why doesn't 5DmarkII have 50D's 9 cross AF sensor?

[carlk]

Don't care what (matter of fact I do know what) but I know it won't do for shooting super fast aperture and super thin DOF. Try it yourself before making those tilt your camera bs.

--
Erik

 

[Rcihardc]

Assuming this article is correct (is there an article directly from Canon?):

The high precision sesnors is accurate up to 1/3 DOF of the lens while normal precision is within 1 DOF (which is pretty bad, if this is the standard!). But the article states the same sensors are either single-axis normal precision at f/5.6 or high precision cross type at f/2.8. there is no mention of high precision yet single axis sensor.

Assuming the case that f/2.8 sensor is indeed high precision compared to f/5.6, then 40D wins again because it has not only the f/2.8 sensors mounted at 45 degress but it also has two of them. 5D/5DII have one.

Erik,
Can you forward the article that states 1/3 DOF for f/1.8 vs 1 DOF
for f/5.6 sensor?

This was first mentioned with the description of the EOS-3. Here is one:
http://www.bobatkins.com/photography/eosfaq/eos3af.html

'What do we mean by "normal-precision" vs. "high-precision" AF? A
"normal-precision" AF sensor produces focusing accuracy within the
depth of focus for the lens' maximum aperture. A high-precision" AF
sensor produces focusing accuracy within 1/3 the depth of focus for
the lens' maximum aperture. Therefore, when activated, the
high-precision AF sensors have up to 3 times the precision of the
normal-precision sensors.'

Can you show me where in the 5D white paper it mentions that assist
points are "high precision" ?

Page 8 shows the diagram of which sensors are f/2.8 sensors. The
first paragraph on Page 9 is the one that says "a switch is made to
the f/2.8-sensitive vertical
line-sensitive sensors for high-precision focusing."

But this is the center point, there is no mention that the ASSIST points are high precision- Also note that two of the assist points in 5D are cross type while they are not on 5DII.

--
Erik

 

[john pringle]

Restructuring the EOS 5D may have meant retooling in some areas and saving in others.

It is quite possible from a business standpoint, that it is cost effective to use the same tooling from the last version instead of making up a whole new tool set for a new model, at least for the areas that are still working well from the first model...

Canon also may have made this decision based on all the flack that they got for having to correct focusing issues in their flagship model. So, to return the customer loyalty and respect, it was cost effective to keep the working focus system and put the newest 21MP chip in the working body, with a twist of video mode.

Kind of a safe investment all the make it fall into the price point that pro's can recover their heart for having their awesome 1D MKIII wrapped in a 5D MKII body.
--
I love photography
http://jpringlephoto.com/

 

[RedFox88]

You missed the point and are being picky.

It is NOT going to take another 9 years for and AF upgrade and it is
NOT unreasonable to have expected more from Canon.

I wrote elsewhere recently that people expected the 5D to grow into a 3D over 3 years. People are expecting the sun and moon with every new model and that simply can't happen.

Here's what people want:

d300 body
5D sensor
d3 AF
1Ds3 viewfinder

All for $2000.

 

[Erik Magnuson]

Distance is important but again looking at the diagram,
http://forums.dpreview.com/forums/read.asp?forum=1032&message=29355216

the distance between the f/2.8 sensors is no greater than f/5.6 sensor.

Each sensor consists of two segments. See those all those horizontal lines in the middle of the photo of the sensor? The middle lines are for the f/5.6 segments and the outer horizontal lines are for the f/2.8 segments. That's the increased distance I'm referring to.

well the sensor we are discussing is at the center of the frame so it
is looking at a subject on optical axis

No, there are prisms and lenses in the AF assembly. Please read the following before commenting further:

Principle of the Split Image Focusing Aid and the Phase Comparison Autofocus Detector in Single Lens Reflex Cameras
http://doug.kerr.home.att.net/pumpkin/Split_Prism.pdf

yes, but I am saying the "whole" center point is called high
precision since it has an extra sensor.

As rrcphoto said, it's ability not precision. The cross sensors work under more conditions, but simply rotating the camera can compensate.

it is more likely to focus o na feature you are aiming it at, I think
that is precision.

That's what I meant by you using that word differently than how Canon and I have been using it.

