How accurate should one expect autofocus to be?

Started Mar 30, 2013 | Questions
tiko
Regular MemberPosts: 151Gear list
Like?
Re: How accurate should one expect autofocus to be?
In reply to mironv, Mar 30, 2013

mironv wrote:

tiko wrote:

mironv wrote:

AF-S like it or not will do own small ajustmants when will detects that object moves but in less rapid sucsesions than AF-C. .

-- hide signature --

Mironv

Are you sure about this? I mean wouldn't that make the single servo concept pointless?

@AT

No it dose not as time between holding lucking shutter and keeping AF point and time you trip a shutter your object can move and where there is detectable movemant AF-S dose track this and ajust it self you can read it in every Nikon description of AF function of camera. They in past few generations of AF make sych an issue of tracking thak we don't have a trap focus anymore. You know what trap fosus is??? My N90s when set to af but I switch lens to manual and preset at distance I need will not fire untill AF detects object in focuas so this was preffere method for macro AF shooting. You just moved camera a bit holding shutter button fully pressesd and camera trip it when AF modul sees object in focus. You can pre focus at empty space like above obstycels for Jumper Horses and just wait till gets there and have perfect shot. Fuji S5 is last camera that did this but only with manual lenses.

-- hide signature --

I respectfully disagree. Nikon's documentation states that the shutter will not fire in AFS unless focus is achieved, and that once focus is achieved it locks until you fire or disengage the shutter release/Af-on. So once the lock is achieved it stays there and if your subject moves an inch forward such that they're not covered by DOF, it'll be OOF.  And while I'm very familiar with trap focus, it's about "monitoring" the focus under the selected point in AFS as opposed to"making small adjsutments" to the actual focus (which AFC does). If I were to check the distance information on the lens while using AFS, it does not change not even slightly once focus is achieved and locked with a half shutter press.

Maybe we'll agree to disagree

-- hide signature --

@AT

 tiko's gear list:tiko's gear list
Nikon D800 Nikon D810 Nikon AF-S Nikkor 24-70mm f/2.8G ED Nikon AF-S Nikkor 70-200mm f/2.8G ED VR II Nikon AF-S Nikkor 85mm f/1.4G +2 more
Reply   Reply with quote   Complain
Teleboy
Regular MemberPosts: 128
Like?
Re: How accurate should one expect autofocus to be?
In reply to tiko, Mar 30, 2013

Can I just throw my hat into the ring and share some recent experiences on precisely this front - both with my D4 and D800. Good focus at short(ish) distances with telephoto lenses and moving or swaying subjects really demands that you stop down from 2.8; and not only that, but base iso is probably not recommended either. You should, I suggest, increase the sensor's sensitivity to better stop action. I think we often forget this important piece of information: shutter speed is not enough by itself. So much so that I would probably recommend you keep an ISO of 400 or higher running permanently in such situations, with the lens stopped down to 3.5/4/5/5.6 and as high a shutter speed as possible.

....Do that and I believe you (and a lot of people on this forum - some or many of whom, like myself, have only recently started playing with FX) will suddenly find you're getting a hit rate of 80-90% or more (depending on other aspects of technique) - and by that I mean TACK sharp focus: not an approximation of focus achieved by expanding DOF.

Anyway: that's my twopence, after running an experiment over the past couple of days at a wedding I am shooting at (and am still attending here in Austria!).

T.

Reply   Reply with quote   Complain
Just a Photographer
Contributing MemberPosts: 934Gear list
Like?
Re: Do you know what contrast AF really means?
In reply to SixtenB, Mar 30, 2013

SixtenB wrote:

Now to my question: I'm having problems getting sharp photos. One typical example would be taking photos of my 2 year old. I set the camera at 70mm, 1/250 (she's never still), ƒ/2.8, AF-C, d9 area mode and keep the AF-point centered on her eye while keeping AF-ON depressed (I have disabled shutter half-press AF). I then wait a few seconds for the right moment and release the shutter. The result is a perfectly sharp eye (when viewed at 100 %) only in about 10 % of the cases. The rest are either significantly back- or front focused (seemingly at random).

Probably there is nothing wrong with the camera, but user error or better said 'error of user expectation' due to lack of knowledge on how contrast detection AF works.

