EF 50/1.8 AF Experiment?
Wilba
shortcomings, 1) softness at wide apertures (OK from f/2.8), and 2) erratic
focus under difficult conditions (low light, shallow DOF).
Many people claim that 2) is a result of the crudeness of the cheap
focussing motor and electronics in the lens, that those components are not
able to provide the required accuracy and control of motion of the focus
ring.
But I wonder if 2) is actually a result of 1) - if the AF sensors have fuzzy
images to work with, how /could/ the system nail the focus in difficult
conditions?
It would be interesting to see what happens when the AF sensors have sharper
images to work with (e.g. at f/2.8 or f/4), but my 450D refuses to AF when
the DOF preview button is pressed, so I can't test that. External aperture
perhaps?
Any ideas for how these competing hypotheses could be tested? Is there a
consequence of either hypothesis that could be disproved empirically?
RichA
Why beat your head on a wall over this? Dump the garbage, buy a D5000
and a Nikon 50mm f1.8.
Chris Malcolm
The AF sensors pay no attention to the aperture at which you're going
to take the picture. They do their work before the lens is stopped
down. Their construction gives them an effective aperture of their
own. Often this is around f6. That means that when the largest
aperture of a lens is smaller than that they can't get enough light to
work properly. That's why generally speaking you can't make reflex
lenses autofocus, because for technical reasons their best compromise
aperture is often smaller than that, e.g. 500mm f8.
More expensive DSLRs will also have larger aperture AF sensors at the
central position, e.g. around f3, with which they'll be able to get
focus in lower light with lenses which with max apertures which open
that far. It also improves the focus on very fast lenses with
spherical aberration and corresponding aperture related focus drift,
such as the old spherical type of 50mm f1.4 lenses.
Since the DOF gets very thin indeed at wide apertures and close
portrait type distances, which is often what is going on in a dimly
lit interior, the slightest error in AF will leave the image blurred
at the point you wished to focus on, and sharp nearby. For example in
a portrait you might have focused on the eyes, and find that the eyes
aren't in focus, but the tip of the nose, or the ears, are. The reason
for that is often that when DoF gets so sharp it becomes smaller than
the small residual error in the AF of your camera, i.e. your camera
has a slight front or back focus in the AF sensor plane calibration
which is larger than the DoF at these wide apertures.
If you find a systematic error of this type in your camera than you
either must switch to manual focus, or compensate yourself, e.g. by
holding down focus on the eyes and then simply moving your head back
or forwards a few cm to take up the systematic error.
Usually the more expensive DSLRs have better AF sensors so they can
focus better in lower light. The wider aperture AF sensors are also
able to get a tighter focus for wide aperture low light work because
the AF sensor itself has effectively a shallower DoF. That will also
rein in some of the aperture related focus drift of wide aperture
spherical lenses.
The more expensive DSLRs are also sometimes able to read lens-specific
focus compensation factors from the lens, and use that to trim out
systematic errors in autofocus for that specific lens.
The most expensive DSLRs go one better than that. They have user
trimmable tables of focus compensation for specific lenses in order to
get better focus with the more awkward lenses in the more awkward
situation, in which the AF will have slight lens-specific systematic
focus errors.
--
Chris Malcolm
Ofnuts
> The Canon EF 50mm f/1.8 II "Nifty Fifty" has a reputation for two
> shortcomings, 1) softness at wide apertures (OK from f/2.8), and 2) erratic
> focus under difficult conditions (low light, shallow DOF).
>
> Many people claim that 2) is a result of the crudeness of the cheap
> focussing motor and electronics in the lens, that those components are not
> able to provide the required accuracy and control of motion of the focus
> ring.
>
> But I wonder if 2) is actually a result of 1) - if the AF sensors have fuzzy
> images to work with, how /could/ the system nail the focus in difficult
> conditions?
AF sensors work quite well with other lenses that are not that sharp, so
I doubt that the "softness" of the lens at f/1.8 is really a culprit.
> It would be interesting to see what happens when the AF sensors have sharper
> images to work with (e.g. at f/2.8 or f/4), but my 450D refuses to AF when
> the DOF preview button is pressed, so I can't test that. External aperture
> perhaps?
Read this excellent explanation of the DSLR phase-detection AF system:
<http://doug.kerr.home.att.net/pumpkin/Split_Prism.pdf>
To make it short, for AF sensors:
- they are designed to work with a minimum aperture (usually f/5.6 or
better)(and don't benefit from a bigger one)
- the more accurate you want the sensor, the wider the design aperture
has to be.
IIRC in the 450D most AF sensors require f/5.6 minimum, and the central
one is doubled with a "bigger" one that requires f/2.8 and is put in
action when the mounted lens reports that it has a maximum aperture of
f/2.8 or better. This allows a more accurate focus with these lenses,
which is required since the aperture of the lens can lead to very
shallow DoF (some entry level DSLR haven't got that second AF sensor and
cannot be efficiently used with lenses opening at f/2.8 or better).
> Any ideas for how these competing hypotheses could be tested? Is there a
> consequence of either hypothesis that could be disproved empirically?
The 450D is an entry-level camera, so don't expect miracles. With the
50/1.8 the accuracy of its AF system may be a bit pushed to its limits.
And make sure that you are using the central sensor for the the AF.
--
Bertrand, happy owner of a 450D and a 50/1.8
Alan Browne
Offhand, even if it is soft wide open, the AF should settle on the
"sharpest" slightly soft contrast, which on average should be sharp enough.
Could you post a simple target wide open?
In fact could you post the one on p18 of this document, shot at a 45 deg
angle. I'd like to compare it to the Minolta 50 f/1.7.
Alan Browne
> Wilba<use...@cutthisimago.com.au> wrote:
>> The Canon EF 50mm f/1.8 II "Nifty Fifty" has a reputation for two
>> shortcomings, 1) softness at wide apertures (OK from f/2.8), and 2) erratic
>> focus under difficult conditions (low light, shallow DOF).
>
>> Many people claim that 2) is a result of the crudeness of the cheap
>> focussing motor and electronics in the lens, that those components are not
>> able to provide the required accuracy and control of motion of the focus
>> ring.
>
>> But I wonder if 2) is actually a result of 1) - if the AF sensors have fuzzy
>> images to work with, how /could/ the system nail the focus in difficult
>> conditions?
>
>> It would be interesting to see what happens when the AF sensors have sharper
>> images to work with (e.g. at f/2.8 or f/4), but my 450D refuses to AF when
>> the DOF preview button is pressed, so I can't test that. External aperture
>> perhaps?
>
>> Any ideas for how these competing hypotheses could be tested? Is there a
>> consequence of either hypothesis that could be disproved empirically?
>
> The AF sensors pay no attention to the aperture at which you're going
> to take the picture. They do their work before the lens is stopped
> down.
It's clear that Wilba understands that. She wants to know if that if
the fact that it is doing the AF function while wide open is affecting
focus accuracy.
LOL!
<Munching popcorn, watching the usenet comedy show, while I enjoy using my
slightly slower but highly accurate contrast-detection focusing cameras.
Just what any intelligent person wants, phase-detection focusing that
focuses slightly faster but never accurately. All those shots, missed
forever.>
too fuckin' funny
This is such good free entertainment. Beats all the comedy routines on TV.
/me wonders if they ever realize what gigantic fools they continually make
of themselves daily ...
LOL!
Ofnuts
It just so happens that DSLR users are a bit more concerned about focus
than P&S users because:
1) they get to choose which part of the frame is in focus
2) out-of-focus parts aren't obscured by sensor noise or
diffraction-induced blur.
--
Bertrand
Alan Browne
We also like high SNR - in this NG. So it's best not to reply to the
brainless automaton aka the P&S troll.
Ray Fischer
I'd ask you to list all of those cameras that do "contrast-detection
focusing" that can shoot 8 frames per second, but you're just a lying
chickenshit little troll.
--
Ray Fischer
rfis...@sonic.net
Chris Malcolm
The simple answer, as the sentence above which you quote indicates, is
no. But the question was raised in the context of AF which becomes
unreliable in dim lighting at high apertures, and there are a number
of technical problems and issues here, some due to the properties of
spherical lenses (or incompletely aspherical ones :-), some due to the
way AF sensors work in different cameras, and some due to not uncommon
small AF calibration errors only apparent with very shallow DoF.
--
Chris Malcolm
RichA
Not yet, but it could be coming. The Panasonic G1/GH1/GF1 are very
quick when it comes to focusing, and we know that half of the focusing
battle is the lens motor and the weight of glass the motors have to
move. Once the mirror-less cameras hit the semi-pro region (Nikon
D300, Canon 7D) you should see some high frame rate (maybe not 8fps)
cameras appear.
Wilba
Thanks for your effort. Unfortunately, there is no answer to my question
within it. :- )
Wilba
> Wilba wrote:
>>
>> The Canon EF 50mm f/1.8 II "Nifty Fifty" has a reputation for two
>> shortcomings, 1) softness at wide apertures (OK from f/2.8), and 2)
>> erratic focus under difficult conditions (low light, shallow DOF).
>>
>> Many people claim that 2) is a result of the crudeness of the cheap
>> focussing motor and electronics in the lens, that those components are
>> not able to provide the required accuracy and control of motion of the
>> focus ring.
>>
>> But I wonder if 2) is actually a result of 1) - if the AF sensors have
>> fuzzy images to work with, how /could/ the system nail the focus in
>> difficult conditions?
>
> AF sensors work quite well with other lenses that are not that sharp, so I
> doubt that the "softness" of the lens at f/1.8 is really a culprit.
>
>> It would be interesting to see what happens when the AF sensors have
>> sharper images to work with (e.g. at f/2.8 or f/4), but my 450D refuses
>> to AF when the DOF preview button is pressed, so I can't test that.
>> External aperture perhaps?
>
> Read this excellent explanation of the DSLR phase-detection AF system:
>
> <http://doug.kerr.home.att.net/pumpkin/Split_Prism.pdf>
It's a good document which has influenced my thinking.
> To make it short, for AF sensors:
> - they are designed to work with a minimum aperture (usually f/5.6 or
> better)(and don't benefit from a bigger one)
> - the more accurate you want the sensor, the wider the design aperture has
> to be.
>
> IIRC in the 450D most AF sensors require f/5.6 minimum, and the central
> one is doubled with a "bigger" one that requires f/2.8 and is put in
> action when the mounted lens reports that it has a maximum aperture of
> f/2.8 or better. This allows a more accurate focus with these lenses,
> which is required since the aperture of the lens can lead to very shallow
> DoF (some entry level DSLR haven't got that second AF sensor and cannot be
> efficiently used with lenses opening at f/2.8 or better).
>
>> Any ideas for how these competing hypotheses could be tested? Is there
>> a consequence of either hypothesis that could be disproved empirically?
>
> The 450D is an entry-level camera, so don't expect miracles. With the
> 50/1.8 the accuracy of its AF system may be a bit pushed to its limits.
> And make sure that you are using the central sensor for the the AF.
Thanks for sharing your thoughts. Unfortunately, they don't answer my
question.
Wilba
Sure. These were shot a while back for comparison with PD AF shots to show
the calibration error I had at the time -
http://www.users.on.net/~alanw/Usenet/f1.8LiveViewFullsize.jpg (3.6MB)
http://www.users.on.net/~alanw/Usenet/f2.8LiveViewFullsize.jpg (3.4MB).
Wilba
>
> It's clear that Wilba understands that. She...
Do I sound like a woman?! I'll try to butch it up a bit. <hwock ptooey>
:- )
Alan Browne
Sorry dude. Wilba sounds feminine.
Now you're in for it, though.
Wilba
> Wilba wrote:
>> Alan Browne wrote:
>>>
>>> It's clear that Wilba understands that. She...
>>
>> Do I sound like a woman?! I'll try to butch it up a bit.<hwock ptooey>
>> :- )
>
> Sorry dude. Wilba sounds feminine.
It's an alternative (Australian-sounding) spelling of Wilbur.
> Now you're in for it, though.
Eh? What?
Alan Browne
Other than the general softness in 1.8, it looks like it focused
accurately or at worst a hair in front.
2.8 looks fine, perhaps a bit behind.
That softness may fall outside the AF hysteresis band.
I shot my tests further back (for other reasons). Will repeat
eventually closer in (with the 50mm @ f/1.7).
http://photo.net/photodb/photo?photo_id=10372792&size=lg
John McWilliams
Huh? I thought all Strines were called Curly (bald) or Blue (redhead).
WTF???
Good on ya, Wilba.. I think stateside we might spell it with an h at the
end, assuming your spelling is more how it's pronounced- or does that
spelling append a different meaning? Is your given name Wilbur?
--
john mcwilliams
Wilba
Those shots were focussed using contrast detection. Any error you might
perceive is of the order of the typical shot-to-shot variation.
> That softness may fall outside the AF hysteresis band.
More information please.
> I shot my tests further back (for other reasons). Will repeat eventually
> closer in (with the 50mm @ f/1.7).
> http://photo.net/photodb/photo?photo_id=10372792&size=lg
Yeah, hard to tell much with that much DOF.
Wilba
> Wilba wrote:
>> Alan Browne wrote:
>>> Wilba wrote:
>>>> Alan Browne wrote:
>>>>>
>>>>> It's clear that Wilba understands that. She...
