Testing Lenses: Best Individual Focus MTF Curves
OK, in the last post we did something useful, but rather boring: we looked at how MTF changes when lenses are stopped down. Today we’re going to use a more powerful optical bench tool, the MTF vs. Field vs. Focus. Unless you work in a metrology lab (and probably not even then, very few are performing this test) you aren’t familiar with what this is, so I’m going to take a minute to show you.
Standard MTF
When we (or anyone else) does a standard MTF it works like this:
- The machine very accurately focuses on the center point of the lens.
- It then measures the MTF from the center to one edge at that best center focus.
When that is done, you get the MTF graph you are used to seeing.
The standard MTF says “if you focus at the center of the lens, the graph shows you how sharp the image will be from one edge to the other.” The graph above shows the image for the standard MTF for the Zeiss CP.2 35mm T2.1 and shows us it stays pretty sharp from one edge to the other. A lot of lenses don’t stay sharp from one side to the other, though.
MTF vs. Field vs. Focus
We can do things a different way, though, and measure the MTF vs Field vs Focus. We program the MTF bench to do this:
- The machine very accurately focuses on the center point of the lens.
- The machine then backs the focus away 300 microns from that focus.
- The machine then measures the MTF, moves forward a few microns, and measures it again. Basically, at every point it measures the MTF 20 times, changing focus a bit each time.
- The machine repeats this step at 40 points from one side of the lens to the other.
So what we end up with is a huge bunch of data. At each of 21 points across the lens, there are 800 measurements saved compared to 20 measurements per point (sagittal and tangential, each at five frequencies) on a standard MTF. That’s a lot of data; a bit over 16,800 data points per tested lens.
We usually graph it something like this.
That’s good information. It shows you how the field of focus curves if you’re focused on the center point. For the lens above, it also shows you that away from the center, the image will be sharpest at a different focusing distance than it is at the center. The “0” in the middle of the Y axis is the best center focus, the top and bottom of the charts are +/- 0.3mm of focusing distance.
In other words, if you focus at the center, your image will be sharper closer or further away as you look from one side to the other; what we call field curvature.
But back to MTF. Let’s take a close up of the sagittal field on the right. I’m also going to draw a black line across the graph to show the focusing distance chosen as “best focus distance for the center of the image.” That black line shows you where the focus was for the standard MTF, and the MTF values along that line are what the standard MTF showed.
The graph below is that standard MTF, and it shows the lens is quite soft 12 to 16mm away from the center (the outer 1/3 of the image). But if you look at the field curvature above, you know that softness is because the best center focus isn’t the best focus for most other points across the field. The lens could be a lot sharper from 10m to 16mm if we had focused there, instead of at the center.
Why is this practical? When I make an image with the subject in the outer 1/3 of the image, well, I’ll focus in the outer 1/3 of the image. When a lens has strong field curvature, as this one does, the standard MTF curve doesn’t tell me a thing about how sharp the image might be if I focus there instead of the center.
We’ve been busy writing programs that let us get more information from the MTF v Field v Focus test. The one I’m going to show you today is ‘individually focused MTF,’ which will answer the question ‘how sharp is it when I focus there?’ In other words, if I focused the lens at the point 10mm off center instead of at the center, how good would the MTF at 10mm be?
Best Individual Focus MTF (BIF-MTF)
We could search through those 16,000 data points in the MTF v Field v Focus graph from center to edge, plotting ‘what is the best possible MTF at each individual point if we focused there.’ If I plot this ‘best individual focus MTF,’ it looks entirely different than the standard MTF above.
To make it simpler, I’ll put thumbnails side by side: standard MTF on the left, ‘best individual focus MTF’ on the right. (Please ignore the different layout of the graphs, I’m working with two different versions of our software.)
They look like two different lenses, don’t they? In reality, depending on how you use them, they are. If you focus at the center point, you get the standard MTF graph on the left. The standard MTF suggests that the lens is really sharp for about 5mm to either side of center (the middle 1/4 of the image) then falls off.
If you focus at a specific point away from center, though, the ‘best individual focus MTF’ graph shows you can get a sharp subject all the way out to nearly 16mm from center, which is all but the very edge of the image.
Remember, though; the graph isn’t reality. If you did focus at, say 16mm away from the center, the rest of the image wouldn’t be as sharp as the graph says. The center would be quite soft if you focused at the point 16mm away from the center, just like the 16mm point is soft when you focused at the center.
The best individual focus MTF graph tells you is how sharp the image could be at each single point; it says nothing about how the rest of the lens would look if you focused there. To demonstrate, here’s what the Zeiss 50mm T1.5 MTF looks like if I set the focus to be best at 14mm away from the center.
Still, this is good information to have. If you’re framing a scene with an off-center subject, knowing how far from center you can get a sharp image is useful. And it’s different for each lens; there’s no general rule for how far away from center you can stay sharp. This is an extreme example; most lenses don’t change this much if we compare standard MTF and best individual focus MTF. But most lenses do change to some degree.
As an aside, this is a nice illustration showing that when people argue about whether the Zeiss 50mm Super Speed (as an example) is soft or sharp away from center, well, they’re both right. It depends on how they’re using the lens. A landscape or architectural photographer focusing in the center would say the lens is very soft in the outer half. A portrait photographer would counter that the eyelashes of his subject are very sharp even though they are posed in the outer 1/3 of the image.
I’m going to put up the best focus MTF charts for all of the Cinema primes we tested in the last article. On the left is the MTF taken at T4 in the standard way, on the right the ‘best individual focus MTF’. The graphs are slightly different sizes because the best focus MTFs are from a beta version of our new software.
If enough people think this is something they will find useful, we’ll start publishing them for all of the lenses we test.
85mm Lenses
All of the 85mm lenses show some improvement at best focus and perform well far away from the center. And then there’s the Sigma 85mm. I don’t have the words, other than I been telling you all that lens is amazing.
Canon CN-E 85mm T1.3
The Canon 85 can get ‘near center’ sharp out nearly half-way to the edge. Further than that there is some loss of fine detail (higher frequency lines) but it’s not severe at T4. I really wouldn’t hesitate to use this lens all the way to the edge of the field.
Sigma Cine 85mm T1.5
I am unworthy. There is moderately mind-boggling goodness all the way to the edge.
Rokinon Xeen 85mm T1.5
The Xeen 85mm really didn’t drop its MTF until halfway to the edge when we used center focus. There is some improvement in the outer 1/2 with best individual focus, but performance does fall off a bit beginning 2/3 of the way to the edge.
Zeiss CP.2 85mm Super Speed T1.5
The Zeiss Super Speed does improve to some degree with best individual focus, and there’s no sudden drop off.
Zeiss CP.2 85mm T2.1
The two Zeiss 85mm, if I remember correctly, are identical optically and differ only in the mechanical size of the aperture. It’s not surprising that they behave similarly at T4.
50mm Lenses
As the field gets wider, we expect to see more weakness near the edge of the field, even with best possible focus, since there are often more aberrations.
Canon CN-E 50mm T1.3
Well, it is a very old design and the only T1.3 lens at 50mm. It struggles a bit away from the center at higher frequencies (fine detail) no matter where you focus. But it is better off axis than the standard MTF curve leads us to expect.
Sigma Cine 50mm T1.5
OK, it’s not quite as spectacular as the Sigma 85mm, but it’s absolutely excellent to 2/3 of the way to the edge and pretty good even further away.
Rokinon Xeen 50mm T1.5
Like the 85mm, the Xeen 50mm doesn’t have a lot of field curvature, but there is some. It stays quite sharp till just under the halfway-to-the-edge point, but you lose sharpness and keep astigmatism in the outer half of images, even at best focus. It’s performs pretty well stopped down but doesn’t improve quite as much as the others at best focus away from center.
Zeiss CP.2 50mm Super Speed T1.5
This was the lens we used in the introduction because the difference is so amazing. It’s exceedingly sharp out to 3/4 of the way to the edge if you’re focusing out there. It’s really close to the Sigma in quality until you get very near the edge.
