Processing Sky Flats

Very happy that CCDAP now can take my flats. But closer inspection of the flats:

shows that they captured quite a number of stars. Using normal mean calculation, I get this master flat:

Hmmm, not to good. Maybe sigma reject mean:

Better, but the stars are still there. Maybe clip min/max Mean:

Also not.

Do I need to process these better? Or do I need to change the CCDAP parameters to take flats when it's brighter (i.e. lower the min and max exposure time)?

After reading more about sky flats processing, it seemed as if a Poisson reject mean algorithm works well for this. Trying that out:

Now, this looks awesome! Not sure why Poisson reject mean is not offered in CCDStacks Flat Master section...

Imaging automation program

I checked on the ccd-newastro Yahoo! mailing list if there is a program that can take flats at dusk or dawn automatically and was pointed to three programs:

1. ACP
2. CCDAutopilot
3. CCDCommander

They all integrate with TheSkyX! CCDCommander i$99,  CCDAutopilot is $295 (I need the professional version for full TheSkyX support), and ACP is $695 (the cheapest version!)

Unfortunately, CCDCommander does not support focusing and camera control in TheSkyX, so I started with CCDAutopilot.

After some issues with getting a license file, I could install it. When I started it for the first time, it wanted to know how to connect to the different modules (telescope, camera...) I wanted to use TheSkyX for everything, but somehow that didn't work. After I asked on the excellent CCDWare mailing list, I found out that I'd have to start TheSkyX with administration permissions (so that it can run as a server process). Now, CCDAutoPilot connected without any problems.

... but now we suddenly have clouds and even rain! In late June in California!

Last night, I took some images of NGC 6205. At the end of the session, I wanted to use CCDAP for taking flats at dawn. When I tried to connect my equipment, I got a couple of error messages again. Turns out that you can have TheSkyX started, but it can't be connected to Telescope, camera...
After I connected it, it took me several attempts to plate solve - weird. Well, I forgot to turn off PHD - and it kept messing around with alignment...
I then set everything up to take flats at dawn. I used the button "Flat Frames Now" - and CCDAP told me
08:29:01 Waiting for dawn...
08:29:01 Sun Altitude: -15.2, is below set value of -8.0

I figured that it would wait until dawn and then start.

... next morning: no flats - still the exact same message. Turns out that you can use "Flat Frames Now" only to take flats right now. And because it wasn't part of a normal session, it didn't update the sun's altitude.

So, I'll try it again - now at dusk. For flats, I don't need to plate solve to know the exact location.

At dusk, the same situation happened again. CCDAP was "counting down" to Sun Altitude: 21.5, but then it stayed there and didn't update - and hence didn't take any flats. I setup my skope for dawn flats - and now it worked! CCDAP started shortly after 5am. First taking luminance flats. Adjusting the exposure time from initially ~30 seconds further and further down. Dithering between frames to avoid sky artifacts. Then switching to the red filter, taking flats. But then it was already so bright, that it had to expose shorter then 3 seconds. CCDAP stopped red, tried blue and green - but those required shorter then 3 sec exposure too. And finally took Ha flats without problems. So, big lesson learned: Only take 5 (not 20) flat frames to fit in all filters! But now I have to deal with stars in my flats...

But overall it felt like magic (and WAY less effort then doing all this manually).

Unfortunately, when I tried to do this the next evening, it didn't work again - CCDAP got stuck just 10 mins before it should start. This seems to be a problem with TheSkyX - CCDAP queries TheSkyX for that information and somehow it got stuck.

One of the biggest challenges to use CCDAP fully is that I'd have to use TheSkyX autoguiding - despite my best efforts, I could never make it work, but still use PHD... But one step at a time.

Preparing for Golden State Star Party

I want to attend this years Golden State Star Party. Had to figure out how to make my entire setup mobile:

• Need 2 large batteries (with all the equipment, my scope draws several amps now) and a recharger - check
• Need to redo my Pelican case (because of all the equipment changes, the old foam does not fit anymore) - check
• Need another Pelican case for the imaging train (so that I don't have to assemble it, but can take it in one piece) - check
• Need some new equipment: chair, table, head lamp ... - check
• Need to make sure that I bring absolutely everything (tools, eyepieces...) that I might need to repair or fix anything
• A red screen for my laptop to dim it down during the night - check
• And of course everything to stay out there: ropes for fixing the scope during the day (can get windy) aluminet for the scope and my tent to protect it from the sun during the day and all kinds of camping equipment that I don't have anymore - check

I want to use the star party primarily for imaging (imaging time is 10:20pm - 4am). IC 1287, NGC 6888 and NGC 7635. Here is my plan:

1. Night:
• Until Midnight: Collimation (using CCDInspector), Alignment (using PEMPro), White Balance measure (using CCDAP)
• Midnight-4:00: NGC 7635 (13 x 10min Lum, 13 x 7.5 min R = 3h 50min)
2. Night
• Evening Flats
• 10:20 - 12:30: NGC 6888 (13 x 10min Lum = 2h 10min)
• 12:30 - 4:00: NGC 7635 (13 x 7.5 min G, 13 x 7.5 min B = 3h 15min)
3. Night
• Evening Flats
• 10:20 - 4:00: NGC 6888 (13 x 7.5 min R, 13 x 7.5 min G, 13 x 7.5 min B = 4h 55min)
4. Night
• Evening Flats
• 10:20 - 4:00 IC 1287 (10 x 10 min Lum, 10 x 7.5 min RGB = 5h 25min)

With my Nikon/Polarie, I'll do some wide angle shots:

1. Try a nebula region (e.g. north america nebula?) - witht the 50mm lens
2. An ultra-wide angle shots of the milky way - with the ultra-wide lens
3. Cygnus region

Working on my cables

With all the additional equipment, my once so neat cabling became quite the mess. Furthermore, the cables could come between the scope and the mount and impact guiding:

When looking for a USB cable, I found www.l-com.com: they have USB cables in all lengths AND in all angles (up, down, left, right). And for very special needs, there is the custom design tool from usbfirewire.com (but also pretty pricey).

For the serial cable and the dew shield cables, I bought right angle adapters.

Also, the cable bulk in the front of my scope from the robofocus rj11 to serial cable to the serial to USB cable needs to be significantly reduced. I shortened both cables (after ordering a second cable in case I would screw it up).

Next, both the Robofocus control box as the Dew Shield control box are only fixed with velcro and could shift a little. I fixed both of them with the double-sided 3M tape and now they are rocksolid on the plate and the mount.

Now, the scope looks like this:

All neat and tight and no cable can get stuck at the mount or such.

Finally, I replaced my cheap serial to USB adapters with new ones who retain their com port - even when moved to a different USB port. Yei! No port reassignments for the focuser and scope anymore!!

Working on calibration frames

I received some feedback on my image of M51 that I should improve my calibration frames.

First, I took some new darks and bias frames. I removed the camera from the scope, put the 1.25" adapter on it and put a tight cap on it to make sure that no light leaks in. Then I put the whole setup into the freezer to make sure that it stays at -10C. I took 50 dark frames and 100 bias frames.

Here are the results:

Dark Frames:

600 min, 1x1 binned:

450 min, 2x2 binned:

But when I used these new frames on my recent image of M51, I noticed some strange black spots:

Here is the middle part:

I posted on the ccd-newastro mailing list. One explanation was that it could be frost that formed on the camera when it was in the freezer. I should try it again - next time outside when it's dark (and cold enough that I can get the temperature constantly down to -10C).

Bias Frames:

1x1 binned:

2x2 binned:

First observation is how much noisier the 2x2 binned frames are - which is expected. But I wasn't prepared how much it would be.

Second, there is a very bright vertical stripe on the left hand side. I checked with Terry on the starlightxpress Yahoo! mailing list - he said that there is a firmware update that will reduce this. But these are the areas that I usually crop out anyway, so I'm not too concerned. 

Third, there is a vertical gradient in the bias frames. But the difference top to bottom is only less then 20 ADU.

I was first surprised about the maximum readout values of the darks, they were 50000! But these were just hot pixels. In areas without hot pixels, the maximum was 2000-3000. The same as they were for the bias frames - these are expected.

I then also took new flats. This time at night to make sure that no other light is leaking in. On visual inspection they looked very similar to the ones that I previously took.

For better inspection, I subtracted one from the other:

It's hard to see anything - here I stretched the image more:

So, apart from the slightly different level of both images, the only difference I can see is in the upper left corner. This could be caused by stray light (which wasn't there when I took the flats at night).

With regards to flats, I'm not sure what I could do differently with my flat panels. I could try to take sky flats again - but don't want to get up so early (and when I come home, I usually play with the kids).

Summary:

• My bias frames are OK, but I should add a delay between them to make sure that the camera does not heat up.
• Don't take dark frames in the freezer - rather outside when it's colder.
• Doesn't make a difference between taking flats during the day or at night. The only way to improve them is to take sky flats.

Weird reflection ... and tilt!

When I align my scope with bright stars, I often see a reflection:

Not sure if this is normal or if I should worry about it.

Check on the ccd-newastro mailing list and was referred to an interesting web page by Don Goldman that explains this. Very interesting article, but it doesn't seem to provide any ideas of how to fix this (short of buying more expensive optical elements). What it does point out though is that I have some tilt in my imaging train (the halo/reflection should be in the middle when the bright star is in the middle).

The donut has a diameter of 60 pixels, but the image was 8x8 binned, i.e. 480 pixels. The H694 camera has a pixel size of 4.54uM (=0.00454mm) which results in a diameter of 2.18mm. With my focal reducer, my scope has an f ratio of 6.2. That means that the reflection comes from a 6.2*2.18mm=13.5mm.

First, use Don's method to understand where in the image train these are. Here is a sketch of my imaging train:

Using Don's method of calculating at which distance the reflection is created yields to...

But first, check where the reflection comes from. I first checked if it's the same with all filters: yes! So, it's not the filters.

