PDA

View Full Version : Autofocus with ultrawide angles



pmack
28-02-2011, 6:44pm
Hey all

I was testing out how the auto focus behaved on my wide angle.
A 10-22mm on my 7D.

I walked out to my balcony which overlooks a valley, the closest object is a tree 10 meters away off to the side.

With just the center focus point selected, it was consistantly selecting past infinity.
With all focus point selected, it was jumping sporadically between 1meter and the infinity mark.
Even with the camera still, each time i press to focus, it gives a different result.
Occasionally it would jump to 0.5 meters.

Of course black squares were coming up saying it was in focus, even though these squares were over scenery 400 meters away.

Now of course wide angle has a large depth of field so this isn't always a major problem, but what i want to know is if this is normal behaviour, or if there is a problem with my camera/lens.

I tested it out with an old 350D and got simillar results.
Is this just how it is?

Thanks!

Clubmanmc
01-03-2011, 4:38pm
depends on settings and the like...

was the camera in AI servo or in one shot?

as for multiple spots of focus, and the infinity sign, on a 10-22 the infinity is not that far out pretty much any thing past 1m is infinity so it should be sharp thru the entire DOF (see the attached image)

the camera's AF is looking for contrast to focus on, and the 7D has 19 cross type AF points, so it will pick the best one each time..

was the camera on a tirpod? or was it hand held.. as the smallest of movement will make the difference...

M

pmack
02-03-2011, 5:40pm
I was on one shot, and yeah i wasn't using a tripod, but my point being that every single autofocus point was across the valley 400 metters away so it shouldn't matter if i moved the camera an inch to the side.
so picking up the "best" one is more like picking up the one which gives the closest FALSE reading.
And with 19 points, that's a lot of focus points to go wrong and focus too close.
And yes i realise how close the 1 and infinity is, i have the lens

can someone simply to this with their lens and see if they get the same result?
I suspect this is just an autofocus "charachteristic" of wide lenses (don't you love that reason given by manufacturers?)

Arg
07-03-2011, 5:21pm
You got the same result with a 7D and a 350D, the lens is "consistently focusing past infinity".

Take the lens to Canon. Tell us how it goes.

pmack
07-03-2011, 6:06pm
You got the same result with a 7D and a 350D, the lens is "consistently focusing past infinity".

Take the lens to Canon. Tell us how it goes.

well it may be normal behaviour though at 10mm, so I don't want to send it back for nothing. I suspect this is quite normal for ultra wide angles.
They don't need to be super accurate.

Could somone kindly test their 10-22mm @ 10mm like i have, and check what results are on the distance scale?

cheers

Arg
07-03-2011, 11:15pm
What you did is not a test, so neiher I nor anyone else can replicate it.

Why are you using 19-point mode when you posted the other day that it is rubbish? Have you got some kind of axe to grind? Use a mode that works for you! It's common sense.

Read post #2 again. If the image is sharp from 1m to infinity and you aim the focus point at infinity, the camera can't see any difference whether it sets the lens to 1m or infinity. Or beyond infinity.

I saw your other thread that admin closed one day ago on a similar topic. You seem to be running around doing bizarre things that you mistakenly think is testing your equipment, and asking everyone else to draw a proof from it.

Use settings that work MOST reliably with your camera equipment, not the opposite. Learn the BEST way to get the BEST results with your equipment. Then use the knowledge and expertise you have gathered to make some strong photos. That's the path I'm on and, I suspect, most other people here.

Roosta
08-03-2011, 11:29am
Just to follow up with Arg, I used 10-22 on my 50D with no problems at all, set in multi shot mode, single and multi focus points, worked a treat. I did find for some lower light shots, it performed better in MF, no worries there either. I wouldn't have this lens attached and be in either of the servo modes, not a lens for quick focus re focus sport type spots IMO.

Arg
08-03-2011, 12:00pm
Just to follow up with Arg, I used 10-22 on my 50D with no problems at all, .....