35mm Lenses
We generally think of wider-angle lenses as having more field curvature, so we’d expect they should look entirely different with best individual focus. They do, but a bit less so than I expected. Probably aberrations are causing more issues than field curvature away from the center at this focal length.
Again, things look a better away from the center than the standard MTF seems to indicate. It’s not quite as sharp as the longer focal length lenses, but then, it’s not the longer focal length lenses.
Sigma Cine 35mm T1.5
Like I said for the Canon, although perhaps a little better. Remember, the Sigma Cine lenses are much newer designs than the others we’re testing here.
Rokinon Xeen 35mm T1.5
The Rokinons seemed to improve the least at the longer focal lengths, but may improve the most here at 35mm. That’s a dramatic difference.
Zeiss CP.2 35mm T2.1
While we do get some improvement with the Zeiss 35 T2.1, it’s not dramatic.
Wider Lenses
I was less certain what to expect with the wider lenses; they do have field curvature, but also tend to have more aberrations away from center that don’t improve with specific focus.
Canon CN-E 24mm T1.5
The Canon 24mm shows it’s capable of better performance off-axis than the standard MTF curve suggests. It has near-center sharpness to 8mm away from center, and decent performance to about 12mm before falling off.
Sigma Cine 24mm FF T1.5
The Sigma 24mm standard MTF curve looks a bit like the Zeiss 50mm SS, and I was hoping for dramatic improvement at best focus. It does improve, no question, but in the outer 1/3 there is still significant fall off. It does perform better in the middle 1/3 of the image than the standard MTF curve suggests.
Rokinon Xeen 24mm T1.5
Similar to the 35mm, the Rokinon 24mm’s best individual focus stays good a bit further out than most 24mm lenses.
Zeiss CP.2 28mm T2.1
The Zeiss 28mm is another lens that shows decent performance 3/4 of the way to the edge with best individual focus MTF, much better than the standard MTF suggests.
Zeiss CP.2 21mm T2.9
The legendary 21mm Zeiss does improve a little bit with best individual point focus. The ‘most sharp’ area ends about 6mm away from center with both standard and best focus MTF, though.
So What Should We Do About This?
With the last post, it was obvious that we should add MTF tests at apertures other than wide open. The Best Individual Focus MTF (BIF-MTF) is a different thing. For one, it’s strange. You’ve never seen it before because no one’s ever done it before.
For the major piece of data it presents, ‘How far from the center can you focus and get near-center sharpness at that point’, I could give as a number without showing the graph. For example, it would be 6mm (about 1/3 the way to the edge) for the Zeiss 21mm above; or 15mm (3/4 the way to the edge) for the Zeiss CP.2 50mm Super Speed. The graphs, though, do present a better picture of whether the drop off is gradual, or sudden. So I’m open to discussion as to whether the graphs are worth showing in our reports.
To some degree, you can get that information from the Field Curvature graphs, which we are going to show with all lenses going forward, but the field curvature compresses the data a lot. You can’t tell if a red area in the Field Curvature graph is .80 or .89 and that can be a big difference.
There’s another neat trick we can do with the MTF v Field v Focus data that may be more useful for some types of images, but this post is already plenty long so we’ll save that for Part III.
Addendum: As Ilya pointed out in the comments, the beta software BIF-MTF graphs have a soft gray T number in the upper left corner. Please ignore them, it’s an artifact caused by a combination of not-quite-finished software being used by my very tired brain.
Roger Cicala, Aaron Closz, Brandon Dube, Max Bruggerman, and Markus Rothacker
Lensrentals.com
November, 2017
145 Comments
Thomas Geist ·
Very interesting as I have come across field curvature frequently and it may bite you in the rear if you don’t anticipate it.
I know you are super busy and surely not waiting for us to ask you which lenses to test 😉
But the one lens that really took me by surprise by seeming to have some extreme field curvature where I don’t need it is the Olympus 7-14 mm f/2.8 Pro.
It is extremely wide for sure but for architecture or landscapes, even stopped down to f/5.6 it goes REALLY soft towards the edges.
I would love to see this confirmed. Or is mine just a bad apple?
bokesan ·
If you haven’t already done something like this: you can visualize the field curvature quite easily by photographing a flat subject at an angle. It’s a very far cry from what Roger and Brandon are doing here, but will give you a general idea. I hesitate to link a testament to my own sloppiness in this temple of exactitude, but I recently posted some examples: http://bokesan.blogspot.de/2017/11/a-simple-test-for-field-curvature-loca.html
Thomas Geist ·
Very interesting as I have come across field curvature frequently and it may bite you in the rear if you don't anticipate it.
I know you are super busy and surely not waiting for us to ask you which lenses to test ;)
But the one lens that really took me by surprise by seeming to have some extreme field curvature where I don't need it is the Olympus 7-14 mm f/2.8 Pro.
It is extremely wide for sure but for architecture or landscapes, even stopped down to f/5.6 it goes REALLY soft towards the edges.
I would love to see this confirmed. Or is mine just a bad apple?
bokesan ·
If you haven't already done something like this: you can visualize the field curvature quite easily by photographing a flat subject at an angle. It's a very far cry from what Roger and Brandon are doing here, but will give you a general idea. I hesitate to link a testament to my own sloppiness in this temple of exactitude, but I recently posted some examples: http://bokesan.blogspot.de/...
Puska ·
In astrophotography, choosing the optimal focus point off center is nothing new. The optical requirements in astrophotography are more stringent than terrestrial photography because stars are point sources of light and will show optical defects right to the corners of the frame. Even highly corrected apochromatic refractors will have slightly different focal planes for different wavelengths of light, as well as varying degrees of field curvature. To mitigate these issues, it is common practice to choose a focus star that is somewhere between 1/3 and 1/2 the distance toward the frame edge. Moreover, that star may be rendered ever so slightly out of focus to split the difference between the focal planes of differing wavelengths, thus reducing the amount of chromatic aberration. In other words, you sacrifice the absolute best center focus for consistency throughout the frame.
I’m grateful that you posted this thought provoking article, because for some reason it has never occurred to me to use the same technique for terrestrial photography. Instead, I’ve blindly depended on stopping down to mitigate field curvature and CA. Indeed, it seems the process could easily be automated by determining whether the field curvature of a given lens benefits from front or back focus and applying that setting in the camera menu. That way auto focus can continue on its merry way in the central portion of the frame and yet optimal focus for full frame consistency would be achieved. Of course, if auto focus chooses a point off center, then all bets are off. If I’m over simplifying this, someone please let me know.
Thanks, Roger, for your continued informative (and often funny) articles.
Roger Cicala ·
Puska, I don’t think you’ve oversimplified at all. And I think you’re going to LOVE the third post in this series. And yes, I’m shamelessly plugging the next post, but these first two are basically the path heading to the cozy warm cabin of the third post. 🙂
DrJon ·
I think the best APOs are really good, for example here’s the performance of mine:
http://www.astro-physics.com/products/telescopes/curves/130f6edfcolor.jpg
However note several of the lines are non-visible.
Brandon Dube ·
What on earth is that a plot of?
Vincent Bonnet ·
I guess it plots the optical path of light by frequencies (colors) to show that all the “useful” ones focus pretty much in the same spot.
Brandon Dube ·
But it doesn’t make sense. Roughly where “Z” is everything bends, I’m guessing that’s a lens. But each color has a different angle of incidence and …ray height? No idea what -0.3mm to +0.3mm indicates.
Vincent Bonnet ·
I guess it plots the optical path of light by frequencies (colors) to show that all the "useful" ones focus pretty much in the same spot.
Brandon Dube ·
But it doesn't make sense. Roughly where "Z" is everything bends, I'm guessing that's a lens. But each color has a different angle of incidence and ...ray height? No idea what -0.3mm to +0.3mm indicates.
Zak McKracken ·
A reverse image search finds this:
http://www.company7.com/astrophy/refractors/155sf-f7.html
…without much of an explanation, though. It’s apparently showing the properties of some telescope. It might be that this is some sort of standard measurement in astronomical circles but I wouldn’t know about that.
My own guess: Something about the sharpness or size of the airy disk, as a function of wavelength and either focus setting or distance from the eyepiece.
