Dusk or Dawn flats

[andrew s]

If anyone takes these would you be kind enough to share you methods and timing etc.

Thanks Andrew

[Martin Meredith]

Hi

I occasionally take dusk flats with the scope pointing at around 70 degrees and open (e.g. no t-shirt diffuser) and aim for a large number of subs (~100) of up to a couple of seconds each, which I then combine using a percentile-based outlier rejection method to get rid of the inevitable stars that appear (I usually find keeping the central 70-80% works well). Timing is critical, especially if you are planning to collect flats for more than one filter. I reckon there is a sweet spot of 10-15 minutes when you can get good flats this way i.e. when the exposure times are not too short to be noisy and not too long that you can't collect enough subs to apply outlier rejection successfully. I reckon someone should publish 'flats' timetables like the tide tables! Actually, I'm trying to automate the procedure in software at the moment.

Others may have a different process. This is what I use for EEVA-style observing rather than AP and I find it works well.

cheers

Martin

[andrew s]

11 minutes ago, Martin Meredith said:

Hi

I occasionally take dusk flats with the scope pointing at around 70 degrees and open (e.g. no t-shirt diffuser) and aim for a large number of subs (~100) of up to a couple of seconds each, which I then combine using a percentile-based outlier rejection method to get rid of the inevitable stars that appear (I usually find keeping the central 70-80% works well). Timing is critical, especially if you are planning to collect flats for more than one filter. I reckon there is a sweet spot of 10-15 minutes when you can get good flats this way i.e. when the exposure times are not too short to be noisy and not too long that you can't collect enough subs to apply outlier rejection successfully. I reckon someone should publish 'flats' timetables like the tide tables! Actually, I'm trying to automate the procedure in software at the moment.

Others may have a different process. This is what I use for EEVA-style observing rather than AP and I find it works well.

cheers

Martin

Thanks Martin, what time do you start relative to sun angle? 

Regards Andrew 

[Martin Meredith]

Good question! I've no idea... What I do is move the scope outside about 1 hour before it is dark enough to start doing serious observing, and I guess about 30 mins later is when I would start to think about taking flats. Next time I will check how far the sun is below the horizon at that point. Not far below, I reckon.

What I actually do is take test exposures and observe the histogram, aiming to get the peak around 70%. Part of what I hope to do in software is to estimate the best time to start taking flats based on analysing the test exposures automatically. And of course the best time will vary with transmissibility if you're using filters. 

Martin

[tomato]

The problem I come across is that at dusk the light level is changing at quite a rate such that it is not possible to take 30 flats at the same exposure time and if you keep compensating for this the flats obtained are not very repeatable from frame to frame. 
 

I have found it much easier with a light panel over the end of the scope which generates a reasonably consistent light output.

[x6gas]

+1 for a light panel.

Also I try really, really hard to keep everything dust free so that if I use flats (and I have to confess I don't always) they are only correcting for vignetting and not dust-bunnies so I can usually get away with one set of flats for all filters and over multiple imaging sessions.

HTH, Ian

[MarkAR]

+1 for a light panel. Proper astro EPL like Artesky make it a breeze.

[davew]

Hi Andrew,

To answer the question as best I can, I have to rely on my poor memory ! When I started doing flats I read up and carefully copied, as best I could, what the Pros did. 

To select a region of sky I read this - https://iopscience.iop.org/article/10.1086/133817/pdf  This article has many citations.

To work out the exposure time I used Maxim DL and a plug in from John Winfield - http://winfij.com/maximdl/skyflats.html  So much simpler than trying to do it manually.

Start and finish times ( This is where my memory could be out ! ) were around sunset for Ha, then Oiii and for RGB around 5* after sunset and for L around 9* / 10* after sunset. The tighter the narrow band filter, the earlier you can start.

You have to have your target exposures and ADU values already worked out and at sunset start things rolling and Maxim would say, via the plug in, when the target values were reached. The plug in also had a predictive algorithm that was very good. This plug in would only save the exposures you set it to save. I believe PI has a very good normalising routine to sort out any changes in exposure levels.

