Focus shift

[Harpomatic]

Advertisement (gone after registration)

For all the doubters and the curious out there I wrote a short article on focus shift, what it is and what causes it.

I hope you’ll find it interesting!

https://www.47-degree.com/focus-shift/what-is-focus-shift

I’m really happy to discuss it!

[jdlaing]

Focus shift is anomaly with auto focus lenses. The camera opens the aperture automatically to focus an image and when you press the shutter the aperture closes down automatically and the focus plane is different.

Rangefinder manual lenses have either front or back focus at a given aperture.

I,personally don’t know anyone, but there may well be, that fiddles with the aperture on a rangefinder after they focus. I, personally set aperture and the last thing I do before I press the shutter button is focus the lens.

[adan]

Good technical explanation. I'll bookmark this thread so I don't have to keep explaining it myself.

Although a second article on when spherical aberration can be useful in rangefinder photography (no screens or EVFs) might be fun.

I like using the non-ASPH 1980 Mandler 90 f/2 and 75 f/1.4 because their SA and its "caustic surface" spreads out the area of "almost sharp" over a wider field, which covers minor focus/RF errors wide-open. Here's what I mean - top and bottom lenses, with SA, produce a wider margin for error (bright white stripes in the center of each light-ray map) that will look approximately equally sharp throughout their breadth.

Rather like "focus-stacking" built right into those lenses.

https://en.wikipedia.org/wiki/Spherical_aberration#/media/File:Spherical-aberration-slice.jpg

Personally, I have the 35 Nokton v.II, and/or fall into the "wide-open or f/8" camp, so for me, focus shift is mostly something that happens to other people.

The 35 Nokton v.I was the only lens I ever owned where I actually saw it happening.

Edited September 28, 2020 by adan

[Harpomatic]

4 hours ago, jdlaing said:

Focus shift is anomaly with auto focus lenses. The camera opens the aperture automatically to focus an image and when you press the shutter the aperture closes down automatically and the focus plane is different.

Rangefinder manual lenses have either front or back focus at a given aperture.

I,personally don’t know anyone, but there may well be, that fiddles with the aperture on a rangefinder after they focus. I, personally set aperture and the last thing I do before I press the shutter button is focus the lens.

If you read the article I never claim anyone changes the aperture of the lens after focusing on a rangefinder camera. The focus shift issue is inherent with the system. The lens is normally calibrated for precise focus wide open. No matter at what aperture you shoot, your rangefinder still focuses at a “wide-open” focus distance setting. That’s why focus shift is exactly the same on manual and auto-focus lenses.

[Harpomatic]

5 hours ago, adan said:

Here's what I mean - top and bottom lenses, with SA, produce a wider margin for error (bright white stripes in the center of each light-ray map) that will look approximately equally sharp throughout their breadth.

Would you say that, looking at the transverse section of the same spherical aberration, they form an Airy disc that is a lot more dense in the centre?

5 hours ago, adan said:

The 35 Nokton v.I was the only lens I ever owned where I actually saw it happening.

I really had issues with that lens with mid-range apertures. V.II never gave me any problems.

 

5 hours ago, adan said:

Good technical explanation. I'll bookmark this thread so I don't have to keep explaining it myself.

Thanks!

[RayD28]

Good stuff.  Best explanation of Circle of Confusion, especially regarding the diameter of the COC, I've read.  Thanks for posting.  

[Ecar]

Advertisement (gone after registration)

Clear and concise, thanks!

Although I own both the Sonnar 50/1.5 ZM and the CV 50/1.5 VM (knurled focus ring) and prefer the rendering of the former, focus shift notwithstanding 😉

[Harpomatic]

3 hours ago, RayD28 said:

Good stuff.  Best explanation of Circle of Confusion, especially regarding the diameter of the COC, I've read.  Thanks for posting.  

Thanks! I appreciate it!

 

1 hour ago, Ecar said:

Clear and concise, thanks!

