Thoughts on which imaging rigs to concentrate on

[Gina]

It's cloudy now - Q.E.D.

[Gina]

PI has just finished BPP on the Cygnus Loop SII data - 115 subs of 120s.  Weak image which might benefit from more data.  I'll post process the image in Photoshop and post the result tomorrow.  I collected sets of biases and darks to complete my library tonight

Edited November 10, 2016 by Gina

[Gina]

Here's the SII master light, well histogram stretched, rotated and cropped in Photoshop.  It's out of focus as well as faint but the forecast for tonight is looking good so maybe I can capture a replacement set.

[Gina]

Clear sky tonight as forecast   Capturing SII subs of Cygnus Loop.  Tried 120s but got oval stars so the PA has gone off a bit - it was alright up to 5m.  So I've gone down to 60s and increased the gain by 12dB (560 on the gain scale) which is just 4dB below maximum gain.  I think this should work - here is a single sub, histogram stretched, rotated and cropped.

[Gina]

There's patches of cloud coming over but I'm hoping it will go away.

[Gina]

Just processed the SII Cygnus Loop subs I captured the other night.  Firstly, I used Blink in PixInsight to weed out the duff frames of which there were a fair number, mainly due to passing clouds and finally to complete cloud cover.  The result was 52 reasonable frames which I calibrated with the bias and dark frames I took the next day plus my latest master flat using BBP in PI.  Here is the result with further processing in Photoshop to histogram stretch, rotate and crop.

[Gina]

Gone back to 2016-10-13 for Ha subs of the NAN & Pelican.  Processed with BPP in PI and then histogram stretched and slightly cropped in Photoshop.  30s, g=440, T=-30°C.

[Gina]

I'm now capturing OIII subs of the NAN with 60s subs with gain of 500 and -28°C between patches of cloud.

[Gina]

Been going through last night's subs and as I expected far more duds than good ones - I could see the clouds passing over except when they were continuous, then the frames were just white.  So far I've gone through about half of the 400 collected.  Rather a tedious job but considerably helped by Blink in PixInsight.

[Gina]

Well, I have 92 frames that contain some data.  A few of them have fairly good images but more than half are pretty poor so I have a dilemma - take maybe a dozen or so good frames only or take the lot.  My feeling is that "poor" extra data would not improve matters.

[Gina]

Seeing the Jellyfish Nebula (IC 443) in the DSO forum, I thought I'd check just where it is with the idea of imaging it when we next get a clear night (probably be several days yet though) BUT it's right next door to the moon so will have to wait for quite some time yet, until the moon is out of the way.  Even with 3nm filters I think the moon is likely to be a problem.  I might just see though

Here is a screenshot of CdC for 10pm this evening

Edited November 16, 2016 by Gina

[Gina]

First screenshot shows the relative FOV with the 135mm f2.5 lens - my current setup plus position of the moon tonight.  A narrower FOV might be better but there is quite a lot of nebulosity in the neighbourhood and this lens is excellent.  Next up is the 200mm f4 and then the Esprit scope with 400mm f5.  These are shown in the second screenshot with biggest rectangle the FOV with 135mm lens then 200mm and smallest the scope.

[Gina]

Here is an image from Wikipedia which covers a nice FOV similar a bit smaller than my 200mm lens.

Edited November 16, 2016 by Gina

[Gina]

And a couple more akin to the FOV of my 135mm lens.

Edited November 16, 2016 by Gina

[Gina]

More reading of the "Inside PixInsight" book has been very interesting and I have learned a lot more.  I can see better ways of processing my data and learned more about the relationship between bias, dark and light frames as well as shortcuts to taking and calibrating both lights and flats.  I know the calibration I have been using has left a lot to be desired and I can see ways it can be improved including SuperBias. 

[Filroden]

20 minutes ago, Gina said:

More reading of the "Inside PixInsight" book has been very interesting and I have learned a lot more.  I can see better ways of processing my data and learned more about the relationship between bias, dark and light frames as well as shortcuts to taking and calibrating both lights and flats.  I know the calibration I have been using has left a lot to be desired and I can see ways it can be improved including SuperBias. 

Before you go down the superbias route I'd suggest reading alternative views on its usefulness. There is a really good thread on Cloudy Nights. Based on that thread I created a normal master bias using over 250 subs and it has worked much better than the superbias I created. Given how easy it is to take many bias frames, statistically, you're going to get a better result integrating real frames rather than trying to mathematically simulate that same result from fewer frames.

[Gina]

Thank you Ken - that sounds a good idea   One thing I've just learned is that you need hundreds of bias frames and I've been using only 100.  As you say taking several hundred bias frames is no problem though a lot of data.  OTOH I have arranged to cater for vast amounts of data and 10GB for 300 bias frames is no problem.

