Posted by John Freeman on June 23, 2014 at 9:59 AM
Thought I would share one shot I got about 2:30 from the Moraine Park
Overlook Sat. morning. I was able to get the "Bucket List" shot I was hoping for. Just wish I could have turned off the lights in Denver while I was shooting this.
This is a 180 degree to vertical Gigapan made up of 24 individual (4
rows x 6 columns) images using Sigma 35mm f/1.2 lens set to f/2.0 3.2
sec. ISO6400. I did a 16:9 crop down to 22062 x 12422 pixels for a
finished size of 73.5" x 41.4" @ 300ppi.
Posted by Glenn Randall on June 23, 2014 at 12:47 PM
Kudos on the technical achievement of putting such a multi-row panorama together, but I'm afraid I can barely see the Milky Way in this jpeg. You could have gone to about 14 seconds in exposure before the stars began to streak noticeably. That might have given you issues with the Denver skyglow blowing out, but would have produced a much brighter Milky Way.
Posted by John Freeman on June 23, 2014 at 1:32 PM
Glenn, I agree that the jpg reduced to this size leaves the Milky Way barely visible. Looks much better in Raw file.
Nice job capturing the full Milky Way. To bring out the Milky Way more, you might try a Levels adjustment in Photoshop with Tony Kuyper's "Shadow Darks" luminosity mask. The mask will allow you to bring out more contrast in the dark areas, while preventing you from blowing out the highlights. Push the right slider to the left a long ways and then the center slider to the right until it looks dark enough to you. This will also bring out the colors a lot, which may make the blue / cyan cast too pronounced. So you may need to reduce the blues and cyans, maybe even the greens a bit.
Posted by John Freeman on June 23, 2014 at 11:21 PM
Grant, thanks for the suggestion to use TK Masks. I have the Tony Kuyper's Luminosity masks but have only played with it a little. I would be interested to hear your experiences with it as I have noticed that you use it some. My problem is that with a file this size with over 270 Megapixels each layer you add quickly starts bogging down Photoshop. I have 48 GB of memory in my Hex core machine along with many Terabytes of disc drive space and when I start using TK masks it slows to a crawl on files of this size. Your point of just using "Shadow Darks" is noted. That keeps the # of layers to a minimum. As my schedule permits, I plan on working on this and several other panoramas I did last Friday night.
Hello John,
Nice that you could get a 180 degree field of view without major impact from Front Range light pollution. Regarding processing, using the left or right slider easily results in clipping the highs or lows. The use of ISO 6400 saturated the brighter stars resulting in more clipping than needed. Not sure what camera you used, but with most cameras, ISO 1600 is better for maintaining star colors (ISO does not change sensitivity). My preference is more natural colors; the dark night sky while we are at solar max is not blue, and there are no blue stars. Some of the star clouds in your scene are hydrogen alpha light, thus quite red. I find your scene too monochrome blue. At ISO 1600, I pull one stop in exposure during raw conversion to prevent stars from saturating, then boost low signals with curves in post production after mosaicing. Examples (post processing only with curves, and a little usm):
http://www.clarkvision.com/galleries/gallery.greece/web/santorini.night.rnc06.07.2014.IMG_5176-424_54frames.e-bin12x12s.html
Most stars in the night sky are yellow, orange and red. The hottest stars are bluish-white. Colors in the night sky are warm and inviting, not cold and blue. I like the warm mood of the natural light, not the cold unnatural blue so popular with today's digital photographers.
Sigma 35 mm f/1.2? Do you mean 1.4, or is there really an f/1.2?
I do like your composition. I would like to see a longer exposure to give more detail in the sky and some detail in the land.
John, TK's masks definitely do take up more memory. When I had 8 gigs of RAM and did a lot of adjustments on stitched images, I'd switch to working on something else while waiting for it to process the file. I'd sometimes flatten the image and resave it to decrease file size. It wasn't the ideal workflow but I didn't really have any other options. Now I have 32 gigs of RAM and 256 GB of steady state memory and it flies through very large stitched images with lots of layers. I'm surprised it bogs down with 48GB, but I know the steady state drive on mine really helps. If you're working with 16-bit files, you might consider converting to 8-bit to cut the file size in half. The only real issue with 8-bit is that it could create some banding in areas with subtle gradients, but I haven't encountered a problem with this with large stitched images at night.
