Most OOF Blur Lens

[chromophore]

So I dug up my program and here is a sample of the output for a comparison between an ideal 85mm f/1.2 lens versus an ideal 300mm f/2.8 lens. Note I say "ideal" here because in practice, the actual EF 85/1.2L and EF 300/2.8L IS II are not exactly 85mm or 300mm, nor are they exactly f/1.2 or f/2.8. But to a large extent, the model is a fairly good representation of what happens in practice.

What the plot shows is that for a subject magnification of 0.06 (meaning that an object in the plane of focus that is 100mm long will project an image onto the sensor that is 6mm long, which is roughly in the ballpark for portraiture), the red curve expresses the diameter of the blur circle for a 300/2.8 lens as a function of the distance between the subject and the background. In other words, when the subject is in focus and the camera-subject distance is adjusted so that the subject has an apparent magnification of 0.06, a point light source 1 meter behind the subject will be rendered as a blur disk with diameter about 1 mm on the sensor. For a point light source 100 meters behind the subject, the blur disk on the sensor will have a diameter of about 6.15 mm.

The blue curve corresponds to the 85/1.2 lens. To maintain the magnification of 0.06, you'd stand quite a bit closer to the subject than for a 300/2.8. But for an object 1 meter behind the subject, the 85/1.2 would have a blur circle diameter of about 1.6 mm, which is more blur than you would get with the 300/2.8 at this background distance. And then this situation is reversed for objects more than 6 meters behind the subject in focus!

So you can see that the extent of blur depends not just on the lens, but the spatial relationship between the subject and the background. And this plot shows exactly what I have been trying to explain: the 85/1.2 goes blurrier for nearby background objects, but not for faraway objects compared to the 300/2.8.

 

[rfdesigner]

So I dug up my program and here is a sample of the output for a comparison between an ideal 85mm f/1.2 lens versus an ideal 300mm f/2.8 lens. Note I say "ideal" here because in practice, the actual EF 85/1.2L and EF 300/2.8L IS II are not exactly 85mm or 300mm, nor are they exactly f/1.2 or f/2.8. But to a large extent, the model is a fairly good representation of what happens in practice.

What the plot shows is that for a subject magnification of 0.06 (meaning that an object in the plane of focus that is 100mm long will project an image onto the sensor that is 6mm long, which is roughly in the ballpark for portraiture), the red curve expresses the diameter of the blur circle for a 300/2.8 lens as a function of the distance between the subject and the background. In other words, when the subject is in focus and the camera-subject distance is adjusted so that the subject has an apparent magnification of 0.06, a point light source 1 meter behind the subject will be rendered as a blur disk with diameter about 1 mm on the sensor. For a point light source 100 meters behind the subject, the blur disk on the sensor will have a diameter of about 6.15 mm.

The blue curve corresponds to the 85/1.2 lens. To maintain the magnification of 0.06, you'd stand quite a bit closer to the subject than for a 300/2.8. But for an object 1 meter behind the subject, the 85/1.2 would have a blur circle diameter of about 1.6 mm, which is more blur than you would get with the 300/2.8 at this background distance. And then this situation is reversed for objects more than 6 meters behind the subject in focus!

So you can see that the extent of blur depends not just on the lens, but the spatial relationship between the subject and the background. And this plot shows exactly what I have been trying to explain: the 85/1.2 goes blurrier for nearby background objects, but not for faraway objects compared to the 300/2.8.

That is an impressive piece of work.

Can you publish the maths behind this please?.. I'm genuinely interested

 

[K-amps]

if we compare an 85/1.2 against a 300/2.8. If you take headshots with both lenses at the same image magnification, and in both shots, you focus on the eyes, you will find that with the 85/1.2, the subject's nose will be much more blurry than with the 300/2.8. But a very distant background will be more blurry on the same shot at 300/2.8 than at 85/1.2. The slope of blur circle diameter as a function of subject-background separation is greater with the 85/1.2 but the asymptotic behavior of blur circle diameter is such that the 300/2.8 will achieve more blur.

This is great info, Gracias! Makes a lot of sense... it clears many things for me now.

The graph makes it even clearer.

Can I use your information as reference to teach this to a group of budding photographers this concept?

 

[TominNJ]

to get that creamy OOF background you need to work in the minimum focusing distance for the lens (get as close as you can), shoot wide open or the maximum aperture that will give you the DOF you need and position the camera so you have as much distance between the subject and background as possible.

 

[chromophore]

if we compare an 85/1.2 against a 300/2.8. If you take headshots with both lenses at the same image magnification, and in both shots, you focus on the eyes, you will find that with the 85/1.2, the subject's nose will be much more blurry than with the 300/2.8. But a very distant background will be more blurry on the same shot at 300/2.8 than at 85/1.2. The slope of blur circle diameter as a function of subject-background separation is greater with the 85/1.2 but the asymptotic behavior of blur circle diameter is such that the 300/2.8 will achieve more blur.

This is great info, Gracias! Makes a lot of sense... it clears many things for me now.

The graph makes it even clearer.

Can I use your information as reference to teach this to a group of budding photographers this concept?

Yes of course; although this is a very technical topic and the relationships are rather complex. The general rule is that, for a constant subject magnification, in the distant background regime, entrance pupil diameter is the dominating factor. In the near background regime, the f-number is the dominating factor for the size of the blur circle. The subject-background separation distance at which the two switch over depends on the focal lengths being compared, the f-numbers being compared, and the subject magnification.

I would love to be able to show the way the plot changes with the selection of different parameters--it is a fully interactive plot--but this is not possible at this time.

 

[K-amps]

if we compare an 85/1.2 against a 300/2.8. If you take headshots with both lenses at the same image magnification, and in both shots, you focus on the eyes, you will find that with the 85/1.2, the subject's nose will be much more blurry than with the 300/2.8. But a very distant background will be more blurry on the same shot at 300/2.8 than at 85/1.2. The slope of blur circle diameter as a function of subject-background separation is greater with the 85/1.2 but the asymptotic behavior of blur circle diameter is such that the 300/2.8 will achieve more blur.

This is great info, Gracias! Makes a lot of sense... it clears many things for me now.

The graph makes it even clearer.

Can I use your information as reference to teach this to a group of budding photographers this concept?

Yes of course; although this is a very technical topic and the relationships are rather complex. The general rule is that, for a constant subject magnification, in the distant background regime, entrance pupil diameter is the dominating factor. In the near background regime, the f-number is the dominating factor for the size of the blur circle. The subject-background separation distance at which the two switch over depends on the focal lengths being compared, the f-numbers being compared, and the subject magnification.

I would love to be able to show the way the plot changes with the selection of different parameters--it is a fully interactive plot--but this is not possible at this time.

Awesome thanks!

Can you post the graphical relationships between a few Canon L's for example:

50 1.2
85 1.2
100 2.8L
135 F2
180 3.5
200 2.8
200 F2
300 2.8
400 4
400 2.8
600 f4

 

[Valvebounce]

Hi Folks.
I dip into the photography technique threads every so often, this has been among the most interesting I have read, thank you to those that know and to the others that ask and draw that knowledge out!

Chromophore, have you published this program of yours? Would you? How much? Does it work for zooms choose 1 focal length or only primes, only lenses that you have written it for? I love to put my lenses through software like this to have an idea how they behave, I have dof calculator and have spent way too long just putting numbers in to that!

Cheers, Graham.

 

[Tuke]

If you want graphs instead of numbers, why dont you use one of these?

 

[K-amps]

If you want graphs instead of numbers, why dont you use one of these?

Excellent!

 

[takesome1]

which Canon lens would give most Out of focus blur?

I go with the 180mm L macro shot wide open at the minimum focal distance.

 

[Steve Balcombe]

... my first stipulation was for a given frame, which DoF calculators ignore. You are right, fix the framing, and it gets quite complex.

On the contrary, it makes it very simple (see below). It is also the most realistic and generally useful comparison - same sensor (we're comparing lenses, not cameras), and same framing (we're looking at alternative ways to take the same shot).

The general rule is that, for a constant subject magnification, in the distant background regime, entrance pupil diameter is the dominating factor. In the near background regime, the f-number is the dominating factor for the size of the blur circle.

I thought I'd better read all the responses before posting mine, and yours is the one which really gets this right. I'd like to rephrase it though. For a given sensor size and subject framing:

- distant background blur is proportional to entrance pupil size (i.e. focal length/f-number)

- depth of field is the same for the same f-number, regardless of focal length.

This means that if your aim is a head and shoulders portrait with enough depth of field to have the whole face and hair in focus, but with maximum background blur for subject separation, you should choose a long lens as that will give you more background blur at the required f-number. If your aim is to have one eye in focus, choose a shorter lens with a very large aperture.

 

[K-amps]

- depth of field is the same for the same f-number, regardless of focal length.

This means that if your aim is a head and shoulders portrait with enough depth of field to have the whole face and hair in focus, but with maximum background blur for subject separation, you should choose a long lens as that will give you more background blur at the required f-number. If your aim is to have one eye in focus, choose a shorter lens with a very large aperture.

I thought DoF varies with FL, because framing would change with FL. So are we saying that a 50mm F1.2 will have same DoF as a 85mm F1.2 regardless of subject distance?


I take it that background blur then depends both on subject distance and background distance then. Which one of these dominates?

 

[Steve Balcombe]

I thought DoF varies with FL, because framing would change with FL. So are we saying that a 50mm F1.2 will have same DoF as a 85mm F1.2 regardless of subject distance?

I was careful to say that the DoF is constant for the same framing. We adjust shooting distance to make the framing the same - for example you will shoot a head-and-shoulders portrait from further away with an 85/1.2 than with a 50/1.2. That's how you get constant DoF. This is a very realistic comparison, because in terms of the subject it's the "same" shot. Not the same perspective of course, but that is an indirect consequence of using a different lens.

I take it that background blur then depends both on subject distance and background distance then. Which one of these dominates?

It does, but again I was careful to say distant background blur. Another way to look at it is maximum background blur, which is basically the same thing. However there is an underlying assumption that the subject is reasonably close - if your subject is a cityscape and the background is the mountains beyond, you won't get much background blur.

I'm not sure it's possible to answer the question in terms of which one "dominates", they both affect blur.

I think there is a danger of losing the context here. The desire is for maximum background blur, and you are clearly not going to get this if the background is very close. Furthermore we are looking at lens options for the same shot of the same subject, so subject distance is dictated by that. If your choice of subject distance is limited, then your choice of focal length is also limited and the original question no longer applies.

 

[meywd]

If you are shooting portraits - and not candids - then the brenizer method is a good way to overcome the OOF limitations of the lens, I know it can't be used for fast shots, but if you have the time its amazing.

this was made from 86 shots using the 70-200mm f/2.8 IS II @ 200mm f/2.8

 

[chauncey]

Any amount of background blur is achievable with any lens set to it's most wide open setting
via the use of PP...simply take the shot, then take another with whatever blur on whatever portion
of the image that you want the most blur, then merge them in PP. No brainer.

 

[photonius]

If you want graphs instead of numbers, why dont you use one of these?

nice.
There is another site with lots of graphs, but not interactive:
http://www.zen20934.zen.co.uk/photography/dof/dof.htm

 

[photonius]

With this site: http://dofsimulator.net/en/ and "Lock frame" box crossed you get the easiest answer to your question!

50mm, f/1.0
- background blur: 1.138mm
- depth of field: 12.5cm

85mm, f/1.2
- background blur: 1.532mm
- depth of field: 15.7cm

200mm, f/2.0
- background blur: 2.162mm
- depth of field: 26.2cm

600mm, f/4.0
- background blur: 3.243mm
- depth of field: 52.5cm

1200mm,f/5.6
- background blur:
- depth of field:

These are from full body shots. In macro distances the dof is yet another thing/problem.

