For proper dark frames. "The key thing to remember about taking dark frames, is that the images must be the same exposure length, temperature, and ISO as your light frames (the picture files)."
@@antdx316 40 is ideal, 20 at least. I am not convinced that taking dark is a waste of time, I always take dark and bias frames and I never have banding in my images.
You can also take one exposure for the stars and one for the foreground, when shooting the foreground reduce the iso as much as possible and shoot a very long exposure to get no grain and more light, expose for the stars normally and take multiple images and run them through a star stacker program then blend foreground and bacground images in photoshop
Tested this on my Nikon D5300. Lens cap on and wrapped in a blanket: - Test 1: ISO 500, F4, 20 sec = Same purple glow at bottom, but as well some horrible yellow/orange round dots splattered around the sensor, most from center towards the right. - Test 2: Went into menu, did a "Clean Image Sensor", then performed above test again: ISO 500, F4, 20 sec = All the yellow/orange round dots were gone, though the purple band at the bottom remains. - Test 3: ISO 6400, F4, 20 sec = Without increasing exposure nor whites, one can notice a very faint purple glow at the bottom. When both are increased picture is very uniform, and somewhat clean. So learned that I must clean the sensor from the menu, much more frequently due to this, thank you! And lovely channel by the way. (.. Subscribed due to this)
I feel like your testing is flawed. Why did you take first two tests at iso 500 and then third one suddenly at 6400? Also, you should try longer exposures. I just don't see how the sensor cleaning function changes anything as what it does is it shakes the sensor or IR filter which is totally unrelated. Also, most cameras do it automatically when they're turned on and off.
The dark frames need to be taken with the exact same settings (ISO and exposure) as the light frames, or otherwise you need to scale them to match them. From what I saw in the video lights and darks were taken with different settings, which explains why the noise was not completely eliminated. Differences in temperature also change differences in noise levels, which is why professional observatories (and more serious astrophotographers) have cooled, temperature controlled CCDs.
You can also take a "dark frame" and substract it from your images to get rid of regular sensor noise which is what the purple glow is.
4 года назад+7
That's exactly what I was thinking. These dark frames are all repeatable(the important bit) characteristics for any given sensor, so couldn't software 'subtract' that (in a user-controlled way if possible) from the light frame image. Daft question (I'm new to this very fascinating subject) - what software did you use to subtract the dark frame and, of course, did it actually work!
I was thinking the same thing, however, if you do the light sensor test image, you could use that to remove the bands in photoshop as a inverted filtered image over, to neutralize the banding.
Peter your a darling. Just checked my 6D, it's not green it's pinkish but with next to no banding on the sensor. Very even throught out. Thanks mate, you just restored my "faith" in the 6D that I bought for £350!
Great video. Awesome to see how the different sensors perform. You may want to do some reading on why each of the different types of frames are taken, and the difference between random noise and average noise. These can both be applied to shot noise, read noise, thermal noise etc. Random noise in these categories will only be benefited from more light frames as you concluded from your pragmatic approach. The sensor read noise you keep stretching out would have to be dithered out. But these are technicalities. Overall if someone is doing wide field, the best bang for their buck (time) is more light frames. There is enough light in a wide angle Astro (landscape or deep sky) that with fairly limited light frames the amount of stretching required to balance the photo won’t bring out the problems, but obviously needs more than one shot. With higher focal lengths and much dimmer objects, the stretching required is going to bring out every little problem hidden. In that case, understanding all the different kinds of noise is ++helpful to make some of the best shots. Clear Skies! Thanks for making these videos!
I have that exact band on the bottom of my D750 images, I wondered why I was getting it, is such a pain! i just checked a dark frame and can see the correlation. I just got a z6 so will have to test that out! Thanks Peter, very useful info! My Z6 sensor is silky smooth and clean as!
First of all let me say this: AWESOME video, what I like especially much is that you have presented a ton of examples of different shots, flat, light, dark/offset frames, from different cameras. (Offset for me is a very quick dark frame, so sometimes they're mixed up by ppl) Nevertheless, I wanna make some comments here as well. One of your major points in this video is that the banding/spots of color/lit corners remain no matter if you use a dark frame in the stacking process etc. So this is kinda true, and kinda obvious, but in fact you are committing a slight maths mistake. Let me try to do some short talk to present this. 1) SNR: probably everybody knows best about SNR here, so true: more light -> more signal -> read noise, electric noise, thermal noise contribute to a lesser percentage -> better SNR 2) but what happens to signal/background when stacking images? well 3:1 ratio + 3:1 ratio + 3:1 ratio = 3:1 ratio. It stays more or less constant. that is the main problem with stacking and that is why you need the background. There is a slight difference between signal/bgd and signal/noise. that's why we need to subtract the background for. 3) how to estimate background noise and what is the problem? 3a) if the background had zero average (which is completely not the case in photography) then by doing many many experiments we should come closer and closer to zero (law of large numbers). but wait? how does a dark frame look -> it is a collection of pixels (and 3 channels) from which each has a NON-ZERO average luminosity and some deviation from it, which is random, or so we hope. 3b) thus by taking a humongous amount of frames (try it, number goes to 1000 rather than 25) you establish a map of the AVERAGE VALUES of noise in each pixel and channel. Maths: the estimation of the average is accurate to about sigma/sqrt(number of tries); so for 100 tries you have only 1/10th of the initial variance in noise luminosity, for 1000 tries you have 1/30th, sort of, of the variance, and only for 10000 tries you get down to 1% of the original noise variance. 3c) the same estimation rules apply to flat, dark, light frames etc 4) so now we know that by stacking we don't change the SNR, tho we hope that after stacking 100 pictures we have the average noise (times 100!!) in our background (coz how could it possibly go away) and we can try to bring it's AVERAGE level down, it will NEVER reduce the grain coz this is the randomness that you would need thousands of shots to remove, and even then by exposing the interesting part of the histogram really much you'd amp it up again! 5) you are using the exposure slider "just so", don't really putting care in how much you slide it, so you don't notice the decrease in AVERAGE noise level, but you still notice the band, coz it's there! you cannot reduce background noise this way! it's just smaller on average but has the same shape, and you can always amp it up again and show it on screen. 5a) overblown example: suppose you have taken 1000 offset, dark, flat frames each (what a titan). and then you took just ONE light pic. The light pic has a RANDOM instance of the noise, and when you subtract/divide etc by your calibration frames you bring the AVERAGE level down, but not the actual variance. So the random instance of the noise is just there, it's average has only been brought down to almost zero. 5b) if you took 1000 identical pix (hard to do of course -> triple titan) you'd bring the noise variance down a lot and by subtracting the stacked picture you'd get a comparable noise variance level, but if you exposed it HARD ENOUGH you'd still see the same banding (tho quantitatively it now constitutes less of the signal) 6) to sum up: by even saying that you should expect that band to go away, you are making a statistical mistake, you can never get rid of that banding etc, you can just reduce the standard deviation around the average by averaging it many times, same goes for any other calibration frame. sort of encapsulation try: (sig+noise) x100 stacked = (same snr) 100x signal + (random_noise x100 different instances summed up) = 100x signal + 100x average_noise + 10x noise deviation (sqrt(100) x orig deviation)! (same average snr 100sig/100av_noise) ( noise ) x100 averaged = av_noise (10% of noise variance as compared to 1 frame)
now we subtract 100x average noise from stacked pic -> we still are left with 10x noise deviation! the band are actually more visible and easier to expose! pre-final result: 100x signal + 10x (one-frame) noise deviation! so if we are ok with 1x noise deviation level, we can then DIVIDE by 10 and we get final result FINAL = 10x signal + 1x noise deviation so we converted a godzillion calibration frames and 100 light frames into a ten-tuple signal with the initial noise variance, thus all be the bands etc. will be visible when sufficiently exposed.
MORE SNR, yes but only on average!, grain is EXACTLY THE SAME, but could be 10 times more wild! So what you say is dead true and kinda obvious and I want to add that by stacking (adding) signal you INCREASE the noise variance, and you are only able to get rid of its average!
I hope that wasn't just mathematician's/theoretical physicist's mumbo-jumbo coz I am a big fan of channels such as your and astronomy! If that seems instructive and has helped you and you want me to clarify more and are interested in actual computations I will be more than happy if any1 contacts me on my channel or later via facebook/gmail. Clear skies! EDIT: I used variance in the common sense, as something being changeable. in fact variance = deviation^2 for a mathematician, so I edited back to "deviation" as a more accurate wording.
To show how averaging works in pictures I prepared examples with controlled exposure bumping: drive.google.com/open?id=1xQH6w4xnAI9D-AZOl0hRheu98h4BmKk0 drive.google.com/open?id=1kIAkedV-q9QxZc_kktPxI4YZShKhWHX6 The bands, splashes and noises are now virtually out of the Master files, but if they were present in the stacked light frame, they're just there.
Interesting video. I discovered something quite interesting regarding banding. I have a Canon 80D and have been using it for DSO imaging for the past couple of years. Out of all of the images I've taken I've only ever seen banding once, in one stack of images. I couldn't understand what was causing it as I'd never seen it before. After doing some tests I realised what it was... I use APT for imaging, and I always choose to save my images to 'Card and PC' in APT (obviously saving it to the camera card, and my laptop hard drive). But for this particular imaging session, on M81 and M82, I accidentally set it to save to Card only. The resulting 7 hour stack had these purple bands across the final stacked image. I then realised the only thing I changed was how I saved the images, so I set it back to save to PC and Card, and since then, and before then, never seen any banding at all. I read somewhere that it can be something to do with saving images to memory cards in the DSLR and that's exactly how it was!
Couple of corrections: you need to take dark frames at the exact settings of the photos you intend to correct ( and possible even at the same temperature) so that the noise pattern is consistent to the one you get in the light frame, otherwise you end up under correcting or over correcting the noise in the original image and that can cause problems. Another thing, what you see over Orion are not Starlink but the "usual" geostationary satellites that orbit that area of the sky, there are so many because it's a special orbit. Starlinks in orbit now would probably appear brigther than that.
V0LDY, it cannot be geostationary, because those do not move accross the sky as they are - geostationary, meaning they are always above fixed point on the Equator.
If you don't have a star tracker, a low cost solution is to take an image series of shorter exposures (where the sum of all exposure times is your total exposure time, i.e. for instance 12 images of 20 seconds result in 4 minutes total exposure time) and stack them using an image stacker like Siril. The advantage in this approach: one bump to your camera during a 4 minute exposure ruins the whole image. With the stack you can sort out the bad apple and still get a decent looking stacked image out of the n-1 images left.
Also you really do need to match settings with your lights. The noise pattern can change noticably at different iso and temp, especially if your camera is dual native iso or highly iso variant.
Dss works really well in terms of dark, flat and bias frames as compared to sequator. I don't have a tracker. I manually track(it damn difficult)... Taking more lights really works... I took 75 lights on first attempt on m42 and then 150 lights the second time... And i can say the second time the results were really awesome... Those dark frames really help in decreasing the noise and mainly useful for stretching the final stacked images
Hey Peter, interesting video as always. Agreed with Voldy that dark frames must be taken with exact same settings (and temperature/humidity etc). Usually do that when packing up at the end of a session. Also, in your example (1:47 vs 2:31) you are comparing single exposures and for those, you are quite right that the more light (exposure time) you get, the better Signal-to-Noise Ratio (SNR) you have. On single exposure, sensor and images are usually not that good in terms of SNR and if stacking doesn't increase the amount of light received, it dramatically increases the SNR by removing/averaging most of sensor/atmospheric noises. That's what allows in post processing to enhance so much the image compared to a single exposure, this large increase in SNR. IMO, I think your comparison of image quality when increasing exposure/white would have been fairer if you compared 13 stack of 20s vs the 260s image. But FYI, there was a test by another astrophotographer (possibly Alyn Wallace ?) who compared the difference of stacking vs single exposure (but with same total exposure time). Single exposure was showing indeed more details as they do get more light, but the difference was much more gentle (and acceptable) than what you showed. Clear skies !
