ICEINSPACE
Moon Phase
CURRENT MOON
Waning Gibbous 88.4%
|
|
08-11-2013, 07:00 PM
|
|
Registered User
|
|
Join Date: Feb 2006
Location: Sydney
Posts: 17,996
|
|
How do you do flats?
I was wanting to start a thread on this as its often a tricky point of processing that can impact on the quality of your image heavily.
Its probably more vital with a long focal length scope but also vital when you image in light polluted areas.
Richard Crisp has done a paper on it. I will try to find the link and post it here.
So what is your best procedure for doing flats - both DSLRs and CCD cameras?
Greg.
|
08-11-2013, 07:25 PM
|
|
Registered User
|
|
Join Date: May 2005
Location: Brisbane
Posts: 4,485
|
|
Use SGPro to calculate an average ADU value of 35,000 for each filter and telescope for flats. completely reproducable and no problems encountered even in light polluted Brisvegas.
Allan
|
08-11-2013, 07:45 PM
|
|
PI cult recruiter
|
|
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
|
|
I do dusk and dawn flats with ACP automation by preference. If that's not an option I use a Gerd Neumann EL panel. I like to get a minimum of around 500K e- total for each filter (about 12 flats at 30K ADU with my U16M camera but it would be a lot more for a camera with shallower wells). The 500K e- number is based on the work by Richard Crisp. Not sure if this the paper you were looking for, Greg?
http://www.narrowbandimaging.com/inc...3_expanded.pdf
Cheers,
Rick.
|
08-11-2013, 08:05 PM
|
Atlas Observatory
|
|
Join Date: May 2012
Location: Canberra
Posts: 268
|
|
Never really done them but with the cdk I've found I now need to. I have a flatman XL on the way that should have been here before the scope but there were delays due to shortages from their inverter supplier.
Will be interesting to see how this thread goes as I have no real idea about this flat stuff. Anyone else using a flatman?
|
08-11-2013, 08:48 PM
|
|
Colour is over-rated
|
|
Join Date: Sep 2005
Location: Newcastle, Australia
Posts: 2,414
|
|
I made a simple cardboard frame for an EL panel, and use SG Pro to sort out exposure times for each filter/binning.... I have 10 sheets of tracing paper in front of the panel to dim it enough....
|
08-11-2013, 10:01 PM
|
Registered User
|
|
Join Date: Nov 2011
Location: Sydney
Posts: 1,454
|
|
Hi Greg
For DSLR, I originally used a light box from Icer EXFSO and used BYE to get the histogram to around 70% from the left hand side. Worked very well.
Am now using Gerd Neumann light panels, one for the frac and one for the RC8, (I may have over capitalised...) and find I need longer subs but aiming for the mid point on the histogram, again through BYE. Still playing but seems to work ok.
When I grow up and get a ccd then will have to change again. Great fun
I checked in Nebulosity and the flats that work for me give around 20,000 to 25,00 adu
Good thread..
|
08-11-2013, 11:22 PM
|
|
Really just a beginner
|
|
Join Date: Jun 2009
Location: Brisbane
Posts: 3,040
|
|
A3 EL panel mounted between 2 sheets of MDF. OZStockman supplied the MDF - one sheet is complete, the other has a hole to sit around the front aperture of the scope and holds the panel in place. It was actually the offcut from around a Bahtinov mask that he cut for me. I just asked him to cut it from an appropriately sized piece of MDF to fit over the A3 EL panel. I had already bought the EL panel so needed something to mount it in. In retrospect, I should have just bought a Gerd Neumann panel from the start. My overall cost was probably about half vs a GN panel, but a lot more fiddling to get into a usable state. The A3 panel is just large enough for a 10 or 11 inch scope.
Seems to work OK. I tell CCD Commander to acquire automatic flats, but not to move the mount or wait until dawn - it works out the exposures. Usually collect 10-20 flats per filter.