Sometimes I'll use two TK masks to really bring out the contrast without clipping the highlights. I suggested Shadow Darks with this particular image but you can try different ones... in general, the darker the sky appears before applying the mask, the darker the mask I'll use.
Posted by John Freeman on June 25, 2014 at 2:54 AM
Roger & Grant,
Thanks for the suggestions. I am shooting with a Canon 5D3. The lens was a Sigma 35mm f/1.4 as Roger correctly noted. I shot the photos with Daylight white balance which tends (to my eyes) to be overly rich in reds. I think this may be a Canon characteristic. I went on a Night Photography workshop last summer with Mike Berenson and noticed that his Nikon D800 tended to have a tendency to be less red than my Canon.
In an attempt to correct for this I applied a Tungsten filter in Camera Raw. I agree this is results in overly blue hues and in a later edit I adjusted hue to the right a bit to reduce the blue effect. The night I shot this there were some thin high level clouds (center, lower third like) that were being illuminated by the Denver lights. With the Daylight white balance these clouds came out bright reddish orange which is not at all what they looked like to my naked eye.
Roger, I find your suggestion to use ISO 1600 (ISO does not change sensitivity) interesting. I would like to better understand your theory behind that statement. As an electronics engineer with a fair amount of Analog/Digital experience I have a fairly good understanding of the theory that goes on behind the electronics inside the camera. Essentially, increasing ISO is turning up the gain in the analog portion of the circuitry. This increases sensitivity and of course results in more noise in the darker regions of the image. Likewise ISO 1600 has less noise but requires longer exposure times. Heating of the sensor during longer exposure can also result in noise. So in a nutshell, as I understand it the trade offs between ISO and exposure for night photography is a mixed bag. I think what your saying is that the sensor itself doesn't change sensitivity with changes in ISO. (I agree.) Rather, the gain of the downstream electronics is turned up before the signal reaches the A/D converter that changes the signal from an analog value to a digital number. So the result is that ISO 1600 has a better Signal to Noise Ratio yielding better color rendition.
In Mike Berenson's workshop he suggested ISO 6400 and exposure times of 20-30 seconds. Here is a shot I did during the workshop. http://abbascreationsphotography.com/Milky_Way_Over_Lake_Irene.htm That was shot at ISO 6400, f/2.8 & 30 sec. Light pollution from Granby didn't get blown out nearly as bad as Denver lights did in my recent photo with settings at f/2.0, ISO 6400 and an exposure time of only 3.2 sec. For the 180 degree shot I spot metered on the brightest Denver lights using the Expose to the Right method setting at +2 stops. This yielded the 3.2 sec. exposure time. Then during RAW processing I moved the Blacks and Shadows sliders to the right to bring out the darker levels. I find it interesting that I am tending to use significantly shorter exposure times than Mike and many other night photographers recommend and still having the brighter stars blown out. What are your thoughts on this?
Regarding ISO, there is only one sensitivity to CMOS/CCD sensors and that is set
by the quantum efficiency. Changing ISO simply changes the post sensor gain,
thus the range of light to be digitized. Higher ISO is higher gain, but
not in light sensitivity.
The noise in images is a combination of:
1) photon noise (Poisson statistics), 2) sensor read + downstream electronics
noise, and 3) noise from dark current (which is pretty minimal in a few
second exposure night image. Read noise at ISOs above 800 is also
minimal and is usually swamped by photon noise from the scene, even
the night sky in most cases. Your short ~3 second exposures may be slightly
impacted by read noise. Thus, the dominant noise source in daytime
images as well as night images you and I show in this thread is from
photon noise. The signal-to-noise ratio (S/N) is proportional to the number
of photons collected by the sensor, and that is entirely controlled by
exposure time and aperture area (not f/ratio), and is independent of ISO.
The reason to use ISO 1600 is because Canon cameras have pattern noise
that becomes a factor in many cameras at ISOs below 1600 (Nikons are
better in that regard, but clip the signal at zero).
For more information on these effects see my (all free and open) articles:
Nightscapes:
http://www.clarkvision.com/articles/nightscapes/
ISO and exposure:
http://www.clarkvision.com/articles/iso/
In metered exposure, when you change ISO, the camera changes exposure time
and/or f/ratio and those change the amount of light recorded, not the
ISO.
