Whay does my Canon 30D's autofocus work so badly on many things?

I'm asking a question.

I used to think that my camera or lenses had a focus problem.

It's a Canon 30D and various lenses. The problem si teh same with
all lenses.

It often simply guesses wrong with autofocus.

So I tested it. I tested it first with flat text material slanted
slightly away from perpendicular to a line from the camera, and
found unreliable autofocus. Then I tried with flat subjects
carefully placed perpendicular to the line of sight, and it works
quite nicely using autofocus ... or manual focus.

I should add that the autofocus is set to use only the center
focus point. The problem is not dependant on whether I'm
using a lens that is faster or slower than f/2.8, which is,
I believe the point at which it switches to "fine focus mode".

But used on real subjects, it simply is poor at autofocus.
Sometimes it will focus in front, sometimes in back.

I read here that this camera in fact lies about the size of the
focus areas: I had assumed that they were the size of the little
squares in the viewfinder, but somebody says that they are
three times larger. That could be the problem, of course, with small
subjects.

How does the autofocus work? I read about "vertical and horizontal lines"
but it seems that the center focus point uses both.

It is simply fooled by things at different distances within its
focusing area? Does it try for an average? The front? The back?
OR is it just junk and I'm going to have to manual focus
on ANYTHING of importance that is not flat and perpendicular
to the line of sight?

What do pros do ... other than of course take types of picture
where the problem does not arise ... it's not a problem
with distant landscapes or portraits (where the person more
than fills any possible center focus area): there the autofocus
works OK.

But say I'm taking picture of geese waddling around in the park.
The goose's head is simply not going to fill the center focus
spot.

Oh yes ... please do not say "you should have bought Nikon".

I'm very happy with the camera otherwise: a lot of the stuff I take
is perfectly obviously not possible for autofocus at all, such
as picture of flowers or bugs, where I have to decide "artistically"
where to focus. But the damn thing should focus on a goose!

Doug McDonald

On Sat, 12 Apr 2008 09:20:49 -0500, lid wrote:

I'm asking a question.

I used to think that my camera or lenses had a focus problem.

It's a Canon 30D and various lenses. The problem si teh same with
all lenses.

It often simply guesses wrong with autofocus.

So I tested it. I tested it first with flat text material slanted
slightly away from perpendicular to a line from the camera, and
found unreliable autofocus. Then I tried with flat subjects
carefully placed perpendicular to the line of sight, and it works
quite nicely using autofocus ... or manual focus.

I should add that the autofocus is set to use only the center
focus point. The problem is not dependant on whether I'm
using a lens that is faster or slower than f/2.8, which is,
I believe the point at which it switches to "fine focus mode".

But used on real subjects, it simply is poor at autofocus.
Sometimes it will focus in front, sometimes in back.

I read here that this camera in fact lies about the size of the
focus areas: I had assumed that they were the size of the little
squares in the viewfinder, but somebody says that they are
three times larger. That could be the problem, of course, with small
subjects.

How does the autofocus work? I read about "vertical and horizontal lines"
but it seems that the center focus point uses both.

It is simply fooled by things at different distances within its
focusing area? Does it try for an average? The front? The back?
OR is it just junk and I'm going to have to manual focus
on ANYTHING of importance that is not flat and perpendicular
to the line of sight?

What do pros do ... other than of course take types of picture
where the problem does not arise ... it's not a problem
with distant landscapes or portraits (where the person more
than fills any possible center focus area): there the autofocus
works OK.

But say I'm taking picture of geese waddling around in the park.
The goose's head is simply not going to fill the center focus
spot.

Oh yes ... please do not say "you should have bought Nikon".

I'm very happy with the camera otherwise: a lot of the stuff I take
is perfectly obviously not possible for autofocus at all, such
as picture of flowers or bugs, where I have to decide "artistically"
where to focus. But the damn thing should focus on a goose!