DT
|
09-11-2013, 06:30 AM
|
|
Registered User
|
|
Join Date: Feb 2006
Location: Sydney
Posts: 17,996
|
|
Quote:
Originally Posted by allan gould
Use SGPro to calculate an average ADU value of 35,000 for each filter and telescope for flats. completely reproducable and no problems encountered even in light polluted Brisvegas.
Allan
|
How many flats do you take? Do you subtract flat darks or use a bias?
Quote:
Originally Posted by RickS
I do dusk and dawn flats with ACP automation by preference. If that's not an option I use a Gerd Neumann EL panel. I like to get a minimum of around 500K e- total for each filter (about 12 flats at 30K ADU with my U16M camera but it would be a lot more for a camera with shallower wells). The 500K e- number is based on the work by Richard Crisp. Not sure if this the paper you were looking for, Greg?
http://www.narrowbandimaging.com/inc...3_expanded.pdf
Cheers,
Rick.
|
500ke is the midpoint he recommends for the 16803 and it varies with model of camera. Richard uses FLI cameras but having had a U16M myself they seem similar to the FLI in many ways. I don't understand the Ke value. The camera softwares do not report Ke. So how do you translate 500ke into an ADU? As I recall from various posts after this paper by Richard he recommend around 40K ADU. In the paper I notice he also states a bighter image needs a brighter flat. That is interesting. I have noticed sometimes I need a brighter flat to correct better.
Why do you say smaller well cameras need more subs for a master flat? I must have missed that implication in the paper. In reality if you take dusk/dawn flats the failing or brightening light puts a practical limit on how many subs you can take. That varies with camera performance (download times). I know Richard uses the video output of KAI sensors to get millisecond download times and gets many flats quickly.
Quote:
Originally Posted by SpaceNoob
Never really done them but with the cdk I've found I now need to. I have a flatman XL on the way that should have been here before the scope but there were delays due to shortages from their inverter supplier.
Will be interesting to see how this thread goes as I have no real idea about this flat stuff. Anyone else using a flatman?
|
If your CDK is similar to my CDK17 you will need flats big time or your images will suffer. Less so for a smaller chip. On the 16803 its vital hence the thread so we can get access to latest data about flats.
Quote:
Originally Posted by Lee
I made a simple cardboard frame for an EL panel, and use SG Pro to sort out exposure times for each filter/binning.... I have 10 sheets of tracing paper in front of the panel to dim it enough....
|
These panels are popular.My concern about panels (perhaps groundless) would be there spectrum bias. For example an LED torch looks very white at night. Take a long exposure photo of one and you'll see its quite purple in colour. That may impact on the accuracy of your flats??
Quote:
Originally Posted by DJT
Hi Greg
For DSLR, I originally used a light box from Icer EXFSO and used BYE to get the histogram to around 70% from the left hand side. Worked very well.
Am now using Gerd Neumann light panels, one for the frac and one for the RC8, (I may have over capitalised...) and find I need longer subs but aiming for the mid point on the histogram, again through BYE. Still playing but seems to work ok.
When I grow up and get a ccd then will have to change again. Great fun
I checked in Nebulosity and the flats that work for me give around 20,000 to 25,00 adu
Good thread..
|
Have you tried brighter flats like 40,000 ADU?
Quote:
Originally Posted by DavidTrap
A3 EL panel mounted between 2 sheets of MDF. OZStockman supplied the MDF - one sheet is complete, the other has a hole to sit around the front aperture of the scope and holds the panel in place. It was actually the offcut from around a Bahtinov mask that he cut for me. I just asked him to cut it from an appropriately sized piece of MDF to fit over the A3 EL panel. I had already bought the EL panel so needed something to mount it in. In retrospect, I should have just bought a Gerd Neumann panel from the start. My overall cost was probably about half vs a GN panel, but a lot more fiddling to get into a usable state. The A3 panel is just large enough for a 10 or 11 inch scope.
