PDA

View Full Version : 7D 19 point autofocus - what's the point?



pmack
05-03-2011, 2:58pm
I typically always use the single center point for autofocus, however there are times when multi point is usefull.

Yet from what i've experienced, the default Auto select 19 point AF is dreadfull.

It seems to always front focus.


An example:

Photo 1 - Auto select 19 point AF
http://i7.photobucket.com/albums/y284/surfphotoac/IMG_4273-screen.jpg

50% crop
http://i7.photobucket.com/albums/y284/surfphotoac/IMG_4273res.jpg



Photo 2 - Single Point
http://i7.photobucket.com/albums/y284/surfphotoac/IMG_4274-screen.jpg

50% crop
http://i7.photobucket.com/albums/y284/surfphotoac/IMG_4274res.jpg


Shots were taken on a tripod with cable release, 1/800s, shot wide open (f/4) to help show focus issues.


My only explanation is that because there are so many active focus points searching for a lock, if one point is only going to be correct only 90% of the time (for arguements sake), then you only have a 13.5% chance of all of the 19 points being correct with each shot.

But yeah i dunno.

Is this just how it is supposed to be? useless?

thanks

etherial
05-03-2011, 4:09pm
I'm not sure of what you are trying to say with all those percentages? Anyway, in any focus mode the camera will be trying to achieve focus on the closest thing it can find on ANY of the focus points which looks like what has happened with your example. The first is focussed closer because the camera has found something closer to focus on. In the second you have limited where it can search to just one point and forced it to focus where you want.

For that reason most recommend not using 19pt auto and selecting something more narrow like zone selection or point expansion which are my favourites.

I @ M
05-03-2011, 4:37pm
In photo 2, why is the crop showing us a tree that is not under the centre focus point used and would appear to be in front of where the camera should have focussed?

pmack
05-03-2011, 4:51pm
in any focus mode the camera will be trying to achieve focus on the closest thing it can find on ANY of the focus points which looks like what has happened with your example. The first is focussed closer because the camera has found something closer to focus on. In the second you have limited where it can search to just one point and forced it to focus where you want.

For that reason most recommend not using 19pt auto and selecting something more narrow like zone selection or point expansion which are my favourites.

errrrm no there is nothing in focus in the first shot :Doh:
(Did you not look at the photo i posted?)

and the point about the percentages was that any errors are exaggerated when you use multiple autofocus opints, as if just one of the 19 points gives a false reading, if this false reading gives a closer distance, then this false reading will be used for the focus, even if the 18 other points functioned correctly.

pmack
05-03-2011, 4:58pm
In photo 2, why is the crop showing us a tree that is not under the centre focus point used and would appear to be in front of where the camera should have focussed?

Look to the left of the tree in the crop, it's covers the same area that the focus point is on. I've just cropped a bit to the right.
But anyway it is irrelevant, the image may as well be a brick wall, this is ~500 meters away. when one house is in focus, just about all of them are.
the 1st image has focused between those two trees in the forground (10 meters away), and the rest of the image (500 meters away).
so if there was a bee somewhere in between, it would probably be in focus, but i highly doubt the camera picked up any insects no more than a pixel in size...

I @ M
05-03-2011, 5:13pm
In the second example where you have placed the centre focus point as the only focus point in use I would have expected the tree that seems to have filled a large amount of that focus point ( in the full photo at least ) to be substantially sharper than the house behind it yet we are unable to see it in the crop.

As for the first photo there appears to be an extremely large depth of field between the focus points that seem to be in use and searching for something to lock focus on.
I don't think that a scene like this is one in which the camera should be allowed to get it right as far as focus goes.
The real benefits of having multiple focus points in a situation comes from being able to select one of those 19 points and use it for your composition without having to frame, focus, recompose and shoot. Canon bodies can do that can't they? Nikon manage it with 51 points and have a similar setup where using predictive tracking of objects using the camera chosen points is best suited to moving objects and particularly those that are moving towards the camera.

pmack
05-03-2011, 5:28pm
In the second example where you have placed the centre focus point as the only focus point in use I would have expected the tree that seems to have filled a large amount of that focus point ( in the full photo at least ) to be substantially sharper than the house behind it yet we are unable to see it in the crop.
huh? no. in the second photo, everything on the hill is in focus, both the trees and houses.
At 500 meters, your DOF (even af f/4), is relatively huge.
in the first photo, nothing on the hill is in focus. neither tree or house.



As for the first photo there appears to be an extremely large depth of field between the focus points that seem to be in use and searching for something to lock focus on.

as i said, this is a long way away, it might as well be a flat image. none of the focus points are correct as not one tree or house on that hill is in focus.




I don't think that a scene like this is one in which the camera should be allowed to get it right as far as focus goes.
The real benefits of having multiple focus points in a situation comes from being able to select one of those 19 points and use it for your composition without having to frame, focus, recompose and shoot. Canon bodies can do that can't they? Nikon manage it with 51 points and have a similar setup where using predictive tracking of objects using the camera chosen points is best suited to moving objects and particularly those that are moving towards the camera.

yes i agree that that is one of the main benefits of the multiple points (selecting your own points), but that is not an excuse for the camera not being able to focus when all points are active. As for saying this is a scene shich is should not "be allowed to get it right", come on are you serious? this is the most simple of subject matters.

ricktas
05-03-2011, 5:43pm
Have you tested your lens using a focus test chart to see if it is focusing correctly? You might just find you need to do some micro-adjustments for the lens and the AF, on the 7D. I had to do so with me 70-200 on my D3, as it was front focusing. Once the micro adjustments to the AF system were done it is spot on.

pmack
05-03-2011, 5:47pm
Have you tested your lens using a focus test chart to see if it is focusing correctly? You might just find you need to do some micro-adjustments for the lens and the AF, on the 7D. I had to do so with me 70-200 on my D3, as it was front focusing. Once the micro adjustments to the AF system were done it is spot on.

