Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

How often do you dither?

  • Please log in to reply
33 replies to this topic

#1 Marco 104

Marco 104

    Sputnik

  • *****
  • topic starter
  • Posts: 35
  • Joined: 10 Nov 2018

Posted 22 January 2020 - 09:19 AM

Hi all,

 

I'm working on Melotte15 and I'm checking intermediate integration results after each imaging night

 

I see that my integrations are way more noisy than they were weeks ago when I captured M42 (similar integration time), the only difference being that with the latter I did apply dithering on each frame, while with the former once every 3 frames and that Melotte15 is being done with SHO, instead of RBG

 

The camera is the ASI1600MM Pro, temp set at -20°C, gain/offset 139/21

 

To me dithering every 3 frames should already be high frequency, what am I doing wrong? Is this all about narrowband?

 

Thanks

Marco

 


  • ks__observer likes this

#2 DrGomer

DrGomer

    Ranger 4

  • *****
  • Posts: 344
  • Joined: 07 Aug 2019

Posted 22 January 2020 - 09:22 AM

I dither every frame which are typically 2 minute subs.
  • AstroFrankMontana and Marco 104 like this

#3 RaulTheRat

RaulTheRat

    Viking 1

  • -----
  • Posts: 678
  • Joined: 09 Oct 2018

Posted 22 January 2020 - 09:27 AM

In narrowband I dither every frame - the individual subs are long enough that the dither time is acceptable, and there's less subs to stack for a given total integration time.

With LRGB I'd dither every 4.

Let's say for example 30s L and 60s RGB, and that I take 2x as many luminance as each of RGB, then that's a total of about 5 minutes per "set" of frames, Vs 5 minute narrowband subs.

So I'm dithering about every 5 minutes in either case, and I have the same number of total dithers in a given total integration time.
  • okiedrifter, Robby9newbie and Marco 104 like this

#4 bobzeq25

bobzeq25

    ISS

  • *****
  • Posts: 21,035
  • Joined: 27 Oct 2014

Posted 22 January 2020 - 10:55 AM

Hi all,

 

I'm working on Melotte15 and I'm checking intermediate integration results after each imaging night

 

I see that my integrations are way more noisy than they were weeks ago when I captured M42 (similar integration time), the only difference being that with the latter I did apply dithering on each frame, while with the former once every 3 frames and that Melotte15 is being done with SHO, instead of RBG

 

The camera is the ASI1600MM Pro, temp set at -20°C, gain/offset 139/21

 

To me dithering every 3 frames should already be high frequency, what am I doing wrong? Is this all about narrowband?

 

Thanks

Marco

A key thing is sub length and how many you shoot.  With a C8 RASA, F2, in Bortle 7, I shoot hundreds of subs at 10-20 seconds.  Example, 662X10".  I've yet to find someone willing to suggest how often I should dither.

 

I'm thinking dithers per X time might be a more useful measure.  If someone dithers every frame at 120", maybe I should dither every 12 frames?


Edited by bobzeq25, 22 January 2020 - 10:56 AM.

  • Miguelo, Thundermoon1994 and Marco 104 like this

#5 OhmEye

OhmEye

    Viking 1

  • *****
  • Posts: 781
  • Joined: 15 Sep 2019
  • Loc: Western NY Southern Tier

Posted 22 January 2020 - 11:46 AM

It boils down to what percentage of imaging time you are willing to give up for dithering. I'm fine with dithering every sub when my subs are 120s or longer. Under 2m subs I might dither every other frame. Dithering takes about 20s on my rig, so it's not hard to keep dithering time to 10% or less.
  • Marco 104 and NuclearRoy like this

#6 nimitz69

nimitz69

    Mercury-Atlas

  • *****
  • Posts: 2,611
  • Joined: 21 Apr 2017
  • Loc: A barrier island 18 miles south of Cocoa Beach

Posted 22 January 2020 - 12:03 PM

Dither or die as someone once said .......