Seems to work OK. I tell CCD Commander to acquire automatic flats, but not to move the mount or wait until dawn - it works out the exposures. Usually collect 10-20 flats per filter.
DT
|
Does CCD Commander reduce the exposure times of the flats at dawn as they progress?
If you sit and watch you flat downloads (as I do) you'll see the ADU change between flats at dawn and dusk. I manually compensate by decreasing/increasing the exposure time to try to reach a target ADU reasonably closely. For example a 10 second flat that gives say 25,000ADU will gie 24,750 next sub then 24,500 next one etc. I am surprised often at how quickly the light fades at dusk. Its faster than we think I suppose because our eyes adjust to the failing light levels masking how quickly it is fading.
Martin Pugh mentioned at AAIC when I asked him about ADU flat levels he averages several sets of flats taken at different ADU. I have not done this but it sounded like a good idea so you are evening out the various reponses.
I knew about this 40K ADU by Richard yet I have encountered times when a bright flat damaged the image. So that confused me on this point. It may have been some other error that confused the result. As his paper says a more intense image requries brighter flats. This may have been a not very bright image. He seems to not mention a specific adu in his paper yet in posts he did. I will send him an email to clarify on this.
Greg.
|
09-11-2013, 09:47 AM
|
|
Really just a beginner
|
|
Join Date: Jun 2009
Location: Brisbane
Posts: 3,040
|
|
CCDC does change exposures during dusk/dawn flat runs.
DT
|
09-11-2013, 10:33 AM
|
|
Registered User
|
|
Join Date: Jan 2009
Location: Adelaide
Posts: 9,966
|
|
I have been using Peter's light boxes for quite a few year. I normally expose to get around 29,000. I have also used CCDAP to do dusk and dawn flats. I have found somewhere around 25,000 to 30,000 works well overall. Sometimes combining flats from 21,000 then 25000 and then 31000 can give a good flat too. Everything is about experimentation.
|
09-11-2013, 01:05 PM
|
|
Registered User
|
|
Join Date: Feb 2006
Location: Sydney
Posts: 17,996
|
|
Hi Richard,
Thanks for writing your paper on flats. Very infomative.
I remember various posts that were made after you published this paper. Your recommendation
for a typical Proline 16803 camera if I recall correctly was to go for about 40K ADU levels. Your
paper indicates around 5-10 subs should do (for an ML4022).
---I don't recall the ADU levels but I described how to determine the camera gain by making a photon transfer plot to character the main performance parameters of the camera. I always work in terms of electron units and to be able to convert from ADU to electrons you need to know the gain (units of e-/ADU)
One thing I didn't understand was the recommendation for 250Ke to 1000Ke. I use CCDSoft and the reported
values are in ADU. How do you convert Ke to ADU?
----1Ke- = 1000e-
ADU * gain (e-/ADU) = e-. When you multiply these two items together the ADU in the numerator cancels the ADU in the denominator leaving behind only e-. That, by the way, is called "dimensional analysis" and is taught in the first semester of the freshman year in engineering school which is where I learned it way back at Texas A&M in 1972 (dating myself).
---Last time I measured a Proline 16803 I found a gain of about 1.5e-/ADU. So for 40,000 ADU that means you have about 60,000e- or 60Ke-
I use a CDK17 often and it has bad vignetting and tends to have a hot spot in the illumination in the centre.
I have minor light pollution when imaging rising objects in the east. This results in hard to process images
and flats do not seem to get rid of all the vignetting but a lot of it.
What would you consider best practice for flats in this situation?
---Flats are easy in principle but difficult in practice. If you are imaging a rising object in the East and have Light Pollution, why not wait until it is higher and then begin shooting? If that doesn't work for you then you may need to first calibrate those low shots, remove the gradient with a gradient removal tool and then set them aside for stacking later when you have all the data you want to stack. My open tube home built Classical Cass is hard to shoot flats for because I cannot do the Mid-Day method I can use on my refractors because the OTA is not light-tight. I found a workaround by using a ML4022 camera which has a KAI series interline sensor. Using the electronic shutter I can take flats that are 100milliseconds (0.1 second) long and not have any mechanical shutter transition artifacts. That is a good method for an interline chip. The problem with removing the gradient from individual frames is that the SNR is usually low and when you start doing mathematical manipulations on low SNR images you can make a noisy mess out of the result.