For your 5D Mark 3, see my sensor analysis at:
http://www.clarkvision.com/articles/evaluation-canon-5diii/
At iso 6400, the apparent read noise is 2.5 electrons and the range digitized
is 0 to 1000 photons (photoelectrons) with a dynamic range of 8.6 stops.
At ISO 1600, the read noise is 3.6 electrons and the range digitized is
0 to 4050 photons for a dynamic range of 10.1 stops. The A/D converter
has enough precision to digitize the smallest signals at iso 1600, and light
from the night sky will dominate the red noise in only a couple of second
exposure with your Sigma lens. Thus, you gain 1.5 stops of dynamic range with
effectively no loss in low light recording.
Table 2 on the above page shows the pattern noise for the 5D3. By ISO 1600,
it pattern n noise is small enough so that it can be ignored and there is really
no benefit for going to higher ISO, and a detriment in saturating more stars and
losing color.
If you expose a scene at ISO 1600 and 6400 with the same exposure time, and ignore
an saturated pixels, the S/N will be essentially the same.
For nightscapes, the larger the aperture area times the longer exposure time
will make for the recording the faintest light. I always us the Sigma 35 f/1.4
wide open for night photos. Stars will begin to show trails after about 6 to
8 seconds unless pointed toward the celestial poles. (See Table 3 on my
nightscapes page.)
Using sliders in photoshop too easily can result in clipping, whether in ACR
or in photoshop. Same with other image processing systems. I only use curves
and never change the lower left or upper right points.
Regarding the color of the night sky, your Lake Irene image has a pretty
natural color balance. The dark, moonless night sky is dominated by airglow,
chemiluminescence induced by solar UV which excites oxygen, hydroxyl and other
molecules in our upper atmosphere creating the exact same colors that we see
in polar aurora, only airglow is all over the globe. Usually green or red
emission lines dominate from oxygen and hydroxyl like in polar aurora. Unless
scattered moonlight dominates, making blue due to Rayleigh scattering, the night
sky is green, or red or combinations of the two.
External to our Solar System, there are many processes creating colors, from brown
dust in the Milky way, to red and green emission nebula to scattered starlight off
of interstellar clouds, making blue from Rayleigh scattering. If one torques
the white balance so everything is a shade of blue, one does not see all these
wonderful colors. There are no blue or green stars. The hottest stars are
bluish-white. And the human eye can see colors in the brighter stars and
celestial objects if well dark adapted. (Often I see photographers with
bright red flashlights, and that strong color warps color perception and
destroys night sensitivity; lights should be dim even if red in order to
not hurt night vision.) If one wanted to make night images like what one sees,
it would be mostly gray, not blue. The true color of the night sky is warm,
inviting, and full of wonderful colors.
More on the color of the night sky:
http://www.clarkvision.com/articles/color.of.the.night.sky/
And finally, protecting your night vision:
http://www.clarkvision.com/articles/night.observing.lights/
Hope this helps.
Roger
Posted by John Freeman on June 26, 2014 at 1:11 AM
Roger,
Yes your detailed post really helps. Thanks for the extensive critique and detailed explanations. I am sure many of us will benefit. After I posted my response last night I went to your web sight and discovered your many well written articles. Thanks for being willing to share your knowledge with the rest of us.I will be reviewing them in detail. This coming weekend I will be in Eastern rural WY on the ranch where I grew up. I plan on attempting again to capture some 180 & 360 degree sequences. Will be experimenting with your recommendations.
Discussion
9 repliesOverlook Sat. morning. I was able to get the "Bucket List" shot I was hoping for. Just wish I could have turned off the lights in Denver while I was shooting this.
This is a 180 degree to vertical Gigapan made up of 24 individual (4
rows x 6 columns) images using Sigma 35mm f/1.2 lens set to f/2.0 3.2
sec. ISO6400. I did a 16:9 crop down to 22062 x 12422 pixels for a
finished size of 73.5" x 41.4" @ 300ppi.