I have not used a chart as such, however as it focuses correctly with the center point, micro adjustments will not help this issue.

ricktas
05-03-2011, 5:51pm
OK, was just asking!

I @ M
05-03-2011, 5:57pm
huh? no. in the second photo, everything on the hill is in focus, both the trees and houses.
At 500 meters, your DOF (even af f/4), is relatively huge.

Whilst the DOF ( even at F/4 ) may be large at that distance I would hesitate to say that everything is in focus, even of a pretty soft looking image like this there appear to be parts that are better focussed than others.


in the first photo, nothing on the hill is in focus. neither tree or house.

Quite possibly nothing on the hill is in focus but to me the branches of the gum tree intruding into the bottom r/h corner of the pic look substantially more in focus than the rest of the elements in the image. Perhaps that is where the focus locked when the camera was left to choose the point to focus on.
Perhaps one of those branches was blown into the focus area . :confused013




yes i agree that that is one of the main benefits of the multiple points (selecting your own points), but that is not an excuse for the camera not being able to focus when all points are active. As for saying this is a scene shich is should not "be allowed to get it right", come on are you serious? this is the most simple of subject matters.

Yes, I am serious, and that "simple scene" is obviously one that is not so simple for the camera to get correct. It just is not the type of scene where the camera should be allowed to override user selectable focus points.
Whilst modern DSLR cameras are extremely technologically advanced and accurate in a great number of cases they do require user input to determine the best method of obtaining an image.
Focus and when to use which method are simply one more area that need to be learned thoroughly and applied appropriately because to put it quite simply the camera does not know which particular part of the scene that you want in focus without being told.

pmack
05-03-2011, 5:58pm
OK, was just asking!
all good. personally i think it is just a poor algorithm that canon use when multiple points are active.
it's like the camera just gets confused, too many choices and so it gives a result that is wrong for all focus points.
very frustrating (you can probably tell from my replies I am very frustrated! I am frustrated. fact! :p)

agb
05-03-2011, 6:20pm
There is no doubt that you are not a happy 7D owner pmack, as I think this is your third thread about focus with this camera, and it seems to me that you are no nearer to being satisfied. It is a pity that you cannot get with some other users of this model and compare notes about its functions and results. I have one, and I set it up to my own satisfaction after reading some articles about other users and how they have set it up, and can say that I am very happy with the focus that I achieve. If I lived closer to you I would be happy to share notes and let you have a go with mine, sorry I am too far away.
Mind you I disabled the 19point ( ) focus as soon as I bought the camera and have never used it that way. I have however used spot, single , expanded and zone , and have used most of the 19 points to achieve focus, usually singly or spot moving my focus point with the joy stick as required. I have never had an issue. I have it set to back button focus with the * button, and I have the AF-on button set to lock metering, so the shutter button only meters and takes the photo for me.

pmack
05-03-2011, 6:32pm
Whilst the DOF ( even at F/4 ) may be large at that distance I would hesitate to say that everything is in focus, even of a pretty soft looking image like this there appear to be parts that are better focussed than others.

two objects at the same distance can appear to be focused differently due to colours and shape giving a different perceived sharpness, so not related to focus, but easily confused with. I can assure you from looking at the original file @100%, all of the hill in focus, perhaps at the very top it starts to fall out of the DOF marginally.
can we not digress?



Quite possibly nothing on the hill is in focus but to me the branches of the gum tree intruding into the bottom r/h corner of the pic look substantially more in focus than the rest of the elements in the image. Perhaps that is where the focus locked when the camera was left to choose the point to focus on.
Perhaps one of those branches was blown into the focus area . :confused013

nope the tree in the foreground on the bottom right is not in focus.
i would have taken test images without trees to distract the camera (or forum repliers), but that would have meant taking a photo of a friends house, and they may have found it kinda creepy posting that on the web... haha.



Yes, I am serious, and that "simple scene" is obviously one that is not so simple for the camera to get correct. It just is not the type of scene where the camera should be allowed to override user selectable focus points.
Whilst modern DSLR cameras are extremely technologically advanced and accurate in a great number of cases they do require user input to determine the best method of obtaining an image.
Focus and when to use which method are simply one more area that need to be learned thoroughly and applied appropriately because to put it quite simply the camera does not know which particular part of the scene that you want in focus without being told.
The camera is not overiding anything.
If some kid picks up an SLR for the first time, puts it in full auto mode, and half presses the shutter button, if a red square comes up over an object, that object should be in focus IMO.
In the image i took, the camera has focussed in the middle of the valley, into empty space.
It really should not require user input to overide a retardedly stupid decision that was made by the camera IMO.

pmack
05-03-2011, 6:43pm
Hi agb,
yes this is my third question about autofocus with this camera (all about different issues though), but I was hoping I would get others telling me if their camera acts like this or not, rather than argueing about nothing!
I think i said in the other thread that i was going to end up sending the camera to canon anyway to get it checked out, but i just wanted to hear experience/opinions of others first.

Regarding the 19 point, realistically the only time i ever enable all points is when i have an ultra wide angle and am taking photos of moving objects close up. So in which case, it does not need to perform like you would expect it to perform in my test (~infinity focus).
So the only reason i am posting this topic, is just to find out if this is normal behaviour.
So if the camera can't handle having 19 points active (as I belive this topic prooves), then how much can you trust it when you reduce it to 9 points? (rhetorical question as i realise you personally do not have any problems in other zones)

I @ M
05-03-2011, 6:45pm
errrrm no there is nothing in focus in the first shot :Doh:
(Did you not look at the photo i posted?)



In the image i took, the camera has focussed in the middle of the valley, into empty space.


I am starting to have troubles with this, first you say that there is nothing in focus in the pic and then you say that the camera has focussed half way in.

You can't have it both ways, either the camera focussed or it didn't and I still reiterate that focus and the correct methods of when and where to apply it are things that need to be learned and not left up to the camera.

pmack
05-03-2011, 7:00pm
I am starting to have troubles with this, first you say that there is nothing in focus in the pic and then you say that the camera has focussed half way in.