Your camera works with contrast detection, which means that the AF (selected AF point) of your DSLR looks for the biggest contrasts on the subject you are pointing at. Eg. if you point at the eye, it will set sharpness on the eye as it will see the iris (dark) and the eyeball (white).

Now you also say that you have set your camera to AF-C and 9-point mode.
This means that you have enabled your camera to be allowed to focus on any of those 9-points around the center point you've choosen. So you allow your camera to automatically let it focus on any of those 9 points that are active.

If your 2 year old is physically moving (eg. moving his/hear head) all the time, your camera will continuously detect different contrasts. So at times your camera might see the biggest contrast in e.g.. hair/ear or head/background instead of the eye you thought you were focusing upon. This does not mean your camera is faulty, but it just selects the AF points that has the biggest difference in contrast at the moment you press the shutter.

This can be any of those 9 AF points that are active as you have set your camera to allow it to do this. I would therefore advise you to set AF-S or AF-C 1-point as default and first try if this solves your problem.

 Just a Photographer's gear list:Just a Photographer's gear list
Nikon D4 Nikon D800E Fujifilm X-T1 Nikon AF-S Nikkor 70-200mm f/2.8G ED VR Nikon AF-S Nikkor 14-24mm f/2.8G ED +8 more
Reply   Reply with quote   Complain
Teleboy
Regular MemberPosts: 128
Like?
Re: How accurate should one expect autofocus to be?
In reply to Teleboy, Mar 30, 2013

Teleboy wrote:

Can I just throw my hat into the ring and share some recent experiences on precisely this front - both with my D4 and D800. Good focus at short(ish) distances with telephoto lenses and moving or swaying subjects really demands that you stop down from 2.8; and not only that, but base iso is probably not recommended either. You should, I suggest, increase the sensor's sensitivity to better stop action. I think we often forget this important piece of information: shutter speed is not enough by itself. So much so that I would probably recommend you keep an ISO of 400 or higher running permanently in such situations, with the lens stopped down to 3.5/4/5/5.6 and as high a shutter speed as possible.

....Do that and I believe you (and a lot of people on this forum - some or many of whom, like myself, have only recently started playing with FX) will suddenly find you're getting a hit rate of 80-90% or more (depending on other aspects of technique) - and by that I mean TACK sharp focus: not an approximation of focus achieved by expanding DOF.

Anyway: that's my twopence, after running an experiment over the past couple of days at a wedding I am shooting at (and am still attending here in Austria!).

T.

I'm doing this with my 24-70 2.8 too (not only my telephoto lenses), and finding it really helps.

A.

Reply   Reply with quote   Complain
Just a Photographer
Contributing MemberPosts: 934Gear list
Like?
Re: How accurate should one expect autofocus to be?
In reply to PHXAZCRAIG, Mar 30, 2013

PHXAZCRAIG wrote:

Another suggestion - try a multitude of different AF settings.   What worked on previous cameras may not be the best choice on a D800.    Might as well try 51-pt 3d and see if the face recognition tech in the camera works better than single point...

He is not using single point mode as he has set his camera to AF-C d9. Which means there are 8 active focus points around the selected middle focus point available, making 9 active focus points.

The camera can pick any of those 9 AF points and will then take the AF point with the greatest difference in contrast when he presses the shutter. This biggest contrast difference might be something else then the eye when pressing the shutter with moving objects.

Single point AF would probably work better for him and only if the subject moves in strange patterns (like birds or dancers) I would advise to use 51-pt 3d.

 Just a Photographer's gear list:Just a Photographer's gear list
Nikon D4 Nikon D800E Fujifilm X-T1 Nikon AF-S Nikkor 70-200mm f/2.8G ED VR Nikon AF-S Nikkor 14-24mm f/2.8G ED +8 more
Reply   Reply with quote   Complain
SixtenB
New MemberPosts: 6
Like?
Re: How accurate should one expect autofocus to be?
In reply to tiko, Mar 30, 2013

tiko wrote:

Try using AFC single point instead of 9 point dynamic while still keeping the focus point on the eye. In the dynamic mode the camera will initially use the point you selected but MAY change it to another point if  it finds something with better contrast under ANY of the other (9, 21 or 51) dynamic points. While this is not how Nikon states it in the manual, I found it to be very true

Thank you tiko. Switching from 9 points to single point AF-C seems to improve the focus consistency significantly.  Didn't notice any further improvement moving to AF-S though.