 

I dither every frame ... it doesn’t cost you much of anything and makes a huge difference ... control the little things


  • Marco 104 likes this

#7 bobzeq25

bobzeq25

    ISS

  • *****
  • Posts: 21,035
  • Joined: 27 Oct 2014

Posted 22 January 2020 - 12:07 PM

Dither or die as someone once said .......

 

I dither every frame ... it doesn’t cost you much of anything and makes a huge difference ... control the little things

It costs you a great deal with 662 subs.  <grin>


  • Miguelo likes this

#8 RaulTheRat

RaulTheRat

    Viking 1

  • -----
  • Posts: 678
  • Joined: 09 Oct 2018

Posted 22 January 2020 - 12:26 PM

A key thing is sub length and how many you shoot. With a C8 RASA, F2, in Bortle 7, I shoot hundreds of subs at 10-20 seconds. Example, 662X10". I've yet to find someone willing to suggest how often I should dither.

I'm thinking dithers per X time might be a more useful measure. If someone dithers every frame at 120", maybe I should dither every 12 frames?


This isn't really a suggestion so much as it is a mulling over of an idea, but I think, for example comparing f/4 to f/2 you would take subs a quarter the length for each sub to be equally exposed, and you'd need to take the same number of subs as the f/4 system (requiring a quarter of the total integration time) to reach the same SNR in the stack.

So with the same number of subs, you'd need to dither the same number of times to get the same suppression of pattern noise. You'd be dithering 4x as often going by the clock, but at the same frequency going by the number of subs.

Thinking about two other cases, say you decided to keep sub lengths the same (maybe not if hitting saturation, but perhaps for narrowband you might) so you got deeper subs with the f/2 system, you'd only need a quarter as many of them for the same total integration time so then I think you might not be able to dither enough - the f/4 system might have 60 5 minute subs in 5 hours with a dither every frame, and you'd have 15 5 minute subs in 1.25 hours and couldn't dither more than 15 times.

Conversely assume you kept the per-sub ADU count similar by exposing a quarter as long per sub, but just used the whole 5 hours and got a deeper total integration, now you'd have 240 subs of 1.25 minutes each, and I think then you could dither every 4 subs to have the same total dithers in the stack (and same dithers per hour of clock time) with no increase in pattern noise compared with the f/4 stack.

Again, nothing particularly scientific about that, but something to think about and possibly refute or tell me why it ain't so.

#9 jhayes_tucson

jhayes_tucson

    Fly Me to the Moon

  • *****
  • Posts: 7,388
  • Joined: 26 Aug 2012
  • Loc: Bend, OR

Posted 22 January 2020 - 01:13 PM

Hi all,

 

I'm working on Melotte15 and I'm checking intermediate integration results after each imaging night

 

I see that my integrations are way more noisy than they were weeks ago when I captured M42 (similar integration time), the only difference being that with the latter I did apply dithering on each frame, while with the former once every 3 frames and that Melotte15 is being done with SHO, instead of RBG

 

The camera is the ASI1600MM Pro, temp set at -20°C, gain/offset 139/21

 

To me dithering every 3 frames should already be high frequency, what am I doing wrong? Is this all about narrowband?

 

Thanks

Marco

 

Marco,

First off, the reason that your "integrations are way more noisy than they were weeks ago when I captured M42" is because of photon noise--not because of the lack of dithering.  M42 is very bright compared to Molette 15.  With so much talk about sources of noise in sensors around here, it's easy to forget that light itself has its own inherent noise.  This is due to the quantum nature of light, which is why light arrives in discrete particles called photons (as I'm sure you know.)  The probability of an arrival per unit area per time is governed by Poisson statistics, which shows that the signal grows linearly with time (as N photons) and noise grows as SQRT(N) so that the S/N ratio grows by the SQRT(N).  Since Molette 15 is so faint relative to M42, your image will appear to be a LOT more "noisy" (which is what the lower SNR causes you to see) using the same integration time.  In Astrophotography, we are able to directly observe effects due to the quantum nature of light--and that's what you are seeing here.