You can't have it both ways
Yes you can. There is no visible object in focus.
The camera has told the lens to select a focus point that is somewhere between the trees in the foreground (10 meters away), and the hill which is 500 meters away.
so it appears to have focused somewhere between 10-500meters, which is what i mean by half way. half way across the valley.

etherial
05-03-2011, 7:09pm
Hi agb,
yes this is my third question about autofocus with this camera (all about different issues though), but I was hoping I would get others telling me if their camera acts like this or not, rather than argueing about nothing!
I think i said in the other thread that i was going to end up sending the camera to canon anyway to get it checked out, but i just wanted to hear experience/opinions of others first.


Obviously you aren't happy with this camera and have made your mind up what you think. People here are just trying to help you but it seems you don't want to listen. Do some decent focus tests with a focus chart (there are plenty of examples here about microadjustment and how much difference it can make). If you tick all the boxes and still aren't happy send it to Canon or sell it!

If you don't want to listen to people trying to help you and just want to argue everything, don't bother asking the question!

I @ M
05-03-2011, 7:19pm
pmack, either the camera focussed or it didn't.
If, as you have said nothing is in focus then the camera failed to achieve focus.
You have said that the camera focussed "somewhere between the trees in the foreground (10 meters away), and the hill which is 500 meters away" so therefore something must be in focus.

How many attempts did you have at allowing the camera to achieve focus based on the camera body's guesstimation of a focus point?
How many attempts did you have at allowing the camera to accurately focus when you selected the point to focus on?



The camera has told the lens to select a focus point

No, the lens does not select a focus point, the camera body is wholly and solely responsible for determining focus ( autofocus ) from the information about light, contrast and colour that is transmitted through the lens.

As a final comment on this issue, I urge you to set up a definitive and technical focus test under controlled conditions with repeated examples and then determine the outcome for your self, either that or send the camera to Canon for testing and repair / rectification so that it falls within their specified tolerance.
Of course, the repair / rectification will either be free if outside of tolerances and the camera is under a local warranty or you will have to expect to pay for adjustments if the camera is out of warranty period or is a grey imported body.

pmack
05-03-2011, 7:26pm
Obviously you aren't happy with this camera and have made your mind up what you think. People here are just trying to help you but it seems you don't want to listen. Do some decent focus tests with a focus chart (there are plenty of examples here about microadjustment and how much difference it can make). If you tick all the boxes and still aren't happy send it to Canon or sell it!

If you don't want to listen to people trying to help you and just want to argue everything, don't bother asking the question!
once again, are you actually following what is going on in this thread?
I'll just post my response to ricktas who has already suggested this
"as it focuses correctly with the center point, micro adjustments will not help this issue"

I'm not arguing for fun, i'm arguing to keep suggestions valid and usefull.
If I see that someone says something wrong, i'll say it. a spade is a spade, etc.

MarkChap
05-03-2011, 7:43pm
Don't forget that the actual area of the sensor used for auto focus is NOT limited to that tiny red square that is on the screen, it's actually quite a bit larger, the camera will always focus on anything closest to the camera
In your first image I am actually inclined to believe that you have some amount of camera shake, that has blurred the image, hence why nothing looks in focus.
The exif data has been stripped from the images, what are the tech specs on the shot, and what tripod is the camera attached to ?

The camera can obviously focus, at least to a point, you say so your self, in single point AF, you seem to be happy.

I own and LOVE my 7D, do I ever use "auto select", as in let the camera select my focus point, not on your nelly, I can not think of 1 single time where it would be prudent, beneficial, practical or acceptable to allow my camera to choose a focus point.
In fact, in the custom functions, I had turned of the option to use 19 point auto select, just not going to happen. For this reason I cannot say if mine does have a problem with focus when using that feature or not.
IF I had time, which I won't have for at the least the next week, I would give it a try just to check.

So whilst owning and operating a 7D for a bit over 12 months and some 10 000 frames, I can honestly say, any out of focus images have been down to me and me alone.

Art Vandelay
05-03-2011, 8:06pm
Some good points already made.

As a side note, they look way more than 50% crops.
And the first set looks suspiciously like a bit of motion blur.

pmack
05-03-2011, 8:43pm
pmack, either the camera focussed or it didn't.
If, as you have said nothing is in focus then the camera failed to achieve focus.
You have said that the camera focussed "somewhere between the trees in the foreground (10 meters away), and the hill which is 500 meters away" so therefore something must be in focus.

oh dear... are we arguing semantics now?
The elements inside the lens must have been in a position that corresponded to a focal point of somewhere between 10-500meters. I can estimate this by observing the bokeh of the foreground, and the background. yes there is probably a bit of dust that is in focus, but no you cannot see this spec of dust... :Doh:





How many attempts did you have at allowing the camera to achieve focus based on the camera body's guesstimation of a focus point?
How many attempts did you have at allowing the camera to accurately focus when you selected the point to focus on?
for each instance, i took about 5 photos, and results were consistant with the examples posted.




No, the lens does not select a focus point, the camera body is wholly and solely responsible for determining focus ( autofocus ) from the information about light, contrast and colour that is transmitted through the lens.

well you're just arguing semantics again which is hardly usefull.
when i said the camera tells the lens to select a point, i was simply meaning a position that corresponds to a point, or more accurately, a 2 dimensional plane of focus.