Not sure what the x axis whould mean exactly, as zero or negative numbers don’t make much sense if it was really the disk size. Maybe it’s rather the displacement of the disk’s center, for an off-axis point source?
Zak McKracken ·
A reverse image search finds this:
http://www.company7.com/ast...
...without much of an explanation, though. It's apparently showing the properties of some telescope. It might be that this is some sort of standard measurement in astronomical circles but I wouldn't know about that.
My own guess: Something about the sharpness or size of the airy disk, as a function of wavelength and either focus setting or distance from the eyepiece.
Not sure what the x axis whould mean exactly, as zero or negative numbers don't make much sense if it was really the disk size. Maybe it's rather the displacement of the disk's center, for an off-axis point source?
SpecialMan ·
What’s this??? A chart that Roger and Brandon don’t already do???!!!!?!?
Shocking…
Puska ·
In astrophotography, choosing the optimal focus point off center is nothing new. The optical requirements in astrophotography are more stringent than terrestrial photography because stars are point sources of light and will show optical defects right to the corners of the frame. Even highly corrected apochromatic refractors will have slightly different focal planes for different wavelengths of light, as well as varying degrees of field curvature. To mitigate these issues, it is common practice to choose a focus star that is somewhere between 1/3 and 1/2 the distance toward the frame edge. Moreover, that star may be rendered ever so slightly out of focus to split the difference between the focal planes of differing wavelengths, thus reducing the amount of chromatic aberration. In other words, you sacrifice the absolute best center focus for consistency throughout the frame.
I'm grateful that you posted this thought provoking article, because for some reason it has never occurred to me to use the same technique for terrestrial photography. Instead, I've blindly depended on stopping down to mitigate field curvature and CA. Indeed, it seems the process could easily be automated by determining whether the field curvature of a given lens benefits from front or back focus and applying that setting in the camera menu. That way auto focus can continue on its merry way in the central portion of the frame and yet optimal focus for full frame consistency would be achieved. Of course, if auto focus chooses a point off center, then all bets are off. If I'm over simplifying this, someone please let me know.
Thanks, Roger, for your continued informative (and often funny) articles.
Roger Cicala ·
Puska, I don't think you've oversimplified at all. And I think you're going to LOVE the third post in this series. And yes, I'm shamelessly plugging the next post, but these first two are basically the path heading to the cozy warm cabin of the third post. :-)
DrJon ·
I think the best APOs are really good, for example here's the performance of mine:
http://www.astro-physics.co...
However note several of the lines are non-visible.
So Roger, ever felt really really ambitious and fancied hooking one of these up to the kit? (They are used for photography.)
SpecialMan ·
What's this??? A chart that Roger and Brandon don't already do???!!!!?!?
Shocking...
l_d_allan ·
So does your article suggest that “focus and recompose” is more flawed than we already suspected?
With my Canon 6d with antiquated AF sub-system but very good center AF point, I have tended to focus with the center AF and then recompose. My bad?
It also perhaps suggest that using native lenses on our Sony a7x cameras has that much more advantage over using adapters … I have more of the AF points to use with native lenses like the FE55 f/1.8 than when adapting a Canon EF lens.
Or am I yet again … “unclear on the comcept”?
Roger Cicala ·
You’re getting it. There’s more to it than, but yes.
If the lens has a high BIF and a lower standard MTF at, say, 14mm off axis, you’ll get a much sharper image focusing there than you would with focus recompose. If standard and BIF are about the same, you’ll be fine with focus recompose.
l_d_allan ·
So does your article suggest that "focus and recompose" is more flawed than we already suspected?
With my Canon 6d with antiquated AF sub-system but very good center AF point, I have tended to focus with the center AF and then recompose. My bad?
It also perhaps suggest that using native lenses on our Sony a7x cameras has that much more advantage over using adapters ... I have more of the AF points to use with native lenses like the FE55 f/1.8 than when adapting a Canon EF lens.
Or am I yet again ... "unclear on the comcept"?
Roger Cicala ·
You're getting it. There's more to it than, but yes.
If the lens has a high BIF and a lower standard MTF at, say, 14mm off axis, you'll get a much sharper image focusing there than you would with focus recompose. If standard and BIF are about the same, you'll be fine with focus recompose.
Andre Yew ·
I would love to see the BIF-MTF chart included with the standard MTF chart for future posts! As someone who loves and uses the Nikon 24/3.5 PC-E for landscapes, BIFing it is kind of necessary.
Brandon Dube ·
“BIF-MTF” is something you can never achieve in a real image. Each field pt has its own unique focus position. Unless you make a final image by mosaicing a bunch focused at different distances, you can’t replicate an image that corresponds to that MTF curve.
Andre Yew ·
Yes, that is true. What I meant is that because of the 24PC-E’s relatively large field curvature, I have to carefully choose which part of the frame I want to focus on. I can’t rely on picking the center point, and having other things in the same plane be in focus.
Andre Yew ·
I would love to see the BIF-MTF chart included with the standard MTF chart for future posts! As someone who loves and uses the Nikon 24/3.5 PC-E for landscapes, BIFing it is kind of necessary. I also adjust the focus so the astigmatism that's noticeable on features in the frame is either less noticeable or more harmonious with the composition.
Brandon Dube ·
"BIF-MTF" is something you can never achieve in a real image. Each field pt has its own unique focus position. Unless you make a final image by mosaicing a bunch focused at different distances, you can't replicate an image that corresponds to that MTF curve.
Andre Yew ·
Yes, that is true. What I meant is that because of the 24PC-E's relatively large field curvature, I have to carefully choose which part of the frame I want to focus on. I can't rely on picking the center point, and having other things in the same plane be in focus.
Matt ·
First off, I disagree with “OK, in the last post we did something useful, but rather boring: we looked at how MTF changes when lenses are stopped down.” I found that article to be quite interesting, but perhaps I’m easily amused.
Second, thanks for this article. I’ve often wondered how worthwhile obsessing about center sharpness is when so many photographs frame the most important elements (at least insofar as needing sharpness) away from center. This affirms that notion and reveals that the Sigma 85 is somehow an even better portrait lens than we previously thought.
As far as what to do, at the start of the article I was expecting it to be interesting but something that once we saw the charts we could make mental note about how this affects lenses and going forward we’d maybe only need to see the charts for some specialty lenses. After seeing them and the variation that occurs, I’m greedy and I want to see it for everything.
Matt ·
First off, I disagree with "OK, in the last post we did something useful, but rather boring: we looked at how MTF changes when lenses are stopped down." I found that article to be quite interesting, but perhaps I'm easily amused.
Second, thanks for this article. I've often wondered how worthwhile obsessing about center sharpness is when so many photographs frame the most important elements (at least insofar as needing sharpness) away from center. This affirms that notion and reveals that the Sigma 85 is somehow an even better portrait lens than we previously thought.
As far as what to do, at the start of the article I was expecting it to be interesting but something that once we saw the charts we could make mental note about how this affects lenses and going forward we'd maybe only need to see the charts for some specialty lenses. After seeing them and the variation that occurs, I'm greedy and I want to see it for everything.
Mike Aubrey ·
That Sigma 85…
Roger Cicala ·
IKR??? Sometimes things are bigger so they can be better.
Arthur Meursault ·
Insanely good. I used to think Sigma reversed engineered and then improved Zeiss designs. Looks like Zeiss needs to do the same with Sigma now.
Mike Aubrey ·
That Sigma 85...
L G ·
Very useful. It takes looking at a lot of real world photos to get the level of understanding of lens performance that these new graphs show. At least regarding sharpness. Thank you taking the time to study and present this.
Philip ·
‘whether the graphs are worth showing in our reports.’
Indeed they are, and if it is easy to implement and you are happy to do so, pls include them. If one is setting up a near-far using a subject you wish to highlight, knowing the lens’s performance cut-off gradient is very valuable (as shown so eloquently in the 50/1.5). There are indications makers are much more aware of full frame image quality (even at the expense of center IQ), whereas for simpler designs in the low res era it was a lower priority. Modern cameras will increasingly enable far off-center focusing and the trend to ‘bokeh shooting’ (esp fast wide angles) both point to the value of this information. I imagine video shooters would also benefit greatly. Finally, it would reduce trial and error. cheers.
