How does EOS AF work? You think you know?

It strikes me that your samples show a slanted field with vertical
lines, which is about the worst thing to feed a camera's
autofocussing system.
Somewhere in the back of my mind something tells me that such a
shot has no chance unless you focus on a HORIZONTAL line maybe next
to what you want to shoot and then reframe for the real shot whilst
keeping that same focus.
Give it a try because it seems to me that you don't give your
camera any chance at all.
It depends on which AF arrays you're using. In the xxxD and xxD cameras, the horizontally oriented rectangles see vertical liines only--they are blind to horizontal lines.

The xD cameras have cross sensors toward the center that can see vertical lines.
--
RDKirk
'TANSTAAFL: The only unbreakable rule in photography.'
 
But what is the "service CD" and where do one get it?
The service CD is the CD full of software that Canon uses to do such things as:

-AF calibration
-exposure and flash exposure adjustments (in 1/6th stops as I recall)
-check (and clear) the real total shutter count

-various sensor level calibrations... this is all unique to the DSLRs whereas the stuff above hasn't really change between old film SLRs and the ones of today

If you have a newer camera, you're not likely to find the service CD in the wild as Canon realized people were getting them and put a stop to it. Canon USA doesn't even let any of its authorized centers except for Irvine and New Jersey to have them. If someone were to decode the USB protocol (which they've already done for basic functions like seeing files on the internal flash drives) they could do all the same functions without this software.

Jason
 
of your subject in order to find necessary contrast?
This will confuse the AF into focusing on the backround, you have
to find contrast within the plane of the area you want in focus,
ie, somewhere in the middle, avoid the edges.
I always paid attention to whether there is only one contrasty thing within AF sensor, so I tried to give AF the best possible conditions. But I just learned that AF sensors are actually bigger than the indocators. I'll have to do my tests again.

--
Regards, Wojtek
 
Because it's faster this way.
The camera knows that if the lens does what it's told to do, the
focus will be "good enough".
No... We're in MF mode, there's no need to do anything fast. But
I'm not talking about the MF confirmation. I don't know how that
works...
If someone did know, it could throw some light on the issue. But I'd expect that AF check of whether in focus or not is the same in both modes, AF or MF. So if it confirms wrongly good focus in MF, it probably does the same in AF, resulting in random focus results.
Anyway, to me it's quite evident that the camera does not actually
check focus using the AF sensors on the final resting point, or it
would not fail on some lenses while having reliable results with
others!
Yes, that theory does make sense.
The AF sensors are actually quite straightforward devices:
basically a line (or two crossed lines) of cells that detect the
contrast across their length. When maximum contrast is achieved
(expecially easy to understand if you imagine the sensor line
crossing an edge of the projected image), the sensor gives maximum
output; when all the cells of the sensor give similar readings (low
contrast), the sensor gives a low output.
Yes, I know that, and in MF mode it also should work this way.

--
Regards, Wojtek
 
I thought everyone knew SLRs used phase detection instead of the
contrast (aka sharpness) detection you describe. It's right in the
specifications for sure.
Jason, thanks for your input! I'll have to read it carefully yet to make sure I understand it all.

But knowing all this about how AF works, could you risk an interpretation why 300D + 24-70/2.8EX confirms focus in MF mode seeing both images below?





This is 100% crop of the image, roughly the area around central AF sensor, which was the only one activated, no reframing after focusing.

--
Regards, Wojtek
 
If someone did know, it could throw some light on the issue. But
I'd expect that AF check of whether in focus or not is the same in
both modes, AF or MF. So if it confirms wrongly good focus in MF,
it probably does the same in AF, resulting in random focus results.
Sounds right. I did some tests on my 18-125 comparing true AF and MF with focus confirmation, and indeed the AF errors are very similar: the lens frontfocuses in AF, and gives erroneous focus confirmation in MF when I focus in front of the subject...

I resorted to MF by eye (I'm good enough at that) with this lens at > 50mm focal length, but for sure it isn't a solution, only an uncomfortable patch. :(

Fernando
--
Portfolios: http://www.usefilm.com/photographer/18417.html
http://www.bytephoto.com/photopost/showgallery.php?ppuser=1468
 
If someone did know, it could throw some light on the issue. But
I'd expect that AF check of whether in focus or not is the same in
both modes, AF or MF. So if it confirms wrongly good focus in MF,
it probably does the same in AF, resulting in random focus results.
Sounds right. I did some tests on my 18-125 comparing true AF and
MF with focus confirmation, and indeed the AF errors are very
similar: the lens frontfocuses in AF, and gives erroneous focus
confirmation in MF when I focus in front of the subject...
Yes, so far, everything understandable. But what is not understandable in my case, is that 300D + 24-70 confirms focus not in specific point, but in some range. This is what I cannot understand...
I resorted to MF by eye (I'm good enough at that) with this lens at
50mm focal length, but for sure it isn't a solution, only an
uncomfortable patch. :(
Unfortunately, I don't get impressive results in MF by eye only :)

--
Regards, Wojtek
 
I thought everyone knew SLRs used phase detection instead of the
contrast (aka sharpness) detection you describe. It's right in the
specifications for sure.
Jason, thanks for your input! I'll have to read it carefully yet to
make sure I understand it all.

