Deep Sky Astrophotography
- Thread starter jrista
- Start date
Hi Jon.
That is really cool, loving the smoking trail, and the fact that you took the effort to share it with us, thank you.
Cheers, Graham.
That is really cool, loving the smoking trail, and the fact that you took the effort to share it with us, thank you.
Cheers, Graham.
jrista said:
Upvote
0
jrista said:
I was checking out that camera. there were some reports that it has an IR filter stack attached - was that the case with yours? there was confusion over whether or not that the was the color versus monochrome imager though.
I was thinking of getting the color - simply because of hyperstar compability.
Upvote
0
rrcphoto said:
Hyperstar III has the filter drawer system. The mono is vastly more efficient than the color, and with the filter drawer, you could do narrow band f/2 imaging. I know a couple people have done that so far.
As for IR filtering, no, not on this camera. There is an AR (anti-reflective) window built into the body, and there is just plain glass covering the sensor for protective purposes. But the sensitivity range of the sensor is full, reaching deep into the NIR range. I think that the original camera design included an IR cutoff window instead of just an AR window, but that was changed for the final design, IIRC from early discussion I had with Sam Wen of ZWO.
ZWO offers their own set of ASI1600-optimized LRGB filters. The RGB filters of course cut off outside of the color channel bandwidth. The L filter cuts off hard at the UV and IR ranges. Same goes for most LRGB filter sets from Baader, Astronomik, AstroDon, etc.
Upvote
0
Experimenting with the 80D and the 135mm f/2. This is a interesting combo as it puts more pixels on target than my Tamron 70-200m with the 6D. The 80D also has a new low read-noise sensor which I thought could be good for deep sky. To compensate for the lower full-well capacity of the smaller pixels I used ISO 800 instead of my typical ISO 3200 or ISO 6400. Not happy with the amount of highlight detail. Maybe the lens is just not sharp enough for 3x Drizzle??? Hopefully my local currency will improve a bit more then I can get a tracking mount.
The most notable feature in the LMC is the Tarantula Nebula https://en.wikipedia.org/wiki/Tarantula_Nebula, which is the most active star-forming region in our local group of galaxies. The 200 parsec size of this extra-galactic object is hard to fathom. To put it in context, if it were as close as the Orion nebula then it would appear 40 times larger than the full moon in the night sky.
2016-11-09 - Large Magellanic Cloud by Omesh Singh, on Flickr
The most notable feature in the LMC is the Tarantula Nebula https://en.wikipedia.org/wiki/Tarantula_Nebula, which is the most active star-forming region in our local group of galaxies. The 200 parsec size of this extra-galactic object is hard to fathom. To put it in context, if it were as close as the Orion nebula then it would appear 40 times larger than the full moon in the night sky.
2016-11-09 - Large Magellanic Cloud by Omesh Singh, on Flickr
Upvote
0
StudentOfLight said:
This is a great start, Student! Deep sky astrophotography is quite challenging. Especially if you are imaging with any kind of light pollution (city light, bright moon, even bright flares from oil & gas drilling.) The Magellanic Clouds are actually relatively faint overall, with a few bright spots like Tarantula.
My recommendation is NOT to bother with drizzling. Drizzling requires a lot of data to really be done effectively. If your sub count is too sparse, drizzling will actually increase noise, and it could introduce artifacts. This is particularly true with 3x drizzling. On top of that, you are needlessly expanding memory and storage usage by expanding resolution like that...and I can tell you you don't need it. It is best to use your full frame, along with effective calibration (particularly dark calibration, as DSLRs, even modern ones, have higher dark current as they are not cooled to sub-zero temperatures). You can calibrate a number of ways...however one of the simplest is to use the in camera LENR (Long Exposure Noise Reduction). This will double the length of each sub, however it is the best way to get well-matched dark calibration. There will be a slight increase in random noise, however that is usually preferable to keeping all the nasty pattern noise around, which tends to be particularly egregious in Canon DSLRs (the 7D II seems to be the lone exception so far.)
I would also recommend you go back to the higher ISO. I would actually recommend ISO 1600, as that seems to be the sweet spot for most Canon DSLRs (although to be honest, I don't know the noise profile of the 80D off the top of my head.) Anyway, with ISO 1600, you can use shorter subs, which will kind of balance out the waiting game if you opt to use LENR. Depending on where you are, you may only be using 30-60 second subs. With LENR, they would be 60-120 seconds. Then it is just a matter of getting as many of them as you possibly can. You want LOTS of them. I like to go in lock-step with noise reduction factors, which is the square root of the subs stacked. I would say get no less than 36 subs (reduces noise by a factor of 6), but you could go for 49 (7x reduction), 64 (8x reduction), 81 (9x reduction) or 100 (10x reduction). Reducing noise through stacking is key to getting good astrophotography results. With shorter subs, say 30 seconds, plus LENR so 60 seconds, you could get 100 subs in 1h40m.
