I understand why the same lens produces a smaller FOV on a small sensor. I also understand that 250mm gives 5X magnification and 400mm gives 8X magnification (given that 50mm is normal) and that lens characteristics do not change regardless of sensor size. Since I lack the lenses with which to test myself, I am asking the question:

Do images taken with a 250mm on a 1.6 crop factor sensor really appear identical to images taken with 400mm on a full frame? I'm assuming that subject size in the frame is identical. Stated another way, can I stand in the same place with those two lenses, shoot the same subject and get the same result? I'm only talking about apparent magnification - all other IQ issues are irrelevant.

I have read so much conflicting blather all over the net that I thought I would ask here, where there seems to be more fact and reasoned opinion than in most places.

TIA

Its not just about subject size it is about the rendering out of focusness of subjects outside the field of focus ie the foreground or background

Larger sensors have less depth of field for a given aperture and field of view

or put another way using a specific lens lenght aperture and distance combo you get a fixed look, but with a larger chip you get more of that look

(ie a 50 is a tele on a D80, a normal on a D3 and a wide on an H1 but crop the H1 images in a quarter and the look is the same as the D80 shot)

Whether a large chip look can be replicated by using wider apertures on smaller chip cameras is really the debate

It is my opinion. that i am convinced is correct, that one can not recreate different looks from different sensor sizes by varying aperture because diffferent lenses and apertures drop in and out of focus at different RATES

ie you can perform a test on two sensor size aperture, combos and get point A sharp and point B equally defocussed with different chip sizes but if you bring a point C into the equation/test the amount of defocussing will not be holdable consistent across different sensor sizes

ANd in a three dimensional world there is nearly always a 'point C'

--------------

My real world summary,

big chip = smooth creamy look that is great for portraits but it can be difficult to get huge DOF for architecture or landscape when using a big chip

My personal preference - I hate small chip cameras exept for using long telephotos - where weight/size/cost is an issue

ps I own Hassy H1, D3, D80 (different chips) and many lenses from 10.5 fisheye to 600F4 - so I might have more real world experience than some 'web blatherers'

If you are into long lenses a small chip camera is a good thing because it is much cheaper and more practical to buy and use if you are after as much magnification as possible

With longer lenses the differences become less noticable

A 14mm on a D80 is wildly different to a 35 on an H1 in terms of DOF

Whereas the look of 200 on the D80 is not wildly different to using a 300 on a D3

The sweet point of MF cameras seems to be when trying to shoot fashion - one typically needs the model to have sharp face and clothes but desires an out of focus background - hence many fashion shooters use MF cameras even when the target repro is only magazine size so MP is not really a factor

S

Theoretically, the answer is yes, provided you also increase the F stop (stop down)on the 400mm lens by a factor of 1.6x. Example, F8 on the 250mm lens is equivalent to f13 on the 400mm lens. (Actually f12.8 but let's not quibble).

This characteristic is one of the advantages of the smaller format. You can get the same FoV with a shorter lens and the same DoF with a wider aperture on that shorter lens. But the reverse is also true. You can get a shallower DoF with a larger format at the same f stop, shooting from the same position with a lens that gives you the same FoV.

There is a myth about lens performance at really wide apertures. The shallow DoF tends to give the impression that the part that's in focus is sharper than it really is, at least with 35mm lenses. MF lenses on a DB can be used at about a one stop numerically larger aperture to produce a similarly OoF background, but the part that's in focus (the lady's eyeball) can be really sharp if it's a good MF lens designed for the DB. Example, a good MF 120mm lens at f4 can be sharper than a good 85mm lens on 35mm format at f2.8. As a consequence of the greater sharpness at the plane of focus, the DB shot appears to have greater 3-dimensionality.

Morgan,
I have to say, but without any intention of insulting you, I hope you understand , that I find your answer very convoluted and confusing.

There is no doubt that larger sensors tend to have an image quality advantage because they collect more photons. The greater the difference in format size, the more pronounced such differences as 'creaminess of texture' can be and the greater the pixel count the greater the likelihood that what's in focus will be sharper, although it's perhaps illuminating to read in Michael's review of the Canon G7, that he occasionally confused (momentarily at least) G7 prints with Leica M8 prints.

To try and add some clarification here, I would say that one has to be aware of the differences between system performance and lens performance. Larger formats have the advantage that they require less resolution from the lens (ie. lower lp/mm at the same MTF). Conversely, smaller formats need higher resolving lenses to equal the system resolution of larger formats.

Another issue is that standard DoF calculators do not take lens quality factors into consideration. There is an implied assumption that all lenses are equal in quality. Now we know that is just not true. All lenses are not equal (in sharpness) and to the extent they are not equal, the DoF formulas and calculators are imprecise.



An H1 is a film camera body, isn't it? What DB did you have on that H1 when you discovered that a 14mm lens on a 10mp D80 is wildly different to a 35mm lens on the H1 in terms of DoF? What is the quality of the 14mm lens on the D80? What is the quality of the 35mm lens on the H1?

The OP's question related to the difference between a 1.6 cropped format and FF 35mm. The D80 in relation to a DB represents a 2.25x crop factor and the difference in pixel count is likely huge. That's a different ball game entirely.