Although I own both the Sonnar 50/1.5 ZM and the CV 50/1.5 VM (knurled focus ring) and prefer the rendering of the former, focus shift notwithstanding 😉

Of course these are opinions! The only reason I haven’t bought the C Sonnar is exactly the focus shift. I think ergonomics and size are my ideal there...and image quality plenty good for my photography. But it always feels good to rant! 😝

[Rick in CO]

Nice thought-stimulating article!  Brings up several questions.

It is oft-repeated that most lenses increase in "performance" two stops down from their maximum aperture.  This can't be a DOF effect?  Still true even for fast lenses calibrated wide-open? The new fast lenses for the SL2 are supposedly optimal wide open, I have not seen this to be factually confirmed by resolution testing, only by calculated MTF.

"Bokeh" is often said to be dependent upon some residual spherical aberration.  Some lenses that can have "nice" Bokeh wide open show a transition to a "ring" effect when stopped down.             Could that be the result of the cut-off of the more peripheral light rays, perhaps reflecting off the aperture blades?

The Zeiss ZM lenses were reportedly designed to have optimal contrast (not the "retro" 50 ZM Sonnar).  My 50 f2 ZM planar was sharp, no focus shift, and very pleasing Bokeh.  The result of very little residual spherical aberration along with the circular aperture?  Perhaps Double Gauss lenses perform well within certain design limitations, once that is exceeded it falls apart.

There is a factor of focus distance on CoC.  This is the Hyperfocal distance for a given focal length?  The combination of an aspherical FLE element that adjusts light ray transmission with focusing distance, but not with aperture, apparently solves the focus shift issue?  Is this true at only close focusing distance?

Your description of the effect of pixel size on CoC nicely reiterates Peter Karbe's recent interview regarding the need for better performing lenses for the higher resolution sensors, to get the maximum performance possible.

Thanks!

[SrMi]

16 hours ago, jdlaing said:

Focus shift is anomaly with auto focus lenses. The camera opens the aperture automatically to focus an image and when you press the shutter the aperture closes down automatically and the focus plane is different.

Rangefinder manual lenses have either front or back focus at a given aperture.

I,personally don’t know anyone, but there may well be, that fiddles with the aperture on a rangefinder after they focus. I, personally set aperture and the last thing I do before I press the shutter button is focus the lens.

Many who use EVF or LCD focus wide open and close the aperture once focus has been achieved.  

[Chaemono]

EVF focus peaking starts to become an issue at f/2.8 and above is my experience.  

[adan]

8 hours ago, Harpomatic said:

Would you say that, looking at the transverse section of the same spherical aberration, they form an Airy disc that is a lot more dense in the centre?

Airy discs I wouldn't know about, specifically. But here is a comparison of the 75 Summilux at f/1.4 (top) and the 75 APO-Summicron-ASPH at f/2 (bottom).

Note that the Summilux's spherical aberration produces more effective depth of field (readable letters), even at f/1.4. Overlaid sharp and fuzzy images in different focus planes, but a lot of them still "sharp enough" to be legible. In terms of the Wikipedia diagram I linked to above, the 75 Summilux is like the top and bottom ray-map sections.

Compared with the 75 ASMA, which at f/2 has one well-defined focus plane, rapidly dropping off to illegible blur. Like the central image in those ray sections.

Interestingly, I have never seen focus shift with the 75 Summilux. Which may explain why Dr. Mandler was so proud of this lens - he worked some kind of magic with the element spacing or glass types or spherical radii such that the "point of best focus" does not shift with aperture, even while revealing that there are many different stacked light-cone-sections (fuzzy over sharp) landing in the image plane on top of each other.

Or maybe used a tradeoff or balancing with the longitudinal chromatic aberration (of which the Summilux has a lot) so that one gets sharp and fuzzy red and cyan images (chromatic bokeh) from the SA, but the "combined white" burns through as readably sharp. The field of "best reasonable focus" is deeper, but acquires rainbow tints. Leitz still tended to "think in B&W" back in 1980, at least for the M lenses.

For proxy-focusing with a rangefinder, the Summilux allows more room for small focus errors. The "soft-sharpness" of the contemporary 90 Summicron-M v.3 is similar, just less extreme.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[UliWer]

First of all, I am glad that Harpomatic's explanation speaks out clearly that focus shift is inherent to the optical system of a lens. It is the result of optical rules and nothing else.