>>>>
>>>> Do I sound like a woman?! I'll try to butch it up a bit.<hwock ptooey>
>>>> :- )
>>>
>>> Sorry dude. Wilba sounds feminine.
>>
>> It's an alternative (Australian-sounding) spelling of Wilbur.
>>
>>> Now you're in for it, though.
>>
>> Eh? What?
>
> Huh? I thought all Strines were called Curly (bald) or Blue (redhead).
>
> WTF???
Yeah, we're not all called Bruce.
> Good on ya, Wilba.. I think stateside we might spell it with an h at the
> end, assuming your spelling is more how it's pronounced-
Right, just like it looks - alba, amoeba, caramba, samba, tuba, etc.
> or does that spelling append a different meaning? Is your given name
> Wilbur?
It's a nickname from 1969, after the cartoon character Wilbur the Worm. It
stuck so I made it my own by changing the spelling. It matches my character
better than my given name. :- )
Chris Malcolm
Your qestion doesn't make sense because you don't understand enough
about what might be going on with your specific camera and your
specific focus difficulties. Your questions need to be revised in the
light of an improved understanding, part of which must come from doing
some experiments of your own to discover exactly what with your
specific camera and focus problems the most important problems
are. For example, your problems could be simply not enough light for
your camera's AF sensors to work properly. If so there's nothing you
can do with the way you use your lens to improve that. But there are
several ways of helping your AF to work better in poor light. On the
other hand your difficulties could be due to a small AF calibration
error which starts to matter under those circumstances with your
lens. If so there are a number of things you can do about that.
But it's not worth going into all the details of all the possibilities
until you have found out more about which particular limitation of the
several possibilities your camera and lens are coming up against.
--
Chris Malcolm
Alan Browne
In order to avoid hunting near focus, the system is designed with an
error band such that once the focus lies within that band the focus
action (control loop) can stop. Otherwise there could be hunting. For
most lenses, the sharpest focus falls somewhere inside the hysteresis
band. With your 50, the AF might determine the focus as in there or
slightly to either side making the focus error range larger. This due
to the softness wide open. The control loop must have several stop
(exit) conditions including 'best case' (to cover this case as well as
dim light).
There may also be more mechanical slop built in to the lens to allow for
a smaller motor to drive it. That's harder to pinpoint.
Wilba
I'm not having focus difficulties. The question is stated in the 5th
paragraph. All the rest is preamble.
Wilba
> Wilba wrote:
>>> That softness may fall outside the AF hysteresis band.
>>
>> More information please.
>
> In order to avoid hunting near focus, the system is designed with an error
> band such that once the focus lies within that band the focus action
> (control loop) can stop. Otherwise there could be hunting. For most
> lenses, the sharpest focus falls somewhere inside the hysteresis band.
> With your 50, the AF might determine the focus as in there or slightly to
> either side making the focus error range larger. This due to the softness
> wide open. The control loop must have several stop (exit) conditions
> including 'best case' (to cover this case as well as dim light).
Thanks, no worries with all that, I just hadn't heard the term before.
What you say fits with my thinking, the problem is finding a way to test the
theories.
> There may also be more mechanical slop built in to the lens to allow for a
> smaller motor to drive it. That's harder to pinpoint.
It sure feels and sounds like a crude sloppy mechanism with the 50/1.8, and
I'm sure that has a lot to do with people inferring that that causes the
focus performance they observe.
rwalker
<use...@CUTTHISimago.com.au> wrote:
snip
>
>Yeah, we're not all called Bruce.
>
Only the philosophers.
Paul Furman
> The Canon EF 50mm f/1.8 II "Nifty Fifty" has a reputation for two
> shortcomings, 1) softness at wide apertures (OK from f/2.8), and 2) erratic
> focus under difficult conditions (low light, shallow DOF).
>
> Many people claim that 2) is a result of the crudeness of the cheap
> focussing motor and electronics in the lens, that those components are not
> able to provide the required accuracy and control of motion of the focus
> ring.
>
> But I wonder if 2) is actually a result of 1) - if the AF sensors have fuzzy
> images to work with, how /could/ the system nail the focus in difficult
> conditions?
>
> It would be interesting to see what happens when the AF sensors have sharper
> images to work with (e.g. at f/2.8 or f/4), but my 450D refuses to AF when
> the DOF preview button is pressed, so I can't test that. External aperture
> perhaps?
>
> Any ideas for how these competing hypotheses could be tested? Is there a
> consequence of either hypothesis that could be disproved empirically?
If you are using the center AF point, sharpness shouldn't be a problem.
In fact the shallower DOF should improve accuracy.
One factor could be the phenomenon where focus shifts when stopping
down, but I suspect they've included corrections for that in the
firmware. There could also be issues with curvature of field, if you are
concerned about something off to the side, etc.
--
Paul Furman
www.edgehill.net
www.baynatives.com
all google groups messages filtered due to spam
Wilba
> Wilba wrote:
>>
>> The Canon EF 50mm f/1.8 II "Nifty Fifty" has a reputation for two
>> shortcomings, 1) softness at wide apertures (OK from f/2.8), and 2)
>> erratic focus under difficult conditions (low light, shallow DOF).
>>
>> Many people claim that 2) is a result of the crudeness of the cheap
>> focussing motor and electronics in the lens, that those components are
>> not able to provide the required accuracy and control of motion of the
>> focus ring.
>>
>> But I wonder if 2) is actually a result of 1) - if the AF sensors have
>> fuzzy images to work with, how /could/ the system nail the focus in
>> difficult conditions?
>>
>> It would be interesting to see what happens when the AF sensors have
>> sharper images to work with (e.g. at f/2.8 or f/4), but my 450D refuses
>> to AF when the DOF preview button is pressed, so I can't test that.
>> External aperture perhaps?
>>
>> Any ideas for how these competing hypotheses could be tested? Is there a
>> consequence of either hypothesis that could be disproved empirically?
>
> If you are using the center AF point, sharpness shouldn't be a problem.
I can interpret that to mean several different things. What do you mean
exactly?
> In fact the shallower DOF should improve accuracy.
I think that's part of why the "crude mechanism" theory is so popular, but I
can't think of any way to test it. I'm not interested in theory or
speculation except that which leads to an experiment which proves something.
> One factor could be the phenomenon where focus shifts when stopping down,
> but I suspect they've included corrections for that in the firmware.
If that were a factor I would expect to see a consistent mis-focus, but
that's not what I get.
> There could also be issues with curvature of field, if you are concerned
> about something off to the side, etc.
I'm not. :- )
Paul Furman
> Paul Furman wrote:
>> Wilba wrote:
>>> The Canon EF 50mm f/1.8 II "Nifty Fifty" has a reputation for two
>>> shortcomings, 1) softness at wide apertures (OK from f/2.8), and 2)
>>> erratic focus under difficult conditions (low light, shallow DOF).
>>>
>>> Many people claim that 2) is a result of the crudeness of the cheap
>>> focussing motor and electronics in the lens, that those components are
>>> not able to provide the required accuracy and control of motion of the
>>> focus ring.
>>>
>>> But I wonder if 2) is actually a result of 1) - if the AF sensors have
>>> fuzzy images to work with, how /could/ the system nail the focus in
>>> difficult conditions?
>>>
>>> It would be interesting to see what happens when the AF sensors have
>>> sharper images to work with (e.g. at f/2.8 or f/4), but my 450D refuses
>>> to AF when the DOF preview button is pressed, so I can't test that.
>>> External aperture perhaps?
>>>
>>> Any ideas for how these competing hypotheses could be tested? Is there a
>>> consequence of either hypothesis that could be disproved empirically?
>> If you are using the center AF point, sharpness shouldn't be a problem.
>
> I can interpret that to mean several different things. What do you mean
> exactly?
Sharpness shouldn't be a problem in the center. I was doing some lens
tests with macro stacking software and just for curiosity tried some
really fast primes, much closer than they were designed for, not
reversed or anything... and was amazed that the center performance was
excellent wide open. That quickly degraded to 'hideous' outside the
middle third though.
http://www.flickr.com/photos/edgehill/4212576776/sizes/o/in/set-72157603231101723/
If you have extension tubes or a bellows, try it at extreme closeup.
That's a good way to make flaws much more obvious. It may not apply to
normal operation though.
>> In fact the shallower DOF should improve accuracy.
>
> I think that's part of why the "crude mechanism" theory is so popular, but I
> can't think of any way to test it. I'm not interested in theory or
> speculation except that which leads to an experiment which proves something.
Lots of ways to test. Do a series at the same spot but defocus manually,
first several starting from infinity, then several from closest focus,
then from just slightly off in each direction. Do that for something up
close then something 30 feet away.
You could try your stopped down theory by putting a piece of black paper
with a hole punched on the front, but might try testing first that it
doesn't make things less sharp due to non-optimal aperture placement.
Put the paper aperture in the right place in back and the lens becomes
telecentric, meaning light rays travel straight out the front and the
field of view is no wider than the front element.
>> One factor could be the phenomenon where focus shifts when stopping down,
>> but I suspect they've included corrections for that in the firmware.
>
> If that were a factor I would expect to see a consistent mis-focus, but
> that's not what I get.
The pattern could be complex though... depending how much you are
stopped down, maybe depending if it's close up or infinity...
>> There could also be issues with curvature of field, if you are concerned
>> about something off to the side, etc.
>
> I'm not. :- )
>
>
Wilba
No, it is a problem for the nifty fifty. You can see it in the shots I
posted a while back -
http://www.users.on.net/~alanw/Usenet/f1.8LiveViewFullsize.jpg (3.6MB)
http://www.users.on.net/~alanw/Usenet/f2.8LiveViewFullsize.jpg (3.4MB)
> If you have extension tubes or a bellows, try it at extreme closeup.
> That's a good way to make flaws much more obvious. It may not apply to
> normal operation though.
>
>>> In fact the shallower DOF should improve accuracy.
>>
>> I think that's part of why the "crude mechanism" theory is so popular,
>> but I can't think of any way to test it. I'm not interested in theory or
>> speculation except that which leads to an experiment which proves
>> something.
>
> Lots of ways to test. Do a series at the same spot but defocus manually,
> first several starting from infinity, then several from closest focus,
> then from just slightly off in each direction. Do that for something up
> close then something 30 feet away.
What consequence of which theory would this test, and how?
> You could try your stopped down theory by putting a piece of black paper
> with a hole punched on the front, but might try testing first that it
> doesn't make things less sharp due to non-optimal aperture placement.
How could I do that?
> Put the paper aperture in the right place in back and the lens becomes
> telecentric, meaning light rays travel straight out the front and the
> field of view is no wider than the front element.
You lost me somewhere there. :- )
Put the external aperture _behind_ the lens? And light comes out the
_front_?! I don't get it. :- )
>>> One factor could be the phenomenon where focus shifts when stopping
>>> down, but I suspect they've included corrections for that in the
>>> firmware.
>>
>> If that were a factor I would expect to see a consistent mis-focus, but
>> that's not what I get.
>
> The pattern could be complex though... depending how much you are stopped
> down, maybe depending if it's close up or infinity...
Yeah, let's just stick with what happens wide open.
Paul Furman
Huh? "No" what? It's not sharp in the center?
> You can see it in the shots I posted a while back -
> http://www.users.on.net/~alanw/Usenet/f1.8LiveViewFullsize.jpg (3.6MB)
> http://www.users.on.net/~alanw/Usenet/f2.8LiveViewFullsize.jpg (3.4MB)
>>
>> If you have extension tubes or a bellows, try it at extreme closeup.
>> That's a good way to make flaws much more obvious. It may not apply to
>> normal operation though.
This will tell you if it's sharp in the center.
>>>> In fact the shallower DOF should improve accuracy.
>>>
>>> I think that's part of why the "crude mechanism" theory is so popular,
>>> but I can't think of any way to test it. I'm not interested in theory or
>>> speculation except that which leads to an experiment which proves
>>> something.
>> Lots of ways to test. Do a series at the same spot but defocus manually,
>> first several starting from infinity, then several from closest focus,
>> then from just slightly off in each direction. Do that for something up
>> close then something 30 feet away.
>
> What consequence of which theory would this test, and how?
Testing the 'slop' theory. To see if it stops at different positions on
repeated runs of the same movement or different movements.
>> You could try your stopped down theory by putting a piece of black paper
>> with a hole punched on the front, but might try testing first that it
>> doesn't make things less sharp due to non-optimal aperture placement.
>
> How could I do that?
Just side by side comparison shots to see if the new aperture makes
things softer.
>> Put the paper aperture in the right place in back and the lens becomes
>> telecentric, meaning light rays travel straight out the front and the
>> field of view is no wider than the front element.
>
> You lost me somewhere there. :- )
>
> Put the external aperture _behind_ the lens? And light comes out the
> _front_?! I don't get it. :- )
Heh, sorry, off on a tangent. The idea is that adding an aperture like
that can do weird things, so maybe not sharper than stopping down the
lens' own aperture and who knows, maybe softer than wide open.
>>>> One factor could be the phenomenon where focus shifts when stopping
>>>> down, but I suspect they've included corrections for that in the
>>>> firmware.
>>>
>>> If that were a factor I would expect to see a consistent mis-focus, but
>>> that's not what I get.