Reply   Reply with quote   Complain
Mojn
Contributing MemberPosts: 882Gear list
Like?
Re: How accurate should one expect autofocus to be?
In reply to SixtenB, Mar 30, 2013

1. calibrate your lens (lots of good advice to be found here)

2. practise - no high-end camera/lens will spit out winners on moving subjects in low-light automatically.

My NEX 5N has face detection. It's great for that kind of shooting.

Reply   Reply with quote   Complain
SixtenB
New MemberPosts: 6
Like?
Re: Do you know what contrast AF really means?
In reply to Just a Photographer, Mar 30, 2013

Just a Photographer wrote:

If your 2 year old is physically moving (eg. moving his/hear head) all the time, your camera will continuously detect different contrasts. So at times your camera might see the biggest contrast in e.g.. hair/ear or head/background instead of the eye you thought you were focusing upon. This does not mean your camera is faulty, but it just selects the AF points that has the biggest difference in contrast at the moment you press the shutter.

This can be any of those 9 AF points that are active as you have set your camera to allow it to do this. I would therefore advise you to set AF-S or AF-C 1-point as default and first try if this solves your problem.

Thanks, I tried AF-C single point, and that seems to give much more consistent results. From the manual, I had assumed that the d9 mode would initially lock focus on what is behind the selected focus point, then keep tracking that object as it moves under all 9 points. From your and tiko's posts, it sounds like it is much less sophisticated than that.

Reply   Reply with quote   Complain
Robin Casady
Forum ProPosts: 11,433
Like?
Re: How accurate should one expect autofocus to be?
In reply to Teleboy, Mar 30, 2013

Teleboy wrote:

Can I just throw my hat into the ring and share some recent experiences on precisely this front - both with my D4 and D800. Good focus at short(ish) distances with telephoto lenses and moving or swaying subjects really demands that you stop down from 2.8; and not only that, but base iso is probably not recommended either. You should, I suggest, increase the sensor's sensitivity to better stop action. I think we often forget this important piece of information: shutter speed is not enough by itself. So much so that I would probably recommend you keep an ISO of 400 or higher running permanently in such situations, with the lens stopped down to 3.5/4/5/5.6 and as high a shutter speed as possible.

Just for clarification, are you saying shoot at f/4 or smaller, and raise ISO so you can get a faster shutter speed at that f/stop?

When you say, "...shutter speed is not enough by itself." do you mean that you need a smaller f/stop with the shutter speed? Or, are you saying that raising ISO, by itself, has some effect on getting sharp images?

-- hide signature --

Robin Casady
http://www.robincasady.com/Photo/index.html
"The whole problem with the world is that fools and fanatics are always so certain of themselves, and wiser people so full of doubts."
— Bertrand Russell

Reply   Reply with quote   Complain
mironv
Veteran MemberPosts: 9,704Gear list
Like?
Re: How accurate should one expect autofocus to be?
In reply to tiko, Mar 30, 2013

tiko wrote:

mironv wrote:

tiko wrote:

mironv wrote:

AF-S like it or not will do own small ajustmants when will detects that object moves but in less rapid sucsesions than AF-C. .

-- hide signature --

Mironv

Are you sure about this? I mean wouldn't that make the single servo concept pointless?

@AT

No it dose not as time between holding lucking shutter and keeping AF point and time you trip a shutter your object can move and where there is detectable movemant AF-S dose track this and ajust it self you can read it in every Nikon description of AF function of camera. They in past few generations of AF make sych an issue of tracking thak we don't have a trap focus anymore. You know what trap fosus is??? My N90s when set to af but I switch lens to manual and preset at distance I need will not fire untill AF detects object in focuas so this was preffere method for macro AF shooting. You just moved camera a bit holding shutter button fully pressesd and camera trip it when AF modul sees object in focus. You can pre focus at empty space like above obstycels for Jumper Horses and just wait till gets there and have perfect shot. Fuji S5 is last camera that did this but only with manual lenses.

-- hide signature --

I respectfully disagree. Nikon's documentation states that the shutter will not fire in AFS unless focus is achieved, and that once focus is achieved it locks until you fire or disengage the shutter release/Af-on. So once the lock is achieved it stays there and if your subject moves an inch forward such that they're not covered by DOF, it'll be OOF.  And while I'm very familiar with trap focus, it's about "monitoring" the focus under the selected point in AFS as opposed to"making small adjsutments" to the actual focus (which AFC does). If I were to check the distance information on the lens while using AFS, it does not change not even slightly once focus is achieved and locked with a half shutter press.