Doug McDonald

Don't know much about your camera in particular, but generally
speaking, focus goes for areas of contrast differentials. So, if
you're focusing on a black cat on a pile of coal, it isn't going to
*autofocus*

If you're focusing on a snowshoe rabbit on a snowdrift drift on a
cloudy day, it probably won't autofocus.

If you're focusing on a zebra, it should snap to position (the correct
position) in the blink of an eye.

Not seeing your original photo, the one you're having a problem with,
leaves us at a disadvantage. Otoh, there could be a problem with the
camera mechanism itself: mechanical, electronic, whatever.

One suggestion I can make is reload the FIRMWARE for the camera,
and/or *reset* it (which is like a reboot on a computer) and see if
your problem goes away.

Lg

It's a similar technology to the split prisms in an old manual camera.
I'm not sure why you are having problems but perhaps this explains it
better:
http://en.wikipedia.org/wiki/Autofoc...sive_autofocus
"Phase detection"

lid wrote:
I'm asking a question.

I used to think that my camera or lenses had a focus problem.

It's a Canon 30D and various lenses. The problem si teh same with
all lenses.

It often simply guesses wrong with autofocus.

So I tested it. I tested it first with flat text material slanted
slightly away from perpendicular to a line from the camera, and
found unreliable autofocus. Then I tried with flat subjects
carefully placed perpendicular to the line of sight, and it works
quite nicely using autofocus ... or manual focus.

I should add that the autofocus is set to use only the center
focus point. The problem is not dependant on whether I'm
using a lens that is faster or slower than f/2.8, which is,
I believe the point at which it switches to "fine focus mode".

But used on real subjects, it simply is poor at autofocus.
Sometimes it will focus in front, sometimes in back.

I read here that this camera in fact lies about the size of the
focus areas: I had assumed that they were the size of the little
squares in the viewfinder, but somebody says that they are
three times larger. That could be the problem, of course, with small
subjects.

How does the autofocus work? I read about "vertical and horizontal lines"
but it seems that the center focus point uses both.

It is simply fooled by things at different distances within its
focusing area? Does it try for an average? The front? The back?
OR is it just junk and I'm going to have to manual focus
on ANYTHING of importance that is not flat and perpendicular
to the line of sight?

What do pros do ... other than of course take types of picture
where the problem does not arise ... it's not a problem
with distant landscapes or portraits (where the person more
than fills any possible center focus area): there the autofocus
works OK.

But say I'm taking picture of geese waddling around in the park.
The goose's head is simply not going to fill the center focus
spot.

Oh yes ... please do not say "you should have bought Nikon".

I'm very happy with the camera otherwise: a lot of the stuff I take
is perfectly obviously not possible for autofocus at all, such
as picture of flowers or bugs, where I have to decide "artistically"
where to focus. But the damn thing should focus on a goose!

Doug McDonald

On Sat, 12 Apr 2008 15:20:50 GMT, Paul Furman
wrote:

It's a similar technology to the split prisms in an old manual camera.
I'm not sure why you are having problems but perhaps this explains it
better:
http://en.wikipedia.org/wiki/Autofoc...sive_autofocus
"Phase detection"

Or maybe this
http://www.cs.mtu.edu/~shene/DigiCam...uto-Focus.html
"Contrast detection"

Lawrence Glickman wrote:


Not seeing your original photo, the one you're having a problem with,
leaves us at a disadvantage. Otoh, there could be a problem with the
camera mechanism itself: mechanical, electronic, whatever.


Its not ONE photo .... its lots and lots of photos. For example, these:


http://polaris.scs.uiuc.edu/~mcdonald/IMG_4795.jpg
and
http://polaris.scs.uiuc.edu/~mcdonald/IMG_4819.jpg

These are full resolution, slightly cropped.

Note that the first one, autofocus, is focused in back of the
critter of interest ... despite there being plenty of
grass and stuff in front of it to provide contrast. (100 mm lens)

The second, manual focus, is properly focused (300 mm lens).


These results are typical.

It's painful standing there spending time on manual focus, knowing that
the critter will pop down the hole at any instant, and that the mere
motion of my hand doing the focusing may be the cause!