>>
>> The pattern could be complex though... depending how much you are stopped
>> down, maybe depending if it's close up or infinity...
>
> Yeah, let's just stick with what happens wide open.
But you wanted to test if things improved stopped down. If I understand
correctly, the view in the optical viewfinder and presumably what the AF
sensors see, is already stopped down to maybe f/2.8; it doesn't benefit
from faster lenses, due to obstructions I guess. So there is already
some potential for focus shift there... but as you say, that should be
consistent, so again this points toward slop in the mechanism. If you
tested various final taking apertures, at least that's some kind of data
pointing to whether that might be playing a role.
The simplest test is that focus chart manually focused then stop down to
various amounts & see if the focus shifts. It will get wider DOF but
might not be centered if you count from each side. Then see if the AF
places the focus differently from the manual fixed focus test at
different apertures. If there is focus shift and the camera firmware is
correcting for it, there should be little differences in where the
camera decides to focus.
PS I still don't get how the AF calibration thing works on the cameras
that have that. I can see one master set screw for coordinating the
distance of the AF sensors to match the distance to the picture sensor
and or viewfinder ground glass but calibrating differently for different
lenses would seem to me to have to be relying on this idea of focus
shift when stopping down. If so, that's going to have to be checked at
various apertures to establish a curve for the correction, not just one
test shot. Does that make sense?
Wilba
Correct. You can see it in the shots I posted a while back -
http://www.users.on.net/~alanw/Usenet/f1.8LiveViewFullsize.jpg (3.6MB)
http://www.users.on.net/~alanw/Usenet/f2.8LiveViewFullsize.jpg (3.4MB)
>>>>> In fact the shallower DOF should improve accuracy.
>>>>
>>>> I think that's part of why the "crude mechanism" theory is so popular,
>>>> but I can't think of any way to test it. I'm not interested in theory
>>>> or speculation except that which leads to an experiment which proves
>>>> something.
>>>
>>> Lots of ways to test. Do a series at the same spot but defocus manually,
>>> first several starting from infinity, then several from closest focus,
>>> then from just slightly off in each direction. Do that for something up
>>> close then something 30 feet away.
>>
>> What consequence of which theory would this test, and how?
>
> Testing the 'slop' theory. To see if it stops at different positions on
> repeated runs of the same movement or different movements.
From previous tests for other purposes, I can say yes, it does exactly that.
But the problem is, that behaviour is just as well explained by the "soft at
f/1.8" theory - if the AF sensor doesn't have sharp fine detail to compare,
I would expect it to show that small-scale variability.
>>> You could try your stopped down theory by putting a piece of black paper
>>> with a hole punched on the front, but might try testing first that it
>>> doesn't make things less sharp due to non-optimal aperture placement.
>>
>> How could I do that?
>
> Just side by side comparison shots to see if the new aperture makes things
> softer.
Ah, that's what you mean, OK, got it. :- )
Yes, I might have a play with some external apertures today. We know that
f/2.8 gives a sharp image, so I will find an external aperture which gives
the same exposure, and if the image is as sharp with the lens on f/1.8, then
I can try focussing through it. Alternatively, if the "equal exposure"
method doesn't work, I could try to find an external aperture which gives
the same or more sharpness as the naked lens at f/2.8.
>>> Put the paper aperture in the right place in back and the lens becomes
>>> telecentric, meaning light rays travel straight out the front and the
>>> field of view is no wider than the front element.
>>
>> You lost me somewhere there. :- )
>>
>> Put the external aperture _behind_ the lens? And light comes out the
>> _front_?! I don't get it. :- )
>
> Heh, sorry, off on a tangent. The idea is that adding an aperture like
> that can do weird things, so maybe not sharper than stopping down the
> lens' own aperture and who knows, maybe softer than wide open.
OK, I'll try to erase all that from my mind.
>>>>> One factor could be the phenomenon where focus shifts when stopping
>>>>> down, but I suspect they've included corrections for that in the
>>>>> firmware.
>>>>
>>>> If that were a factor I would expect to see a consistent mis-focus, but
>>>> that's not what I get.
>>>
>>> The pattern could be complex though... depending how much you are
>>> stopped down, maybe depending if it's close up or infinity...
>>
>> Yeah, let's just stick with what happens wide open.
>
> But you wanted to test if things improved stopped down.
Just to make sure no-one misunderstands that statement, I would like to be
able to auto-focus with the aperture stopped down.
> If I understand correctly, the view in the optical viewfinder and
> presumably what the AF sensors see, is already stopped down
> to maybe f/2.8; it doesn't benefit from faster lenses, due to
> obstructions I guess.
Absolutely not. The aperture is wide open except for the duration of the
shutter opening. To see what I mean, in aperture priority mode press the DOF
preview button while you change the aperture.
> So there is already some potential for focus shift there...
I might come back to focus shift later if nothing else works. :- )
> PS I still don't get how the AF calibration thing works on the cameras
> that have that. I can see one master set screw for coordinating the
> distance of the AF sensors to match the distance to the picture sensor and
> or viewfinder ground glass but calibrating differently for different
> lenses would seem to me to have to be relying on this idea of focus shift
> when stopping down. If so, that's going to have to be checked at various
> apertures to establish a curve for the correction, not just one test shot.
> Does that make sense?
It does. There is a lot of mystery in this area.
Doug McDonald
>
> From previous tests for other purposes, I can say yes, it does exactly that.
> But the problem is, that behaviour is just as well explained by the "soft at
> f/1.8" theory - if the AF sensor doesn't have sharp fine detail to compare,
> I would expect it to show that small-scale variability.
>
Another test, which I did.
You take pictures using manual focus (at f/1.8) with the focus
indicator (the little dot at lower right in my 30D) as the criterion
of correct focus. Try to get it centered in the middle of
the "in focus is indicated" range. Make several tries. If this
gives better focus than real autofocus, something is wrong with the
"focus movement prediction" system.
My results say that the indicator is better than real autofocus,
but not by a lot. This is using only the center spot on text
7 feet away.
In focus this lens is quite sharp at f/1.8, in the center that is.
Doug McDonald
Wilba
Ah, this is exactly the kind of idea I need! That experiment attempts to
separate the detection (AF sensor) and actuation (lens motor) functions
within the system. I'll have a play with it today. Thanks a bunch.
> In focus this lens is quite sharp at f/1.8, in the center that is.
Not in my experience. If the lens was as sharp in the centre at f/1.8 as it
is at f/2.8, then I would see at f/1.8 a narrow band of decent sharpness
_somewhere_ in images of slanted subjects with fine textured detail, but I
don't, and not just in my own images.
Wilba
Well, it sure has been an interesting day. :- )
Here are some preliminary results with a 45 degree target. I need to repeat
with a different target before I would state the following with confidence,
and it could be quite different for other units depending on their
calibration. But anyway...
Sharpness in the very centre at f/1.8 appeared fine, practically
indistinguishable from f/2.8 or f/4. In previous tests I might not have
looked hard enough at the very centre of the image, and may have been misled
by other sources of variation. So that's suggests a disproof of my "soft at
f/1.8" theory.
I found that my phase detect AF sensor has sidedness. If I start with the
lens focussed closer than the subject, the results are uniformly excellent,
whether autofocussing or manually focussing using the AF confirmation (as
Doug described above).
If I start with the lens focussed behind the subject, and I manually focus
using the AF confirmation, focus is always off by the same tiny amount (one
click towards infinity in the EOS Utility will bring it into optimal focus).
With initial focus behind the subject and PD autofocus, about seven shots
out of ten are out by the same one click as the manual focus, and the rest
are optimal, like when starting from the nearside. I assume that the good
ones come about from the lens overshooting and then the system corrects
towards infinity (so it ultimately approaches focus from the nearside).
With an external aperture which gives similar exposures to f/2.8 (with the
lens at f/1.8), focus is still excellent starting from the near side, and it
improves the performance when starting from the far side (but still not
optimal).
Those results, and the fact that the lens is able to focus perfectly via
contrast detect, suggest to me that the "crude and sloppy mechanism" theory
is bogus. I suspect the truth is about the performance of the PD AF sensor
with this lens, specifically how much latitude it has for confirming a
focus. I noticed when manually focussing with a macro rail that I could move
the camera quite some (micro) distance and still get a focus confirmation
beep.
Wilba
> Wilba wrote:
>> Doug McDonald wrote:
>>>
>>> You take pictures using manual focus (at f/1.8) with the focus
>>> indicator (the little dot at lower right in my 30D) as the criterion
>>> of correct focus. Try to get it centered in the middle of
>>> the "in focus is indicated" range. Make several tries. If this
>>> gives better focus than real autofocus, something is wrong with the
>>> "focus movement prediction" system.
>>>
>>> My results say that the indicator is better than real autofocus,
>>> but not by a lot. This is using only the center spot on text
>>> 7 feet away.
>>
>> Ah, this is exactly the kind of idea I need! That experiment attempts to
>> separate the detection (AF sensor) and actuation (lens motor) functions
>> within the system. I'll have a play with it today. Thanks a bunch.
>
I've been thinking a lot about the width of the "beep band" and what it
means, so I measured it with a macro rail. With the lens at its closest
focus (430mm from the sensor), the far beep is 3.0mm from the near beep. A
conventional DOF calculation gives around 4.3mm, and a focus from the
nearside appears to put the subject bang in the middle of that. So that
means a farside focus would put the subject something like 0.8mm outside the
DOF, which sounds about right.
I also tried it at that distance with an EF-S 18-55 at about 50mm (f/5.6).
The beep band is 12mm, the DOF is 13mm, and where you start from makes very
little difference to the focus achieved (although from the nearside is best
again). I wish I had another wide-aperture FFL lens to compare.
Any thoughts about why the 50/1.8 figures make sense (farside focus... 0.8mm
outside the DOF), but it doesn't work like that for the 18-55?
David J Taylor
news:00c4cc60$0$15581$c3e...@news.astraweb.com...
[]
> Any thoughts about why the 50/1.8 figures make sense (farside focus...
> 0.8mm outside the DOF), but it doesn't work like that for the 18-55?
As others have said, the focal point will shift with f/number, and the
focus sensors may only accept the smaller cone of rays rather than the
full f/1.8 cone.
Cheers,
David
Paul Furman
|4.3mm| theoretical
->|3mm|<- measured
.8mm
<- 430mm -> [camera]
> I also tried it at that distance with an EF-S 18-55 at about 50mm (f/5.6).
> The beep band is 12mm, the DOF is 13mm, and where you start from makes very
> little difference to the focus achieved (although from the nearside is best
> again). I wish I had another wide-aperture FFL lens to compare.
|13mm| theoretical
->|12 |<- measured
> Any thoughts about why the 50/1.8 figures make sense (farside focus... 0.8mm
> outside the DOF), but it doesn't work like that for the 18-55?
Did I get those diagrams right? I'm really not following what 'makes
sense' about the f/1.8 error. Are these distances confirmed with photo
results in a focus test pattern? For one thing, I wouldn't bother with
calculated DOF, especially for close up.
I would predict results kinda like this due to the AF sensors getting a
restricted aperture view of things with a bit more DOF:
->| 4mm |<- where it beeps
|3mm| confirmed test chart dof
Then if you were taking the final shot at f/8, focus shift as a result
of stopping down could produce something like this:
->| 4mm |<- where it beeps
| 30mm | confirmed test chart dof
or, since you are getting better results from in front:
->| 4mm |<- where it beeps
| 30mm | confirmed test chart dof
But without stopping down, I can't explain the difference from behind or
in front... however, that's at least a useful conclusion from all this:
if you start focused from in front, better focus can be achieved. So
maybe that's the only relevant lesson from all this, and it's a good one.
OK, here's a possible explanation for the front/back difference. The
character of the out of focus area in front & back is indeed different.
In terms of bokeh, the background is usually smoother and the foreground
harsh. There are a few specialized 'defocus control' lenses which have
an extra ring to change this relationship. I'm not sure quite how that
would effect the AF confirm but it could be a factor.
Wilba
> Wilba wrote:
>>
>> Any thoughts about why the 50/1.8 figures make sense (farside focus...
>> 0.8mm outside the DOF), but it doesn't work like that for the 18-55?
>
> As others have said, the focal point will shift with f/number,
How is that a factor when shooting at widest apertures?
> and the focus sensors may only accept the smaller cone of rays rather
> than the full f/1.8 cone.
I don't understand any of that. :- )
Wilba
Um, I don't understand your diagrams. :- ) Let me try again.
With f/1.8 and a subject in optimal focus at 430mm, a conventional DOF
calculation gives 4.3mm, so lets assume 2.2mm each side (I know we can't
rely on this figure, but it's not a million miles from what I see). Starting
with the camera at 425mm and moving it away from the subject until it beeps
focus confirmation, we get optimal focus again at 430mm. Starting with the
camera at 435mm and moving in, it beeps at 433mm, which is 0.8mm outside the
theoretical DOF and is clearly out of focus.
> I'm really not following what 'makes sense' about the f/1.8 error.
What makes sense is that a focus 3mm beyond the optimal camera to subject
distance would put the subject outside the DOF (3.0 - 4.3/2 = 0.8).
> Are these distances confirmed with photo results in a focus test
> pattern?
