Testing Lenses: Finding The Best Average Focus Point

You have been patient, my friends. You have made it through the first article, where we looked at how MTFs improve stopped down and saw it was a bit different for every lens. You enjoyed the second article, where we looked at the Best Individual Focus MTF (BIF MTF) which showed how good the MTF could be at different points from center to edge if you focus at that point.

By the end of this third article, you will be able to use terms like “the BIF-BAF point”, “the VLOR”, and the modestly named Roger’s Point to both amaze and intimidate your online forum of choice for months to come. It’s like an early Christmas.

In more concrete terms, we’re going to work with all that data we get with the MTF versus Field versus Focus test to show you something that may also be useful for certain types of imaging; the times when you want as much of the image in focus as possible. That might be a landscape, a group photograph, architectural shooting, or anytime you just want the best overall sharpness possible, and don’t want to shoot stopped-down very far.

Let’s go back to the MTF v Field v Focus curve I used in the last article, the Zeiss CP.2 50mm Super Speed lens at T5.6.

Olaf Optical Testing, 2017

In that article, I pointed out that the normal MTF, which is measured along the black line across the center of the image didn’t show how sharp the lens could be at each position, if we refocused the machine to the sharpest position for each point. That’s the difference between the standard MTF (left, below) and the Best Individual Focus MTF (BIF MTF; below right.)

Olaf Optical Testing, 2017

When we take the measurements for the BIF MTF, we also get the exact focus position used to get the sharpest MTF at each point. We can plot that out, too.

Olaf Optical Testing, 2017

If you scroll back up to the MTF v Field v Focus graph, you can see the best focus position is similar to the curve of one side of the lens (the red area is the area of maximum sharpness). We can run a few more calculations and determine the focusing distance gives the best mean MTF from one side of the field to the other, which we do in the graph below.

Olaf Optical Testing, 2017


To give a more intuitive picture, I’ll draw a second bar across the Field graph from above in green, at +0.052mm.

Olaf Optical Testing, 2017

As you can see, the image will be a bit softer in the center, but reasonably sharp from side-to-side. This Best Average Focus distance is different for every lens but is easily measurable.

Of course, if you’re thinking ahead, you’re probably saying, “Well, that’s cool sciencey stuff, but how the hell am I supposed to set my camera to focus the equivalent of 0.05mm in front of center focus?” It’s easy; we go back to that ‘best focus’ chart and find the point where ‘Best Individual Focus Distance’ (red curve) crosses the “Best Average Focus Position” (gray line). This, of course, is the soon-to-be-famous BIF-BAF point.

Here’s where things get real. If we drop a vertical line from the BIF-BAF point (we will call the Vertical Line of Roger {VLOR}) it intersects the image height. (The image height is the distance from the center of the image to the edge of the image.)

Olaf Optical Testing, 2017

The VLOR points to a position some distance between the center and the side edge of the image (we can call this, oh, I don’t know, Roger’s Point). If you focus your camera at this point, you will be focusing at the Best Average Focus Position, for edge-to-edge sharpness.

A full-frame camera is 17.5mm from center to side. In the case of the lens above, focusing your camera just over 1/3 of the way to either edge (+/- 7.5 mm from the center) will work. You can do it by choosing an AF point in that area or manually in live view. Obviously, you’ll estimate a bit; you might pick 7mm or 8mm but that will be pretty close.

The distance from center for Best Average Focus will vary for each lens. It may also vary a little bit in some lenses depending on aperture (because field curvature can change slightly with aperture) but that difference is probably too small to matter for this technique.

Does This Work?

Yep. Obviously, it isn’t magic. We’re trading a little bit of center sharpness to get a larger part of the side-to-side image reasonably sharp. Let’s see how it works.

We took a Zeiss CP.2 50mm SS off the shelf and ran its standard MTF, side to side. It’s really sharp for +/- 6mm from the center(about the middle 1/3), then it falls off pretty steeply.

Olaf Optical Testing, 2017

Then we focused 0.05mm further than the best center point focus and ran the MTF again.

Olaf Optical Testing, 2017

You can see now that the best MTF is at about +/- 6 to 8mm from the center, just as predicted. The center isn’t quite as sharp, particularly at the higher frequencies (fine detail) but an MTF of 0.5 at 50 lp/mm is still good. Out at 10 to 14mm away from the center, the image is now usably sharp (high-frequency MTF > 0.3).  With center focus, this area had almost lost the high-frequency MTF. With Best Average Focus, the center 2/3 of our image is now reasonably sharp, where with center focus only about 1/3 of the image was.

