noisy flats

[Daniel-K]

is it possible to get noisy flats? i have 3 hours worth of data and with matching darks and bias and flats. i dont think my flats have worked very well so im currently taking some new ones  to see if i can get a better result but with 3 hrs worth of data i was hoping the nise would be alot less than this. below is a stretched image before processing, as you can see theres a wicked gradient. and the the other image is my final result which although is ok im not happy with it, i think there is still more blue stuff to pull out. 

[ollypenrice]

Sure you can have noisy flats. How to avoid it?

1) Take a lot. The palaver is in setting them up so take 30 to 40 if you suspect noise.

2) Calibrate them by subtracting a master bias used as a dark. This is very important.

3) Give the masterflat a bit of Gaussian blur. Pro observatory flats contain pixel by pixel information but for the amateurs I think you can forget that.

But why do you thnk the flats are to blame? 3 Hrs is not a lot of data if you have the best amateur M45s in mind.

Olly

[IanL]

Assuming this is taken with the 450D?  First question is how did you make the flats? If you're using an artificial light source, the culprit may be just one of the channels.  Since you are shooting OSC, you only have one exposure length to try to evenly illuminate all three channels, whereas with mono and filters you can adjust the exposure length for each filter.  If your light source is under-illuminating in one (or two) channels, you may end up with very noisy flats.

This is especially the case if you are using AV mode to determine the correct exposure, but even if you set the exposure lengths manually by trial and error, you may not be able to get the least illuminated channel well exposed without over-exposing the others.

I've had plenty of trouble with noisy flats using an LED light box that I built (and rebuilt twice), especially in the red channel.  I was finding that the green channel would reach the correct exposure (histogram peak somewhere between 1/3rd and 2/3rds of the way to the right of the graph), the blue would be a bit behind it but still okay and the red would be underexposed.  Looking at a single flat (stretched) visually I could see the green and blue channels looked okay, but the red was basically just noise (hardly any sign of the vignette even) and the master flat was much the same.

I tried adjusting the intensity of the R, G and B LEDs by rebuilding the box with better matched LEDs, adding mini-pots to adjust the output of each channel and also tried 'white' LEDs.  It helped a bit but the red channel still wasn't great, and there was a significant colour cast in the image at the end of the stacking process.  It is possible to fix that of course, but it certainly doesn't make life any easier.

Depending on your light source, you may have a similar issue.  Just because a light source looks white doesn't mean it is full spectrum. Single colour LEDs only emit light at one wavelength, and 'white' LEDs only do so at a few wavelengths.  Flat panels and laptop/iPad screens can also have significant gaps in their output. The human eye/brain doe a remarkably good job of re-calibrating the scene to make it look 'normal' so you can't really tell just by looking.

I have only recently managed to get decent flats by taking them indoors using a white(ish) curtain illuminated from behind by the sun, plus the acrylic diffuser from my light box on the scope (with a bit of shielding around it to prevent stray light from the sides creating a gradient).  These worked pretty well for me with decent exposure in all three channels.  You could do much the same with a diffuser and a cloudy sky (to avoid gradients from direct sunlight).

I suspect it is more beneficial to have pixel by pixel flats with a OSC camera.  That said, you might try extracting a single channel (least noisy) from the master flat and then using that as the mater flat.  You'd need to debayer (using the superpixel method is probably best), extract the best channel and then resample the image back to its original size (since superpixel debayering will reduce the flat by half on each axis).  It still might not work perfectly if you are using due to chromatic aberration. I certainly notice the vignette in a stretched flat is slightly different in all three channels, but using a reflector you may get better results.