This is important as you'll find so many postings in this forum claiming that their special copy of a certain lens design does not show any focus shift, perhaps because it has a silver chrome lens mount or any other kind of magical device to cure the malady. It is also important to remember that focus shift is the result of optical rules to understand that any posturing like "it doesn't happen to me but only to other people" is just bad metaphysics. Nobody is beyond the rules of physics. One may say, it doesn't bother me, and it may be reasonable to say so, but it happens to everybody using a lens.

(We are lucky that we are just talking about lenses and optics. Unfortunately there are some topics much more important where you find the same sort of talking: it is not something caused by nature's laws and may be cured by magical thinking; it only happens to other people but never to me. For optics this sort of thinking is just false, for other topics it is false and dangerous). 

Back to Harpomatic's explanation.

I am not sure wether his description of the "Circle of Confusion" helps much to understand or to cope with focus shift.  Of course a non-perfect lens system will show a point as a circle. But this is not - only - the result of spherical aberration. We may imagine a lens with (almost) no spherical aberration but with a lot of coma and/or astigmatism: it will not show a point, but a blurred circle. These lenses exist in reality: The 1:1.5/85mm Summarex is a very critical lens for exact sharpness if you use it fully opened: by theory the depth of focus is so thin, that you would notice any focus shift immediately if you looked properly at the results. Though in practice it does not show any relevant focus shift - because it is ever so "soft" at f/1.5. There are so many "faults" you have to accept using this lens fully opened that you don't bother about focus shift. You see some difference between totally unsharp and - relatively - sharp and accept the best as "sharpness" even if it is not sharp but only less blurred. Max Berek described the same for the 1:2.5/5cm Hektor when he wrote, that in theory it had a rather thin depth of focus (compared to the Elmar). Small depth of focus was not popular at the early times of photography. But Berek wrote that it's "natural softness" compensated for this. That's the point where the CoC comes in: with a "bad" lens you accept a larger circle of confusion, since it is not able to produce a smaller one. The diameter of the CoC (0.01mm or 0.03mm) may be used as a psychological definition of what you accept as sharp in relation to unsharp, it is no device to measure focus shift.

Another remark about the 1:1.5/50mm Sonnar. The lens was originally produced for the Zeiss Ikon Contax with f/11 as smallest aperture, some years later there was a change of the optical system and it had f/16 as smallest opening, this later design was kept for the post-war version of the Sonnar. I have both versions and with an adapter I can use them on an M body with viewfinder coupling and live view. When I compared both versions for the first time I was almost shocked: on a longer distance with live view the older version (pre war) was a lot better than the newer post war one: colours had more contrast, resolution in the center as well in the corners were better. Though when I compared at closer distances and used the rangefinder to focus both version, I found out that the older version has enormous, very massive, never seen before focus shift, while the newer post-war version has some, but it is not disturbing. The massive focus shift of the old version must have been obvious even on film. I am rather sure that Ludwig Bertele, the Sonnar's designer, got aware of this problem of his genial first version. Bertele realized that he went too far to optimize contrast and resolution. So he reduced its very high overall performance just a little bit, still having a very good lens at its time, but with much less focus shift than its - "better" - precedessor.  If we look at the modern Zeiss ZM (Cosina) Sonnar, we may realize that it is a simplification of the original design.  It is no weak performer as far as overall contrast and resolution are concerned - but the optical system is stretched too far, same as Berteles first version was stretched too far. Same for the 1:1.4/35mm Summilux asph. in its older version: very high performer - but a lot of focus shift. Leica had newer means to "cure" it than Bertele had in the mid thirties: they used floating elements and perhaps some other modern means. If Zeiss would apply similar means on the original Bertele design of the Sonnar, they could achieve an outstanding lens without focus detectable shift - though it would be horribly expensive, because the original Sonnar design with modern means applied would be much more difficult to produce than a modern Double-Gauss design like the Summilux. 

 

 

 

 

[Harpomatic]

First of all a big thank you everyone for participating in the discussion, I really enjoy this.

