Is it normal to see focus inconsistency frame-to-frame with an f/1.4 lens?

Question

I have the Nikon Sigma 85mm 1.4 and I've noticed some inconsistency with focus. I setup a tripod, and a lens calibration card in an attempt to micro adjust the focus. However, what I noticed was that when the aperture is wide open, frame to frame the focus changes slightly from back focusing to front focusing. Its only slight, but at 1.4 it's very noticeable.

I'm curious if anybody has seen something similar? Would you expect the focus to be exactly the same frame to frame with nothing else moving?

Answer 1

Is it normal to see focus inconsistency frame-to-frame with an f/1.4 lens?

It is normal to see slight AF inconsistency from frame-to-frame with every AF lens ever made. More precisely, it is normal to see at least slight AF inconsistency from frame-to-frame for every cameras/lens combination ever made.

The million dollar question is: How much frame-to-frame inconsistency should I expect? How much is acceptable?

The first, covered more below, can be measured and some generalizations made.

The second all depends upon the intended use of the camera/lens system, the size at which images will be displayed, and the budget one is willing to spend to reduce the amount of variability. Like many things, it costs a lot more to get from, say, 90% to 95% than it does to get from 80% to 90%. Getting from 95% to 98% is even more costly, as is getting from 98% to 99%, or from 99% to 99.5%.

In general, imaging sensor based focusing systems do better in terms of shot-to-shot consistency than systems that use a dedicated sensor for AF. But even with imaging sensor based systems, there will be some variation based on the limits of the system's ability to measure and the limits of the lens' smallest increment that it is capable of moving. You may also have slight movement of the focusing part of the lens that may occur between focusing and taking the picture due to things such as mechanical slack and gravity or other forces due to camera movement.

There are some tradeoffs to using imaging sensor based AF methods. By using much larger "pixels", dedicated PDAF sensors can often be more sensitive in lower light than CDAF or hybrid AF systems based on the camera's imaging sensor. They can often be faster because there is much less information that needs to be processed from a dedicated PDAF sensor than an imaging sensor.

Roger Cicala, the founder and chief lens guru at lensrentals, did a fascinating blog series several years ago. In it he compared several different camera bodies all using the same 24mm f/1.4 lens wide open with the center AF point. He also tested a few other lenses with varying introduction dates. He discovered that different camera bodies had varying amounts of standard deviation with the same lens. He also discovered that the better, newer bodies, which did better with newer lenses, did no better than the older cameras with older lenses. His conclusion was that it took both the newest bodies and the newest lenses that both supported a new "semi-closed loop" AF protocol to get the advantage from either the camera or the lens.

Roger Cicala's Autofocus Reality series is very insightful.

• Autofocus Reality Part 1: Center-Point, Single-Shot Accuracy
• Autofocus Reality Part 2: One vs. Two, Old vs. New
• Autofocus Reality Part 3A: Canon Lenses
• Autofocus Reality Part 3B: Canon Cameras
• Autofocus Reality Part 4: Nikon Full Frame.

And: How Auto Focus (Often) Works
Also: Are zooms always sharper at one end than the other?

There are also a few questions here at Photography SE that you might find informative:

How to get more pin sharp shots with a lens known for inaccurate autofocus?
Should Canon 5D mk II autofocus be accurate enough for a f/1.2 lens?
Do all telephoto zoom lenses backfocus on the wide angle end?
Why isn't my Canon 70D autofocus accurate in manual zone AF mode with a 50mm f/1.8 lens?
Autofocus points in Mirrorless Cameras
Do the issues with sharpness I am seeing require AF fine-tuning?

Answer 2

I would not be surprised at all to see slight variation in very precise measurement of focus from shot to shot. The market for digital cameras is a fiercely competitive one, and making a system with the degree of precision you're looking for would increase cost. Even relatively high-end consumer (and I'm including professional photography in "consumer" here) cameras don't put a lot of emphasis on this, because if they did, they'd have to raise prices or sacrifice features and functionality that matter in more situations.

I recently saw a Facebook post complaining that we can send a probe to Ultima Thule but "can't" make a cell phone that doesn't drop calls. But of course we could make a cell phone and network that never dropped calls — it would just be enormously more expensive than the "usually doesn't drop calls" system we have now.

You will probably see less variance with contrast-detect autofocus (e.g., in live view) than with phase-detect, because phase-detect systems are usually not a closed loop. (See What is being set with Micro Focus Adjustment? on the closed-loop issue, and How does autofocus work? for general background.)

Answer 3

I actually suspect the issue may be related to using lenses not made by the camera manufacturer.

These lenses are always reverse-engineered, and thus, may not perform as well as Nikon lenses would perform in a Nikon body.

Autofocus is exactly where I would suspect issues, because it's the most complex thing that the camera body has to do and it has to be done together with the lens. Especially for open loop phase detect autofocus (PDAF), the camera body may not know the exact characteristics of third-party lenses. Thus, it might guess slightly incorrectly, and therefore, the focus will be slightly off. How much it will be off can vary depending on the original position of the focus motor and its final position.

I have a Canon 85mm f/1.8 and I have not noticed any autofocus issues in it even though I have the second cheapest Canon body. However, I am aware that Sigma 30mm f/1.4 (which I don't have) has poor autofocus in these Canon bodies.

There are three variants of autofocus:

• Phase detect autofocus: fast, when shooting through the viewfinder
• Contrast detect autofocus: slow, when using live view
• Dual pixel autofocous: fast, when using live view, only in some of the Canon bodies

You could try switching the type of autofocus if that would help. Nevertheless, I suspect autofocus, which might work in contrast detect mode, will nevertheless be slower than it would be with a Nikon lens.