Nice that you could get a 180 degree field of view without major impact from Front Range light pollution. Regarding processing, using the left or right slider easily results in clipping the highs or lows. The use of ISO 6400 saturated the brighter stars resulting in more clipping than needed. Not sure what camera you used, but with most cameras, ISO 1600 is better for maintaining star colors (ISO does not change sensitivity). My preference is more natural colors; the dark night sky while we are at solar max is not blue, and there are no blue stars. Some of the star clouds in your scene are hydrogen alpha light, thus quite red. I find your scene too monochrome blue. At ISO 1600, I pull one stop in exposure during raw conversion to prevent stars from saturating, then boost low signals with curves in post production after mosaicing. Examples (post processing only with curves, and a little usm):
http://www.clarkvision.com/galleries/gallery.greece/web/santorini.night.rnc06.07.2014.IMG_5176-424_54frames.e-bin12x12s.html
http://www.clarkvision.com/galleries/gallery.nightscapes/web/mesa.arch.nightscape.rnc05.02.2014.IMG_1685-1742.g-bin10x10-c1s.html
Most stars in the night sky are yellow, orange and red. The hottest stars are bluish-white. Colors in the night sky are warm and inviting, not cold and blue. I like the warm mood of the natural light, not the cold unnatural blue so popular with today's digital photographers.
Sigma 35 mm f/1.2? Do you mean 1.4, or is there really an f/1.2?
I do like your composition. I would like to see a longer exposure to give more detail in the sky and some detail in the land.
Roger Clark
Sometimes I'll use two TK masks to really bring out the contrast without clipping the highlights. I suggested Shadow Darks with this particular image but you can try different ones... in general, the darker the sky appears before applying the mask, the darker the mask I'll use.
Thanks for the suggestions. I am shooting with a Canon 5D3. The lens was a Sigma 35mm f/1.4 as Roger correctly noted. I shot the photos with Daylight white balance which tends (to my eyes) to be overly rich in reds. I think this may be a Canon characteristic. I went on a Night Photography workshop last summer with Mike Berenson and noticed that his Nikon D800 tended to have a tendency to be less red than my Canon.
In an attempt to correct for this I applied a Tungsten filter in Camera Raw. I agree this is results in overly blue hues and in a later edit I adjusted hue to the right a bit to reduce the blue effect. The night I shot this there were some thin high level clouds (center, lower third like) that were being illuminated by the Denver lights. With the Daylight white balance these clouds came out bright reddish orange which is not at all what they looked like to my naked eye.
Roger, I find your suggestion to use ISO 1600 (ISO does not change sensitivity) interesting. I would like to better understand your theory behind that statement. As an electronics engineer with a fair amount of Analog/Digital experience I have a fairly good understanding of the theory that goes on behind the electronics inside the camera. Essentially, increasing ISO is turning up the gain in the analog portion of the circuitry. This increases sensitivity and of course results in more noise in the darker regions of the image. Likewise ISO 1600 has less noise but requires longer exposure times. Heating of the sensor during longer exposure can also result in noise. So in a nutshell, as I understand it the trade offs between ISO and exposure for night photography is a mixed bag. I think what your saying is that the sensor itself doesn't change sensitivity with changes in ISO. (I agree.) Rather, the gain of the downstream electronics is turned up before the signal reaches the A/D converter that changes the signal from an analog value to a digital number. So the result is that ISO 1600 has a better Signal to Noise Ratio yielding better color rendition.
In Mike Berenson's workshop he suggested ISO 6400 and exposure times of 20-30 seconds. Here is a shot I did during the workshop. http://abbascreationsphotography.com/Milky_Way_Over_Lake_Irene.htm That was shot at ISO 6400, f/2.8 & 30 sec. Light pollution from Granby didn't get blown out nearly as bad as Denver lights did in my recent photo with settings at f/2.0, ISO 6400 and an exposure time of only 3.2 sec. For the 180 degree shot I spot metered on the brightest Denver lights using the Expose to the Right method setting at +2 stops. This yielded the 3.2 sec. exposure time. Then during RAW processing I moved the Blacks and Shadows sliders to the right to bring out the darker levels. I find it interesting that I am tending to use significantly shorter exposure times than Mike and many other night photographers recommend and still having the brighter stars blown out. What are your thoughts on this?
Regarding ISO, there is only one sensitivity to CMOS/CCD sensors and that is set
by the quantum efficiency. Changing ISO simply changes the post sensor gain,
thus the range of light to be digitized. Higher ISO is higher gain, but
not in light sensitivity.