Philip ·
'whether the graphs are worth showing in our reports.'
Indeed they are, and if it is easy to implement and you are happy to do so, pls include them. If one is setting up a near-far using a subject you wish to highlight, knowing the lens's performance cut-off gradient is very valuable (as shown so eloquently in the 50/1.5). There are indications makers are much more aware of full frame image quality (even at the expense of center IQ), whereas for simpler designs in the low res era it was a lower priority. Modern cameras will increasingly enable far off-center focusing and the trend to 'bokeh shooting' (esp fast wide angles) both point to the value of this information. I imagine video shooters would also benefit greatly. Finally, it would reduce trial and error. cheers.
Ilya Zakharevich ·
First: your BIF-MTF are marked with something like T/NUMBER at the upper left corner. I think the meaning of this number changed somewhere in the middle of data collection. At least sometimes it is the same as the open T-number, and sometimes it is the same as as-tested T-number.
Moreover, for 50mm T1.5, it is T/2, which is neither of two. Given that the as-tested T-number is not shown in the header either, this makes it pretty confusing…
Brandon Dube ·
“Beta software.” The companies that make our machines are not very good at metadata in their software, which we are forced to pull from.
Roger Cicala ·
Thank you, Ilya. That is my fault: the Beta software I used for the new graphs isn’t ready for public use yet — you may have noticed I had to make graph titles by hand. I neglected to pull off the T# on some, which, as Brandon mentioned, the new software puts a default # if the data is looking for isn’t there. I should have cleared that off when I retitled the graphs but I usually end up doing this stuff late at night and the combination of late night and my brain doesn’t work as well as it did 20 years ago.
Roger
Ilya Zakharevich ·
First: your BIF-MTF are marked with something like T/NUMBER at the upper left corner. I think the meaning of this number changed somewhere in the middle of data collection. At least sometimes it is the same as the open T-number, and sometimes it is the same as as-tested T-number.
Moreover, for 50mm T1.5, it is T/2, which is neither of two. Given that the as-tested T-number is not shown in the header either, this makes it pretty confusing…
Brandon Dube ·
"Beta software." The companies that make our machines are not very good at metadata in their software, which we are forced to pull from.
Roger Cicala ·
Thank you, Ilya. That is my fault: the Beta software I used for the new graphs isn't ready for public use yet -- you may have noticed I had to make graph titles by hand. I neglected to pull off the T# on some, which, as Brandon mentioned, the new software puts a default # if the data is looking for isn't there. I should have cleared that off when I retitled the graphs but I usually end up doing this stuff late at night and the combination of late night and my brain doesn't work as well as it did 20 years ago.
Roger
Ilya Zakharevich ·
Second, your spoilers for the part III led me to spell out how would *I* do the next part! 😉
Start with a nice number which summarizes the MTF curves in a meaningful way: the total MTF=0.5 resolution. (I already described it once here?—?and it is not rocket science.) It is very easy to calculate: interpolate/extrapolate curves to find the tangential and saggital resolution for MTF=0.5 at a particular point. The product of these numbers gives the numbers of the “matching point-spread ellipses” per mm² at this point. Taking the total over the whole sensor area gives “the total MTF=0.5 resolution” (up to a multiplicative constant which does not matter for what follows).
Now: for every displacement of teh focal plane one gets this number. Choose the base displacement not to maximize resolution at center, but the total resolution.
Suggested Part III: Plot MTF curves for this displacement.
Brandon Dube ·
I’ll have to quit before Olaf deals in MTF-50 😉
Ilya Zakharevich ·
Olaf does not need to. I’m pretty sure it is not Olaf what produces the MTF graphs, right? They are just results of post-processing Olaf’s data.
Likewise, this number is also a result of post-processing the transfer functions. And I’m pretty sure that the results (“the focal planes of maximal resolution”) are going to be very similar for MTF=70%, MTF=50%, and MTF=30%…
Brandon Dube ·
Olaf is a company and a machine.
The company will never use MTF-50, or any other metric that inverts MTF, so long as I am involved. They are not correlated to perceptual image quality, are completely unrelated to how an optical designer would evaluate or focus the lens in design software, are not related to the camera’s autofocus algorithms, and are not shown to be an accurate method to reduce an MTF vs freq curve to a single value without loss of detail in the data.
There are (and we have already) implemented better algorithms.
Ilya Zakharevich ·
This “we” in the last sentence is the company behind Olaf? Is it a trade secret, or is it (or is it going to be eventually) published? I do not remember seeing investigations of scalar correlates of useful objective or subjective measures…
And: is this antipathy to “inverse measures” somehow related to putting curves for so many “intermediate” frequencies on your graphs? My (naive) expectations would be that the curves for 10, 40 and 100 lp/mm would contain all the relevant info in these graphs (and some more…).
Brandon Dube ·
Please email technicalsupport@olafoptical.com if you want to discuss more.
Roger Cicala ·
Ilya, we’re heading to the same place, although we’re trying to get there a different way. I’ll be interested in whether our method holds up mathematically when you see it. I don’t quite have Brandon’s complete antipathy to inverse measurements, probably because at my age I just don’t have much antipathy left, period. I try to stay away from MTF50 type measurements mostly because so many people think MTF50 and MTF at 50 lp/mm are the same thing.
Ilya Zakharevich ·
Second, your spoilers for the part III led me to spell out how would *I* do the next part! ;-)
Start with a nice number which summarizes the MTF curves in a meaningful way: the total MTF=0.5 resolution. (I already described it once here — and it is not rocket science.) It is very easy to calculate: interpolate/extrapolate curves to find the tangential and saggital resolution for MTF=0.5 at a particular point. The product of these numbers gives the numbers of the “matching point-spread ellipses” per mm² at this point. Taking the total over the whole sensor area gives “the total MTF=0.5 resolution” (up to a multiplicative constant which does not matter for what follows).
Now: for every displacement of teh focal plane one gets this number. Choose the base displacement not to maximize resolution at center, but the total resolution.
Suggested Part III: Plot MTF curves for this displacement.
Brandon Dube ·
I'll have to quit before Olaf deals in MTF-50 ;)
Ilya Zakharevich ·
Olaf does not need to. I’m pretty sure it is not Olaf what produces the MTF graphs, right? They are just results of post-processing Olaf’s data.
Likewise, this number is also a result of post-processing the transfer functions. And I’m pretty sure that the results (“the focal planes of maximal resolution”) are going to be very similar for MTF=70%, MTF=50%, and MTF=30%…
Brandon Dube ·
Olaf is a company and a machine.
The company will never use MTF-50, or any other metric that inverts MTF, so long as I am involved. They are not correlated to perceptual image quality, are completely unrelated to how an optical designer would evaluate or focus the lens in design software, are not related to the camera's autofocus algorithms, and are not shown to be an accurate method to reduce an MTF vs freq curve to a single value without loss of detail in the data.
There are (and we have already) implemented better algorithms.
Ilya Zakharevich ·
This “we” in the last sentence is the company behind Olaf? Is it a trade secret, or is it (or is it going to be eventually) published? I do not remember seeing investigations of scalar correlates of useful objective or subjective measures…
And: is this antipathy to “inverse measures” somehow related to putting curves for so many “intermediate” frequencies on your graphs? My (naive) expectations would be that the curves for 10, 40 and 100 lp/mm would contain all the relevant info in these graphs (and some more…).
Brandon Dube ·
Please email technicalsupport@olafoptical.com if you want to discuss more.
Roger Cicala ·
Ilya, we're heading to the same place, although we're trying to get there a different way. I'll be interested in whether our method holds up mathematically when you see it. I don't quite have Brandon's complete antipathy to inverse measurements, probably because at my age I just don't have much antipathy left, period. I try to stay away from MTF50 type measurements mostly because so many people think MTF50 and MTF at 50 lp/mm are the same thing.