The only theoritical advantage that I can see is that the assist
points will help track the subject in AI-servo mode.

Which is what Chuck Westfall said.

for static shots, 40D/50D will have an edge because of being sensitive to both
horizontal and vertical detail at all points and 45 degree contrast
on center so it is more likely that it will achieve focus.

Only if you are naive about how the AF system works. The workaround is rather trivial.

One more point is that two of the assist points are only active at
f/2.8 so lenses like 70-200 f/4, 24-104, f/4, 100-400L ,400 f/5.6,
300 f/4 and 500 f/4 will not use these points, these are my main
lenses.

No, the center two assist points have f/2.8 horizontal segments but f/5.6 vertical segments. They will still be active in one direction.

--
Erik

 

[Rcihardc]

Distance is important but again looking at the diagram,
http://forums.dpreview.com/forums/read.asp?forum=1032&message=29355216

the distance between the f/2.8 sensors is no greater than f/5.6 sensor.

Each sensor consists of two segments. See those all those horizontal
lines in the middle of the photo of the sensor? The middle lines are
for the f/5.6 segments and the outer horizontal lines are for the
f/2.8 segments. That's the increased distance I'm referring to.

OK, in this case if we assume the center horizontal f/2.8 sensor is high prescision and thus further spaced in X then the two assist points are spaced at the same distance as the center horizontal f/5.6 standard precision sensor, so they are standard precision.

well the sensor we are discussing is at the center of the frame so it
is looking at a subject on optical axis

No, there are prisms and lenses in the AF assembly. Please read the
following before commenting further:

There are lenses and beam splitter (it is actually a 50% beam splitter) but the point is that the path btween the lens and the CMOS sensor is optically equal to the path between lens and AF sensor and thus focus on AF sensor implies focus on image sensor.

Principle of the Split Image Focusing Aid and the Phase Comparison
Autofocus Detector in Single Lens Reflex Cameras
http://doug.kerr.home.att.net/pumpkin/Split_Prism.pdf

yes, but I am saying the "whole" center point is called high
precision since it has an extra sensor.

As rrcphoto said, it's ability not precision. The cross sensors work
under more conditions, but simply rotating the camera can compensate.

yes, if you want to put it that way, but you may not want to take a portrait shot, what do you do? rotate, focus, rotate take the shot? imagine you're on a tripod.

it is more likely to focus o na feature you are aiming it at, I think
that is precision.

That's what I meant by you using that word differently than how Canon
and I have been using it.

The only theoritical advantage that I can see is that the assist
points will help track the subject in AI-servo mode.

Which is what Chuck Westfall said.

Has anyone actually tested to confirm it tracks better?

for static shots, 40D/50D will have an edge because of being sensitive to both
horizontal and vertical detail at all points and 45 degree contrast
on center so it is more likely that it will achieve focus.

Only if you are naive about how the AF system works. The workaround
is rather trivial.

I'm saying that it has 2X high precision sensors at center vs one compared to 5D so at least on paper it is better, and yah you can't rotate the camera everytime you take a shot...

One more point is that two of the assist points are only active at
f/2.8 so lenses like 70-200 f/4, 24-104, f/4, 100-400L ,400 f/5.6,
300 f/4 and 500 f/4 will not use these points, these are my main
lenses.

No, the center two assist points have f/2.8 horizontal segments but
f/5.6 vertical segments. They will still be active in one direction.

that is 5D, in 5DII two of the assist points do not work with the mentioned lenses see below. So if you are a bird shotter and use 500 f/4 you have 4 assist points, still better than 0 in 40D/50D but how much can they help?

http://forums.dpreview.com/forums/read.asp?forum=1032&message=29355216

--
Erik

 

[Erik Magnuson]

Assuming this article is correct (is there an article directly from Canon?):

I think it was in the EOS3 literature. It's in several different non-Canon sources.

The high precision sesnors is accurate up to 1/3 DOF of the lens
while normal precision is within 1 DOF (which is pretty bad, if this
is the standard!).

AF -- like AE -- has never been perfect; the goal was to be "good enough for 8x10 prints" under most circumstances with mid-1980's technology. At least now all Canons except the 1000D have at least one high-precision sensor which offers better performance. But many expect AF to do all of the work. Nope, sorry, you will have to help it out. Even the Nikon 51-pt system has similar limitations (outer points are neither cross nor high-precision.)

because it has not only the f/2.8
sensors mounted at 45 degress but it also has two of them. 5D/5DII
have one.

