Dither settings

[KathyNS]

I have never really paid much attention to dithering. I know it is a good thing, so I turned it on. Now I'm thinking that I should probably tune it up a bit. I have no idea what good settings are.

In SGP, dithering is set to "Medium", whatever that is. In PHD2, Dither Scale is set to 0.1, whatever that means. I had RA Only selected in order to speed up settling, and I know I should turn that off.

What does the combination of Medium and 0.1 do for me? Is there a better combination that I should use? My imaging scale is 1.2"/px. My guiding scale is 6.6"/px.

[DEnc]

My understanding is that the SGP dither settings, small – extreme, give displacements of 0.5, 1, 2, 3, and 5 pixels; and the PHD2 settings are multipliers.

How much to dither? The only concrete statement I’ve run across (rather than opinion and experience of “what works”, which of course are perfectly valid) is from John Hayes over on CN:--

Dithering has to be greater than the autocorrelation distance of the image. This is the distance where features are statistically uncorrelated (i.e. random) across the image. In simple terms that means that the dithering distance has to be larger than the spatial scale of the features that you want to eliminate. This distance varies with the type of the sensor. On my KAI-16803 sensor, 6-9 pixels works pretty well. CMOS sensors typically require a larger distance but again it will depend on the sensor. I'd recommend trying something in the range of 9-15 pixels with an OSC CMOS camera.

I took an empirical approach and extracted the lock position values from the PHD2 GuideLog in Excel, and calculated the Pythagorean distance between successive locks. I did this for combinations of SGP and PHD2 settings. Here’s an example of one session. These are guide camera pixels—I didn’t convert to the equivalent displacements on the imaging camera. There's quite a range of distances! I went with settings that gave at least 5 pixel displacement on the imaging camera, predominantly.

[JayTee]

I think [mention]STEVE333[/mention] is really good at this answer.

[Juno16]

Yes, JT, Steve is really good at dither settings. Looking forward to his response.

Kathy, My imaging scale is similar to yours at 1.33”/px with a guiding camera image scale of about 6”/px.With my dslr, I used a dither setting in APT of 3 with a random dither scale of 1.0 in PHD2. By my calculations, the guide camera dither of 3 pixels (3x1) Times the ratio of guide camera image scale to imaging camera image scale (6/1.33) gave me dithers of about 15 pixels at the imaging camera. I dithered every frame. This is what it took to eliminate the “rain” in my images.
With my new camera (ASI 533), I use a dither setting of 1 (px) in APT x the 1.0 scale in PHD2 to give me a 5 px dither (roughly) at the imaging camera. I dither every other frame. The lower and less frequent dithers are enough to eliminate “walking noise” in my imsges and is much less of a recovery time for my mount.

For me, less is better for mount and also less cropping when processing.

[yobbo89]

do you use sgp kathy ? there should be more brief settings in there. i use the maximum this helps to dither out internal reflections

[KathyNS]

do you use sgp kathy ? there should be more brief settings in there. i use the maximum this helps to dither out internal reflections

Yes, I use SGP. My options there are Small, Medium, High, Very High, and Extreme. Someone said these correspond to 0.5, 1, 2, <something more> and <something even more> guide camera pixels. Which is then multiplied by the setting in PHD. There is a mount guiding rate as well, which I have set to 0.8, though I don't know if that applies to dithering or not.

After starting to research this more, I have set SGP to Medium (= 1 guide camera pixel), PHD2 to 1.1, which should give me a dither of about 6 imaging camera pixels. Which may or may not actually be 4.8 pixels if the guiding rate applies. Does that make sense?

I am guessing that my dithering to date has been fractions of a pixel, which is good for the settling time, but not so useful for noise reduction.

[starfield]

I remember posting a similar question and seem to remember the rule of thumb was to shoot for 10 to 15 imaging pixels. I used a Very High Dither setting and no multiplier in PHD2 which takes gets me close to a 10 px dither.

[DEnc]

Someone said these correspond to 0.5, 1, 2, <something more> and <something even more> guide camera pixels. Which is then multiplied by the setting in PHD.

"Someone"? "Something"?

I'm out of here.