How do you battle with LP gradients

[souls33k3r]

Hi All,

Just trying to pick brains here and maybe help some who might be going through the same issue that i currently am.

What are your top tips for battling with LP gradients? 

I will be found using NB filters and process my image using Pixinsight.

My camera is the ASI1600MM shooting at 5m subs.

Thank you

 

[RolandKol]

I afraid, you will get advice to go even narrower... 3nm Astrodons... and etc

[souls33k3r]

43 minutes ago, RolandKol said:

I afraid, you will get advice to go even narrower... 3nm Astrodons... and etc

No chance, got to make do with what I've currently got which is Astronomik 6nm filters.

[Allinthehead]

I have issues where i live depending on the area of the sky i'm pointing and i found reducing my sub length in rgb helped. I never really had much of an issue with Ha, but found the Oiii to have large gradients. If i were you i would experiment with either shorter subs or a lower gain when shooting Oiii and Sii.

[DaveS]

I have to use multiple iterations of gradient removal in AstroArt 5, sometimes it's a matter of try it and see.

Oh yes, and 3nm Astrodons.

[souls33k3r]

16 hours ago, Allinthehead said:

I have issues where i live depending on the area of the sky i'm pointing and i found reducing my sub length in rgb helped. I never really had much of an issue with Ha, but found the Oiii to have large gradients. If i were you i would experiment with either shorter subs or a lower gain when shooting Oiii and Sii.

I don't want to go down shorter subs if i'm being honest and it's purely because of the file size per sub. Currently i'm on unity gain. Do you think i should go "Highest dynamic range" (Gain 0 and Offset 10) or should i try "Low Read Noise" (Gain 300 Offset 50)?

16 hours ago, DaveS said:

I have to use multiple iterations of gradient removal in AstroArt 5, sometimes it's a matter of try it and see.

Oh yes, and 3nm Astrodons.

It really is a pain, doing multiple iterations of gradient removal can't be the only fix other than going 3nm (which if i'm being honest is super out of the question due to the price point).

What about introducing a LP filter in front of the FW? I might sound daft by saying this but surely there are plenty of imagers out there in the wild who live in LP areas or Red zones in US who still take cracking images.

[Allinthehead]

1 minute ago, souls33k3r said:

I don't want to go down shorter subs if i'm being honest and it's purely because of the file size per sub. Currently i'm on unity gain. Do you think i should go "Highest dynamic range" (Gain 0 and Offset 10) or should i try "Low Read Noise" (Gain 300 Offset 50)?

You'll have to experiment, but i found lowering the gain when using an Oiii filter to help with gradients. So you could stick with 300 second subs but with 0 gain. Start from there and raise if necessary.

[souls33k3r]

20 minutes ago, Allinthehead said:

You'll have to experiment, but i found lowering the gain when using an Oiii filter to help with gradients. So you could stick with 300 second subs but with 0 gain. Start from there and raise if necessary.

I can definitely try that. This then means a new dark library for OIII  

[Allinthehead]

11 minutes ago, souls33k3r said:

I can definitely try that. This then means a new dark library for OIII  

Yes and don't neglect the flats.

[souls33k3r]

1 minute ago, Allinthehead said:

Yes and don't neglect the flats.

I'm struggling to understand how can flats help with LP gradients? aren't flats meant to only deal with the imperfections of the imaging train (Dust motes, vignetting)?

[Allinthehead]

Flats provide even illumination across the field and bad flats or none at all can show up as gradients. You should always use flats.

No Flat

Flat

Both images auto stretched to show the difference. Granted this is an extreme example as i had an issue in my imaging train but you get the idea.

[souls33k3r]

59 minutes ago, Allinthehead said:

Flats provide even illumination across the field and bad flats or none at all can show up as gradients. You should always use flats.

No Flat

Flat

Both images auto stretched to show the difference. Granted this is an extreme example as i had an issue in my imaging train but you get the idea.

Ah yes, this makes much more sense. Cheers for the share mate. 

[AstroAndy]

Hi

Flats are the way to go, sometimes, I create a synthetic flat in PS by duplicating the image, using the clone tool to clone out the brighter parts (or the dust and scratches filter, then the clone tool), darkening it a bit (to keep all the details of the object), then applying it to to the main img., and playing with the settings to get an even field; works for lum., and color. Disregarding the noise ?, see below example (although this had regular flats applied).

Andy

[ollypenrice]

Your LP must be pretty bad if it's still a big issue in NB. First I'd want to be sure that your flats are good and not introducing gradients of their own. OK, next, let's assume the flats are, indeed, OK.

