What is considered best practice in regard to how often to dither while capturing subs?
Once every sub?
Every other sub?
Once every 3 subs?
Thanks,
Jerry
Once every sub?
Every other sub?
Once every 3 subs?
Thanks,
Jerry
Dithering
Well Written Engaging
For narrowband, I do after each sub.
For broadband, I do after 2 subs.
For broadband, I do after 2 subs.
Bad pixel removal routines will perhaps work reasonably if they have 20 or more shifted frames. I normally set it up as a such that I get to that condition. So, if I plan less than 40 frames per filter in the session, will dither after each frame, for more than 40, every second frame for 60 and more, every third and so on…
Helpful Concise
Never
With the IMX571 I dither every 10 minutes regardless of subexposure time.
Well Written
Generally I dither every 3.
That being said, everytime you dither, the mount takes a few seconds to settle. Just keep that in mind, excessive dithering adds unnecessary time to your sequence.
OSC dithering often is a must. Mono can get away with less.
That being said, everytime you dither, the mount takes a few seconds to settle. Just keep that in mind, excessive dithering adds unnecessary time to your sequence.
OSC dithering often is a must. Mono can get away with less.
Helpful Concise
I shoot in OSC colour and only dither every 3 frames, I've never noticed any problems.
My old mount used to take ages to recover from a dither so every 3 seemed like a good balance for me and minimised the time I lost between exposures. Although I've just upgraded to an AM5 which settles much more quickly than my old EQ5 ever did, now might be the time to dither more frequently.
As Stjepan has said, I've been led to believe that as long as there are a sufficient number of dithered subs in the stack then the stacking algorithms will be able to do their job.
My old mount used to take ages to recover from a dither so every 3 seemed like a good balance for me and minimised the time I lost between exposures. Although I've just upgraded to an AM5 which settles much more quickly than my old EQ5 ever did, now might be the time to dither more frequently.
As Stjepan has said, I've been led to believe that as long as there are a sufficient number of dithered subs in the stack then the stacking algorithms will be able to do their job.
Well Written Concise
As much as necessary, as little as possible. I dither every 10 minutes for less than 5-6 hrs of integration, every 15-20 minutes for longer projects. Same sort of idea as Stjepan. Each dither takes time and even more time if your mount isn't quick to settle. The less it's wasting precious clear sky on dancing around the more data you get.
Helpful Concise
On the AM5 / Asirair, I just dither 5 pixels every 5 minutes. With narrowband for me that is every frame. For Broadband I usually take 60 second exposures so every 5 frames. AM5 settles really fast, just a few seconds needed to do this. So over a whole night session it does not consume more than 5-10 minutes of imaging time.
Helpful
andrea tasselli:
Never
Never? Why is that?
Sean Mc:
Never? Why is that?
Never did, never will. Waste of time, as far as I am concerned. Mount drift will do all the dithering you may ask for. My mounts anyway (all of 5 of them).
Mount drift will do all the dithering.
You what.
You what.
Proper dithering is done in two ways.
1. With OSC or NO OFFSETS: After every frame.
2. With filter offsets: Once per loop iteration, so each filter is dithered relative to the last image it took. Aka, SHO dither SHO dither.
Most of these replies really show how little the community knows about basic imaging fundamentals. zzz
1. With OSC or NO OFFSETS: After every frame.
2. With filter offsets: Once per loop iteration, so each filter is dithered relative to the last image it took. Aka, SHO dither SHO dither.
Most of these replies really show how little the community knows about basic imaging fundamentals. zzz
I'm sure you do.
andrea tasselli:
Never did, never will. Waste of time
If proper astrophotography is a waste of time to you then why are you here to tell us what to do? You have over 2k image hours and I'd expect someone like you to have a much better understanding of imaging fundamentals.
Well Written
When I use filter offsets, every rotation. When I don't usually anything that adds up to 5-6 minutes, could be one sub on NB, could be 3 on BB. Ideal is every sub. Dithering is not pointless. People that say they don't need it are either not stretching their data hard enough to see why one dithers, or simply don't know why. I've never found an instance where I could go without it.
For me it depends on the exposure time & number of total subs I'm planning. Once per 10 minutes as others have mentioned seems about right.