Yeah ditto Roosta. Forgot to mention I have the 7D and I have the EF-S10-22, no problems at all. Look, I disabled the 19-point AF mode and I have recommended to new 7D owners to do the same because it just increases one's failed image rate. It isn't faulty, but it is not so reliable as smaller focusing areas because it has to do more guesswork. I don't like it; I don't use it.

pmack
08-03-2011, 5:48pm
What you did is not a test, so neiher I nor anyone else can replicate it.

well i beg to differ, but last time I disagree'd with people i started world war 3. sigh.

It's not hard to replicate what i did.
Go outside to a scene where there are no objects within 10 meters of range.
Enable 19 point auto select.
With the lens @ 10mm, point camera to said scene, and autofocus on the scene.
Observe position on distance scale, and repeat multiple times to observe the behaviour.


Why are you using 19-point mode when you posted the other day that it is rubbish? Have you got some kind of axe to grind? Use a mode that works for you! It's common sense.

Read post #2 again. If the image is sharp from 1m to infinity and you aim the focus point at infinity, the camera can't see any difference whether it sets the lens to 1m or infinity. Or beyond infinity.

I saw your other thread that admin closed one day ago on a similar topic. You seem to be running around doing bizarre things that you mistakenly think is testing your equipment, and asking everyone else to draw a proof from it.


This thread was actually made before that one, it's been bumped up however... by you actually. So yes since I played around with the 19point mode for that more recent thread, I have learnt to avoid it completely. Out of the 40,000 odd photos on my computer (mostly sports in case you were wondering why so many), I cannot recall any of them where I used the "all points" mode for a serious photo, so what i learnt does not really affect me anyway, but still good information to have.

Good point about if the lens can give an image almost all in focus from 1m to infinity, then the camera will also not be able to see much of a difference when focusing.

Regarding by "bizzare" tests which you don't think are tests, well they are, I am just testing certain behaviour of my equipment to get a better understanding of it. Yes you should focus on what is the best way to get the best results, but it is also important to understand the cameras limitations.

Cheers

arthurking83
08-03-2011, 7:41pm
......

With all focus point selected, it was jumping sporadically between 1meter and the infinity mark.
Even with the camera still, each time i press to focus, it gives a different result.
.....

Whilst I will disagree with arg's summary of whether this is or isn't a test(which it certainly is), I agree with the intent in his comment.

A test is a test, regardless of the outcome of the experiment, I think what arg may have been trying to surmise with his comment is that the test results are inconclusive.

That is, the results you recorded were chaotic and random, so while you did in fact carry out an experiment of some kind which is still regarded as a test, the randomness of the results are obviously inconclusive.

Each scene is going to give different results, and my personal opinion is that you are not really doing this 'test' in a scientific manner to reproduce accurately repeatable results.

I did in fact test my D300's ability to focus using the auto focusing feature, but I maintained the 51 point focus points, and I dare say in reducing this to a lower figure would still have given me back the results I got.
I did two different tests, and recorded two opposing results, where I could repeatedly get a similar set of random focusing distances, AND in my other test I could easily get the same focus point to focus, where both test runs were done 10 times to ensure the broadest set of results to make any sense of.

In the quote above, I'm still not convinced that you are doing this in a methodical manner, and in not doing so, you will generally get inconclusive results.

Can you explain every detail of your method, to ensure that you are testing in a proper manner.
Note: The way in which I regard the term 'proper manner' is based on what I think you are trying to achieve? I'm still unsure as to what this is, but my assumption is that you want 19 point auto to focus on an area or point and stick with that?
Or are you testing for some other 'result'?
The method of testing is the single most crucial aspect of doing any testing, otherwise the test is going likely to be futile and return random chaotic results where no conclusion can be made.

NOTE!!!! if any more of your replies are abrasive and simply another counterargument to any other again, I personally will not reply again, and will seek to have these threads removed(as they don't add to the site in any way, other than to showcase your disrespect for others opinions).
One last important point to note!!.... and this is even more important than your testing method.... if you do not want to read others opinions of and respect their experience, then why bother even posting your results.
It's obvious that almost all respondents are trying to explain something to you, but for some reason, you simply refuse to accept these opinions.