To get back to a more realistic comparison, for example the recently announced 12mp Canon 450D with a 250mm lens compared with a 5D with 400mm lens, the answer is that the 450D with 250mm lens at f8 will provide very similar DoF and over all quality to the 5D with 400mm lens at f13.

The smaller sensor of the 450D requires a higher resolving lens and it's got it at f8, which is the aperture at which most 35mm lenses are usually sharpest. All lenses are roughly equal at f13.

On the other hand, if you change the conditions and compare a 450D using a Canon 50/1.4 at f1.4, with a 5D and 85/1.2 at f2.2, the results could be wildly different. The 50/1.4 needs to be sharper at f1.4 than the 85/1.2 is at f2.2, in order for the 450D to produce similar quality results. I think it would be no understatement to say that the opposite is true. I think there is no doubt that the 85/1.2 is much sharper at f2.2 than the 50/1.4 is at f1.4 and this fact will skew the results.

Which bit of my post dont you understand ?


ps I have a sinar 54LV digital back on my H1 camera and a nikkor 14/2.8

S

This bit:



What on earth does that mean?

Ok for this example figures are approximate

Right. stand at a point with two cameras lets say for simplicity a FUll frame with a 300mm and a Crop with a 200m lets say we have f8

Your subject say a person is far enough to be framed head to toe

(the cameras are upright)

Take two image anc compare them -

both images are at first glance identical - a person framed head to toe in focus.

Then you look more carefully and realise that there is a telegraph pole 5 meters behind the person

It is distracting

But you notice that the FF shot on the 300 the pole is more out of focus - less distracting

Five meters behind the first pole is a second pole even more out of focus but still a distraction

You go back to the scene and re-shoot the image with the small sensor camera, having chosen a wdier aperture than the inital f8

You play with the aperture until the first pole has an equal 'confusion' to the FF image by opening the aperture

You then print the new wider aperture image

Still it is not the same as the FF image, yes the first pole and the subject look the same but the second pole is differently rendered still from the FF image -

You then go back and shoot loads of aperture combinations and realise that what ever you do you cant get the small chip image to look the same as the FF image

because the focus is dropping off at different RATES with the different chip sizes, irrelevant of aperture

You may be able to get more blur with the Small chip by using the wider aperture but you will never get the same blur RATIOS between the different subjects

-----------

Anyone knows that the DOF drops off faster with a lens of a longer focal lenght - look at the grass on the pitch in a sports photo say shot with a 300 2.8

If one had a 300 2.8 10 plate camera lens (Yes that lens would be the siaze of a car)

One could photograph a good chunk of the stadium (300 is quite wide on a 10.8 plate) but would still have that 'focus cut' though the grass same as the sports picture

a totally differnt look to a shot of the stadium shot say with a 35mm on a canon

S

Morgan,
This sounds to me like lens quality variation and nothing necessarily to do with format. Lenses are very imperfect products. Just look at any real world MTF chart. The sharpness (contrast at a specific spatial frequency) is often all over the place. In the centre of its image circle a lens is usually sharpest, but not always. It might sometimes be sharpest 2mm from its centre, fall in sharpness towards a point 10mm from its centre, rise in sharpness from 10mm to 18mm from the centre then fall drastically towards the corner at 22mm (in the case of a 35mm lens). And this pattern is likely to be different at different apertures and different spatial frequencies.

There's an interesting comparison of the Canon 14/2.8, Nikkor 14/2.8 and Sigma 12-24 zoom at http://www.16-9.net/lens_tests/14mm/14mm_test_2.html

The Nikkor 14/2.8 has some very weird performance at wide apertures, in this test, which gradually disappear, however, as the lens is stopped down.

If your copy of the Nikkor 14/2.8 is anything like the one tested in the above review, it's no wonder you found it wildly different to your 35mm lens on the H1 .

(Original reply deleted after reading OP again)

Lenses are different enough in characteristics that it should never be exactly the same (ignoring resolution and noise, as you seem to have requested). The bokeh is different with different lenses, and bokeh is something that people often get sensitive to when they look at more and more pictures from various lenses. The bokeh on a lot of my photos when I use too many TCs, and of many images linked to by other people often makes me cringe; bad bokeh makes me feel uncomfortable with its spatial distortion.

Actually, if you want to be really precise about this, you can never get the same effect with 2 different lenses irrespective of format differences. I'm sure that even the quality of bokeh can be different between two lenses of the same model.

I got the impression the OP didn't really mean, "do images really appear identical". No two things are identical. Not even two Blank DVD discs on the same spindle.

One has to allow for a certain interpretive license here. I interpreted the question as meaning, "Are the two images broadly identical in respect of FoV and DoF?", and I think the answer is yes.

And I think the answer is no !

But it only really becomes obvious when comparing a D80 to an H1 the second having about four times the recording area of the first

The practical differences between cameras is most important

like the pleasure of vieing though an H1, the AF on a D3 or the cheap telephoto options on crop cameras

S

Assuming "perfect" lenses and sensors, it is impossible to get identical photos in two different size formats.

You can get damn close, and if you take a photo of a 2-dimensional plane (perfectly aligned with the film plane), I suppose they would match.

