Saturday, February 06, 2016

Moon, Venus, and Mercury - Feb 6, 2016

I got up really early again this morning to catch the Moon, Venus, and Mercury dancing in the pre-dawn sky.  Here are a couple of my best shots.  The first one was taken from Larimer County Road 38E, where it curves into the foothills.  It was taken at 5:55 AM with a Canon 7DmkII and an EF 100-400 f/4.5-5.6L II lens, at 100mm f/4.5.  It is a 1.6 second exposure at ISO 800:

For the curious, here is a map of the foreground showing what is visible in the 12.8 degree field-of-view of the lens:

I drove home, and snapped one more picture above the neighbors' rooftops.  Same camera/lens/exposure settings:

OK, *now* I can stop waking up early in the morning.

Thursday, February 04, 2016

Five Planets and the Moon

I got up this morning very early to make an animation of Jupiter, Mars, Saturn, the Moon, Venus, and Mercury as they rose in the early morning sky.  Here is the finished product:

It was very cold - 7 degrees Fahrenheight - and very early.  This is a 30-minute sequence starting at 5:39 AM and finishing at 6:09 AM.  All of the images have the same parameters:
  • Canon 7DmkII camera
  • Canon 8-15mm f/4L lens, locked at 10mm (widest useful zoom with this lens on a 1.6x crop camera)
  • 8 second exposures at ISO 3200
  • Image sequencing performed by Canon's EOS Utility.  A capture was started every 15 seconds.

That camera/lens combo yields a 180-degree diagonal field-of-view.  Jupiter and Mercury are about 120 degrees apart in these images, so this gives some extra room for animation and cropping.

Capturing all five planets in the pre-dawn skyglow is surprisingly tricky, especially when city light pollution is contributing to a murky lower atmosphere.  I was located at the north end of the parking lot for the Foothills Trail, above Horsetooth Reservoir on County Road 23.  Here is the best of the single frames I captured.  It is not bad - but the skyglow completely obscures Mercury, and even mostly hides Venus - even though the image was being processed as a 16-bit-per-component TIFF:

To fix this, I created a correction layer in Photoshop.  I first began by blurring the entire image with a Gaussian blur of about 25 pixels.  I then selected the region near the treeline / horizon, and did a "Maximum" filter.  This shrank the size of the trees down so they did not interfere with the mask I was making.  I then selected the whole sky, and copied it to a new layer.  This new layer was then hit with a "Minimum" filter over the entire sky.  This removed the stars and most of the planets.  I had to use the "Healing Brush" tools to finish removing the Moon and Jupiter.  A little touch-up on the skyline, and I had this mask:

I then recorded an Action which selected this mask, pasted it into each image, changed its blend mode to "Subtract", decreased its opacity to 50%, merged the layers, and then cropped and resized the image to 1920x1080 (HD resolution).  I replayed the Action with Photoshop's automation tools, then assembled the whole animation in Premiere Pro.

Here is that same shot as the first one, cleaned up pretty much as described (with some extra care taken with the skyline, and some additional contrast boost to bring out the planets):

And here are the labels to show the named stars and planets visible in the picture (not counting the Earth, of course!):

Fun trivia: You can see the trails of 5 early morning runners at about 0:10 and 0:27 in the video.  The last shot was an accidental selfie - it is hard to tell where the edge of the frame is when you are shooting with a 180-degree field-of-view!  I count six cars that passed on County Road 23.

Edit (Feb 5, 2016): Here is a different way to visualize the sequence of image I took - star trails.  Without labels:

And with labels:

Edit 2 (Feb 6, 2016): I refined my processing formula.  I gave GradientXterminator a try, ,but was not very happy with it.  So I spent the time to make a very detailed foreground selection mask, and then changed the formula to:

  • Duplicate the image into a new layer
  • Load the foreground selection mask
  • Select the new (topmost) layer
  • Do a Maximum filter of 25 pixels width ("preserve roundness")
  • Remove the selection
  • Do a Gaussian Blur of 20 pixels width
  • Do a Minimum filter of 25 pixels width ("preserve roundness")
  • Adjust Levels to a gamma of 0.75 and a maximum of 192
Here is the updated video:

And, inspired by my brother-in-law Edward Plumer, I made a version of my best image with constellation lines and labels.  Here is the reprocessed best image:

And here it is with the labels:

Edit 3 (Feb 7, 2016): Check out my brother-in-law Edward Plumer's blog - we have a friendly competition going for this kind of thing.