If proper astrophotography is a waste of time to you then why are you here to tell us what to do? You have over 2k image hours and I'd expect someone like you to have a much better understanding of imaging fundamentals.
I'm not here to tell YOU what YOU have to do but to state what I do. And I don't, most emphatically, dither. And there is no such thing as "proper astro-photography". That ship has long sailed away.
Depends on aperture and your preference. If I have exposure time of 60 seconds I do it after 180 seconds.
Dithering only takes a couple of seconds with each of my two mounts. I dither every single frame. Combining dithering with image calibration is the best way to minimize spatial variations in dark current along with FPN due to PRNU.
John
John
Well Written Insightful Concise
Jerry Gerber:
What is considered best practice in regard to how often to dither while capturing subs?
Once every sub?
Every other sub?
Once every 3 subs?
Thanks,
Jerry
The 1st question to answer is why should one dither?
The 2nd question is how to I select the dither frequency based on my setup and calibration? Because how much you dither really depends on many factors.
One dithers in part to enable removal of hot and cold pixels and other fixed pattern noises from your sensor. The idea is the fixed pattern noise is fixed with respect to the sensor. So if you move the sky with respect to your sensor such that the pattern ends up in different parts of your image, those fixed patterns become random in your final image. With multiple subs and dither these fixed pattern noises become random and will increase with the sqrt(# subs), while your signal increases by #subs. So your S/N relative to fixed pattern noise increases. If you don't dither, fixed pattern noises increase just as fast as your signal, and you never get rid of them. Fortunately, as was pointed out by someone else in this thread, if you use a signal rejection method in your stack, as long as you have enough dithers to enable rejection, this should remove these fixed pattern noises. In theory as few as 12 dithers should work, but I agree with the other post, at least 20 or more would be better for rejection filters.
The second point is that darks add noise to your image! And if you are going to use flats, you need to subtract a dark from your lights, otherwise your flat correction will be wrong. Of course, you also need a flat dark for the flat. Also, darks can help somewhat with pattern noise, so certainly another reason to do them. However, we cannot ignore that darks add noise, both read noise (NR) and thermal noise (NT). So if we take an equal number of darks and lights, we will increase NR and NT by sqrt(2) =1.41x! To ensure we don’t increase read noise significantly by using darks, we really should have about 3X as many darks as lights! Going back to the example, if we have equal number of darks and lights, then we need to dither every 3rd frame to meet the above criteria. If we had twice as many darks as lights, we only need to dither every 6th frame.
Now fortunately, in most cases sky noise will dominate read and thermal noise in your lights, if your exposures are sufficiently long with a modern low noise camera and cooling. So one need not need to stick to this 3X criteria, since these noises are not the dominant noise. So you can relax this criteria depending on your situation.
Finally, flats do not contribute as strongly to noise, since the signal in flats is strong compared to read and thermal noises. Still I usually do 100 flats just to be sure, with a panel 100 flats takes about 30 seconds, so not a big deal. Darks unfortunately take a lot of time for long exposures, but if you have a dark library that you use for a long time this can be well worth the effort to take a lot of dark subs.
So to reiterate:
1) For pattern noise, a total of at least 12 dithers per final image, preferably more, assuming you are using rejection, which you should.
2) If you have a significant contribution from read noise+thermal noise in each sub then:
# dark frames * dither interval >= 3* # light frames
3) If your read+ thermal noise is relatively small compared to sky noise, then your can reduce the requirement in 2). How much depends on the relative noise levels.
I suggest for anyone who is really serious about optimizing their imaging, you need to know you noises. Read noise and thermal noise are in the camera spec sheet. Sky noise can be measured on one of your subs—measure the adu and then convert to electrons using your gain (in e/adu).
Finally, if you are not guiding and you are getting drift in your image, you need not necessarily dither very much. I do not guide my images, and I have a portable setup with exposure to wind and vibrations, as well as some inevitable errors in polar align, tracking errors, and meridian flips. With just a few random dithers over a night’s imaging, and these natural variations, I do get effective dithering and it is random enough. If it is not random enough you will see walking noise, which is a pattern of noise that lines up in one direction, a result of fixed pattern noise that is not averaged enough.
Hope this helps
Rick
Helpful Insightful