I don't have the proof on me anymore, but you can prove it yourself with some basic algebra. It's actually in interesting exercise.

Find the equation for the circle of confusion based on aperture, focal length and focus distance and distance from the sensor. Let's assume that we're comparing 35mm with aps-c (1.6x crop factor).

First pick lenses for the sensor; how about a 160mm lens for the 35mm sensor and 100mm lens for the aps sensor. This will give them the same angle of view.

Now pick a focus distance. You could use a a constant "c" here or plug in a number (like ten meters).

We need to solve to find an aperture aperture that makes the CoC curves (image) the same, here's where it gets tricky. You have 2 three-dimensional functions. The variable is aperture, and the resultant is a function of distance and size of the CoC.

It is impossible to solve this equation without imputing a specific CoC (just like those DoF calculators do). This is because the CoC, as a function of distance from the sensor, curves are not the same.

These curves are very useful in photography. They allow you to calculate desired hyperfocal distances, Depth of Field, and blur. Exploring them can tell you a lot about your photos.

You said that the images would be theoretically identical. Which is definitely untrue. Practically similar, yes. Lens variability and other IQ issues would overshadow the effect of CoC curves. But projecting theoretically identical images on different sized media is impossible with current optical technology.

When someone says something is theoretically true, I understand him to mean it's true according to some theory but not precisely or necessarily true in practice.

Are you saying it would be impossible to design a lens for a smaller format, which would behave with regard to DoF exactly the same as a longer focal length on a larger format after equalising physical aperture sizes on both lenses?

In other words, setting aside the difficulties of manufacturing a lens precisely according to design specifications, are you saying there's no theory which would support the design of two such lenses without contravening the laws of physics?

I don't know enough about optics or lens design to answer that question, but if this is indeed true, then of course I am wrong in asserting that the images are theoretically identical.

What would concern me here regarding a general principle of sound methodology, is that someone such as Morgan (and others) might have a bunch of lenses and a bunch of different format cameras, but not a bunch of lenses of the same focal length that fit the same format. I generally try to avoid duplicating focal lengths, but I will sometimes do that when I discover a lens which appears to be much better than what I already have.

If one does a lot of testing of lenses of equivalent FoV on different formats, equalising the physical aperture size used by adjusting the f stop accordingly, and one finds that out-of-focus areas on the different formats are surprisingly different, how can one be sure that such differences are not due to the normal variation amongst lenses?

In order to check this properly, one would need to have a number of different lenses of the same focal length for the formats one was comparing, preferrably lenses that differed in quality and model or brand, say a Mamiya 80mm, a Hassy 80mm a Rodenstock 80mm and perhaps a few zooms for the DB, and a Nikon, Canon, Sigma and Tamron 50mm for the 35mm format.

If one then found that there was very little difference in the OoF areas amongst lenses of the same focal length on the same format, but found there was a significant difference when using lens FL and DoF equivalents with different formats, then you would have proved your point.

Perhaps I should add, just in case anyone has the lenses and can find the time to do such tests, that the lenses of equal FL should be tested at various apertures. Any variation in DoF behaviour is likely to be greatest at wide apertures.

Great Explaination of what I have been trying to say





Indeed


---------------

I actually happen to have an identical lens for for three formats..

A mamiya 80mm which I have a home made T/S adapter for nikon so the lens fits MF, D3 and APS size cameras

I terms of practicality I think there are real wrold visible benefits to using MF systems versus smaller formats if ones desire is a steep DOF curve - the effect that makes the subject 'pop' off the background

Even if a close theoretical match can be found between different theoretical lenses one finds that with the smaller sensors you must be either shooting avbsolutely wide open or at an aperture that does not exist and is not available like F1.4 or F1

in the real world the only lenses that can come close to creating MF style look are the most expensive canon or nikon primes (35 1.4 50 1.2 85 1.2etc) used on a FF camera

I did a test yesterday with a model shooting some pretty tiddly apertures like F13 - when shooting tight faces one still sees a situation wher the eyeball is sharp and the nose and ears distinctly blurred (using the H1)

This reminded me that the effect is very real world

Conversly shooting kids the other day with a D3 on a commercial job I found the 80-200 didnt really have the reach I was used to compared with shooting on the D200 with that lens, really I needed a 300 2.8 on the D3, no zoom and much more money and weight, If jobs of that nature are ones bread and butter then APS can provide great practicality and value with no client dicernable differnces in the final product

Also the auto focus points on the H1 and the D3 are poorly spaced compared to nikon APS offerings like D2x and D300 which gives those camers a huge real world disadvantage over APS

SMM

Dobson,
That is interesting. I didn't know that. But how can I be sure you know what you are talking about ? Mathematicians tell me one can prove some very absurd notions at times and that such proofs seem very logical. Paul Dirac proved mathematically that anti-matter should theoretically exist, but kept quiet about for some time fearing ridicule and loss of reputation.

Can you provide a link to any authoritative source on the net that might explain in detail this algebraic proof and what the current status of such a proof might be amongst mathematicians? It seems a bit negative to me, like some of those attempts before airplanes were developed to prove that heavier-than-air objects could not fly.