If you explain to us what it is you're trying to achieve, maybe we can help. After all that's a part of the reason a lot of us are here. Another part is to listen and learn. YEP!! even tho we experienced folks know a lot of the stuff we think we need to know, we still listen and learn too!

My opinion is that your testing methods are wrong(if my prior assumption is correct). If so, I know what you may be doing wrong, and what to do. But you have to explain bit for bit, every single aspect of your test, to the finest detail.

Speedway
08-03-2011, 8:33pm
Here we go again. What was the photo and the EXIF info so we can have a look and maybe help you.
Keith.

pmack
08-03-2011, 9:20pm
my personal opinion is that you are not really doing this 'test' in a scientific manner to reproduce accurately repeatable results.

Each scene is going to give different results, and my personal opinion is that you are not really doing this 'test' in a scientific manner to reproduce accurately repeatable results.
Correct, I was not attempting to perform any precise testing for purposes of drawing a bulletproof conclusion etc, I was simply observing the behaviour qualitatively, and was wondering if people had made simillar observations. (or could via their own innacurate testing)

A response that I might have expected to hear (and which the first two posters basically covered) would have been "yes when you are shooting with an ultra wide angle, due to the small difference in the scale between 1m and infinity, you will find that the lens does jump around, particularly when using the less accurate AF point auto select. As always, sticking with the center point for focusing will give the most accurate results"

That's really all i was after.

But if you can see more to add, please do! (and yep am keeping myself in check :o )

What i was trying to achieve, was an understanding of general autofocus issues and characteristics with ultra wide angles. The reason for wanting this understanding was due to the behaviour which i noticed, and recorded in the original post. So with my rough testing, I thought it was unusual behaviour, and it got me thinking that maybe ultra wide angles are commonly like that, i.e they often will jump around to focusing from 1m to infinity when focusing on distant scenes.

I think that clears up what you were asking about what I was trying to test?

There's not really much more detail to how I did it. And as i said this wasn't a very technical test, i wasn't even using a tripod, but i'll explain as much as i can anyway.

Looking out from my balcony, with the camera in one shot auto focus, with all 19 AF points active. I pointed the camera across the valley half a kilometer away. All of the focus points corresponded with this scene half a kilometer away (i.e there was no forground obstructing any AF points). I half depressed the shutter to start autofocus. Being day light it had no trouble getting a lock instantly (typically half or most of the AF points lit up to indicate focus), so i removed my finger. Then i looked at the top of the lens to observe what distance had been selected. I did not take a photo at all, as i was only interested in this position of the lens. I repeated this maybe 20 times, so pointing the camera at approximately the same scene as before. The reason i felt no need to do this on a tripod, is that for every time i did this, If i were taking a photo, each time i would have wanted it to focus almost right on the infinity mark (certainly not the 1m mark). So whether it was pointed at a scene an average of 490 meteres away, or 510 meters away, It didn't matter to my way of thinking. It should never have selected a 1 meter distance (had all the technology been perfect, which if course nothing is).


Thanks for doing your own (by the sounds of it methodical) testing, what exactly did you find?
And what lens?

pmack
08-03-2011, 9:25pm
What was the photo and the EXIF info so we can have a look and maybe help you.
Keith.

Keith, no photos were taken as i was simply inspecting the lens distance scale

Arg
08-03-2011, 10:26pm
Looking out from my balcony, with the camera in one shot auto focus, with all 19 AF points active. I pointed the camera across the valley half a kilometer away. All of the focus points corresponded with this scene half a kilometer away (i.e there was no forground obstructing any AF points). I half depressed the shutter to start autofocus. Being day light it had no trouble getting a lock instantly (typically half or most of the AF points lit up to indicate focus), so i removed my finger. Then i looked at the top of the lens to observe what distance had been selected.A camera is not a distance measuring instrument. That is the function of a tape measure or a laser distance thingy. A camera doesn't have the equipment to measure distance. (although a few historical cameras did use active AF with sonar or IR, but that's not relevant to this thread)