As a final comment on this issue, I urge you to set up a definitive and technical focus test under controlled conditions with repeated examples and then determine the outcome for your self, either that or send the camera to Canon for testing and repair / rectification so that it falls within their specified tolerance.
Of course, the repair / rectification will either be free if outside of tolerances and the camera is under a local warranty or you will have to expect to pay for adjustments if the camera is out of warranty period or is a grey imported body.
I did do my test under controlled conditions, and do have multiple examples. There's just no point in posting them all as I didn't think i needed to prove this to anyone, i came here for help, not for people to tell me that "it looks like it has focused a bit here". I'm no idiot, i've done all this testing to be assured the results are repeatable. And yep I do plan on sending my camera to canon anyway, even just for piece of mind. If they say it's fine, they will probably say something like "when trying to focus close to infinity, it is best to have only one or several focus points active as the camera gets confused in this situation"
In which case, i'm fine with that, it should just be published in the manual, or known as a common fact if it is also the case with other cameras.

pmack
05-03-2011, 9:08pm
<>
hey mark,
yes i know the actual area used for the focus is larger than the squares show, that shouldn't have affected the results though.
Here's a shot where it has misfocused to the front (i pressume) again. Nothing could have been close to the squares:

http://i7.photobucket.com/albums/y284/surfphotoac/IMG_4270-screen.jpg
(it actually doesn't look too bad reduced for web, particularly down the bottom, but trust me it's not in focus)

I took this same shot with single point and it of course had no problems.

I can assure you that this is nothing to do with camera shake, i have done this multiple times and it only misses focus when i have it on auto select.

As i have already stated, I do usually only use a single point, but i'll give you an example of where you might want to have all of your focus points active. Just say you're trying to get a close up shot (with a wide angle) of a dog running around. he's constantly moving so you can't easily frame and predict where you want the focus point to be in the frame, however as you know you want the the closest part (i.e the dog) to be in focus, you put it in auto select, and bobs your uncle. ANd i just thought of this one now, so not sure if it is valid as i've not tried it, but another example would be where you are tracking a moving object that fills almost the entire frame, any points that fall off the subject will be the background, hence it should only ever use the focus points on your subject. This should give you much better tracking ability, in theory. however my testing proves that auto select is useless, at least for subjects close to infinity

Max
05-03-2011, 9:09pm
I think main the point of the 19 point system is to keep the closest subject in focus, particular when using it to track multiple moving objects in servo mode. Think a number of runners close together or cyclists in a bunch.

oops, I think you said all this while I was typing, beat me to it.

pmack
05-03-2011, 9:18pm
Some good points already made.

As a side note, they look way more than 50% crops.
And the first set looks suspiciously like a bit of motion blur.

nope they are 50, this is what a 100% crop of pic 1 looks like:
http://i7.photobucket.com/albums/y284/surfphotoac/IMG_4273-100pc.jpg

and as i said, definitely not motion blur, unless autoselect generates it's own internal vibrations... ;)

kiwi
05-03-2011, 9:31pm
I always chuckle with these sort of focus tests where people use just ridiculous subjects such as this with so many different contrast points and items in different focal planes in the same rough focus zone, and miles away to boot!!! If you want to have a credible whine about the camera do some credible tests

pmack
05-03-2011, 9:32pm
I think main the point of the 19 point system is to keep the closest subject in focus, particular when using it to track multiple moving objects in servo mode. Think a number of runners close together or cyclists in a bunch.

oops, I think you said all this while I was typing, beat me to it.

exactly, see in theory, it has it's uses. it's only useless because it doesn't work as it should. (as least in my case)
It's a cop out to just say "select your own point, give up on 19 point auto select". the question is why doesn't the camera behave as you would expect it to.
people are so afraid to use Auto select, because that's the setting that "newbies" and amateurs use.
People need to get off their high horses once in a while and stop worrying about others judging them.
If you need to take a photo in a split second, and your camera is not ready, and all the settings are inappropriately set up for that night shoot you did last night, put it in "green" or P mode and snap that photo. You will not lose your soul if you commit this mortal sin.
Too much misguided elitism. what was that article on smh the other day about narcissism....

MarkChap
05-03-2011, 9:33pm
Tech Specs ??
Camera - 7D
Shutter Speed
ISO ?
Lens ?
Aperture ?
Exposure Mode ? (Av, Tv, M, Auto)
Tripod ?

Just asking ?

Max
05-03-2011, 9:37pm
I gather you took these with single shot focus?

pmack
05-03-2011, 9:45pm
I always chuckle with these sort of focus tests where people use just ridiculous subjects such as this with so many different contrast points and items in different focal planes in the same rough focus zone, and miles away to boot!!! If you want to have a credible whine about the camera do some credible tests

Glad you got a chuckle.

It's nice to know that if I print out a focus test chart, and in lab conditions, I can get my camera to focus properly on a test peice.

Now it's a pitty that in the real world, there is very little market for images of focus charts.
Heaven forbid a camera has to take an image as dramatic as this:


http://i7.photobucket.com/albums/y284/surfphotoac/IMG_4270-screen.jpg

:eek: :eek: :eek:

pmack
05-03-2011, 9:52pm
Tech Specs ??
Camera - 7D
Shutter Speed
ISO ?
Lens ?
Aperture ?
Exposure Mode ? (Av, Tv, M, Auto)
Tripod ?

Just asking ?

sorry i thought i left exif data on the images, must have selected save for web.

details of images 1 and 2 are:
f/4
1/800s
ISO200
24-105L (@105mm)
shot in Av
tripod, with shutter release
single shot, but shutter release carefully half depressed to ensure a focus lock.

MarkChap
05-03-2011, 9:57pm
exactly, see in theory, it has it's uses. it's only useless because it doesn't work as it should. (as least in my case)
It's a cop out to just say "select your own point, give up on 19 point auto select". the question is why doesn't the camera behave as you would expect it to.
people are so afraid to use Auto select, because that's the setting that "newbies" and amateurs use.
People need to get off their high horses once in a while and stop worrying about others judging them.
If you need to take a photo in a split second, and your camera is not ready, and all the settings are inappropriately set up for that night shoot you did last night, put it in "green" or P mode and snap that photo. You will not lose your soul if you commit this mortal sin.
Too much misguided elitism. what was that article on smh the other day about narcissism....

