Pages: 1 [2]   Go Down

Author Topic: New Article - Sony a7r III Pixel Shift  (Read 7276 times)

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: New Article - Sony a7r III Pixel Shift
« Reply #20 on: November 23, 2017, 02:07:47 pm »

Hi Kevin,

Thanks for constructive feedback!

It was interesting to see the Phase One shot side by side, that camera can yield a lot of image quality.

It is an interesting time we live in, with smaller systems getting more and more capable.

I don't know about Phase, but it seems that Hasselblad really made a success story with X1D and the Fuji GFX also seems to be very popular. Makes a lot of sense to optimize the system for the smaller sensor size.

Best regards
Erik

Erik and Jonathan
Thanks for your input and further explanation.  I included the Phase One image just because I had the camera with me and thought it would be interesting to compare. I should have shot at a smaller f/stop great DOF on Phase One. Frankly, I didn't see a lot of advantage with Pixel Shift.  The way you guys explained it one would hardly ever use it.  It does reduce artifacts, stair stepping in diagonals in an image if you really pixel peep.  The feature is there if photographers want to use it.  Many photographers aren't going to be aware of much what you discuss regarding f/stops and diffraction. 

Pixel shifting aside the Sony a7r III for $3100.00 delivers a beautiful file.  I have now shot a lot with the camera.  Although we were told we were shooting with production cameras the firmware on mine was .91.  I ordered the camera and when it is received I'll share a lot more images in another article.  The camera is a big step in features even though it has the same chip as the a7r II.  I shot 1500 plus images and was still above half on battery power.  The image Stabilization really works great.  I was shooting handheld at 1/8th of a sec. with good results. Dynamic range is also pretty impressive. 

The thing with Sony is you know it will only get better from here.

Thanks to both of you for good posts here.
Logged
Erik Kaffehr
 

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: New Article - Sony a7r III Pixel Shift
« Reply #21 on: November 24, 2017, 11:34:44 am »

In real-world picture taking using optimal f/stops is not always possible.  I'll never argue with science but when needed you use the f/stop that delivers the DOF needed to get the shot.  And, I made it very clear in the article that the result of the Pixel Shift is a perceived improvement in image quality.  The test I did was typical of conditions many photographers would use Pixel Shift.  While I love the Physics and Science of photography I am a photographer and thus sometimes I have to use not so optimal f/stops like f/22 god forbid.  As many know I am a Capture One user and there are tools in C1 that help minimize many of the issues of diffraction and other lens aberrations.  Bottom line is that in the end, most people will never know the difference between an image shot at f/5.6 or f/11 or even f/22.  Unless you have something to compare to and in most cases, the results will be hard to see.  For me, it is getting the best image possible using the tools I have.  And, yes many times when I need to leave the sweet spot of my lens I say to myself - ouch, this is going to hurt but at least I got a photograph and usually, I can make the best of it.

Your point that one must compromise between resolution and depth of field by shooting at less than the optimum aperture for resolution is well taken. George Douvos (author of some excellent DoF and stacking aps for iOS) has written a very helpful article on depth of field, diffraction and high resolution sensors. The article was about the Nikon D800, but also applies to the Sony a7r III and Nikon D850. To resolve to the limit of the sensor one must keep the allowable blur circle to 10 μ which allows an aperture of no smaller than f/5.

I don't know the shooting distance and magnification of your shots of the engine compartment of an automobile which were taken with a 50mm lens at f/8 or f/11, but for the purpose of discussion I will assume that the front and back distances are 4 and 6 feet respectively. George has an ap, Optimum CS-Pro, that calculates the optimum focusing distance and aperture for given near and far depth of field. For depth of field extending from 4 to 6 feet with a 50 mm lens, the ap shows the optimum f/stop is f/16 and one should focus at 4' 10" as shown in this screen capture. Blur is about 32 μ, much larger than the 10 μ necessary to resolve to the limit of the sensor.



One can use Georges DoF calculator, True DoF-Pro, to calculate the depth of field for a blur of 10 μ as shown. The depth of field for a focusing distance of 4' 10" at f/8 is only 2.4 inches. Obviously, one must stop down.



The only way to achieve more optimal resolution is with focus stacking. George also has a focus stacking ap, Focus Stacker. Results for the the assumed parameters are shown. The ap suggests taking 5 shots at f/7.1. This is larger than the diffraction limited aperture of f/5.