On technical issue like this I sometimes refer to Roger Clark's website. He addresses the DoF/format issue at http://www.clarkvision.com/photoinfo/dof_myth/ and concludes the following:

Morgan,
If you were to use this lens on the 3 formats with the purpose of trying to get identical FoV and DoF, you'd be shooting from a different perspective and using different apertures. I think one would expect lenses to have at least a slightly different performance at different apertures. What I had in mind was testing say 3 different lenses of the same focal length, at the same aperture, and on the same format, to see if there was any differences in OoF effects. I suspect that there would be, depending on how critical you want to be about such differences.

For my own benefit, this afternoon I took a few test shots using my Canon 100-400 zoom with 40D and 5D. I was curious if such effects you refer to were noticeable with this lens. I used the 40D at 250mm and F8, and the 5D at 400mm and f13. I wasn't sure whether to use f13 or f11 on the 5D, but figured f13 was closer although not providing an exact DoF equivalence.

I did notice some slight differences in the degree of out-of-focusness behind the plane of focus in the two images. The 5D appeared to be showing very slightly shallower DoF. On the other hand, it could be that the 40D was not focussed as accurately as the 5D. Perhaps focussed just a little behind the focus point. I find it difficult to believe my 5D at f13 is sharper than the 40D at f8 with this zoom lens. It might be worth my redoing these tests using LiveView with the 40D.

For what they're worth, here are the results with 100% crops. No processing other than auto adjustment in ACR, default sharpening of 25, and downsizing the 5D shot to the same size as the 40D for ease of comparison.

Click to view attachment Click to view attachment Click to view attachment Click to view attachment Click to view attachment Click to view attachment

I took a few more shots later in the day using f11 instead of f13 on the 5D, but the lighting was poor so I bumped up the 5D ISO to 800 leaving the 40D at ISO 400. The OoF parts in these shots seem very close to me. Are you referring to differences of this magnitude?

Click to view attachment Click to view attachment

The focus point was supposed to be the power pole.

Thses are geniunely interesting tests

I am amazed you can be bothered with such an activity !

They would appear to completely disprove my theory because for what I can see the images look near enough the same

I think choosing a small aperture, focus near infinity you are maringalising the effect if it exists at all !

That doesnt mean a bigger sensor is not better (if your mission is shallow DOF)

At F13 you were able to open the small camera up to F8,

If however you were shooting with a H1 at f2.8 you could have to open up an APS camera about four stops - this is not possible becuase no such lens exists

As I said before some fast DSLR prime lenses on FF DSLRs can get less DOF than MF cameras

I still think I can 'see' MF pictures as being different - particularly 67 and larger formats like 54 - of course there are no practical digital solutions for these formats

SMM

I have sketed a graph of circle of confusion verses distance from point of focus for two lenses,

obviouly changing the aperture will change the curve steepness

I still dont exactly know if..

changing aperture can suitably affect those curves to equalize them

OR

changing aperture can do this but only if those apertures are available

I hate to admit I would like to see some maths wiki is a start..

I agree completely with that. As regards shallow DoF, people using P&S cameras are stuck. Thre's no way of achieving it, except in macro shots. Some 35mm lenses seem to be faster than their MF counterparts, ( Do you have any F1.2 or F1.4 MF lenses?), so one might think it's possible to get equally shallow DoF with FF 35mm.

However, my experience is that fast 35mm lenses are simply not sharp at full aperture. Is the 85/1.2 an exception? Some people claim it is sharp at full aperture, but I suspect they mean it's surprisingly good considering how wide f1.2 really is. I know for sure my own copy of the Canon 50/1.4 is not sharp at full aperture.

I would not expect the 50mm lens on a 40D at F1.4 could produce the same effect as a 110mm lens at F3.2 on a DB, yet theoretically they should have the same FoV and DoF.

Hahaha. Conflicting blather can be found everywhere on the net. It seems Ray and Morgan missed or chose to ignore that you said, "all other IQ issues are irrelevant," which would include apparent DOF. To answer your question regarding apparent magnification: yes.

And he was right that anti-matter exists.

Back to the matter at hand.
Using the CoC equation found on wikipedia (not the best source, I know):

C= abv(s2-s1)/s2 * F^2/(N(S1-F))

C is the CoC, S1 is the focus distance, S2 is the blurred obj distance, F is focal length and N is the F-stop.

Substitute values corresponding to the different formats and an arbitrary focus distance (if you'd like):

Full Frame
C= abv(s2-10000mm)/s2 * 160mm^2/(N(10000mm-160mm))

Crop Sensor
C= abv(s2-10000mm)/s2 * 100mm^2/(N(10000mm-100mm))

Assuming identical print sizes, you have to add a 1.6x coefficient to the crop sensor equation.

C= 1.6*abv(s2-10000mm)/s2 * 100mm^2/(N(10000mm-100mm))

Set the equations equal to one another and solve for N. Or at least try to solve. They don't cancel. This proves that you cannot achieve identical CoC curves in different formats by changing aperture.

"digital cameras with different sized sensors will produce images with identical depths-of-field"

This statement is correct, but over-simplified. Given a threshold CoC, you can indeed solve for aperture. This, however only means that only the 2 points at the threshold CoC will have identical blurring, and the object in sharp focus of course. These 3 distances do not make up the entire image, so while "DoF" is the identical using a strict definition, the blur throughout the image will not be identical.