Do I use flat dark subtract when making the master flat or simply use bias subtract when applying the flats?
--you can do either. I usually use darks (no RBI mitigation!). My flats will be exposed from 5 to 12-13 seconds typically so I may shoot 20 darks of 6 seconds and use that to calibrate each raw flat prior to stacking (using normalization turned ON).
I seem to get better results using bias subtract when applying the min/max clipped master flat. But perhaps I am not
doing something 100% there. I use dusk flats usually pointing to the opposite side of the sky to the sunset area
and usually only 3 flats per filter (I see you recommend more than that). It is a bit of a rush to capture dusk flats
but I find also I get the same results by using a white cotton cover over the scope and take flats during the day
in my internally black painted observatory as a small amount of light filters in at the roof/wall junction.
---That is precisely why I recommend having a fast downloading camera. You are screwed on flats otherwise. The best flats use the sky as the light source, your task is to come up with a scheme that works. 3 is entirely too few. Look over my treatise on flats again and see how the FFPTC shows you how effective your flats are at removing the FPN (can be sensor based FPN or can be optical non-uniformity including the center hot spot (really cosine to the fourth power rollof) and dust motes.
Do small well cameras require more subs in their flats than deep well cameras?
---No: again looking at the FFPTC information (flat field photon transfer curve) you can immediately see that for low signal level in the images you don't need much in the way of flats. For high signal level you need really good flats. So the moon is a worst case but a faint galaxy may be the best case from the perspective of making do with a small number of flats. Signal level is what is important.
Show message history
Best regards,
Greg.
Greg Bradley
Reply, Reply All or Forward | More
Me Hi Richard, Brilliant. Thank you for taking the time to answer me there. Much appreciated. Best regards, Greg. Greg Bradley To Richard Crisp
Today at 12:58 PM
Hi Richard,
Brilliant.
Thank you for taking the time to answer me there. Much appreciated.
Best regards,
Greg.
Greg Bradley
Show message history
On Saturday, 9 November 2013 9:50 AM, Richard Crisp <rdcrisp@sbcglobal.net> wrote:
answers inline below.
I am including our new buddy Mark Striebeck of Google on this reply because I suspect he will gain some benefit from this discussion.
rdc
From: Greg Bradley <bradgregley10@yahoo.com>
To: "rdcrisp@sbcglobal.net" <rdcrisp@sbcglobal.net>
Sent: Friday, November 8, 2013 11:46 AM
Subject: re; Flats
Hi Richard,
Thanks for writing your paper on flats. Very infomative.
I remember various posts that were made after you published this paper. Your recommendation
for a typical Proline 16803 camera if I recall correctly was to go for about 40K ADU levels. Your
paper indicates around 5-10 subs should do (for an ML4022).
---I don't recall the ADU levels but I described how to determine the camera gain by making a photon transfer plot to character the main performance parameters of the camera. I always work in terms of electron units and to be able to convert from ADU to electrons you need to know the gain (units of e-/ADU)
One thing I didn't understand was the recommendation for 250Ke to 1000Ke. I use CCDSoft and the reported
values are in ADU. How do you convert Ke to ADU?
----1Ke- = 1000e-
ADU * gain (e-/ADU) = e-. When you multiply these two items together the ADU in the numerator cancels the ADU in the denominator leaving behind only e-. That, by the way, is called "dimensional analysis" and is taught in the first semester of the freshman year in engineering school which is where I learned it way back at Texas A&M in 1972 (dating myself).
---Last time I measured a Proline 16803 I found a gain of about 1.5e-/ADU. So for 40,000 ADU that means you have about 60,000e- or 60Ke-
I use a CDK17 often and it has bad vignetting and tends to have a hot spot in the illumination in the centre.