I checked my images with CCDInspector which confirmed this result. Here is the curvature map of a recent image:

The tilt direction is exactly what I could see in the image above.

The most obvious candidate is the adapter of the Lepus reducer as it is not screwed into the reducer, but fixed with 3 screws. I replace it with the celestron t-adapter.
... and no reflection...

Thinking about trying the Celestron reducer - which gets screwed in ...

First image with new setup: M51

After spending a lot of time to improve my setup (hypertuning, guiding parameters, camera setup...) I was finally ready for my first imaging session - even though the moon was pretty bright and high right now. I chose an easy target: M51.

Over 2 nights, I took 7.5 hours worth of data (10x600sec Luminance, 10x450sec RGB). I used these images to learn and understand CCDStack much better (with the help of the excellent tutorial by Adam Block). I had a lot of questions with regards to colors and calibration and Bob Franke helped me a lot too.

Overall, I'm pretty happy with this image: the mount finally seems to guide well (the stars are round) and the colors are quite good too.

Measuring white balance

With my scope finally working again, I started by determining the white balance in my hyperstar setup using the instructions from starizona. Took 3 images of 20LMi (a G3 star). It turned out to too tricky to get the mean brightness value for the stars in those images.

I wanted to do an alternative way and used eXcalibrator. eXcalibrator needs coordinates stored in the fits image. TheSkyX should store these automatically. But somehow they are not. With the help of http://nova.astrometry.net/, I plate solved my images and could download a fit image with the coordinates written in. With that, I get the following parameters:

Red: 1
Blue: 1.970
Green: 1.662

This is still high in red - but that could be the result of being in such a light polluted area where we have a lot of the orange glow from the street lights. Well, let's try it.

I took 10x7.5min RGB exposures (binned 2x2) and 10x10min L exposure (binned 1x1) of M51.

Here are the RGB images and histograms:

Red:
Green:
Blue:

You can clearly see the sky glow in the red image and in the histogram. Now, when I combine all three colors with the luminance image without any color correction, I get the following:

It wasn't clear (at least to me) from the histograms how dominant the red would be.

If I use the color correction values from above when combining the images I get this one:

So, as extreme as these color correction values seemed, in my light flooded area, they seem to be pretty accurate!

I posted about this on the ccd-newastro Yahoo! group and some folks pointed out that I shouldn't fix my sky glow issues with color calibration. Ron Wodaski gave a very good explanation of the difference of color imbalance and bias.

The problem with my (very high) calibration values was that they were derived from captured images, i.e. they did not only contain the color imbalance from my imaging setup but also the bias from the sky glow!

So, for now, I have to combine the RGB images with the same weights and then neutralize the background. Doing that results in this:

This image seems to have more natural colors then the previous one (which was very blue).

Setting PEC

To improve guiding more, I want to finally train and set my PEC (Periodic Error Correction). Hope that it will reduce the RA error more. I am evaluating PEMPro for this. I could connect it fairly easy to my mount, but the camera was a problem. It only works with video cameras, Maxim, CCDSoft or Astrosoft. I have none of these. So, I installed a test version of Astrosoft - that works.

So, getting ready. Starting the Calibration Wizard. Slewing to the correct place in the sky works. First, taking a picture without tracking and measuring star trails works. But the next step (taking a picture while slewing the scope) does not work. I always get the error message "Telescope not connected - Line: 1". Tried everything - nothing :-(

Posted a question in the CCDWare forum - waiting for an answer...

After I turned the camera, I tried again. This time with the "ASCOM RA" option and it worked. Ran through the calibration wizard and could finally start measuring PEC. The good news is that the lines in PEMPro looked very straight. The bad news is that they looked too straight and that the guide star slipped out of the image too quickly. In fact, it looked as if the scope did not track at all.

The next run worked better - I don't think I did anything different though :-S

Here are the recorded graphs:

The resulting PE curves are:

This doesn't look useful at all ...

... I asked about this on the PEMPro forum and Ray Gralak pointed out that I have the infamous CGEM 8/3 frequency here that can't be fixed by PEC. But that PEMPro should be able to improve the other (integer-based) frequencies.

This can be seen when looking at the frequency spectrum:

There are several spikes at full frequencies. There is the 8/3 frequency (2.66) which is inherent in the CGEM mount. But there is also a really larger spike at ~0.2. I have no idea what this is...

I loaded the PEC curve into the mount and did another PEMPro run to see if and how tracking improved. The result wasn't too convincing. The overall RMS error stayed at 4.5 and I couldn't see any improvement when I guided with PHD.

I asked again at the PEMPro forum and Ray pointed out the individual amplitudes roughly doubled (e.g. for the 1 frequency, the amplitude went from 2.053 to 3.739). Which indicates that my curve has to be inverted (which can be done in the upload functionality of PEMPro) in order to cancel out the periodic errors.

OK, doing that...

... and now errors from all frequencies are below 1 !!!

So, now to increase accuracy, I will run this for a few hours, program my mount with the new curve and then check again.

And this is what I got:

Nice! All the integer frequencies are way down. But the big distortion at ~0.2 is still there :-(