Mike Earussi ·
Excellent info and very useful, and you’re right, totally different than any one else’s tests. The next question then is how do we adjust the AF to take advantage of this information?
Roger Cicala ·
For me it literally is a list of “mm off axis maintaining sharpness” for each lens. I chose about 90% of center sharpness at 30 lp/mm arbitrarily OR the point where sag-tan separation becomes significant. So I’ll shoot the 28mm Zeiss 10mm off center (half way to the edge), the 50mm Zeiss T1.5 to 3/4 the way to the edge, etc.
Mike Earussi ·
Excellent info and very useful, and you're right, totally different than any one else's tests. The next question then is how do we adjust the AF to take advantage of this information?
Roger Cicala ·
For me it literally is a list of "mm off axis maintaining sharpness" for each lens. I chose about 90% of center sharpness at 30 lp/mm arbitrarily OR the point where sag-tan separation becomes significant. So I'll shoot the 28mm Zeiss 10mm off center (half way to the edge), the 50mm Zeiss T1.5 to 3/4 the way to the edge, etc.
Ilya Zakharevich ·
Anyway: I realized that I did not write down how much I appreciate what you are doing in general, and, especially, these particular “closed-down” data points. A lot of thanks!
Guido ·
«If enough people think this is something they will find useful, we’ll start publishing them for all of the lenses we test.»
Of course it’s useful and we want it 😉
And by the way: If people want to understand lens technology to a certain degree, then the Lensrentals log is by far the best website on the entire internet. There is no other website, that explains topics like this so good. And there is no other website, that tests usually 10 or more pieces of each lens, which means the other lens tests are some kind of a lottery and less reliable.
Søren Stærke ·
In optical simulation software OpticStudio (formerly known as Zemax) it was called MTF through focus.
Here’s a sample lens, with the MTF vs field plot on top as we all know and have seen before, followed by 2 MTF through focus plots. MTF10 and MTF30 are plotted, and though this is non-optimized lens design of my own, it shows the displacement of best focus.
https://uploads.disquscdn.com/images/632e63b1172fbec4e5b05e18f3d033ba6cc5231b30919fd8354e95b45f7ba6da.jpg
Brandon Dube ·
Sure, MTF Thru-Focus is certainly nothing new. Code V and OSLO both do that, too.
Søren Stærke ·
In optical simulation software OpticStudio (formerly known as Zemax) it was called MTF through focus.
Here's a sample lens, with the MTF vs field plot on top as we all know and have seen before, followed by 2 MTF through focus plots. MTF@10lines/mm and MTF@30lines/mm are plotted, and though this is non-optimized lens design of my own, it shows the displacement of best focus.
https://uploads.disquscdn.c...
Samuel H ·
You’re nailing it. I don’t know what’s in store for part 3, but the three pieces of information your are checking and nobody else ever worried about should be enough for most users to get an extremely accurate idea of how sharp the lens will be for their kind of use.
If I could ask for stuff, I’d love to see your MTF reviews including a three-graphs row for each lens.
* standard graph, wide open, with copy-to-copy variation, which is awesome to get a general idea about the lens
* f/4 MTF from a good copy, which tells me stuff I care about when shooting a landscape
* best-individual-focus, wide open, which tells me stuff I care about when shooting portraits
Oh, and move the T/x to the lower left corner 🙂
PS: Sigma couldn’t possibly ask for better marketing than what you’re giving them for free. That graph for the 85mm is going to sell a lot of lenses.
Roger Cicala ·
It sold one to me 🙂 Those guys are generally killing it lately.
We’re on the same page for MTF reviews, but I think MTF wide open, f/2.8 (for the zoom-prime comparisons) and either f/4 or f/5.6 (haven’t decided yet).
Samuel H ·
So, you’re feeling generous!! Nice. ^_______^
(I’d be happy to do my zoom-prime comparisons at f/4)
Roger Cicala ·
More like I get nauseous every time someone says this zoom is just as good as a prime because the zoom at f/2.8 is as good as the prime at f/1.4 (and actually even that is rarely true).
Samuel H ·
As someone who owns 15 prime lenses and not one single zoom*, I completely understand that feeling. But somehow, over the last few years, my mind went from “how you dare, you uneducated monkey” to “use whatever you want, if you insist on a bad choice, well, my stuff will look better in comparison”. Goes for camera brands as well.
* OK, my RX100 IV comes with a zoom, but I can’t change that. Though I wish I could buy two versions of this camera, with f/1.4 primes, one wide-angle and one for portraits.
Roger Cicala ·
I’ve got one of those too, and love it.
Samuel H ·
As someone who owns 15 prime lenses and not one single zoom*, I completely understand that feeling. But somehow, over the last few years, my mind went from "how you dare, you uneducated monkey" to "use whatever you want, if you insist on a bad choice, well, my stuff will look better in comparison". Goes for camera brands as well.
* OK, my RX100 IV comes with a zoom, but I can't change that. Though I wish I could buy two versions of this camera, with f/1.4 primes, one wide-angle and one for portraits.
Roger Cicala ·
I've got one of those too, and love it.
Samuel H ·
You're nailing it. I don't know what's in store for part 3, but the three pieces of information your are checking and nobody else ever worried about should be enough for most users to get an extremely accurate idea of how sharp the lens will be for their kind of use.
If I could ask for stuff, I'd love to see your MTF reviews including a three-graphs row for each lens.
* standard graph, wide open, with copy-to-copy variation, which is awesome to get a general idea about the lens
* f/4 MTF from a good copy, which tells me stuff I care about when shooting a landscape
* best-individual-focus, wide open, which tells me stuff I care about when shooting portraits
Oh, and move the T/x to the lower left corner :)
PS: Sigma couldn't possibly ask for better marketing than what you're giving them for free. That graph for the 85mm is going to sell a lot of lenses.
Roger Cicala ·
It sold one to me :-) Those guys are generally killing it lately.
We're on the same page for MTF reviews, but I think MTF wide open, f/2.8 (for the zoom-prime comparisons) and either f/4 or f/5.6 (haven't decided yet).
Samuel H ·
So, you're feeling generous!! Nice. ^_______^
(I'd be happy to do my zoom-prime comparisons at f/4)
grubernd ·
You are onto something here. Practical MTF numbers for real world photographers and not for brickwallers or siemenssterners. Best Focus MTF + the focus heatmap describes the possibilities a lens has when taking images. It feels like you printed that combination of charts directly from my brain.
Roger Cicala ·
Thank you! The goal of all this stuff is to show that lab tests DO have let us use the tool that is our lens better, and choose the tool better. It’s more complicated than this one is 78 and this other one is 81, but it’s way more useful.
Roger Cicala ·
Thank you! The goal of all this stuff is to show that lab tests DO have let us use the tool that is our lens better, and choose the tool better. It's more complicated than this one is 78 and this other one is 81, but it's way more useful.
Miros?aw St?pi?ski ·
Just a note to my stupid self: use the other focus points instead of using just the central one and recomposing, you fool!
Thank you Roger and the whole team for getting that far and for introducing these interesting concepts that stimulate my nerdy brain. I caught myself on checking this space for new posts a few times a day, it is some kind of addiction. I just can’t wait for part III.
I owe you all, Gentlemen, a bottle or two, or even more. Due to the simple fact of my non-US residence I cannot be your customer and support your research this way, but our national tradition forces me to take some necessary steps.
Roger Cicala ·
Miroslaw, someday when I make my grand trek across Europe, I will take you up on that!
Mireaux ·
Just a note to my stupid self: use the other focus points instead of using just the central one and recomposing, you fool!
Thank you Roger and the whole team for getting that far and for introducing these interesting concepts that stimulate my nerdy brain. I caught myself on checking this space for new posts a few times a day, it is some kind of addiction. I just can't wait for part III.
I owe you all, Gentlemen, a bottle or two, or even more. Due to the simple fact of my non-US residence I cannot be your customer and support your research this way, but our national tradition forces me to take some necessary steps.
