Help me understand the hype behind dithering

Yes, you must dither to prevent walking noise and other noise issues which do not calibrate out otherwise. If you want a quality result, especially on faint targets where you’ll stretch aggressively, dithering is not optional. I’m sure someone is going to come up with a counter example and show some image where they didn’t dither and got IOTD or something, but unless that someone is you and you know exactly why dithering doesn’t apply to you, you should definitely do it.

AstroShed:
Well I have stated why I don’t do it, a good set of temp matched darks, at least 50 of them combined into a master get rid of all my walking noise, so why would I need to dither too, that is the question, I understand that some people no longer use darks, so they would need to dither to remove walking noise, but in my case I do use them.
What am I missing here, obviously something..🤔

Or not...

Before I incorporated guiding in my setup I would have to periodically re-center the image because of drift.  I shot a lot of frames (still do) and always re-centered to a slightly random point. I wasn't trying to dither because I didn't know about it at that point. It was just faster to not be too precise. When I started guiding, my images were suddenly swamped with walking noise, that was with well calibrated data. Dithering fix that issue so for me, it's something I have to do. That said, it does reduce total integration so I try to minimize it as much as possible. With 15 sec subs I dither on every 20th frame, that seems to be the right balance for my setup. I would love to not do it but that's just not an option for me.

I should mention that I shoot in a  Bortle-8 zone.

Well, as stated in the beginning, the real benefit is drizzle in many cases. 

However, even with a good dark frame calibration, the dark frames will always add some noise on their own into the light frames. If you have significantly more lights than darks, which happens often for me, you might end up with a lower amplitude fixed pattern noise in your frames again. Again, dithering helps here. 

Also, even with good flats, there will always be some non uniform responsiveness in the pixels. Dithering helps again. 

So in conclusion, besides drizzling, most things that dithering died are better done by other techniques. Dithering is just a simple yet extremely effective option to get rid of all the residuals that are inevitable and make sure you don't miss anything. 
For me, it is a very cheap and easy fix for otherwise complex problems. 

CS Gerrit 
​

AstroShed:
Well I have stated why I don’t do it, a good set of temp matched darks, at least 50 of them combined into a master get rid of all my walking noise, so why would I need to dither too, that is the question, I understand that some people no longer use darks, so they would need to dither to remove walking noise, but in my case I do use them.
What am I missing here, obviously something..🤔

What you are missing may be a clear understanding of the statistics of image calibration.  First, you don't need a stack of 50 subs to create a master dark.  16 subs are generally fine for just about any size stack.  (I worked out the math to show this a number of years ago but I won't reproduce it here.)   Second, dithering is a method of mitigating the effects of fixed pattern noise (FPN), which is due to PRNU (photo response non-uniform) across the sensor.  FPN is proportional to signal strength.  PRNU is pretty small in most modern CMOS sensors but it is not zero.  In order for dithering to work well, you need to set the dither distance to be more than the autocorrelation distance of the FPN, which isn't always easy to determine.  I typically use 10 pixels but that's probably just barely enough.  15-30 pixels with most CMOS sensors is probably better.  If you want to learn more about this stuff you might read Janesick's book, "Photon Transfer".  (https://www.amazon.com/Photon-Transfer-Press-Monograph-PM170/dp/0819467227/ref=sr_1_2?crid=BYMB3OSWR046&dib=eyJ2IjoiMSJ9.bsMWECwhdB6KFZs5b3-nbYW83L0m5-sTbAPVExU5uuABNLPs5XBfDxf7z0bjG6z9YW1qYtFmwwiaP4qYzCpb4_LEp63bYCrCGB4IIxi0gIfQQMG_AlmV-RYTdnZQnRYZmRwHaeeNkFy0ZXeY2E-Yg0042pXcHCzERBkXos5843rmGLz6lwFPJsrX2BYq_jiMMP3ajUcqsBwaXFeQJiWj7xkv_izX3LLjaR_MVbS3cMMT-fTKukzyFj6UU1dxUyUp.CfaMa7kK8Oj4Bjgvmvq1McGFga6Gv5H8ugd7JWVNzh8&dib_tag=se&keywords=photon+transfer&qid=1741533777&sprefix=photo+transfer%2Caps%2C188&sr=8-2)

Even better,  Berry and Burnell's "Handbook of Astronomical Processing" provides an outstanding review of image calibration.  Unfortunately, it is out of print.  I found a copy here but it is super pricy:   US_Shopp_Textbookproduct_id=COM9780943396828USEDkeyword=&gad_source=1&gbraid=0AAAAAD3Y6gvnc32UkfTSPB8LqD-EcBrTY&gclid=Cj0KCQiAlbW-BhCMARIsADnwasoAoZFb3aH6AcIiNPzLP7ssQ0ocek42liE5Ps1W1XJlc7YoQN1o-74aAmRBEALw_wcB">https://www.abebooks.com/servlet/BookDetailsPL?bi=31953293771&dest=usa&ref_=ps_ggl_11147913055&cm_mmc=gglUS_Shopp_Textbookproduct_id=COM9780943396828USEDkeyword=&gad_source=1&gbraid=0AAAAAD3Y6gvnc32UkfTSPB8LqD-EcBrTY&gclid=Cj0KCQiAlbW-BhCMARIsADnwasoAoZFb3aH6AcIiNPzLP7ssQ0ocek42liE5Ps1W1XJlc7YoQN1o-74aAmRBEALw_wcB

You can ignore all this stuff and produce pretty good images, but if you want the very best results, proper image calibration is critical.

John

In order for dithering to work well, you need to set the dither distance to be more than the autocorrelation distance of the FPN, which isn't always easy to determine.  I typically use 10 pixels but that's probably just barely enough.  15-30 pixels with most CMOS sensors is probably better. 

John

Hi John,

Thanks for pointing this out.  >10 pixels is also my observations, although I don't think I really understand it.  Do you know what does cause this correlation length of >10 pixels?  In CMOS, every pixel should be independent.  Even on interpolated images from Bayer arrays, the correlation shouldn't spread over 5 pixels.  What can cause correlation over >10 pixels?