If you want to look at it another way, here are the Full Frame MTF Displays at 40 lp/mm, which is a pretty high resolution. Red we consider unacceptably soft, yellow is borderline.

Center Focus MTF

Olaf Optical Testing, 2017

Best Average Focus MTF

The FFD shows it nicely. We’ve sacrificed the center sharpness from excellent to good, but now have the entire image in the acceptable category. This was at T4; things would be a bit better at T5.6.

Olaf Optical Testing, 2017

Again, it’s not a tool you’d want to use for all photography. But when the goal is to get as much side-to-side sharpness as possible, it’s a nice trick to have. It could be useful for landscapes, group photos, architectural shots, and some types of street shooting at any rate.

Another Example

I tried the same test on the Sigma Cine 85mm T1.5. I wasn’t sure that this would make a difference since this lens stays pretty sharp so far from center. I was also interested in this lens because the Roger’s Point was way out at 11mm from center, more than half-way to the edge, which is fairly extreme. Because this lens is so good at T4, I dropped the aperture down to T2.8, so we had a narrower depth of field.

The difference is more subtle, but it’s there. Follow the 0.9, 0.8, and 0.7 MTF lines out, and you’ll lens has a higher MTF at each frequency away from the center. Here is the standard MTF on top, and the Best Average Focus MTF (BAF-MTF) below.

Olaf Optical Testing, 2017

Olaf Optical Testing, 2017


It’s more apparent if I use the Full Frame Display MTF. I’m doing this one at a higher frequency, 50 lp/mm, because the Sigma 85mm T1.5 can handle that, which many lenses can’t. Again, center focus MTF is on top, BAF MTF on the bottom.

Olaf Optical Testing, 2017

Olaf Optical Testing, 2017

BAF MTF Position for the Cine Lenses

I’m not going to (at least right now) show you the MTF curves for every single Cine lens we’ve tested. I’m just going to put up a table of the focus position that will give you the Best Average MTF. One of the questions I have is, do people really want the BAF MTF curves? I think the field curvature graph and BAF Position would be sufficient.

If people really want to see it, we could either just show the BAF Position and the VLOR graph (scroll way back up). And I guess we could show give the BAF MTF curve for each lens. Personally, I think the FFD MTF graphs (the colored circle showing the image field) are best for that purpose.

Until our software upgrades are finished, none of this is automated, though. So, for now, I’m just going to give you the BAF Positions for the Cinema Primes we’ve tested. (They are the same for the photo lens version of each lens.) Since a full frame sensor is 17.5 mm from center to side, it should be fairly easy to estimate where to focus to obtain BAF. You don’t need to count mm, we ran these a bit to either side, and there was no detectable change, so ‘just short of halfway to the edge’ or ‘about 1/4 of the way to the edge’ is accurate enough.


Focal Length

BAF (mm from center)
Canon CN-E14mm6
Canon CN-E24mm7
Canon CN-E35mm5
Canon CN-E50mm9
Canon CN-E85mm5
Sigma Cine24mm8
Sigma Cine35mm9
Sigma Cine
Sigma Cine85mm12
Zeiss CP.2 21mm21mm14
Zeiss CP.2 28mm28mm6
Zeiss CP.2 T2.135mm0
Zeiss CP.2 T2.150mm8
ZEiss CP.2 SS T1.550mm8
Zeiss CP.2 T2.185mm16
Zeiss CP.2 SS T1.585mm16

I’ve made the table sortable so you can look for a pattern, but I don’t see one myself.

So What Does This Mean?

I have no idea. I can see it being a sometimes useful tool. If I did a lot of weddings or group photos, I’d keep the BAF in the back of my mind. It’s kind of counterintuitive to focus on the 3rd lady from center, but it might keep everyone in even focus in the group photo. For landscape photographers it might help when just stopping down isn’t getting everything as sharp as you want. Or perhaps to let you get the shot at f/8 instead of f/16.

Whether that means there’s enough demand for us to publish either the BAF number or the focus curve graph when we do reviews, I don’t know. A lot of times you could probably eyeball it by looking at the MTF v Field v Focus graphs. So I’ll wait for your input on making that decision.

But I think it’s an excellent example of lenses are tools, and the more you know about the tool you’re using, the better you’ll be able to use it. Knowing how far from center the BIF-MTF stays sharp, and where the BAF MTF point is are useful things, no matter how geeky the technique we use to find them is.


Roger Cicala, Aaron Closz, Brandon Dube, Max Bruggeman, and Markus Rothacker

November, 2017

Author: Roger Cicala

I’m Roger and I am the founder of Hailed as one of the optic nerds here, I enjoy shooting collimated light through 30X microscope objectives in my spare time. When I do take real pictures I like using something different: a Medium format, or Pentax K1, or a Sony RX1R.