How can choosing, what I, am going to focus on be classed at "elitist"
And for the record, I don't believe I judged anybody for what settings they choose, I simply stated what I use and the reason I couldn't either confirm or deny if my 7D exhibited any thing similar.

Seriuosly, if you are convinced there is an issue with the AF system on your camera, pack it up and send it to Canon.

zolaxi
05-03-2011, 9:57pm
In the many years I've had a Canon 1D/5D type of camera, I've never used 19 point (and in the case of the 5D, 9 point) auto focus.

I can't see why anyone would want to?

And it's not a question of "why doesn't the camera behave as you would expect it to", but rather coming to an understanding of the functions of the camera. It's behaving quite normally. It can't read your mind.

pmack
05-03-2011, 10:08pm
How can choosing, what I, am going to focus on be classed at "elitist"
And for the record, I don't believe I judged anybody for what settings they choose, I simply stated what I use and the reason I couldn't either confirm or deny if my 7D exhibited any thing similar.

Seriuosly, if you are convinced there is an issue with the AF system on your camera, pack it up and send it to Canon.
I wasn't reffering to you or anyone specifically, just the general responses, and the general "fear" people seem have of the 19 point Auto Select.
(but in my testing, founded fear... go figure)
Oh and of course I'm not saying that choosing your own focus point is elitist. That's just common practice, i'm only referring to people who say you should NEVER use 19 point auto select. As long as it actually worked properly, there are definitely examples of where it would be the preferred setting.

pmack
05-03-2011, 10:12pm
In the many years I've had a Canon 1D/5D type of camera, I've never used 19 point (and in the case of the 5D, 9 point) auto focus.

I can't see why anyone would want to?

And it's not a question of "why doesn't the camera behave as you would expect it to", but rather coming to an understanding of the functions of the camera. It's behaving quite normally. It can't read your mind.
please see below:

I do usually only use a single point, but i'll give you an example of where you might want to have all of your focus points active. Just say you're trying to get a close up shot (with a wide angle) of a dog running around. he's constantly moving so you can't easily frame and predict where you want the focus point to be in the frame, however as you know you want the the closest part (i.e the dog) to be in focus, you put it in auto select, and bobs your uncle. ANd i just thought of this one now, so not sure if it is valid as i've not tried it, but another example would be where you are tracking a moving object that fills almost the entire frame, any points that fall off the subject will be the background, hence it should only ever use the focus points on your subject. This should give you much better tracking ability, in theory. however my testing proves that auto select is useless, at least for subjects close to infinity


I think main the point of the 19 point system is to keep the closest subject in focus, particular when using it to track multiple moving objects in servo mode. Think a number of runners close together or cyclists in a bunch.

oops, I think you said all this while I was typing, beat me to it.


exactly, see in theory, it has it's uses. it's only useless because it doesn't work as it should. (as least in my case)
It's a cop out to just say "select your own point, give up on 19 point auto select". the question is why doesn't the camera behave as you would expect it to.
people are so afraid to use Auto select, because that's the setting that "newbies" and amateurs use.
People need to get off their high horses once in a while and stop worrying about others judging them.
If you need to take a photo in a split second, and your camera is not ready, and all the settings are inappropriately set up for that night shoot you did last night, put it in "green" or P mode and snap that photo. You will not lose your soul if you commit this mortal sin.
Too much misguided elitism. what was that article on smh the other day about narcissism....

pmack
05-03-2011, 10:18pm
And it's not a question of "why doesn't the camera behave as you would expect it to", but rather coming to an understanding of the functions of the camera. It's behaving quite normally. It can't read your mind.
You have got to be kidding me. Clearly you're not paying attention to this topic.
The camera should not have to read the operators mind to focus in the image i've shown.
I don't care which square it chooses to select to focus on in this case.
It claims that it made focus on about 7 focus points, but in reality, the entire image is out of focus as it has focused on a plane in front of the photographed scene.
how hard is this to understand? If you honestly think that should be normal operation... i really don't know what to say then.

this is honestly the most frustrating topic i've even made on any forum, gah.

MarkChap
05-03-2011, 10:21pm
That would be as Darren replied, I think your testing methodology is flawed.

But that is just my humble opinion

pmack
05-03-2011, 10:28pm
That would be as Darren replied, I think your testing methodology is flawed.


but the only thing my testing was trying to test, was the ability to focus on a generic scene at great distance, whilst in 19 point Auto select mode.
and it failed that test.
which tells me that either 7D's cannot focus well in that situation, meaning you need to ensure a only a single focus point is used, OR, there is an issue with my camera.
Only way for me to know would be for someone to do a simillar test to mine, or i send it off to canon.

zolaxi
05-03-2011, 10:43pm
Hey, if the 19 point auto focus works for you, keep using it. Maybe I have a closed mind on this topic (I admit!) and should explore it further.

But in the the case of a pooch/cyclist/runner and even BIF, I always use single point focus. In the case of a cyclist, I don't want the camera deciding to focus on say the front wheel. So I use a single focus point aimed at the body or preferably head/face area.

For things like birds, if the eye/head is not in focus, the shot is not in focus to my way of thinking. So I aim there with a single focus point.

As for talk of "misguided elitism", how ridiculous. It's about applying your photographic technique and getting effective results from your camera, not what you do.

If the 19 point auto focus works for you? Hallelujah, and keep using it!! I won't be.

strictfunctor
05-03-2011, 11:49pm
My 2c would be that for any static scene, eg a landscape, manual focus can always
beat any autofocus system, but in a dynamic situation, AF will win, and AF with more
points will be better than AF with fewer.

arthurking83
06-03-2011, 11:44am
pmack, I think you don't have a full understanding of how the focusing system works.
I think you have a basic understanding, of it, but if you expect the camera to focus on a particular on a scene like that and in a consistent manner... well you've just figured out for yourself that you have less understanding of the camera's mode of operation. You just have yet to fully realise this fact.