I am not so sure about this having thought about it and seen Rays images - a bit of an about face - no bad thing

Some thoughts/questions (all ignoring IQ), ..

Simply 1)

-If I shrunk my hasselblad and lens in the wash why would the image change ?

-A d80 could with a 30mm could be described approximately as a shrunken blad with an 80

Mathsly 2)

The COC curves (and I have tried plotting some this weekend - anything to avoid client Post Production) are exponential or parabolic or some other curve of a cartesian nature if I remember the lingo right from shcool

The point of cartesian algebra is that given point a and point b one can extrapolate point C

(an example of the use of this algebra could be "how much do I need to invest now in my 6% bank account to have a million in 10 years?")

Therefore there is a consistant performance as one varies the input values, (Fstop focal lenth etc)

If there is consistent factors then one can tinker until the eqivilent Fstop and Flenght are found for a different chip size

One could therefore recreaste the look with a smaller chip if a fast enough lense is avaiable

I am trying to write a bit of software theat can do the graphs - it may take some time !

SMM

ps no cheek about not putting my blad in the spin dryer please

Dobson,
Thanks for taking the trouble to find the formula. Unfortunately, I'm neither a mathematician nor a physicist so cannot dispute the accuracy of the formula you've provided. For all I know, the formula itself could be an over-simplification but perhaps to a lesser degree than the over-simplification of DoF formulas.

However, even though I'm not a physicist, if someone were to state that the distance between point A and point B is a straight line, just to be contentious, I could argue that according to Einstein there are no straight lines. However, we all know that for terrestrial purposes such errors are usually insignificant. Nevertheless, one could dogmatically stick the view that in theory there are no straight lines.

Whilst this is an extreme example, I think it demonstrates a principle that the maths used for an application does not have to be (and should not be)cumbersomely more accurate than the application requires.

As I understand, in lens design and construction, there are all sorts of compromises and trade-offs that have to be made. One cannot get all aspects of performance equally good. If a lens is optimised for best performance at f2.8 it might be too difficult for it also to have best performance at f8, although performance at f8 might still be good. (There actually are 35mm lenses that are sharper at f2.8 than at f8).

As Morgan has implied, how relevant are such inaccuracies in practice regarding uniformity of DoF with different size sensors, considering my test results comparing 250mm at f8 with 400mm at f11 and f13?

By the way, in connection with the formula you provided, you mentioned multiplying one side by 1.6 for equal size prints. You are still thinking film . Nowadays, print size is dependent on pixel count, not sensor size.

I agree that there may be very little real-world relevance to this exercise. I originally did the calculations to improve my understanding of the concepts of sensor size, depth of field, and hyperfocal distance. The formula I started with is indeed an over-simplification; it is true only for symmetrical lenses with identical-sized entrance and exit pupils. The calculations for asymmetrical lenses are very complicated and specific to each lens; logic shows that it would be extremely unlikely for two independently-designed lenses to show similar characteristics in two different formats. Not necessarily impossible, though (and I really don't like math enough to figure that one out).

Boy this is getting complicated for the beginner's forum. Sure is interesting, though.



The formula I provided is independent of recording medium. This thought experiment will show why.

Let's say I just invented a new quantum sensor/film, one that recorded perfectly every photon that hits it. I have two cameras with this sensor, one with a 35mm frame and one at APS-C. These cameras also boast perfect lenses.

I am photographing 2 points of light at different distances from the film-plane. The cameras are set up in such a way that distances to these points and angle of view are identical. The cameras focus on the primary dot while the secondary dot is blurred out of focus. We match the apertures of the two lenses so the dot is blurred identically, lets say it projects a circle 1mm in diameter (the CoCs are equal).

Now I make a 1 meter wide print of each image. This doesn't matter if the image was enlarged optically or with a printer, the result is the same; two dots in each image, one infinitesimally small and one a blurred circle. The center points of each dot lie in exactly the same place for each print.

These blurred circles are not the same size, however. Because the APS-C sensor is smaller it had to be enlarged more to achieve the desired print size. This made the blurred circle in the APS print 1.6 times the size of the circle in the 35mm print.

As a result, the APS-C image appears more out of focus. Since our goal of the first experiment was to achieve identical images, and adjustment of CoC was required to make the 2 formats appear equivalent when printed (or viewed in any manner).

Crap - I just noticed a typo in my first proof.

The 1.6 coefficient should be added to the full frame side. I did the calculations correctly on paper, I just typed them wrong. Sorry if it caused any confusion.

Apparently he was, but in order to find out if he was right someone had to go looking for it. I went looking for these DoF differences that you assert should exist according to the maths. I haven't found them yet. Should I continue looking ?

Here are the rest of the crops comparing the 5D at f11, ISO 800, 400mm, and the 40D at f8, ISO 400, 250mm.

Click to view attachment Click to view attachment Click to view attachment Click to view attachment Click to view attachment

Now you have my curiosity. I would really like to see exactly how much difference, (if any),is visible between the two formats.

May I propose an experiment, I do not have cameras with different sized sensors.

The only issue I have with your previous tests is the difficulty I'm having measuring the blur spots.

The experiment I propose requires 3 point light sources and a dark background. Point lights could be small LED, or a hole cut in a backlit screen, or anything snall and bright. The blur circle of a point light will be very easy to measure in a print or on a monitor.

The experiment would be best performed at night (or in a dark studio) to eliminate unwanted distractions. The primary point light will be the object of focus, and closest to the camera. The secondary and tertiary points will be at different distances and out of focus.To accentuate any discrepancy, the secondary and tertiary points should be very far apart. Placing the tertiary point at infinity (such as a street light in the distance), will maximize this distance.

Then shoot the scene just as you did last time. Keeping the field of view identical and adjusting aperture to match depth-of-field. A wide aperture will make the blur spots bigger and easier/more-reliable to measure.

To measure the blur dots, resize the images to the same pixel size and measure (pixel diameters or mm). Or make identical-sized prints and measure. You will know you've done it right when either both secondary or both tertiary spots are exactly the same size (or maybe both pairs will match).

If you could perform this experiment, I would be extremely grateful. I've kicked back into engineer-mode and I am really interested in what the real-world results are. If you measure the distance to the points, we can even see how closely the calculations hold up in real-life.

Thanks,
Phillip

The curve of focus drop off is not striaght - we are talking about the route between A and B - and where is C if one contiues taking that route !

S

Phillip,
When I thought it might be worthwhile to check this matter, I chose the scene at the foot of my driveway because it contained a number of posts and readily identifiable objects at various distances receding towards infinity.

Instead of setting up an elaborate experiment, how about I just provide 800% enlargements of key features, such as the red reflectors on the white posts as they recede towards infinity? The differences you refer to appear to be so small that they might be completely obscured by other lens imperfections and quality variations. As a general principle, when trying to prove the existence of such subtle variations, as the size of blur in this case, one would have to do lots of experiments with lots of different lenses at different sets of equivalent apertures, in order to identify a trend. That task is not only too onerous for me, there just doesn't appear to be a practical issue of concern here as you can see from the following crops which, on my screen at least, represent a print size of about 30ftx20ft.

The crops from left to right are in order of their distance from the focus point. The ruler divisions at the top of each image should give you an indication of the size of the blur and any variation in the magnification of the objects I've selected.

Click to view attachment Click to view attachment Click to view attachment Click to view attachment Click to view attachment Click to view attachment

That might well be true. I was making a general point about theory accuracy. It could be argued that any theory that includes the concept of a straight line is inaccurate because straight lines do not exist. However, the inaccuracy is so small it can be irrelevant.

As I sit here in my studio, typing away to foreign people on the internet because I'm retired and have time to spare, I am ageing at a slightly greater rate than all those buzy people buzzing around from place to place in the city. I'm getting older than them, because I'm mostly stationary, at a faster rate of about one billionth of a second per year, but it doesn't worry me .

You might have noticed I sometimes revert to extreme examples to get a point across .

Im not getting all philosopical here,

When one puts $100 in the bank at 10% interest you have, (simply)

$110, at the end of the first year
$121 et the end of the second year
$132 at the end of the third year

and $259 at the end of the 10th year, (not $200)

A non linear increase - your interest is not $10 every year

while one can use a smaller principal investment with a higher intereste rate to get to the same wealth at year 10 the route will not be the same - or will it? - I dont think so

If we swap 'COC' for 'total money' and 'Duration of investment' for 'distance from focus point' a similar non linear graph is created



The question is can one create identical graphs by tinkering with chip size aperture and focal lenght

I am too stupid to be abe to answer this currently but I think with lenses the answer is it can be done

If the answer is yes then the answer is that one can recreate the same look across manys sensor sizes (excluding IQ factors)

S

Wait a minute! You've just delivered an oxymoron here. The same look, excluding IQ factors? What sort of look is that?

What I could do, I suppose, is go out onto the roadside and measure the various distances between key objects in the scene I shot, then feed the results into Phillip's formula to see if they corresponded with the visual results on my monitor with respect to CoC size. If friendly neighbours happened to enquire why I appeared to be measuring their boundaries, I could try to explain that I was just checking DoF variations and they would just assume I was crazy .

-your test is too close to infinity to show significant variance (if such exists)

-BY IQ I mean, resolution, lenses distortion etc things that dont exist in a simplified theoritical world - maybe not the right definition

S

I'm also curious how the math on this shakes out, but my guess is that the theoretical curves will be the same once you equalize for aperture and format size.

I agree with Sam that there is a definite difference in how DOF and falloff appear when shooting larger sized formats. I shoot 35 FF Digital with fast primes (35 f/1.4, 85 f/1.2), MF Film, and 4x5 film so I have some experience.

I think it is rooted in a couple things.

1) For a given print size, larger formats require less enlargement. This requires less lp/mm of the lens, which means that the MTF (microcontrast) will be better, assuming equal lens quality. As a rule, the higher the lp/mm measurement the lower the MTF (contrast). This means when you enlarge less or demand less lp/mm of the lens, the contrast at the plane of focus improves. Improved contrast = increased perceived sharpness (even if resolution is equal). Increased perceived sharpness at the plane of focus contributes to the sense of DOF and subject separation, even if it is not measured by a COC calculation.