Posted in Equipment
  • Adrian Bujor

    Impressive and useful info, thank you a very big lot for this series.

    By reading this now i’m thinking that for general photography one might be much better to focus at 1/2 BAF distance from center. Center still very sharp, extends the sharp area less than full BAF, but still a good amount. I’m thinking this can be done (with phase detect) by micro-focus adjustment.
    Do you know if there is any correlation between the adjustment values provided by the manufacturers in camera for adjusting (those +/-N things) and actual focus values (from the MFT vs Field vs Focus graphs)?

  • It would differ for different lenses. The more interesting (to me) part is that depth of field is narrower as frequency increases. So it’s easy to get depth of field at 10 or 20 lp/mm that covers the whole field. But at 50 (fine detail) it’s not happening ever.

    So you can get a nice contrasty look over the whole image, but you can’t simply use depth of field to pick up fine detail on a high-resolution sensor.

  • Someone actually loaned me one of those a while back to test. It was fascinating you could definitely change the shape of the field very easily. I don’t think I kept the files, I sent them to the person who asked me to test it.

  • We are testing focus shift going forward. It occurs everywhere, although some lenses it might be more significant at closer focusing.

  • Poorly worded indeed. I’ll try to get that more specific sometime today.

  • boeck hannes

    unfortunately my nikkor Ai 85mmf2 is missing in your list!

    very valuable information since i actually shoot flat things a lot! thank you so very much!

  • Ilya Zakharevich

    “…determine at what focusing distance the largest area of the image is within that sharp area.”

    Am I alone in being disappointed by vagueness of the part «that sharp area»? (But: a lot of thanks for the rest of the post!)

  • User Colin

    This is interesting however I am wondering how it interacts with the accuracy of focus points. I assume contrast-detect focusing will be accurate, if slow, across the frame. But there must be a reason why the centre point is usually given all the bells and whistles wrt phase-detect focus ability. And in some DSLRs, the grid of PD focus points only fills the centre third. So would it be wise to use one of those outward focus points if it wasn’t actually as accurate?

    Are you planning to look at focus shift? From what I understand, it is more of a problem with closer focusing, and you always seem to test at infinity.

  • Gomez Garcia Gonzalez

    I was just thinking how this testing backed up my high opinion of the Zeiss 21mm, with it’s score of 14 being close to the best!

  • maximum

    Wow, now thats an interesting lens!

  • Ionut Macri

    Looks like 40 years ago Minolta had a odd lens based on that

  • Not THAT Ross Cameron

    Missed opportunity for another of Rogers’ Laws – maybe Roger’s law Of Focusing Lenses – ROFL 🙂
    Re aperture testing, I was going to suggest point at which diffraction starts to kick in (f8 on 24mp FF bodies), but is there a point at which DOF from aperture overrides VLOR & Rogers Point? For landscape, it could still be used to help with determining where in frame to focus in conjunction with hyperfocal distance…
    Tis all good for thought. Luv ya work!

  • If you look back at old, wet-plate, large-format camera images you’ll notice the wise photographers of the day did just that. 🙂

  • Andrew Milne

    OMG that is so awesome!

  • Brandon Dube

    MTF and field curvature are not two different things. “deconvolving” (this is not the correct process) the two does not give you a distinct “BAF” and “focus curvature.” BIF + focal surface sag would separate the two.

    In your wedding photo example, if you arrange the people along a curve, the photo will probably look weird, compositionally. We are accustomed to a flat line.

  • Tom Stermitz

    So MTF and Focus Curvature are two different things, but the ubiquitous MTF lens charts we see are always a mixture of MTF & Focus Curvature? What if you deconvolve the MTF from the focus curvature into a “BAF” measure, and a Focus Curvature measure.

    Internet lens measure-bating does seem to put an inordinate amount of importance on high-resolution lenses with flat fields for taking pictures of newspapers taped to the wall.

    I guess if we know the the lens has forward curving focus, we could ask the wedding party edges to move in a bit so everybody would be in focus.

    In landscape you are screwed every which way (fore-, mid- and back-ground) unless you get a tilt shift lens or focus stack. But it might be useful to know how screwed you are.

  • We can, and that’s what we’ll start showing soon. Also the Tan-Sag difference. So it will look like this:

  • Brandon Dube

    Just T and S are not enough to produce a good measure of “resolving power” across the frame and determine the shape of the best focus region. You can average the two, but it is not a very good way of doing it. If we had, say, 36 azimuths in MTF-space, it would be enough. 2 is insufficient.

  • If I measure the Milvus 35, will you give up on sharpness scores? 🙂

  • Andrew Milne

    If you can combine T and S information in a principled way, can you then combine them in a joint MTF v Field v Focus curve? That, to me, would be interesting, because you could read off both resolving power across the image and also the shape of the best-focus region, which would provide the same kind of information about where to focus.

  • Brandon Dube

    The VLOR is determined from a combination of T and S information, neither one in isolation. We could make it a trust-region like algorithm where it chooses to ignore some field points if they cannot be made to be sharp, but this would substantially raise the computation cost from something that can be done within 5ms via array operations to something that can be done in potentially minute(s) with a smart nonlinear optimizer. Most lenses are not so extremely astigmatic that the difference is substantial, too.

  • Andrew Milne

    I guess I am back to the same questions that I had on the last post. I love the idea here, but I’m not sure it is quite telling us what we might want (if any number ever could…).

    What would happen if you repeated the process for the Zeiss using the tan instead of the sag? Looking at the relationship between the best focus and the field graph, instead of going down on the right the tan graph keeps going up. Which means that when you minimize the area under the graph, the vlor will be pulled to the right.. And if you followed the advice it gives, you would move your focal point further to edge of the frame.

    And that might be the right thing to do, practically, but it does seem (1) you are moving it to improve focus in an area (the edge) where the tan MTF isn’t very good anyway and (2) you are moving it to a point where tan and sag are fighting each other, because the field graph for each is moving in opposite directions.

    I completely support the overall idea, and I am sure this number is very useful and a vast improvement over what we have. But I have a niggling feeling that there is a parameter that is missing that can’t really be left out. It seems to me that the question “where should I focus to maximize my amount of in-focus” is exactly the right question, and that the answer is “not the center” is really important. But if we want to mathematicize (!) it, then we have to answer the question “how good a focus do I demand?” before we can answer the first question. Because otherwise we could end up focusing in a way which gets us the best (compromise) focus at the edges, but where the level of resolving at those edges is not acceptable anyway, and we have lost the best (compromise) middle. In other words, vlor will be pulled to the left, and further to the left the worse (in terms of curvature) the edges are.

    Again, I’m ignorant and don’t know what I am talking about. But I am imagining the following decision tree: I want to resolve to at least X. Well, turns out that my lens will resolve to X only to 12mm across the sensor. Now, where should I focus so as to maximize my X resolution?

  • Phil Service

    Thank you very much for this. I can easily see how to apply the BAF position to images of 2-dimensional subjects. Three-dimensional scenes seem more problematic. Do you move your focusing point to the BAF position, use that to focus on the most important part of the scene, and then recompose? Or, as you say, do you focus on the 3rd lady from center, and not recompose? Do those two methods yield different results? It seems to me it will depend on whether the 3rd lady is in the same plane as the center lady.
    For future tests, I would really like to know the BAF position (wide-open and at, say, f/4 or f/5.6 of course :-), unless aperture has no appreciable effect on BAF). It would also be good to have some idea of the cost in lost center sharpness, and the benefit in increased edge/corner sharpness. These can be seen easily enough in the graphics.

  • Daniel

    Great, please repeat all tests on all lenses (if necessary) to include the new bifbaf data and as a bonus, integrate the mtf, bif and baf curves, weigh it according to the size of the circle around the center (that’s another integral, right?) and subtract the part, that’s cut off the sensor. Tada: the result is your sharpness score (or scores, if you do it for the different mtfs separately). That would be the most scientific and meaningful sharpness score ever. Cmon, admit it, you like scores 😉

    ps: are you going to measure the zeiss milvus 35mm 1.4?

  • Andre, we’ll definitely be including MTFvFvF charts in our tests going forward.

  • If you knew that’s where you always wanted to focus you could do it. Kind of like a lot of people walk around with their lens set at hyperfocal distance (granted there’s tons of argument about where that is).

  • Ed Hassell

    Then, would it be advantageous to incorporate your BAF point in setting AF fine tuning on those camera which offer the function?

  • This is awesome, and I’m glad you guys are exploring MTF in ways that haven’t (publicly) been done. I think an MTF-vs-focus-vs-field chart would be very useful already since we could guesstimate the BAF point from it. It would be useful for finding other ways to optimize the focus if absolute MTF wasn’t the goal, like if you wanted to find a point where astigmatism is minimized, or if you cared more about MTF in one direction than the other.

  • Phil it is at infinity focus and with modern, non-macro lenses it should be accurate until at least right near minimum focusing distances.

  • Phil

    Thanks for this fascinating series of posts! One question… this is measured at infinity focus right? Any idea whether the BAF is likely to be a function of focus distance?

Follow on Feedly