Doug McDonald

On Sat, 12 Apr 2008 11:01:24 -0500, lid wrote:

Lawrence Glickman wrote:


Not seeing your original photo, the one you're having a problem with,
leaves us at a disadvantage. Otoh, there could be a problem with the
camera mechanism itself: mechanical, electronic, whatever.


Its not ONE photo .... its lots and lots of photos. For example, these:


http://polaris.scs.uiuc.edu/~mcdonald/IMG_4795.jpg

See where the maximum contrast is? Where the Groundhog shadow falls
across the sticks to his left side. Perfect focus, if that shadow
edge were the subject ;-( Unfortunately, the groundhog is the
subject.

The DOF (depth of field) is pretty shallow on this photograph, and the
rule of thumb is focus will extend 1/3 in front of the plane-of-focus
and 2/3rds behind it. So what we see, is that the focus zone
(technically, where the circles of confusion are less than 1/2000th of
an inch), the focus zone should extend 2/3rds away from the shadow
line and 1/3rd in front of it. Now let's look at the photo again and
see if this is true:

No, it isn't true. The zone of focus is the strand of light to the
groundhog's left. IOW, the zone of focus is _behind_ the subject.
However! The zone of focus IS the zone (area) with the most contrast.
Look and see. The groundhog isn't even directly illuminated by the
sun? He/she is in a shadow area. There is no direct sunlight on the
groundhog...plenty of direct sunlight on the lighted area to his/her
LEFT.

and
http://polaris.scs.uiuc.edu/~mcdonald/IMG_4819.jpg

In this picture, the fur patter of the animal is camoflage. I.e. it
is the same color and pattern and _intensity_ (hue/saturation) as the
surrounding environment. In biology this is called *protective
coloration.*

Where the max contrast is, is on the forelegs of the creature, and
that is in -perfect- focus. Look at the white hairs on the right
foreleg of our friendly furball.

What I am seeing, is a camera doing what it is designed to do when on
*autofocus.* It is WHY the manufacturer supplies lenses with Manual
Focus Over ride, for situations just like these.

I'm not sure that my *new* (3 week old now out of the box) Nikon would
do any differently in the exact same circumstances.

These are full resolution, slightly cropped.

Note that the first one, autofocus, is focused in back of the
critter of interest ... despite there being plenty of
grass and stuff in front of it to provide contrast. (100 mm lens)

I've noted this, but the max illumination differential is behind the
animal.

The second, manual focus, is properly focused (300 mm lens).


These results are typical.

It's painful standing there spending time on manual focus, knowing that
the critter will pop down the hole at any instant, and that the mere
motion of my hand doing the focusing may be the cause!

Doug McDonald

It takes time to master the machinery. There is no substitute of
experience. Read books forever, but only hands-on teaches these
painful lessons.

My idea is that your brain is a lot more in focus than the camera is
capable of being, and in such circumstances, go to manual over ride.
I suspect there may be some testings you can do in the studio to
verify correct camera function, but these images demonstrate exactly
one of the weaknesses of autofocus, not necessarily a camera defect;
just an inability of the camera to decide what the intended subject
matter is.

Lg

Lawrence Glickman wrote:
On Sat, 12 Apr 2008 15:20:50 GMT, Paul Furman
wrote:

It's a similar technology to the split prisms in an old manual
camera. I'm not sure why you are having problems but perhaps this
explains it better:
http://en.wikipedia.org/wiki/Autofoc...sive_autofocus
"Phase detection"

Or maybe this
http://www.cs.mtu.edu/~shene/DigiCam...uto-Focus.html
"Contrast detection"

The DSLR (such as the 30D of the subject) uses phase-detection rather than
contrast-detection (except sometimes when a live-view function is being
used, which the 30D doesn't have).

David

Lawrence Glickman wrote:
On Sat, 12 Apr 2008 11:01:24 -0500, lid wrote:

Lawrence Glickman wrote:

Not seeing your original photo, the one you're having a problem with,
leaves us at a disadvantage. Otoh, there could be a problem with the
camera mechanism itself: mechanical, electronic, whatever.