That's a new one I haven't seen before.
This one has been around for many years, but not well known:
http://kingfisher.in.coocan.jp/boke2/bokekeisan2e.html

 

[PavelR]

That's a new one I haven't seen before.
This one has been around for many years, but not well known:
http://kingfisher.in.coocan.jp/boke2/bokekeisan2e.html

Thank you very much for the link! It is finally the model that correspond to my experience that 400/5.6 produce bigger background blur than 200/2 = focal length is the most influencing variable (with decent distance subject ~ background about 10m+).

 

[chromophore]

At present, I don't have the time to go through and generate plots for all the common EF primes, but I might be able to do that over the weekend. The problem is that there are a LOT of plots that would need to be generated, because for each lens, I would need to show a plot for a variety of subject magnifications and distance regimes. The curves *look* similar, but the axes and scales change. If I could publish it as a Wolfram Demonstration Project (the program is written in Mathematica), then it would be a simple matter for anyone to interact with it.

Anyway, the other reason I wanted to post here is because there are still some lingering misconceptions that I would like to clear up.

First, depth of field is not the same as background blur. Depth of field tells you what is in acceptable focus. If you have studied the plots I have provided thus far, you should understand by now that what the in-focus areas of the field does not say much, if anything, about what the out-of-focus areas will look like. Quantitatively, you can read off the DOF behind the subject in focus by drawing a horizontal line at the CoC (circle of confusion) diameter on the plot, and seeing where it intersects the curve. This intersection point corresponds to the distance behind the subject in focus for which objects will still appear "acceptably sharp."

The problem of course, is that by definition, objects within the DOF are "acceptably sharp," so it is difficult to ascertain with precision the actual distance at which a real-world lens is focused within the DOF. To a first-order approximation, for "non-distant" subjects, this distance is around the midpoint of the DOF (especially in the macro regime).

But back to blur circle diameter. For a given subject magnification, in the near regime, the DOF is approximately the same. It is not exactly equivalent. But in practice, any differences are not likely to be noticed for two reasons: (1) they are tiny; (2) the DOF is an abstraction that simplifies what is in fact a continuous phenomenon of increasing blur as a function of distance from the plane of sharpest focus, into a dichotomous outcome ("acceptably sharp" vs. "not acceptably sharp").

What we need to remember is that all our discussions in this thread pertain to the condition that all comparisons are to be made at a constant subject magnification; that is to say, regardless of focal length, the subject-camera distance is adjusted so that the subject in focus appears the same size. This may or may not be possible for all lenses, since some lenses cannot achieve the magnification factor that other lenses can (non-macro vs. macro being an obvious example).

So, if we want to make a fair comparison of what lenses give the greatest background blur for a given real-world scene, then we impose the following conditions: (1) We cannot change the subject, background, or relationship between them; (2) we are unwilling to do post-processing trickery; (3) we are allowed to choose the lens; (4) we are allowed to choose where we stand relative to the subject; (5) we are not concerned with differences in camera perspective as a consequence of subject-camera distance choice; (6) we do not model individual lens aberrations but model them as ideal lenses.

Then, under the above conditions, the appropriate comparison is done with the plots I have described. Again, what we see in the out-of-focus areas is a combination of factors. That an 85/1.2 can achieve more blur than a 300/2.8 in the near regime is not quantitatively explained any other way, because, as I must remind the reader of what I wrote at the beginning of this already very long response, the DOF tells us what is in focus. It doesn't tell us how rapidly something goes out of focus, at least not directly.

 

[meywd]

Any amount of background blur is achievable with any lens set to it's most wide open setting
via the use of PP...simply take the shot, then take another with whatever blur on whatever portion
of the image that you want the most blur, then merge them in PP. No brainer.

huh?