 

Second, I dither every single frame, but my exposures are always 20 minutes.  In my view, the ratio of the amount of time spent dithering to the amount of time spent gathering data should be kept below about 15% of the total integration time.  Be careful to understand that I am most definitely not saying to dither every ~9th frame!  I am saying that if your dithering routine takes say 30 seconds (including the dither, a pause to dissipate vibration and to allow guiding to stabilize,) then you should dither every ~200 seconds or so.  And if your exposures are longer than that, then you should consider dithering every frame.  This assumes that the subs are being properly calibrated to remove dark signal and flat fielding to minimize the effects of PRNU modulation (which produces FPN.) The whole point of dithering is to spatially average the nose added to the image by the calibration process--not to simply smear out FPN or dark signal.

 

John


Edited by jhayes_tucson, 22 January 2020 - 11:54 PM.

  • Miguelo and B 26354 like this

#10 iwannabswiss

iwannabswiss

    Apollo

  • *****
  • Posts: 1,296
  • Joined: 14 Feb 2014
  • Loc: Somewhere in Space

Posted 22 January 2020 - 02:49 PM

I dither after every frame


  • Marco 104 likes this

#11 Der_Pit

Der_Pit

    Surveyor 1

  • *****
  • Posts: 1,576
  • Joined: 07 Jul 2018
  • Loc: La Palma

Posted 22 January 2020 - 03:55 PM

A key thing is sub length and how many you shoot.  With a C8 RASA, F2, in Bortle 7, I shoot hundreds of subs at 10-20 seconds.  Example, 662X10".  I've yet to find someone willing to suggest how often I should dither.

 

I'm thinking dithers per X time might be a more useful measure.  If someone dithers every frame at 120", maybe I should dither every 12 frames?

I've settled to wanting at least something like 50-60 different dither positions, to properly handle the fixed pattern noise.  So with 662 frames (for one wavelength, or OSC, resp.) a dither every 10-12 would IMO still be OK.  The other constraint is dither overhead - I don't want that to exceed 20-25% (like 2.5h wall-clock for 2h of exposure time).  That of course depends on how fast the mount settles after dither....


  • sharkmelley, bobzeq25 and Marco 104 like this

#12 sharkmelley

sharkmelley

    Soyuz

  • *****
  • Posts: 3,555
  • Joined: 19 Feb 2013
  • Loc: UK

Posted 22 January 2020 - 05:49 PM

I've settled to wanting at least something like 50-60 different dither positions, to properly handle the fixed pattern noise. 

Agreed.  For me 30 is an absolute minimum number of dithers during an imaging run.  60 is preferable.  You definitely need a good number for the sigma rejection to work well.  For instance PixInsight's Winsorized sigma clipping works well for 20 different subs but any subs without an intervening dither need to be considered as pretty much identical from the point of view of sigma rejection because their fixed pattern noise will line up.

 

Mark


Edited by sharkmelley, 22 January 2020 - 05:52 PM.

  • freestar8n, ks__observer and Marco 104 like this

#13 kyle528

kyle528

    Viking 1

  • *****
  • Posts: 704
  • Joined: 02 Feb 2019
  • Loc: SE Ohio

Posted 22 January 2020 - 06:08 PM

Some nights with my guiding it's like I'm dithering every half a sub lol.gif


  • Fernando134, wavo, joshman and 1 other like this

#14 kathyastro

kathyastro

    Vanguard

  • *****
  • Posts: 2,147
  • Joined: 23 Dec 2016
  • Loc: Nova Scotia

Posted 22 January 2020 - 06:10 PM

I dither every frame.  My exposure times range from 120" (RGB) to 600" (Ha)


  • Marco 104 likes this

#15 RaulTheRat

RaulTheRat

    Viking 1

  • -----
  • Posts: 678
  • Joined: 09 Oct 2018

Posted 22 January 2020 - 06:20 PM

Some nights with my guiding it's like I'm dithering every half a sub lol.gif


Finally we have a definitive answer. This is how often you dither.
  • Fernando134 likes this

#16 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 9,764
  • Joined: 12 Oct 2007

Posted 22 January 2020 - 06:21 PM

Dithering is similar to subexposure time: more is always better except when other factors dominate.  The main factor with dithering is the loss of sky exposure time due to dead time between exposures.