I also used a white panel to replicate dome flats and they actually seemed to work better for me. ( End panel from a kitchen unit  )

I did end up using a flat panel as the lenses I used to image with were fairly small and it meant I could take flats at any time. I went off sky flats because clouds drifting through made a mess ! If you are in a less cloudy place then you're sorted.

I can't remember much more but you can ask !

Dave.

 

[davew]

I think the most qualified person on this site to answer your question could well be dph1nm . 

I hope he sees this post.

Dave.

[Davey-T]

37 minutes ago, davew said:

To work out the exposure time I used Maxim DL and a plug in from John Winfield

I used that for a while but as said clouds were a pain and eventually built a light box for my SCT much easier doing them in the obs'y at any time.

Dave

[andrew s]

Thanks for the replies.  A light panel is a possibility by I have  400mm scope so ther are expensive and as its remote in a shared observatory not that easy logistically. 

I am trying out other options  first.

Thanks again.

Regards Andrew 

[Whirlwind]

Sky flats should be fine for the telescope.  Gradients only tend to become an issue on widefield setups, but I'm guessing the focal length we are looking at here is 1600mm+ and hence the field of view is small.  Just make sure you point the telescope in the right direction (i.e. about 180 degrees from the sun, i.e. west at dawn, east at dusk and about 45-60 degrees altitude).  

The real trick is that exposure times need to be constantly adjusted.  If you have fast downloads then you may be able to run a sequence of 10 or so and then change exposure.  Some programs will do this automatically. You want to aim for the exposure of about half the full depth (so a count of about 25000-35000 for 16 bit cameras, adjust accordingly for 12/14 bit CMOS etc).  It doesn't matter that you are altering exposure length as long as the counts stay around these figures (though very short exposures and long exposures raise the risk of shutter effects or stars becoming visible so should be avoided).

There is also the risk of point source (stars) and there are two ways to avoid this (noting though you should always avoid bright stars).  Firstly you can turn off tracking, stars will then trail and hence the area of the sky being observed changes.  Hence when you combine multiple flats these blurred areas are averaged away.  The alternative is to take a sequence of say 10 images, then slew the telescope slightly and then another sequence (easier with faster downloads).

Because conditions change quickly it can take time to build up an appropriate sequence of flats especially when you factor in multiple filters as well.  A lot of professional observatories cycle their flats i.e. every day they will take dawn and dusk flats for their filters and then for images use those generated x weeks before.  This way they have a continually relevant master flat.  They have the same issue in that you can't have a white panel once you get above a certain size.

Finally don't use flats when there are any clouds.  This will really mess up the flats as you no longer have consistent even illumination.

[andrew s]

On reflection I am not sure how practical they are with a CMOS camera that does not allow for scaled flats due to the non linear amp glow. However, thanks for all the input.

I am experimenting with night time sky flats.

Regards Andrew 

[bottletopburly]

20 hours ago, andrew s said:

Thanks for the replies.  A light panel is a possibility by I have  400mm scope so ther are expensive and as its remote in a shared observatory not that easy logistically. 

I am trying out other options  first.

Thanks again.

Regards Andrew 

Look at using an led ceiling light you get Big ones these days up to 600mm square 

[andrew s]

50 minutes ago, bottletopburly said:

Look at using an led ceiling light you get Big ones these days up to 600mm square 

Thanks for the idea but logistically it would be difficult in the remote ROR observatory. 

My trial with sky flats last night looks promising even with the near full moon.

Regards Andrew 

Edited June 5, 2020 by andrew s

[bottletopburly]

2 hours ago, andrew s said:

Thanks for the idea but logistically it would be difficult in the remote ROR observatory. 

My trial with sky flats last night looks promising even with the near full moon.

Regards Andrew 

Could you not fix to the wall then create a custom park so it points at light source.

[andrew s]

3 minutes ago, bottletopburly said:

Could you not fix to the wall then create a custom park so it points at light source.

Possibly it's something I will look into if the sky flats look problematic.  In theory night sky flats shoul be ideal. The main issue, it seems, is having the time to take them but this is not an issue for me.