First up!!... your mode of testing is totally flawed because you haven't given us a reference image to show what the lens is capable of producing at f/4. That is, have you taken a similar image at f/4 focsed to a similar distance to show us(and yourself) that the lens is capable of producing sharp image into the corner at f/4?
You then completely ignore Andrew's(I@M) comments that the image not in focus must be in focus somewhere, and argue your point without fully realising the consequences of the test.
You say that no point is in focus, but don't offer any alternative explanations(such as soft corners due to wide open aperture). You haven't offered any clue as to what lens either. Is it a kit lens? Is it a pro L lens with known fabulous across the frame sharpness. Is it an f/1.4 or f/1.8 lens shot at f/4.. more likely to be sharp across the frame, than an f/4 lens.
Is it a zoom lens shot at the longest focal length.. known to be softer in the corners than it would otherwise be at the shorter end of the focal range?

You're replies to others help are usually counter productive to your quest of working out how to best utilise the auto point focusing mode on this camera, and due to that the inevitable consequence is that you are doomed never to figure this out. Simple solution is to drop it and stop banging your head against some brick wall of technical mystery.. OR!!!! open your mind a little, stop arguing against other's helpful replies and do the test in a more methodical manner.. not some higgledy piggeldy botch job.. you get nowhere this way.

Zolaxi basically explained exactly what I was going to initially comment too.. the camera doesn't read human thought. While you know what you want, the camera only knows what to do as it's programming allows it to do.. no more.

Quick hint. Try the non focused test image again at something like f/8(not knowing what lens you've used) and see if any part of the scene is rendered sharper.(if you say that it's front focused in that scene, then there should be something in focus). Does the lens have a distance scale? If so, before you change any setting, at least have a peek at the distance scale and confirm that the lens has focused to a specific distance
Having done that in the first place, and given that data, I wouldn't have replied at all to your thread, as it would have implied that you have some idea on the technical aspect on what you're trying to achieve.

As it stands, all you have confirmed is that you have limited understanding of the process of testing on a technical level.

OK!! and whoops on my part! I initially made this reply having read a good part of this thread,and then skimmed and missed a few of the final few replies. It seems you have given us the info as to the lens.. and quite simply not surprising!! an F/4 lens shot wide open at the longest focal length.. can you see the error in your testing method. Did you really expect there to be the same image sharpness across the entire frame? Do you know that this lens has an entirely flat focus plane, or is there some curvature?

Your best bet is to either trash all the data you have gathered already, forget that you have this prejudice against this auto point focusing mode and start all over again with an open mind. If you don't, your tainted opinion of this mode of the camera's feature will lead you to see faults where faults probably don't exist.

agb
06-03-2011, 12:12pm
Have another read of this if you have not already done so.
http://www.usa.canon.com/uploadedimages/FCK/Image/2010/QuickGuides/CDLC_EOS7D_AF_QuickGuide.pdf

pmack
06-03-2011, 4:39pm
pmack, I think you don't have a full understanding of how the focusing system works.
I think you have a basic understanding, of it, but if you expect the camera to focus on a particular on a scene like that and in a consistent manner... well you've just figured out for yourself that you have less understanding of the camera's mode of operation. You just have yet to fully realise this fact.


Well no I certainly do not have a full understanding of how the focus system works as I am not one of the engineers who designed it. And I don't hesitate to say that there are likely very few people on this forum that do. At least certainly none who have posted in this thread so far. Yes I have read most material (and yes have RTFM) that canon have released, as have a lot of you.

I expect the camera to focus on the image that it tells the user. It's not hard to understand. This testing has simply shown me a weakness in one function of the autofocus system, nothing more.


First up!!... your mode of testing is totally flawed because you haven't given us a reference image to show what the lens is capable of producing at f/4. That is, have you taken a similar image at f/4 focsed to a similar distance to show us(and yourself) that the lens is capable of producing sharp image into the corner at f/4?

Excuse me? Did you actually look at the original post which has the same image in focus, at f/4?
And why on earth are you overcomplicating things and now talking about corner sharpness? This post of mine was so amazingly straight forward, yet I get people like you trying to prove my lack of understanding, and then bring up completely irrelevant points.



You then completely ignore Andrew's(I@M) comments that the image not in focus must be in focus somewhere, and argue your point without fully realising the consequences of the test.
You say that no point is in focus, but don't offer any alternative explanations(such as soft corners due to wide open aperture). You haven't offered any clue as to what lens either. Is it a kit lens? Is it a pro L lens with known fabulous across the frame sharpness. Is it an f/1.4 or f/1.8 lens shot at f/4.. more likely to be sharp across the frame, than an f/4 lens.
Is it a zoom lens shot at the longest focal length.. known to be softer in the corners than it would otherwise be at the shorter end of the focal range?


Ah no I did not completely ignore any of Andrews comments, I believe I addressed everything he said. Argue the point without realising the consequence? Sorry not sure what you are getting at with that, it was pretty self explanatory what was being talked about, it was simply a trivial misunderstanding of my explanation of how the lens focused to a plane in front of the photographed scene. Once again, corner sharpness? Why are you bringing up useless points. It as shot wide open to exaggerate any missed focus. Pic #1 @ f4 has nothing visible in focus. Pic # 2 @ f/4 has the entire hill in acceptable focus. I can't see how that is hard to follow.
And yep as you're realised, I've posted details of my shots, and the lens was a canon 24-105L lens.



You're replies to others help are usually counter productive to your quest of working out how to best utilise the auto point focusing mode on this camera, and due to that the inevitable consequence is that you are doomed never to figure this out. Simple solution is to drop it and stop banging your head against some brick wall of technical mystery.. OR!!!! open your mind a little, stop arguing against other's helpful replies and do the test in a more methodical manner.. not some higgledy piggeldy botch job.. you get nowhere this way.

