APT dithering distance

[BrendanC]

Hi all,

So, dithering distance when using APT dithering...

I'm not guiding (yet), but I am using APT's built-in dithering feature because it helps me prevent walking noise.

However, I'm slightly confused about the dithering distance setting.

From the user guide, I understand the setting is specified in pixels. However, the number of pixels covered, will equal a different number of arcseconds covered, depending on my focal length and camera.

For example, my 130PDS has a focal length of 650mm, while my EOS1000D has a pixel size of 5.71 microns. According to this page (https://astronomy.tools/calculators/ccd), the number of arcseconds covered by one pixel is (5.71/650)*206.265=1.8 arcseconds per pixel. So, if I specify, say, 10 pixels for the dithering distance, that would cover 1.8x10=18 arcseconds within the image.

Let's say I then use my 0.9x coma corrector, which reduces my focal length to 590mm. Now, each pixel accounts for (5.71/590)*206.265=2 arcseconds per pixel, so now 10 pixels of dithering distance covers 2x10=20 arcseconds of image.

Then, if I use my 3x Barlow, with no coma corrector, my focal length becomes 1950mm, which radically changes things. Now, each pixel is (5.71/1950)*206.265=0.6 arcseconds per pixel. So, if I specify 10 pixels in dithering distance, this is now 0.6x10=6 arcseconds of image. Quite a difference!

So, could someone clarify please: when I specify the pixel distance, am I right about the difference in arcseconds of image that would be covered depending on my optical train? If so, does APT auto-adjust to compensate for this in some way? Or, is the pixel distance constant, because what really matters is in fact the actual number of pixels of dithering, regardless of the arcseconds covered?

Or am I not making sense here? Should I just carry on and stop worrying?

Oh, and while I'm at it, what would people recommend for the optimal distance? I've seen references to at least 12 pixels (see https://stargazerslounge.com/topic/253335-how-much-dithering/ so I'm currently at 13 pixels to be on the safe side, but I'd like at least to understand why I'm doing this, and whether I could go lower, because higher dithering distances will mean less usable image after stacking.

Thanks, Brendan

 

[michael8554]

2 hours ago, BrendanC said:

is the pixel distance constant, because what really matters is in fact the actual number of pixels of dithering,

Correct. But if you were Guiding then the arcsec/pixel relationship between main and guide cameras becomes inportant.

2 hours ago, BrendanC said:

higher dithering distances will mean less usable image after stacking.

Why ? You Dither to make images MORE usable after stacking.

But 12 pixels seems to be the norm.

If it's a Random Dither setting with a maximum of 12, then most Dithers will be less than 12 pixels.

Michael

[BrendanC]

Sorry, I didn't mean a less usable image, I meant less image - as in, the more the image is dithered, it effectively crops out more of the edges of the image. 

Anyway, looks like this is something I don't need to worry my little head about (yet). 

Thanks.

[symmetal]

Dithering is done to remove or reduce defects in the camera, primarily fixed pattern 'noise'. It can also remove hot/cold pixels though other methods can help with hot/cold pixels like darks and/or bad pixel maps. These defects are in the same position on each frame and are not related to what target the camera is pointing at, or its size, so the image scale has no bearing on how much to dither. 🙂

The most obvious fixed pattern noise in DSLRs is the colour mottling effect where clumps of adjacent pixels tend to have a distinct colour cast. These seem to extend up to a maximum of around 12 pixels wide which is why 12 pixels is chosen as the dither amount for DSLRs. Cooled OSC CMOS cameras seem to have a similar mottling effect but it's less pronounced, probably due to being cooled.

Other fixed pattern noise effects are based on slight sensitivity changes between pixels but these don't seem to appear in such large clumps as the colour mottling so dithering by just one or two pixels can usually be sufficient to combat them. So mono cameras are usually fine with around 2 pixel dithers.

I've found using 'random' dither in PHD2 seems to prefer a specific direction so the final stacked image needs cropping quite a bit on just 2 adjacent sides. Using circular dither seems to lessen the amound of cropping needed to around 6 pixels a side. 🙂

Alan

 

[Merlin66]

I don't use drizzle, but I'm interested in the possible imapct on resolution. Moving your image +/- a few pixels infers to me that you're spreading the star FWHM and reducing the possible resolution.

Is this correct?

 

 

[Seelive]

Did you mean dither rather than drizzle? (2 completely different techniques). The stacking software realigns the stars in the images so the FWHM will not be degraded.

[Merlin66]

Yeah, 

I was thinking of the discussions on dithering.....

I’d like to think that the stacking software is capable of re-aligning the individual images.....but not sure. (I haven’t seen examples to quantify)

 

[Seelive]

4 hours ago, Merlin66 said:

Yeah, 

I was thinking of the discussions on dithering.....

I’d like to think that the stacking software is capable of re-aligning the individual images.....but not sure. (I haven’t seen examples to quantify)

 

Any stacked image relies on the stacking software to align the images to correct for any misalignment between the individual images. The fact that the misalignment may be caused by the unintended mount polar alignment error or by the intended random movement created by dithering is irrelevant to the stacking software. There are plenty of images out there that state dithering was used (and they are normally the better ones).

[symmetal]

8 hours ago, Merlin66 said:

Yeah, 

I was thinking of the discussions on dithering.....

I’d like to think that the stacking software is capable of re-aligning the individual images.....but not sure. (I haven’t seen examples to quantify)

 

The individual images have to be aligned by the stacking software. Unless you have perfect polar alignment, a perfect tracking mount, or perfect guiding, and take all your subs in one session on one side of the meridian, you'll end up with misaligned images. 🙂

A slight polar alignment error and taking a series of short unguided subs gives the 'walking noise' effect which is a result of a form of dither but in one direction only so the sensor fixed pattern noise shifts by a small but consistant amount between subs. Making these dither movements between subs in a random direction and distance means the fixed pattern noise is 'randomized' and so can be removed by stacking enough subs.

A meridian flip is an extreme example of a dither, as well as taking subs over different sessions as centreing the target at the start of a new session is the same as a large dither of the amount of your centreing precision.

Polar alignment errors also means each image will be slightly rotated from the others, so good stacking software will rotate each image by a sub pixel amount as well as shift it in x and y to best align the images.

This of course means the edges of the stacked images have to be cropped off and large dither movements generally means more cropping but the benefit of dithering outweighs this.

Alan

 

Edited September 20, 2020 by symmetal

[BrendanC]

Textbook stuff!