Regards Andrew 

[Davey-T]

The main issue I found was the surprisingly rapid light level changes so need software to do it.

Dave

[andrew s]

26 minutes ago, Davey-T said:

The main issue I found was the surprisingly rapid light level changes so need software to do it.

Dave

I am not referring to dusk and dawn flats now but to full on night flats. I have found a reference to some "blank" areas of the sky and imaged one last night. Not ideal given the moon but with a large dither and median combine it looks like it will give a good quality low noise flat.

Obviously you need lots if data so several hours exposure but I can do it on nights of intermittent cloud which are not ideal for my work on flare stars.

Regards Andrew 

 

[dph1nm]

On 04/06/2020 at 16:49, davew said:

I think the most qualified person on this site to answer your question could well be dph1nm . 

I hope he sees this post.

Dave.

Err, well it took me a while to see it, and I am not sure I can help much. With my DSLR I usually point the telescope relatively near the zenith at dusk and wait until 0.5s exposures just saturate the back of the camera histogram (in blue of course, not green or red). Note that on my 1000D 'saturation' in the histogram is actually when the jpeg saturates - there is another two stops (4x) to play with in RAW. I usually aim for about 40 before I get bored (or it gets too dark).

Professionally it is easy - you just look up a suitable 'blank' field and then ask the night assistant when to start observing or sneak a look at the logs to see when the previous observers did theirs)!!

I have also used night flats professionally - they waste too much time amateur-wise given how little clear sky I get round here. If you targets are small and you dither (or you take lots of different fields) then they can work extremely well. Obviously they do not work well if you are imaging a large nebula!

NigelM

[andrew s]

@dph1nm thanks. I have taken several sets of sky flats pointing at a "blank" field over four nights. About 8hrs worth in total. My night assistant Voyager did a fine job and they were ready each morning.

This has produced a good quality V flat  that works well with my vignetted field.  Now trying the SA200.

Fortunately,  my scope is in Castillejar Spain and I (Voyager) fits them in before my targets are hight enough to image or after they get too low.

Regards Andrew 

 

Edited June 17, 2020 by andrew s

[Martin Meredith]

I understand you have found a solution using sky flats but in case anyone else comes across this thread, here is my experience of collecting dusk flats last night and monitoring the timings.

I have an automatic routine that checks the exposure required to reach a target ADU (in my case I set it to 80% of the ADU range based on a robust estimate -- i.e. at the top of the expected linear range of the sensor, but I allow a range so e.g. even 30% of the ADU range is deemed OK). Essentially it collects 5 exposures in the range 0.5-2.5 seconds and checks whether it is too early, too late, or just about the right time to collect the flats.

I was able to collect H-alpha flats with the sun 3 degrees below the horizon, but couldn't start blue until it had reached 5 degrees below. Then it was a bit of a rush to do blue, green and red but there was time to collect 91@1s for blue, 39@2.25s for green and 70@2.5s for red (not the most sensible distribution I know). Then I had to wait until the sun was around -7 below to collect luminance flats (98@1.5s). 

Yes, there were some stars -- pretty much unavoidable with dusk flats -- but retaining the central 80% of values at each pixel was able to remove them effectively. It has been mentioned above that changing exposure times is one approach to handling the rapid drop in light levels at dusk. An easier approach, for me at least, in software, is to use the same exposure for each sub, but to scale the individual subs so that they have the same overall level (for normalisation I use an outlier rejection scheme operating in the central third of the image, but I'm sure other approaches would work as well).

It is important to either scale in this way, or adjust exposures during the capture, not for aesthetic reasons or ensuring theoretically the same SNR in each sub, but for the practical reason that outlier rejection techniques don't work well if the subs are not normalised. What is an outlier in one sub and ought to occupy an 'extreme' brightness value might not do so if the brightness level is changing across the set of subs. Actually, this doesn't affect the obvious outliers, but it does affect the fringes of stars that creep into the flats, leaving ghost trails on the flats. 

In sum, it is possible to collect flats at twilight for multiple filters in sufficient quantities to use outlier rejection to remove stars, but the timing is quite tight for the RGB filters...

Martin