Back to forums

Dithering amount

ZWO AM5 Acquisition Dithering Deep Sky
7 replies•537 views
Anthony (Tony) Johnson avatar
Anthony (Tony) Johnson avatar
I have a question, and truly I’m not finding my answer on the internet so I’m hoping someone can answer it for me. My question; is there a formula I can use to figure out the amount of dithering movements needed for a given focal length camera combination? I’m new to this aspect of astrophotography, just purchased an AM5 mount, my previous mount was an LX200 which did not allow me to guide so guiding and dithering are new to me. I seem to be finding out a lot of what is is but nothing when it comes to figuring it out. I have 2 scope that I use, one of which I may not because of its size and weight. I found this formula on cloudynights, but it’s seems backward to me. Here’s the formula I found;K = 2 X (Main Scope FL / Guide-Scope FL) X (Guide Camera Pixel Size / Main Camera Pixel Size)
now when I plug my numbers into it, I get 5pixels for my 82ed 530mm and 15 pixels for my 12” SCT running at f/6.3, around 1768m. Which to me seems backward. Wouldn’t you need to dither a larger amount on a wider field and less on a narrow field of view, to get the same amount of dithering. Can someone please help because I’m confused as to which is right. Thanks.
Engaging
Anthony (Tony) Johnson avatar
Anthony (Tony) Johnson avatar
Oh crap, the light bulb just went off. The narrower field of view has to move more and in my example 3x more than the wider field to effect the same amount of dithering duh. Maybe that formula will help someone else out.
D. Jung avatar
D. Jung avatar
If you want to use dithering to remove FPN (fixed pattern nose), you should dither on the scale of your noise in pixels. That should be independent of the optics used. Probably the main thing you want to remove is the impact of dead pixels, which are usually 1 to several pixels clustered.
Helpful Concise
John Hayes avatar
John Hayes avatar
As D. Jung said, dithering has nothing to do with the optical system.  The whole idea is to use stacking filters to reduce the effects of unwanted signals that are fixed with respect to the sensor.   For most sensors, the most prominent unwanted signal is dark current.  FPN is the result of something called PRNU, which stands for “Pixel Response Non-Uniform.”   For most modern sensors, PRNU is pretty small and it’s calibrated out with flat data so that’s not why you dither. (To be clear, I’m using the most common definition of FPN as defined by Janesick.  There are other definitions of FPN floating around, which make it hard to discuss this stuff so that’s why I’m calling out the source of this definitions.  It’s precise and it’s well known.)  The amount of dither is determined by a mathematical quantity called the auto-correlation distance, which is the distance across the sensor where the signal at one position is statistically unrelated to other points on the sensor.   This isn’t a difficult quantity to compute but you don’t really have to go to that effort.  For most modern sensors with small pixels, dithering in the range of 8-20 pixels will be sufficient.  I personally get good results with my QHY600M-PH using a value of 9 pixels.   My advice is that if you are uncertain about what value to use, error on the high side.

John
Well Written Helpful Insightful
lunohodov avatar
lunohodov avatar
Anthony, NINAs documentation on dithering has helped me in the past.

The general recommendation is for dithering to shift the main imaging camera by about 10px.

First, you'll need to calculate the guiding and imaging resolution(s) with the following formula:
(Pixel_Size ÷ Focal_Length) × 206.265

Given your setup, the numbers are as follows:
Guiding_Resolution = (3.75 ÷ 190) × 206.265 = 4.07 arcsec/px
Imaging_Resolution = (4.3 ÷ 520) × 206.265 = 1.67 arcsec/px

So what is the amount of dithering pixels that will shift the imaging camera by 10px?
(Imaging_Resolution ÷ Guiding_Resolution) × 10 = (1.67 ÷ 4.07) × 10 = 4.1

Dithering with 4.1px (on the guide camera) will result in 10px shift for your image camera.

I hope this helps a bit.

Yanko
Well Written Helpful Engaging Supportive
Jan Erik Vallestad avatar
Jan Erik Vallestad avatar
Anthony (Tony) Johnson avatar
Anthony (Tony) Johnson avatar
John Hayes:
As D. Jung said, dithering has nothing to do with the optical system.  The whole idea is to use stacking filters to reduce the effects of unwanted signals that are fixed with respect to the sensor.   For most sensors, the most prominent unwanted signal is dark current.  FPN is the result of something called PRNU, which stands for “Pixel Response Non-Uniform.”   For most modern sensors, PRNU is pretty small and it’s calibrated out with flat data so that’s not why you dither. (To be clear, I’m using the most common definition of FPN as defined by Janesick.  There are other definitions of FPN floating around, which make it hard to discuss this stuff so that’s why I’m calling out the source of this definitions.  It’s precise and it’s well known.)  The amount of dither is determined by a mathematical quantity called the auto-correlation distance, which is the distance across the sensor where the signal at one position is statistically unrelated to other points on the sensor.   This isn’t a difficult quantity to compute but you don’t really have to go to that effort.  For most modern sensors with small pixels, dithering in the range of 8-20 pixels will be sufficient.  I personally get good results with my QHY600M-PH using a value of 9 pixels.   My advice is that if you are uncertain about what value to use, error on the high side.

John

Thanks John. I appreciate your in depth description of this. Helps a lot. Not saying I totally understand everything you wrote but I get the idea behind it, and understand it a lot better, I know another friend of mine on here @Jeff Kisslinger told me a similar thing, also gave me an idea of what he does which is very close to your answer.
Brian Puhl avatar
Brian Puhl avatar
Think of it this way.   Image stacking relies on pixel rejection.   Hot pixels will always be hot.   The neighboring pixel is probably fine and has a normal ADU value.   Now, if you don't dither, that pixel will always be in the same place relative to the stars around it, and that pixel will be perceived as signal and thus it will get stacked. 

If you dither, that hot pixel will move to another location relative to the stars around it.  Because it has moved, pixel rejection will recognize it as an outlier and reject it from the final stack.  

For mono cameras, i dither every 3, and move by 3 pixels.   Moving by greater amounts is fine, but it adds settle time to your guider because of the larger movement.  

For color cams, I dither every frame by 3 pixels.  The reasoning for added dither is because of the Bayer matrix.   

There are other reasons behind it, but this is the easiest way to explain it.  Excessively large dither movement is pointless IMO and can require more settle time…. but to each his own.   My settings have proved reliable, and dither times are under 4 seconds.
Well Written Helpful Insightful Engaging Supportive