You have PI so you have the best software flattening tool in the business, DBE. I follow Rogelio Bernal Andreo, Harry Page and other PI gurus in thinking that fewer and better background markers work best. You are trying to find real background sky rather than gradient so it can be worth looking at a dark sky image of your target to help you identify where such background exists in your framing, even if, in your data, a patch of background is bright. You then need to adjust the tolerance to be sure that your markers on bright regions have been accepted. Always look at the gradient model, too. It should not, in any way, resemble the nebulosity in the image. It should resemble the gradient. Keep this firmly in mind.

I'm talking to Harry on here about something else at the moment so perhaps he'll be along.

Olly

[harry page]

9 minutes ago, ollypenrice said:

Your LP must be pretty bad if it's still a big issue in NB. First I'd want to be sure that your flats are good and not introducing gradients of their own. OK, next, let's assume the flats are, indeed, OK.

You have PI so you have the best software flattening tool in the business, DBE. I follow Rogelio Bernal Andreo, Harry Page and other PI gurus in thinking that fewer and better background markers work best. You are trying to find real background sky rather than gradient so it can be worth looking at a dark sky image of your target to help you identify where such background exists in your framing, even if, in your data, a patch of background is bright. You then need to adjust the tolerance to be sure that your markers on bright regions have been accepted. Always look at the gradient model, too. It should not, in any way, resemble the nebulosity in the image. It should resemble the gradient. Keep this firmly in mind.

I'm talking to Harry on here about something else at the moment so perhaps he'll be along.

Olly

Hi

you called

as you will find out , full calibration of your images are a must if you are going to get the best from your images - flats -bias and darks if you need them  , true flats are the only reliable way of rmoving

dust bunnies as well.

This will give you a good starting point , leaving only LP gradients to deal with.

The mighty Olly has given good advice on the use of DBE - I regularly say its only with a evenly illumined image  ( for most after using DBE) can you extract the most from your image

And if there is 1 tool in Pixinsight you learn well its DBE-- everything starts there , get it wrong and you are fighting a loosing battle for the rest off your processing

All hail the might DBE

Harry

[souls33k3r]

Excellent advice @AstroAndy, @harry page & @ollypenrice. This is truly very helpful. I guess time to go back to the drawing board and learn the wizardry of DBE properly

DBE is a beast and like Harry mentioned that get it wrong and i'll be fighting a losing battle. I don't want to apply DBE but just putting in the numbers and hope for the best but like to learn what numbers to put and why to put them. 

Still slightly confused about when to apply subtraction and division. If someone can explain that in layman terms, that would be great. Trying to wrap my head around the additive and multiplicative gradients is a bit of a mind numbing thought process.

Much appreciated everyone

[ollypenrice]

1 hour ago, souls33k3r said:

Excellent advice @AstroAndy, @harry page & @ollypenrice. This is truly very helpful. I guess time to go back to the drawing board and learn the wizardry of DBE properly

DBE is a beast and like Harry mentioned that get it wrong and i'll be fighting a losing battle. I don't want to apply DBE but just putting in the numbers and hope for the best but like to learn what numbers to put and why to put them. 

Still slightly confused about when to apply subtraction and division. If someone can explain that in layman terms, that would be great. Trying to wrap my head around the additive and multiplicative gradients is a bit of a mind numbing thought process.

Much appreciated everyone

I'm content to take the maths experts at their word when they tell me that vignetting is a multiplicative process and light pollution an additive one, so when I apply flats to correct vignetting I use division and when I use a gradient map created in DBE I subtract it. Since I have already applied flats before going into DBE I'm left only with LP gradients. I make the gradient model in DBE and subtract it.

I dare say that, if you linked to one of your stacks, members could screen grab the process they would use in DBE to remove gradients.

Olly

[Scooot]

I have spent much more time trying to understand the DBE process this last week. In particular using fewer placement points (with a larger size if possible) in hopefully smarter positions. Harry’s tutorial & others are very helpful. I still struggle to follow how near to a gradient change I need to place the points but I’m getting there. I have found that doing a “boosted auto stretch” can help me to view the gradient more clearly. 

[harry page]

23 minutes ago, Scooot said:

I have spent much more time trying to understand the DBE process this last week. In particular using fewer placement points (with a larger size if possible) in hopefully smarter positions. Harry’s tutorial & others are very helpful. I still struggle to follow how near to a gradient change I need to place the points but I’m getting there. I have found that doing a “boosted auto stretch” can help me to view the gradient more clearly. 