[Gina]

I shall need to read the Calibration and Integration section again to take it all in.  So far I've gathered that several hundred bias frames are required and I've just read the suggestion of twice as many darks as lights.  Now one of my sets of lights amount to 250 frames which would indicate 500 darks.  500 x 1m darks is over 8 hours - this is getting ridiculous!  Maybe the 250 lights want pruning to select the very best.  I gather that adding extra poorer quality lights can degrade rather than improve the result - so fewer really good lights should be better.

Seems PixInsight works rather differently from DSS et al and uses different ideas to obtain improved results.  This is all very interesting and encouraging but I have an awful lot to learn!!  Good job I like learning

Edited November 19, 2016 by Gina

[Horwig]

15 hours ago, Gina said:

More reading of the "Inside PixInsight" book has been very interesting and I have learned a lot more.

Didn't know this was out, Christmas is sorted then 

[Filroden]

13 hours ago, Gina said:

I shall need to read the Calibration and Integration section again to take it all in.  So far I've gathered that several hundred bias frames are required and I've just read the suggestion of twice as many darks as lights.  Now one of my sets of lights amount to 250 frames which would indicate 500 darks.  500 x 1m darks is over 8 hours - this is getting ridiculous!  Maybe the 250 lights want pruning to select the very best.  I gather that adding extra poorer quality lights can degrade rather than improve the result - so fewer really good lights should be better.

Seems PixInsight works rather differently from DSS et al and uses different ideas to obtain improved results.  This is all very interesting and encouraging but I have an awful lot to learn!!  Good job I like learning

I think he assumes you're taking few, long exposure lights so double darks could be right. For many short lights I wouldn't go more than about 30-50 darks.

[wimvb]

13 hours ago, Gina said:

I shall need to read the Calibration and Integration section again to take it all in.  So far I've gathered that several hundred bias frames are required and I've just read the suggestion of twice as many darks as lights.  Now one of my sets of lights amount to 250 frames which would indicate 500 darks.  500 x 1m darks is over 8 hours - this is getting ridiculous!  Maybe the 250 lights want pruning to select the very best.  I gather that adding extra poorer quality lights can degrade rather than improve the result - so fewer really good lights should be better.

Seems PixInsight works rather differently from DSS et al and uses different ideas to obtain improved results.  This is all very interesting and encouraging but I have an awful lot to learn!!  Good job I like learning

Why would you need so many darks?

Each light frame is calibrated with a master dark. The master dark is the integration of all dark frames. The noise in the master dark will add to the noise in each light frame during calibration, according to

noise_cal^2 = noise_uncal^2 + noise_md^2

where noise_cal is the noise in the calibrated light frame, noise_uncal is the noise in the uncalibrated light frame, and noise_md is the noise of the master dark.

As long as the noise in the master dark is substantially less than in the uncalibrated light frame, it shouldn't decrease the quality much.

IMO, since the noise scales down as SQRT(number of frames), the master dark noise should be substantially lower than single light frame noise, even with relatively few dark frames.

500 dark frames for a master dark seems excessive.

BTW, the above argument is for random shot noise, not for fixed pattern "noise", which can not be integrated out in a master dark, no matter how many dark frames are used in creating it.

 

Or am I missing something (again)??

[Gina]

I gather fixed pattern noise is handled by bias frames. 

I agree with your reasoning about number of darks.  Yes, this book is based on either DSLR cameras or CCD astro cameras - the ZWO cameras designed for astro use are a pretty new innovation and a different beasty from either, so a bit of "intuition" is needed in interpreting the information.  This is a new experience for me after a year or so of DSLR imaging followed by a couple of years of high quality astro CCD imaging.  Big and noisy to small and quiet and now to big and less noisy.

So latest thoughts are - several hundred biases, maybe 50 or so darks and as many decent lights as possible.  I see it's unnecessary to take special darks for flats as the "Optimize" parameter can be used to correct for the shorter time used for flats but I'm not sure this will work with the lower gain used when taking flats - I don't remember reading anything about this.

[Gina]

I've just completed 300 bias frames but I'm wondering if anything would be gained by taking yet more.  The 300 didn't take long.  I guess I could take another 300 and do a comparison between the integration of 300 versus 600 and see if I can see any difference.

[Gina]

Now downloading the bias frames to my Linux Mint desktop ready to integrate in PI.  Doesn't take long with gigabit Ethernet link

[Gina]

Now running ImageIntegration on 300 bias frames using the settings recommended in the book for a large number.  Each frame is taking just over a second (about 4 frames in 5s) so something like 6-7 minutes processing time.