One can use Jonathan's DoF ap for the Android or Windows to calculate stacking parameters for a smaller blur circle as shown below. I increased the number of shots to 12 in order to use an aperture of f/5 and decrease blur to about 10 microns.



Lloyd Chambers, Diglloyd, has written that automated focus stacking as implemented with the Nikon D850 is more applicable and useful than pixel shift as implemented with the Sony a7r III for Landscape work, even though both can be valuable. Nikon calls this automation focus shift, a poorly chosen name which can cause confusion with the shift in focus one can observe when stopping down with lenses having excessive spherical aberration. Using this automation, one can take stacks with minimal effort and time. Perhaps Kevin should review the D850.

Regards,

Bill
Logged

Kevin Raber

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1339
  • Kevin Raber
    • Kevin Raber
Re: New Article - Sony a7r III Pixel Shift
« Reply #22 on: November 24, 2017, 12:19:05 pm »

Bill...Thanks....It all gets rather convoluted when in the field.  The issue with focus stacking which I did an article on a while back is the same with Pixel Shift.  You can't have movement and doing focus stacking outdoors has been hit and miss for me due to the environmental challenges. (being subject movement - water, leaves, grass etc).  Even with some of the auto-alignment features it sometimes fails.  Phase One XF cameras have a great stacking feature built in.  I hope Sony does this someday.  I do manual focus stacking when I can.  If I tether into C1 I can use the focus mask to make sure I have the right overlap.  Otherwise, I overshoot and work it out later.

Once again though, even knowing all the math etc., I doubt if I made an image at F/16 and then focus stacked of say a landscape and then printed them that most people could see a difference.  I have done this and you really need to look hard. 

When I am in the field I don't like to think scientifically or mathematically as it tears me away from the vision and creative part of photography.  Knowing you have good glass and a capable camera allows one to get good images without overthinking. Plus if you know the capabilities of your RAW software you find that 95% of the time you are good to go.

I do want to download the app though and look at it.  It does look pretty cool.


Logged
Kevin Raber
kwr@rabereyes.com
kevin@photopxl.com
rockhopperworkshops.com
photopxl.com

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: New Article - Sony a7r III Pixel Shift
« Reply #23 on: November 24, 2017, 01:35:31 pm »

Hi,

I would agree that focus stacking often does not work. Sometimes I take several images with different focus and blend them manually.

In many cases, using Scheimpflug can be a good solution. But, Scheimpflug only achieves sharpness i single plane of focus.



When I started shooting Pentax 67, on film, I quickly found that relying of focus scales did not yield the image sharpness that was the reason I bought into MF for. So, I modified my shooting to concentrate on the important detail and not care to much about the rest.

Human vision is more sensitive to low frequency detail than to high frequency detail. Stopping down reduces sharpness and looses fine detail. But, the loss of sharpness can be compensated by sharpening. Diffraction can be estimated reasonably well with a a Gaussian, that is not the case with out of focus images, so diffraction responds well to deconvolution sharpening.

Once MTF goes below say 10% (or so) for certain size of image structure, the structure cannot be restored. So we can restore "sharpness" but not fine detail.

My take is that we need to be aware of those things. In the real world, everything is a compromise. Stopping down gives up on some fine detail.

The P45+ back I have has relative large pixels, 6.8 microns. At optimal aperture it has heavy aliasing, like shown below:


Stopping down to f/16 gets rid of almost all aliasing, which essentially means that MTF at 77 lp/mm is near zero. A generous amount of sharpening still yields a decent quality image:


Going to a 4.5 micron device, the aperture corresponding to f/16 would be 16*4.5/6.8 -> 10.6.

Best regards
Erik

Bill...Thanks....It all gets rather convoluted when in the field.  The issue with focus stacking which I did an article on a while back is the same with Pixel Shift.  You can't have movement and doing focus stacking outdoors has been hit and miss for me due to the environmental challenges. (being subject movement - water, leaves, grass etc).  Even with some of the auto-alignment features it sometimes fails.  Phase One XF cameras have a great stacking feature built in.  I hope Sony does this someday.  I do manual focus stacking when I can.  If I tether into C1 I can use the focus mask to make sure I have the right overlap.  Otherwise, I overshoot and work it out later.