Maybe we'll agree to disagree

-- hide signature --

@AT

Go found a flower outside and use AF-S on it luck your focus and see what happen when wind blows and move that flower.

-- hide signature --
 mironv's gear list:mironv's gear list
Fujifilm FinePix S5 Pro Nikon D700 Nikon AF-S Nikkor 70-200mm f/2.8G ED VR Nikon AF Nikkor 105mm f/2D DC +10 more
Reply   Reply with quote   Complain
jackpro
Regular MemberPosts: 368Gear list
Like?
Re: How accurate should one expect autofocus to be?
In reply to mironv, Mar 31, 2013
 jackpro's gear list:jackpro's gear list
Canon PowerShot S95 Canon EOS 5D Mark II Canon EOS 450D Canon EF 35mm f/1.4L USM Canon EF 85mm f/1.2L II USM +5 more
Reply   Reply with quote   Complain
Teleboy
Regular MemberPosts: 128
Like?
Re: How accurate should one expect autofocus to be?
In reply to Robin Casady, Mar 31, 2013

Teleboy wrote:

Can I just throw my hat into the ring and share some recent experiences on precisely this front - both with my D4 and D800. Good focus at short(ish) distances with telephoto lenses and moving or swaying subjects really demands that you stop down from 2.8; and not only that, but base iso is probably not recommended either. You should, I suggest, increase the sensor's sensitivity to better stop action. I think we often forget this important piece of information: shutter speed is not enough by itself. So much so that I would probably recommend you keep an ISO of 400 or higher running permanently in such situations, with the lens stopped down to 3.5/4/5/5.6 and as high a shutter speed as possible.

Just for clarification, are you saying shoot at f/4 or smaller, and raise ISO so you can get a faster shutter speed at that f/stop?

When you say, "...shutter speed is not enough by itself." do you mean that you need a smaller f/stop with the shutter speed? Or, are you saying that raising ISO, by itself, has some effect on getting sharp images?

-- hide signature --

Robin Casady
http://www.robincasady.com/Photo/index.html
"The whole problem with the world is that fools and fanatics are always so certain of themselves, and wiser people so full of doubts."
— Bertrand Russell

Hi Robyn,
Sorry for the delayed response: I'm travelling and writing to you from Dusseldorf Airport! Basically, yes: 3.5 or higher and yes, higher ISO itself helps freeze action and thereby sharpen shots to the extent the camera is able to control noise. But judging by your galleries, I don't think you needed me to tell you that lol! :). From what I can see on my iPhone, seems like you have some nice product.
T.

Reply   Reply with quote   Complain
Truman Prevatt
Veteran MemberPosts: 3,638Gear list
Like?
Re: How accurate should one expect autofocus to be?
In reply to Just a Photographer, Mar 31, 2013

I was down at the race track (horse) yesterday.  I have always used AF-C 9 point when shooting horses galloping and that works fine.  I also usually shoot at about f8.  Back at the barn I tired an experiment.  I shot the horses first with 9 point and then with AF-C single point. I focus using the AF-on and not the shutter release.

One gray horses or horses with a white stripe, star, etc. the 9 point works fine.  On the bays and chestnuts with no white markings, the 9 point is always out of focus from where I pointed the selected (center) point.  The AF-C single point is dead on.  So the manual is a bit misleading or else I did not interpret it correctly. In reality, if there is not much contrast like a dark horse under a shed row then it will find the best point (which in my case was usually someone's garment, the white wall of the barn, etc. ) for it to focus.

So of these types of subjects AF-C single is the best option by far.  It is not really a fault of the camera - it is doing what it was programmed to do and in many cases that is the right thing to do like galloping horses.  However, the camera cannot read you mind and if you give it an option it will take it.  I expect that outdoors in good light it won't matter.  It would nail the focus in the selected point and not need to look around.  In marginal light - that is a different story.