Yes, the above is just a mathematical expression of what I'm seeing from
doing the beep tests. When I said, "a farside focus would put the subject
something like 0.8mm outside the DOF, which sounds about right", I meant
that the numbers agree with what I'm seeing - the subject is definitely a
small distance outside the DOF in applicable images.
What's interesting is that the beep tests give the same images as AF. That
tells me that the focus errors that people see with this lens when focussing
are probably not due to a lack of precision in the lens's focus mechanism
(as is commonly believed), but may be due to the fact that the focus sensor
will confirm a focus when the subject is clearly not in focus. That happens
100% of the time when I focus by moving my camera, and something in the
order of 70% of the time when I AF, with the subject further away than where
the lens is initially focussed. (YMMV with a differently calibrated body and
lens.)
> For one thing, I wouldn't bother with calculated DOF, especially for close
> up.
Right, that's just to have some numbers that fit with what I'm seeing.
> I would predict results kinda like this due to the AF sensors getting a
> restricted aperture view of things with a bit more DOF:
I don't follow that, sorry. (All tests and calculations were done at widest
aperture - f/1.8.)
> But without stopping down, I can't explain the difference from behind or
> in front... however, that's at least a useful conclusion from all this: if
> you start focused from in front, better focus can be achieved. So maybe
> that's the only relevant lesson from all this, and it's a good one.
Yes. :- ) (Again, results may be quite different with other combinations
of bodies and lenses.)
> OK, here's a possible explanation for the front/back difference. The
> character of the out of focus area in front & back is indeed different. In
> terms of bokeh, the background is usually smoother and the foreground
> harsh. There are a few specialized 'defocus control' lenses which have an
> extra ring to change this relationship. I'm not sure quite how that would
> effect the AF confirm but it could be a factor.
Very interesting idea, thanks very much. This -
http://www.stacken.kth.se/~maxz/defocuscontrol/ (and the links in it to
http://www.luminous-landscape.com/essays/bokeh.shtml) - gave me enough to
get what you're saying.
That brings us back again to the notion that it is the optical performance
of the 50/1.8 that causes the erratic focus from the farside, not the
perceived crudeness of the focus mechanism.
What's different for me now is that I believe the erratic farside focus is
not about the sharpness wide open, but more likely about the different
"character of the out of focus area" on the nearside and farside of the
plane of focus.
Wilba
>
> Um, I don't understand your diagrams. :- ) Let me try again.
>
> With f/1.8 and a subject in optimal focus at 430mm, a conventional DOF
> calculation gives 4.3mm, so lets assume 2.2mm each side (I know we can't
> rely on this figure, but it's not a million miles from what I see).
> Starting with the camera at 425mm and moving it away from the subject
> until it beeps focus confirmation, we get optimal focus again at 430mm.
> Starting with the camera at 435mm and moving in, it beeps at 433mm, which
> is 0.8mm outside the theoretical DOF and is clearly out of focus.
Try viewing this text with a fixed width font (e.g. copy into Notepad) -
|< 430mm >| sensor to subject distance for optimal
focus
| |< 2.2mm >|< 2.2mm >| theoretical
DOF
|< 430mm >| nearside
focus
|< 433mm >| farside
focus
| |< 3.0mm >| the "beep
band"
| |<0.8mm>| outside the
DOF
Chris Malcolm
> David J Taylor wrote:
>> Wilba wrote:
>>>
>>> Any thoughts about why the 50/1.8 figures make sense (farside focus...
>>> 0.8mm outside the DOF), but it doesn't work like that for the 18-55?
>>
>> As others have said, the focal point will shift with f/number,
> How is that a factor when shooting at widest apertures?
Because with a specific lens and AF sensor the focusing takes place
always at the effective aperture of the AF sensor, whereas the
aperture with which the photograph is taken can vary. Many cameras
have a max AF sensor aperture of around f6. Some go down as far as
f2.8. Hence if you're using a lens at f1.8, and it happens to be a
spherical lens design with aperture related focus drift (as many of
the golden oldie 50mms are), then this is an important factor.
>> and the focus sensors may only accept the smaller cone of rays rather
>> than the full f/1.8 cone.
> I don't understand any of that. :- )
It means the AF sensor has a smaller effective aperture than is being
used to take the photograph. If you want the best results from a wide
aperture 50mm of spherical design then you need to understand
this. Alternatively you could buy a modern aspherical design. One of
the great benefits of modern technology is that it enables people who
don't understand what they're doing to do it well.
--
Chris Malcolm
Paul Furman
Your follow-up post diagram makes sense.
You have the camera & subject reversed from mine and enough room to show
things with more detail.
I see my diagrams were mistaken so ignore.
> With f/1.8 and a subject in optimal focus at 430mm, a conventional DOF
> calculation gives 4.3mm, so lets assume 2.2mm each side (I know we can't
> rely on this figure, but it's not a million miles from what I see). Starting
> with the camera at 425mm and moving it away from the subject until it beeps
> focus confirmation, we get optimal focus again at 430mm. Starting with the
> camera at 435mm and moving in, it beeps at 433mm, which is 0.8mm outside the
> theoretical DOF and is clearly out of focus.
>
>> I'm really not following what 'makes sense' about the f/1.8 error.
>
> What makes sense is that a focus 3mm beyond the optimal camera to subject
> distance would put the subject outside the DOF (3.0 - 4.3/2 = 0.8).
>
>> Are these distances confirmed with photo results in a focus test
>> pattern?
>
> Yes, the above is just a mathematical expression of what I'm seeing from
> doing the beep tests. When I said, "a farside focus would put the subject
> something like 0.8mm outside the DOF, which sounds about right", I meant
> that the numbers agree with what I'm seeing - the subject is definitely a
> small distance outside the DOF in applicable images.
>
> What's interesting is that the beep tests give the same images as AF. That
> tells me that the focus errors that people see with this lens when focussing
> are probably not due to a lack of precision in the lens's focus mechanism
> (as is commonly believed), but may be due to the fact that the focus sensor
> will confirm a focus when the subject is clearly not in focus.
Excellent.
Yeah, that's the only thing that makes sense, though it's still a bit
sketchy. In the middle of this page:
http://www.stacken.kth.se/~maxz/defocuscontrol/
the crops of dark blurry blobs show harsh rimmed OOF circles and perhaps
the AF sensor is grabbing those edges. I don't think that's quite right
but the best I can come up with.
Paul Furman
> Wilba <use...@cutthisimago.com.au> wrote:
>> David J Taylor wrote:
>>> Wilba wrote:
>>>> Any thoughts about why the 50/1.8 figures make sense (farside focus...
>>>> 0.8mm outside the DOF), but it doesn't work like that for the 18-55?
>>> As others have said, the focal point will shift with f/number,
>
>> How is that a factor when shooting at widest apertures?
>
> Because with a specific lens and AF sensor the focusing takes place
> always at the effective aperture of the AF sensor, whereas the
> aperture with which the photograph is taken can vary. Many cameras
> have a max AF sensor aperture of around f6. Some go down as far as
> f2.8. Hence if you're using a lens at f1.8, and it happens to be a
> spherical lens design with aperture related focus drift (as many of
> the golden oldie 50mms are), then this is an important factor.
Yeah, this could still be the explanation. Which direction does the
focus shift when stopping down? -that would provide a clue.
http://diglloyd.com/diglloyd/free/FocusShift/index.html
"With every lens I�ve tested to date, the focus moves farther away. For
example, if focus at f/1.4 is centered at 1.00 meters, then by f/2.8 it
might now be centered at 1.02 meters."
On average (ignoring front/back approach), Wilba's camera is focusing
closer (when 'stopped down' through obstructions to the AF sensor), then
opening fully to the sensor for taking the pic; the actual point in
focus is further back. That's the opposite of what this effect would
suggest.
Theoretically the camera could correct for this given that it has the
lens data from the cpu connection. It is my understanding that Nikons
correct exposure for vignetting when stopping down and this is the
rational for not allowing entry level Nikon bodies to meter without a
cpu chip telling the camera which lens it has, where the pro models
allow inputing the lens specs manually.
>>> and the focus sensors may only accept the smaller cone of rays rather
>>> than the full f/1.8 cone.
>
>> I don't understand any of that. :- )
>
> It means the AF sensor has a smaller effective aperture than is being
> used to take the photograph. If you want the best results from a wide
> aperture 50mm of spherical design then you need to understand
> this. Alternatively you could buy a modern aspherical design. One of
> the great benefits of modern technology is that it enables people who
> don't understand what they're doing to do it well.
>
--
David J Taylor
See the ray diagram here:
http://en.wikipedia.org/wiki/Spherical_aberration
With spherical aberration, for example, rays from the edge of the lens
(i.e. full aperture) focus at a different point from those from the middle
part of the lens, so the focus "point" can shift with aperture. The focus
sensors tend to have a narrower acceptance angle than f/1.8, hence they
will adjust the lens so that rays from nearer the centre will be focussed,
leaving the outer rays focussing at a different, incorrect position.
David
OldBoy
Keep in mind that according to Canon, AF accuracy is within 1 CoC for
non-pro EOS camera's and within 1/3 CoC for pro EOS camera's.
(About +/- 0.02 mm resp. +/- 0.010mm at the sensor)
Ofnuts
> Many cameras
> have a max AF sensor aperture of around f6. Some go down as far as
> f2.8.
Actually AFAIK most DSLRs have at least one dual AF sensor, that
switches between one sensor for lenses opening at least at f/5.6 and one
for lenses doing f/2.8 or better. The accuracy of the f/2.8 sensor is
required to focus within the DOF of the more open lenses, and
entry-level cameras lacking these dual sensors cannot use the more open
lenses effectively.
--
Bertrand
Wilba
> Wilba wrote:
>> David J Taylor wrote:
>>> Wilba wrote:
>>>>
>>>> Any thoughts about why the 50/1.8 figures make sense (farside focus...
>>>> 0.8mm outside the DOF), but it doesn't work like that for the 18-55?
>>>
>>> As others have said, the focal point will shift with f/number,
>
>> How is that a factor when shooting at widest apertures?
>
> Because with a specific lens and AF sensor the focusing takes place
> always at the effective aperture of the AF sensor, whereas the
> aperture with which the photograph is taken can vary. Many cameras
> have a max AF sensor aperture of around f6. Some go down as far as
> f2.8. Hence if you're using a lens at f1.8, and it happens to be a
> spherical lens design with aperture related focus drift (as many of
> the golden oldie 50mms are), then this is an important factor.
I thought aperture related focus drift happened if the shot was at an
aperture other than wide open. You're saying when I focus and shoot at the
lens's widest aperture, that's when I get focus drift? How does that work?
>>> and the focus sensors may only accept the smaller cone of rays rather
>>> than the full f/1.8 cone.
>
>> I don't understand any of that. :- )
>
> It means the AF sensor has a smaller effective aperture than is being
> used to take the photograph. If you want the best results from a wide
> aperture 50mm of spherical design then you need to understand
> this.
What do I have to do differently because of it?
> Alternatively you could buy a modern aspherical design. One of
> the great benefits of modern technology is that it enables people who
> don't understand what they're doing to do it well.
I think that often cuts both ways. :- )
Wilba
> Wilba wrote:
>>> OK, here's a possible explanation for the front/back difference. The
>>> character of the out of focus area in front & back is indeed different.
>>> In terms of bokeh, the background is usually smoother and the foreground
>>> harsh. There are a few specialized 'defocus control' lenses which have
>>> an extra ring to change this relationship. I'm not sure quite how that
>>> would effect the AF confirm but it could be a factor.
>>
>> Very interesting idea, thanks very much. This -
>> http://www.stacken.kth.se/~maxz/defocuscontrol/ (and the links in it to
>> http://www.luminous-landscape.com/essays/bokeh.shtml) - gave me enough
>> to get what you're saying.
>>
>> That brings us back again to the notion that it is the optical
>> performance of the 50/1.8 that causes the erratic focus from the farside,
>> not the perceived crudeness of the focus mechanism.
>>
>> What's different for me now is that I believe the erratic farside focus
>> is not about the sharpness wide open, but more likely about the different
>> "character of the out of focus area" on the nearside and farside of the
>> plane of focus.
>
> Yeah, that's the only thing that makes sense, though it's still a bit
> sketchy. In the middle of this page:
> http://www.stacken.kth.se/~maxz/defocuscontrol/
> the crops of dark blurry blobs show harsh rimmed OOF circles and perhaps
> the AF sensor is grabbing those edges. I don't think that's quite right
> but the best I can come up with.
It's a credible idea. :- )
Thanks a bunch for your help!
Wilba
> Chris Malcolm wrote:
>> Wilba wrote:
>>> David J Taylor wrote:
>>>> Wilba wrote:
>>>>>
>>>>> Any thoughts about why the 50/1.8 figures make sense (farside focus...
>>>>> 0.8mm outside the DOF), but it doesn't work like that for the 18-55?
>>>>
>>>> As others have said, the focal point will shift with f/number,
>>
>>> How is that a factor when shooting at widest apertures?