People (like yourself) think they are being useful by telling me to read the manual, as though I don't have a good understanding of how to use the autofocus system. Well I believe I do have a reasonably good understanding (not full) of this camera, and I don't think anyone one has posted anything in this thread that I did not already know. The only reason I am doomed to figure this out (on this forum) is that there are so many people in this internet age full of knowledge, but without the intelligence to realise their knowledge is irrelevant, or simply wrong.

And I have done other methodical tests, and the camera has worked. The reason I have posted these particular results are because it is in this particular scenario (multipoint AF at an entire scene close to infinity), that a weakness of the camera is revealed. All I am doing is testing how reliable the camera is in that situation, and the results are repeat-ably POOR. Fact.

MarkChap
06-03-2011, 4:49pm
What make/model is your tripod, I am still of the thought that the first image particularly is affected by camera shake, not just simply being out of focus.

If the camera has front focused to a point close enough to the camera, to not be able to render anything in that scene sharp, there is definitely a serious issue with the AF system. If it was that bad, it would not render anything sharp, ever.

pmack
06-03-2011, 4:51pm
Zolaxi basically explained exactly what I was going to initially comment too.. the camera doesn't read human thought. While you know what you want, the camera only knows what to do as it's programming allows it to do.. no more.

Did you bother reading my response to Zolaxi's post? The camera should not need to mind read in this situation. I'll post my reply again:
“The camera should not have to read the operators mind to focus in the image i've shown.
I don't care which square it chooses to select to focus on in this case. (as all squares cover a region within the available depth of field at this distance)
It claims that it made focus on about 7 focus points, but in reality, the entire image is out of focus as it has focused on a plane in front of the photographed scene.
how hard is this to understand? ”


Quick hint. Try the non focused test image again at something like f/8(not knowing what lens you've used) and see if any part of the scene is rendered sharper.(if you say that it's front focused in that scene, then there should be something in focus). Does the lens have a distance scale? If so, before you change any setting, at least have a peek at the distance scale and confirm that the lens has focused to a specific distance
Having done that in the first place, and given that data, I wouldn't have replied at all to your thread, as it would have implied that you have some idea on the technical aspect on what you're trying to achieve.

Well I'm sorry, next time I start a thread I will state that I graduated from my mechanical engineering degree with honours and have a solid understanding of the technical aspects of photography, due to fact I am more passionate about photography that I am with designing... :Doh:

And of course the image will be sharper at f/8. I walked out to my balcony just now and took the image again at f/4 and f/8. The result at f/8 with 19 point auto select, was very similar to the result at f/4 with center point autofocus. As expected. That was a good suggestion about looking at the distance scale however, as it showed me that the error is only marginal (at least from observing the scale). When it gets the focus correct (with the center point), the mark on the lens is just shy of infinity. In Auto select, it moves approximately 0.5mm away from infinity. A very fine adjustment, but obviously noticeable at f/4. So I according to the scale and the results, I would approximate that when in 19 point Auto select, it is focusing approximately 200 meters or more short of the scene which is approximately 500 meters away.




As it stands, all you have confirmed is that you have limited understanding of the process of testing on a technical level.


And how might that be so, all I am testing is the accuracy of the auto focus in 19 point Auto select mode when taking a photo of an image close to infinity. I think you have limited understanding of what this topic is actually about.


OK!! and whoops on my part! I initially made this reply having read a good part of this thread,and then skimmed and missed a few of the final few replies. It seems you have given us the info as to the lens.. and quite simply not surprising!! an F/4 lens shot wide open at the longest focal length.. can you see the error in your testing method. Did you really expect there to be the same image sharpness across the entire frame? Do you know that this lens has an entirely flat focus plane, or is there some curvature?


Once again I think you have limited understanding of what this thread is about. Corner sharpness and the wide aperture have very little to do with what I am talking about. The wide aperture was used to help illustrate the differences in focus results. When focuses correctly at f/4, the image is acceptable. It seems you missed the photos in my original post.



Your best bet is to either trash all the data you have gathered already, forget that you have this prejudice against this auto point focusing mode and start all over again with an open mind. If you don't, your tainted opinion of this mode of the camera's feature will lead you to see faults where faults probably don't exist.
Look as canon tell users that the center/single point is more accurate, I suppose i can just accept that and not ever use Multi point due to it being largely innaccurate. I simply expected it to be more accurate than it is, but clearly that is not the case. I suspect it is largely due to the fact that probability works against you with the design of this system (the 13.5% number from my first post)

pmack
06-03-2011, 5:51pm
What make/model is your tripod, I am still of the thought that the first image particularly is affected by camera shake, not just simply being out of focus.

If the camera has front focused to a point close enough to the camera, to not be able to render anything in that scene sharp, there is definitely a serious issue with the AF system. If it was that bad, it would not render anything sharp, ever.
nope it's definetly not the tripod, i can repeat this test as often as i want, with or without tripod.
The only reason why nothing is sharp in the scene is because the middle ground is large and empty!

I did this little test, i focused to the right at the hill in 19 point auto select, locked the focus, then i moved my camera to the left towards a crane, and took the photo.
By a matter of coincidence, the crane is approximately at the distance where the lens is focused at, as this image shows:
http://i7.photobucket.com/albums/y284/surfphotoac/IMG_4343res.jpg
100% crop
http://i7.photobucket.com/albums/y284/surfphotoac/IMG_4343crop50jpg.jpg

As you can see even though the focus was locked when the crane was out of view, that is still the rough distance it has focused the lens at.

Meaning it is actually focussing even closer than i thought (the crane is only 200 meters away from me)

pmack
06-03-2011, 6:03pm
Just to throw a spanner in the works, i did some testing with the lens @ 24mm.

In auto point select, almost all focus points lit up for this photo.
After taking the photo, i looked at the scale, it selected a distance of 3 meters! despite the closest object being about 40 meters away (trees on the bottom)

http://i7.photobucket.com/albums/y284/surfphotoac/IMG_4332res.jpg

When i changed it to single/center point focus, and aimed it at the crane, it improved only slightly to a distance of 5 meters, so the shot was still out of focus.
My 10-22mm lens doesn't seem to have trouble getting an infinity focus at 10mm, not sure why the 24-105L can't get an infinity focus zoomed out.

etherial
06-03-2011, 6:18pm
Depth of Field for 24mm, f4 40m is infinite so if it isn't sharp then, you have other issues.