 

I mainly use 5m exposures and the time between exposures with download, filter change and dither is about 10 to 20 seconds - so I dither every exposure.  I could reduce the time by reducing the time allowed to settle the guiding and make sure it is steady, but it is about 5 seconds as it is.  I am guiding with video so I have a live view of the guide error and I check it every second for settling.  For people who use 5-15 second guide exposures I assume that would greatly extend the time needed to check for settling.

 

For the OP - the difference between RGB of M42 and narrowband of a relatively faint nebula is huge.  You don't state the exposure times, but I would use much shorter exposures for M42 than for Melotte15.  The pattern noise will be much more visible with narrowband since there is so much less signal and the pattern noise stands out in the background, particularly if you stretch.  You are probably fine dithering every 3rd frame as long as you have many frames and the dither pattern is large enough to randomize the pattern noise.  If the dither pattern isn't large enough then even dithering every frame won't help much.

 

So, it isn't so much the frequency of dithering as the total number of dithers in the stack - and how big the dither pattern is.  And the only penalty for dithering too much is loss of sky imaging time - which won't be too bad if the exposures are long and the dither recovery is quick.

 

Frank


  • ks__observer and Marco 104 like this

#17 Stelios

Stelios

    Cosmos

  • *****
  • Moderators
  • Posts: 9,939
  • Joined: 04 Oct 2003
  • Loc: West Hills, CA

Posted 22 January 2020 - 10:09 PM

I dither every 3 frames with exposures of 90" or less (I take at least 100), every 2 frames with exposures of 120" to 180" and every frame with exposures 240" and up. 

 

If I wasn't lazy I would test whether it's better to small-dither every frame than medium-dither every 3 frames. 


  • Marco 104 likes this

#18 Cfreerksen

Cfreerksen

    Apollo

  • *****
  • Posts: 1,119
  • Joined: 18 Aug 2018
  • Loc: Tooele, Ut

Posted 22 January 2020 - 11:42 PM

Ok, maybe on my narrow-band stuff I should be dithering more (every 3rd frame now). I am shooting 7 minute subs right now. Thinking on going to 9min (bortle 4+ish) But how much do you dither? I am doing 20 pixel random. My sampling is 0.53 Arc-seconds/pixel. Am I being too aggressive? I have my minimum settle time set to 30 seconds. Dither & focus eats a lot of time. Focusing every 0.75C and every 6-7 frames narrow-band. 6 steps each side of focus, 15 sec exposures.

 

Recommendations please.

 

Chris


Edited by Cfreerksen, 22 January 2020 - 11:56 PM.

  • Marco 104 likes this

#19 jhayes_tucson

jhayes_tucson

    Fly Me to the Moon

  • *****
  • Posts: 7,388
  • Joined: 26 Aug 2012
  • Loc: Bend, OR

Posted 23 January 2020 - 12:04 AM

I dither every 3 frames with exposures of 90" or less (I take at least 100), every 2 frames with exposures of 120" to 180" and every frame with exposures 240" and up. 

 

If I wasn't lazy I would test whether it's better to small-dither every frame than medium-dither every 3 frames. 

The dither distance should be greater than the autocorrelation length of the FPN in your image.  Precisely determining that distance can be tricky so if you have to guess, it's better to dither at a "moderately large" distance.  If it's too big, it won't hurt anything but if it's too small, you won't get much, if any, benefit from dithering.  On my system, I have the dither parameter set to 9 pixels and it seems to work pretty well with my sensor (which has 9 micron pixels.)