2) Larger formats typically are higher resolution (more megapixels or larger/higher resolution scans). A higher resolution file will show more fine detail when viewed at 100%. When you increase print size (which is what you are doing when viewing at 100%) then you decrease DOF. There is a point at which you cannot enlarge a print further and reduce DOF - when you run out of resolution. To me this means that a higher resolution file has the potential for shallower DOF, when printed large enough.

As an example, compare a head and shoulders portrait taken Canon 8mp digital crop with a 35mm f/1.4 L lens wide open to a 39mp MFDB shot with an 80mm f/2.8 lens wide open (or whatever the equivalent aperture is). On a 5x7 or 8x10 print the DOF may appear the same. However, on a 16x20 print the crop DSLR will be soft at the plane of focus, but the MFDB will show very fine detail on the eyelashes, with the plane of focus softening progressively as you move away from the eyes. The MFDB shot obviously has less DOF at this print size, even though the apertures were "equalized". This holds true purely on the basis of the camera's capture resolution, even if both lenses are equally sharp at their chosen apertures.

3) Tonality, highlight/shadow detail, etc. All the other image quality intangibles help things out, too.

Since photographers typically view their photos at 100% (effectively the largest possible print size), it's no wonder that all the MFDB guys insist that there is something different about MF digital.

This is very true. I've made this point a number of times in discussions about DB versus high pixel count 35mm DSLRs such as the 1Ds3.

Even when the smaller format has the same pixel count, it cannot compete unless its lens is proportionally better than the larger format lens. If what's in focus is sharper in one image but what's OoF is about the same, the images will inevitably look different.

Even when comparing different formats which both use 35mm lenses, such as the 40D and 5D, results are likely to vary depending on which set of equivalent apertures are used. F8 on the 40D compared with f11 or f13 on the 5D is a fairly good match because we would expect the 35mm lens to be at least marginally sharper at f8 than at f11 or f13, and the 40D needs a sharper lens for the plane of focus to appear as sharp as it does on the 5D.

However, if we open up the apertures a bit and compare the 40D at f2.8 with the 5D at f4.5, then we could get some very different results.

This why I think it would be a very onerous exercise to try and prove the soundness or accuracy of Phillip's CoC formula by experimenting with various lenses and formats at different apertures. One would need to eliminate all the factors not related to format size that might skew the results. Pixel count would have to be the same on both formats, and lens resolution at the apertures chosen would need to be proportionally better with whatever lenses were used on the smaller format, better to the same degree that the pixel pitch is smaller, ie. system resolution, at the apertures chosen would ideally need to be the same for both formats.

DoF calculations are ultimately based on the simple thin-lens equation:

1/s + 1/d = 1/f, where

s = distance from lens to subject,
d = distance from lens to image,
f = focal length of lens.

The relevant dimensionless quantity which affects focal length scaling is the ratio f/s, which will be small for scenes with objects far from the lens. The rest of the calculation is ray tracing geometry which can be found on many sources on the web.

Someone already reproduced the formula for the blur diameter of an out of focus point object, derived from the lens equation. With a little rearrangement, the formula becomes:

C = (s2-s1)/(s1s2) * (f^2/N) / (1-f/s1), where

C is the blur diameter (circle of confusion),
s1 is the (sharp) focus distance,
s2 is the (blurred) object distance,
f is the lens focal length, and N is the lens f/stop.

If this formula did not contain the pesky factor (1-f/s1) at the end, scaling the f/stop with the square of the focal length would produce identical blur across the image for objects at any distance. The f/s1 correction makes this scaling imperfect. In practice, this will be evident only in photographs where f/s1 is relatively large, meaning focus is set fairly close to the lens.

So do you think the look is 'recreatable' or not ...?

S

What I might do tomorrow, if I have the time, is compare a close-up taken with my 10D and 17-55/2.8 zoom at 17mm and f3.5, with the same scene taken with my 5D and Sigma 15-30 zoom at 27mm and f5.6. I'm interested in comparing the sharpness and detail of these two cameras/lenses in any case.

I dont think youl see a lot if diffewrence wich such a wide lens

imo the differences, if there are any, show in the 50-135 region mainly and at f4 or wider

S

Too late, Morgan. Already done! 18mm at f3.5 on the 40D compared with 30mm at F5.6 on the 5D. As you predict, little difference.

Click to view attachment Click to view attachment Click to view attachment Click to view attachment Click to view attachment

Depth of focus will be similar on different formats, provided the f/stop is properly scaled to the lens focal length squared. The ratio (lens focal length) / (distance to subject) can be used to determine when this equivalence might be expected to break down. Using a 30mm lens, focusing on a subject as close as 300mm (roughly 12 inches), this ratio is still only 1/10. Such a small effect will likely be unobservable, or confused by other effects like diffraction. None of the images posted so far probe the regime where depth of focus differences might be seen.

Wow ... what an over-the-top thread, and entirely missing the original question!

To cut to long story short---Morgan is right. With different image formats, depth-of-field sharpness indeed does change differently across the depth of the field. The gradient will be smoother with larger formats. However, with 35-mm format vs. APS-C, the difference is hardly visible (at long range even less than at close range); it'll be more obvious with more dissimilar formats like APS-C vs. medium format, or 35-mm vs. 4×5", for instance.