I have minor light pollution when imaging rising objects in the east. This results in hard to process images
and flats do not seem to get rid of all the vignetting but a lot of it.
What would you consider best practice for flats in this situation?
---Flats are easy in principle but difficult in practice. If you are imaging a rising object in the East and have Light Pollution, why not wait until it is higher and then begin shooting? If that doesn't work for you then you may need to first calibrate those low shots, remove the gradient with a gradient removal tool and then set them aside for stacking later when you have all the data you want to stack. My open tube home built Classical Cass is hard to shoot flats for because I cannot do the Mid-Day method I can use on my refractors because the OTA is not light-tight. I found a workaround by using a ML4022 camera which has a KAI series interline sensor. Using the electronic shutter I can take flats that are 100milliseconds (0.1 second) long and not have any mechanical shutter transition artifacts. That is a good method for an interline chip. The problem with removing the gradient from individual frames is that the SNR is usually low and when you start doing mathematical manipulations on low SNR images you can make a noisy mess out of the result.
Do I use flat dark subtract when making the master flat or simply use bias subtract when applying the flats?
--you can do either. I usually use darks (no RBI mitigation!). My flats will be exposed from 5 to 12-13 seconds typically so I may shoot 20 darks of 6 seconds and use that to calibrate each raw flat prior to stacking (using normalization turned ON).
I seem to get better results using bias subtract when applying the min/max clipped master flat. But perhaps I am not
doing something 100% there. I use dusk flats usually pointing to the opposite side of the sky to the sunset area
and usually only 3 flats per filter (I see you recommend more than that). It is a bit of a rush to capture dusk flats
but I find also I get the same results by using a white cotton cover over the scope and take flats during the day
in my internally black painted observatory as a small amount of light filters in at the roof/wall junction.
---That is precisely why I recommend having a fast downloading camera. You are screwed on flats otherwise. The best flats use the sky as the light source, your task is to come up with a scheme that works. 3 is entirely too few. Look over my treatise on flats again and see how the FFPTC shows you how effective your flats are at removing the FPN (can be sensor based FPN or can be optical non-uniformity including the center hot spot (really cosine to the fourth power rollof) and dust motes.
Do small well cameras require more subs in their flats than deep well cameras?
---No: again looking at the FFPTC information (flat field photon transfer curve) you can immediately see that for low signal level in the images you don't need much in the way of flats. For high signal level you need really good flats. So the moon is a worst case but a faint galaxy may be the best case from the perspective of making do with a small number of flats. Signal level is what is important.
Best regards,
Greg.
|
09-11-2013, 01:12 PM
|
|
Registered User
|
|
Join Date: Jan 2012
Location: Melbourne
Posts: 3,687
|
|
My method to take flats:
I use my 27" LCD monitor with the screen made all white in Photoshop.
I put the Newt. on 2 pillows - a white T shirt taped across the front aperture.
I make sure the camera has not moved in rotation or focus.
I put the monitor about 2 feet away from the front aperture
& line it up to be central.
I turn all the lights out & put a black T-shirt over the camera
but not on the heatsink - to stop stray light getting in.
I also block the bottom end of the Newt. with a cover & black cloth.
I bring the camera to the correct identical temperature.
( probably not necessary)
I then take short exposures to go to about half the 16 bit brightness
for all LRGB & Ha filters - 3 frames of each.
|
09-11-2013, 10:09 PM
|
|
Registered User
|
|
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
|
|
Why take flats though?
I can see how they correct for vignetting and dust, but can't see how they help otherwise - flats can also get rid of fixed pattern noise (FPN), but only in the bright parts of the image where it doesn't matter. I am not too concerned that my saturated stars should have high SNR - I want high SNR in the dimmest bits where FPN is negligible.