But the camera only uses one or the other (whichever has a higher confidence solution.) This is just the cross vs. single orientation argument again.

But this is the center point, there is no mention that the ASSIST
points are high precision-

It's the only way to read the plural used in that sentence.

Also note that two of the assist points in
5D are cross type while they are not on 5DII.

The AF systems are likely identical. The specs just don't go into this level of detail.

--
Erik

 

[Erik Magnuson]

Don't care what (matter of fact I do know what) but I know it won't
do for shooting super fast aperture and super thin DOF.

No current phase AF in an affordable full-frame system will do what you ask. You will just have to learn how to use your equipment and depend less on the automation. Or use live view AF where you can put a high-precision point anywhere independent of the lens aperture.

--
Erik

 

[genotypewriter]

Face it... even if it had all 25 or so cross type sensors, it won't be taken seriously as a sports camera because it shoots almost under 4 fps. But with the controversial AF design it'd get the right kind of attention... i.e. by photographers who're mostly wedding/non-action shooters recognise the 5D2 as a camera made especially for them. Take the D3/D700/D300/1D3 etc. with all their supposedly superior AF, they still won't make that particular market group feel like they have a device catered for them.

GTW
--
http://www.flickr.com/genotypewriter

 

[Rcihardc]

Assuming this article is correct (is there an article directly from Canon?):

I think it was in the EOS3 literature. It's in several different
non-Canon sources.

EOS3 is pretty old, I'd search for more recent resources.

The high precision sesnors is accurate up to 1/3 DOF of the lens
while normal precision is within 1 DOF (which is pretty bad, if this
is the standard!).

AF -- like AE -- has never been perfect; the goal was to be "good
enough for 8x10 prints" under most circumstances with mid-1980's
technology. At least now all Canons except the 1000D have at least
one high-precision sensor which offers better performance. But many
expect AF to do all of the work. Nope, sorry, you will have to help
it out. Even the Nikon 51-pt system has similar limitations (outer
points are neither cross nor high-precision.)

But nikon has a dense grid of 15 cross points at center which can be programed in various ways to act as assist points in user defined groups, this helps tracking a lot.

because it has not only the f/2.8
sensors mounted at 45 degress but it also has two of them. 5D/5DII
have one.

But the camera only uses one or the other (whichever has a higher
confidence solution.) This is just the cross vs. single orientation
argument again.

even if it uses only one at a time (where is this from?) it still has as many as 5D and it is a better layout because 45 degree can detect both horizontal and vertical detail-5D is in fact identical to 20D and 40D was an upgrade to 20D.

But this is the center point, there is no mention that the ASSIST
points are high precision-

It's the only way to read the plural used in that sentence.

"The center AF point has a special hybrid design. With f/2.8 or faster lenses, focusing is a two-step process. First, the f/5.6-sensitive cross-type sensor components are used to focus. When focus is almost achieved, a switch is made to the f/2.8-sensitive vertical line-sensitive sensors for high-precision focusing"

The plural here does not refer to assit points, it refers to the f/2.8 center horizonta sensors, which come in a pair and thus, sensor"s"-the assist points are mentioned later and there is no mention of high precision, Canon marketing would not hesitate to use high-precision if that was a fact

"The two Supplemental AF points directly above and below the center are also
f/2.8-sensitive when a lens of f/2.8 or brighter is used."

Also note that two of the assist points in
5D are cross type while they are not on 5DII.

The AF systems are likely identical. The specs just don't go into
this level of detail.

That is our or at least my problem, I wanted a better AF than 5D.

--
Erik

 

[Hugo1234]

Nothing

 

[Erik Magnuson]

OK, in this case if we assume the center horizontal f/2.8 sensor is
high prescision and thus further spaced in X then the two assist
points are spaced at the same distance as the center horizontal f/5.6
standard precision sensor, so they are standard precision.

That's a lot of work and assumptions to deny the obvious interpretation. Why would canon bother making longer baseline f/2.8 sensors if there were not going to use the precision?