That is a decent start...however if you are imaging from or near a city, less than two hours is usually not enough to really go deep on a faintish object like the Magellanic Clouds. Generally speaking, I encourage beginners in or near the city to aim for four hours at a minimum. The deeper into the city you are, the more data you will really need to get decent results. I acquired 10, 14, in a couple cases 18 hours of data with my 5D III when I was still imaging in the city!
If you are using a DSLR for imaging, the BEST solution is to just get the heck out of the city. City light pollution is the bane of astrophotographers. There is really only one solution to it: More and more and more and ever more exposure time. You can sometimes use a larger lens with a larger physical aperture, however that usually comes with a change in FoV as well, which can require mosaicing if you really want a big wide field, and that increases the time to produce a final image as well. The solution is to just get rid of the LP, which means getting away from it. Driving 30-40 minutes out from the edge of town is usually sufficient to get you to nice, dark rural skies. From there, you will usually only need about two hours of data, and you'll be golden!
Upvote
0
I generally do not have sufficient focal length for deep sky objects so I usually end up drizzling the interesting portion of frame. I previously shot the LMC under similar LP conditions with my 6D and Tamron 70-200 at 70mm (102x6s).
LMC (102x 6s) by Omesh Singh, on Flickr
The most notable difference is that the stars were not round with the 6D+Tamron image. I suspect it is due to some residual tilt in the IS mechanism as I always get stretching in my images from top left to bottom right.
Regarding ideal settings for the 80D, I read this on DPReview:
https://www.dpreview.com/forums/post/57481095
I'm interested to see how it does at ISO100-200 with long exposure tracked images.
Wouldn't you get the best results on an 80D with an ETTR approach at a lower ISO setting?
LMC (102x 6s) by Omesh Singh, on Flickr
The most notable difference is that the stars were not round with the 6D+Tamron image. I suspect it is due to some residual tilt in the IS mechanism as I always get stretching in my images from top left to bottom right.
Regarding ideal settings for the 80D, I read this on DPReview:
https://www.dpreview.com/forums/post/57481095
I'm interested to see how it does at ISO100-200 with long exposure tracked images.
Wouldn't you get the best results on an 80D with an ETTR approach at a lower ISO setting?
Upvote
0
StudentOfLight said:I generally do not have sufficient focal length for deep sky objects so I usually end up drizzling the interesting portion of frame. I previously shot the LMC under similar LP conditions with my 6D and Tamron 70-200 at 70mm (102x6s).
LMC (102x 6s) by Omesh Singh, on Flickr
The most notable difference is that the stars were not round with the 6D+Tamron image. I suspect it is due to some residual tilt in the IS mechanism as I always get stretching in my images from top left to bottom right.
Regarding ideal settings for the 80D, I read this on DPReview:
https://www.dpreview.com/forums/post/57481095
I'm interested to see how it does at ISO100-200 with long exposure tracked images.
Wouldn't you get the best results on an 80D with an ETTR approach at a lower ISO setting?
Hmm. I am not sure that Jerry nor Roger are correct in that post. Canon has always, to my knowledge, used a 512 ADU 14-bit offset. I've been measuring DSLR data for years now, and I always make sure to measure in the proper bit depth, and across countless Canon DSLRs, I've measured a bias offset of 512 ADU (14-bit).
I think people make the common mistake of converting the original RAW DSLR data into 16-bit, or the mistake of processing it in 16-bit space. Interestingly, the conversion factor between 14-bit and 16-bit is 4 (2^2), and 512*4 = 2048. I do not believe, at least not with any cameras since the 5D II or around there, that Canon has used a 2048 ADU offset.
If Roger is correct that Canon is using a lower offset at ISO 100 and 200, my guess is it would be a 128 ADU 14-bit offset. With lower read noise, that would certainly be possible. It also seems logical to keep a larger offset at higher ISO settings, which amplify noise more, and need more "buffer" room to avoid clipping the signal anyway. I haven't read anything specific to that effect myself, so I honestly cannot say what Canon may have really done with the 80D.