>>
>> Because with a specific lens and AF sensor the focusing takes place
>> always at the effective aperture of the AF sensor, whereas the
>> aperture with which the photograph is taken can vary. Many cameras
>> have a max AF sensor aperture of around f6. Some go down as far as
>> f2.8. Hence if you're using a lens at f1.8, and it happens to be a
>> spherical lens design with aperture related focus drift (as many of
>> the golden oldie 50mms are), then this is an important factor.
>
> Yeah, this could still be the explanation. Which direction does the focus
> shift when stopping down? -that would provide a clue.
> http://diglloyd.com/diglloyd/free/FocusShift/index.html
> might now be centered at 1.02 meters."
>
> On average (ignoring front/back approach), Wilba's camera is focusing
> closer (when 'stopped down' through obstructions to the AF sensor),
> then opening fully to the sensor for taking the pic; the actual point in
> focus is further back. That's the opposite of what this effect would
> suggest.
My version of this jigsaw puzzle has several significant pieces missing.
:- )
"Obstructions to the AF sensor" - what is that? How is that "stopped down"
(for the AF sensor?)?
What part of the camera+lens system is "opening fully to the sensor for
taking the pic"? That sounds like something you'd say about the aperture,
but that isn't closing at any stage so it can't then open...?
Completely not getting what your saying. :- )
Wilba
Sure. So when the aperture *doesn't* change (focus and exposure both occur
at the same aperture), how do you get focus shift in the centre of the
image?
Wilba
> Wilba wrote:
>> Wilba wrote:
I thought that was about the high-precision AF sensor kicking in with lenses
that have a maximum aperture of f/2.8 or better...?
If it is about the max. aperture of the lens (and whether it's narrower than
f/2.8), then the 18-55 should perform worse, not better. Having thought some
more about this, I'm tending to think that it's probably more of a DOF thing
(harder to see it with deeper DOF) than the effect being absent with the
18-55.
OldBoy
It's more a switch to higher sensitivity and by that more accuracy (no
hunting).
> If it is about the max. aperture of the lens (and whether it's narrower
> than f/2.8), then the 18-55 should perform worse, not better. Having
> thought some more about this, I'm tending to think that it's probably more
> of a DOF thing (harder to see it with deeper DOF) than the effect being
> absent with the 18-55.
And in-lens stabilization helps also since the image on the AF sensor is
also stabilized :-)
Paul Furman
> Paul Furman wrote:
>> Chris Malcolm wrote:
>>> Wilba wrote:
>>>> David J Taylor wrote:
>>>>> Wilba wrote:
>>>>>> Any thoughts about why the 50/1.8 figures make sense (farside focus...
>>>>>> 0.8mm outside the DOF), but it doesn't work like that for the 18-55?
>>>>> As others have said, the focal point will shift with f/number,
>>>> How is that a factor when shooting at widest apertures?
>>> Because with a specific lens and AF sensor the focusing takes place
>>> always at the effective aperture of the AF sensor, whereas the
>>> aperture with which the photograph is taken can vary. Many cameras
>>> have a max AF sensor aperture of around f6. Some go down as far as
>>> f2.8. Hence if you're using a lens at f1.8, and it happens to be a
>>> spherical lens design with aperture related focus drift (as many of
>>> the golden oldie 50mms are), then this is an important factor.
>> Yeah, this could still be the explanation. Which direction does the focus
>> shift when stopping down? -that would provide a clue.
>> http://diglloyd.com/diglloyd/free/FocusShift/index.html
>> might now be centered at 1.02 meters."
>>
>> On average (ignoring front/back approach), Wilba's camera is focusing
>> closer (when 'stopped down' through obstructions to the AF sensor),
>> then opening fully to the sensor for taking the pic; the actual point in
>> focus is further back. That's the opposite of what this effect would
>> suggest.
>
> My version of this jigsaw puzzle has several significant pieces missing.
> :- )
>
> "Obstructions to the AF sensor" - what is that? How is that "stopped down"
> (for the AF sensor?)?
>
> What part of the camera+lens system is "opening fully to the sensor for
> taking the pic"? That sounds like something you'd say about the aperture,
> but that isn't closing at any stage so it can't then open...?
>
> Completely not getting what your saying. :- )
The view from the AF sensor isn't as clear as the view to the image
sensor so it's effectively stopped down a little. I'm not sure how exactly.
The view on the ground glass is also. Supposedly anything faster than
perhaps f/2.8 doesn't improve the brightness, etc.
But this focus shift theory seems to predict the opposite behavior from
what you are getting.
Wilba
Yeah, by using a more accurate (i.e. wider "base") AF sensor than the f/5.6
sensor.
>> If it is about the max. aperture of the lens (and whether it's narrower
>> than f/2.8), then the 18-55 should perform worse, not better. Having
>> thought some more about this, I'm tending to think that it's probably
>> more of a DOF thing (harder to see it with deeper DOF) than the effect
>> being absent with the 18-55.
>
> And in-lens stabilization helps also since the image on the AF sensor is
> also stabilized :-)
Yeah, could do. I had it off for all the tests for that reason.
Wilba
> Wilba wrote:
>> Paul Furman wrote:
>>> Chris Malcolm wrote:
>>>> Wilba wrote:
>>>>> David J Taylor wrote:
>>>>>> Wilba wrote:
>>>>>>>
>>>>>>> Any thoughts about why the 50/1.8 figures make sense (farside
>>>>>>> focus... 0.8mm outside the DOF), but it doesn't work like that for
>>>>>>> the 18-55?
>>>>>>
>>>>>> As others have said, the focal point will shift with f/number,
>>>>>
>>>>> How is that a factor when shooting at widest apertures?
>>>>
>>>> Because with a specific lens and AF sensor the focusing takes place
>>>> always at the effective aperture of the AF sensor, whereas the
>>>> aperture with which the photograph is taken can vary. Many cameras
>>>> have a max AF sensor aperture of around f6. Some go down as far as
>>>> f2.8. Hence if you're using a lens at f1.8, and it happens to be a
>>>> spherical lens design with aperture related focus drift (as many of
>>>> the golden oldie 50mms are), then this is an important factor.
>>>
>>> Yeah, this could still be the explanation. Which direction does the
>>> focus shift when stopping down? -that would provide a clue.
>>> http://diglloyd.com/diglloyd/free/FocusShift/index.html
>>> might now be centered at 1.02 meters."
>>>
>>> On average (ignoring front/back approach), Wilba's camera is focusing
>>> closer (when 'stopped down' through obstructions to the AF sensor),
>>> then opening fully to the sensor for taking the pic; the actual point in
>>> focus is further back. That's the opposite of what this effect would
>>> suggest.
>>
>> My version of this jigsaw puzzle has several significant pieces missing.
>> :- )
>>
>> "Obstructions to the AF sensor" - what is that? How is that "stopped
>> down" (for the AF sensor?)?
>>
>> What part of the camera+lens system is "opening fully to the sensor for
>> taking the pic"? That sounds like something you'd say about the aperture,
>> but that isn't closing at any stage so it can't then open...?
>>
>> Completely not getting what your saying. :- )
>
> The view from the AF sensor isn't as clear as the view to the image sensor
> so it's effectively stopped down a little. I'm not sure how exactly.
>
> The view on the ground glass is also. Supposedly anything faster than
> perhaps f/2.8 doesn't improve the brightness, etc.
>
> But this focus shift theory seems to predict the opposite behavior from
> what you are getting.
But how do you get aperture-related focus shift if you're not stopping down
for the exposure?
Paul Furman
> Paul Furman wrote:
>> Wilba wrote:
>>> Paul Furman wrote:
>>>> Chris Malcolm wrote:
>>>>> Wilba wrote:
>>>>>> David J Taylor wrote:
>>>>>>> Wilba wrote:
>>>>>>>> Any thoughts about why the 50/1.8 figures make sense (farside
>>>>>>>> focus... 0.8mm outside the DOF), but it doesn't work like that for
>>>>>>>> the 18-55?
>>>>>>> As others have said, the focal point will shift with f/number,
>>>>>> How is that a factor when shooting at widest apertures?
>>>>> Because with a specific lens and AF sensor the focusing takes place
>>>>> always at the effective aperture of the AF sensor, whereas the
>>>>> aperture with which the photograph is taken can vary. Many cameras
>>>>> have a max AF sensor aperture of around f6. Some go down as far as
>>>>> f2.8. Hence if you're using a lens at f1.8, and it happens to be a
>>>>> spherical lens design with aperture related focus drift (as many of
>>>>> the golden oldie 50mms are), then this is an important factor.
>>>> Yeah, this could still be the explanation. Which direction does the
>>>> focus shift when stopping down? -that would provide a clue.
>>>> http://diglloyd.com/diglloyd/free/FocusShift/index.html
>>>> might now be centered at 1.02 meters."
>>>>
>>>> On average (ignoring front/back approach), Wilba's camera is focusing
>>>> closer (when 'stopped down' through obstructions to the AF sensor),
>>>> then opening fully to the sensor for taking the pic; the actual point in
>>>> focus is further back. That's the opposite of what this effect would
>>>> suggest.
>>> My version of this jigsaw puzzle has several significant pieces missing.
>>> :- )
>>>
>>> "Obstructions to the AF sensor" - what is that? How is that "stopped
>>> down" (for the AF sensor?)?
>>>
>>> What part of the camera+lens system is "opening fully to the sensor for
>>> taking the pic"? That sounds like something you'd say about the aperture,
>>> but that isn't closing at any stage so it can't then open...?
>>>
>>> Completely not getting what your saying. :- )
>> The view from the AF sensor isn't as clear as the view to the image sensor
>> so it's effectively stopped down a little. I'm not sure how exactly.
>>
>> The view on the ground glass is also. Supposedly anything faster than
>> perhaps f/2.8 doesn't improve the brightness, etc.
>>
>> But this focus shift theory seems to predict the opposite behavior from
>> what you are getting.
>
> But how do you get aperture-related focus shift if you're not stopping down
> for the exposure?
It would be opening up for the exposure. Like if you held a smaller
aperture over the front for focusing then removed it to take the shot.
Wilba
> Wilba wrote:
>> Paul Furman wrote:
>>> Wilba wrote:
>>>> Paul Furman wrote:
>>>>> Chris Malcolm wrote:
>>>>>> Wilba wrote:
>>>>>>> David J Taylor wrote:
>>>>>>>> Wilba wrote:
>>>>>>>>> Any thoughts about why the 50/1.8 figures make sense (farside
>>>>>>>>> focus... 0.8mm outside the DOF), but it doesn't work like that for
>>>>>>>>> the 18-55?
>>>>>>>>
>>>>>>>> As others have said, the focal point will shift with f/number,
>>>>>>>
>>>>>>> How is that a factor when shooting at widest apertures?
>>>>>>
>>>>>> Because with a specific lens and AF sensor the focusing takes place
>>>>>> always at the effective aperture of the AF sensor, whereas the
>>>>>> aperture with which the photograph is taken can vary. Many cameras
>>>>>> have a max AF sensor aperture of around f6. Some go down as far as
>>>>>> f2.8. Hence if you're using a lens at f1.8, and it happens to be a
>>>>>> spherical lens design with aperture related focus drift (as many of
>>>>>> the golden oldie 50mms are), then this is an important factor.
>>>>>
>>>>> Yeah, this could still be the explanation. Which direction does the
>>>>> focus shift when stopping down? -that would provide a clue.
>>>>> http://diglloyd.com/diglloyd/free/FocusShift/index.html
>>>>> "With every lens I�ve tested to date, the focus moves farther away.
What the flipping heck are you talking about!?! :- )
How can the aperture go wider than its widest?! Why/how/who/when would you
hold a smaller aperture over the front for focussing?
Are we actually talking about the same thing? The only aperture-related
focus shift I know about is like this -
http://diglloyd.com/diglloyd/free/FocusShift/index.html. What you're saying
only sounds like it might make some kind of sense in relation to something
completely different. :- )
Paul Furman
:-)
The AF sensors are off to the side, looking through mirrors and/or
prisms or something, like the viewfinder. When you block part of their
view, the obstruction acts the same as aperture blades: making the
opening narrower.
When the actual photo is taken, all the mirrors get out of the way and
the full f/1.8 projects in all it's glory onto the photo sensor.
I could be wrong but that's my understanding.
Wilba
>>>>>>> "With every lens I�ve tested to date, the focus moves farther away.
Ri-ight.
> When you block part of their view, the obstruction acts the same as
> aperture blades: making the opening narrower.
What does that?! How's it relevant to _anything_ FFS? :- )
> When the actual photo is taken, all the mirrors get out of the way and the
> full f/1.8 projects in all it's glory onto the photo sensor.
Yeah... so at that point there's no light from the optical path falling on
the AF sensor. Why does it matter what the AF sensor is doing then?!! :- )
Just give me any clue, anything, about what this has to do with my
experiements and this kind of focus shift -
http://diglloyd.com/diglloyd/free/FocusShift/index.html - that _could_ be a
factor, _if_ I was stopping down the aperture from its maximum. :- )
Paul Furman
The trend this suggests is the opposite of what you are experiencing if
I'm following this right, so consider this a tangent. But if it were a
factor, it would be because there is something blocking the view that
the AF sensors see, so they are locking onto the target based on a
stopped down view.
It's like the AF sensor(s) are sitting at the side of the stage, peering
through the curtains at an angle through a small mirror <g>.
Happy New Year (my time) in one minute...
Wilba
>>>>>>>>> "With every lens I�ve tested to date, the focus moves farther
Ah, OK. As long as it's not related to any known phenomenon of relevance, I
won't let it won't bother me. :- )
> Happy New Year (my time) in one minute...
Yeah? We did that last night. <yawn> :- )
David J Taylor
[]
> But how do you get aperture-related focus shift if you're not stopping
> down for the exposure?
Focus sensors use the rays from the centre portion of the ray bundle
coming out of the back end of the lens (effectively f/2.8 or f/4.0 - for
example), and may therefore focus on a different axial location than where
the rays from the extreme of the lens (f/1.8, say) converge, if the lens
has spherical aberrations.
David
David J Taylor
[]
> Sure. So when the aperture *doesn't* change (focus and exposure both
> occur at the same aperture), how do you get focus shift in the centre of
> the image?
I already explained that the focus sensor may only have a working aperture
of f/4, for example, so the aperture /is/ different.
David
Doug McDonald
>>
>> With spherical aberration, for example, rays from the edge of the lens
>> (i.e. full aperture) focus at a different point from those from the middle
>> part of the lens, so the focus "point" can shift with aperture. The focus
>> sensors tend to have a narrower acceptance angle than f/1.8, hence they
>> will adjust the lens so that rays from nearer the centre will be focussed,
>> leaving the outer rays focussing at a different, incorrect position.
>
> Sure. So when the aperture *doesn't* change (focus and exposure both occur
> at the same aperture), how do you get focus shift in the centre of the
> image?
>
>
At least in Canon cameras, it is because the focus sensors use at most
the light of an f/2.8 lens. IF this is the problem cause, the focus should be
correct at f/2.8 using the central cross autofocus sensor.
TYPICALLY, for Gauss lenses like this one, no aspheric elements, the central part
of the lens (say that part used at f/11) and the very outer part focus at the
same distance. It is the part at one f-stop from maximum (i.e. in this
lens near f/2.8) that suffers worst from focus shift. Thus the problem.
However, this is a small problem.
Doug McDonald
Wilba
> Wilba wrote:
>> But how do you get aperture-related focus shift if you're not stopping
>> down for the exposure?
>
> Focus sensors use the rays from the centre portion of the ray bundle
> coming out of the back end of the lens (effectively f/2.8 or f/4.0 - for
> example), and may therefore focus on a different axial location than where
> the rays from the extreme of the lens (f/1.8, say) converge, if the lens
> has spherical aberrations.
Let me rephrase my question for clarity -
How do you get aperture-related focus shift
(http://diglloyd.com/diglloyd/free/FocusShift/index.html), in the centre of
the image, if you're not stopping down for the exposure?
Wilba
> Wilba wrote:
>> Sure. So when the aperture *doesn't* change (focus and exposure both
>> occur at the same aperture), how do you get focus shift in the centre of
>> the image?
>
> I already explained that the focus sensor may only have a working aperture
> of f/4, for example, so the aperture /is/ different.
That's nothing to do with the only form of aperture-related focus shift that
I'm familiar with - http://diglloyd.com/diglloyd/free/FocusShift/index.html.
If you are talking about some other phenomenon, please provide a link to a
good description.
I can't see how what you're talking about would be relevant anyway, since
the AF system is calibrated to produce a good focus under exactly those
conditions.
Wilba
> Wilba wrote:
>>> With spherical aberration, for example, rays from the edge of the lens
>>> (i.e. full aperture) focus at a different point from those from the
>>> middle part of the lens, so the focus "point" can shift with aperture.
>>> The focus sensors tend to have a narrower acceptance angle than f/1.8,
>>> hence they will adjust the lens so that rays from nearer the centre will
>>> be focussed, leaving the outer rays focussing at a different, incorrect
>>> position.
>>
>> Sure. So when the aperture *doesn't* change (focus and exposure both
>> occur at the same aperture), how do you get focus shift in the centre of
>> the image?
>
> At least in Canon cameras, it is because the focus sensors use at most
> the light of an f/2.8 lens.
Do you mean that you don't get any advantage from an aperture wider than
f/2.8?
I'd say what I'm thinking more like, a Canon high-precision AF sensor needs
f/2.8 or better to match its "base" (as in a rangefinder).
> IF this is the problem cause, the focus should be
> correct at f/2.8 using the central cross autofocus sensor.
Depends what problem you're talking about. I can't imagine how it could
matter to this problem -
http://diglloyd.com/diglloyd/free/FocusShift/index.html. If you're talking
about a different problem, please give me a link to a description so that I
have a clue what it is.
> TYPICALLY, for Gauss lenses like this one, no aspheric elements,
> the central part of the lens (say that part used at f/11) and the very
> outer part focus at the same distance. It is the part at one f-stop from
> maximum (i.e. in this lens near f/2.8) that suffers worst from focus
> shift.
> Thus the problem.
>
> However, this is a small problem.
What _is_ the problem? :- )
Wilba
>
> Bertrand, happy owner of a 450D and a 50/1.8
Bertrand, would you be willing to spend a few minutes doing a test for me?
I have come up with a simple direct way to test the "crude mechanism"
theory, but it requires something like a macro focus rail to move the camera
in increments of say 0.5mm. Is that something you can do?
Anyone else able to help? Thanks.
Ofnuts
I'd be willing to help, but I haven't got a macro rail... I have to
think if I can cobble up something to move the subject instead.
What do you have in mind?
--
Bertrand
David J Taylor
[]
> Let me rephrase my question for clarity -
>
> How do you get aperture-related focus shift
> (http://diglloyd.com/diglloyd/free/FocusShift/index.html), in the centre
> of the image, if you're not stopping down for the exposure?
.. because the focus sensor /is/ stopping down when measuring the focus.
It's only looking at a small f/4 (or whatever) cone, not the full f/1.8
cone from the lens. Were you to stop down when taking to the cone the
focus sensor is using, the focus would then be exactly as the focus sensor
measured. Use a larger ray bundle when taking, i.e. by using an f/1.8
lens at full aperture, the rays may focus at a different point, and the
resulting image on the sensor may appear out of focus. Likely to be a
small effect unless the lens has excessive aberrations and the ratio
between focus f/number and taking f/number is significant.
The calibration of the AF system will be at a certain aperture (possibly
more than one aperture on more expensive cameras). Use a different lens
aperture and you are at the mercy of how much the lens focal point shifts
when it works at a different aperture to that which the focus sensor uses.
Cheers,
David
Wilba
Yeah, as long as the distance between them changes in a controlled way, it
doesn't matter which moves.
> What do you have in mind?
The crude mechanism theory goes like this - because the lens uses a stepping
motor to move the focus ring, the focus ring must move in perceptible steps,
and therefore one step is big enough to make the difference between nicely
in focus (within the inner 1/3 of the DOF), and clearly out of focus
(outside the DOF). IOW, each step moves the plane of focus by a distance of
the order of the DOF.
That theory predicts that if we take a series of shots over a range larger
than the DOF, at intervals much smaller than the DOF, we would see a
saw-tooth pattern in the focus performance, caused by the focus ring ending
up one step from where it was on the previous step. But if each shot in the
series was equally well focussed, it would mean that the steps are much
smaller than the DOF. At around the minimum focus distance, the DOF is in
the order of 2-4mm.
Here is the method I have used -
1. Camera focussing by phase detect in one shot mode, well-lit high-contrast
subject 50-100mm beyond the minimum focus distance.
2. Take 5-10 shots without moving the camera or subject, to get a feel for
the shot-to-shot variation in focus performance. *Before each shot, prefocus
the lens closer than the subject.*
2. Repeat over an 8mm range of camera-to-subject distance -
a. Increment the camera-to-subject distance, e.g. 0.5mm or finer if you
can. It doesn't matter which way you go as long as it's the same direction
for each increment. You don't have to move exactly the same distance each
time (in some ways it's a good thing if you don't), so don't worry if one
step is like 0.4 and the next is 0.7.
b. Prefocus closer than the subject.
c. Auto-focus - make sure you get a confirmation beep.
d. Either take the shot to critically assess the focus later, or check
the focus now - 10x in Live View is okay if you have good eyesight, but I
use the remote shooting function of the EOS Utility to view a real-time 100%
view on my PC monitor.
It is critical to prefocus closer than the subject before each auto-focus.
You could put something just in front of the subject then AF, or (if you
don't believe the myth that you can break your lens by turning the focus
ring while it's in AF mode), just turn the ring all the way clockwise as
seen from behind.
It sounds like a lot of faffing around when it's decribed in words, but once
you get it set up it only takes a few minutes to complete a series.
Please report on what you observed about the focus through the series of
shots. Thanks!
Wilba
> 1. Camera focussing by phase detect in one shot mode, well-lit
> high-contrast subject 50-100mm beyond the minimum focus distance.
Oops, I left out one critical thing - lens at f/1.8, so NO APERTURE-RELATED
FOCUS DRIFT! :- )
Wilba
> Wilba wrote:
>> []
>> (http://diglloyd.com/diglloyd/free/FocusShift/index.html), in the centre
>> of the image, if you're not stopping down for the exposure?
>
> .. because the focus sensor /is/ stopping down when measuring the focus.
> It's only looking at a small f/4 (or whatever) cone, not the full f/1.8
> cone from the lens. Were you to stop down when taking to the cone the
> focus sensor is using, the focus would then be exactly as the focus sensor
> measured. Use a larger ray bundle when taking, i.e. by using an f/1.8
> lens at full aperture, the rays may focus at a different point, and the
> resulting image on the sensor may appear out of focus. Likely to be a
> small effect unless the lens has excessive aberrations and the ratio
> between focus f/number and taking f/number is significant.
Na, I don't buy it. I can't find any write up apart from speculation in this
thread (which seems to me to be based on a misunderstanding of the original
mention of the effective aperture of the AF sensor), that describes what
you're talking about in relation to aperture-related focus shift. Can you
give me some references?
> The calibration of the AF system will be at a certain aperture (possibly
> more than one aperture on more expensive cameras). Use a different lens
> aperture and you are at the mercy of how much the lens focal point shifts
> when it works at a different aperture to that which the focus sensor uses.
Yep, so if the AF system is calibrated at f/1.8 (as mine is), and I expose
at f/1.8 (as I do), I won't get any shift (as I don't). :- )
Paul Furman
> David J Taylor wrote:
>> Wilba wrote:
>>> []
>>> How do you get aperture-related focus shift
>>> (http://diglloyd.com/diglloyd/free/FocusShift/index.html), in the centre
>>> of the image, if you're not stopping down for the exposure?
>> .. because the focus sensor /is/ stopping down when measuring the focus.
>> It's only looking at a small f/4 (or whatever) cone, not the full f/1.8
>> cone from the lens. Were you to stop down when taking to the cone the
>> focus sensor is using, the focus would then be exactly as the focus sensor
>> measured. Use a larger ray bundle when taking, i.e. by using an f/1.8
>> lens at full aperture, the rays may focus at a different point, and the
>> resulting image on the sensor may appear out of focus. Likely to be a
>> small effect unless the lens has excessive aberrations and the ratio
>> between focus f/number and taking f/number is significant.
>
> Na, I don't buy it. I can't find any write up apart from speculation in this
> thread (which seems to me to be based on a misunderstanding of the original
> mention of the effective aperture of the AF sensor), that describes what
> you're talking about in relation to aperture-related focus shift. Can you
> give me some references?
overview:
http://www.sensorcleaning.com/whatisanslr.php
detail of pro canon AF system (awfully complicated):
http://a.img-dpreview.com/reviews/CanonEOS5DMarkII/Images/features/af04.gif
more detailed explanation:
http://community.the-digital-picture.com/forums/t/2012.aspx
"Many cameras have a special autofocus sensor that becomes active when
the maximum f-number of the lens is f/2.8 or faster. It has better
performance than the normal autofocus sensors because it has a wider
baseline, as illustrated in this diagram from the Canon 1D3 manual:"
more on the last page of their pdf link:
http://doug.kerr.home.att.net/pumpkin/Split_Prism.pdf
Perhaps your 450D lacks this extra sensor, the pdf says the 20D has it.
>> The calibration of the AF system will be at a certain aperture (possibly
>> more than one aperture on more expensive cameras). Use a different lens
>> aperture and you are at the mercy of how much the lens focal point shifts
>> when it works at a different aperture to that which the focus sensor uses.
>
> Yep, so if the AF system is calibrated at f/1.8 (as mine is), and I expose
> at f/1.8 (as I do), I won't get any shift (as I don't). :- )
There is a shift in your tests, it's the other direction though if I was
following correctly. If it turns out to be a more or less random error,
it could be the lack of extra AF sensor for wide lenses.
Wilba
That's all good stuff, but unfortunately none of it has anything to do with
aperture-related focus shift (hint: search for "shift" and "aperture" in all
those documents - you won't find any reference to the focus changing as the
aperture of the lens is stopped down during the exposure).
I see nothing in those documents that suggests that using an f/1.8 lens with
an f/2.8 sensor will cause any kind of errror, shift, perturbation, skew,
imbalance, fluctuation, deviation, or perversion of any kind (that's the
kind of thing you'd have to come up with to convince me that you're onto
something).
> Perhaps your 450D lacks this extra sensor, the pdf says the 20D has it.
It does have it. The high-precision central sensor just means that
wide-aperture lenses will focus more accurately (within the inner 1/3 of the
DOF rather than just within the DOF), not that the focus will change during
exposure at an aperture tighter than wide open.
>>> The calibration of the AF system will be at a certain aperture (possibly
>>> more than one aperture on more expensive cameras). Use a different lens
>>> aperture and you are at the mercy of how much the lens focal point
>>> shifts when it works at a different aperture to that which the focus
>>> sensor uses.
>>
>> Yep, so if the AF system is calibrated at f/1.8 (as mine is), and I
>> expose at f/1.8 (as I do), I won't get any shift (as I don't). :- )
>
> There is a shift in your tests, it's the other direction though if I was
> following correctly.
I would bet lots of money that you haven't. :- )
> If it turns out to be a more or less random error, it could be the lack of
> extra AF sensor for wide lenses.
Absolutely not. AF behaves erratically because the AF sensor confirms focus
at different points depending on the spatial relationship between the
subject and the initial plane of focus. There is a random component in the
single case of auto-focussing when the subject is closer to the camera than
the initial plane of focus, but that randomness is completely removed by
using the confirmation beep method. The focus accuracy is extremely
consistent (it's either spot-on or off by the same small amount), even when
you can't predict which of the two it will be.
David J Taylor
[]
> Na, I don't buy it. I can't find any write up apart from speculation in
> this thread (which seems to me to be based on a misunderstanding of the
> original mention of the effective aperture of the AF sensor), that
> describes what you're talking about in relation to aperture-related
> focus shift. Can you give me some references?
I already pointed you to the Wikipedia article showing how lens
aberrations could cause a shift of focal point as the f/number was
changed:
http://en.wikipedia.org/wiki/Spherical_aberration
Look at this diagram:
http://upload.wikimedia.org/wikipedia/commons/9/92/Spherical_aberration_2.svgand see how the rays from the edge of the lens are brought to a focusnearer the lens than those from the centre part of the lens. Of course, agood lens design will try to minimise this effect. Depending on the exactlens design, the shift of minimum circle of confusion as the aperturechanges may be either towards or away from that at smaller apertures.Remember that the auto-focus on some cameras can be open-loop rather thanclosed loop, and therefore subject to varying degrees of error. Thesensor says the focus is off by so much, the firmware says move the lensby a certain amount, the lens is moved, and that's that. There's nosecond check to see how much error remains. In continuous-focus moderather than single-shot-focus mode, performance may differ again.Cheers,David
Wilba
> Wilba wrote:
>> []
>> this thread (which seems to me to be based on a misunderstanding
>> of the original mention of the effective aperture of the AF sensor),
>> that describes what you're talking about in relation to aperture-related
>> focus shift. Can you give me some references?
>
> I already pointed you to the Wikipedia article showing how lens
> aberrations could cause a shift of focal point as the f/number was
> changed:
>
> http://en.wikipedia.org/wiki/Spherical_aberration
>
> Look at this diagram:
>
> http://upload.wikimedia.org/wikipedia/commons/9/92/Spherical_aberration_2.svg
> and see how the rays from the edge of the lens are brought to a
> Of course, a good lens design will try to minimise this effect.
> Depending on the exact lens design, the shift of minimum circle
> of confusion as the aperture changes may be either towards
> or away from that at smaller apertures.
So you're saying that changing the aperture for the exposure can cause the
focus to shift with a spherically-aberrant lens. Lovely. Wonderful. That's
aperture-related focus shift as we know it
(http://diglloyd.com/diglloyd/free/FocusShift/index.html).
So I ask the exact same question again - if there is no, N-O, none, nil,
nought, and zero aperture change, how then can you get a focus shift due to
this phenomenon?
If you don't understand the question, ask one. :- )
If you are talking about some other kind of aperture change or focus shift,
I need a name and references to _that_. Thanks.
> Remember that the auto-focus on some cameras can be open-loop
> rather than closed loop, and therefore subject to varying degrees of
> error. The sensor says the focus is off by so much, the firmware
> says move the lens by a certain amount, the lens is moved, and that's
> that. There's no second check to see how much error remains.
> In continuous-focus mode rather than single-shot-focus mode,
> performance may differ again.
This is a closed-loop system, so not relevant.
David J Taylor
[]
> So you're saying that changing the aperture for the exposure can cause
> the focus to shift with a spherically-aberrant lens. Lovely. Wonderful.
> That's aperture-related focus shift as we know it
> (http://diglloyd.com/diglloyd/free/FocusShift/index.html).
Correct.
> So I ask the exact same question again - if there is no, N-O, none, nil,
> nought, and zero aperture change, how then can you get a focus shift due
> to this phenomenon?
If the focus sensors have a different f/number than that of the lens, they
may focus at the incorrect position, i.e. where the convergence of the ray
bundle that the focus sensors can see is minimum.
> This is a closed-loop system, so not relevant.
It may not be closed loop if you rely on the auto-focus and don't use
continuous mode. If you are using manual focus there is still any
backlash in the system.
Cheers,
David
Ofnuts
Unfortunately, I didn't find anything precise enough in my tool box. The
thread of my adjustable clamps feature is a bit coarse (4mm/turn) and
has too much slack.
I did put the camera on a tripod, using a wire remote (with pre-AF
contact of course) at 50cm in front of a high contrast subject (black on
white cross on my laptop screen) with aperture set at f/1.8.
By gently pushing the top of the screen a few mm (hence about half that
for the test cross which is half-screen), and repeatedly pressing the
pre-AF, and checking front lens rotations:
- the AF "step" corresponds to about 2mm of subject distance
- in a few cases, I could get the AF to settle on two different
positions at random. However, I couldn't find any significant difference
in focus in the two resulting shots.
Of course, given the accuracy of my jury-rigged optical bench, this
should be taken with all due care and suspicion.
--
Bertrand
Paul Furman
Hey! <smack!>
For one thing, I'm not trying to convince you that this phenomenon *is*
the source of your problems, only that it's conceivably a factor in the
complex situation and to explain it fully.
Check it yourself in the diglloyyd link. I think I got it right. If you
think I got it wrong, tell me why.
You are experiencing a back focus error and this theory predicts a front
focus error given that we are talking about the opposite effect he
describes.
"With every lens I�ve tested to date, the focus moves farther away
[after stopping down]. For example, if focus at f/1.4 is centered at
1.00 meters, then by f/2.8 it might now be centered at 1.02 meters."
note, also that focus shift does not appear to be corrected for in the
camera's brains:
"Some cameras today offer a fine-tuning autofocus adjustment on a
lens-specific basis. It would be a modest step for Canon and Nikon to
allow this functionality for the aperture range of the lens, allowing
the user to dial in focus compensation for f/2, f2.8, f/4, etc; the
camera could interpolate intermediate partial apertures."
Now, check this one out:
http://diglloyd.com/diglloyd/free/FocusAccuracy/autofocus.html
"Focusing from infinity to 2 meters
Excellent consistency is seen when the the lens starts at the infinity
mark, with very little detectable difference between shots. This is an
excellent performance in terms of consistency. Even so, none of the
shots are accurate enough... All of the shots... are focused about 1
millimeter in front of the correct focus point...
...
Focusing from 1 meter
In this case, the lens must focus in the opposite direction.
When forcing the lens to start at the 1 meter mark instead of infinity,
focus becomes both less consistent and less accurate"
So this contradicts your results about front/back approach, although he
doesn't suggest a reason for the results.
>> If it turns out to be a more or less random error, it could be the lack of
>> extra AF sensor for wide lenses.
>
> Absolutely not.
OK, I just had to ask.
> AF behaves erratically because the AF sensor confirms focus
> at different points depending on the spatial relationship between the
> subject and the initial plane of focus. There is a random component in the
> single case of auto-focussing when the subject is closer to the camera than
> the initial plane of focus, but that randomness is completely removed by
> using the confirmation beep method. The focus accuracy is extremely
> consistent (it's either spot-on or off by the same small amount), even when
> you can't predict which of the two it will be.
I'll paste your results again below. After reviewing again, and
scratching my head some more, I'll make three observations
1) You say focus is correct coming from close focus but what I see is a
range of accuracy the sensors are able to detect and that range is off
by being back focused a little so that approaching from the foreground,
the first spot where acceptable focus is seen happens to be in the
middle of true focus, then it continues to confirm through some back
point which is not really in focus. Approaching from focus behind the
subject, the most faulty reading in the range of acceptable tolerance is
encountered first.
2) Using the af confirm beep shouldn't matter which direction you are
coming from, it's a live view type reading. Using the camera's af motor,
it *could* matter because the camera takes a snapshot from the initial
position, then jumps to where it calculates the patterns will match.
Referring to the doug.kerr pdf, the af sensors take two snapshots of the
same location on the subject but split in different directions at the af
aperture plate such that they will be offset if mis-focused, then
matches them through cross-correlation and jumps to that spot. The
character of the bokeh might account for differences. Focused behind the
subject, the bokeh is likely to be harsh with hard edged rings, focusing
from in front of the subject, the bokeh of the subject should be softer.
BUT... I don't see why the af prediction match would grab onto the first
closest match, it should line up the two snapshots in the best aligned
situation, not the first found 'close-enough' spot. So the af motor
could have less accuracy starting from close focus due to a softer bokeh
in the snapshot but this would not contribute to a back or front focus
error, just more randomness.
3) Your back focus problem appears to be a mis-calibration issue, for
lack of a better explanation. It's just a coincidence that focusing from
the front hits the edge of the acceptable range in the right spot. I
can't explain why the af motor/brains would decide to stop there rather
than centering in the acceptable zone.
pasted from above for reference:
your diagram with some edits for clarity:
(must use fixed width font or this jpeg:
http://edgehill.net/1/temp/Clipboard01.jpg)
sensor subject background
_________________________________________________________________
|< 430mm >+ sensor to subject distance for optimal focus
_________________________________________________________________
|< 430mm >+ starting from near-side focus
_________________________________________________________________
|< 433mm >+ far-side focus
_________________________________________________________________
| |< 3.0mm >| the "beep band"
_________________________________________________________________
| |< 0.75mm >| back-focus error from center of beep band
I removed the theoretical dof & at the bottom gave a 0.75mm distance for
the back-focus error from the center of the beep zone.
and your text description:
>> Well, it sure has been an interesting day. :- )
>> >
>> > Here are some preliminary results with a 45 degree target. I need to
>> > repeat with a different target before I would state the following with
>> > confidence, and it could be quite different for other units depending on
>> > their calibration. But anyway...
>> >
>> > Sharpness in the very centre at f/1.8 appeared fine, practically
>> > indistinguishable from f/2.8 or f/4. In previous tests I might not have
>> > looked hard enough at the very centre of the image, and may have been
>> > misled by other sources of variation. So that's suggests a disproof of my
>> > "soft at f/1.8" theory.
>> >
>> > I found that my phase detect AF sensor has sidedness. If I start with the
>> > lens focussed closer than the subject, the results are uniformly
>> > excellent, whether autofocussing or manually focussing using the AF
>> > confirmation (as Doug described above).
>> >
>> > If I start with the lens focussed behind the subject, and I manually focus
>> > using the AF confirmation, focus is always off by the same tiny amount
>> > (one click towards infinity in the EOS Utility will bring it into optimal
>> > focus).
>> >
>> > With initial focus behind the subject and PD autofocus, about seven shots
>> > out of ten are out by the same one click as the manual focus, and the rest
>> > are optimal, like when starting from the nearside. I assume that the good
>> > ones come about from the lens overshooting and then the system corrects
>> > towards infinity (so it ultimately approaches focus from the nearside).
>> >
>> > With an external aperture which gives similar exposures to f/2.8 (with the
>> > lens at f/1.8), focus is still excellent starting from the near side, and
>> > it improves the performance when starting from the far side (but still not
>> > optimal).
>> >
>> > Those results, and the fact that the lens is able to focus perfectly via
>> > contrast detect, suggest to me that the "crude and sloppy mechanism"
>> > theory is bogus. I suspect the truth is about the performance of the PD AF
>> > sensor with this lens, specifically how much latitude it has for
>> > confirming a focus. I noticed when manually focussing with a macro rail
>> > that I could move the camera quite some (micro) distance and still get a
>> > focus confirmation beep.
>
> I've been thinking a lot about the width of the "beep band" and what it
> means, so I measured it with a macro rail. With the lens at its closest
> focus (430mm from the sensor), the far beep is 3.0mm from the near beep. A
> conventional DOF calculation gives around 4.3mm, and a focus from the
> nearside appears to put the subject bang in the middle of that. So that
> means a farside focus would put the subject something like 0.8mm outside the
> DOF, which sounds about right.
>
> I also tried it at that distance with an EF-S 18-55 at about 50mm (f/5.6).
> The beep band is 12mm, the DOF is 13mm, and where you start from makes very
> little difference to the focus achieved (although from the nearside is best
> again). I wish I had another wide-aperture FFL lens to compare.
>
> Any thoughts about why the 50/1.8 figures make sense (farside focus... 0.8mm
> outside the DOF), but it doesn't work like that for the 18-55?
