What is focus shift?

Focus shift is an issue mostly known to rangefinder camera users. It is a results of an optical aberration called spherical aberration. It will be responsible for a good number of out of focus shots. Let’s dive in.

Focus shift is inherent in every optical design and is impossible to fully eliminate. It can be reduced to insignificant levels though. Some optical designs are famously plagued by it, others much less. One of the worst offenders is the Zeiss C Sonnar T* 50mm F1.5 ZM, a love it or hate it fast prime lens famously difficult to use because of its wild focus shift. That lens needs a very careful and experienced user to get the most out of it on a digital Leica M. The other 50mm prime lens from the same manufacturer, the Zeiss Planar T* 50mm F2 ZM, is famously devoid of focus shift in real world use. That is a Double Gauss design. Then again, you have the Double Gauss design Leica Summilux 35mm 1.4 first version and the lens inspired by it, the Voigtlander Nokton Classic 35mm 1.4 first version. That is focus shifting at its worst.

But what is it?

Focus shift does what is says on the tin: it moves the focus plane backwards when you close the diaphragm to a smaller stop. Almost all lenses are spot-on wide open (the C Sonnar is an exception, but we will talk about that later); when you stop down the focus precision can suffer. The faster the lens, the more prone to focus shift.

Rangefinder users are more aware of it, DSLR users rarely are, mirrorless users aren’t. There are good reasons for this: those of us that use fast lenses and use the full range of aperture stops available will see it, rangefinder or DSLR user alike. Mirrorless users will see it only when the camera focuses wide open, which in some cases doesn’t happen.

The main reason is that rangefinder lenses can be a lot older, and often have extreme designs to keep the size down. DSLR lenses tend to have less extreme designs because they don’t have the same size constraints. Mirrorless lenses are seldom small and the newest designs out there.

We will clarify these points more when we have understood what causes focus shift.

What causes it then?

Spherical aberration is the short answer.

But what is spherical aberration?

It is the inability of the lens to focus all rays of light of a point source in the same spot.

I will be using a diagram found on Wikipedia throughout this article, on which I drew myself to better explain this. Here are the credits: Made by Mglg, uploaded to English Wikipedia, Public Domain, https://commons.wikimedia.org/w/index.php?curid=3655426

This would be a perfect lens: all light rays are focused perfectly at the same distance from the lens, forming the perfect focus plane. Obviously, no perfect lens can exist. The refraction of glass, its curvature and properties will always bend light differently from the center to the periphery of the lens. Like this:

As you can see, the light rays passing through the periphery of the lens focus a lot closer than the ones from the centre. What does this mean for the final image? Ideally, we would like all light rays to focus exactly on the image sensor, like with the ideal lens. In reality, things work like this:

At each stop of the aperture the focus plane will find itself in a different place.

This was much less of a problem in the film days: photographic film has a thickness of approximately 0.14mm. While this might seem very thin, it actually allows for quite some flexibility for the focus plane position. The diagrams show dramatic differences in focus plane shifting: in reality these are really minimal. But they can have a big impact on the final image.

Those 0.14mm of thickness meant that the focus plane could move by a few hundredths of a millimetre in front or beyond the ideal position and still be recorded by the silver halide crystals in the film. That is not the case with digital photography: the imaging sensor is positioned with high precision and has no tolerance at all for the focus plane position. Thus even a few hundredths of a millimetre do make a difference.

Lenses that never were problematic with analog photography became noticeably inaccurate at mid-range apertures. They are far from unusable though: they are focusing precisely wide open and usually work well stopped down to F8.

But why only wide open and F8?

Let’s first consider that focus shift is worse on faster lenses. Why, you ask? Back to the diagram:

Look at the angle of incidence of the peripheral rays compared to the paraxial (close to the axis) ones and where the respective focus point falls. The bigger the lens element, the steeper the angle of incidence of the peripheral rays, thus the closer the focus plane. Faster lenses have bigger lens elements.

When the diaphragm is wide open, the whole lens element will contribute to the image. When you stop the aperture down, you gradually cut off the more peripheral rays and only use the less peripheral and paraxial ones. Towards the smallest aperture values only the paraxial rays will reach the sensor.

Faster lenses will also have a very small depth of field (DOF) when wide open and close to maximum aperture. This means that the step backwards of the plane of focus when you stop down from widest aperture can shift the whole DOF away from the intended focus point:

Let’s assume we are shooting a portrait with a 50mm 1.4 prime lens and we are focusing on the model’s pupil. Wide open the lens will be spot on. If you look at the diagram the DOF at F1.4 will cover the eyelashes and the whole eye socket. We decide that we want more in focus and stop down to F2: now the pupil is slightly blurry because at the closer edge of the DOF, but the cheek and ear will be nice and sharp. At F4 we have a beautifully sharp and detailed ear and a very blurry eye. The DOF is completely behind the eye. But if we stop down to F8 the eye will be in focus again. Wait, what? Yep, you read that right. That is because at F8 the DOF will be so deep that it will reach back to include again the eye and will be getting a lot of the background detail as well (maybe not quite all the way to infinity but that was just to give the idea).

How do you solve the problem?

The simplest way? Wide open or F8.

Slower lenses are another simple solution: they will show the focus shift a lot less because they always have more DOF, even from wide open. The intended plane of focus will always fall within the DOF.

