How much dithering ?

[alan4908]

I've recently implemented dithering and I've started at 1.5 pixels in both the RA and DEC axes. I've noticed that I get an improved image quality but on the down side I also have to wait about 120s for the guider to settle. The settle time is always much longer on the DEC axis which I understand is consistent with mount backlash. To reduce the guider settling time, I was thinking about reducing the amount of dithering on both the RA and DEC axis and maybe even turning off DEC dithering altogether.  Is this is good idea ?

Given that the actual settle time is variable, I was also thinking of switching on the option in MAXIM to set the settling criteria to be dependent on the measured guider error (eg say 0.3 guide pixels) rather than a fixed time.  Given that my main camera is imaging at 2.39 arc second/pixel and my guide camera is 4.42 arc seconds/pixel, if I select 0.3 guide pixels then this would be 0.3 x 4.42 = 1.33 arc seconds or about half a main camera pixel. 

Alan 

[focaldepth]

I am not yet what I would call an expert in all this but in PHD2 and Nebulosity 3 I use High Dither and settle at < 1.0. (imaging at aprox 1 sec per pix and guide at aprox 2 sec per pix)

 

My thinking is to remove hot pixels and aeroplanes etc. we need to spread the offending pixels over as many registered pixels as possible, so I try to make sure the dithers are in RA and DEC and several pixels.

Increasing the settling tolerance speeds up the dither process.

[alan4908]

Hi Focaldepth

That's interesting, I find that I can very effectively eliminate very large defects eg aeroplane streaks etc without dithering if I use an SD Clip stack algorithm (which eliminates outlying pixels in the stacking). For me, I find dithering useful for eliminating colour pixel streaks (several pixels long without dithering) which I presume are due to camera imperfections.

Since it is now clear I thought I'd try 1 pixel dithering and see what happens 

Alan

[IanL]

Berry and Burnell recommend dithering by at least 12 pixels. I am not sure that a 1 or 2 pixel dither achieves much. A decent rejection method and sufficient frames will deal with any transient phenomena whether you dither or not. So really you are trying to address camera imperfections here, and a small dither risks the same piece of sky being covered by the same pixel multiple time in the stack, thus skewing the final result. Furthermore any drift between frames may cause those skewed pixels to build in to aa streaky pattern or similar. A big dither will eliminate both risks, and should not take any longer for guiding to settle than aa small dither would.

[Zakalwe]

Given that the actual settle time is variable, I was also thinking of switching on the option in MAXIM to set the settling criteria to be dependent on the measured guider error (eg say 0.3 guide pixels) rather than a fixed time.  Given that my main camera is imaging at 2.39 arc second/pixel and my guide camera is 4.42 arc seconds/pixel, if I select 0.3 guide pixels then this would be 0.3 x 4.42 = 1.33 arc seconds or about half a main camera pixel. 

Alan 

Sequence Generator Pro also uses a measured guider error when settling rather than a fixed time period. Heck of a lot cheaper than Maxim too!

[alan4908]

Zakalwe - yes, tried this with MAXIM with a 0.3 pixel guider error settling criteria last night and it works.... plus it cuts down the overall settling time dramatically - I think I will use this from now on. 

Ian - Thanks for reply. I understand what you are saying and I'm more than willing try much larger pixel dithering (at least 12 pixels) but I'm confused due to conflicting information.  To give you two examples:

Astrophotography by Thierry Legault, page 187, states whilst describing dithering "...A slight shift of a few pixels in a random direction between successive images because some defects may remain even after calibration..."

Starizona's guide to CCD images in the section on dithering states "The amount of offset can be selected by the user and several pixels is usefully sufficient"

Now, I'm really confused 

Alan 

t of offset can be selected by the user, and several pixels is usually su

[Zakalwe]

A big dither seems to work well if you are using a DSLR, as these seem to generate a heavy "mottling" effect. For a CCD, a much smaller dither will work

[MartinB]

Hot pixels are a piece of cake to remove, you don't need to dither to do this. However, for reasons I don't fully understand, dithering is reported to improve signal to noise ratio. I've read explanations as to why but they are beyond my limited brain. I think it is worth dithering provided it doesn't take too long for guiding to settle. Too long varies, if doing short binned RGBs then too long for me might be 10 seconds whereas with 30min narrowband subs I'm not going to worry about 1 minute.