My opinion on which ISO is best is this, and actually in line with Roger's opinions most of the time:
It is better to "oversample" each electron of true signal, than to "undersample" them. By that I mean, you want your gain to be high enough that you are amplifying each electron by a factor of two, if not more. At low ISO, you are, for lack of a better term, "attenuating" the signal, since in most cases you must convert many electrons into each output ADU (analog to digital unit). This is low gain, high quantization. You are quantizing, or grouping together and effectively "squashing", many electrons together and losing a certain amount of information they held when they were separate and distinct.
This is not ideal for astrophotography, since we are chasing extremely faint details, and every single electron matters. It is entirely possible that at a low ISO, say ISO 100, you are converting 2, 3, 5, maybe even more electrons (depends on the pixel size and architecture) into one ADU. Technically speaking, it is possible to recover precision by stacking lots and lots of subs. However that is at odds with using a lower ISO, which requires longer exposures, which means you tend to get fewer of them.
While from a read noise standpoint alone, you may have low read noise at low ISO with a supposedly "isoless" camera, from an actual functional standpoint, ISO 100 is absolutely NOT the same as ISO 800, or 1600, etc. ISO 100 is going to have poor representation of fine, faint details. ISO 1600 is going to have good representation of fine details. Quantization error is also usually going to be lower at higher gain.
To preserve the full fidelity of the signal, and not lose any of the fineness of detail that each and every individual electron holds, you want to use a higher ISO. You generally want to "sample" each electron by a factor of 2-3x, meaning for each electron "input" into the ADC unit, you want to get 2-3 ADU out. There is always going to be some amount of quantization error (not even unity gain is free of quantization error, and it usually isn't selectable anyway), however relative to the scale of each electron at a higher gain, quantization error tends to be much smaller than at lower gain. So there are benefits all around when using higher gain, and all of them are more conducive to acquiring the best data for the kind of ultra faint signals you get with astrophotography.
I don't ever recommend using low ISO for AP anymore. I also don't generally recommend using extremely high ISO settings, as you not only lose dynamic range, but FPN has a gain component and it can get much worse at higher ISO settings. This is particularly true of any secondary amplifiers are used in downstream electronics, which will amplify any signal and noise coming out of the pixels (and Canon definitely uses a secondary amp at high ISO settings). The sweet spot is usually ISO 800, 1600, maybe 3200 on some select few cameras with low FPN.
Upvote
0
Could I share a couple of RAW files with you to help determine the best ISO to use with the 80D? There is so little info out on this camera. Just tell me what you need.jrista said:StudentOfLight said:I generally do not have sufficient focal length for deep sky objects so I usually end up drizzling the interesting portion of frame. I previously shot the LMC under similar LP conditions with my 6D and Tamron 70-200 at 70mm (102x6s).
LMC (102x 6s) by Omesh Singh, on Flickr
The most notable difference is that the stars were not round with the 6D+Tamron image. I suspect it is due to some residual tilt in the IS mechanism as I always get stretching in my images from top left to bottom right.
Regarding ideal settings for the 80D, I read this on DPReview:
https://www.dpreview.com/forums/post/57481095
I'm interested to see how it does at ISO100-200 with long exposure tracked images.
Wouldn't you get the best results on an 80D with an ETTR approach at a lower ISO setting?
Hmm. I am not sure that Jerry nor Roger are correct in that post. Canon has always, to my knowledge, used a 512 ADU 14-bit offset. I've been measuring DSLR data for years now, and I always make sure to measure in the proper bit depth, and across countless Canon DSLRs, I've measured a bias offset of 512 ADU (14-bit).
I think people make the common mistake of converting the original RAW DSLR data into 16-bit, or the mistake of processing it in 16-bit space. Interestingly, the conversion factor between 14-bit and 16-bit is 4 (2^2), and 512*4 = 2048. I do not believe, at least not with any cameras since the 5D II or around there, that Canon has used a 2048 ADU offset.
If Roger is correct that Canon is using a lower offset at ISO 100 and 200, my guess is it would be a 128 ADU 14-bit offset. With lower read noise, that would certainly be possible. It also seems logical to keep a larger offset at higher ISO settings, which amplify noise more, and need more "buffer" room to avoid clipping the signal anyway. I haven't read anything specific to that effect myself, so I honestly cannot say what Canon may have really done with the 80D.
My opinion on which ISO is best is this, and actually in line with Roger's opinions most of the time:
It is better to "oversample" each electron of true signal, than to "undersample" them. By that I mean, you want your gain to be high enough that you are amplifying each electron by a factor of two, if not more. At low ISO, you are, for lack of a better term, "attenuating" the signal, since in most cases you must convert many electrons into each output ADU (analog to digital unit). This is low gain, high quantization. You are quantizing, or grouping together and effectively "squashing", many electrons together and losing a certain amount of information they held when they were separate and distinct.
