Question about image streaks

[astropixel]

I think the streaks is a bias calibration problem?

EDIT: No! I am wrong.

[Nocturnal]

It's streaky because you have differential flexure which causes your images to shift from exposure to exposure. The points that cause the streaks are probably caused by unmatched dark frames.

http://deepskystacker.wikispaces.com/Measuring+differential+flexure

[astropixel]

Thanks Sander. I read the article, and it mentions guiding. The 106 frames that make up the image are unguided, the longest at 4 minutes - managed a good alignment for a change.

Would this also be related to an optical train slightly off centre? Or just unguided. The difference in the frames after integration was only a few pixels.

You are probably right about the unmatched dark. Not always having the time to take darks during the session I tend to take them the next morning or evening.

Temperature may be the cause of the unmatched dark - is that a linearity issue?

[Nocturnal]

No, I don't think it has anything with anything being off-center although I'm not sure what you mean by it. Your drift is caused by PAE (Polar Alignment Error) and PE (Periodic Error). 4 Minutes is probably pushing it without guiding.

I'm not sure what you mean with linearity issue. If your darks don't match your lights then your dark subtraction doesn't work properly, pretty simple You either leave dark signal in your frames or you're removing real signal. All that depends on your sensor and how it behaves under different temperatures.

[astropixel]

Rereading your first reply, and while I am not using auto guiding, I see your point about alignment. Though all the stars are very close to round - probably the best I have ever managed to achieve. But the issue is that the dark is not matched. I'm not going to ask now these two factors play out in that scenario to produce streaking, because proper calibration and registration should take care of all that. Too complex for now. It's a calibration problem.

[Yuriy Toropin]

I think the streaks is a bias calibration problem?

Looks like you've used Canon DSLR?
If so then "streaks" is natural feature for them that could not be fully eliminated by perfect callibration only.
As was pointed out by Nocturnal, differencial flexure extends "points", individual defects in calibration, to "streaks".

It could be effectively removed by introducing dithering, small random shifts between exposures.
In this case the same "streak" will take different position (relative to stars) on different frames and will be removed by sigma clipping.
If you can't automate dithering than you can do it manually between frames, at least several (6+) times shooting the same object.

[astropixel]

Thank you Yuriy and Sander That all falls into place now. On this particular image run, alignment was acceptable for short exposures and therefore I made no mid stream adjustments, such as framing or small corrections to alignment - which would do as you suggest - I often make small changes midstream. And, as I remember, a set from last year to which no adjustments were required - framing or alignment - had the same streaking. I couldn't work out what was different and why some image sets were not affected.

I guess it is possible to dither the individual frames with software - a combination of calibration and dithering. Is that on the right track?

[astropixel]

According to the threads I've read, it seems that PI does not employ dithering - is that correct. Aside from all the other issues associated with the streaks on the image, which I understand, how should I go about dealing with the lack of dithering during the image run?

[Yuriy Toropin]

What you are reffering to is called "drizzling", attempt to improve resolution in stacking of undersampled images during postprocessing. In "drizzling" images are scaled up by factor 2 or 3 than alligned and stacked. It will not help in your case.

"Dithering" is a different thing, it is a trick used during images acquisition.
Idea is in small shifts of optical axis of imaging setup relative to imaged object, in this case stellar images will be exposed on different parts (pixels) of CCD. Technically, it's just small movements of mount, several pixels, done in arbitrary directions between individual exposures. As a result, small local uncertainties related to CCD will be averaged and sigma-clip'ed out when aligned images will be stacked.

[astropixel]

Thanks Yuriy.

I have a previous set of similar images taken a few weeks ago. During that run I made several adjustments and the streaks are not present. Could I interpose/alternate the images - possibly drizzle with the  unaffected images. I'm just guessing here - being creative - a work around?

[astropixel]

OK. I combined a few images from one set with the other, please ignore the oval stars and satellite, I'll vet the images later and use only the best. The streaking is not as evident.

[pfile]

it would not be the first time that i've misunderstood something, but... isnt the fact that the mount is drifting very similar to having done dithering on purpose? (aside from any oval stars that might result from the drift...)

in other words, the streaks are actually a good thing in the sense that those hot pixels are not stacking right on top of one another, and could be rejected by tweaking the stacking parameters... right?

still yes, it could be that the calibration was a bit off in the first place, leading to excessive dark or bias noise being left in the calibrated frame.

[Nocturnal]

That is correct pfile (Rob, right?). With enough subs and a sigma clip type stacking method the streaks may be minimized. It's better to prevent them with proper calibration though. Sigma clipping sounds good but every pixels it throws away lowers the SNR of the pixel. Ideally each pixel has properly correlated (no outliers) values so they can be averaged for best SNR.

[astropixel]

Just to be sure, I have created a new master bias and dark - Vicent's tutorial. The last set of lights integrated with no failure of MRS noise evaluation.

I'm not sure of what parameters to apply to pixel rejection other than the defaults, keeping in mind that the average frame is 15 seconds and the total over 2 nights = approx 68 minutes.

Sky gradients are outer suburban, where the MW and bright features are discernable - LMC - Eta Carina - M8 and so on. I have attempted to limit the exposures so that sky glow is not overwhelming, tracking errors are minimized and features and stars are not blown out. So lots of frames - 266 so far. 200mm, f/2.8 (first night), f/3.2 (second night) iso 800.

I can get away with up to 5 minute exposures at times, unguided. An old EM-200B, PEC = +/- 5 arc seconds - tracking appears to be excellent, even unguided.

[pfile]

hi sander, yah, that's my name, don't call me late for dinner!  8)

astropixel - what stacking method are you using? in general making the 'sigma' values lower rejects more pixels from the stack. you can experiment with different values reasonably quickly after the first time you've done an integration. just tell ImageIntegration to work over a small but interesting 'region of interest'.