I did not take a photo at all, as i was only interested in this position of the lens. ... It should never have selected a 1 meter distance (had all the technology been perfect, which if course nothing is). I disagree. Remember that a camera's AF is not a distance measuring instrument, it is an image sharpness measuring instrument, via phase detection or contrast detection. What it SHOULD do is what it DID do: choose a setting that creates sufficient sharpness. Where the lens barrel's focusing distance indicator points is irrelevant because it's a case of JOB DONE. If the distance scale marker had been pointing at exactly infinity on the scale, would you have complained because your subject is 400m away and not infinity so it 'should' have pointed at a hair's breadth less than infinity? No you wouldn't (I hope), because it has detected sufficient sharpness. It cannot detect any sharper sharpness, which might even be a function of the optics of the lens, not necessarily the AF system's resolution.

You are asking your camera to measure distance and display that on a gauge, which it is not designed to do. The function of the distance scale is to allow you to manually set focus when you know the distance to the focal subject, without using an optical focusing aid. (And even then you should calibrate the scale manually rather than trust it :rolleyes:).

MarkChap
08-03-2011, 10:39pm
I only really have 1 single comment to add, and I am not 100% sure why I am bothering but here goes.

In the other thread, you told us that your 10-22 worked as expected

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.

So which is it, does this lens focus properly or not ?

pmack
08-03-2011, 10:41pm
You are asking your camera to measure distance and display that on a gauge, which it is not designed to do.
Well not quite, i was using the distance scale as a performance indiactor, not to plan my trip to the moon.
I know exactly what you are saying, and if the image that my camera + lens combo produces when set to 1m is identical to the image produced at infinity, then yes i would agree with everything else you are saying.
But I'm not convinced on that, and now i regret not taking sample photos as i can't check this at the moment, being night time :Doh:

pmack
08-03-2011, 10:52pm
I only really have 1 single comment to add, and I am not 100% sure why I am bothering but here goes.

In the other thread, you told us that your 10-22 worked as expected
.

So which is it, does this lens focus properly or not ?

The 10-22 hit infinity just about every second time. So It was able to get an infinty result. The 24-105 at the wide end, was not able to get an infinity result at all.
The 24-105, when it should have selected infinity, instead selected around 3 to 5 meters, which resulted in a completely blurred image:
http://i7.photobucket.com/albums/y284/surfphotoac/IMG_4332res.jpg

anyway lets just talk about the 10-22 please

arthurking83
08-03-2011, 11:36pm
.........

What i was trying to achieve, was an understanding of general autofocus issues and characteristics with ultra wide angles. The reason for wanting this understanding was due to the behaviour which i noticed, and recorded in the original post. So with my rough testing, I thought it was unusual behaviour, and it got me thinking that maybe ultra wide angles are commonly like that, i.e they often will jump around to focusing from 1m to infinity when focusing on distant scenes.

.....


Thanks for doing your own (by the sounds of it methodical) testing, what exactly did you find?
And what lens?

D300, Sigma 10-20 at 10mm.

first off, you said you wanted an understanding of what or how auto focus in general works with UWA lenses. Using this testing method doesn't actually achieve your desire to understand this tho.

What I did: First up, I did it what I believed to be 'your way'. Set up the camera so that there is something at infinity to focus on. In general anything at or beyond 3m on a 10mm lens is considered to be infinity, or close enough to it.
First test involved focusing using the feature Called 3D focus tracking, which is the same or similar to your 19point auto focusing feature. That is, the camera chooses where to focus, all the photographer has to do is to compose. So the first test was a simple focus, with a wait for the confirmation beep, and focus again and wait for the confirm beep. The lens did appear to focus at random points in the scene doing it this way. The way I have my camera set up, is that I always use continuous mode focus, but this 3D auto focusing requires the use of single shot focus mode(called AF-S mode, continuous mode is called AF-C mode). So camera is set to single AF-S mode to operate in auto focus point mode. in just about all of the ten attempts at focusing, all were randomly focused. There may have been a few similarly focused distances, but the highlighted focus points always changed from one focus attempt to the other.
This test involved one major deviation from the next test, and that was for me, the operator, to not touch the focus ring on the lens in any way. Because this is the crucial difference in the next test.