This is not ideal for astrophotography, since we are chasing extremely faint details, and every single electron matters. It is entirely possible that at a low ISO, say ISO 100, you are converting 2, 3, 5, maybe even more electrons (depends on the pixel size and architecture) into one ADU. Technically speaking, it is possible to recover precision by stacking lots and lots of subs. However that is at odds with using a lower ISO, which requires longer exposures, which means you tend to get fewer of them.
While from a read noise standpoint alone, you may have low read noise at low ISO with a supposedly "isoless" camera, from an actual functional standpoint, ISO 100 is absolutely NOT the same as ISO 800, or 1600, etc. ISO 100 is going to have poor representation of fine, faint details. ISO 1600 is going to have good representation of fine details. Quantization error is also usually going to be lower at higher gain.
To preserve the full fidelity of the signal, and not lose any of the fineness of detail that each and every individual electron holds, you want to use a higher ISO. You generally want to "sample" each electron by a factor of 2-3x, meaning for each electron "input" into the ADC unit, you want to get 2-3 ADU out. There is always going to be some amount of quantization error (not even unity gain is free of quantization error, and it usually isn't selectable anyway), however relative to the scale of each electron at a higher gain, quantization error tends to be much smaller than at lower gain. So there are benefits all around when using higher gain, and all of them are more conducive to acquiring the best data for the kind of ultra faint signals you get with astrophotography.
I don't ever recommend using low ISO for AP anymore. I also don't generally recommend using extremely high ISO settings, as you not only lose dynamic range, but FPN has a gain component and it can get much worse at higher ISO settings. This is particularly true of any secondary amplifiers are used in downstream electronics, which will amplify any signal and noise coming out of the pixels (and Canon definitely uses a secondary amp at high ISO settings). The sweet spot is usually ISO 800, 1600, maybe 3200 on some select few cameras with low FPN.
Upvote
0
I can actually do a full evaluation if you want. It is actually pretty easy. I can evaluate for several different ISOs if you wish.
What I would need for each ISO setting:
2 flats
2 biases (dark frame, shortest exposure possible)
2 darks (dark frame, one at say 30 seconds, one at say 300 seconds, so I can measure dark current rate)
If you get me several sets of such files, one set at least for each ISO, I can get you info about the true gain (e-/ADU), true read noise, actual dark current rate, etc. I can also get FWC, which will allow me to determine dynamic range.
For most accurate results, three distinct sets of data for each ISO would allow me to give you an average of the three sets, which would give you more realistic expectations.
What I would need for each ISO setting:
2 flats
2 biases (dark frame, shortest exposure possible)
2 darks (dark frame, one at say 30 seconds, one at say 300 seconds, so I can measure dark current rate)
If you get me several sets of such files, one set at least for each ISO, I can get you info about the true gain (e-/ADU), true read noise, actual dark current rate, etc. I can also get FWC, which will allow me to determine dynamic range.
For most accurate results, three distinct sets of data for each ISO would allow me to give you an average of the three sets, which would give you more realistic expectations.
Upvote
0
ISO64 said:
Thanks!
I am definitely not the first to create an animated astro photo. There are actually guys who "hunt the smoke" I guess you could say, who have a number of little GIF videos of ultra wide field milky way images with smoking meteors in them.I think I may be the first to actually capture a sequence in a narrow field DSO image, though. New technology is a wonderful thing!
Upvote
0
Andromeda Galaxy
One of my latest images. This is a classic object, but one of the first LRGB images I've made with my new mono camera. Full fill factor for both the luminance (L) channel, which is where most of the integration time goes and where you optimize SNR, detail, etc. And full fill factor (means 100% of the pixels are used) for all three color channels. This is in contrast to a normal DSLR, which has a 50% green and 25% red/blue fill factor, which greatly reduces the sensitivity in those channels.
The data is still light polluted, however I was a bit surprised at how nice it turned out despite that fact:
One of my latest images. This is a classic object, but one of the first LRGB images I've made with my new mono camera. Full fill factor for both the luminance (L) channel, which is where most of the integration time goes and where you optimize SNR, detail, etc. And full fill factor (means 100% of the pixels are used) for all three color channels. This is in contrast to a normal DSLR, which has a 50% green and 25% red/blue fill factor, which greatly reduces the sensitivity in those channels.
The data is still light polluted, however I was a bit surprised at how nice it turned out despite that fact:
Upvote
0