Andre Yew ·
This is probably more of a question for Brandon, but I remember from past discussions that your optical bench uses multiple frequencies of light? If so, could the field curvature map be separated out for individual frequencies (eg. R, G, B), and be used to plot a LoCA graph? Sort of like what Jim Kasson does with his motorized rail, but over the entire frame instead of a central patch?
Brandon Dube ·
Could we do it – yes. Do we need to do it through focus and take the measurement time from 5 minutes to 5 hours – no.
Will we do it – no; we don’t have a motorized filter wheel, so we can’t automate (i.e. economize) testing at different color configurations.
Andre Yew ·
I was wondering if it’s possible to extract the individual spectral components from the white light measurements, so you wouldn’t need to filter the light for the measurement.
Brandon Dube ·
Unfortunately, that is not possible.
Brandon Dube ·
Could we do it - yes. Do we need to do it through focus and take the measurement time from 5 minutes to 5 hours - no.
Will we do it - no; we don't have a motorized filter wheel, so we can't automate (i.e. economize) testing at different color configurations.
Andre Yew ·
I was wondering if it's possible to extract the individual spectral components from the white light measurements, so you wouldn't need to filter the light for the measurement.
Brandon Dube ·
Unfortunately, that is not possible.
Andrew Milne ·
I have a question, I think just as clarification, but maybe more than that.
If I understand what has been done, you measure the MTFs at many spots across the frame at many distances. The standard MTF graph picks the distance with the best MTF at the center, then tells you the MTFs off center. Your new BIF-MTF chooses the best MTF for that spot instead.
My question is: which MTF? Looking at the graphs, I’m guessing sagittal?
My follow-up questions revolve around this choice. I’m not smart enough to look at a photo and tell what blurring is sagittal and what tangential. The blurring I detect is some combination of both, I am assuming. At the center, normally we can ignore this, because that is usually where both are best. But on some of these lenses, they are quite different. So look at the Sigma 24, where the best focus tangential drops off much more steeply. Or, to put it another way, if we made a best focus tangential graph, the sagittal would probably drop off much more steeply. Is there a reason to pick one over the other?
So: what is the joint effect of the two MTFs? Also, isn’t that often the biggest issue for many lenses – because the shape of the focus plane is different for both, you may not be able to find a good compromise – you could focus for one, and not the other, and end up fuzzy anyway. The lens works best when the focal planes overlap, right? So a good lens (off center) needs both to resolve sharply across the frame and also at the same distance (for both MTFs) across the frame.
Am I missing something?
Roger Cicala ·
Both sagittal (dotted lines) and tangential (solid lines). You can see the point where you can’t get both the same when the lines separate widely (that can also be lateral color; MTF looks the same whether its astigmatism or lateral color).
Andrew Milne ·
Thanks! Is there a general way of combining the two to get a “overall” mtf, or is it just sum/average? Does that have some correspondence to a physical/visual phenomenon?
Roger Cicala ·
You can (and we sometimes do) average MTF. But it’s kind of a slipper slope. A lens with an MTF of xx at a certain point because sag is xx+10 and tan is xx-10 looks really different than a lens with MTF xx because both sag and tan are xx. Still, it can be useful to average like that.
Andrew Milne ·
I have a question, I think just as clarification, but maybe more than that.
If I understand what has been done, you measure the MTFs at many spots across the frame at many distances. The standard MTF graph picks the distance with the best MTF at the center, then tells you the MTFs off center. Your new BIF-MTF chooses the best MTF for that spot instead.
My question is: which MTF? Looking at the graphs, I'm guessing sagittal?
My follow-up questions revolve around this choice. I'm not smart enough to look at a photo and tell what blurring is sagittal and what tangential. The blurring I detect is some combination of both, I am assuming. At the center, normally we can ignore this, because that is usually where both are best. But on some of these lenses, they are quite different. So look at the Sigma 24, where the best focus tangential drops off much more steeply. Or, to put it another way, if we made a best focus tangential graph, the sagittal would probably drop off much more steeply. Is there a reason to pick one over the other?
So: what is the joint effect of the two MTFs? Also, isn't that often the biggest issue for many lenses - because the shape of the focus plane is different for both, you may not be able to find a good compromise - you could focus for one, and not the other, and end up fuzzy anyway. The lens works best when the focal planes overlap, right? So a good lens (off center) needs both to resolve sharply across the frame and also at the same distance (for both MTFs) across the frame.
Am I missing something?
Roger Cicala ·
Both sagittal (dotted lines) and tangential (solid lines). You can see the point where you can't get both the same when the lines separate widely (that can also be lateral color; MTF looks the same whether its astigmatism or lateral color).
Andrew Milne ·
Thanks! Is there a general way of combining the two to get a "overall" mtf, or is it just sum/average? Does that have some correspondence to a physical/visual phenomenon?
Roger Cicala ·
You can (and we sometimes do) average MTF. But it's kind of a slipper slope. A lens with an MTF of xx at a certain point because sag is xx+10 and tan is xx-10 looks really different than a lens with MTF xx because both sag and tan are xx. Still, it can be useful to average like that.
Lars Kvinge ·
I found the in focus mtf to be very useful, more useful than common mtf curves. Especially when shooting wide open/large apertures, I always focus at the point that I want in focus, often off center. In such cases in focus mtf is what matters. Thank tou for another great article. I hope you continue publishing in focus mtf curves.
Not THAT Ross Cameron ·
The optics & maths are more than my feeble caffeine-free brain can cope with this early in the morning, so dumb Q – for those using AF fine tune to get best AF from ILC system (or even best focus assist for MF lens), does this imply / state that for a given lens the optimal focus point may be off-centre, and be able to determine that point (using a representative lens). Noting it depend’s upon one’s definition of ‘optimal’, I.e. Central area only or corner to corner average.
Roger Cicala ·
Well, we’re going to the next level in the next article. But it depends on your definition of optimal. Most lenses are at their sharpest right in the center and that’s where you AF fine tune. BUT that might not be the best way to focus a subject that’s off-center, with many lenses you should focus at that point, with some it doesn’t matter much. There’s also a point xx from center where no matter where you focus you can’t get it razor sharp. That can be as close as only 1/3 of the way from center, or as far as 3/4 of the way to the edge.
Not THAT Ross Cameron ·
No worries, hint taken. Will hold my horses until the next article 🙂
Many thanks for the response, and thanks for these articles – they are greatly appreciated, even if I don’t fully understand the tedious time taken to get the data in the first place. Enjoy your weekend.
Not THAT Ross Cameron ·
The optics & maths are more than my feeble caffeine-free brain can cope with this early in the morning, so dumb Q - for those using AF fine tune to get best AF from ILC system (or even best focus assist for MF lens), does this imply / state that for a given lens the optimal focus point may be off-centre, and be able to determine that point (using a representative lens). Noting it depend's upon one's definition of 'optimal', I.e. Central area only or corner to corner average.
Roger Cicala ·
Well, we're going to the next level in the next article. But it depends on your definition of optimal. Most lenses are at their sharpest right in the center and that's where you AF fine tune. BUT that might not be the best way to focus a subject that's off-center, with many lenses you should focus at that point, with some it doesn't matter much. There's also a point xx from center where no matter where you focus you can't get it razor sharp. That can be as close as only 1/3 of the way from center, or as far as 3/4 of the way to the edge.
Not THAT Ross Cameron ·
No worries, hint taken. Will hold my horses until the next article :)
Many thanks for the response, and thanks for these articles - they are greatly appreciated, even if I don't fully understand the tedious time taken to get the data in the first place. Enjoy your weekend.
picbod ·
The burning question I have is more to do with what happens on the other side of the lens – that is, in front of the camera with the focus-field curvature.
Translating the MTF vs Field vs Focus graphs into a real-world understanding of the lens’s characteristics should be on every photographer’s radar.
For example, my antique Canon 16-35 2.8 focuses concavely at the wide end, making it ideal for interiors (the corners of the image are focused closer than the centre). At the 35mm end of the zoom range, the focus field flips to converse, with the edges focusing further back than the centre, making it suitable for 3 dimensional solids, like the exterior of a house.
While it is well past due for replacement, I continue to use it because I know and take advantage of these characteristics, and as a result get more predictably sharp and focused images from it.