As for how much you dither and whether you dither by integers or fractions of pixels, I've read all sorts. Berry and Burnell are out on a limb with 12 pixels as they are with some of their other suggestions. 12 pixels for most set ups will take significantly longer than 1.5 pixels especially if you are using guidestar exposures of several seconds and moderate aggression.

I think dithering has it's place and I plan to use it with my backlash free Mesu 200 but I only ever bothered using it with long exposures with my EM200 and could never really see much benefit.

I think your best bet Alan will be to experiment, you can do this fairly painlessly since it won't interfere with your imaging time. Use what works for you.

[glowingturnip]

not sure if you're using PHD or not, but if you are, PHD has a 'fast recenter after dither or calibration' option in the latest versions, so you could use that, I do.

I would keep the dither in Dec option turned on, it gives you a whole dimension more of random shift.  Especially if you're doing a dither and then drizzle integration in Pixinsight then if you'd only dithered in RA only, you'd have streaks of missing information in the Dec direction

[alan4908]

Zakalwe - yes, on your DSLR dithering point, I did come across this video by Tony Hallas (https://www.youtube.com/watch?v=PZoCJBLAYEs#t=39) which talks at length about large dithering to remove "colour mottle" in DSLR as opposed to CCD cameras. 

Martin - thanks for this - OK I will experiment with dithering around the 1.5 pixel level but keep an eye on my guider settling time. 

Stuart - yes, I was a bit concerned about just Dithering in the RA direction (in an attempt to circumvent my mount DEC backlash) - on reflection I don't think I'll do this since as you say, it is likely to seriously degrade the value of dithering since will destroy one dimension of randomness.  I use MAXIM with Photoshop CC and do want to try Drizzling at some point - perhaps not quite yet. 

Alan

[focaldepth]

I am not yet what I would call an expert in all this but in PHD2 and Nebulosity 3 I use High Dither and settle at < 1.0. (imaging at aprox 1 sec per pix and guide at aprox 2 sec per pix)

My thinking is to remove hot pixels and aeroplanes etc. we need to spread the offending pixels over as many registered pixels as possible, so I try to make sure the dithers are in RA and DEC and several pixels.

Increasing the settling tolerance speeds up the dither process.

I agree with others that I was talking nonsense about removing aeroplane streaks. Obviously dithering is for removing camera defects.

[IanL]

Zakalwe - yes, tried this with MAXIM with a 0.3 pixel guider error settling criteria last night and it works.... plus it cuts down the overall settling time dramatically - I think I will use this from now on. 

Ian - Thanks for reply. I understand what you are saying and I'm more than willing try much larger pixel dithering (at least 12 pixels) but I'm confused due to conflicting information.  To give you two examples:

Astrophotography by Thierry Legault, page 187, states whilst describing dithering "...A slight shift of a few pixels in a random direction between successive images because some defects may remain even after calibration..."

Starizona's guide to CCD images in the section on dithering states "The amount of offset can be selected by the user and several pixels is usefully sufficient"

Now, I'm really confused 

Alan 

t of offset can be selected by the user, and several pixels is usually su

Mathematically, this is a class of random walk problem. My statistics are not up to the job of working out probabilities, but a couple of quick tests in Excel make it clear that dither scales of 3 pixels or less produce a lot of overlap. A series of 30 subs can have a spot in the sky imaged by parts of the same four or five different pixels, leading to potential skewing when performing rejection. As the dither scale increases the probability of this occurring drops rapidly and at scales of perhaps 5 pixel dither there appears to be much less chance.

I honestly don't know where the figure of 12 comes from mathematically. I do know with my DSLR that a dither of about 15 eliminated any regular patterns in the sky background. At smaller dithers it would become quite pronounced. This won't be such an issue for CDDs but worth knowing.

[Herra Kuulapaa]

My DSLR dithering settings are usually ~5 pixels and I've noticed it's enough.

Of course we can nowadays spare much more dithering pixels with typical DSLR resolutions. Better more than less..