Copyright note: this page, the video, and the images linked are Copyright 2016 by Ross Cunniff under Creative Commons.  Any use is fair use as long as attribution is made.  The music to the video is "Starway" by Jaja - - also licensed under Creative Commons.

Monday, September 28, 2015

Total Lunar Eclipse, September 27-28 2015

Jill and I went eclipse chasing this evening.  I did initial research on the Clear Dark Sky site to see what conditions were going to be like.  After digging, I decided that somewhere south and east of Briggsdale was likely to have clear skies.  We ended up on the grounds of an old church called Osgood Church.  We were greeted by this raptor, which we took as an auspicious omen:

We were rewarded with gorgeous skies - only a few wispy clouds floated through occasionally.  Here is an all-sky shot taken with my Canon EF 8-15mm f/4.0L lens on a Canon 5DmkII camera that shows how nice the skies were:

Jill and I are there in the foreground next to the telescopes.  You can see the eclipsed Moon, as well as the Milky Way and the skyglow from Fort Morgan, Sterling, and smaller towns off past the horizon.

For the eclipse photos, I used a Canon 7DmkII camera on an Orion ED80T CF refractor - an 80mm diameter, f/6.0 480mm focal length triplet.  I mounted it all on my handy portable iOptron MiniTower Pro.  We were nearly faced with disaster when my inverter failed to power on - fortunately, our Jeep had plenty of 12V power to drive the mount.

I took shots every two minutes - and will eventually make an animation - but here is one of the better shots at mid-eclipse:

This was a 4 second exposure at ISO 3200.  In between my every-two-minute exposures, I wandered around and took pictures of our surroundings.  Here is another selfie of Jill and me with the eclipse - 30 second exposure, ISO 3200, Canon EF100-400mm f/5.6L lens on the 5DmkII - the long exposure streaked the moon and stars):


Here is a context view of our trusty steed, my telescope setup, and Jill taking pictures with her camera, with the church in the background.  Taken by moonlight, hand-held (I love image stabilization...)

I got one shot of the ending partial phase near the chimney of the church - I used a flash to fill in the foreground and tightened down the aperture to f/18 to increase the depth of field:

Staying up for another couple of hours, I made this collage and then went to bed.  Some time later this week I'll try for an animation.

Edit: I must have been sleep-deprived last night.  I re-used a template I had made for the 2010 Lunar Eclipse and forgot that it had a left-right reversal built in due to the fact that those photos had been taken with a mirror diagonal.  Here is the corrected collage:

Edit: here is the YouTube time-lapse video I made:

Wednesday, August 05, 2015

Armchair science

I noted with interest this update from the DSCOVR mission:

Cool!  The DSCOVR mission is an Earth-imaging mission that takes a continuous stream of Earth images to monitor it for climate and other changes over time.  It happened to catch the Moon crossing the face of the Earth (go to the link for the animation - you'll be glad you did).  This will happen twice a year, when the plane of DSCOVR's orbit intersects the plane of the Moon's orbit in line with the Earth.

The DSCOVR web page says that the satellite orbits "a million miles" from Earth.  It turns out, you can calculate that distance just from this image and from the known sizes of the Earth and the Moon.  The Earth has a mean radius of 3,959 miles, and the Moon has a mean radius of 1,079 miles.  The ratio between the two is 3.67 - meaning the Earth is 3.67 times as big as the Moon.  But if you open up that image in Photoshop and measure the Earth and Moon in pixels, you will find this:

The Earth is 1595 pixels big, and the Moon is 596 pixels big.  This is a ratio of 2.72 - what's up with that?  The answer is, perspective.  The Earth is farther away in this picture than the Moon, so it looks smaller.  We know that the Earth and the Moon are roughly 250,000 miles apart.  So, how far is DSCOVR from the Earth?  Based on these measurements, this diagram, and a little trigonometry, we can calculate it:

Point "S" is the spacecraft.  Point "E" is the north pole of the Earth.  Point "M" is the north pole of the Moon.  Point "X" is the center of the Moon, and point "Y" is the center of the Earth.  We know the following distances:

  • Point X to point M: 1,079 miles (and 596 pixels)
  • Point Y to point E: 3,959 miles, (and 1,595 pixels)
  • Point X to point Y: 250,000 miles

The dashed line shows the projection of the edge of the moon from the spacecraft's point-of-view. The pixel measurements are taken at the same distance - since they are both projected against the same plane (in reality, the plane of the imaging system).  The point labeled "Mp" is that projected point in the image.  We have similar triangles which can help us figure out the distance D:

  • Triangle Y-S-Mp
  • Triangle D-S-M

The image shows the calcuations, which gets a distance of 965,000 miles.  Close enough to "one million miles" for government work!  Especially given the precision with which I specified the Earth-Moon distance (the actual distance varies between 221,457 miles and 252,712 miles).

Finally, here is a picture showing how big the Earth would have looked if it had been at the same distance as the Moon in the DSCOVR photo (I show it as a dim ghost behind the original image):

Of course, we would all be having a very bad day if the Earth and Moon had been at the same distance at that time...

Edit (Aug 5 7:15PM MDT): I had an email exchange with Phil Plait, the Bad Astronomer, and he pointed out I can get the exact Earth-Moon distance on the day in question (July 16) from the United States Naval Observatory website.  On that date, it was 243,388 miles.  So, plugging that into the formula I get that DSCOVR was 940,000 miles from Earth on that date.  L1 is about 929,000 miles from the Earth-Moon barycenter, which in this photo is about 2900 miles closer to the spacecraft than Earth is, making the L1 point 932,000 miles from the center of the Earth.  Which is very close to what I calculated - less than 1% difference.  Part of this is difference is attributable to the inaccuracy of measuring the pixel sizes of the Earth and Moon, and part is attributable to the fact that DSCOVR is not actually *at* the L1 point; rather, it is orbiting about it in a complex pattern known as a "Lissajous Orbit".

Given that complex orbit, this method is probably the most accurate way I have of conveniently determining DSCOVR's distance from the Earth.  At least on the days that the Moon occults or is occulted by the Earth.

Phil also pointed out that the moon will look smaller due to perspective when it is on the other side of its orbit (the back of my envelope tells me it will be 59% of the apparent size) - so it might look something like this by comparison:

And  yes, I'm showing the other face of the moon - that's the one DSCOVR will see in that geometry.  We'll see...

Saturday, August 01, 2015

*Big* Pluto Icosahedron

I made a Pluto icosahedron from my previous post, and I decided it was nice - but it needed to be bigger.  So I spent today figuring out the geometry to maximize size while minimizing paper.  Here's what I came up with - an icosahedron twice the size as the previous one, taking 4 sheets of paper.  It will be a little more complicated to assemble - you have to match all of the tabs just so - but it makes a nice model when it is complete.

It is in PDF format, since it is four pages.  Without further ado - Pluto.

Thursday, July 30, 2015

More Icosahedral Models

A few years ago, I converted various planet maps to icosahedral models.  You can print these out, cut along the solid lines, fold along the dotted lines, and have a nice icosahedron to play with.  Since then, two more robot spacecraft have created global maps of planetary bodies.  Most recently, the New Horizons spacecraft flew past Pluto.  Here is an icosahedral projection of the data we have so far:

Also fairly recently, the Dawn spacecraft has orbited both Vesta and Ceres, two of the largest asteroids (or "dwarf planets" as the new nomenclature has it) in the Solar System.  Here are icosahedral projections of those two.  First, Vesta:

Next, Ceres - note the mysterious white spots toward the upper right:

All of the data used to create these projections was taken from Steve Albers' Planetary Maps Page, which I highly recommend.

Edit: A PDF file of helpful hints on assembling these can be found here.

Sunday, April 19, 2015

Addendum - new vs. old Canon 100-400mm

In my previous post on this lens, I used a 1.4x teleconverter on both lenses.  I got a request to repeat the test without the teleconverters, so here are the results.