More modern designs reduce the focus shift dramatically with better calculations, or in the case of the Voigtlander Nokton Classic 35mm II a different type of glass in a lens element seems to have done the trick. Have a look at the comparison between the first and second version: the difference is phenomenal! One lens that really impressed me for the lack of focus shift is the TTArtisan 50mm 0.95: it doesn’t get faster than that as lenses go and it’s an extreme design, but focus shift is negligible.

Getting even more advanced technologically floating lens elements (FLE in Leica jargon) have been introduced in the designs. What are they? They are lens groups in the optical cell that move independently from the focusing group. They will optimise the image quality at all focus distances, especially close focus, and minimise focus shift. The current Leica Summilux 50mm 1.4 and 35mm 1.4 are renown for this design solution, and the 35mm is even called the FLE version by the user community to distinguish it from the previous one. This kind of design is also present in many modern lens designs for other systems.

The Zeiss C Sonnar T* 50mm 1.5 ZM case

This lens has a cult status in the Leica M mount user community. It is a reissue of a design dating back to 1932, with better lens coatings and a small design variation in the optical cell (two lens elements separated instead of cemented). Apparently its rendering and bokeh are something to die for according to its fans. From what I can see from the many images available on the web it’s just an older lens design showing its age. But that’s a personal opinion. I would be glad to review it if I could get hold of a copy. I won’t be buying it though, it’s vastly overpriced in my opinion - every lens I review here it’s bought with my own money (correction: I did buy it for the big 50mm comparison and here is its review as well!).

Image rendering aside, it focus shifts like a champ.

What did Zeiss do to tame this issue? When it was released back in 2004 apparently the lens was calibrated to have perfect focus at F2.8. Which meant that at F1.5 you had very slim chances of getting an eye in sharp focus unless you leant forward just the right amount after acquiring focus. Then the user forum lore tells tales of the calibration changing to F2, so you have to lean forward a bit less wide open and backwards a bit more at F4. Now Zeiss calibrates for you at either F2 or F1.5 according to your preference. Obviously the owners of the lens don’t see this as an issue, they are proud of having acquired the skill of rocking backwards or forwards just the right amount according to the aperture chosen, looking like they might be on the spectrum while shooting. It just makes no sense to me.

If the lens focus shifts that bad it’s not a reliable tool. And the fabled rendering might just be a crutch for helping mediocre pictures with a perceived “classic” look.

There are much, much better lenses out there for less money. The Voigtlander Nokton 50mm 1.5 VM is one of them: same speed, better image quality, negligible focus shift. It costs 2/3rds the price of the C Sonnar. But hey, to each his own.

Just as a mention of honour, there is a couple of lenses I know of that have known issues with focus shift outside the rangefinder world: one in the DSLR world, one in the mirrorless world.

The DSLR lens is the Canon EF 50mm 1.2 L USM, a beautiful lens but known to trouble some users with its focus shift. Obviously this comes from the fact that Canon DSLR focus the lens wide open and stop it down at the time of shutter release.

In the mirrorless world I know of the Panasonic Lumix G 25mm 1.7 ASPH (a 50mm equivalent) for the Micro Four Thirds system, which is quite problematic when stopped down a bit. Again, focusing wide open and stopping down at shutter release.

Geeking out a bit more

One question that popped in my mind looking at the diagram was this: how is it that when the lens is wide open, even though the peripheral light rays focus closer to the lens, we can’t see the focused paraxial rays in the image? They will be still focused back there, won’t they? But everything is a big blur where the sharpness coming from the paraxial rays should be.

That took me a while to fully research. It turns out there is very little information about that unless you have a book on optics lying around.

This is where the term Circle of Confusion (CoC) comes forward, and those light rays get…confused!

What is the CoC? It is an optical spot caused by a cone of light rays from a lens not coming to a perfect focus when imaging a point source.

This means that no point of a detail of a scene will ever be rendered as a point on the sensor. The spherical aberration will always spread those into spots. The larger the aperture, or the less corrected the optical design, the larger the spot. This spot is the CoC.

The size of the CoC will determine the amount of detail recorded by the lens on the sensor. The CoC will be constant per aperture on the same lens. If the diameter of the CoC is smaller than a single pixel on the sensor, the detail will be fully recorded. If the CoC is larger than a pixel, the detail will gradually lose definition. This is why a sensor with a higher pixel count, having smaller pixels, will test the resolution of a lens: the CoC will easily exceed the pixel size.

But why do we not see the paraxial focus points after the focus plane?

Because rays intersecting which each other rob of any definition such paraxial focus points. In technical terms they form a Caustic surface of spherical aberration, or Caustic surface with waist. The waist diameter corresponds to the CoC.

Within this caustic surface there is no defined point of focus at all because of the intersection of all the rays coming from the whole surface of the lens element. The only resolution available is defined by the radius of the waist of the caustic surface, also known as the CoC.

Now I understand why outside of the DOF everything is blurry. Will that make me a better photographer? Nope. Will that make my images better? Not at all. Will it make me understand my lenses better in order to get the most out of them? Mmhh, maybe, maybe not.

But it definitely satisfied my curiosity. Understanding something that didn’t make sense before always enriches me. I hope you found it as interesting as I did!

Thanks for reading, especially if you did read the whole thing: kudos to you!

A penny for your thoughts (in the comments obviously)!