There are lenses and beam splitter (it is actually a 50% beam
splitter) but the point is that the path btween the lens and the CMOS
sensor is optically equal to the path between lens and AF sensor and
thus focus on AF sensor implies focus on image sensor.

Sigh. Look at what's inside the AF assembly. Read the reference I gave.

yes, if you want to put it that way, but you may not want to take a
portrait shot, what do you do? rotate, focus, rotate take the shot?
imagine you're on a tripod.

If I'm on a tripod, I'll use live view AF or MF. As long as I have a good idea of what the AF system will or will not focus on, I can compensate. (Hint: the rectangles in the viewfinder tell you which way the sensor works.)

Has anyone actually tested to confirm it tracks better?

RG's comment following the Westfall quote suggested that was his experience.

I'm saying that it has 2X high precision sensors at center vs one
compared to 5D so at least on paper it is better,

A lot of this is comparing paper specs. 51 > 15, so it must be better. 11 cross-points is better than 8 non-cross + 1 cross. While this may be true to a limited extent, it does not cover all of the potential differences in the AF systems. It also leaves the question "how much better and when" mostly unanswered.

that is 5D, in 5DII two of the assist points do not work with the
mentioned lenses see below.
http://forums.dpreview.com/forums/read.asp?forum=1032&message=29355216

The photo of the 5D2 AF sensor shows the same configuration as the 5D sensor. So either the 5D2 diagram is wrong or the original 5D diagram is wrong or one the photos are wrong. It's not a big deal in any case.

--
Erik

 

[Erik Magnuson]

EOS3 is pretty old, I'd search for more recent resources.

It's the system that first introduced the "high-precision" sensors.

But nikon has a dense grid of 15 cross points at center which can be
programed in various ways to act as assist points in user defined
groups, this helps tracking a lot.

Sure it probably helps tracking. But it doesn't help off-center single point focus with shallow DOF lenses which is the biggest complaint voiced in this thread. Nor does the 50D AF help with tracking.

it is a better layout because 45 degree can detect
both horizontal and vertical detail

It's better on paper. In typical usage, the reports are mixed. There is more to AF than just a few specifications of number of points and orientation.

The plural here does not refer to assit points, it refers to the
f/2.8 center horizonta sensors, which come in a pair and thus,
sensor"s"-the assist points are mentioned later and there is no
mention of high precision, Canon marketing would not hesitate to use
high-precision if that was a fact

As I said, that's a rather contrived reading. You've yet to explain what other reason there might be for f/2.8 sensors.

That is our or at least my problem, I wanted a better AF than 5D.

Well, you didn't get it this time. No camera has everything.

--
Erik

 

[Rcihardc]

OK, in this case if we assume the center horizontal f/2.8 sensor is
high prescision and thus further spaced in X then the two assist
points are spaced at the same distance as the center horizontal f/5.6
standard precision sensor, so they are standard precision.

That's a lot of work and assumptions to deny the obvious
interpretation. Why would canon bother making longer baseline f/2.8
sensors if there were not going to use the precision?

I am saying that the distance between the assist f/2.8 sensors is the same as the center horizontal f/5.6 sensor, you are saying that they are high prescision because they are further spaced, but they are not. My point is the assist sensors are not high prescion, and it doesn't make sense for them to be, because they are only active in AI-servo, if they are too sensitive they will just lock on BG.

There are lenses and beam splitter (it is actually a 50% beam
splitter) but the point is that the path btween the lens and the CMOS
sensor is optically equal to the path between lens and AF sensor and
thus focus on AF sensor implies focus on image sensor.

Sigh. Look at what's inside the AF assembly. Read the reference I
gave.

I did and did not find any argument contradicting what I say.

yes, if you want to put it that way, but you may not want to take a
portrait shot, what do you do? rotate, focus, rotate take the shot?
imagine you're on a tripod.

If I'm on a tripod, I'll use live view AF or MF. As long as I have a
good idea of what the AF system will or will not focus on, I can
compensate. (Hint: the rectangles in the viewfinder tell you which
way the sensor works.)

I don't like the idea of rotating the camera, there is not enough time if you want to capture a critical moment. and if I am to manually compensate I just use manual focus why bother with AF?

Has anyone actually tested to confirm it tracks better?