And before you lens-nerds start berating me for using such an awful (yes it is) lens – I am a photojournalist not a landscape photographer so ultimate sharpness has never been at the top of my wish list. As I shoot fast and frequently, I need to trust that my photographs will be focused at the place and moment they need to be, and that makes for a much better result than a sharply detailed photo of yet another dull subject.
Needless to say, the 16-35 2.8 mkIII is on my to-buy list but may get pipped by the delightful-to-use Sigma 24-35 f2.
picbod ·
The burning question I have is more to do with what happens on the other side of the lens – that is, in front of the camera with the focus-field curvature.
Translating the MTF vs Field vs Focus graphs into a real-world understanding of the lens's characteristics should be on every photographer's radar.
For example, my antique Canon 16-35 2.8 focuses concavely at the wide end, making it ideal for interiors (the corners of the image are focused closer than the centre). At the 35mm end of the zoom range, the focus field flips to converse, with the edges focusing further back than the centre, making it suitable for 3 dimensional solids, like the exterior of a house.
While it is well past due for replacement, I continue to use it because I know and take advantage of these characteristics, and as a result get more predictably sharp and focused images from it.
And before you lens-nerds start berating me for using such an awful (yes it is) lens – I am a photojournalist not a landscape photographer so ultimate sharpness has never been at the top of my wish list. As I shoot fast and frequently, I need to trust that my photographs will be focused at the place and moment they need to be, and that makes for a much better result than a sharply detailed photo of yet another dull subject.
Needless to say, the 16-35 2.8 mkIII is on my to-buy list but may get pipped by the delightful-to-use Sigma 24-35 f2.
Philip Service ·
First, many thanks for sharing all that data. And, yes, i’d very much like to see BIF-MTF data in future lens tests. My take-away from your results is that focus stacking could be useful even for flat (2-dimensional) subjects, if one wanted maximum resolution across the entire frame. That is, combine several images made with different focus points.
Philip Service ·
First, many thanks for sharing all that data. And, yes, i'd very much like to see BIF-MTF data in future lens tests. My take-away from your results is that focus stacking could be useful even for flat (2-dimensional) subjects, if one wanted maximum resolution across the entire frame. That is, combine several images made with different focus points.
Claudia Muster ·
Hey, but this is EXACTLY what I had asked you some time ago in a comment! Thank you very much for these interesting measurements. Yes, I think they are useful.
That one plot of a “normal” MTF with the focus point set to 14mm is very useful, too. I didn’t think of this before, and it’s a strong warning that one should be well aware of the side effects.
And while I’m at it: Yet another thing that would be interesting is MTF curves at various focus distances. As I understand, your MTF curves (and published MTF curves in general) are measured with the focus set to infinity. But this is only of limited interest for e.g. a portrait photographer. I suspect that different lenses are optimized for different focus distances.
Roger Cicala ·
Claudia, you did indeed!!! We can only test at infinity, but the focus and field curvature in general should not change. Absolute MTF curves and sharpness can, but only at really close distances with any modern lens. There might be a difference if we could test at 12 feet versus infinity, but it should be small, and well, we can’t anyway. 🙂
Claudia Muster ·
Hey, but this is EXACTLY what I had asked you some time ago in a comment! (I think it was when you presented the field curvature plots for the first time and I realized that the field curvature plays a major roe in the MTF curves.) And you even found a name for it! Thank you very much for these interesting plots. Yes, I think they are useful.
That one plot of a "normal" MTF curve with the focus point set to 14mm off center is very useful, too. I didn't think of this before, and it's a strong warning that one should be well aware of the side effects.
And while I'm at it: Yet another thing that would be interesting is MTF curves at various focus distances. As I understand, your MTF curves (and published MTF curves in general) are measured with the focus set to infinity. But this is only of limited interest for e.g. a portrait photographer. I suspect that different lenses are optimized for different focus distances, and it would be interesting to see how big (or small) the effect ist.
Roger Cicala ·
Claudia, you did indeed!!! We can only test at infinity, but the focus and field curvature in general should not change. Absolute MTF curves and sharpness can, but only at really close distances with any modern lens. There might be a difference if we could test at 12 feet versus infinity, but it should be small, and well, we can't anyway. :-)
Jim A. ·
Seems like another application for focus stacking? Maybe instead of multiple exposures for more dynamic range, we have multiple exposures for highest MTF. If your frame rate and focusing mechanics were fast and accurate enough, you might make field curvature less of an issue and make sharper photos too…we just need the camera electronics to handle the image stacking calculations and output the final file. Obviously not gonna work for race car photos, but maybe still life and other slow things, wide, sweeping landscapes that look like they were shot on medium format. Hmmm?
Jim A. ·
Seems like another application for focus stacking? Maybe instead of multiple exposures for more dynamic range, we have multiple exposures for highest MTF. If your frame rate and focusing mechanics were fast and accurate enough, you might make field curvature less of an issue and make sharper photos too...we just need the camera electronics to handle the image stacking calculations and output the final file. Obviously not gonna work for race car photos, but maybe still life and other slow things, wide, sweeping landscapes that look like they were shot on medium format. Hmmm?
boeck hannes ·
now we should find a way to render space according to our lens field curvature and nothing will ever be out of focus again!
but i have to admit i shoot lots of flat surfaces actually. paintings. and i take good care that i am absolutely paralell to them. if possible i even do it with my 85mm. i stop it down to f8 to even things out. havent yet seen a difference in sharpness no matter where i focus in the frame at f8 if the camera is properly aligned. maybe i could get a sharper result at f4 and the way you describe, but it seems unpractical in the field for me.
but i think i´ll keep shooting at f8 because if i´m not perfectly aligned it still gives me deep enough focus to get away with it. at f4 the area of acceptable sharpness gets pretty thin at 85mm.
anyway, interesting material again!
Roger Cicala ·
I didn’t mean to say people should shoot at f/8. I was leaning toward the portrait photographer who wants to know how far from center he can get a sharp image at wider apertures.
boeck hannes ·
i think your findings are relevant to more than just the portrait photographer. i for one am searching for peak sharpness and a uniform field at the same time. the sigma 85 as you have shown seems to provide that. i have to test if it makes a meaningful differenece on my d800 compared to my venerable Ai 85f2.
Roger Cicala ·
Oh, you’ll definitely like the next article then. 🙂
Roger Cicala ·
Oh, you'll definitely like the next article then. :-)
boeck hannes ·
now we should find a way to render space according to our lens field curvature and nothing will ever be out of focus again!
but i have to admit i shoot lots of flat surfaces actually. paintings. and i take good care that i am absolutely paralell to them. if possible i even do it with my 85mm. i stop it down to f8 to even things out. havent yet seen a difference in sharpness no matter where i focus in the frame at f8 if the camera is properly aligned. maybe i could get a sharper result at f4 and the way you describe, but it seems unpractical in the field for me.
but i think i´ll keep shooting at f8 because if i´m not perfectly aligned it still gives me deep enough focus to get away with it. at f4 the area of acceptable sharpness gets pretty thin at 85mm.
anyway, interesting material again!
Roger Cicala ·
I didn't mean to say people should shoot at f/8. I was leaning toward the portrait photographer who wants to know how far from center he can get a sharp image at wider apertures.
Valentin Alexiss ·
U could post another practical graph : not far from BIF-MFT, but something near that’s BPF-MFT.
Best Practical Focus-MFT : That is, what happens when you choose a focus that gives u the best overall IQ
on all the frame, but not exactly in the center.
I guess that’s what ur eye is used to, when manually focusing in general, for standard purpose framing : so BPF-MFT would be the practical quality u can get from a lens most of the time.
In case of the Zeiss CP.2 35mm T2.1, @5.6 : it seems a +0.05mm focus correction (from a perfect focus in the center) would give the best overall results, for example. Then u could compare the Zeiss chart with what the Sigma is doing @5.6, with it’s own BPF-MFT @5.6.
Brandon Dube ·
ding ding ding, we have a winning guess 🙂
Roger Cicala ·
Spoiler alert!