Hi

this is what I do with a difficult target and off course with the STF active as required

1)  place samples on the image in as good as place as you can i.e best guess if need be

2) apply this image and create a new image ( say temp 1 ) , usually a decent correction has been made even if not perfect

3) drag dbe icon to desktop to create new process ( process 1) ( this will have the dbe settings saved )

4) close DBE on your original image

5) start up the saved process 1 DBE on the desktop and apply to the temp image

6) on the temp image refine and or add new samples as req ( quality not quantity ), which is hopefully easier with most of the errors removed

7) now save as before this process to the desktop again ( process 2 )

8 ) close DBE again and discard Temp 1 image

9) open process 2 dbe

10 ) apply to your original uncorrected image ( this will place samples in your refined way )

11) check samples have the correct tolerance ( i.e not red )

 

Then apply away as normal ( subtraction for LP )  hopefully with a perfect result , but on a very difficult image 3 or 4 try's may be required before you get it right

 

Regards

Harry

 

 

[souls33k3r]

1 hour ago, harry page said:

Hi

this is what I do with a difficult target and off course with the STF active as required

1)  place samples on the image in as good as place as you can i.e best guess if need be

2) apply this image and create a new image ( say temp 1 ) , usually a decent correction has been made even if not perfect

3) drag dbe icon to desktop to create new process ( process 1) ( this will have the dbe settings saved )

4) close DBE on your original image

5) start up the saved process 1 DBE on the desktop and apply to the temp image

6) on the temp image refine and or add new samples as req ( quality not quantity ), which is hopefully easier with most of the errors removed

7) now save as before this process to the desktop again ( process 2 )

8 ) close DBE again and discard Temp 1 image

9) open process 2 dbe

10 ) apply to your original uncorrected image ( this will place samples in your refined way )

11) check samples have the correct tolerance ( i.e not red )

 

Then apply away as normal ( subtraction for LP )  hopefully with a perfect result , but on a very difficult image 3 or 4 try's may be required before you get it right

 

Regards

Harry

 

 

Excellent short tutorial on this Harry. Much appreciated mate.

If i'm not mistaken, i believe i saw Adam Block use the same technique in one of his free videos.

Regardless of which, i am definitely going to be doing that. Love this software. 

4 hours ago, ollypenrice said:

I'm content to take the maths experts at their word when they tell me that vignetting is a multiplicative process and light pollution an additive one, so when I apply flats to correct vignetting I use division and when I use a gradient map created in DBE I subtract it. Since I have already applied flats before going into DBE I'm left only with LP gradients. I make the gradient model in DBE and subtract it.

I dare say that, if you linked to one of your stacks, members could screen grab the process they would use in DBE to remove gradients.

Olly

Cheers Olly. Unfortunately (i'm going to have the courage to say this) i have been avoiding flats up until now but i will start to use it.

Now saying that, i do not have vignetting issue. Do i really need to use flats or will it help even out the image in order to process the data for LP gradients?

I live in Bortle 8/9 (Red/White) zone which obviously doesn't help much either.

[harry page]

Hi

I am sure Mr Block has never watched any of my vids  , but I had a look at some of his recently and I am not a huge fan and certainly not aimed at newbies

I would recommend me ?or Warren kellers first  

Unless you have a small chip , every image I have seen has vignetting  and only flats will rid you of dust bunnies - so just get used to them ?

I remember a certain Olly saying he does not have light pollution or gradients to deal with ?

Regards

Harry

[ollypenrice]

1 hour ago, harry page said:

Hi

I remember a certain Olly saying he does not have light pollution or gradients to deal with ?

Regards

Harry

I don't think you do, Harry! I don't have significant LP, certainly, but I've always been intrigued by the fact that I still get colour gradients. I always run my RGB through DBE. I may have used it once or twice on Ha but that is almost invariably gradient free and I rarely use it on luminance, though sometimes I do. But colour - always.

Where does the colour gradient come from, though? When I used OSC it seemed to take the form of a red-green imbalance either side of a diagonal across the chip, however the chip was orientated. Beats me, but I can always nail it with DBE. It also gives a good colour balance at the higher brightnesses, I find.

Some folks use DBE first on the individual colour channels. So far I've never done this but I keep meaning to try. What's your view?

Olly

PS Flats are essential, as you say. I resisted in the early days with a small chip but this was just a process of denial! You need them.

[harry page]

2 minutes ago, ollypenrice said:

I don't think you do, Harry! I don't have significant LP, certainly, but I've always been intrigued by the fact that I still get colour gradients. I always run my RGB through DBE. I may have used it once or twice on Ha but that is almost invariably gradient free and I rarely use it on luminance, though sometimes I do. But colour - always.