Good stuff....Even on a dark frame w/cap..... the sensor only sees what the lens allows it to see. so taking the lens into consideration is a must. Change the lens to find the best lens to camera combination also.
I'd like to say thank you. You have gained a subscriber and provided useful information. There are numerous videos on each and every subject out there. It can be true that the overwhelming majority does not provide useful information. Many of which are popular channels. However, through digging around or quite possibly appearing on your recommended page lies informative knowledge. You are an excellent representation of just that. You got straight to the point and did not overly represent your channel with a couple of minutes of essentially stating like and subscribe. No, you provided usable knowledge in a relatively quick video that I will actually use. You were understandable, knowledgable in this subject, and the title held to be true throughout the extent of the video. It's uncommon to find channels like this where you truly appreciate the content and their hardworking creators. Your an example of exactly that. Keep working hard, and thank you.
Possible solution to the banding problem - Take a few photos then rotate the camera on the tripod by a few degrees and recompose the photo with the important feature in the center again. Repeat this several times while taking the whole set for the stack. This will require using an app that can align the stars before doing the stack. This might work because all the banding will not be close to the same positions. Averaged over a larger set, if the bands are in six directions, each band should not contribute as much to the finished result. Disclaimer: I am just getting into night photography (generally too much light pollution, haze, and clouds where I live) and my camera is not a Canon. Thanks for the video. You give us much to consider.
Excellent video Peter. I wrote to you perhaps about a year ago as I saw this image of your SUV parked in a dark sky site with the Milky Way sky in the background. You had displayed it in a popular Facebook adtrophotography site to explain the importance of longer exposures. I had to concur with your advice, as an astrophotographer since my early teen years in the mid 1960's using crude consumer-grade cameras, actually my parents Kodak Box camera, starting in about 1956 as they handed it to me at around age 2 to take phototograghs of them on my birthdays. I'm backtracking here to the Dark Ages, so to speak, for most of the younger generations now entering astrophotography with digital cameras today. If only to make the point, that I dropped all the photography about the time I left High School, to instead become an art major in college in 1973 at the age of 18. I actually also purchased a new 35mm Mamiya/Sekor camera and took Photography and darkroom courses to learn how to process the wet-film-base photography that we were are all using long before digital photography came along a generation later. This may be a rather lengthy comment to leave but I later learned in recent years why I admired the great film directors, say out of Italy, or other foreign film directors such a Stanley Kubrick who left America to live in England the remainder of his life, as many people know he's the director producer of the great film 2001: A Space Odyssey, that is still admired today for it's groundbreaking photographic work. The early motion film directors too, had a similar childhood process of attrition as their parents gave them advanced photographic film cameras when they were very young. Many great figures in the history books today, that are known for their hand drawn and painted artwork also started out in photography and hand-drawn art work. As I began producing very large pastel sketches, starting with the sun, produced from live observation, with h-alpha filtered solar telescopes in 1999, during the daytime, and then the moon at night, also took me away from the photography for a while as I contributed these to the Cloudy Nights Astronomy Forums for competition in sketching, which I held the records there, for quite a while, after these large hand drawn pastel sketches first appeared in the NASA websites, Spaceweather several times, and Astronomy Picture of the Day. Now Sky & Telescope magazine is asking to display some of my large hand drawn pastels of the Moon, which will appear in the upcoming May 2020 issue. So finally today discovering your video here, I'm so appreciative of the fact that you explained in fine detail, the failure that many new astrophotographers make by under exposing their images and wondering why they have so much noise and color banding, even after using stacking programs and shooting dark frames, in hopes to cancel out the visual noise and color banding errors. To this day I'm only using a Sony NEX 5r consumer grade Mirrorless Camera with APS-C sensor and an older Sony NEX 5, that I photographed the famous American Solar Total Eclipse with in 2017 in the Oregon desert through my Celestron Nexstar 5i Cassegrain telescope, not far from my home in Portland Oregon. I was asked by the Oregon State Park Administration Directors to give a few lectures to the public in a large State Park about an hour drive from The Painted Hills Oregon for a few nights before the actual Eclipse occurred on that Monday morning August 21st 2017. I also produce very large panorama photo stitches of the Milky Way over beautiful dark landscapes such as at the Oregon Coast Beaches. As you may know, that these large panorama's, up to a hundred photographs in each finished image, can be reproduced into very large prints without any resolution loss up to perhaps 6, or even or 10 feet wide. Where most photographers today will only display their work as digital online images that are only viewed in cell phone screens of only 2 by 4 inches. I'm also employing my old world camera lenses from the 1970s and 80s adapted onto my newer Sony digital mirrorless camera to allow much more light through, say, an antique 55mm f/1.2 lens, so this goes without saying, in taking your advice on initially capturing more light, with only the most modest equipment available, that of a free camera lens from my college days. Many people are now discovering this by utilizing these huge old and fine quality camera lenses from the 1960s, 70s, and 80s, found and rediscovered in their parents closets, garages, and attics, collecting dust, and adapted onto their new digital cameras to capture the most light from the Milky Way. So my subjects of astrophotography, starting from the deep dark sky of my teen years at our parents front yard in the far Northeast edges of Portland Oregon under fairly dark skies with my parents crude Kodak box camera strapped to the top of my Tasco 60mm refractor, that I bought at age 13 with my berry picking money, in my first efforts to take a hand guided while through the eyepiece, as a 5-minute exposure of the Orion Nebula, which I still have that negative today somewhere stored away, and now ranging to the bright solar sunlight of the solar total eclipse a couple years ago, with my modestly advanced digital cameras attached to my modern equatorially tracking motor-driven Celestron Cassegrain telescope, that was given to me for free by the Celestron company, as I was invited to teach astronomy overseas in the South Pacific in 2004, which are all still my most modestly advanced equipment that I have today. I've often thought about taking the leap to the newest astrophotography cameras, like the ZWO that you mentioned in your videos, but living now in retirement at age 65, on a diminished Social Security, and retirement pension income, I've trimmed down my living expenses, to be satisfied with this very modest astrophotography equipment, because my greatest efforts for the past 20 years have been dedicated to the public, in providing thousands of hours of sidewalk astronomy which I later spoke about on National Public Radio many times, and I've been covered in other news articles Across the Nation and overseas in the Fiji islands later, about this service that I provide for the public and schools students for free, still today. If this isn't enough of a spectrum to tackle over 50 years, I also became a very active sidewalk astronomer for the purpose of Public Service, to the public community, spanning the entire United States and Canadian continent in 2000, when I borrowed from our astronomy club in Portland, a large h-alpha solar telescope, taking it onto a 10,000 mile cross-continental solo road trip over six weeks, to allow thousands of the public on sidewalks, to view the Sun's solar sunspot maximum and huge solar flares, safely, and directly, live through this instrument. I audio recorded thousands of responses from the public and assembled them into a best excerpts 15-minute video in my RUclips channel, as my top cover page displayed video. Three trips a few years later in 2003~2004 to the Fiji Islands, I did not record audio of the public responses but only relied on their hand written responses, in a large guest registry book, which now includes over 3,000 handwritten comments in many different languages and locations across the United States and Canada, and also in the South Pacific Fiji Islands in 2004. All these ventures are covered in my Facebook albums, with many candid photographic images on location. Also some of the higher resolution photographs of the sky, I have uploaded into the famous Digital Photography Review site gallery. Today I'm sharing your great video here with others that are getting started in astrophotography because you've covered so many great points that they need to know about basics in a proper Time Exposure. Thanks for posting and providing this.
the dark frames from my fujifilm xt2 is a lovely uniform green like the canon 6d! was actually really worried it would have a terrible sensor. i havent gotten into astrophotography yet but now i am definitely reassured that i have the sensor capability! thanks for the great information!
Hello. I have a X-T2 and I'm going to get a X-T3. Which stacking software do you use? Because DeepSkyStacker doesn't work with the RAF files. Thank you.
To do proper calibration on your lights you need to first integrate your bias, then calibrate your darks with your bias to remove the bias signal from the darks, without doing that you will get the issues that you are displaying. The problem is that you are not handling calibration properly. Everything that you have displayed as problematic can be fixed with doing proper calibration. I also suggest using pixinsight. Great video, but you really should learn the way to handle sensor issues. Inside pixinsight it a great book to start with.
Yeah, PixInsight has a lot of great features! However, my core audience is not using PixInsight. That's for the hardcore folks that really want to take things to the next level. It's a steep learning curve just getting used to that interface, let alone all the advanced concepts. My goal is to teach people how to create great images with what they already have, and are familiar with - Lightroom, Photoshop, DSLR, star tracker, telephoto lens. There's plenty of other channels that will focus on PixInsight.
Longer exposures won’t get rid of banding. As you said it’s baked into the sensor. As others have been saying, dithering is the only way to get rid of that horrible banding.
Those are Geosynchronous satellites in the photo of the Orion Nebula (M42). M42 is almost right on the celesital equator, and is the primary orbit that the geosynchronous satellites get launched into. In fact, you can prove this if you measure how many arcseconds long those satellite trails are. The trail length will be close to, if not exactly the same value you get if you were to multiply your exposure duration in seconds by 15 (because the earth rotates at 15 arcseconds per second).
I use the Pentax K-1 and is so amazing camera, the sensor have astro tracer, means the sensor ibis follow the stars without use of a star tracker and can make long exposures from more then 2 minutes, i'm so in love of my pentax K-1, really recomended camera for astro photography
that is absolutely amazing. i was just sitting behind the computer the other day thinking, stacking cannot possibly be the only way of getting good results. and sure enough i find this video of you doing this extensive test and thanks for sharing the result. this is pure photography. spend more time in the field and get the right shot rather than wasting so much time in post processing.
It’s not stacking that’s the issue, he is talking about dark and bias frames if you don’t need them, a part of most true astrophotography is always enhanced or at least stacked, even when done with dedicated Astro cams and serious telescope rigs, at its simplest form, it’s essentially just stacking multiple images you took yes? bringing the image into much greater view, this guy (great photographer I must add) created these really “popping” Astro images by stacking and post enhancements, like these ones at 18:10 for example. stacking is simply stacking images taken, it doesn’t really take away anything to with it being pure... it’s a crucial part of the majority of most exciting Astrophotography you always see.
@7:00 coming from a Photoshop approach to fix the sensor noise issue, in theory (I haven't done this my self but will try it soon), you could clean up the noise by importing the 'dark frame' and main exposure into Ps: Place the dark-frame layer over the main exposure shot and set the dark-frame blending mode to one of the 'canceling ' filters : Difference, Exclusion, Subtract or Divide.
I just did a quick-and-dirty test of my X-T20 out of curiosity; I took a photo with the lens cap on at those settings. Unusual result. There’s a light leak in one spot on the left, not sure if that is the lens cap or very dim internal light (I know some cameras can’t be modified for Astro because of internal lights), I’d have to test without lights in the room. Secondly, while it has an even flat noise profile overall, no colour noise, banding or light/dark spots, there’s a box in the middle that is lighter. Most unusual. It has never turned up in any photo though, even with heavy Milky Way editing (neither has the light leak, which is most likely the lens cap anyway)
Hello. I have a X-T2 and I'm going to get a X-T3. Which stacking software do you use? Because DeepSkyStacker doesn't work with the RAF files. Thank you.
Btw, I did another quick test after, turning off the lights as well as a lens cap and putting it under my jacket for the exposure to cut out all external light. No internal light leaks, which is comforting. The overall noise profile is very pleasing, no colour noise or banding, a prime reason for getting my X-T20 over the Canon M (I have a Canon 80D, so seemed logical at first), the other main reason are lenses. Canon hasn’t supported their APS-C mirrorless system well, despite it being around for years, and Fuji’s primes are just so good. If I was looking at FF, my choice would have been different no doubt, but neither Canon nor Nikon had FF mirrorless at the time. Despite the X-T20 not being designed for night photography (I bought it for walk around street shooting; the small body doesn’t make it as easy at night with no light), I have tried it out twice with excellent results. Great dynamic range too for an APS-C camera.
Juan Pablo Revert - Process your Raw files to Tiff first; use 16-bit Tiff, not sure any stacking software supports Raw. You want all your files with the same basic processing (white balance, lens correction, chromatic aberration, etc). Only after they have been stacked do you worry about stretching the file; I’m not an expert, I’ve only done night scapes, and used either Photoshop or Affinity Photo to process. There are plugin packs for Photoshop to help develop Astro photos. Some great tutorials here on RUclips, but they can vary a bit in their technique. As I was doing landscapes, I used Starry Landscape Stacker on the Mac.
Peter, great video! Stacking does work, that's the only way people who use telescopes and dedicated astro cameras do it. For the stacked image you showed, you need to stretch the images first setting white and black points with thresholds. Bring all the channels (RGB) to the same level (let's say 40 for each channel and set your white point to ~ 240 just a bit less than 255. This will neutralize the background. Next use curves to stretch the image (use a traditional Astro curve). Go back to thresholds, reset the levels so they are all the same then back to curves. Do this a few times and your images will look very well. Tip: doing this with a star tracker will allow you to create "deep" deep sky images. But, great video!
Yeah, I guess my point is that dedicated astro cameras and DSLRs are different, and require slightly different workflows. What works well for one camera, might not work as effectively for another camera. As we saw, Canon has terrible banding problems while Nikon has bad "amp glow". dedicated astro cameras tend to have even worse amp glow, based on what I've seen. A "one size fit's all" approach is therefore not really applicable for all cameras. Yeah, I know. I didn't want to get too sidetracked though, so I didn't include that method in this video. I like using that method, or even the Black Point dropper in Curves. It usually does a good job, and takes a fraction of the time.
@@PeterZelinka Many dedicated astro people do use DSLR cameras (you should check out Trevor at Astrobackyard) and the CMOS market for astro cameras has really taken off. As I'm sure you know CMOS astro cameras are essentially DSLR cameras only with cooling. With all that said, stacking is a great way to beat down the noise and dedicated telescope astrophotography people use it 100% of the time to image deep-sky objects. You make wonderful content and I enjoy your channel that's why I subscribe... keep up the good work.
Using multiple lights for the total exposure rather than a single long exposure, and darks, bias, flats, dark flats, stacking done correctly using proper processing workflow and astrophoto software like DSS rather than photoshop or Sequetor) will help. I might have totally misunderstood what you were saying, but you seemed to indicate that you ADDED the stacked darks to the stacked lights? That would simply DOUBLE to effect of the dark (noise). Proper astrophoto processing SUBTRACTS master (averaged) dark from each light being stacked, to remove the typical noise. Similarly, master bias is used to adjust for sensitivity of each sensor pixel, and the flats remove dust/blemishes from the lights (but have to be taken close to when the lights were taken, as dust moves around between sessions). And flat darks are used to minimise noise impact on the flats (have to be done separately from the std darks, as the exposure used for lights is different to exposure used for the flats). ps. You don't just stack lights and darks, you use to appropriate software to subtract the master dark from the stacked lights. And also correctly process re the master bias and the flats/dark flats. Perhaps you need to use proper astrophoto processing software rather than just some random std photo processing software? pps. darks won't fix noise problems if the noise varies over time for the same temp, exposure etc settings, but in reality the noise for a particular sensor is fairly constant over months or years (so often astrophotographers build up 'libraries' of master darks for different exposure times and ISO setting, so can reuse matches master darks for subsequent light sets, and not have to always take a set of darks for each set of lights (as time may not permit).
It’s a physical limitation for camera sensors and basically all sensors. The higher the gain, the more sensitive each pixels are, to the point where it starts to pick up onboard circuitry(heat) and other random things. I’m guessing that Nikons have heat coming from below the sensor, while the canons are a result of the dual pixel circuitry. The only way to reduce the noise is to have better algorithms.
Yes, those are definitely starlinks. I saw 16 or more of them back to back in two parallel lines with binoculars at around 4:15 am a couple of months ago and they can ruin a photo.
Excellent technical explanation using the dark frames. I immediately took some dark frames with my Sony A7RIII and Canon EOS R5. They both have areas of splotches and banding but the Sony is much cleaner. Great stuff Peter, thank you!
1) Those are not starlink satellites, Those a geostationary satellites (mostly weather) 2) Dither, even if it's one axis, Dithering in RA helps heaps, Any imperfections in PA will take care of DEC dither. 3) Darks and bias frames help with dark current, but not completely remove it, that's where dither lands the final punch. 4) If you want REALLY clean shots, Do all that with a Cooled astro camera adapted to DSLR lenses.
Brilliant....thank you.....and you didn't push the point too much....it was perfect....a true clear lesson.....and haven't done any astro yet....but wow....thanks for the heads up....well done !
Very Detailed explanation about light and dark and why you should take them, THANK YOU FOR THIS VIDEO! i will carry this information into my next astro shoot.
I tested the sensor on my D610. It looks gooood, also as flat as the D6's sensor from your video, except a small triangle orange-ish shape top-midle and one in the top-right corner. But barely visible even with the exposure set to max and the whites cranked. I am pleased with the results, thank you. I didn't know these things
Thermal currents are important when calibrating a sensor. Remember that heat or thermal temps can trick a photo-site on a sensor that a photon struck that pixel. More noise can be present with higher temperatures. Calibration frames require that your light frames are at the same temps as your darks, bias, or flats. Good calibration has nothing to do with what brand of camera you use. Some CCD or CMOS ships have whole dead rows and can be calibrated out. Astro-imagers usually use thermal electric coolers(TEC) to control a consistence temp of their CCD or CMOS chips. Do this experiment with any camera. Go out on a cold night let the camera cool down for 1 hours or put it in a freezer. Take a 5 min dark frame. Let the camera warm up for a few hours and do the same 5 min dark. Compare the cold 5 min dark to the warm 5 min dark. More noise will be present with the warmer temps. If you are calibrating a light frame that was taken at 10C and using a dark frame at 30C. In the calibration process, its will over subtract and add noise to a light frame. If even temps are not solving noise issues, it would then be the number of samples. The more samples of darks, bias and flats create better "masters." Masters are just many dark frames averaged together.
Thanks a lot Peter. This was very informative. On some cameras, some of the peripheral blotches can be fixed with specific camera settings or minor hardware modifications. My canon 6D was leaking light both from Life View and from the red led indicator, leading to horrible magenta and red blotches when stretching long exposure photos. The purple blotches were removed by disabling silent LV in the menu. The red blotches by opening the camera and putting some black silicon on the led.
I remember my 5dIII also having obscene amounts of magenta blotches + banding; Just did the "dark frame" test on my 90d - It looks like your 6d image (even/ NO BANDING!!!), so could be worse! Will be interesting to redo it at the end of the night when theoretically the senor is warmer after a few hours of timelapse work. THank you!
Thanks for the video. The discussion of the dark frames is really very interesting Not the one with the hot pixels, but the one with the banding and the light and dark areas. I took my Fuji X-T20, made a dark frame and was pleasantly surprised. No banding, no differently bright surfaces. A homogeneous surface. Purple, but homogeneous. :-)
Cool video! I do deep sky AP but love shooting time lapses of my scope as it images the night sky behind (i include those time lapses in my videos). I use an a6400 and want to try milkyway shot in the summer! Clear skies!
Not that it matters much after almost two years, but you did say cameras of the same make/model can act differently in these tests. I mirrored the test with my 5D MKiii (the worst example you've seen) and it definitely was closer to the 6D with it's uniform grain pattern/structure than it is to the 5D MKiii example you provided. I had a little color variation on the right side of the sensor/frame but it was a smooth transition, with virtually zero banding. I did ISOs from 800 to 6400 with exposures from 5 minutes to 20 seconds. All similar patterns. I also did f2.8 (if it matters) after mimicking the 5.6 1600 for 120 seconds. Very clean pattern(s). (Lightroom classic and I also torched it all the way to +5 exposure in addition to the +3.85 in your example and pushed the whites further) Cool test though.
You could always check out the website" photons to photos" this will give you a heat map for practically any sensor of the last 15 years also will give you optimal iso setting to use. Or just buy a CCD camera preferable mono then add a filter wheel with individual R,G,B filters, your signal will be amplified many times.
Totally on side about not listening blindly to what you hear online. It would have been nice to actually show the DSS image that you processed offline. I also totally agree about capturing more light. Thanks for the video
Just found about about this video. Did a test with my new Canon R6 and it's got the same purple glow as the d750 just around the edges. Other than that it pretty much looks like the 6D sensor. No banding at all and very flat in the middle of the frame. Very light cropping and the purple will be gone! :)
I just did this with my Sony A6000 with a 30 second dark. It's a little blotchy but surprisingly even. Not at all like some of the bad ones you were showing.
Did you have long exposure compensation on? (By default its on) After the actuall foto is taken, camera takes one more photo also for 30s (It shows on display Processing Image) but with closed shutter to compensate for heat pixels etc.. Sometimes it can be a bit annoying as you have a long downtime between pictures. I should try the black image on my own camera to see ho much different images come out.
I just tryed it :) with both options image is very even (much better then the 5DIII shown on video) when long exposure compensation is turned off, "purple" is still nice and consistent but more random heat pixels are there.
@@danielridzon Good catch, I didn't think of that off the top of my head, but you are correct. Fortunately I have that feature turned off in the settings by default.
Long exposure compensation should be turned on. Essentially it is automatic dark frame subtraction immediately after each exposure using the same settings, at the same temperature etc.
My Nikon D750 has exactly the sort of purple artifact he describes. I found a simple fix for this in another video using the Dust & Scratches filter. Description starts at about 1:19:20
Just remever that it’s not just the pattern, but also the level of the noise that matters. A uniform noise pattern that’s high is typically way worse than a freaky low noise pattern, since your signal-to-noise ratio will be much higher in the latter! So my point is that you can’t really compare just the patterns and say that one camera/sensor would be better or worse, the level is at least equally important.
Another awesome, informative video - thank you!. I was in Joshua Tree this January 23-25 and got the EXACT same streaks across Orion...... 🤨 One other thing.... as an actual photographer, I would NEVER shoot higher than a 1600 ISO - EVER - even for asto. Think of ISO as gain for noise regardless of what the camera makes try to sell....... That is why your image is cleaner at the longer exposure. That noise is still one that I would dump for anything other than a RUclips video.
Good point, a higher ISO will amplify any grain in the photo. However, even if you use a lower ISO, you'll need to brighten the image in post-processing, which will have largely the same effect. Ultimately the ISO you should use depends on whether your camera is ISO Invariant or not. If it is ISO Invariant, you can freely use lower ISO's then brighten as needed in post-processing. This will allow you to retain the most dynamic range, especially in the highlights. However, if the camera is ISO Variant, it's usually better to use a higher ISO in-camera. The more you push the Exposure up in post-processing, the worse the results. Lonely Speck has a great example of this on his website. Ultimately it's the amount of light you capture that determines the grain in the photo. If you only capture 20 seconds of light, regardless of ISO, you will have a grainy photo. If you capture 4 minutes of light, regardless of ISO, you will have a much cleaner image. The camera is essentially generating white noise at all times. If we don't capture enough light, that white noise is visible in the photos. Once you capture enough light though, it will cover up that white noise. At least, that's one way to think of it.
@@PeterZelinka Beautiful put and it makes sense. I have noticed as well, as common sense as it seems..... the darker the sky, the better the image. What I capture here from my back porch under Bortle class 7 sky in Los Angeles at 3 minutes - with a filter - are still not as clean as 2 minutes in Joshua Tree. I am also starting to prefer camera lenses over a telescope that has other problems. My 200 f2.8 and 300 f2.8 are amazing..... at least for me at this point. Again, thank you for the awesome videos - I have learned a great deal from you!
It's pretty obvious that you don't get that much noise at ISO 800 as on ISO 6400, so for noise it doesn't matter how much minutes you are gathering on your shot (if this would be the case then dark frames wouldn't show any noise), also (if you can) dither, it moves around the mount slightly and helps to reduce even more noise. And for the banding issue: if you have PixInsight , there is a tool under Scripts/Utilities/CanonBandingReduction that eliminates pretty much any banding issue from any camera, not only Canon
Great demo Peter. I just finished stacking 6 hours worth of 4 minute subs with an old Canon 450D and noticed some horizontal banding on mine. I'll test my dark frames now but I bet I know what I'll find. Great information.
Hello. I have a X-T2 and I'm going to get a X-T3. Which stacking software do you use? Because DeepSkyStacker doesn't work with the RAF files. Thank you.
Very interesting, my D7500 was smooth across the image with only a small amount of purple glumped in one corner when I did the Darks test. I was very surprised as this is not a high end camera. I recently (June 2021) took 48, two min, ISO 1600, exposures of Andromeda with a Kit lens and got the most amazing picture. Thanks for this great information !
This is excellent insight Peter. I guess what we should do is just take some dark frames at the camera shop before committing to a body if using it specifically for astro imaging. From your samples, the Canon 6D performed miles better than every other in this sample anyway. I was thinking about buying a Canon EOS Ra for night scapes instead of my Canon 5DSr which is great for daylight work but struggles with night scapes. So next chance I get, I'm going to Edinburgh and take some darks using the Ra before dumping a couple of grand on a sensor that might just have the same banding issues anyway, unless someone can comment below with their own test results? Great video peter. Thanks, its a big help.
Yep! I ordered a D780, mainly for my RUclips videos, but it has a BSI sensor which should theoretically perform better than my D750. As soon as I get the camera I'm going to do this test and quickly find out how it is going to perform in low-light scenarios. Hopefully it looks comparable to that 6D frame. I know Trevor did a video on the Ra. If he still has it, maybe he can try this test. Then again, there may be a method to remove those bands during the photo stacking that I'm not aware of.
@@PeterZelinka By the way, I ordered the Star Adventurer last year after watching your videos and after a lot of trial and error and some truly dreadful results, i managed to get a Zeiss 15mm f2.8 Distagon T* on a Canon 5DSr tracking the Milky way for a whopping 30 minutes above Callanish 5,000 yr old neolithic stones on the Isle of Lewis, unguided with pin sharp stars; ISO 200 at f4. It quickly became my best selling image. I guess I should send you a cheque? lol
@@grahamhgraham Beautiful images Graham! I've never felt much of an urge to travel to Europe, until now! I'd love to see those landscapes and ancient landmarks!
Peter, I took several black frames, and they are all different! but you should see the canon 60 D sensor, it seams much better than the 6 D! I could send you a black frame if you are interested.
That was incredibly informative! ✨ My D810 does that purple-reddish "glow" -- guess it's a Nikon thing haha -- but now thanks to this video, I know it's easily fixable (I'll just composite/blend shots) 🙌🏽
I know this video is a year old but after watching it I checked my darks and saw some banding but it doesn't show up in my photos. I checked my master dark and there were only the hot pixels which is what darks are supposed to remove. I think the difference is I also use bias frames which accounts for the noise created when data is read from the sensor. DSS combines darks and bias to create the master dark if I remember correctly which eliminates the unevenness you see in dark frames and the final pictures.
There is an app for android called " Eagle Image stacker " which do alignment+stacking for both astrophotography and normal photography, also supports stacking most RAW images. Great when you do smartphone astrophotography, also when you want to stack you dslr images.
Great video with good information! I’ve purchased one of your astro courses and it was great! I’ll test out my D500 and D850 tonight. Sad to think how much time I’ve wasted getting dark and bias frames. I’m also not too thrilled with my star adventurer as it’s not very good at handling my 150-600 lens. I’ve spent this whole last year learning all this stuff and you’ve been very very helpful. I need to spend some learning how to do that blending you talked about in this video. Thanks!
How long are you able to shoot with the 150-600mm? We've been using the Star Adventurer with that lens this week and getting up to 2 minute exposures. Although, we are using an auto-guider too
Really fantastic video, thank you. Got me thinking, while checking my Darkframe, I also checked the Lens Profile Correction in Camera Raw and could see even more distortions added to the Darkframe noise... Now I know where that circular banding was coming from; I knew it was lens related as it was circular... but good to know!
Yeah! I tend to get a weird banding problem too, usually at the very end of creating a star trails image. I realized that applying the lens distortion corrections causes that problem apparently. That's why I normally just allow the vignette correction now.
Oh yeah the bloody sensor problems. People are not aware that there is still another type of even more annoying problem with sensors. Sometimes the sensor affects the signal itself. This creates color/white balance shifts which can be visible in low contrast images independent of exposure time!! It is completely impossible to remove, my new D850 is at service for this reason even if it is a little better than my old D700. The problem plagued me for years on my D700 affecting even normal images. But I found it to late... It was already out of warranty. Some of these problems are due to uneven heating of the sensor and can be removed by proper calibration with darks and biases but it is difficult to do everything perfectly. Also on sensor and other image processing circuits might interfere with the calibration software. So people check your sensors before it is to late!
Thank you for this video. As you suggested at the last minute I experimented. I made not only one dark, but sets of dark frames. I made sets of darks at different ISOs from 100 to 6400 at the steps normally used in the cameras. For each ISO I used a shutter speed of 30 seconds, 60 seconds, 120 seconds, 180 seconds, 240 seconds and 300 seconds. Not being a camera guru I found the results interesting and a little confusing. The reason I made so many darks is because at first when I made my experimental darks at ISO 100 I was disapointed at the ISO 100 results and almost panicked because of the banding. I thought I might decide to get a different camera. I'm not at all sorry I went further and made more darks because I now have several settings to use for my EXIF and can avoid banding. Several EXIF settings give me a fairly smooth pattern where there is no banding. At the end I made notes so I can avoid using a combination of ISO and shutter speeds that give banding. I used my 800D inside at room temperature throughout the experiment. I haven't a clue whether the results are temperature dependent. I saw that at some of the lower ISO settings I got banding some of the time at longer shutter speeds and no banding at a few of the shorter shutter speeds. At 800, 1600 and 3200 ISOs I started getting almost no banding. Also at those ISOs i started getting a much more even on screen pattern and although I never lost the purple color as I increased the ISO the purple became much darker. At ISO 6400 I had little if any banding, but I started to see the right half of the screen is notably darker than the left half. Because of my results Imade notes to carry with me in the field. That will be helpful when I start making subs with different EXIF so I can use layers to overcome washout on pictures of Orion. Thanks again for all the time you spend sharing.
I repeated e experiment with lights. I don't even have to stretch to see a lot of banding. It doesn't matter much what the ISO is I still get banding at every ISO and shutter speed up to 90 seconds. Next chance I get I'll make some 2, 3 and 4 minute subs. I'll likely have to get a guide scope to precvent trais on sube at linger than 90 seconds. Then banding is not consistant, predictable. The nanding is sometimes wide and sometimes narrow. I have banding on anywhere from 30% of the subs up to 75%.
I see what you're saying with the whole 4 minute long exposure. It's a really great way to get a universal lit foreground, however there's still a lot of noise (for my taste), for high res photos like canvas prints. I tend to do a lot of astrophotography with partly lit foreground using a flash or similar lighting source to light up the foreground image, then stack or merge the various parts of the image session to get my long exposured background with the shorter exposure, sharper, foreground.
4:17 or you could use a hasselblad x1d which has an autofocus feature that allows you to focus different parts of an image. And yes I know no one wants to spend that much money on a camera, it would make it easier.
Thank You! Checked out my gear. The winners Sony A7iii, Olympus OM-D-EM1X. They both beat my Canon 6D and Mark III. Decided to take the Sony and Olympus to AK. We will see what the results will be.
Thank You! Checked out MY gear. The winners Sony and Red. They both beat my Sony A7iii and OM-D-EM1X toys. Decided to take the professional gear to a professional shoot. We will see what the results will be.
Really cool investigative video Peter, Traditionally a Canon shooter I've recently picked up a Fuji after hearing the odd comment about them having a inherently weak IR filters which might be great for nebulae? After seeing your video I'm also keen to fire off a dark and stretch the histogram to see how uniform the sensor is on my XT100! I used to dabble in stacked shots with darks, lights and bias and never noticed the darks weren't doing there job, so no I have just got my hands on a star tracker again I'm ALSO keen to double check your findings here. Thanks for the upload, really interesting stuff and some of the shots you demoed were fab, thumbs up :)
Yeah, most DSLRs have an aggressive IR Cut Filter which blocks about 80% of the H-Alpha (red) light coming from the nebulae. It's possible the Fuji camera's have a better IR Cut filter, which allows more of that H-Alpha light through.
Hey, I have an X-T2 & I did the same test & it's a completely black frame...I have just started to try Capture1 as my Adobe annual subscription ended & thought maybe something is wrong with C1..or maybe I am doing something wrong 😂 the reason Fuji is better is because of it's X-trans Sensor, which is more sensitive to the H-Alpha here's a link to the video ruclips.net/video/KNloULA7A7w/видео.html
I owned a 450D in my early days of astrophotography and the banding was atrocious so I bought a 6D and haven't looked back. Great DSLR with low noise, especially below 10deg celcius ambient temperature.
My sony alpha a7R 3 doesn't have those banding issues. I just started taking astrophotography and the guys that said to get the best you need longer photos like you said. I don't have a start tracker so I'm limited to 45 seconds before I get star streak. Thanks for the info man. You definitely have a handle on astrophotography. I love it but it's so expensive haha
For proper dark frames.
"The key thing to remember about taking dark frames, is that the images must be the same exposure length, temperature, and ISO as your light frames (the picture files)."
how many dark frames do you need
@@antdx316 40 is ideal, 20 at least. I am not convinced that taking dark is a waste of time, I always take dark and bias frames and I never have banding in my images.
Love the mass effect music
Me too ☺
was just about to mention it, galaxy map tune.
*shivers with nostalgia*
I thought I was losing my mind. This dude is too cool.
@@hbastronomer517 ☺👌
@@hbastronomer517 do you own a telescope!
OH MY GOD.. Mass Effect music fits so perfectly.. Love it!
You can also take one exposure for the stars and one for the foreground, when shooting the foreground reduce the iso as much as possible and shoot a very long exposure to get no grain and more light, expose for the stars normally and take multiple images and run them through a star stacker program then blend foreground and bacground images in photoshop
Tested this on my Nikon D5300. Lens cap on and wrapped in a blanket:
- Test 1: ISO 500, F4, 20 sec = Same purple glow at bottom, but as well some horrible yellow/orange round dots splattered around the sensor, most from center towards the right.
- Test 2: Went into menu, did a "Clean Image Sensor", then performed above test again: ISO 500, F4, 20 sec = All the yellow/orange round dots were gone, though the purple band at the bottom remains.
- Test 3: ISO 6400, F4, 20 sec = Without increasing exposure nor whites, one can notice a very faint purple glow at the bottom. When both are increased picture is very uniform, and somewhat clean.
So learned that I must clean the sensor from the menu, much more frequently due to this, thank you! And lovely channel by the way. (.. Subscribed due to this)
I feel like your testing is flawed. Why did you take first two tests at iso 500 and then third one suddenly at 6400? Also, you should try longer exposures. I just don't see how the sensor cleaning function changes anything as what it does is it shakes the sensor or IR filter which is totally unrelated. Also, most cameras do it automatically when they're turned on and off.
The dark frames need to be taken with the exact same settings (ISO and exposure) as the light frames, or otherwise you need to scale them to match them. From what I saw in the video lights and darks were taken with different settings, which explains why the noise was not completely eliminated. Differences in temperature also change differences in noise levels, which is why professional observatories (and more serious astrophotographers) have cooled, temperature controlled CCDs.
How do you 'scale' them? That sounds like an important detail.
Omg thank you so much for explaining what dark frames do. So many channels talk about how to take dark frames but don’t explain what they actually do.
@2'35": 'no real grain anymore' ... love your optimism :)
It’s a lot less though :)
You can also take a "dark frame" and substract it from your images to get rid of regular sensor noise which is what the purple glow is.
That's exactly what I was thinking. These dark frames are all repeatable(the important bit) characteristics for any given sensor, so couldn't software 'subtract' that (in a user-controlled way if possible) from the light frame image.
Daft question (I'm new to this very fascinating subject) - what software did you use to subtract the dark frame and, of course, did it actually work!
You can use the free software called Deep Sky Tracker to do this, but you can also use flat frames to remove vignetting
I was thinking the same thing, however, if you do the light sensor test image, you could use that to remove the bands in photoshop as a inverted filtered image over, to neutralize the banding.
Deep sky stacker does all this for you, or any software for that matter. There's no need to do any manual stuff
The purple glow looks to be amp glow
Peter your a darling. Just checked my 6D, it's not green it's pinkish but with next to no banding on the sensor. Very even throught out. Thanks mate, you just restored my "faith" in the 6D that I bought for £350!
Great video. Awesome to see how the different sensors perform.
You may want to do some reading on why each of the different types of frames are taken, and the difference between random noise and average noise. These can both be applied to shot noise, read noise, thermal noise etc. Random noise in these categories will only be benefited from more light frames as you concluded from your pragmatic approach. The sensor read noise you keep stretching out would have to be dithered out. But these are technicalities. Overall if someone is doing wide field, the best bang for their buck (time) is more light frames. There is enough light in a wide angle Astro (landscape or deep sky) that with fairly limited light frames the amount of stretching required to balance the photo won’t bring out the problems, but obviously needs more than one shot. With higher focal lengths and much dimmer objects, the stretching required is going to bring out every little problem hidden. In that case, understanding all the different kinds of noise is ++helpful to make some of the best shots.
Clear Skies!
Thanks for making these videos!
I have that exact band on the bottom of my D750 images, I wondered why I was getting it, is such a pain! i just checked a dark frame and can see the correlation. I just got a z6 so will have to test that out! Thanks Peter, very useful info!
My Z6 sensor is silky smooth and clean as!
hello , does the z6 also suffer from same purple issue?
@@nissaarwz4645 yes
i love how you used the Mass Effect 3 Galaxy Map song
That's what I thought, I was like this sounds familiar lol
First of all let me say this: AWESOME video, what I like especially much is that you have presented a ton of examples of different shots, flat, light, dark/offset frames, from different cameras. (Offset for me is a very quick dark frame, so sometimes they're mixed up by ppl)
Nevertheless, I wanna make some comments here as well. One of your major points in this video is that the banding/spots of color/lit corners remain no matter if you use a dark frame in the stacking process etc. So this is kinda true, and kinda obvious, but in fact you are committing a slight maths mistake. Let me try to do some short talk to present this.
1) SNR: probably everybody knows best about SNR here, so true: more light -> more signal -> read noise, electric noise, thermal noise contribute to a lesser percentage -> better SNR
2) but what happens to signal/background when stacking images? well 3:1 ratio + 3:1 ratio + 3:1 ratio = 3:1 ratio. It stays more or less constant. that is the main problem with stacking and that is why you need the background. There is a slight difference between signal/bgd and signal/noise. that's why we need to subtract the background for.
3) how to estimate background noise and what is the problem?
3a) if the background had zero average (which is completely not the case in photography) then by doing many many experiments we should come closer and closer to zero (law of large numbers). but wait? how does a dark frame look -> it is a collection of pixels (and 3 channels) from which each has a NON-ZERO average luminosity and some deviation from it, which is random, or so we hope.
3b) thus by taking a humongous amount of frames (try it, number goes to 1000 rather than 25) you establish a map of the AVERAGE VALUES of noise in each pixel and channel. Maths: the estimation of the average is accurate to about sigma/sqrt(number of tries); so for 100 tries you have only 1/10th of the initial variance in noise luminosity, for 1000 tries you have 1/30th, sort of, of the variance, and only for 10000 tries you get down to 1% of the original noise variance.
3c) the same estimation rules apply to flat, dark, light frames etc
4) so now we know that by stacking we don't change the SNR, tho we hope that after stacking 100 pictures we have the average noise (times 100!!) in our background (coz how could it possibly go away) and we can try to bring it's AVERAGE level down, it will NEVER reduce the grain coz this is the randomness that you would need thousands of shots to remove, and even then by exposing the interesting part of the histogram really much you'd amp it up again!
5) you are using the exposure slider "just so", don't really putting care in how much you slide it, so you don't notice the decrease in AVERAGE noise level, but you still notice the band, coz it's there! you cannot reduce background noise this way! it's just smaller on average but has the same shape, and you can always amp it up again and show it on screen.
5a) overblown example: suppose you have taken 1000 offset, dark, flat frames each (what a titan). and then you took just ONE light pic. The light pic has a RANDOM instance of the noise, and when you subtract/divide etc by your calibration frames you bring the AVERAGE level down, but not the actual variance. So the random instance of the noise is just there, it's average has only been brought down to almost zero.
5b) if you took 1000 identical pix (hard to do of course -> triple titan) you'd bring the noise variance down a lot and by subtracting the stacked picture you'd get a comparable noise variance level, but if you exposed it HARD ENOUGH you'd still see the same banding (tho quantitatively it now constitutes less of the signal)
6) to sum up: by even saying that you should expect that band to go away, you are making a statistical mistake, you can never get rid of that banding etc, you can just reduce the standard deviation around the average by averaging it many times, same goes for any other calibration frame.
sort of encapsulation try:
(sig+noise) x100 stacked = (same snr) 100x signal + (random_noise x100 different instances summed up)
= 100x signal + 100x average_noise + 10x noise deviation (sqrt(100) x orig deviation)! (same average snr 100sig/100av_noise)
( noise ) x100 averaged = av_noise (10% of noise variance as compared to 1 frame)
now we subtract 100x average noise from stacked pic -> we still are left with 10x noise deviation! the band are actually more visible and easier to expose!
pre-final result: 100x signal + 10x (one-frame) noise deviation!
so if we are ok with 1x noise deviation level, we can then DIVIDE by 10 and we get final result
FINAL = 10x signal + 1x noise deviation so we converted a godzillion calibration frames and 100 light frames into a ten-tuple signal with the initial noise variance, thus all be the bands etc. will be visible when sufficiently exposed.
MORE SNR, yes but only on average!, grain is EXACTLY THE SAME, but could be 10 times more wild! So what you say is dead true and kinda obvious and I want to add that by stacking (adding) signal you INCREASE the noise variance, and you are only able to get rid of its average!
I hope that wasn't just mathematician's/theoretical physicist's mumbo-jumbo coz I am a big fan of channels such as your and astronomy!
If that seems instructive and has helped you and you want me to clarify more and are interested in actual computations I will be more than happy if any1 contacts me on my channel or later via facebook/gmail. Clear skies!
EDIT: I used variance in the common sense, as something being changeable. in fact variance = deviation^2 for a mathematician, so I edited back to "deviation" as a more accurate wording.
To show how averaging works in pictures I prepared examples with controlled exposure bumping:
drive.google.com/open?id=1xQH6w4xnAI9D-AZOl0hRheu98h4BmKk0
drive.google.com/open?id=1kIAkedV-q9QxZc_kktPxI4YZShKhWHX6
The bands, splashes and noises are now virtually out of the Master files, but if they were present in the stacked light frame, they're just there.
@@przemekmajewski1 Thank you for this, Przemek.
Interesting video. I discovered something quite interesting regarding banding. I have a Canon 80D and have been using it for DSO imaging for the past couple of years. Out of all of the images I've taken I've only ever seen banding once, in one stack of images. I couldn't understand what was causing it as I'd never seen it before. After doing some tests I realised what it was... I use APT for imaging, and I always choose to save my images to 'Card and PC' in APT (obviously saving it to the camera card, and my laptop hard drive). But for this particular imaging session, on M81 and M82, I accidentally set it to save to Card only. The resulting 7 hour stack had these purple bands across the final stacked image. I then realised the only thing I changed was how I saved the images, so I set it back to save to PC and Card, and since then, and before then, never seen any banding at all. I read somewhere that it can be something to do with saving images to memory cards in the DSLR and that's exactly how it was!
Couple of corrections: you need to take dark frames at the exact settings of the photos you intend to correct ( and possible even at the same temperature) so that the noise pattern is consistent to the one you get in the light frame, otherwise you end up under correcting or over correcting the noise in the original image and that can cause problems.
Another thing, what you see over Orion are not Starlink but the "usual" geostationary satellites that orbit that area of the sky, there are so many because it's a special orbit. Starlinks in orbit now would probably appear brigther than that.
1 more correction. He was wrong when he said the earth rotates. truth is the stars travel around the flat earth.
@@hughmongus1233 please don't vote !!
@@RPMac please don't think.
@@hughmongus1233 flat brainer huh?
V0LDY, it cannot be geostationary, because those do not move accross the sky as they are - geostationary, meaning they are always above fixed point on the Equator.
If you don't have a star tracker, a low cost solution is to take an image series of shorter exposures (where the sum of all exposure times is your total exposure time, i.e. for instance 12 images of 20 seconds result in 4 minutes total exposure time) and stack them using an image stacker like Siril. The advantage in this approach: one bump to your camera during a 4 minute exposure ruins the whole image. With the stack you can sort out the bad apple and still get a decent looking stacked image out of the n-1 images left.
Yes. A lot of folks out there use stacking for wide field photography i.e milkyway, and trackers for the distant objects.
Also you really do need to match settings with your lights. The noise pattern can change noticably at different iso and temp, especially if your camera is dual native iso or highly iso variant.
Dss works really well in terms of dark, flat and bias frames as compared to sequator. I don't have a tracker. I manually track(it damn difficult)... Taking more lights really works... I took 75 lights on first attempt on m42 and then 150 lights the second time... And i can say the second time the results were really awesome... Those dark frames really help in decreasing the noise and mainly useful for stretching the final stacked images
Do you remember how long the exposure for each photo was? Tell me please🙂
@@michael_1601 It depends on the focal length. Since it was untracked I took 1 sec exposures
Hey Peter, interesting video as always. Agreed with Voldy that dark frames must be taken with exact same settings (and temperature/humidity etc). Usually do that when packing up at the end of a session. Also, in your example (1:47 vs 2:31) you are comparing single exposures and for those, you are quite right that the more light (exposure time) you get, the better Signal-to-Noise Ratio (SNR) you have. On single exposure, sensor and images are usually not that good in terms of SNR and if stacking doesn't increase the amount of light received, it dramatically increases the SNR by removing/averaging most of sensor/atmospheric noises. That's what allows in post processing to enhance so much the image compared to a single exposure, this large increase in SNR. IMO, I think your comparison of image quality when increasing exposure/white would have been fairer if you compared 13 stack of 20s vs the 260s image. But FYI, there was a test by another astrophotographer (possibly Alyn Wallace ?) who compared the difference of stacking vs single exposure (but with same total exposure time). Single exposure was showing indeed more details as they do get more light, but the difference was much more gentle (and acceptable) than what you showed. Clear skies !
It's the most useful astrophotography video I've ever watched. Every minute is useful. And your channel is HIGH QUALITY :)
Good stuff....Even on a dark frame w/cap..... the sensor only sees what the lens allows it to see. so taking the lens into consideration is a must. Change the lens to find the best lens to camera combination also.
I'd like to say thank you. You have gained a subscriber and provided useful information. There are numerous videos on each and every subject out there. It can be true that the overwhelming majority does not provide useful information. Many of which are popular channels. However, through digging around or quite possibly appearing on your recommended page lies informative knowledge. You are an excellent representation of just that. You got straight to the point and did not overly represent your channel with a couple of minutes of essentially stating like and subscribe. No, you provided usable knowledge in a relatively quick video that I will actually use. You were understandable, knowledgable in this subject, and the title held to be true throughout the extent of the video. It's uncommon to find channels like this where you truly appreciate the content and their hardworking creators. Your an example of exactly that. Keep working hard, and thank you.
This is ground breaking, no one has done anything like this. Thank you!
Possible solution to the banding problem - Take a few photos then rotate the camera on the tripod by a few degrees and recompose the photo with the important feature in the center again. Repeat this several times while taking the whole set for the stack. This will require using an app that can align the stars before doing the stack. This might work because all the banding will not be close to the same positions. Averaged over a larger set, if the bands are in six directions, each band should not contribute as much to the finished result. Disclaimer: I am just getting into night photography (generally too much light pollution, haze, and clouds where I live) and my camera is not a Canon. Thanks for the video. You give us much to consider.
That Mass Effect map music tho. Man, it's just so timeless and fits astrophotography so well
Excellent video Peter. I wrote to you perhaps about a year ago as I saw this image of your SUV parked in a dark sky site with the Milky Way sky in the background. You had displayed it in a popular Facebook adtrophotography site to explain the importance of longer exposures.
I had to concur with your advice, as an astrophotographer since my early teen years in the mid 1960's using crude consumer-grade cameras, actually my parents Kodak Box camera, starting in about 1956 as they handed it to me at around age 2 to take phototograghs of them on my birthdays.
I'm backtracking here to the Dark Ages, so to speak, for most of the younger generations now entering astrophotography with digital cameras today. If only to make the point, that I dropped all the photography about the time I left High School, to instead become an art major in college in 1973 at the age of 18. I actually also purchased a new 35mm Mamiya/Sekor camera and took Photography and darkroom courses to learn how to process the wet-film-base photography that we were are all using long before digital photography came along a generation later.
This may be a rather lengthy comment to leave but I later learned in recent years why I admired the great film directors, say out of Italy, or other foreign film directors such a Stanley Kubrick who left America to live in England the remainder of his life, as many people know he's the director producer of the great film 2001: A Space Odyssey, that is still admired today for it's groundbreaking photographic work.
The early motion film directors too, had a similar childhood process of attrition as their parents gave them advanced photographic film cameras when they were very young. Many great figures in the history books today, that are known for their hand drawn and painted artwork also started out in photography and hand-drawn art work. As I began producing very large pastel sketches, starting with the sun, produced from live observation, with h-alpha filtered solar telescopes in 1999, during the daytime, and then the moon at night, also took me away from the photography for a while as I contributed these to the Cloudy Nights Astronomy Forums for competition in sketching, which I held the records there, for quite a while, after these large hand drawn pastel sketches first appeared in the NASA websites, Spaceweather several times, and Astronomy Picture of the Day. Now Sky & Telescope magazine is asking to display some of my large hand drawn pastels of the Moon, which will appear in the upcoming May 2020 issue.
So finally today discovering your video here, I'm so appreciative of the fact that you explained in fine detail, the failure that many new astrophotographers make by under exposing their images and wondering why they have so much noise and color banding, even after using stacking programs and shooting dark frames, in hopes to cancel out the visual noise and color banding errors. To this day I'm only using a Sony NEX 5r consumer grade Mirrorless Camera with APS-C sensor and an older Sony NEX 5, that I photographed the famous American Solar Total Eclipse with in 2017 in the Oregon desert through my Celestron Nexstar 5i Cassegrain telescope, not far from my home in Portland Oregon. I was asked by the Oregon State Park Administration Directors to give a few lectures to the public in a large State Park about an hour drive from The Painted Hills Oregon for a few nights before the actual Eclipse occurred on that Monday morning August 21st 2017. I also produce very large panorama photo stitches of the Milky Way over beautiful dark landscapes such as at the Oregon Coast Beaches.
As you may know, that these large panorama's, up to a hundred photographs in each finished image, can be reproduced into very large prints without any resolution loss up to perhaps 6, or even or 10 feet wide. Where most photographers today will only display their work as digital online images that are only viewed in cell phone screens of only 2 by 4 inches. I'm also employing my old world camera lenses from the 1970s and 80s adapted onto my newer Sony digital mirrorless camera to allow much more light through, say, an antique 55mm f/1.2 lens, so this goes without saying, in taking your advice on initially capturing more light, with only the most modest equipment available, that of a free camera lens from my college days. Many people are now discovering this by utilizing these huge old and fine quality camera lenses from the 1960s, 70s, and 80s, found and rediscovered in their parents closets, garages, and attics, collecting dust, and adapted onto their new digital cameras to capture the most light from the Milky Way.
So my subjects of astrophotography, starting from the deep dark sky of my teen years at our parents front yard in the far Northeast edges of Portland Oregon under fairly dark skies with my parents crude Kodak box camera strapped to the top of my Tasco 60mm refractor, that I bought at age 13 with my berry picking money, in my first efforts to take a hand guided while through the eyepiece, as a 5-minute exposure of the Orion Nebula, which I still have that negative today somewhere stored away, and now ranging to the bright solar sunlight of the solar total eclipse a couple years ago, with my modestly advanced digital cameras attached to my modern equatorially tracking motor-driven Celestron Cassegrain telescope, that was given to me for free by the Celestron company, as I was invited to teach astronomy overseas in the South Pacific in 2004, which are all still my most modestly advanced equipment that I have today.
I've often thought about taking the leap to the newest astrophotography cameras, like the ZWO that you mentioned in your videos, but living now in retirement at age 65, on a diminished Social Security, and retirement pension income, I've trimmed down my living expenses, to be satisfied with this very modest astrophotography equipment, because my greatest efforts for the past 20 years have been dedicated to the public, in providing thousands of hours of sidewalk astronomy which I later spoke about on National Public Radio many times, and I've been covered in other news articles Across the Nation and overseas in the Fiji islands later, about this service that I provide for the public and schools students for free, still today.
If this isn't enough of a spectrum to tackle over 50 years, I also became a very active sidewalk astronomer for the purpose of Public Service, to the public community, spanning the entire United States and Canadian continent in 2000, when I borrowed from our astronomy club in Portland, a large h-alpha solar telescope, taking it onto a 10,000 mile cross-continental solo road trip over six weeks, to allow thousands of the public on sidewalks, to view the Sun's solar sunspot maximum and huge solar flares, safely, and directly, live through this instrument. I audio recorded thousands of responses from the public and assembled them into a best excerpts 15-minute video in my RUclips channel, as my top cover page displayed video. Three trips a few years later in 2003~2004 to the Fiji Islands, I did not record audio of the public responses but only relied on their hand written responses, in a large guest registry book, which now includes over 3,000 handwritten comments in many different languages and locations across the United States and Canada, and also in the South Pacific Fiji Islands in 2004. All these ventures are covered in my Facebook albums, with many candid photographic images on location. Also some of the higher resolution photographs of the sky, I have uploaded into the famous Digital Photography Review site gallery.
Today I'm sharing your great video here with others that are getting started in astrophotography because you've covered so many great points that they need to know about basics in a proper Time Exposure. Thanks for posting and providing this.
WOW that test of sensor is mind blowing. Thanks for sharing!
the dark frames from my fujifilm xt2 is a lovely uniform green like the canon 6d! was actually really worried it would have a terrible sensor. i havent gotten into astrophotography yet but now i am definitely reassured that i have the sensor capability! thanks for the great information!
Hello. I have a X-T2 and I'm going to get a X-T3. Which stacking software do you use? Because DeepSkyStacker doesn't work with the RAF files. Thank you.
To do proper calibration on your lights you need to first integrate your bias, then calibrate your darks with your bias to remove the bias signal from the darks, without doing that you will get the issues that you are displaying. The problem is that you are not handling calibration properly. Everything that you have displayed as problematic can be fixed with doing proper calibration. I also suggest using pixinsight. Great video, but you really should learn the way to handle sensor issues. Inside pixinsight it a great book to start with.
Yeah, PixInsight has a lot of great features! However, my core audience is not using PixInsight. That's for the hardcore folks that really want to take things to the next level. It's a steep learning curve just getting used to that interface, let alone all the advanced concepts. My goal is to teach people how to create great images with what they already have, and are familiar with - Lightroom, Photoshop, DSLR, star tracker, telephoto lens. There's plenty of other channels that will focus on PixInsight.
I think his point is was that you’re not doing calibration properly. The pixinsight was a side topic
Longer exposures won’t get rid of banding. As you said it’s baked into the sensor. As others have been saying, dithering is the only way to get rid of that horrible banding.
The Star map theme from Mass Effect
Mass effect music from Noveria makes everything special❤
Those are Geosynchronous satellites in the photo of the Orion Nebula (M42). M42 is almost right on the celesital equator, and is the primary orbit that the geosynchronous satellites get launched into.
In fact, you can prove this if you measure how many arcseconds long those satellite trails are. The trail length will be close to, if not exactly the same value you get if you were to multiply your exposure duration in seconds by 15 (because the earth rotates at 15 arcseconds per second).
the only thing that rivals your excellent star photo techniques is your narration and presentation. Both super-clear!!
I use the Pentax K-1 and is so amazing camera, the sensor have astro tracer, means the sensor ibis follow the stars without use of a star tracker and can make long exposures from more then 2 minutes, i'm so in love of my pentax K-1, really recomended camera for astro photography
Interesting feature! Never heard of it before.
that is absolutely amazing. i was just sitting behind the computer the other day thinking, stacking cannot possibly be the only way of getting good results. and sure enough i find this video of you doing this extensive test and thanks for sharing the result. this is pure photography. spend more time in the field and get the right shot rather than wasting so much time in post processing.
It’s not stacking that’s the issue, he is talking about dark and bias frames if you don’t need them, a part of most true astrophotography is always enhanced or at least stacked, even when done with dedicated Astro cams and serious telescope rigs, at its simplest form, it’s essentially just stacking multiple images you took yes? bringing the image into much greater view, this guy (great photographer I must add) created these really “popping” Astro images by stacking and post enhancements, like these ones at 18:10 for example.
stacking is simply stacking images taken, it doesn’t really take away anything to with it being pure... it’s a crucial part of the majority of most exciting Astrophotography you always see.
@7:00 coming from a Photoshop approach to fix the sensor noise issue, in theory (I haven't done this my self but will try it soon), you could clean up the noise by importing the 'dark frame' and main exposure into Ps: Place the dark-frame layer over the main exposure shot and set the dark-frame blending mode to one of the 'canceling ' filters : Difference, Exclusion, Subtract or Divide.
I just did a quick-and-dirty test of my X-T20 out of curiosity; I took a photo with the lens cap on at those settings. Unusual result. There’s a light leak in one spot on the left, not sure if that is the lens cap or very dim internal light (I know some cameras can’t be modified for Astro because of internal lights), I’d have to test without lights in the room. Secondly, while it has an even flat noise profile overall, no colour noise, banding or light/dark spots, there’s a box in the middle that is lighter. Most unusual. It has never turned up in any photo though, even with heavy Milky Way editing (neither has the light leak, which is most likely the lens cap anyway)
Hello. I have a X-T2 and I'm going to get a X-T3. Which stacking software do you use? Because DeepSkyStacker doesn't work with the RAF files. Thank you.
Btw, I did another quick test after, turning off the lights as well as a lens cap and putting it under my jacket for the exposure to cut out all external light. No internal light leaks, which is comforting. The overall noise profile is very pleasing, no colour noise or banding, a prime reason for getting my X-T20 over the Canon M (I have a Canon 80D, so seemed logical at first), the other main reason are lenses. Canon hasn’t supported their APS-C mirrorless system well, despite it being around for years, and Fuji’s primes are just so good. If I was looking at FF, my choice would have been different no doubt, but neither Canon nor Nikon had FF mirrorless at the time. Despite the X-T20 not being designed for night photography (I bought it for walk around street shooting; the small body doesn’t make it as easy at night with no light), I have tried it out twice with excellent results. Great dynamic range too for an APS-C camera.