These are all shot at ISO 400, f/5.6 at 400mm, 1/1600s, autofocus with the center point, RAW.  It is later in the day, and it looks like light clouds and shadows moving in the breeze are slightly changing the lighting between the shots, so caveat emptor.  First the whole image with the old lens:

Next, the new lens:

Here are the locations of the crops I took:

First, the center crop.  The old lens:

The new lens - again, sharper even in the center (although the moving shadows from the windy day make it a little harder to tell):

The left edge crop, the old lens:

The new lens - significantly sharper:

And now, the corner crop, the old lens:

And the new lens.  It is no contest:

The chromatic aberration is not as bad with the old lens without the teleconverter - but the sharpness is still much worse than the new lens.  And autofocus is slower, and the IS is not as capable, and...

Edit - one last thing.  Reading the MTF charts, it seems as if the new lens with a teleconverter should be better than the old lens without a teleconverter.  So, here is a direct comparison - I adjusted the white balance and brightness a bit to try to make the images match a bit better, but otherwise did not modify the images.  I resized the image from the new lens with a 1.4x teleconverter by 1/1.4 and took a crop from the lower left corner of that:

Here is the old lens, at 400mm f/5.6:

And here is the new lens, at 560mm f/8.0 with the 1.4x teleconverter, scaled as mentioned by 1/1.4 or 71.43% to match the pixel scale (without using a sharpening downsample filter):

I'm sold.

Old versus new Canon 100-400mm zoom

Late last year, Canon introduced a new lens, the EF 100-400mm f/4.5-5.6L IS II USM (say that three times fast!).  Reviews held that it was a significant improvement over its predecessor (which is named the same, but does not have the "II" between "IS" and "USM"), and, in fact, was equivalent to their 400mm prime in image quality, but with tons of feature advantages.  I acquired the new lens this week, and have put it through a test to compare, stealing some techniques from my brother-in-law Edward Plumer.

Note - I have an addendum without the teleconverter in the mix.  It does not change the conclusion.

I set the lenses up on a tripod, pointed at a fork in the tree.  The contrast and colors in the tree are similar to those I find when taking wildlife shots.  Image stabilization was disabled in both lenses.  The new lens was attached to a Canon Expander 1.4x III and the old was attached to a Canon Expander 1.4x II (the previous model).

The lenses were attached to a Canon 7D Mark II camera, which was allowed to autofocus with the center point (a significant advantage of this camera - it autofocuses up to a minimum aperture of f/8).  Exposure parameters were manual - ISO 400, 1/640s exposure, f/8 at a 560mm net focal length - and images were shot RAW to avoid in-camera lens correction (although it may still have happened, I did not dig through the menus to find out).  I put the camera on a 10s timer to reduce lens shake from pressing the shutter button.  I took three images with each camera, and selected the best of the three for this analysis.

Here are the results.  First, the old lens:

 Not a bad image - the bokeh is reasonable, the contrast seems OK.  Now the new lens:

Again, at this size, not a bad image.  The bokeh seems a little softer, which is good.  If I only ever looked at pictures this size, it would be a hard choice (although the new lens auto-focuses significantly faster than the old).  However, with wildlife, the images are almost always cropped since the birds are so tiny and so far away.  So, I took some crops.  Here are the locations of the crops:

First, the center crop.  This is the point at which autofocus was locked.  Here is the old lens:

Hmm.  A little soft.  Let's take a look at the new lens:

Obviously better.  Maybe 50% better?  Take a look now at the left edge crop from the old lens:

Definitely soft.  Is that a little chromatic aberration creeping in?  Here comes the new lens:

It's no contest.  And for wildlife, edge performance is important, even though I autofocus for the center.  Sometimes the animal is moving so fast that my picture is snapped with it close to the edge of the frame.  I've lost many shots to the softness of the old lens.  And now, for the coup de grace, the corner crop.  The old lens:

This crop is, simply put, a mess.  Chromatic aberration is all over the place, and the image is just ugly.  Here is the new lens:

The detail in the corner on the new lens looks about as sharp as the detail in the center.  In fact, the detail in the corner is shaper than the center on the old lens.

I'd say we have a winner.  Other improvements in the new lens include: a 9-blade iris (versus 8 in the old lens), image stabilization supposedly good to 4 stops (vs. 1.5 in the old lens), improved anti-reflection coatings, rotating zoom ring versus push-pull, improved weather sealing, and an improved lens cap which makes it possible to put it on and take it off with the lens hood attached.

In summary - well worth the 50% price premium.  This is probably a 10x better lens.