Valentin Alexiss ·
U could post another practical graph : not far from BIF-MFT, but something near that's BPF-MFT.
Best Practical Focus-MFT : That is, what happens when you choose a focus that gives u the best overall IQ
on all the frame, but not exactly in the center.
I guess that's what ur eye is used to, when manually focusing in general, for standard purpose framing : so BPF-MFT would be the practical quality u can get from a lens most of the time.
In case of the Zeiss CP.2 35mm T2.1, @5.6 : it seems a +0.05mm focus correction (from a perfect focus in the center) would give the best overall results, for example. Then u could compare the Zeiss chart with what the Sigma is doing @5.6, with it's own BPF-MFT @5.6.
Brandon Dube ·
ding ding ding, we have a winning guess :)
Tim ·
Thanks for yet another great read. Definitely count me as one of the readers interested in seeing BIF plots in future posts.
You could also consider making user-tunable MTF plots, which should be straightforward to do from the data you are collecting. What I have in mind is a little slider at the bottom of the plot that selects the location of focus point, with far left being center and far right being focused at the edge. This would allow one to see how the behavior (MTF) of the lens changes as the focus position is changed. I would find this a lot more informative and intuitive than a BIF plot. Of course, it doesn’t quite work for print publishing, but for print, one could restrict to just four key MTF plots (e.g. center, 25%, 50%, and 75% to the edge).
Tim ·
Thanks for yet another great read. Definitely count me as one of the readers interested in seeing BIF plots in future posts.
You could also consider making user-tunable MTF plots, which should be straightforward to do from the data you are collecting. What I have in mind is a little slider at the bottom of the plot that selects the location of focus point, with far left being center and far right being focused at the edge. This would allow one to see how the behavior (MTF) of the lens changes as the focus position is changed. I would find this a lot more informative and intuitive than a BIF plot. Of course, it doesn't quite work for print publishing, but for print, one could restrict to just four key MTF plots (e.g. center, 25%, 50%, and 75% to the edge).
Michael T. ·
First of all Imwant to thank you for the amazing work you are doing and for sharing it with us. And I really like your humor/sarcasm that makes all the most interesting but rather dry facts much more entertaining.
Just on minor issue: as far as I know (and read the graphs) 0.3 mm equals 300 microns not 30 microns (top of the article).
I guess (and read the graphs) that even with very sharp lenses the focus doesn’t change that much within +- 30 microns (0.03 mm)
And please, give us more of the good stuff.
Thanks
Michael
Roger Cicala ·
Thank you, Michael. Fixed that.
Michael T. ·
First of all I want to thank you for the amazing work you are doing and for sharing it with us. And I really like your humor/sarcasm that makes all the most interesting but rather dry facts much more entertaining.
From reading through the comments I think I know what you are up to in Part III. So no need for me to make suggestion on how to represent the data. If you would want to use only one number for this you could in the MTF vs field vs focus plots use the depth in mm (or microns) of the largest rectangle (total width) with a MTF of say 85 or higher. Wasn't this something you mentioned in a comment below? Yes, I know this simple number isn't your thing (and the idea I think you have is way nicer, too).
Just on minor issue: as far as I know (and read the graphs) 0.3 mm equals 300 microns not 30 microns (top of the article).
And please, give us more of the good stuff.
Thanks
Michael
kirkmoon ·
Very interesting information. Thank you!
It strikes me that another very useful way to slice this data would be to use it to define a focus plane that optimizes focus for the most points across the focus plane. In other words, to define the plane that produces the highest average focus integrated across the entire plane. For example, for the Zeiss 50mm lens, it appears that placing the focus plane approximately 0.1 mm forward of the zero position on the tangential plot and 0.08 mm forward on the sagittal plot would produce the highest average focus across the plane.
If one were willing to sacrifice a bit of sharpness in the center of the image for the sake of overall better focus this could potentially be valuable (assuming that sharpness was the desired goal.) Or are these differences so small that it wouldn’t be possible to actually make such fine adjustments when manually focusing the lens?
Can your software do this? Would be interesting.
Roger Cicala ·
You win the guessing game. Stay tuned tomorrow.
Tim ·
Presumably the optimal focus location is defined as something like the position that maximizes the sum of the areas under all the MTF curves? I suppose I’ll find out tomorrow!
Tim ·
Presumably the optimal focus location is defined as something like the position that maximizes the sum of the areas under all the MTF curves? I suppose I'll find out tomorrow!
Someone ·
With “Focus fine tuning” feature of modern lenses and cameras, you could even dial such numbers as to automatically use “best across the frame” focusing when you focus with the central point. Could be useful for landscape, for example, but probably to awkward to use and unusable for other situations.
kirkmoon ·
Very interesting information. Thank you!
It strikes me that another very useful way to slice this data would be to use it to define a focus plane that optimizes focus for the most points across the focus plane. In other words, to define the plane that produces the highest average focus integrated across the entire plane. For example, for the Zeiss 50mm lens, it appears that placing the focus plane approximately 0.1 mm forward of the zero position on the tangential plot and 0.08 mm forward on the sagittal plot would produce the highest average focus across the plane.
If one were willing to sacrifice a bit of sharpness in the center of the image for the sake of overall better focus this could potentially be valuable (assuming that sharpness was the desired goal.) Or are these differences so small that it wouldn't be possible to actually make such fine adjustments when manually focusing the lens?
Can your software do this? Would be interesting.
Someone ·
With "Focus fine tuning" feature of modern lenses and cameras, you could even dial such numbers as to automatically use "best across the frame" focusing when you focus with the central point. Could be useful for landscape, for example, but probably to awkward to use and unusable for other situations.
Professional Photographer ·
I totally LOVE this article. This is a genius idea! So relevant to the way I actually need and use my lenses in day-to-day life! How often do I focus dead center, anyway? Thank you, thank you, thank you!!!!!
Dragon ·
This is very useful data when choosing a lens, particularly for those who like to use off-center focus points to place their subject. At the end of the day, flat field isn’t all that important for a lot of subjects (e.g. portraits) because the vast majority of the image is intentionally out of focus. This data clearly shows how good the focal point of the image can be when that focal point is off center. Combined with the field curvature chart, you can get a good idea of the area that will be in or near optimum focus when using an off-center focus point. The ideal tool would be an algorithm that has access to all those test points that would handily draw you an MTF chart around a chosen focus point. I see birth of a new software business :-).
Dragon ·
This is very useful data when choosing a lens, particularly for those who like to use off-center focus points to place their subject. At the end of the day, flat field isn't all that important for a lot of subjects (e.g. portraits) because the vast majority of the image is intentionally out of focus. This data clearly shows how good the focal point of the image can be when that focal point is off center. Combined with the field curvature chart, you can get a good idea of the area that will be in or near optimum focus when using an off-center focus point. The ideal tool would be an algorithm that has access to all those test points that would handily draw you an MTF chart around a chosen focus point. I see birth of a new software business :-).
Zak McKracken ·
Brilliant! This is what I originally believed MTF graphs to show, and I think it’s exactly what is the more representative thing in many real situations. Exceptions are of course focus-and-recomposing or “wall photography” (test charts, reproductive photos of paintings… PIV) where all the things that need to be in focus are actually in a plane.
As such, I’d say that the BIF graphs are even more relevant than the regular ones (although, of course those still aren’t useless). I wonder how many cheaper lenses with good center sharpness but bad off-center numbers in regular MTF tests may actually look much better in BIF tests. I guess there will be some lenses that will suddenly look a lot more interesting, and some who won’t.
Zak McKracken ·
Brilliant! This is what I originally believed MTF graphs to show, and I think it's exactly what is the more representative thing in many real situations. Exceptions are of course focus-and-recomposing or "wall photography" (test charts, reproductive photos of paintings... PIV) where all the things that need to be in focus are actually in a plane.
As such, I'd say that the BIF graphs are even more relevant than the regular ones (although, of course those still aren't useless). I wonder how many cheaper lenses with good center sharpness but bad off-center numbers in regular MTF tests may actually look much better in BIF tests. I guess there will be some lenses that will suddenly look a lot more interesting, and some who won't.