Focus Shift, a General Discussion
Get L-Mount lenses at B&H Photo.
I thought I would make a few points about focus shift, since it is poorly understood, even by very experienced photographers, and yet it is an increasingly important issue for optimal results with high resolution digital. See Focusing Zeiss DSLR Lenses For Peak Performance (parts one and two) for more as well as pages mentioning focus shift.
Without understanding focus shift and adjusting appropriately, photographers are guaranteed to have sub-optimal results on a regular basis. It is about the odds, which are affected by and mitigated by a variety of factors. Add in the pixel density of an iMac 5K, and it can “disappear”. when photographers evaluate their work.
Focus shift is a serious challenge at 36 megapixels or more (even at 24MP with some lenses) and can be massively destructive of image sharpness using 187 megapixel Multi-Shot High-Res mode, which has focus tolerances twice as tight as at 47MP.
What focus shift is, what it does, what causes it
Focus shift is a change in distance of the center of the zone of sharp focus associated with stopping down. Focus shift can be rearward (more distant) or forward (closer) or both in the same lens in the same frame, e.g., rearward shift in the center and forward shift in the outer zones, varying by degree and by position in the frame.
The effects of focus shift can be:
- Mild — the gain in depth of field with stopping down exceeds the shift enough to make it subtle;
- Troublesome to visual impact — partially canceling out the gain in depth of field, resulting in subject matter at the wrong distance being favored (e.g., ears instead of nose when focusing on eyes). What can be confusing is that coarse and medium structures improve even while fine detail degrades;
- Sharpness damaging — the shift exceeds gains in depth of field for the first stop or two, so that even three stops down the losses are scarcely regained.
The foregoing applies mainly to the first three apertures, but once the zone shifts, it really applies to all apertures because the entire zone of focus has shifted. Depth masks focus shift strongly by three stops down, which is one reason it is so confusing.
The root cause of focus shift is/are balancing aberrations introduced as part of the optical design to solve other issues—various kinds of spherical aberration typically. These aberrations rapidly disappear with stopping down—half gone one stop down, 3/4 down two stops down, etc. That is, peripheral rays are increasingly cut off with stopping down, leaving only central rays. Because central rays come to dominate and these focus at a slightly different distance than peripheral rays, the center of the zone of sharpest focus can be significantly different wide open than one or two or three stops down. Incidentally, the elimination of peripheral rays is why lenses tend to improve in performance with stopping down; rigorous efforts of optical design and build tolerances are required for high performance at full aperture.
Focus shift can vary a little by lens sample (build tolerances), but in my experience it never changes between good samples in any significant way. And a “bad sample” has other problems that jump out.
Consider a 90mm lens focused 3.00 meters away at f/2 with the depth (thickness) of the zone of critical sharpness of ±0.01 meters on a high resolution sensor (roughly 1 cm before and behind plane of focus). It might be that by f/4, the center of its zone of sharpness shifts to 3.03 meters, with the depth of the zone of critical sharpness of ±0.02 meters. Thus the point of sharpest focus goes out of focus by 1cm and the entire zone of sharp focus moves to an unintended distance, such as the interior of a solid wall versus its surface. While 1cm does not sound like much, on a high-resolution digital sensor a world class lens at f/2 might now perform like a mediocre one at f/4 at the intended focusing distance. See the example below.
For an example of focus shift that makes a lens all but unsuitable for landscape work, see Sharpness Losses from Field Curvature at Distance: Grassy Meadow and the other examples. Or see all the pages with focus shift content.
Observations about focus shift made over a decade of lens evaluations
Evaluating focus shift properly is a non-trivial exercise. It is not always obvious, for various reasons and thus quite confusing, but after a decade, I’ve tested so many lenses that I usually spot it quickly because it is so destructive to lens comparisons in particular, and to lens evaluation in general.
It is rare to hear from anyone who understands focus shift and knows how to test for it. Typically I am admonished with “my lens is tack sharp”, “my test shows no shift”, etc. In such cases, I’ve always found a methodology error, suggested how to evaluate better, and gotten an affirmative response. Why should that be? There are various reasons, some technical, and some psychological. The tolerance for optimal versus pretty-good varies with people.
Here then are some key points I have found to be truisms, some obvious, some less so:
- Anyone shooting wide open won't notice, as there is no focus shift wide open, since the lens is focused at the shooting aperture (or usually is, there are exceptions). Similarly focusing and shooting stopped down eliminates the issue, but may induce other errors if focused too-far stopped-down.
- Few photographers will notice focus shift when shooting at f/8 or f/11, since things will be sharp where focused—shrugging off the subtle oddity of things being sharper at distance (or closer) in the center (or outer zones) than expected—perhaps the thought “weird, I must have focused a little wrong” is likely the most that will occur even from observant photographers.
- Handheld shooting is rife with error potentials (focus error, movement of the photographer or subject) that it just becomes a matter of odds—blind squirrels find acorns and those with one good eye might still get 70% of the acorns.
- It takes only a few razor sharp shots from a high performance lens to form a cognitive commitment that the lens is perfect—all contrary evidence thereafter gets excluded. Often this is also a psychological commitment tied to one’s choice of brand. At least two things result from this: (1) a cognitive blindness develops, making it impossible to consider contrary evidence, and (2) impaired ability to consistently obtain optimal results, since technique is not adjusted appropriately.
- If the camera tends to frontfocus wide open and the lens has rearward focus shift, then the shift cancels it out at least partly. See my notes in Lundy Canyon Portraits and related pages for examples of that.