But knowing all this about how AF works, could you risk an
interpretation why 300D + 24-70/2.8EX confirms focus in MF mode
seeing both images below?





This is 100% crop of the image, roughly the area around central AF
sensor, which was the only one activated, no reframing after
focusing.

--
Regards, Wojtek
--
RDKirk
'TANSTAAFL: The only unbreakable rule in photography.'
 
Here you are, downresed to 25%:

in focus:





out of focus:





If transfer limit on my free account runs out and you cannot see pics, I'm sorry, but for some time there should be no problems.

--
Regards, Wojtek
 
AFAIK, AF confirmation is determined at focus point of maximum contrast. When there is little or no contrast (e.g. picture of only white wall), contrast is hard to detect and AF system might not lock/confirm or might be fooled.

For the sample pics in previous post which has large areas of low contrast, the AF sensors might be positioned over the low contrast areas, making it hard for AF system to distinguish truly focused from merely the computed point of maximum contrast, but still within calibrated thresholds.
In both cases, camere confirmed correct focus in MF mode. It may
lie within DoF theoretically, but it is clearly out of focus when
you take a look at 100% magnification.
Depth of field varies inversely with the final viewing
magnification. Canon designs its af systems around a final viewing
magnification of A5 to A6. Viewing an image on a monitor at 100
percent of the pixel resolution is like burying your nose into a A1
print.

Check out the tips for judging AF accuracy in Canon's own words:

http://www.photoworkshop.com/canon/EOS_Digital.pdf

--
RDKirk
'TANSTAAFL: The only unbreakable rule in photography.'
--
[email protected]
 
Here you are, downresed to 25%:

in focus:





out of focus:





If transfer limit on my free account runs out and you cannot see
pics, I'm sorry, but for some time there should be no problems.

--
Regards, Wojtek
I don't even see a near miss on the second image, which leads me to believe that something else went wrong.

--
RDKirk
'TANSTAAFL: The only unbreakable rule in photography.'
 
Where do you get these service CDs?
Never mind, I read the posts on this topic. It's too damn bad that Canon keeps them under wraps. I really wish these companies would take a more open approach. Yes, it would cost them some money, but there would also be significant benefits.

--
Ben (an open-source person)
 
Seems you got a hold of one, any chance I could get a copy?
But what is the "service CD" and where do one get it?
The service CD is the CD full of software that Canon uses to do
such things as:

-AF calibration
-exposure and flash exposure adjustments (in 1/6th stops as I recall)
-check (and clear) the real total shutter count
-various sensor level calibrations... this is all unique to the
DSLRs whereas the stuff above hasn't really change between old film
SLRs and the ones of today

If you have a newer camera, you're not likely to find the service
CD in the wild as Canon realized people were getting them and put a
stop to it. Canon USA doesn't even let any of its authorized
centers except for Irvine and New Jersey to have them. If someone
were to decode the USB protocol (which they've already done for
basic functions like seeing files on the internal flash drives)
they could do all the same functions without this software.

Jason
 
It's great to see such a detailed explanation. The one part I didn't see is the loop. If the focus is off by much but detectable by the AF the AF algorithm does not rely on the lens making the correct move in one go. The lens is ordered to move some (most of what is needed) and then the focus is checked and the lens is ordered to move again. With slower moving AF you can actually sometimes see this as the move/stop/move/stop behavior.

The crux of the problem with some lenses (Sigma for instance) is that the final step is always a move command, followed by the camera confirming focus without checking focus. When the lens performs the wrong move (because the lens manufacturer doesn't know exactly what the body's command means) the camera will confirm focus while the image is still not focused correctly.

Now there's more details that we don't understand, because if it were all as simple as this we wouldn't see strange things like Sigma lenses focusing being completely off when the AF assist on an external flash is used. But the explanation helps a lot to understand that the AF is no magic wand and to understand what to expect from it.

--
Still learning to use the DRebel (only around 11.000 shots)
Public pictures at http://wwwis.win.tue.nl/~debra/photos/
 

Keyboard shortcuts

Back
Top