Popular Post Astrobdlbug Posted September 18, 2021 Popular Post Share Posted September 18, 2021 (edited) All, so following my 'reset' with my RPi / Astroberry / KStars/EKOS and I am temporarily using a separate power tank to power the dew straps until I source a separate power supply. I had 5 hours of relatively clear skies last Wednesday night and no RPi crashes - it all worked as it should. I decided to image NGC7000. Although it is near the zenith, which is good astro imaging wise, not so good when I have a SkyShed POD and no zenith table, I did my usual push off the dome and balance on my shed roof, but this time the shed moved and the dome fell off or rather dropped down between the shed and POD wall 😮 - Ooops - no harm done to dome or shed and I now had a full 180deg of sky to work with during the imaging session, but I had to climb in and out of the dome over the skyshed wall as the dome had blocked the door.. all sounds a bit mad writing this up , but I got my image and thats what counts 👍 A bit of leverage and some bits of wood and dome was all back in place and functional the following afternoon.. Anyway on to the image. So this image is 4 hrs of 120s subs taken with ASI2600MC-Pro using an Optolong L-Extreme filter and 1 hr of RGB subs taken though a IDAS-D2 filter. The raw L-Extreme image had loads of detail, but as I am now finding out very flat and mostly red. I used the tutorial https://www.stastrophotography.com/creating-a-hubble-palette-image-from-osc-dual-band-data/ This is a Pixinsight based process, and works exceptionally well although it does take most of the green out of the backgound which makes black look odd - so I used a processed view of L-Extreme as a luminance and added it to the Synthetic SHO with a mask layer and that corrected the backgound reasonably well. I also recently recently managed to buy a Nvidia graphics card (RTX 3060 ) and have used the Starnet 'Hack' to use CUDA/Tensorflow GPGPU processing to remove stars - it works amazingly well, Starnet now removes stars at stride 32 on this 26Mpixel image in about 10mins !! So using this improved Starnet process I removed the stars and replaced them with the stars from the RGB image, in hindsight I think 120s on RGB was too long an exposure and I should have halved the exposure to get better star colour as 120s has rather overexposed the stars and lack the colour depth I wanted to acheive. I am very pleased to get such a close rendition to SHO from dual band filter on a OSC camera Bryan UPDATED : I have watched the youtube video referenced by many by lukomatico - How to get Hubble style prictures from OSC data - Easy Pixinsight Guide This is the result, hard to believe both images came from same base dataset Edited September 24, 2021 by Astrobdlbug 10 Quote Link to comment Share on other sites More sharing options...
TerryMcK Posted September 19, 2021 Share Posted September 19, 2021 A great write up Bryan. Also a good save with the dome dropping off maybe a zenith table as a next construction project. I’ve been after a faster nVidia card for ages so great that you managed to source one. All the bitcoin miners have pushed up the scarcity of such cards and if you can get them the price is at a premium. 1 Quote Link to comment Share on other sites More sharing options...
MarkAR Posted September 21, 2021 Share Posted September 21, 2021 Excellent work Bryan, Stars look well tamed and the contrast really lifts the image. 1 Quote Link to comment Share on other sites More sharing options...
Sonyme Posted September 21, 2021 Share Posted September 21, 2021 The sky looks fractured and slowly revealing whatever is behind it. Lovely deep image. Gary 1 Quote Link to comment Share on other sites More sharing options...
SkyShed Observatories Posted September 21, 2021 Share Posted September 21, 2021 Beauty Bryan! It makes me want to go back and re-process mine! I did a quick pass when I took the shot. Sorry to hear about your dome falling! From time to time we ship Quick Relaease Brackets to the UK/EU for those who make their own Zenith Table. If you send me your address I can cant get a mailing price? CS! Wayne 1 Quote Link to comment Share on other sites More sharing options...
Astrobdlbug Posted September 23, 2021 Author Share Posted September 23, 2021 On 9/21/2021 at 10:18 PM, SkyShed Observatories said: Sorry to hear about your dome falling! From time to time we ship Quick Relaease Brackets to the UK/EU for those who make their own Zenith Table. If you send me your address I can cant get a mailing price? Thanks Wayne, as I said no harm done to dome etc , I’ll send PM. I use this process flow to create the false SHO https://www.stastrophotography.com/creating-a-hubble-palette-image-from-osc-dual-band-data/ But I gave the blue/Oiii a slight histogram stretch to help colour boost. I also found the background has no green so is a very Deep Purple ( Ha!! Space truck’n and so much more) anyway I add in a masked copy of the original L-extreme image to mix the original background to get it Back to Black ! If you use ezDeconv it generates the required mask. Bryan Quote Link to comment Share on other sites More sharing options...
SkyShed Observatories Posted September 23, 2021 Share Posted September 23, 2021 (edited) 12 hours ago, Astrobdlbug said: Thanks Wayne, as I said no harm done to dome etc , I’ll send PM. I use this process flow to create the false SHO https://www.stastrophotography.com/creating-a-hubble-palette-image-from-osc-dual-band-data/ But I gave the blue/Oiii a slight histogram stretch to help colour boost. I also found the background has no green so is a very Deep Purple ( Ha!! Space truck’n and so much more) anyway I add in a masked copy of the original L-extreme image to mix the original background to get it Back to Black ! If you use ezDeconv it generates the required mask. Bryan Yes, nicely done. I use Lukomaticos "Hubble for OSC" (Youtube) w/ IDAS NBZ for f.2 instead of L-Extreme, which many use with higher F ratios. Here's my 2 hour shot. You got more detail and "black". Taken with 14" Edge , Hyperstar, and 26C. Plus I like your "non-halo" stars. A trade off with 120 second subs and the Hyperstar for fast acquisition, is Halos around brighter stars. I'll "take it" though for two hours of acquisition. CS! Edited September 23, 2021 by SkyShed Observatories 1 Quote Link to comment Share on other sites More sharing options...
SkyShed Observatories Posted September 23, 2021 Share Posted September 23, 2021 (edited) Oh! Forgot to say. Thx for mentioning the GPU thing for PI. I went out and bought a got suped-up Nvidia just to do that, and couldn't get it to work as per instructions. A buddy said the same so I didn't try again. I will now that you mention it works for you. 😉 Edited September 23, 2021 by SkyShed Observatories Quote Link to comment Share on other sites More sharing options...
Astrobdlbug Posted September 23, 2021 Author Share Posted September 23, 2021 (edited) 46 minutes ago, SkyShed Observatories said: couldn't get it to work as per instructions If its a RTX30 series then there is a thread on Pi forum https://pixinsight.com/forum/index.php?threads/no-luck-with-starnet-on-rtx-3070.15585/ post #13 has the info I’ve copied below but in summary you have to use updated libraries that support RTX 30 series GPU got it working with the 30xx video cards. Install Tensorflow GPU 2.4 Install Cuda 11.0 (not 11.1 or 11.2) Install Cudnn 8.x for Cuda 11 https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow-gpu-windows-x86_64-2.4.0.zip CUDA Toolkit 11.0 Download | NVIDIA Developerhttps://developer.nvidia.com/cuda-11.0-download-archive cuDNN Download | NVIDIA Developerhttps://developer.nvidia.com/rdp/cudnn-download The windows environment variable then has to point to C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\bin Edited September 24, 2021 by Astrobdlbug Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.