- Less than optimal sharpness is written off as focusing errors, often followed by Fine Focus Adjust, which cannot address focus shift except for one aperture, but thereafter sets up a cognitive commitment that all is well with good odds of subsequent cognitive blindness, as discussed above.
I greatly prefer lenses with strictly controlled focus shift, meaning that I need take no special steps to work with the lens—I can focus wide open and shoot stopped down with confidence. Lenses with focus shift can offer world-class performance but it becomes challenging to reliably capture that performance.
Zeiss goes to great pains to avoid focus shift in its Zeiss Otus line and very strong efforts in the Zeiss Milvus line. Zeiss understands that in a practical working sense that a lens is only as sharp as it can be accurately focused at the shooting aperture. Zeiss has thus adapted to the high resolution digital age in a practical and important way. Other lens manufacturers (most) seem to have ignored the point entirely—rather shocking since accurate focus is the single most important aspect of image sharpness.
In the image below, the lens was focused once at f/2 with world-class sharpness as shown (crop from 187 megapixel image). I was shooting an aperture series at f/2, f/2.8, f/4, f/5.6. Even at f/5.6, the sharpness is inferior to f/2. The cause is focus shift, as cross-confirmed in several ways. The shift looks to be about two meters at a distance of about 30 meters, which at 187 megapixels (or 47MP) is enough of a shift to seriously degrade the quality, let alone improve upon f/2.
Actual pixels crop from 187 megapixel image
Roy P writes regaring the Leica 90mm f/2 APO-Summicron-M ASPH:
Hi Lloyd, your mosaic comparison at f/2 and f/4 was quite surprising. I would not have expected that, and I could not recall having seen that kind of behavior before. But then I realized I had never done a systematic comparison. As you said in your excellent article, you can’t shoot hand-held and expect to flush out focus shift issues.
So I thought I’d take a few test shots with my 90 APO and see how my lens fared. I was bracing for bad news, but I was very pleasantly surprised. As far as I can tell, everything is as it should be. I don’t see any focus shift at all. See crops from one of my test shots attached.
Center sharpness improves noticeably from f/2 to f/2.8, then further marginally at f/4 and f/5.6. Corner sharpness improves more dramatically from f/2 to f/2.8, and again strongly at f/4, reaching peak sharpness at f/5.6. At f/5.6, sharpness is at a max and uniform across the field. There is very little further improvement from f/5.6 to f/6.3 or f/8, but no degradation either.
I checked all four corners, in case the sensor plane was not parallel to the subject plane. My setup is fine – I see exactly the same behavior at all four corners, and I randomly picked the top right corner to send you.
The only inexplicable thing is the exposure time at f/2. I don’t understand why it was reported as 0.8s and not 0.6 or 0.7s. But apart from that little bit of mystery, everything else looks good! Maybe I lucked out with a very good copy of the lens??!
DIGLLOYD: I wrote back:
Lenses don't change by sample unless there is something way off.
Looking at the center, what I see is an out of focus image to start (or one badly sharpened), and one that gets worse at f/4, but has attributes that are fooling you. Remember that it is fine detail we are talking about and that coarse and medium structures will “clean up” with stopping down.
What I see is that f/4 has greatly enlarged the blur spots vs f/2. The area at left looks sharper but it just has more crisply defined blur spots (less aberrational softness) with no more detail, and the area at right is notably more blurred at f/4 than at f/2.
To which Roy P replied:
I’ll be darned! You’re absolutely right. See the attached f/2 vs. f/4 comparison. The little white dots at f/2 are printing as tiny circles at f/4, like bokeh, for crying out loud!
Well, it is bokeh, anyway. These tiny circles created enough additional bogus structure that at a lower magnification as in my screen capture with four images, that the f/4 shot looked sharper than the f/2. But at actual pixels, it’s clear what’s really going on!
And it’s also easy to see the solid areas in the patterns (easily seen in the red, green and orange areas) are mushy at f/4 compared to f/2. I know the title for your next article: “Can you ******* see?” Sigh.
BTW, no sharpening was applied. Focusing at f/2 is really tough with this pattern. I should probably start using a newspaper as my test subject, instead of this rug hanging on the wall.
Thanks for the clarification.
DIGLLOYD: there are various visual traps and methodology errors in assessing focus shift—it’s hard to get it right. But when blur circles in non-vignetting areas stay the same size or increase in size or change only a little with stopping down, that is absolute proof of focus shift—because stopping down one stop should always cut the blur circle diameter by 1.4X, by definition.
The mosaic shot is really good for showing focus shift without any ambiguity provided that the shift is large enough. The foreground brick pavers also show the transition from unsharpness to sharpness, very useful for confirmation. The problem with most focus shift evaluations is that many lenses shift, say, 1/2 or 2/3 or 3/4 or even a full stop (in DoF terms) at one stop down. In such cases, a 2/3 stop shift still means a 1/3 stop gain in DoF—but the entire zone has moved rearward. So a planar target like the mosaic can be really good for troublesome cases (as above), or it can mislead because the sharpness (just) holds. Typically though, some area of the frame degrades enough, which is grounds for further investigation.
In magnified Live View it is possible with some lenses on some subject matter for a slightly defocused image to look better than ideal focus, typically with lenses not well corrected for secondary color. Stopping down can eliminate the secondary color, which can sometimes look very similar to focus shift, which it is in a way (differential focus by color). But most modern lenses now control violet fringing and secondary color really well.
Newspapers don't work very well for focus shift tests unless at just the right size, because the eye responds to acutance, which typically improves with stopping down, and there is not good fine detail in printed newspaper unless the paper itself is resolved. It can work, it all depends on size and texture.