On to some replies:

4 hours ago, Rick in CO said:

It is oft-repeated that most lenses increase in "performance" two stops down from their maximum aperture.  This can't be a DOF effect?  Still true even for fast lenses calibrated wide-open? The new fast lenses for the SL2 are supposedly optimal wide open, I have not seen this to be factually confirmed by resolution testing, only by calculated MTF.

I don’t know about the optimum being necessarily two stops down from wide open: I believe the optimum is the smallest Caustic surface waist (or CoC) before diffraction sets in and enlarges the CoC again, whatever the aperture might be. This will be dependent on the pixel pitch as well: on Micro Four Thirds diffraction can set in as early as F8, and so it might be on the Sony A7r mIV with its high pixel density.

The new fast lenses might have great performance wide open but it is impossible for them to be better than stopped down. You can’t cheat physics. Stopping down creates a smaller CoC thus better resolution.

For the other questions I am not sure, I’m a long way from fully understand optics. I’ll get closer bit by bit.

[Harpomatic]

3 hours ago, adan said:

Note that the Summilux's spherical aberration produces more effective depth of field (readable letters), even at f/1.4. Overlaid sharp and fuzzy images in different focus planes, but a lot of them still "sharp enough" to be legible. In terms of the Wikipedia diagram I linked to above, the 75 Summilux is like the top and bottom ray-map sections.

That is so interesting, thanks for the illustration! I will have to understand this a bit better technically!

[Harpomatic]

55 minutes ago, UliWer said:

I am not sure wether his description of the "Circle of Confusion" helps much to understand or to cope with focus shift.  Of course a non-perfect lens system will show a point as a circle. But this is not - only - the result of spherical aberration. We may imagine a lens with (almost) no spherical aberration but with a lot of coma and/or astigmatism: it will not show a point, but a blurred circle.

The point of the CoC was to understand why we can’t see the other focused points from the paraxial rays when wider open.

It’s true that the CoC is not only the result of spherical aberration, you are absolutely right, but I tried to simplify the article talking about the point focused on axis, where astigmatism and coma have no effect. Wanting to get deeper in the theory then we definitely have to talk about other aberrations.

59 minutes ago, UliWer said:

Another remark about the 1:1.5/50mm Sonnar.

Thanks for this, it’s incredibly interesting information I didn’t know at all.

I’d love a C Sonnar corrected for focus shift though...

[Pyrogallol]

I thought that what a circle of Confusion related to was your own eye’s ability to distinguish a point of light in the final print that you are viewing. I guess it would not be the same if you are viewing a picture on a computer screen as that is determined by the little spots of light/pixels created by the screen ?

[BrianS]

From an article that I wrote a while ago- 

The Asymmetric layout and compact design of the Sonnar gives rise to its most difficult to master design flaw: focus shift. The focal length of the center of the Sonnar is longer than the focal length of the edges. This means the best focus point of light entering the Sonnar from the center is behind that of light entering from the edge. Used wide-open, the image is dominated by light coming in at the shorter focal length of the edges. This is responsible for the lower-contrast/spread-out depth-of-field of the Sonnar used wide-open. Stopping down the aperture eliminates contributions from the edge; the image that remains is the product of the longer focal-length center of the lens. It “shifts” towards infinity.

 

https://www.35mmc.com/29/04/2020/zeiss-jena-5cm-sonnars-the-magic-of-the-prewar-uncoated-sonnar-by-brian-sweeney/

 

Edited September 29, 2020 by BrianS

[lct]

Depends upon lens calibration. My (superb) ZM 50/1.5 is calibrated from scratch at f/1.5 for instance, whilst my (gorgeous) M 35/1.4 v2 is calibrated at f/2.8. Learning how to use one's gear does help for sure but an EVF is hard to beat when used properly, i mean when focusing at real aperture but it is another topic. 

[Advertisement]

Hallo,

Look here: Leica M

[BrianS]

I currently have 8 CZJ 5cm F1.5 Sonnars in LTM.

 

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

Eight is Enough by fiftyonepointsix, on Flickr

Enough where I set the shims for some for F1.5, others for F2.8, and a couple to account for focus shift using deep filters for the M Monochrom...

Eight is Enough. Does not explain the 12 Jupiter-3's.