Juan Pablo Revert - Process your Raw files to Tiff first; use 16-bit Tiff, not sure any stacking software supports Raw. You want all your files with the same basic processing (white balance, lens correction, chromatic aberration, etc). Only after they have been stacked do you worry about stretching the file; I’m not an expert, I’ve only done night scapes, and used either Photoshop or Affinity Photo to process. There are plugin packs for Photoshop to help develop Astro photos. Some great tutorials here on RUclips, but they can vary a bit in their technique. As I was doing landscapes, I used Starry Landscape Stacker on the Mac.
@@msandersen Thank you very much.
Peter, great video! Stacking does work, that's the only way people who use telescopes and dedicated astro cameras do it. For the stacked image you showed, you need to stretch the images first setting white and black points with thresholds. Bring all the channels (RGB) to the same level (let's say 40 for each channel and set your white point to ~ 240 just a bit less than 255. This will neutralize the background. Next use curves to stretch the image (use a traditional Astro curve). Go back to thresholds, reset the levels so they are all the same then back to curves. Do this a few times and your images will look very well. Tip: doing this with a star tracker will allow you to create "deep" deep sky images. But, great video!
Yeah, I guess my point is that dedicated astro cameras and DSLRs are different, and require slightly different workflows. What works well for one camera, might not work as effectively for another camera. As we saw, Canon has terrible banding problems while Nikon has bad "amp glow". dedicated astro cameras tend to have even worse amp glow, based on what I've seen. A "one size fit's all" approach is therefore not really applicable for all cameras.
Yeah, I know. I didn't want to get too sidetracked though, so I didn't include that method in this video. I like using that method, or even the Black Point dropper in Curves. It usually does a good job, and takes a fraction of the time.
Peter Zelinka Peter, agree sort of. Many people like me, use a DSLR and stack. Checkout Astrobackyard and Trevors videos and processing
@@PeterZelinka Many dedicated astro people do use DSLR cameras (you should check out Trevor at Astrobackyard) and the CMOS market for astro cameras has really taken off. As I'm sure you know CMOS astro cameras are essentially DSLR cameras only with cooling. With all that said, stacking is a great way to beat down the noise and dedicated telescope astrophotography people use it 100% of the time to image deep-sky objects. You make wonderful content and I enjoy your channel that's why I subscribe... keep up the good work.
@@M31glow That's true, most of the new astro cameras are CMOS. I just bought my first one and am looking forward to testing it out!
Using multiple lights for the total exposure rather than a single long exposure, and darks, bias, flats, dark flats, stacking done correctly using proper processing workflow and astrophoto software like DSS rather than photoshop or Sequetor) will help.
I might have totally misunderstood what you were saying, but you seemed to indicate that you ADDED the stacked darks to the stacked lights? That would simply DOUBLE to effect of the dark (noise). Proper astrophoto processing SUBTRACTS master (averaged) dark from each light being stacked, to remove the typical noise. Similarly, master bias is used to adjust for sensitivity of each sensor pixel, and the flats remove dust/blemishes from the lights (but have to be taken close to when the lights were taken, as dust moves around between sessions). And flat darks are used to minimise noise impact on the flats (have to be done separately from the std darks, as the exposure used for lights is different to exposure used for the flats).
ps. You don't just stack lights and darks, you use to appropriate software to subtract the master dark from the stacked lights. And also correctly process re the master bias and the flats/dark flats. Perhaps you need to use proper astrophoto processing software rather than just some random std photo processing software?
pps. darks won't fix noise problems if the noise varies over time for the same temp, exposure etc settings, but in reality the noise for a particular sensor is fairly constant over months or years (so often astrophotographers build up 'libraries' of master darks for different exposure times and ISO setting, so can reuse matches master darks for subsequent light sets, and not have to always take a set of darks for each set of lights (as time may not permit).
It’s a physical limitation for camera sensors and basically all sensors. The higher the gain, the more sensitive each pixels are, to the point where it starts to pick up onboard circuitry(heat) and other random things. I’m guessing that Nikons have heat coming from below the sensor, while the canons are a result of the dual pixel circuitry. The only way to reduce the noise is to have better algorithms.
Yes, those are definitely starlinks. I saw 16 or more of them back to back in two parallel lines with binoculars at around 4:15 am a couple of months ago and they can ruin a photo.
Excellent technical explanation using the dark frames. I immediately took some dark frames with my Sony A7RIII and Canon EOS R5. They both have areas of splotches and banding but the Sony is much cleaner. Great stuff Peter, thank you!
1) Those are not starlink satellites, Those a geostationary satellites (mostly weather)
2) Dither, even if it's one axis, Dithering in RA helps heaps, Any imperfections in PA will take care of DEC dither.
3) Darks and bias frames help with dark current, but not completely remove it, that's where dither lands the final punch.
4) If you want REALLY clean shots, Do all that with a Cooled astro camera adapted to DSLR lenses.
be careful about pixel scale, too!
This was such good info! I always wondered about the banding in night photography! I need to do my dark frame!
Some ❤ for the Mass Effect background music! 😁
Hehe... the Mass Effect soundtrack threw me for a few minutes until I realized it was playing in your video... nice touch!
Brilliant....thank you.....and you didn't push the point too much....it was perfect....a true clear lesson.....and haven't done any astro yet....but wow....thanks for the heads up....well done !
EVEN AS A MONUMENTAL UNINFORMED AMATEUR, THIS VID IS AWESOME AND I THANK YOU FOR SHARING
Very Detailed explanation about light and dark and why you should take them, THANK YOU FOR THIS VIDEO! i will carry this information into my next astro shoot.
I tested the sensor on my D610. It looks gooood, also as flat as the D6's sensor from your video, except a small triangle orange-ish shape top-midle and one in the top-right corner. But barely visible even with the exposure set to max and the whites cranked. I am pleased with the results, thank you. I didn't know these things
Thermal currents are important when calibrating a sensor. Remember that heat or thermal temps can trick a photo-site on a sensor that a photon struck that pixel. More noise can be present with higher temperatures. Calibration frames require that your light frames are at the same temps as your darks, bias, or flats. Good calibration has nothing to do with what brand of camera you use. Some CCD or CMOS ships have whole dead rows and can be calibrated out. Astro-imagers usually use thermal electric coolers(TEC) to control a consistence temp of their CCD or CMOS chips. Do this experiment with any camera. Go out on a cold night let the camera cool down for 1 hours or put it in a freezer. Take a 5 min dark frame. Let the camera warm up for a few hours and do the same 5 min dark. Compare the cold 5 min dark to the warm 5 min dark. More noise will be present with the warmer temps. If you are calibrating a light frame that was taken at 10C and using a dark frame at 30C. In the calibration process, its will over subtract and add noise to a light frame. If even temps are not solving noise issues, it would then be the number of samples. The more samples of darks, bias and flats create better "masters." Masters are just many dark frames averaged together.
Thanks a lot Peter. This was very informative. On some cameras, some of the peripheral blotches can be fixed with specific camera settings or minor hardware modifications. My canon 6D was leaking light both from Life View and from the red led indicator, leading to horrible magenta and red blotches when stretching long exposure photos. The purple blotches were removed by disabling silent LV in the menu. The red blotches by opening the camera and putting some black silicon on the led.
I remember my 5dIII also having obscene amounts of magenta blotches + banding; Just did the "dark frame" test on my 90d - It looks like your 6d image (even/ NO BANDING!!!), so could be worse! Will be interesting to redo it at the end of the night when theoretically the senor is warmer after a few hours of timelapse work. THank you!
Love the Mass Effect music.
The images specially the last one, you took are blown my mind. Seriously! And Background music took me into that world while watching these images!😱😇😳
Thanks for the video.
The discussion of the dark frames is really very interesting
Not the one with the hot pixels, but the one with the banding and the light and dark areas.
I took my Fuji X-T20, made a dark frame and was pleasantly surprised.
No banding, no differently bright surfaces.
A homogeneous surface.
Purple, but homogeneous. :-)
Cool video! I do deep sky AP but love shooting time lapses of my scope as it images the night sky behind (i include those time lapses in my videos). I use an a6400 and want to try milkyway shot in the summer! Clear skies!
Not that it matters much after almost two years, but you did say cameras of the same make/model can act differently in these tests. I mirrored the test with my 5D MKiii (the worst example you've seen) and it definitely was closer to the 6D with it's uniform grain pattern/structure than it is to the 5D MKiii example you provided. I had a little color variation on the right side of the sensor/frame but it was a smooth transition, with virtually zero banding. I did ISOs from 800 to 6400 with exposures from 5 minutes to 20 seconds. All similar patterns. I also did f2.8 (if it matters) after mimicking the 5.6 1600 for 120 seconds. Very clean pattern(s). (Lightroom classic and I also torched it all the way to +5 exposure in addition to the +3.85 in your example and pushed the whites further) Cool test though.
You could always check out the website" photons to photos" this will give you a heat map for practically any sensor of the last 15 years also will give you optimal iso setting to use. Or just buy a CCD camera preferable mono then add a filter wheel with individual R,G,B filters, your signal will be amplified many times.
I just bought my first mono camera with filters so I'm excited to see the difference! I'll have to check out that website, thanks for the link!
Totally on side about not listening blindly to what you hear online. It would have been nice to actually show the DSS image that you processed offline. I also totally agree about capturing more light. Thanks for the video
18:03 Wow!!! that's an awesome photo!!!
Just found about about this video. Did a test with my new Canon R6 and it's got the same purple glow as the d750 just around the edges. Other than that it pretty much looks like the 6D sensor. No banding at all and very flat in the middle of the frame. Very light cropping and the purple will be gone! :)
I just did this with my Sony A6000 with a 30 second dark. It's a little blotchy but surprisingly even. Not at all like some of the bad ones you were showing.
Did you have long exposure compensation on? (By default its on)
After the actuall foto is taken, camera takes one more photo also for 30s (It shows on display Processing Image) but with closed shutter to compensate for heat pixels etc..
Sometimes it can be a bit annoying as you have a long downtime between pictures.
I should try the black image on my own camera to see ho much different images come out.
I just tryed it :) with both options image is very even (much better then the 5DIII shown on video) when long exposure compensation is turned off, "purple" is still nice and consistent but more random heat pixels are there.
@@danielridzon Good catch, I didn't think of that off the top of my head, but you are correct. Fortunately I have that feature turned off in the settings by default.
Long exposure compensation should be turned on. Essentially it is automatic dark frame subtraction immediately after each exposure using the same settings, at the same temperature etc.
My Nikon D750 has exactly the sort of purple artifact he describes. I found a simple fix for this in another video using the Dust & Scratches filter. Description starts at about 1:19:20
This is such a great video! I can’t believe your camera is not modified!
Thank you for this video
Just remever that it’s not just the pattern, but also the level of the noise that matters. A uniform noise pattern that’s high is typically way worse than a freaky low noise pattern, since your signal-to-noise ratio will be much higher in the latter! So my point is that you can’t really compare just the patterns and say that one camera/sensor would be better or worse, the level is at least equally important.
Another awesome, informative video - thank you!. I was in Joshua Tree this January 23-25 and got the EXACT same streaks across Orion...... 🤨 One other thing.... as an actual photographer, I would NEVER shoot higher than a 1600 ISO - EVER - even for asto. Think of ISO as gain for noise regardless of what the camera makes try to sell....... That is why your image is cleaner at the longer exposure. That noise is still one that I would dump for anything other than a RUclips video.
Good point, a higher ISO will amplify any grain in the photo. However, even if you use a lower ISO, you'll need to brighten the image in post-processing, which will have largely the same effect. Ultimately the ISO you should use depends on whether your camera is ISO Invariant or not. If it is ISO Invariant, you can freely use lower ISO's then brighten as needed in post-processing. This will allow you to retain the most dynamic range, especially in the highlights. However, if the camera is ISO Variant, it's usually better to use a higher ISO in-camera. The more you push the Exposure up in post-processing, the worse the results. Lonely Speck has a great example of this on his website.