On day I stayed for the races and photographed people both outside and inside.  I tried AF-C Automatic.  It worked great outside.  It grabbed faces, etc. and the exposure was great.  Inside it did not work so well.  Of course it told me where it was focusing and where it was focusing was not the face I wanted.  It was the highest contrast area in the scene which could have been a shirt, etc.  It was doing what it was programmed to do but it was not reading my mind.

It gets down to a matter of working with the camera and understanding what it will do in particular situations and not expect the camera to read your mind about the focus point.  This is particularly true in more difficult environments like indoors or otherwise lower ambient light. If in doubt use AF-C S or AF-S.

Truman

Just a Photographer wrote:

PHXAZCRAIG wrote:

Another suggestion - try a multitude of different AF settings.   What worked on previous cameras may not be the best choice on a D800.    Might as well try 51-pt 3d and see if the face recognition tech in the camera works better than single point...

He is not using single point mode as he has set his camera to AF-C d9. Which means there are 8 active focus points around the selected middle focus point available, making 9 active focus points.

The camera can pick any of those 9 AF points and will then take the AF point with the greatest difference in contrast when he presses the shutter. This biggest contrast difference might be something else then the eye when pressing the shutter with moving objects.

Single point AF would probably work better for him and only if the subject moves in strange patterns (like birds or dancers) I would advise to use 51-pt 3d.

-- hide signature --

Truman
www.pbase.com/tprevatt

 Truman Prevatt's gear list:Truman Prevatt's gear list
Nikon D800E Nikon AF Nikkor 50mm f/1.4D Nikon AF Nikkor 85mm f/1.4D Nikon AF Nikkor 135mm f/2D DC Nikon AF Nikkor 180mm f/2.8D ED-IF +6 more
Reply   Reply with quote   Complain
John4
Contributing MemberPosts: 791
Like?
Re: How accurate should one expect autofocus to be?
In reply to SixtenB, Mar 31, 2013

I have perhaps the same or a similar problem.  When I got my D800 I tested the extreme focus points are they were right on.  However in subsequent photographs of my grandson some photos were were noticeably and consistently soft.

I just did an interesting experiment.   I used extreme left focus point and focused on a bar in a window.  The vertical bar went right through the center of the focus area.  In the background, across the street, were some cars.  The camera focused on the distant cars.  Doesn't this violate Greatest Contrast captures focus, or Nearest Object captures focus?  Yet when I focus on something with texture which occupies the whole focus rectangle, the focus is good.

Advise and opinion welcome.

Reply   Reply with quote   Complain
Anker BergSonne
Regular MemberPosts: 314Gear list
Like?
Re: How accurate should one expect autofocus to be?
In reply to John4, Mar 31, 2013

The outer focus points work with targets that have good vertical contrast. A vertical bar doesn't have that attribute and it will have difficulty focusing on it. To use a vertical bar as a focus target you should use one of the inner cross-focus points.

Anker

John4 wrote:

I have perhaps the same or a similar problem.  When I got my D800 I tested the extreme focus points are they were right on.  However in subsequent photographs of my grandson some photos were were noticeably and consistently soft.

I just did an interesting experiment.   I used extreme left focus point and focused on a bar in a window.  The vertical bar went right through the center of the focus area.  In the background, across the street, were some cars.  The camera focused on the distant cars.  Doesn't this violate Greatest Contrast captures focus, or Nearest Object captures focus?  Yet when I focus on something with texture which occupies the whole focus rectangle, the focus is good.

Advise and opinion welcome.

 Anker BergSonne's gear list:Anker BergSonne's gear list
Nikon D800 Nikon AF-S DX Nikkor 10-24mm f/3-5-4.5G ED Nikon AF-S Nikkor 70-200mm f/2.8G ED VR II Nikon AF Nikkor 50mm f/1.8D Nikon AF-S Micro-Nikkor 105mm f/2.8G IF-ED VR +3 more
Reply   Reply with quote   Complain
jtan163
Contributing MemberPosts: 939Gear list
Like?
Re: How accurate should one expect autofocus to be?
In reply to SixtenB, Mar 31, 2013

I don't have a D800 - in fact have never had a high end DSLR, so I might be off here, but might changing the release mode help?

I believe that for AF-S and AF-C you can choose to set the shutter to only release when the subject under the focus point is in focus, or you can set it to release as soon as the button is pressed, regardless of focus.

Checking that you are on focus priority instead of release priority might help.