my assumption of your test is that it never involved defocusing after each focus run. :confused013(you havn't specifically mentioned it, so I assume you haven't)
What may be happening is nothing to do with what lens you choose, nor the actual focus mode you choose or the distance to subject matter, but some engineered programming in the focus mode feature.
Basically as the operator of the camera, you command that the camera focuses, and so even tho the camera has focused, you once again demand it to focus again, so the camera is trying to refocus .. because that's what you are telling it to do. I'm assuming that between each focus attempt you haven't defocused the lens? That is, use this 19point focus mode, but after each run defocus the lens manually so that nothing is 'in focus'. If that means infinity or MFD, then which ever way you need to turn to defocus the image. But if you haven't defocused the lens, then the camera is simply obeying the command. As you want to focus again(without defocusing) then the camera may be assuming that this focus is no good, I'll try another!
Without any evidence to the contrary, we can (safely) assume this is the case. In my non defocused test, I basically hit the same randomness of focus distances.

When I defocused the lens after each focus attempt, then every thing was then straight forward and focus was at the same point, confirmed by the highlighted red square to be at the closest subject distance.
10 out of 10 focus runs, it focused on the closest detail it could reliably focus on, even though these details were black on black. exact same focus point(by the cameras programming) used in each run, and that point was deliberately set by me to see how this 3D auto focusing mode would work.
From this, I can also assume that Nikon have programmed this feature to focus on the closest feature available if it can. 10 out of 10 is reliable enough in my book to confirm this line of thought.

So what I assume now is:
Nikon have programmed some 'intelligence' in this focus mode(which I never use, other than for a few test runs to see what it does.. I'll explain another part this Nikon feature later)
So, back to the assumption, and because I'm not a Nikon DSLR software/firmware engineer, I can only assume base don my testing. If the lens is not focused on anything in particular, or is focused at a specific point, and this auto mode focus is required, it may focus to the closest available subject, but if the command is then to focus again, without any user input, it may then try to focus again to a different distance. After all, if the camera has already acquired focus(as it believes it has), then it should obviously try to focus to another point, if the command to do so again is made.
Without considering other(or all) aspects of a test like this, and the results turn out to be random, then there is nothing unusual to report, as the camera is simply obeying a command ... to refocus. Every focus attempt made by the operator is in effect a command to focus. When the camera is set to have some level of automation, such as this auto mode focusing, then the results it returns simply have to be accepted(unless you are the programmer and you know these results are incorrect!)
When you specifically set the camera not to automate any aspect of the operation, such as single focus point, then the camera has no option other than to focus at that point. Your result will have a level of consistency about them(depending on the subject matter chosen).

With your test, I see nothing wrong with either your camera, nor your lenses. Only problem I saw was in the reason for the test in the first place? Why this test. if it was to simply see what was happening, then sure.. why not? I'd also do it.
But if part of your expectations were to see a probable random event with any sense of repeatability, then I hope know you can see.. why all these replies and posts directed at you that your testing is flawed. Because it appears to us that it is(based on what we thought you were expecting to find).

FWIW, most UWA lens,(that I've had experience with) seem to stop their distance scales at 3m, which I usually find to be a bit long.
1-1.5meters and beyond at f/5.6 usually gives an 'unlimited DOF' at 10mm. If the Canon engineers were smart, they'd have programmed this kind of flexibility into this auto mode focusing feature, and have the lens set to focus at about from 1-3m in most circumstances.

I'm sure if you try this focus mode even with an UWA lens, and position some subject matter set to one of the periphery focus points, at a distance of between 50cm- 1m, and defocus the lens after each focus test run, you may also find a level of consistency about the auto mode focus(generally programmed to focus on the closest subject).
If you don't get the same 10/10 consistent results, as I got in my Nikon with the 10mm lens, then one of two probable explanations can be drawn:
1. Canon do it differently than Nikon
2: Nikon do it differently to Canon :p.. no! ... actually, I meant to say... there may be an issue with your camera.