Ultimately it's the amount of light you capture that determines the grain in the photo. If you only capture 20 seconds of light, regardless of ISO, you will have a grainy photo. If you capture 4 minutes of light, regardless of ISO, you will have a much cleaner image.
The camera is essentially generating white noise at all times. If we don't capture enough light, that white noise is visible in the photos. Once you capture enough light though, it will cover up that white noise. At least, that's one way to think of it.
@@PeterZelinka Beautiful put and it makes sense. I have noticed as well, as common sense as it seems..... the darker the sky, the better the image. What I capture here from my back porch under Bortle class 7 sky in Los Angeles at 3 minutes - with a filter - are still not as clean as 2 minutes in Joshua Tree. I am also starting to prefer camera lenses over a telescope that has other problems. My 200 f2.8 and 300 f2.8 are amazing..... at least for me at this point. Again, thank you for the awesome videos - I have learned a great deal from you!
It's pretty obvious that you don't get that much noise at ISO 800 as on ISO 6400, so for noise it doesn't matter how much minutes you are gathering on your shot (if this would be the case then dark frames wouldn't show any noise), also (if you can) dither, it moves around the mount slightly and helps to reduce even more noise.
And for the banding issue: if you have PixInsight , there is a tool under Scripts/Utilities/CanonBandingReduction that eliminates pretty much any banding issue from any camera, not only Canon
I am seriously becoming your fan Peter! Nice videos!
Great demo Peter. I just finished stacking 6 hours worth of 4 minute subs with an old Canon 450D and noticed some horizontal banding on mine. I'll test my dark frames now but I bet I know what I'll find. Great information.
Great video - thanks. Got similar results to the 6D with the Fuji XT3 :-)
Hello. I have a X-T2 and I'm going to get a X-T3. Which stacking software do you use? Because DeepSkyStacker doesn't work with the RAF files. Thank you.
@@juanpablorevert I convert the RAF files to TIFFs and then use DSS.
Very interesting, my D7500 was smooth across the image with only a small amount of purple glumped in one corner when I did the Darks test. I was very surprised as this is not a high end camera. I recently (June 2021) took 48, two min, ISO 1600, exposures of Andromeda with a Kit lens and got the most amazing picture. Thanks for this great information !
Thanks so much for a very informative Video Peter
This is excellent insight Peter. I guess what we should do is just take some dark frames at the camera shop before committing to a body if using it specifically for astro imaging. From your samples, the Canon 6D performed miles better than every other in this sample anyway. I was thinking about buying a Canon EOS Ra for night scapes instead of my Canon 5DSr which is great for daylight work but struggles with night scapes. So next chance I get, I'm going to Edinburgh and take some darks using the Ra before dumping a couple of grand on a sensor that might just have the same banding issues anyway, unless someone can comment below with their own test results? Great video peter. Thanks, its a big help.
Yep! I ordered a D780, mainly for my RUclips videos, but it has a BSI sensor which should theoretically perform better than my D750. As soon as I get the camera I'm going to do this test and quickly find out how it is going to perform in low-light scenarios. Hopefully it looks comparable to that 6D frame.
I know Trevor did a video on the Ra. If he still has it, maybe he can try this test. Then again, there may be a method to remove those bands during the photo stacking that I'm not aware of.
@@PeterZelinka By the way, I ordered the Star Adventurer last year after watching your videos and after a lot of trial and error and some truly dreadful results, i managed to get a Zeiss 15mm f2.8 Distagon T* on a Canon 5DSr tracking the Milky way for a whopping 30 minutes above Callanish 5,000 yr old neolithic stones on the Isle of Lewis, unguided with pin sharp stars; ISO 200 at f4. It quickly became my best selling image. I guess I should send you a cheque? lol
@@grahamhgraham Hahaha that's great! Got a link to the photo, I'd love to see it!
@@PeterZelinka Scroll 3rd row down www.ghgraham.com/standard
@@grahamhgraham Beautiful images Graham! I've never felt much of an urge to travel to Europe, until now! I'd love to see those landscapes and ancient landmarks!
Peter, I took several black frames, and they are all different! but you should see the canon 60 D sensor, it seams much better than the 6 D! I could send you a black frame if you are interested.
Your videos are so nice to watch and very informative
That was incredibly informative! ✨
My D810 does that purple-reddish "glow" -- guess it's a Nikon thing haha -- but now thanks to this video, I know it's easily fixable (I'll just composite/blend shots) 🙌🏽
I know this video is a year old but after watching it I checked my darks and saw some banding but it doesn't show up in my photos. I checked my master dark and there were only the hot pixels which is what darks are supposed to remove. I think the difference is I also use bias frames which accounts for the noise created when data is read from the sensor. DSS combines darks and bias to create the master dark if I remember correctly which eliminates the unevenness you see in dark frames and the final pictures.
I love the Mass Effect music...very fitting.
The WHAT? Do you seriously watch shit like this for the entertainment value?
@@johnfrum9676 It's pretty entertaining dude
Great video with very helpful information Peter! Thank you!
There is an app for android called " Eagle Image stacker " which do alignment+stacking for both astrophotography and normal photography, also supports stacking most RAW images.
Great when you do smartphone astrophotography, also when you want to stack you dslr images.
Great video with good information! I’ve purchased one of your astro courses and it was great! I’ll test out my D500 and D850 tonight. Sad to think how much time I’ve wasted getting dark and bias frames. I’m also not too thrilled with my star adventurer as it’s not very good at handling my 150-600 lens. I’ve spent this whole last year learning all this stuff and you’ve been very very helpful. I need to spend some learning how to do that blending you talked about in this video. Thanks!
How long are you able to shoot with the 150-600mm? We've been using the Star Adventurer with that lens this week and getting up to 2 minute exposures. Although, we are using an auto-guider too
Really fantastic video, thank you. Got me thinking, while checking my Darkframe, I also checked the Lens Profile Correction in Camera Raw and could see even more distortions added to the Darkframe noise... Now I know where that circular banding was coming from; I knew it was lens related as it was circular... but good to know!
Yeah! I tend to get a weird banding problem too, usually at the very end of creating a star trails image. I realized that applying the lens distortion corrections causes that problem apparently.
That's why I normally just allow the vignette correction now.
I believe the lines you got near the Orion nebula are geostationary satellites. I get exactly the same everytime I shoot Orion
Oh yeah the bloody sensor problems. People are not aware that there is still another type of even more annoying problem with sensors. Sometimes the sensor affects the signal itself. This creates color/white balance shifts which can be visible in low contrast images independent of exposure time!! It is completely impossible to remove, my new D850 is at service for this reason even if it is a little better than my old D700. The problem plagued me for years on my D700 affecting even normal images. But I found it to late... It was already out of warranty.
Some of these problems are due to uneven heating of the sensor and can be removed by proper calibration with darks and biases but it is difficult to do everything perfectly. Also on sensor and other image processing circuits might interfere with the calibration software.
So people check your sensors before it is to late!
My canon 60d produces almost the same result as the 6d you have here, so that's pleasing!
Thank you for this video. As you suggested at the last minute I experimented. I made not only one dark, but sets of dark frames. I made sets of darks at different ISOs from 100 to 6400 at the steps normally used in the cameras. For each ISO I used a shutter speed of 30 seconds, 60 seconds, 120 seconds, 180 seconds, 240 seconds and 300 seconds.
Not being a camera guru I found the results interesting and a little confusing.
The reason I made so many darks is because at first when I made my experimental darks at ISO 100 I was disapointed at the ISO 100 results and almost panicked because of the banding. I thought I might decide to get a different camera. I'm not at all sorry I went further and made more darks because I now have several settings to use for my EXIF and can avoid banding. Several EXIF settings give me a fairly smooth pattern where there is no banding. At the end I made notes so I can avoid using a combination of ISO and shutter speeds that give banding.
I used my 800D inside at room temperature throughout the experiment. I haven't a clue whether the results are temperature dependent.
I saw that at some of the lower ISO settings I got banding some of the time at longer shutter speeds and no banding at a few of the shorter shutter speeds.
At 800, 1600 and 3200 ISOs I started getting almost no banding. Also at those ISOs i started getting a much more even on screen pattern and although I never lost the purple color as I increased the ISO the purple became much darker. At ISO 6400 I had little if any banding, but I started to see the right half of the screen is notably darker than the left half.
Because of my results Imade notes to carry with me in the field. That will be helpful when I start making subs with different EXIF so I can use layers to overcome washout on pictures of Orion.
Thanks again for all the time you spend sharing.
I repeated e experiment with lights. I don't even have to stretch to see a lot of banding. It doesn't matter much what the ISO is I still get banding at every ISO and shutter speed up to 90 seconds. Next chance I get I'll make some 2, 3 and 4 minute subs. I'll likely have to get a guide scope to precvent trais on sube at linger than 90 seconds. Then banding is not consistant, predictable. The nanding is sometimes wide and sometimes narrow. I have banding on anywhere from 30% of the subs up to 75%.
I see what you're saying with the whole 4 minute long exposure. It's a really great way to get a universal lit foreground, however there's still a lot of noise (for my taste), for high res photos like canvas prints. I tend to do a lot of astrophotography with partly lit foreground using a flash or similar lighting source to light up the foreground image, then stack or merge the various parts of the image session to get my long exposured background with the shorter exposure, sharper, foreground.
That's a great technique too! Plus, that added light will lower the grain in the image and ultimately give a cleaner, more detailed photo
Thanks for your content and your clean language
Just examined my Sony A99ii sensor. Clean and smooth like baby skin.
4:17 or you could use a hasselblad x1d which has an autofocus feature that allows you to focus different parts of an image. And yes I know no one wants to spend that much money on a camera, it would make it easier.
My D850 sensor was not too bad. A little green on the right side. D500 has a thin horizontal band. All those times I thought it was just me.
Thank You! Checked out my gear. The winners Sony A7iii, Olympus OM-D-EM1X. They both beat my Canon 6D and Mark III. Decided to take the Sony and Olympus to AK. We will see what the results will be.
Thank You! Checked out MY gear. The winners Sony and Red. They both beat my Sony A7iii and OM-D-EM1X toys. Decided to take the professional gear to a professional shoot. We will see what the results will be.
Extremely interesting: food for thought, thank you.
Really cool investigative video Peter, Traditionally a Canon shooter I've recently picked up a Fuji after hearing the odd comment about them having a inherently weak IR filters which might be great for nebulae? After seeing your video I'm also keen to fire off a dark and stretch the histogram to see how uniform the sensor is on my XT100! I used to dabble in stacked shots with darks, lights and bias and never noticed the darks weren't doing there job, so no I have just got my hands on a star tracker again I'm ALSO keen to double check your findings here. Thanks for the upload, really interesting stuff and some of the shots you demoed were fab, thumbs up :)
Yeah, most DSLRs have an aggressive IR Cut Filter which blocks about 80% of the H-Alpha (red) light coming from the nebulae. It's possible the Fuji camera's have a better IR Cut filter, which allows more of that H-Alpha light through.
Hey, I have an X-T2 & I did the same test & it's a completely black frame...I have just started to try Capture1 as my Adobe annual subscription ended & thought maybe something is wrong with C1..or maybe I am doing something wrong 😂
the reason Fuji is better is because of it's X-trans Sensor, which is more sensitive to the H-Alpha here's a link to the video
ruclips.net/video/KNloULA7A7w/видео.html
I owned a 450D in my early days of astrophotography and the banding was atrocious so I bought a 6D and haven't looked back. Great DSLR with low noise, especially below 10deg celcius ambient temperature.
My sony alpha a7R 3 doesn't have those banding issues. I just started taking astrophotography and the guys that said to get the best you need longer photos like you said. I don't have a start tracker so I'm limited to 45 seconds before I get star streak. Thanks for the info man. You definitely have a handle on astrophotography. I love it but it's so expensive haha
This really helped educate me