Wilba
> Wilba wrote:
> []
>> the focus to shift with a spherically-aberrant lens. Lovely. Wonderful.
>> That's aperture-related focus shift as we know it
>> (http://diglloyd.com/diglloyd/free/FocusShift/index.html).
>
> Correct.
>
>> So I ask the exact same question again - if there is no, N-O, none, nil,
>> nought, and zero aperture change, how then can you get a focus shift due
>> to this phenomenon?
>
> If the focus sensors have a different f/number than that of the lens, they
> may focus at the incorrect position, i.e. where the convergence of the ray
> bundle that the focus sensors can see is minimum.
OK, again, there is no aperture _change_ in that idea, only a static
difference, and I can't see how, if that difference were significant, it
would not be dealt with by calibration.
Again, since that idea has nothing to do with aperture-related focus shift
as we know it (http://diglloyd.com/diglloyd/free/FocusShift/index.html), I
need a name and references for that phenomenon.
>> This is a closed-loop system, so not relevant.
>
> It may not be closed loop if you rely on the auto-focus and don't use
> continuous mode.
It is.
> If you are using manual focus there is still any backlash in the system.
Irrelevant in this case.
Wilba
Yeah, I'd have to take that as inconclusive, but thanks very much for having
a go.
OldBoy
Nice experiment :-)
I don't have an EF 50 f/1.8 so I can't help.
Some related info about the step number of some USM lenses:
http://redmodz.com/lensmounts/59-red-birger-lens-mount-unofficial-mini-faq-and-thread-compilation
David J Taylor
[]
> OK, again, there is no aperture _change_ in that idea, only a static
> difference, and I can't see how, if that difference were significant, it
> would not be dealt with by calibration.
I don't know the Canon system, but I would imagine that you could
calibrate out such a difference.
>>> This is a closed-loop system, so not relevant.
>>
>> It may not be closed loop if you rely on the auto-focus and don't use
>> continuous mode.
>
> It is.
>
>> If you are using manual focus there is still any backlash in the
>> system.
>
> Irrelevant in this case.
Unless the focus point is infinitely small, there is a dead-zone between
the in-focus indication turning on and it disappearing. You might expect
to find the the correct point of focus somewhere between those two points.
As we seem to have drifted into things which you say are irrelevant,
perhaps a succinct restatement of your original problem might help.
Cheers,
David
Chris Malcolm
How come? It's at f/1.8 and wider that I get aperture related focus
drift on my 50mm lens. As would be expected of a spherical design of
lens. The only aperure where no aperture related focus drift would
occur with a spherical lens would be the same aperture as the
effective aperture of the AF sensor doing the phase detect
focusing. AFAIK there is no AF sensor in a DSLR wider than f2.8 in
effective aperture, and in most DSLRS it is a lot narrower than that,
e.g. in the f/5 f/6 range. Of course some DSLRs use lens specific
focus bias adjustment tables to try to cancel out most of the drift.
--
Chris Malcolm
Paul Furman
He needs a reference proving such a phenomenon. Also, by my reckoning,
focus shift would have the opposite effect for this scenario.
Chris Malcolm
You keep quoting that diglloyd document. It goes into aperture related
focus shift when focussing manually. It doesn't consider the focus
shift problems introduced by AF mechanisms and the fact that they have
fixed apertures which are usually smaller than the max aperture of the
lens. It suggests that if you want to know about that kind of thing
you should pay money to get the fuller more detailed explanatory
document.
> If you are talking about some other kind of aperture change or focus shift,
> I need a name and references to _that_. Thanks.
It's exactly the same kind of focus shift, introduced by the factor
not specifically mentioned in the diglloyd document: the difference
which usually exists between the fixed effective aperture of an AF
phase detection sensor, and the lens aperture which the image exposure in
made at.
>> Remember that the auto-focus on some cameras can be open-loop
>> rather than closed loop, and therefore subject to varying degrees of
>> error. The sensor says the focus is off by so much, the firmware
>> says move the lens by a certain amount, the lens is moved, and that's
>> that. There's no second check to see how much error remains.
>> In continuous-focus mode rather than single-shot-focus mode,
>> performance may differ again.
> This is a closed-loop system, so not relevant.
But the loop is closed on an AF phase detect sensor with a fixed
aperture which is usually less than the widest aperture of the lens in
use.
--
Chris Malcolm
Chris Malcolm
> David J Taylor wrote:
>> Wilba wrote:
>> []
>>> So you're saying that changing the aperture for the exposure can cause
>>> the focus to shift with a spherically-aberrant lens. Lovely. Wonderful.
>>> That's aperture-related focus shift as we know it
>>> (http://diglloyd.com/diglloyd/free/FocusShift/index.html).
>>
>> Correct.
>>
>>> So I ask the exact same question again - if there is no, N-O, none, nil,
>>> nought, and zero aperture change, how then can you get a focus shift due
>>> to this phenomenon?
>>
>> If the focus sensors have a different f/number than that of the lens, they
>> may focus at the incorrect position, i.e. where the convergence of the ray
>> bundle that the focus sensors can see is minimum.
> OK, again, there is no aperture _change_ in that idea, only a static
> difference, and I can't see how, if that difference were significant, it
> would not be dealt with by calibration.
Explain how you deal with it by calibration, given that there are
different kinds of designs of 50mm lenses around, some spherical, some
aspherical, some with large focus drift, some designed to minimise it,
and which are going to be used on DSLRs some of which have their
widest aperture AF sensor as wide as (approx) F2.8, and some only as wide as
(approx) f5.6?
> Again, since that idea has nothing to do with aperture-related focus shift
> as we know it (http://diglloyd.com/diglloyd/free/FocusShift/index.html), I
> need a name and references for that phenomenon.
It needs no new name since it's exactly the same phenomenon. You're
just failing to pick up on what happens when AF is brought into the
picture, which your favourite reference above doesn't consider.
--
Chris Malcolm
Chris Malcolm
How can that be, given that different lens designs have very different
amounts of this problem? How does the AF system distinguish between a
spherical and an aspherical design of 50mm lens? And if your answer is
(which is true of some DSLRs) that the camera reads the lens
calibration data from the in-lens chip, then how does the lens
calibration data deal with the different sizes of camera AF sensors
and their different algorithms?
--
Chris Malcolm
Ofnuts
> AFAIK there is no AF sensor in a DSLR wider than f2.8 in
> effective aperture, and in most DSLRS it is a lot narrower than that,
> e.g. in the f/5 f/6 range.
In most DSLRs you have at least one double f/5.6-f/2.8 AF sensor. The
f/5.6 sensor would not be accurate enough for the shallow DOF of lenses
fastser than f/2.8.
--
Bertrand
Wilba
> Wilba wrote:
>>>
>>> If you are using manual focus there is still any backlash in the system.
>>
>> Irrelevant in this case.
>
> Unless the focus point is infinitely small, there is a dead-zone between
> the in-focus indication turning on and it disappearing. You might expect
> to find the the correct point of focus somewhere between those two points.
I guess you mean that there is a range of focusses over which the system
will confirm focus. Yes, with my 450D's high-precision AF sensor, that range
is about as big as the DOF near the closest focus distance. Best focus is
achieved at one end of that range, making the other end clearly OOF.
> As we seem to have drifted into things which you say are irrelevant,
> perhaps a succinct restatement of your original problem might help.
Okay. My original problem was that I couldn't figure out how to test the
"crude mechanism" theory of erratic focus with the EF 50mm f/1.8 II. Doug
McDonald gave me the idea of the beep test (like trap focussing). Using that
idea I was able to show three things - the AF system will confirm focus over
a range about as big as the DOF, where the system confirms focus depends on
where the lens was initially focussed, and that the smallest focus step the
lens can make is of the order of a few % of the DOF. Thus the "crude
mechanism" theory is disproved, and the erratic focus behaviour is
explained.
The remaining challenge is to explain why the AF system confirms focus over
such a wide range. The idea of asymmetrical bokeh makes sense, so the
obvious thing to do now is compare the near side and far side bokeh of the
50/1.8.
Wilba
No, it says their examples were focussed "using Live View". There are three
ways to focus using Live View, two forms of autofocus and one manual, and
AFAICS they don't say which they used.
> It doesn't consider the focus shift problems introduced by AF
> mechanisms and the fact that they have fixed apertures which
> are usually smaller than the max aperture of the lens.
Correct. That's what I want to find out about. Where can I find documents
describing that?
> It suggests that if you want to know about that kind of thing
> you should pay money to get the fuller more detailed explanatory
> document.
I don't get any hint in that document about the phenomenon you're talking
about.
>> If you are talking about some other kind of aperture change or focus
>> shift,
>> I need a name and references to _that_. Thanks.
>
> It's exactly the same kind of focus shift, introduced by the factor
> not specifically mentioned in the diglloyd document: the difference
> which usually exists between the fixed effective aperture of an AF
> phase detection sensor, and the lens aperture which the image
> exposure in made at.
What happens because of this effect? How would the pre-exposure and exposure
focusses be different if it is and isn't present?
Does this effect change the focus at exposure-time if the exposure-time lens
aperture is the same as the focus-time lens aperture? In other words, if the
same difference is there at focussing and exposure, how is this effect
significant?
How can this effect be demonstrated? In other words, what variable can be
changed between exposures to show this effect alone?
>>> Remember that the auto-focus on some cameras can be open-loop
>>> rather than closed loop, and therefore subject to varying degrees of
>>> error. The sensor says the focus is off by so much, the firmware
>>> says move the lens by a certain amount, the lens is moved, and that's
>>> that. There's no second check to see how much error remains.
>>> In continuous-focus mode rather than single-shot-focus mode,
>>> performance may differ again.
>>
>> This is a closed-loop system, so not relevant.
>
> But the loop is closed on an AF phase detect sensor with a fixed
> aperture which is usually less than the widest aperture of the lens in
> use.
So...? I need more than just statements that there is a mysterious
inexplicable unpredictable effect. I need an explanation of how it works,
what happens because of it, and how that's relevant to the case of exposing
at the lens's widest aperture. If it has been documented, all I need is a
URL or specific name. I haven't been able to find anything about it by
searching for aperture-related focus shift.
Wilba
My 50/1.8 has been calibrated with my 450D body by the Canon service centre
in Sydney. They wouldn't tell me how they did it, but they have made it so
that that combination focusses optimally via PD AF at all apertures (as long
as I start with the plane of focus between the camera and the subject).
Where and how is your effect apparent in that?
>> Again, since that idea has nothing to do with aperture-related focus
>> shift
>> as we know it (http://diglloyd.com/diglloyd/free/FocusShift/index.html),
>> I
>> need a name and references for that phenomenon.
>
> It needs no new name since it's exactly the same phenomenon.
Let me see if I've got this right. You're saying that the focus shift you
get when the aperture stops down during exposure with a spherically-aberrant
lens (an exposure-time effect), is exactly the same shift you get because
the effective aperture of the AF sensor is different to the widest lens
aperture (a focus-time effect)?
That doesn't make sense. The way I see it, the AF sensor's effective
aperture may affect how it "sees" the rays that come through the lens, and
it will focus according to that "view" rather than what the image sensor
would "see" at the widest aperture. So if the AF system is able to focus
optimally like that (which mine does), what's the problem at exposure-time?
The only shift that can happen then is because you expose at a tighter lens
aperture, which causes the only kind of aperture-related focus shift I can
find documented (well, and in many places). What am I missing? What are you
assuming that you haven't stated?
> You're just failing to pick up on what happens when AF is brought into
> the picture, which your favourite reference above doesn't consider.
So explain it to me, or give me a reference.
Wilba
It doesn't have to. For these tests, the AF system in my 450D only has to
work with my 50/1.8, which it does, with consistently excellent results
given the right initial conditions.
If you're saying your effect is a general PD AF system design and/or
calibration problem, that's fine by me, but it doesn't have anything to do
with my original question or anything else that has come from it, including
the results and conclusions.
Wilba
Interesting, thanks.
Wilba
I'm talking about the well-documented and widely understood ARFD, not your
mysterious version that seems to be undocumented and remains unexplained and
illogical.
Chris Malcolm
My cheap DSLRs only have a central f/5.6 AF sensor. But it's quite
accurate enough to give precise consistent results down to f/1.4 on my
50mm lens. The only problem is that at wider apertures than f2.8 on
that lens they back focus due to aperture related focus drift. But
they do so precisely and consistently. I could adjust the camera's AF
sensor plane to give accurate focus between f/1.4 and f/2.8 on that
50mm lens. At smaller apertures there would be slight front focus, but
largely trivial because it would be within the larger DoF. But that
adjustment would throw out accurate focus on other lenses where they have
shallow DoF and critical focus.
--
Chris Malcolm
Chris Malcolm
It's explained in your own favourite reference, if only you'd read it
more carefully with a bit more understanding. It doesn't actually go
so far as to mention it specifically, but it's implied by the clear
explanation given there. And I'm sure it's gone into in detail in the
fuller explanation he hopes you'll be persuaded to buy by hints as to
what he misses out in the short document you keep referring to.
My "mysterious" version is a straightforward logical implication of
that explanation of aperture related focus shift. It follows logically
from the well documented fact which you already accept that the AF
sensor has a fixed aperture which is smaller than the widest apertures
of a fast lens.
--
Chris Malcolm