One other part of the Nikon 3D focus mode, is that it also tracks by colour. So that when enabled, you can choose a primary target(making sure it contrast well against the rest of the scene) and not worry about subject position in the frame.
The target chosen is tracked automatically by the focusing system and the focus point moves about the viewfinder following the subject as best as it can.
This mode of focusing is hit and miss, where the subject moves quickly, but on occasion, it works really well(eg. a red subject against a green background).

ricktas
09-03-2011, 5:27am
Remember that infiinity on a camera focus is usually not infinity, but just past it. Most lenses go past infinity, and the actual infinity point is slightly back from where a lens bumps its end point when manually focusing.

I think rather than doing all these unscientific tests, you should get out and take some photos, of real subjects. You might find you rather enjoy it a lot more than sitting on your balcony doing these tests each and every day. You will soon tire of this wonderful thing called photography if you spend all your time trying to test lenses and gear out, and arguing with members about the results of these tests, day in and day out. Arg is 100% correct, your camera cannot measure distance so the distance scale you keep looking at is a useless exercise. Camera's AF systems, as stated, work on contrast to lock on focus, not detection and measurement of distance. Distance scales on lenses are a hang-over from before AF systems existed, and you should put masking tape over it, and ignore it when using a modern electronic digital camera with an AF system.

You are getting all to worried about something that is completely irrelevant. Go out and enjoy taking photos, I reckon both you and us will enjoy the results more than these threads, that even though people are giving you answers, you choose to ignore or argue against. If you don't, you will find members of AP get sick of seeing your threads and posts and just start ignoring you completely. After all, why would anyone continue to give you advice, when you repeatedly don't accept, acknowlege, or even thank them for the information they are giving you.

Roosta
09-03-2011, 6:30pm
Just quickly, at the end of the day if your lens is taking a quality shoot, good. Most, but not all, will NOT use 10 x + focal point focus for a land/sea scape shot. (Picture a street/city scape sceen from a high vantage point, do you leave it on 19 point focus? or do you use a single point of reference, and like most, maybe even go to MF to get say the wording (Company name LOGO) on the skyscraper in focus)???

I still find that a MF and a steady hand work well, especially if there are objects close and afar your trying to capture, DOF will also become a bigger issue. More so at 10mm.

So, with that said, I don't think there is a concern for your said lens. Does it take a goos shot in MF, and on the older body ?? Have you got an image of the shot where you think the lens is not performing correctly ?

William
09-03-2011, 6:54pm
Was'nt there an old saying, A poor tradsman always blames his tools , I shoot with the Sigma 10-20 on AF and never had a problem with DOF , Always sharp as a tack to the horizon , BTW I never look at the distance scale :D

arthurking83
09-03-2011, 7:39pm
according to the manual:


Best for: Candids and other non-moving subjects where you want to quickly focus on whatever is closest to the camera, or single moving subjects that you expect will move across the composition as you continue to shoot.
Not recommended for: Situations where you need precise control over the point of focus.

If Canon recommend this mode for candids, non moving subjects closest to the camera, or tracking a single subject moving across the frame, how about testing the feature on the types of uses Canon recommend, instead of the types of scenes Canon recommend against!

in2fx
09-03-2011, 7:55pm
I walked out to my balcony which overlooks a valley, the closest object is a tree 10 meters away off to the side.

With just the center focus point selected, it was consistantly selecting past infinity.


Maybe I am reading this wrong, but if you are using the centre spot with the tree off to one side, unless you pointed your centre spot at the tree to focus and then recompose the shot, then the focus would be on the other side of the valley where the centre spot was ?

pmack
09-03-2011, 8:05pm
wowsers, thanks for the extensive reply arthurking

So camera is set to single AF-S mode to operate in auto focus point mode. in just about all of the ten attempts at focusing, all were randomly focused. There may have been a few similarly focused distances, but the highlighted focus points always changed from one focus attempt to the other.
Are you talking about the distance corresponding to the different locations of the highlighted focus points, or are you talking about observing distance displayed on the lens scale. If the latter, what sort of readings were you getting, simillar to mine?