I think it will be custom menu a-1 and a-2 as far as I can tell it seems to be the same on all Nikons.

 jtan163's gear list:jtan163's gear list
Olympus C-740 UZ Nikon D7000 Olympus OM-D E-M5 Nikon AF-S DX Nikkor 16-85mm f/3.5-5.6G ED VR Nikon AF-S DX Nikkor 35mm f/1.8G +13 more
Reply   Reply with quote   Complain
Truman Prevatt
Veteran MemberPosts: 3,638Gear list
Like?
Re: How accurate should one expect autofocus to be?
In reply to John4, Mar 31, 2013

There are two comments.  First is cars might have a lot more contrast - spectral reflections - than a target.  The second comment is something pointed out by Thom.  He stated that for best results the focal point should cover the object you wish to focus on.  If there it is too small and the area contains other objects, it can get confused.

John4 wrote:

I have perhaps the same or a similar problem.  When I got my D800 I tested the extreme focus points are they were right on.  However in subsequent photographs of my grandson some photos were were noticeably and consistently soft.

I just did an interesting experiment.   I used extreme left focus point and focused on a bar in a window.  The vertical bar went right through the center of the focus area.  In the background, across the street, were some cars.  The camera focused on the distant cars.  Doesn't this violate Greatest Contrast captures focus, or Nearest Object captures focus?  Yet when I focus on something with texture which occupies the whole focus rectangle, the focus is good.

Advise and opinion welcome.

-- hide signature --

Truman
www.pbase.com/tprevatt

 Truman Prevatt's gear list:Truman Prevatt's gear list
Nikon D800E Nikon AF Nikkor 50mm f/1.4D Nikon AF Nikkor 85mm f/1.4D Nikon AF Nikkor 135mm f/2D DC Nikon AF Nikkor 180mm f/2.8D ED-IF +6 more
Reply   Reply with quote   Complain
Just a Photographer
Contributing MemberPosts: 934Gear list
Like?
Re: How accurate should one expect autofocus to be?
In reply to jtan163, Mar 31, 2013

You are referring to focus trap setting.
Nikon unfortunately changed this option and now its less useful.

 Just a Photographer's gear list:Just a Photographer's gear list
Nikon D4 Nikon D800E Fujifilm X-T1 Nikon AF-S Nikkor 70-200mm f/2.8G ED VR Nikon AF-S Nikkor 14-24mm f/2.8G ED +8 more
Reply   Reply with quote   Complain
jtan163
Contributing MemberPosts: 939Gear list
Like?
Re: How accurate should one expect autofocus to be?
In reply to Just a Photographer, Apr 1, 2013

Just a Photographer wrote:

You are referring to focus trap setting.
Nikon unfortunately changed this option and now its less useful.

I thought trap focus was a technique that exploits the focus release feature?

 jtan163's gear list:jtan163's gear list
Olympus C-740 UZ Nikon D7000 Olympus OM-D E-M5 Nikon AF-S DX Nikkor 16-85mm f/3.5-5.6G ED VR Nikon AF-S DX Nikkor 35mm f/1.8G +13 more
Reply   Reply with quote   Complain
Brandon birder
Senior MemberPosts: 2,151
Like?
Re: How accurate should one expect autofocus to be?
In reply to Anker BergSonne, Apr 1, 2013

Anker BergSonne wrote:

The outer focus points work with targets that have good vertical contrast. A vertical bar doesn't have that attribute and it will have difficulty focusing on it. To use a vertical bar as a focus target you should use one of the inner cross-focus points.

Anker

John4 wrote:

I have perhaps the same or a similar problem.  When I got my D800 I tested the extreme focus points are they were right on.  However in subsequent photographs of my grandson some photos were were noticeably and consistently soft.

I just did an interesting experiment.   I used extreme left focus point and focused on a bar in a window.  The vertical bar went right through the center of the focus area.  In the background, across the street, were some cars.  The camera focused on the distant cars.  Doesn't this violate Greatest Contrast captures focus, or Nearest Object captures focus?  Yet when I focus on something with texture which occupies the whole focus rectangle, the focus is good.

Advise and opinion welcome.

For clarity that means a contrasting horizontal bar is a good target for the outer focu points.

Vertical contrast  equals horizontal lines.

Reply   Reply with quote   Complain
Keyboard shortcuts:
FForum MMy threads