The noise in images is a combination of:
1) photon noise (Poisson statistics), 2) sensor read + downstream electronics
noise, and 3) noise from dark current (which is pretty minimal in a few
second exposure night image. Read noise at ISOs above 800 is also
minimal and is usually swamped by photon noise from the scene, even
the night sky in most cases. Your short ~3 second exposures may be slightly
impacted by read noise. Thus, the dominant noise source in daytime
images as well as night images you and I show in this thread is from
photon noise. The signal-to-noise ratio (S/N) is proportional to the number
of photons collected by the sensor, and that is entirely controlled by
exposure time and aperture area (not f/ratio), and is independent of ISO.
The reason to use ISO 1600 is because Canon cameras have pattern noise
that becomes a factor in many cameras at ISOs below 1600 (Nikons are
better in that regard, but clip the signal at zero).
For more information on these effects see my (all free and open) articles:
Nightscapes:
http://www.clarkvision.com/articles/nightscapes/
ISO and exposure:
http://www.clarkvision.com/articles/iso/
In metered exposure, when you change ISO, the camera changes exposure time
and/or f/ratio and those change the amount of light recorded, not the
ISO.
For your 5D Mark 3, see my sensor analysis at:
http://www.clarkvision.com/articles/evaluation-canon-5diii/
At iso 6400, the apparent read noise is 2.5 electrons and the range digitized
is 0 to 1000 photons (photoelectrons) with a dynamic range of 8.6 stops.
At ISO 1600, the read noise is 3.6 electrons and the range digitized is
0 to 4050 photons for a dynamic range of 10.1 stops. The A/D converter
has enough precision to digitize the smallest signals at iso 1600, and light
from the night sky will dominate the red noise in only a couple of second
exposure with your Sigma lens. Thus, you gain 1.5 stops of dynamic range with
effectively no loss in low light recording.
Table 2 on the above page shows the pattern noise for the 5D3. By ISO 1600,
it pattern n noise is small enough so that it can be ignored and there is really
no benefit for going to higher ISO, and a detriment in saturating more stars and
losing color.
If you expose a scene at ISO 1600 and 6400 with the same exposure time, and ignore
an saturated pixels, the S/N will be essentially the same.
For nightscapes, the larger the aperture area times the longer exposure time
will make for the recording the faintest light. I always us the Sigma 35 f/1.4
wide open for night photos. Stars will begin to show trails after about 6 to
8 seconds unless pointed toward the celestial poles. (See Table 3 on my
nightscapes page.)
Using sliders in photoshop too easily can result in clipping, whether in ACR
or in photoshop. Same with other image processing systems. I only use curves
and never change the lower left or upper right points.
Regarding the color of the night sky, your Lake Irene image has a pretty
natural color balance. The dark, moonless night sky is dominated by airglow,
chemiluminescence induced by solar UV which excites oxygen, hydroxyl and other
molecules in our upper atmosphere creating the exact same colors that we see
in polar aurora, only airglow is all over the globe. Usually green or red
emission lines dominate from oxygen and hydroxyl like in polar aurora. Unless
scattered moonlight dominates, making blue due to Rayleigh scattering, the night
sky is green, or red or combinations of the two.
External to our Solar System, there are many processes creating colors, from brown
dust in the Milky way, to red and green emission nebula to scattered starlight off
of interstellar clouds, making blue from Rayleigh scattering. If one torques
the white balance so everything is a shade of blue, one does not see all these
wonderful colors. There are no blue or green stars. The hottest stars are
bluish-white. And the human eye can see colors in the brighter stars and
celestial objects if well dark adapted. (Often I see photographers with
bright red flashlights, and that strong color warps color perception and
destroys night sensitivity; lights should be dim even if red in order to
not hurt night vision.) If one wanted to make night images like what one sees,
it would be mostly gray, not blue. The true color of the night sky is warm,
inviting, and full of wonderful colors.
More on the color of the night sky:
http://www.clarkvision.com/articles/color.of.the.night.sky/
And finally, protecting your night vision:
http://www.clarkvision.com/articles/night.observing.lights/
Hope this helps.
Roger
Yes your detailed post really helps. Thanks for the extensive critique and detailed explanations. I am sure many of us will benefit. After I posted my response last night I went to your web sight and discovered your many well written articles. Thanks for being willing to share your knowledge with the rest of us.I will be reviewing them in detail. This coming weekend I will be in Eastern rural WY on the ranch where I grew up. I plan on attempting again to capture some 180 & 360 degree sequences. Will be experimenting with your recommendations.
Thanks,
John