So what I assume now is:
Nikon have programmed some 'intelligence' in this focus mode(which I never use, other than for a few test runs to see what it does.. I'll explain another part this Nikon feature later)
So, back to the assumption, and because I'm not a Nikon DSLR software/firmware engineer, I can only assume base don my testing. If the lens is not focused on anything in particular, or is focused at a specific point, and this auto mode focus is required, it may focus to the closest available subject, but if the command is then to focus again, without any user input, it may then try to focus again to a different distance. After all, if the camera has already acquired focus(as it believes it has), then it should obviously try to focus to another point, if the command to do so again is made.

Hmmm very interesting assumption, makes sense to me. I didn't defocus my camera also, so this all seems pretty consistant.


With your test, I see nothing wrong with either your camera, nor your lenses. Only problem I saw was in the reason for the test in the first place? Why this test. if it was to simply see what was happening, then sure.. why not? I'd also do it.
But if part of your expectations were to see a probable random event with any sense of repeatability, then I hope know you can see.. why all these replies and posts directed at you that your testing is flawed. Because it appears to us that it is(based on what we thought you were expecting to find).

To be honest i completely forget the motivation for the test, maybe i was playing around with my camera out of curiosity sake, and noticed what appeared to be unusual bahaviour. I posted this topic to see if this was a commonly known behaviour or not. I expected the camera to be more repeatable with selecting infinity each time, but your hypothesis or assumption could be the explanation for why that is not the case.

I'll re-do my test after defocusing to see if i get the same results as you.
BTW, did you mention if you used a tripod or not?


One other part of the Nikon 3D focus mode, is that it also tracks by colour. So that when enabled, you can choose a primary target(making sure it contrast well against the rest of the scene) and not worry about subject position in the frame.
The target chosen is tracked automatically by the focusing system and the focus point moves about the viewfinder following the subject as best as it can.
This mode of focusing is hit and miss, where the subject moves quickly, but on occasion, it works really well(eg. a red subject against a green background).

The 7D has a very similar feature. Though i am not sure if it uses colour to follow your object.
You select your starting point, and it follows your target regardless of how you frame it (obviously as long as it stays within the area of the 19 AF points). I haven't tested it out on a proper moving target though so not sure how good it is.

Anyway thanks again, excellent post.
Oh and I am not saying that because of ricktas post.

pmack
09-03-2011, 8:31pm
If Canon recommend this mode for candids, non moving subjects closest to the camera, or tracking a single subject moving across the frame, how about testing the feature on the types of uses Canon recommend, instead of the types of scenes Canon recommend against!
Hmm suppose it depends how you interpret what they say, or more importantly how you are supposed to interpret it.
"non-moving subjects where you want to quickly focus on whatever is closest to the camera" could still be used to explain the scene at infinity. The scene close to infinity may still contain the closest objects to the camera. And "precise control over the point of focus", I think they are referring to control over which part of the frame it focusses on, rather than the accuracy. Although obviously you do get better accuracy with the one point anyway.

Still to test out the how Automatic point selection works in servo mode.

pmack
09-03-2011, 8:34pm
Maybe I am reading this wrong, but if you are using the centre spot with the tree off to one side, unless you pointed your centre spot at the tree to focus and then recompose the shot, then the focus would be on the other side of the valley where the centre spot was ?

Yes, i didn't recompose. I expected it to go close to infinity in that case, instead it went past infinity.
This was done to compare with the Auto point AF select result.

pmack
06-04-2011, 5:48pm
nice bump, champion.

btw, in the OP when i said it was selecting past infinity with center point auto focus, i think this is correct as the other day i found infinity on the lens to be slightly past the vertical infinity mark, so halfway across the horizontal line. so results are only bad when all focus points are selected, which would rarely affect me so no biggie

Clubmanmc
08-04-2011, 8:40am
still testing your lenses??

i find the more you tell the camera to do, ie less auto settings as possible.. the more control over the image you have...

so setting it in auto with all the focus points on is kinda not really a test, specifically when the lens has such a short infinity setting...

M