor (barlows and such) the exposure was at max. I'd imagine that maybe darks could be useful then?

[riklaunim]

If there is significant noise then you can add darks. I used them for methane band images where 0.5 - 1 sec exposures have some hot pixels and stronger noise in general. For Ir/RGB where exposures are way below 0,5 sec noise is much lower.

[narrowbandpaul]

hi tom.

Flats are essential. As I said before after you combine several hundred or whatever frames, the random noise components are virtually zero...that is read noise and photon shot noise are minimal. The only thing left to degrade signal to noise is fixed pattern noise, which is removed using good quality flats....each flat should be around 3/4 full well (or 'maximum signal').

The extent to which FPN degrades an image depends on one factor, PRNU (photoresponse non-uniformity). For typical CCD sensors (I know webcams are CMOS) is can be around 0.5%...ie FPN=0.005*Signal. Although low, this will easily dominate with bright objects like planets and after several hundred have been combined.

As for darks....yes use those. Again its for the same reason. Dark images contain random noise (dark shot noise) and dark FPN! When you take a bunch of darks and average to subtract from an image, you DO NOT remove the shot noise component. You CANT! Its impossible to remove random noise via subtraction. What you actually remove is the Dark FPN. Typically the Dark Signal Non-uniformity is much higher than PRNU...upwards to 20% upto about 100%..ie DFPN=0.2*Dark Signal. This dark signal is the thing that is temperature and exposure length dependent. If you are imaging at a short exp time maybe its not worth it, as the noise will only be around a few electrons.

Now let me respond to riklaunim.

Firstly. Do not think for a minute that because I prefer narrowband imaging that I dont know anything about planetary imaging. Secondly this post regards image sensor technology. Which I am doing for a PHD! So I unlike you know what Im talking about.

Why dont the Planetary gurus use flats. Well I guess it could be that they dont understand the theory? Yes! They may take good planetary images but as to they theory...they simply dont know it from an image sensor point of view. Just because someone takes a good picture doesnt mean that they are worthy of worship.

Its important to think why you are doing something rather than just copy what they do, on the assumption that they actually know what they are doing and why.

anyway, Riklaunim, whether or not you agree, I do know what Im talking about when it comes to image sensors. Im not convinced that you do.

paul

[riklaunim]

But if it works - then why it's bad? If flats can be neglected without what is seems any noticable changes and ability to produce very good images then whats the point of doing flats - only for theory?

And note that the planet on the sensor is drifting slowly, it's not DS where guided image is still. It's a big dihtering.

[narrowbandpaul]

it works clearly, but it could be better. Theory says to do it, so why not...its simple with a webcam at fast framerates.

if you are tracking equatorially, then the planet is not going to move, not over a short period of time anyway. In theory you are limited by the rotation rate of the planet, as separate RGB images will not line up if taken too far apart. Jupiter is about 10mins or so I believe. Not going to cause significant movement.

even then, that dithering is not going to completely remove FPN.....where as good flats will.

[riklaunim]

Moving on the field of view of the sensor, not in the terms of planetary rotation - it will move due to wind, bounce from bad seeing and most common - non perfect mount tracking (4000 - 6000 mm of FL). I sometimes have to pause the AVI to fix the position.

Well - to make a flat I would have to have glowing panel and do flats after imaging in given setup (which likes to change). Maybe I'll try some time

[toml42]

This has certainly been an interesting discussion! When my HEQ5 arrives this week, weather gods permitting, i will put theory to the test and see if there is a noticeable improvement. I may even put comparison pics up on here if they're interesting enough!

Thanks Paul and Riklaunim for your detailed answers

[Tim]

I see you've lost none of your people skills Paul :s

Have you got any side by side comparison shots to illustrate the removal of FPN by any chance? To be honest I am not convinced it would make a massive difference to image quality. Would it even be spotted by the average enjoyer of astropics???

[MartinB]

hi tom.

each flat should be around 3/4 full well (or 'maximum signal').

Does this apply to chips with a non linear response?