Where does the colour gradient come from, though? When I used OSC it seemed to take the form of a red-green imbalance either side of a diagonal across the chip, however the chip was orientated. Beats me, but I can always nail it with DBE. It also gives a good colour balance at the higher brightnesses, I find.

Some folks use DBE first on the individual colour channels. So far I've never done this but I keep meaning to try. What's your view?

Olly

Hi

Ok I yield on your LP free skies ? , But every Image I have ever been sent has a gradient ( exc some narrow band ) the causes are many and will take up a whole thread on its own

I do not do individual DBE on the RGB as it gives no colour balance and a no neutral background  so I do not see the point , infact I have tried it and the results are always inferior

this may be due to the heavy light pollution I have to deal with , most of the time My RGB master is a real shocker and without DBE I could not image broadband from my location

Regards from my fully light polluted shed

Harry

[souls33k3r]

3 hours ago, harry page said:

Hi

I am sure Mr Block has never watched any of my vids  , but I had a look at some of his recently and I am not a huge fan and certainly not aimed at newbies

I would recommend me ?or Warren kellers first  

Unless you have a small chip , every image I have seen has vignetting  and only flats will rid you of dust bunnies - so just get used to them ?

I remember a certain Olly saying he does not have light pollution or gradients to deal with ?

Regards

Harry

I totally agree with you. His videos are most certainly not aimed at newbies. Plus i'm never going to be able to justify the cost of the subscription. Which brings me nicely to your page Harry. Your tutorials are an excellent resource. You have done well for the likes of me mate

I use the ASI1600MM ... when i was choosing my filters, i was told that because the filters sit quite close to the sensor the 1.25" ones are going to be good enough up until F4 scope. I still went ahead and bought the 31mm version filters just to be on the safe side. I must admit, i've never noticed any vignetting on my images. I might be wrong but that's what my untrained eye has told me.

2 hours ago, ollypenrice said:

I don't think you do, Harry! I don't have significant LP, certainly, but I've always been intrigued by the fact that I still get colour gradients. I always run my RGB through DBE. I may have used it once or twice on Ha but that is almost invariably gradient free and I rarely use it on luminance, though sometimes I do. But colour - always.

Where does the colour gradient come from, though? When I used OSC it seemed to take the form of a red-green imbalance either side of a diagonal across the chip, however the chip was orientated. Beats me, but I can always nail it with DBE. It also gives a good colour balance at the higher brightnesses, I find.

Some folks use DBE first on the individual colour channels. So far I've never done this but I keep meaning to try. What's your view?

Olly

PS Flats are essential, as you say. I resisted in the early days with a small chip but this was just a process of denial! You need them.

Point taken Olly. I know i need to do flats, i've just been too lazy to take them and it's mostly because the amount of faffing about doing them is just too much. 

Pardon my naivety but how does one differentiate a colour gradient from a LP gradient? Do colour gradients exist in when using mono camera images?

 

[kirkster501]

What exposure times are your flats?  If they are too short you will get a gradient in the flats caused by the shutter.  Then you are *really* fighting a losing battle from the start.  I had this issue when my flats were very short exposures and had to dim my flats light panel to get longer exposed flats to reduce the effect of the rolling shutter.  

Just to add to the processing discussion, I am a big advocate of Adam Block's and I have subscribed to his "Fundamentals" video series and realise, after watching a few, that I didn't even know what I didn't know.....   Without doubt, he is a wizard in PixInsight.  He goes into great detail not only *what* to do but also *why* you need to do it and, sure at times this can  be quite involved.  However, I think processing of astro images *is* so involved, after all the effort to get the data, that a grounds up understanding of exactly what is going on from first principles - sometimes at the pixel level - is very important and what all the settings, buttons and knobs do is critical to understand.  I think you really need to sit down and invest the time and energy to *really* learn and master PixInsight.   I think I have made the mistake of spending too much time and energy (and money) in the scopes, cameras, mounts and driving them to get reasonably good subs and not enough time to really learn the processing of the acquired data.  I mean, *really* learn it. I've been using PI for four years without much success and straight away learned loads on such basics on how the windows and views work from Adam in his very first video in the series!  He spends nearly 40 minutes on this one subject alone and there is a ton of stuff I never knew about that is incredibly useful.  I have a lot of data sets that I am going to go through with Adam's series and see how I can get better.

I think it's better to be good at processing half decent data sets as opposed to having better subs and being rubbish at processing them (like I am).  I'm really trying hard to get better at this and Adam, Harry and Warhen's resources are incredibly helpful.

 Not sure if I was answering the question or if I went off on a monologue.....!  But my two penneth FWIW!!

Steve