The example camera in Richard Crisp's paper is an 8300 with a fixed pattern factor of 0.77%. My 694 is a bit under 0.4% (I think). With either of them, according to Richard's data, this level of FPN is not significant until the SNR reaches above about 100. This is way above the ISO standard for "excellent" image quality, so FPN removal will be unlikely to improve in any significant way on what is already an excellent+ image quality .
The human visual system is very sensitive to regular patterns, but if this is an issue, the images can be dithered, which turns the FPN into minor random noise.
As far as I can tell, if the camera has very high FPN, flats could be useful for reducing noise - otherwise they can only really help in correcting for uneven illumination (vignetting and dust). Am I missing something?
Last edited by Shiraz; 10-11-2013 at 07:00 AM.
|
09-11-2013, 10:35 PM
|
|
Registered User
|
|
Join Date: Feb 2006
Location: Sydney
Posts: 17,996
|
|
My opinion is flats are more vital if you have any light pollution, use a scope with a smallish corrected circle, have dust issues with the chip or filters, or use a large chipped camera, use longer focal length or want to increase contrast in dim areas plus make the image look more balanced and even.
I see a definite improvement even with the 694 chip on my CDK17. The vignetting is mild to weak but its still there. If a dim part of say a galaxy falls in that zone it loses contrast. There is a gain, not massive and nothing like what is needed with the 16803 chip but still worth it.
Greg.
|
09-11-2013, 10:40 PM
|
|
Registered User
|
|
Join Date: Feb 2006
Location: Sydney
Posts: 17,996
|
|
Quote:
Originally Posted by alpal
My method to take flats:
I use my 27" LCD monitor with the screen made all white in Photoshop.
I put the Newt. on 2 pillows - a white T shirt taped across the front aperture.
I make sure the camera has not moved in rotation or focus.
I put the monitor about 2 feet away from the front aperture
& line it up to be central.
I turn all the lights out & put a black T-shirt over the camera
but not on the heatsink - to stop stray light getting in.
I also block the bottom end of the Newt. with a cover & black cloth.
I bring the camera to the correct identical temperature.
( probably not necessary)
I then take short exposures to go to about half the 16 bit brightness
for all LRGB & Ha filters - 3 frames of each.
|
Sounds good. Taking flats at the same temperature I would regard as quite important. When I had an Apogee U16M the really slow cooldown time used to cause me grief. I would inevitably leave it to dusk to setup for flats and the light would be failing yet the camera took 40 minutes to cooldown. In that time the flats I did take were useless as they were too warm and the noise levels way too high.
You are referring to a DSLR I am sure but regardless it would be best practice to match the temperature. With most CCD if you look at their data sheet they typically double the thermal noise with every 6C rise in temp. That would be true for DSLRs as well. So if you took images at night in 10C and then took flats at 25C at dusk you may not get ideal results.
Per Richard 3 is not enough so I plan to increase the number of flats I take. It may be worthwhile to take more and compare results with yours as well.
Greg.
|
09-11-2013, 10:48 PM
|
|
Registered User
|
|
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
|
|
Quote:
Originally Posted by gregbradley
My opinion is flats are more vital if you have any light pollution, use a scope with a smallish corrected circle, have dust issues with the chip or filters, or use a large chipped camera, use longer focal length or want to increase contrast in dim areas plus make the image look more balanced and even.
I see a definite improvement even with the 694 chip on my CDK17. The vignetting is mild to weak but its still there. If a dim part of say a galaxy falls in that zone it loses contrast. There is a gain, not massive and nothing like what is needed with the 16803 chip but still worth it.
Greg.
|
so you are just using flats to correct for illumination variability - not FPN? If you are not trying to reduce pixel-level fixed pattern noise, have you considered (or tried) noise filtering on the flats?