But---that wasn't the question.

The question, as I understand it, was dealing with the contradiction of two statements. First statement: At a given distance, image magnification depends on focal length only ... i. e. a longer focal length will yield a larger magnification. Second statement: At a smaller image format, you'll need a shorter focal length for the same apparent magnification.

The solution to this contradiction lies in the word 'apparent'. Statement 1 is dealing with absolute, statement 2 with apparent, or relative, magnification. Relative to what? To the camera's frame size, of course. To make a given subject fill the frame at a given distance, for a frame half the size you'll need half the (absolute) magnification, and thus half the focal length. That's all.

So the answer to the original poster's question is: yes.

Of course, magnification is not the only factor determining an image's appearance (although it's the most significant). Depth-of-field is another. At the same distance and the same (relative) magnification---or in other words, the same angle of view---, the larger format will need a smaller aperture for the same DOF. However, even with the same DOF, the images' appearances still won't be perfectly identical because the gradient of the sharpness within DOF turning into blur beyond DOF will be different---that's what the carried-away discussion above mostly was about.



Eric, nobody here was talking about depth of focus; that's a completely different thing still. Please stop adding confusion to an already way-too-complex thread!

-- Olaf

Possibly for the begineners section but it the look thing is certainly important for me to clarify in my head (which I have not)

My experience is awful with cameras like the D80 and the D200 - horrible to use - now I have a d3 I am getting an cool look back to my DSLR images - but the res is not there compared to my blad,

If however I can convince myself that a 1Ds3 can compare to a 22mp Dback - it might be time to change and that is a big thing to think about

For a beginner laerning the whole game understanding the different formats and looks is part of the learning process too I would sugest - like learning DOF or Camera shake etc

S

[quote=Ray,Feb 28 2008, 08:18 AM]
Too late, Morgan. Already done! 18mm at f3.5 on the 40D compared with 30mm at F5.6 on the 5D. As you predict, little difference.


Hi!
Why on earth should anyone waste their time with this completely flawed "test"? Next time you could at least try to get sharp images. The 40D image is a horrible mess. Try using a tripod and mlu. Now it's impossible to tell where it's been focused. No point in comparing anything with these images.

Cheers,
J

If the 40D is like the D200 or D80 - its always a horrible mess - immpossible to tell where its been focussed - its the beacause of the small chip

Too much DOF declining at to shallow a rate !

Sam -

Why don't you bang out a quick test - don't you have a MFDB and a Nikon D3?

Just use a normalish lens, focus on a close subject and include middle distance and far distance background subjects (still life of some sort). Then downrez the MFDB shot to the same resolution as the D3 and compare the quantity (not quality) of the background blur of the middle distance and far distance subjects.

If I were ambitious I'd compare my 5D and 4x5 film, but it would take me a few weeks to have the film processed and scanned.

Depth of focus will be similar ...

Yes! It's over the top, but all my tests have shown that magnification is very similar. The issue has naturally gravitated towards the 'resolution' of that similar magnification. That's now become the issue, and it's the central issue which has flowed on from a beginner's question.

Morgan has suggested that comparisons will vary with focal length. I've tested the extremes of 250mm versus 400mm, and 18mm versus 30mm. I've now gone somewhere in the middle and compared 58mm on the 40D with 105mm on the 5D.
Surprisingly, the results are different. I didn't expect it and I was already composing in my mind some derogatory comments, before I saw the results.
In this range between 50mm and 105mm, with my lenses at least, the results tend to suggest the larger format produces slightly less DoF for the equivalent FL and F stop.

Has this anything to do with the mathematical formula presented as evidence for this effect? Who knows? Once again, it all boils down to 'knowing thy lenses'! Because my Canon 24-105 behaves like this is no guarantee that all lenses will behave similarly.

Of course I offer the results at maximum quality jpeg comprerssion. First the over all scene.

Click to view attachment

Then for dedicated pixel peepers, a 100% crop of of the 5D at f8 versus the 40D at f5, with the same zoom lens. The 5D is marginally sharper and more detailed, but not by much. There are all sorts of issues such as, is the 24-105 sharper at 105mm than at 58mm? (The FoV's are the same but the nomenclature may be erroneous.)

Click to view attachment

Let the results speak for themselves. The image headings should provide sufficient information. You can see from the following 2 images that DoF resolution does in deed vary noticeably with the these focal lengths (but not greatly). Perhaps Roger Clark should re-examine his views.

Click to view attachment Click to view attachment

The bottom left hand corner is slightly nearer than the plane of focus, yet it is still marginally sharper in the 40D shot, confirming that at these focal lengths, something different is happening. The larger format does indeed tend to produce a slightly shallower DoF.

Because Ive been extremely busy shooting real pictures for real clients

I may do now its the weekend


S

We're all busy, whether it's shooting real pictures for real clients or mowing the lawn.

The main point about my latest test comparing the 40D at 58mm and the 5D at 105mm, is that the 40D is actually slightly sharper a little bit in front of the focus point and also slightly sharper a little bit behind the focus point.

At the precise focus point the 5D is slightly sharper. This would be proof that there is no misfocussing.