Strange noise pattern

[EorEquis]

I've begun seeing a noise pattern in some images after integration that really has me baffled.  It looks like so :



This pattern appears in both drizzled and non-drizzled integrations, and it SEEMS to appear regardless of settings in Integration windows.  I get this same pattern using a dark master, or using no darks.

The image above is 14 x 1200" Ha frames, captured in the same session.  PHD2 dithering was enabled.  Frames were calibrated with 30 x Flats, Bias Master, and both with and without Dark Master, with the same pattern.  Frames were integrated with these settings :



Gear is :

• ED80T CF
• Orion DSMI-III CCD
• Orion Ha Filter
• PHD2 Guiding, high dither enabled.  I have confirmed with Blink that dithering was functioning, ~ 2-3 pixel movement each frame.

Autostretched Master Dark :



And finally, a single autostretched frame :

[EorEquis]

This is the result of integrating all 14 light frames with no calibration of any kind, only registration and default ImageIntegration settings.

[EorEquis]

I've experimented more with very aggressive pixel rejection settings, and also tried each of the rejection algorithms.  None of these changes seemed even to attenuate the pattern.

[EorEquis]

Well, hrm...

Seems I had 2-3 images at the tail end of the session show some drift...trailing was slight enough 1-2px maybe) SFS didn't kick the frames, but enough that apparently drift + dither = streaks.

SO...I have some more guiding troubleshooting to do.

[Carlos Milovic]

It looks a bit like field rotation. How accurate is your polar aligment? How many pixels are you dithering?

[Geoff]

And if it is field rotation, the quickest way to check this will be to stack the unregistered images.
Geoff

[EorEquis]

It looks a bit like field rotation.

It does, doesn't it?  However a stack of unregistered images is rock solid...so that makes that unlikely, IMO.  (Heh, just saw Geoff's suggestion...yep, concur)

How accurate is your polar aligment?

Depends on who you ask.     PHD2's Drift Aligner said < 2' of PA error in both axes...and that seemed pretty reasonable, as there was minimal dec drift watching the trend lines.  PHDLab, however, estimated over a degree of PA error from the guidelog.  but, frankly, I've got next to no confidence in that estimate.  FWIW, an early-evening test frame of 180" with no guiding showed no trailing at all...so I'm feeling like the alignment was pretty close.

How many pixels are you dithering?

Don't know.  SGP speaks in terms of "Small" or "High" and so on :   I was using "High" last night.  SGP's manual gives no guidance on this setting, and PHD2's Dither amount is applicable only to manual movement.

Having said all of that, from running Blink on all my frames, I'd say it was 2-3 pixels.


There's some drift and elongation in the final 3-4 frames of the 15 frame set...despite no changes to the guide graph.  I would, of course, immediately suspect differential flex, except I'm using an OAG, so that seems...unlikely to say the least. *heh*  The drift/elongation DOES occur just before, during, and after the meridian (I was guiding through it, did no flip) so I suspect the two are related...but I'm not sure how or why at this point.

To be sure, however, removing those frames that drifted resolves the issue...and the further back into the set I cull, the better the result...

[EorEquis]

It does, doesn't it?  However a stack of unregistered images is rock solid...so that makes that unlikely, IMO.  (Heh, just saw Geoff's suggestion...yep, concur)

I was wrong.

I forgot I'd only stacked the first 10 of the 15 frame set, unregistered, and it was fine.  Adding the final 5, however, where I'd noticed the drift, produces some quite obvious rotation just as you suggested.



So now the question becomes...why did I only start getting drift and rotation 2/3 of the way into the night...hrmmm.

[Alejandro Tombolini]

Hi EorEquis,

The first part of this processing example could be helpful.

Saludos, Alejandro.

[EorEquis]

Hi EorEquis,

The first part of this processing example could be helpful.

I had found that example, but I do not believe it's applicable here for several reasons :

• This pattern appears whether darks are used or not.
• The reference frame was chosen from the group of 10 that stack well without registration

Not only have I not satisfied either condition, the example referenced is clear that BOTH must be satisfied :

using a MasterDark that differs too much from lights in exposure time and temperature AND if you did not have choosen a correct reference frame for integration could be a good cause.

It seems fairly clear at this point that the issue was, indeed, field rotation, introduced in the final 5 frames.  I simply need to determine how that could have happened.


As a followup to above, re: dither amounts, I did finally find some information on that :

From this thread, it appears PHD2 dithers at "+/- 0.5, 1.0, 2.0, 3.0, 5.0" pixels, for the Small, Medium, High, Very High, and Extreme values.  My PHD2 Dither Scale is at the default of 1.0, so the answer would appear to be "2px"...in line with my observation in blink of "2-3 pixels".

While I doubt that's applicable to this particular issue, it probably wouldn't hurt to bump that up to 5 or so...

[Carlos Milovic]

Indeed. To eliminate correlation between pixels, a large dither is needed.

[EorEquis]

Indeed. To eliminate correlation between pixels, a large dither is needed.

Out of curiosity then, is there a "preferred" amount?  Depends on image scale?  Image size?  Target?  All of the above?

[IanL]

Out of curiosity then, is there a "preferred" amount?  Depends on image scale?  Image size?  Target?  All of the above?

Even the largest PHD dither of 5 pixels is not really adequate to avoid this sort of effect.  Aim for something around 15 pixels, so that's 'very high' with a scale of 3x.  Clearly that might cause more issues with guiding settling post-dither so you may need to adjust the PHD / capture application parameters to allow more time between frames.

This sort of streaked background  pattern crops up all the time on various imaging forums.  The diagnostic is the hot pixels though - if your stars are nice and round (which they are) and you have little dotted trails of hot pixels in the same orientation as the background streaks then it's lack of dithering.  You can usually find a way to reject the hot pixels themselves, but the background streaks are much harder to suppress.

[gvanhau]

So now the question becomes...why did I only start getting drift and rotation 2/3 of the way into the night...hrmmm.

Probably because the balancing of your scope/mount was not good.

[mads0100]

It's nice to see you're rude everywhere.  They've given you good advice.