Last edited by Shiraz; 10-11-2013 at 12:13 AM.
|
09-11-2013, 10:53 PM
|
|
Registered User
|
|
Join Date: Apr 2011
Location: Sydney
Posts: 341
|
|
I used dimmable lightbox bought from Peter and MaximDL SkyFlatAssistant plug-in. Very much automated. I do too think vignetting and dust shadows are significantly reduced from flats. I Google searched target ADU, many said 1/3 of depth well is ok, so I used 22k, will try much higher as Greg, Allen and other have used.
|
10-11-2013, 08:39 AM
|
|
Registered User
|
|
Join Date: Feb 2006
Location: Sydney
Posts: 17,996
|
|
Quote:
Originally Posted by Shiraz
so you are just using flats to correct for illumination variability - not FPN? If you are not trying to reduce pixel-level fixed pattern noise, have you considered (or tried) noise filtering on the flats?
|
When I first used the 694 in CCDsoft the first thing that was obvious was FPN. I could see some manufacturing cut marks in some shots (short focus shots) and also a grid like flyscreen so the chip definitely has FPN.
But I like the idea of using PI to correct you mentioned. The idea of noise filtering on the flats I have not heard before and it'd be worth trying as is Martin's approach of multiple level flats and average combine them.
What tool did you use in PI to correct the uneven illumination?
Perhaps a test of the various methods with photos posted showing the results could be useful.
Greg.
|
10-11-2013, 10:25 AM
|
|
Registered User
|
|
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
|
|
Quote:
Originally Posted by gregbradley
When I first used the 694 in CCDsoft the first thing that was obvious was FPN. I could see some manufacturing cut marks in some shots (short focus shots) and also a grid like flyscreen so the chip definitely has FPN.
But I like the idea of using PI to correct you mentioned. The idea of noise filtering on the flats I have not heard before and it'd be worth trying as is Martin's approach of multiple level flats and average combine them.
What tool did you use in PI to correct the uneven illumination?
Perhaps a test of the various methods with photos posted showing the results could be useful.
Greg.
|
thanks Greg.
Under what conditions do you see the grid structure? - my flats and bias images are very smooth at f4. I have seen fixed noise structure in narrow band images, but that is due to thermal noise and cannot be handled with flats (flats can deal with variability of pixel sensitivity, but most cameras seem to have negligible pixel gain variation anyway).
I think that flat smoothing is widely used - for example Nebulosity applies a smoothing filter by default to remove hot pixels. This is different to averaging multiple flats, because filtering blends individual pixel values and removes the ability to do FPN correction on noise. However, it looks to me like FP noise is a non-event anyway with most cameras and certainly on targets with texture, so why not smooth the flats and reduce their noise contribution to the final image.
Martin's method of averaging flats with different ADU would probably help to deal with non-linearity in the CCD - flat fielding falls in a heap if the flats or lights stray into a non-linear region.
The PI dynamic background extraction process will clean up gradients pretty well, including minor vignetting - but not dust.
would be nice to be able to post some comparative images - for the past 6 weeks though, that would have been almost exclusively the undersides of clouds and of limited interest. Maybe tonight if the wind drops below 20kts Cheers Ray
Last edited by Shiraz; 10-11-2013 at 05:07 PM.
|
11-11-2013, 11:15 AM
|
|
Don't have a cow, Man!
|
|
Join Date: Apr 2013
Location: Adelaide, South Australia
Posts: 1,105
|
|
I find flats invaluable. I do have some vignetting and trying to keep the sensor, filters and other glass surfaces dust free is very difficult.
I have a light box by Exfso that came with my scope that has a potentiometer on the side to alter the brightness(have another Exfso light box too, good stuff!), 3/4 for LRGB and flat out for narrow band. I generally aim for about 30,000 which gives a nice flat and cleans my images right up, simplifying my processing.
Generally I let the software take care of the maths, although I am sure it does all the bits as I see it in the processing console as its working(PI is my processing software).
I think the real trick is to work out what the best flat for your camera is using a very even light source.
Interesting read thanks Greg.
|
Posting Rules
|
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
HTML code is Off
|
|
|
All times are GMT +10. The time is now 08:12 AM.
|
|