 

John


  • ks__observer likes this

#20 Tonk

Tonk

    Cosmos

  • *****
  • Posts: 9,278
  • Joined: 19 Aug 2004
  • Loc: Leeds, UK, 54N

Posted 23 January 2020 - 12:06 AM

I never dither - I found it had no noticeable improvement over regular calibration and thus just wastes time.


  • Miguelo and Marco 104 like this

#21 jhayes_tucson

jhayes_tucson

    Fly Me to the Moon

  • *****
  • Posts: 7,388
  • Joined: 26 Aug 2012
  • Loc: Bend, OR

Posted 23 January 2020 - 12:10 AM

Ok, maybe on my narrow-band stuff I should be dithering more (every 3rd frame now). I am shooting 7 minute subs right now. Thinking on going to 9min (bortle 4+ish) But how much do you dither? I am doing 20 pixel random. My sampling is 0.53 Arc-seconds/pixel. Am I being too aggressive? I have my minimum settle time set to 30 seconds. Dither & focus eats a lot of time. Focusing every 0.75C and every 6-7 frames narrow-band. 6 steps each side of focus, 15 sec exposures.

 

Recommendations please.

 

Chris

 

My guess is that dithering 15 - 20 pixels is probably in the right ballpark for the ASI-1600MM.  With 7 minute subs, I'd recommend dithering every frame.  V-curve focusing is the really big time waster.  Are you using filter offsets to speed up the data acquisition time?  That's where real-time focusing provides a very noticeable advantage--not to mention that it's essential for any scope with significant temperature sensitivity.

 

John


Edited by jhayes_tucson, 23 January 2020 - 12:11 AM.

  • Cfreerksen likes this

#22 schmeah

schmeah

    Fly Me to the Moon

  • *****
  • Posts: 5,924
  • Joined: 26 Jul 2005
  • Loc: Morristown, NJ

Posted 23 January 2020 - 06:17 AM

If you are rotating through your filters, for example L-R-G-B, then dithering every fourth is effectively dithering every sub, though most people shoot narrowband in groups, not rotating. So if grouped and shooting short subs (ie 3 minutes or less) I’ll dither every other, every one for long subs. Dithering for me in APT takes only about 5-10 seconds.

 

Derek


  • Ken Sturrock likes this

#23 Coconuts

Coconuts

    Viking 1

  • *****
  • Posts: 575
  • Joined: 23 Sep 2012

Posted 23 January 2020 - 07:52 PM

John:  "If it's too big, it won't hurt anything".  I'm hearing that more than a full field can lead to issues during stacking smile.gif

 

All the best,

 

Kevin


  • RossW likes this

#24 jhayes_tucson

jhayes_tucson

    Fly Me to the Moon

  • *****
  • Posts: 7,388
  • Joined: 26 Aug 2012
  • Loc: Bend, OR

Posted 23 January 2020 - 07:57 PM

John:  "If it's too big, it won't hurt anything".  I'm hearing that more than a full field can lead to issues during stacking smile.gif

 

All the best,

 

Kevin

 

lol.gif lol.gif lol.gif

Touche.  Yeah, if it's more than full field, there might be a problem.

 

John



#25 Alex McConahay

Alex McConahay

    Cosmos

  • *****
  • Posts: 9,419
  • Joined: 11 Aug 2008
  • Loc: Moreno Valley, CA

Posted 23 January 2020 - 11:11 PM

As was pointed out, the nut of the matter is dithering often enough so that you go to stacking you have enough different aiming positions that your rejection algorithm kicks in. The more, the better, although there are diminishing returns after a point. That point differs from algorithm to algorithm.

As noted, it is good to dithervevery time, unless it is cutting into exposure time too much. Maybe what needs to be addressed is the time it takes to dither. If you are taking to long to settle and start the next exposure maybe you need to look at mount mechanics, settling settings, and so forth. For instance, trying to settle at about the level of seeing or for longer than one or two guiding cycles may be a slow process but no better in practice than a slightly higher settling.

Alex


CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics