Undersampling is BAD!!.. Or is it?
Vložit
- čas přidán 4. 07. 2023
- Support me on Patreon!: / cuivlazygeek
If you're planning on buying equipment, it will help me if you click the equipment links below!
Jeffrey Horne's Astrobin: www.astrobin.com/users/Jeffre...
Previous video: • What camera and pixel ...
Dither and drizzle: • A match made in heaven...
M101 with details: • I tried Deep Space Ast...
FOV and resolution/Dawes Limit calculator: astronomy.tools/calculators/f...
Sampling vs Seeing calculator: astronomy.tools/calculators/c...
-----------------------
My equipment
-----------------------
Cooled camera: s.click.aliexpress.com/e/_DCY...
SW Quattro 150P Budget Newt: bit.ly/3BABCxl
Backyard Universe upgrade kit: tinyurl.com/3juwm5sz
OCAL Collimator: s.click.aliexpress.com/e/_Dmu... or s.click.aliexpress.com/e/_Dc6...
Askar V Modular Telescope: bit.ly/3Imoq2M
or bit.ly/3GLuyR8 or tinyurl.com/2h758bvt
ZWO AM5 Mount: bit.ly/3BAS8x5 or bit.ly/3JeVUkn
iOptron CEM70: bit.ly/431Ii2U (successor of my mount)
MiniGuideScope: bit.ly/3MyUJhn
Antlia ALP-T highspeed filter: bit.ly/42RBR2e
Celestron C6: bit.ly/3ocePVs
Hyperstar C6: starizona.com/products/hypers...
Mini Computer (newer version): amzn.to/3LQTgTJ and amzn.to/3K6x2vy
AstroPC Pro: astro-gadget.net/gadgets/astr...
Cable management: www.etsy.com/shop/SmartCompon...
IDAS GNB filter: idas.uno/space/en/IDAS/gnb.htm
Flocking paper/Felt paper: amzn.to/3zASgvu or amzn.to/3m94zwa
----------------------
Affiliate Links
----------------------
NEW Agena Astro Affiliate Link: bit.ly/3Om0hNG
General High Point Scientific affiliate link: bit.ly/3lReu8R
General All-Star Telescope: bit.ly/3SCgVbV
Dwarf 2 Smart Telescope: bit.ly/3SyChXu
General Amazon Link:
amzn.to/3hTB5Ne
Follow me!
My Instagram: / cuivlazygeek
My Facebook page: / cuivlazygeek - Věda a technologie
Support me on Patreon!: www.patreon.com/cuivlazygeek
If you're planning on buying equipment, it will help me if you click the equipment links in the description first!
Jeffrey Horne's Astrobin: www.astrobin.com/users/Jeffreyhorne/
Previous video: czcams.com/video/tzhpTDfJvo8/video.html
Dither and drizzle: czcams.com/video/ItmrAZyLeek/video.html
M101 with details: czcams.com/video/lFNyhrGYlVw/video.html
Hi Cuiv! I watched your video on patreon but after watching it again here on CZcams, you are really speaking gold here as there are many out there who still do not understand these concepts. Your years of astrophotography experience shines here in this video and I hope many watch it several times to get a good understanding. Thanks and Clear Skies!
Thanks so much Dave! I love making these videos, and at the same time I'm kind of forced to because of the rainy season haha
Thanks, Cuiv, your videos on sampling have been awesome. The whole “undersampled = bad” advice always struck me as… oversimplifying things, is, I guess, a good way to put it. It always reminded me of when I was heavily into visual astronomy decades ago and the advice was “never have an exit pupil over 7mm because 7mm is the max size the human pupil can open, on average. Any larger exit pupil than your eye pupil is just wasting light.” But what they didn’t understand is that only children had 7mm pupils. The average adult only had a 5mm pupil, max. So if you lowered the power and the exit pupil became larger than 7mm, the image didn’t get brighter, but it also didn’t get dimmer, you just got more field of view. So if you need a wider field of view, just go for it!
Sorry. I will get off my soap box now. That has been bugging me for 30 years. 😂
Hahaha yes on the exit pupil!! I think people want to be as efficient as possible, and sometimes forget about the big picture!
Charles Bracken has a good discussion on why underampling is preferable to oversampling in his book on astro imaging. A good read. The whole book is in fact. Bottom line is don't sweat it, you are unlikely to be under sampled enough to matter with most scopes and modern cameras anyway. But it's easier to outrun your seeing with long focal length scopes and small pixel cameras if you are trying to get the maximum sampling your rig is theoretically capable of.
Great video. Very informative. Clear explanation and easy to understand.
Thank you Rick, glad you enjoyed!!
Great video Cuiv, watched that through twice. Especially useful the bit about the binning. Like you I think some of my images are in both camps depending on which scope is being used. The tone of the video was really refreshing in that you are not doomed to failure, as some posts suggest, if you stray far from the path of ‘correct’ sampling. We all strive for the perfect image but I suggest more often than not it’s not the sampling being the major hurdle that’s holding us back😂 clear skies.
Thanks Stephen! And yes exactly! There is no one size fits all, we are all free to experiment!!
What I tend to do now is what you mentioned earlier especially with oversampling I shoot for the stars using a ircut or light pollution filter then expose for the nebula with a narrowband dual band as I only have an OSC camera, unfortunately I also have nothing but cloudy stars so have not really had much luck!
Hope you can get clear skies soon!!
a really good video Cuiv. This is a topic that is confusing at first and you have broken it down very well.
I have an Edge HD 11 @ 2800 FL and finding a camera that is perfect for that is a tall order. even with the .73 reducer it still comes in oversampled. Pixinsight has some good resampling options which help as my on camera (colour) binning never looks right.
Thanks so much Andrew!! I love these kinds of topics, making things easier to understand AND having practical use :) Glad you've got your workflow figured out!
Hi, thanks for the video :) I did not quite get the part about well depth and read noise in the oversampling section of the video (and how they impact to the exposure time?), more research to do for me to better understand!
Yep, that's a different topic, I have an older video on that, but I might make an updated version soon!
Brilliant explanation - NOW I understand much more. My undersampling complex is now resolved. 😁 Thank you as always.
Happy to help and glad you liked it!
Hi Cuiv, super video thanks, though it raised a few concerns for me, especially when you said that x2 binning is practically useless in a cmos camera. I have recently got hold of an ASI294mm pro and one of the key features of the camera is it’s inherent two times binning that give it a large well depth and read noise as well. It is also ‘unlocked’ now and can be used in bin 1 mode, but at a cost to the well depth and read noise and High gain conversion ability in bin x2. I guess from what you say these binning capabilities are not as useful as the manufacture claims or hoped for?
Aha this is a great question! The 294MM as I understand has separate ADCs for the two binning modes, so it is a rare camera where binning actually does make sense! But it is unique for that model (and I may be wrong, having never owned it)
Hi Cuiv, thanks fo another great video, I use an Altair 150mm f5 newtonian with an Altair Hypercam 183c fan cooled with 2.4m pixels, which is oversampled so I use 2x2 binning with sharp cap and get reasonable images with no loss of colour, my mount is an old eq5 so guiding is not great, I loose approx 30% of subs which can be frustrating, im in a Bortle 5 zone, so it is possible to run oversampled, I'm not a pixel peeper. 👍
Another awesome video! And thanks for the shoutout! 🙂
Thanks Jeff - and your images are already so awesome they hardly need a shout-out :)
Great explanation. One aspect I never see discussed is the effect of stacking. Averaging muliple observations will improve the estimate of the mean, so it seems that integrating multiple subs should improve the resolution over what’s available in a single image.
Stacking on its own doesn't really help (the better estimate of the mean is good for SNR though), but processes like drizzling make use of the partial pixel offsets afforded by dithering+star alignment to achieve what you're mentioning!
Very good explanation, Cuiv 👍
Thank you! 👍
Cuiv, I swear I think you were reading my search history yesterday as I was just looking up information on this topic. LOL🤣
Mwah ahahaha I have eyes everywhere lol
One exception CMOS sensor (as far as I am aware) that can benefit from in-camera 2x2 binning is the IMX 492 (ASI294MM). That's because the sensor has 2 separate ADCs for each of the 1x1 & 2x2 binning modes.
Great video! The Hubble takes under-sampled imaged with yhe wide-field instruments. NASA developed the Drizzle or variable-pixel linear reconstruction algorithm to process the deep, wide field Hubble images. This process essentially paints a higher resolution image of the originals by drizzling "points" or "blobs" of light from the original images pixels to a higher resolution canvas. These blobs hit multiple pixels on the hi res canvas depending on the alignment of each original to the final canvas. If the originals were dithered then these blobs can add up statistically to create a new, higher resolution image. Phenomenal results can be obtained when dithering, drizzling and deconvolution are used together.
i am typically oversampled with my 115/800 frac and the 585 sensor for my bortle 8-9 ... poor to ok seeing
i recently have started to use Sharpcap and take advantage of lucky imaging. i still struggle to get a good enough balance between exposure length and all else ... i noticed that shorter exposures provide way more sharpness so bypassing the atmospheric wallowing and getting to FWHM of 2.1 which is real good! however it seems to always favour the brigher objects in the sky as it suffer at bringing out the fainter nebulosity....
so i am considering a standard approach and stacking the Sharpcap image on top for more details within the core of the subjects
In Bortle 8-9 you should be overwhelming the read noise quite easily. Make sure your gain is high enough for your short exposures so your read noise is low compared to LP shot noise - it will then give you pretty much the same result as longer sub exposures (assuming same integration time)
Loved this video, Cuiv! I've always been a firm believer that undersampling is rarely ever an actual issue, and not nearly as bad as oversampling can be. Love the shoutout to Jeff Horne, too, his work is incredible!
Thank you so much! And yes, Jeff's work is absolutely mind blowing!
Very good description between over and under sampling of images using long focal length scopes and wide field scopes with different cameras. I think you covered the topic very well.
Clear Skies
Thank you!
Thanks for vid, I’ve been trying to understand this over the last few weeks.
Good video. I tell people the suggested image scale range by a certain website is a guideline, not a hard set rule. :)
Exactly!
Hey Cuiv, off topic but what do you think of the JTW Trident astrophotography mounts? Apparently there's a new GTR upcoming too with 30kg weight and can handle the same 75kg load as the P75 trident, no absolute encoder though like the P75 Trident has though.
On paper it looks amazing! But I haven't really seen any reviews...
I was concerned abt this at the beginning, but then I pushed on with a asi1600mm and canon 50mm lens and since then, I’ve gotten really good images. I’m more so going for large regions of nebulosity (like cygnus or ifn bc that’s more so showing the scale than detail) because that’s what I like, even if I’m way undersampled.
Great vid, Cuiv. But I'll be honest, I'm just here for the pix-elles ... Clear ciels !
Mwahaha, go ahead make fun of my ciels :)
Great video! I have found that oversampling can get problematic with processing... the stars for example might be so large (9 HFR) that the software just does not want to detect them. Have you ran into this at all?
Ooooh no I haven't run into that - there are star scale detection settings that should help though!
Hi Cuiv! I am quite new in astrophotography, about 3 months now.
I have considered binning on my setup. Because i have 0,48 arcsec/px and I do unguided imaging with a TS RC8" on a IOptron HEM44EC.
But I have not tried it yet, because i can get 2-3 minutes subs, depending on the polar alignment.
I think I will try it, then I can get subs with even more than 3 minutes. What is your opinion on that ?
Greetings from Austria and Clear Skies!
If it's a monochrome sensor, go for it!
I have a "properly sampled" setup (C8 SCT + 0.63x reducer + ASI533MC Pro) for OK to excellent seeing (as per astronomy tools website). What I have been reading on some forums is that if I debayer the images using "SuperPixel" (AFAICT, this method reduces the resolution of your images but is more correct as it doesn't do any pixel interpolation), I can take advantage of drizzling and get a better result that using VNG with no drizzling.
Do you happen to have tried or read something about this? Right now I think there's a bug in WBPP that avoids this setup to be easily tested (when superpixel is selected, no matter which drizzle conf you ask, it'll default to 1x), but I guess it could be tried manually, I haven't yet.
I believe you could set debayer to bilinear, drizzling to 1 and square 0.9 and it would achieve this, if I recall correctly
Did I miss something on mosaics? Is this a way to get wider field with acceptable sampling?
Yes, mosaics can of course achieve a wider FOV while still having all the details - I'm lazy though, so rather than do a mosaic (whose point is kind of look at an object as a whole), I'll rather use a very undersampled setup!
Hi Cuiv, could you explain your statement "binning is basically useless" in CMOS??? Dont you improve your SNR by binning pixels? And also allows for adapting your (pseudo)pixel size to seeing, for example ?
Great video again, thx
You can do the same thing more flexibly in post :)
Best Astro tuber out there right now
Thank you!
I recently bought an asi553mc pro, and I'm struggling to decide on a main scope to use with it (currently I have a Skywatcher 72ed and Askar FMA 180). These are my two potential choices:
Askar FRA300 @ 300mm focal length and F5
Sharpstar 61 EDPH iii @ 270mm focal length and F4.4 w/reducer
Not discussing any other elements of the scopes, but the FRA300 would give me a pixel scale of 2.575, and the 61 EDPH iii would be 2.86. This seems relatively close in my mind, but what are people's thoughts on how much different my results would be in terms of undersampling?
It's honestly not that huge of a difference, so with a 533 I'd go with the Sharpstar 61. Note that I will soon be testing and reviewing a small F6.3 scope from Founder Optics!
@@CuivTheLazyGeek - Kind of the answer I was hoping for, as I hike all my gear around the city by bus/walking to get to darker locations since I don't have a car or backyard, and the Sharpstar is a couple pounds lighter even with the reducer. I think for my use case the faster optics may also end up being a bigger benefit than the higher end build quality of the Askar.
Thanks for your feedback!
Great video!
Thanks!
Great explanation Cuiv! I have always been aware of this and its never bothered me in the past, but I have always noticed when zooming in on my RedCat image being undersampled they do get blurry quickly. I am now drizzling which does improve things quite a lot. However I have another system that's properly sampled in good seeing and over sampled in poor my seeing its usually good so its definitely there of over sampled and the penny has dropped here when you said you can zoom in further as I captured just 3 hours on M51 and had to do a severe crop as I skipped the flats, but the details were so sharp even with minimal integration. I didn't put the two together here until now watching this so thanks! Here is my takeaway: Undersampled "typically" looks good zoomed out widefield images. Oversampled "typically" looks good zoomed in which I think is beneficial as you can then pick crops that others my not have tried! Thanks
Good takeaway, and glad the video help crystallize this knowledge you effectively already had!
So with undersampling. Can you just gather more data to make up for the lost details or go to a darker location? Or are they gone forever?
Nope, more data doesn't help -you'll only get better SNR, but never more resolution!
If undersampling with the intention of dithering to drizzle the data I assume you have to increase your guiding accuracy. If undersampling at 1000mm and then doing a 2x drizzle you end up with an effective FL of 2000mm. Any errors in your tracking that were hidden in the undersampled data would be revealed.
My head is spinning, but I grasped the comparisons of undersampling vs oversampling. I finally get it!
Glad it helped!!
My astro camera has tiny 2.4µm pixels and I've been oversampling on everything except my widefield setup. 😅 What a headache of a way to begin the hobby!
Hahaha you're starting the hard way for sure :)
I am huge fan of oversampling - 0.3-0.5”/pixel. Only seeing is limit.
Sounds like dithering is a good idea
Dithering is ALWAYS a good idea :) I see it as compulsory if the equipment is capable of it!
Rule of thumb in digital signal processing (in general): it's way better to do undersampling rather than oversampling.
It's definitely an approach :)
I am not stupid (ok maybe I am it's hard to be objective about your own intelligence}, I have an engineering degree, this still sounds like word salad to me. I wish I could find an understandable explanation of over sample and under sample.
As mentioned the real explanation is in the previous video :)
@@CuivTheLazyGeek I will go back and watch it again.