MarkChap
06-03-2011, 6:19pm
so if your 10-22mm acquires focus ok, maybe, just maybe, here is a thought, wait for it.

Maybe it is the lens and not the camera after all. I am sure some one suggested this very thing to early on.

You have spent the last day or say telling us with out doubt that the camera is at fault, you now say that one lens is good and one is not, so just stop and think for a minute, which part of this equation is wrong.

I still think your methodology is flawed, for the life of my I cannot work out how you can focus on something, move your camera and then claim that the camera didn't focus correctly.
And despite your claim that your tripod cannot be at fault, I am still not convinced that you haven't got blur from movement

Art Vandelay
06-03-2011, 6:49pm
Look as canon tell users that the center/single point is more accurate, I suppose i can just accept that and not ever use Multi point due to it being largely innaccurate. I simply expected it to be more accurate than it is, but clearly that is not the case. I suspect it is largely due to the fact that probability works against you with the design of this system (the 13.5% number from my first post)

I'm having a lazy sunday afternoon so turned my auto multi point selection on and poked it out the front yard. I found it worked awesomely well in that it did exactly as it's supposed to do i.e pick up and use the closest point/s it grabs.

http://i20.photobucket.com/albums/b229/Barry_Mundi/test1.jpg

http://i20.photobucket.com/albums/b229/Barry_Mundi/test2.jpg

Even in this jumble of signs and objects it still picked out the closest object.

http://i20.photobucket.com/albums/b229/Barry_Mundi/test4.jpg

It's not often I go birding with the 17-55, :) but this Ibis flew over so grabbed a shot. It picked up reasonably well considering it was a random in single shot mode, not tracked in servo.

http://i20.photobucket.com/albums/b229/Barry_Mundi/test3.jpg

So the multi point on the 7D itself works as intended. (Which is why most people dont use it and prefer to pick their own point of focus) It maybe time to send yours back if you keep having trouble and have exhausted all other causes.

pmack
06-03-2011, 6:57pm
so if your 10-22mm acquires focus ok, maybe, just maybe, here is a thought, wait for it.

Maybe it is the lens and not the camera after all. I am sure some one suggested this very thing to early on.

You have spent the last day or say telling us with out doubt that the camera is at fault, you now say that one lens is good and one is not, so just stop and think for a minute, which part of this equation is wrong.


The lens focuses fine when the camera uses a single auto focus point @ 105mm. This suggests to me the lens is OK. Just because the camera appears to operate well with one lens, and not so well with another lens, does not mean the lens is at fault. However if I test another copy of the 24-105L on my 7D, and it works better, then yes you could conclude that. Unfortunately I do not have another copy of this lens. Or if someone else with this lens tells me that my lens is not acting as theirs does, then that too would help me conclude that.



I still think your methodology is flawed, for the life of my I cannot work out how you can focus on something, move your camera and then claim that the camera didn't focus correctly.


Heaven forbid your life never depends on the understanding of something so simple.
I recomposed to illustrate where the focal plane actually was (approximately)
This was the photo where I obtained the focus using 19 point auto select:

http://i7.photobucket.com/albums/y284/surfphotoac/IMG_4342res.jpg

Then maintaining the same focus, i moved slightly to the left (notice some of the same houses), and took this photo:

http://i7.photobucket.com/albums/y284/surfphotoac/IMG_4343res.jpg

If the houses were in focus in the first photo, they would also have been in focus in the second photo. If you don't understand that, i give up.



And despite your claim that your tripod cannot be at fault, I am still not convinced that you haven't got blur from movement

Ok I hope you are just trying to frustrate me, and are not honestly still thinking this. Do you own tripods with built in vibrators or something? I'm not going to answer this, i have already explained why this has NOTHING to do with motion blur...

pmack
06-03-2011, 7:04pm
I'm having a lazy sunday afternoon so turned my auto multi point selection on and poked it out the front yard. I found it worked awesomely well in that it did exactly as it's supposed to do i.e pick up and use the closest point/s it grabs.

Hey Art thanks for testing that. My camera and lens would most likely have performed just as yours seems to have, as it does seem to work for objects reasonably close. It just seems to be when the whole scene is close to infinity that it front focuses at this focal length, like the photos i've posted.
What lens did you use by the way?

kiwi
06-03-2011, 7:54pm
I'm glad the crane was approximately at the same focal length and that you also focus and recomposed....this sort of exact testing is most comforting.:cool:

MarkChap
06-03-2011, 8:15pm
I am going to say this ONE more time, if your 10-22 focuses as expected and your 24-105 does not focus as expected, how in blue blazes can you blame the CAMERA

And for your information Tripod quality and stability has everything to do with blur. The one other constant you seem reluctant to reveal is the make and model of said rock solid tripod ??
I look at the images and I see blur, not just out of focus images, for my benefit what tripod are you using, then I can be confident that either I am on the right track or not, you want help, with out ALL of the information that is very hard

pmack
06-03-2011, 8:19pm
I'm glad the crane was approximately at the same focal length and that you also focus and recomposed....this sort of exact testing is most comforting.:cool:

I did that to illustrate a point that would be impossible to show any other way (unless i had controll of the crane and could have rotated it for the second photo.)
It was in no way meant to be any sort of exact test.

But thanks for your second usefull post in this topic that once again shows your great level of intelligence and insight.
I look forward to the third...

ricktas
06-03-2011, 8:23pm
I am closing this thread:

The OP seems to have no interest in reading the replies that could be of benefit, and seems to want to only argue with each poster about the information they are providing. Some posts by members are bordering on baiting others, which breaches site rules.