Its not ONE photo .... its lots and lots of photos. For example, these:


http://polaris.scs.uiuc.edu/~mcdonald/IMG_4795.jpg

See where the maximum contrast is? Where the Groundhog shadow falls
across the sticks to his left side. Perfect focus, if that shadow
edge were the subject ;-( Unfortunately, the groundhog is the
subject.

Well, yes. BUT ... I sure thought that the little black square was
centered on, and essentially filled by, some part of the goundhog. In that case,
were the obvious true (that the squares are were the autoficus is done),
the camera should eventually have found proper focus (there was plenty
of light on him, it's just that there was far more on the twigs on the ground,
or flashed red at me. It was a test. Do you agree it seems to be
saying that the focus area really IS bigger than the square, as somebody said?
Does Canon say EXACTLY where the autofocus areas are? If I knew the exact answer to
that, especially if the marks were accurate, I'd be in a lot better position
in "quick response needed" situations like the 100mm shot (the 300 mm one
was two days later, and I came prepared, so I get him!)

I do wish the camera had a microprism focus viewfinder like my old Minolta SRT201 and
X700 film cameras!

Doug McDonald

On Sat, 12 Apr 2008 11:55:49 -0500, lid wrote:

Lawrence Glickman wrote:
On Sat, 12 Apr 2008 11:01:24 -0500, lid wrote:

Lawrence Glickman wrote:

Not seeing your original photo, the one you're having a problem with,
leaves us at a disadvantage. Otoh, there could be a problem with the
camera mechanism itself: mechanical, electronic, whatever.

Its not ONE photo .... its lots and lots of photos. For example, these:


http://polaris.scs.uiuc.edu/~mcdonald/IMG_4795.jpg

See where the maximum contrast is? Where the Groundhog shadow falls
across the sticks to his left side. Perfect focus, if that shadow
edge were the subject ;-( Unfortunately, the groundhog is the
subject.

Well, yes. BUT ... I sure thought that the little black square was
centered on, and essentially filled by, some part of the goundhog. In that case,
were the obvious true (that the squares are were the autoficus is done),
the camera should eventually have found proper focus (there was plenty
of light on him, it's just that there was far more on the twigs on the ground,
or flashed red at me. It was a test. Do you agree it seems to be
saying that the focus area really IS bigger than the square, as somebody said?

It is certainly possible. A qualified YES is my answer, except I
can't see how exactly you have programmed the autofocus on the DSLR.
On my camera, mabe a little different than your camera, I can decide
what area(s) I want the focus to think about.

Does Canon say EXACTLY where the autofocus areas are? If I knew the exact answer to
that, especially if the marks were accurate, I'd be in a lot better position
in "quick response needed" situations like the 100mm shot (the 300 mm one
was two days later, and I came prepared, so I get him!)

Good questions for Canon technical support. I'm not familiar with
your particular hardware.

I do wish the camera had a microprism focus viewfinder like my old Minolta SRT201 and
X700 film cameras!

Doug McDonald

That would be a nice feature.

Lg

Lawrence Glickman wrote:

Well, yes. BUT ... I sure thought that the little black square was
centered on, and essentially filled by, some part of the goundhog. In that case,
were the obvious true (that the squares are were the autoficus is done),
the camera should eventually have found proper focus (there was plenty
of light on him, it's just that there was far more on the twigs on the ground,
or flashed red at me. It was a test. Do you agree it seems to be
saying that the focus area really IS bigger than the square, as somebody said?

It is certainly possible. A qualified YES is my answer, except I
can't see how exactly you have programmed the autofocus on the DSLR.
On my camera, mabe a little different than your camera, I can decide
what area(s) I want the focus to think about.


As far as RTFM tells me, I can choose which "spot" to use, or to
use all spots, but not how big the spots are, to autofocus.

I can tell it how big I want the central auto *exposure* spot to be.

Doug