Once again though, even knowing all the math etc., I doubt if I made an image at F/16 and then focus stacked of say a landscape and then printed them that most people could see a difference.  I have done this and you really need to look hard. 

When I am in the field I don't like to think scientifically or mathematically as it tears me away from the vision and creative part of photography.  Knowing you have good glass and a capable camera allows one to get good images without overthinking. Plus if you know the capabilities of your RAW software you find that 95% of the time you are good to go.

I do want to download the app though and look at it.  It does look pretty cool.
Logged
Erik Kaffehr
 

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
A schoolbook example of diffraction...
« Reply #24 on: November 25, 2017, 12:55:15 am »

Hi,

Jim Kasson does some very intelligent testing. The way he tests, he uses a Stackshot for making a large set of exposures passing trough maximum focus and measures MTF for the RGB channels. That gives a lot of interesting information about:

  • Longitudional chromatic aberration
  • Focus shift when stopping down
  • Effects of diffraction

Sony has a seemingly very good 90/2.8 macro lens, let's see how capable it is at optimum aperture:



We can see that the lens is very sharp, it reaches around 1800 lp/PH (line pairs per picture height). There is some LoCA, quite small as we will see.

Repeating the experiment at different apertures we can see the effect of focus shift and also the effect of diffraction.


What we see here is that focus shift is virtually none. Regarding diffraction we can see that resolution went from around 1700 lp/mm to around 1000 lp/mm. If we assume that 1700 lp/mm corresponds to 42 MP, at f/11 we would have something like:
42*(1000/1700)^2 ->14.5 MP

Now, lp/PH is normally measured at 50% MTF. Much of what is lost can be regained with sharpening, but sharpening can create artificial detail.

Now, let's look at a famous lens from long ago, the Leica 100/2.8 APO Macro


Peak performance is at f/5.6 and quiet a bit below the Sony lens. LoCA is wider spread, indicating that the Leica lens is not fully corrected for axial chroma for the RGB wavelengths of the Sony sensor. Something like 1400 lp/PH is measured.


Stopping down the Leica lens to f/11 yields around 1000 lp/PH, corresponding to f/11 on the Sony lens. But, the Leica lens has some significant focus shift. Focusing at full aperture would, it would achieve 700 lp/PH at f/11.


The last part of Jim's test shows that all lenses pretty much converge at f/8 and beyond.



In this measurement, the Sony is the sharpest lens, so it has most to loose.

Now, let's look at a real 'champ', the Fuji GFX 120/4 macro at f/5.6:


This is an impressive piece of a lens. LoCA is nil, and it reaches 2800 lp/PH. It i a bit helped by the aspect ratio of the GFX, but truly impressive, anyway.


Jim also tested the Voigtlander APO Lanthar 65/2 APO. An impressive lens that I will probably buy:


It reaches maximum performance at f/2.8, at around 2300 lp/PH on the Sony A7rII.



The focus shift/diffraction plot also indicates very little focus shift. It seems to quiet perform a bit better than the other lenses at f/11, that may be due to physical aperture being a bit larger. Macro lenses have floating elements and vary both focal length and aperture with focusing.

I would say that Jim's posting illustrate a lot of interesting phenomena. It may be that we don't see much of these in real world images. Why? Some explanations:
  • Medium apertures are often used
  • Focusing may be not accurate enough
  • Observation may not be sensitive enough

Also needs to be said that Jim's data is on axis. Most lenses perform weaker of axis, but off axis measurements are much more difficult.

Jim has developed a protocol for qualitative off axis measurements, involving large distance and large targets.


Links:

http://blog.kasson.com/the-last-word/towards-a-macro-mtf-test-protocol/
http://blog.kasson.com/the-last-word/focus-shift-and-loca-in-the-leica-r-1002-8-apo-macro/
http://blog.kasson.com/the-last-word/focus-shift-and-loca-in-the-zeiss-1002-makro-planar/
http://blog.kasson.com/the-last-word/focus-shift-and-loca-in-the-sony-902-8-fe-macro/
http://blog.kasson.com/the-last-word/focus-shift-and-loca-in-the-nikon-1052-8-macro/
http://blog.kasson.com/the-last-word/macro-on-axis-sharpness-and-loca-summary/

Best regards
Erik
« Last Edit: November 25, 2017, 01:58:36 am by ErikKaffehr »
Logged
Erik Kaffehr
 
Pages: 1 [2]   Go Up