RG's comment following the Westfall quote suggested that was his
experience.

I'm saying that it has 2X high precision sensors at center vs one
compared to 5D so at least on paper it is better,

A lot of this is comparing paper specs. 51 > 15, so it must be
better. 11 cross-points is better than 8 non-cross + 1 cross.
While this may be true to a limited extent, it does not cover all of
the potential differences in the AF systems. It also leaves the
question "how much better and when" mostly unanswered.

yah, it is hard to quantify how much faster or slower 5D AF is comapred to 40D, it is harder to think of a valid test

that is 5D, in 5DII two of the assist points do not work with the
mentioned lenses see below.
http://forums.dpreview.com/forums/read.asp?forum=1032&message=29355216

The photo of the 5D2 AF sensor shows the same configuration as the 5D
sensor. So either the 5D2 diagram is wrong or the original 5D diagram
is wrong or one the photos are wrong. It's not a big deal in any
case.

maybe, but then again if white paper is wrong there is no point in discussing this because we might both be wrong.

--
Erik

 

[WalterSrChat]

40D and 50D both have the "new" AF system with 9 AF points having
cross sensor. meanwhile the 5D mark II still keeps the 1 cross sensor
in the center and has 8 points which are not cross type. any reason
for this? any vantages?

so far i couldn't find an answer here on the forum.

Have you tried asking Canon?
Canon will answer your question, if you are a Canon owner.

regards
thomas

--
visit my homepage http://thomas.im
portrait - studio - streetlife

--
I am out to take the perfect picture, if it exits!

 

[Rcihardc]

EOS3 is pretty old, I'd search for more recent resources.

It's the system that first introduced the "high-precision" sensors.

But nikon has a dense grid of 15 cross points at center which can be
programed in various ways to act as assist points in user defined
groups, this helps tracking a lot.

Sure it probably helps tracking. But it doesn't help off-center
single point focus with shallow DOF lenses which is the biggest
complaint voiced in this thread. Nor does the 50D AF help with
tracking.

OK, it's a valid point however I do happen to have a D300 with 17-55 (the only nikon lens I have) most likely I am going to sell these and get a D700 (D300 is noisy), especially if I don't like the new 5D. D300 and D700 use identical AF, perhaps not exactly the same but identical, I can claim with confidence that peripheral points on D300 are very accurate despite not being cross type, I have not had a single OOF shot from this camera that was taken in reasonable conditions, sometimes in dim light it just doesn't lock but when it does it is perfect. My 40D just locks, sometimes it is OOF in bright daylight contrasty scene.

it is a better layout because 45 degree can detect
both horizontal and vertical detail

It's better on paper. In typical usage, the reports are mixed. There
is more to AF than just a few specifications of number of points and
orientation.

yah, but based on my experience with 20D and later 40D, 40D peripheral AF points are more reliable, I can now use them and expect 60-80% hit rate with 20D it was a no go especially in dim light. now if 5D II is like 20D....

The plural here does not refer to assit points, it refers to the
f/2.8 center horizonta sensors, which come in a pair and thus,
sensor"s"-the assist points are mentioned later and there is no
mention of high precision, Canon marketing would not hesitate to use
high-precision if that was a fact

As I said, that's a rather contrived reading. You've yet to explain
what other reason there might be for f/2.8 sensors.

I don't know, but the text doesn't imply that either. explaining what they are for doesn't seem easier to me than explaining why Canon did not say high precision and why they are not spaced further like the center sensor.

That is our or at least my problem, I wanted a better AF than 5D.

Well, you didn't get it this time. No camera has everything.

true, true, but as a faithful Canon user I was expecting more from this company. Imagine Canon had made 5DII with 1D AF, small shutter lag and 5fps, how many people would stand in the line to buy this camera?

--
Erik

 

[Erik Magnuson]

doesn't make sense for them to be, because they are only active in
AI-servo, if they are too sensitive they will just lock on BG.

They are only getting a longer baseline look at the same subject. They are neither more or less likely to lock onto the background.

I did and did not find any argument contradicting what I say.

Why do you think Mr. Kerr's paper combined the descriptions of split-prism and phase AF?

I don't like the idea of rotating the camera, there is not enough
time if you want to capture a critical moment.

Fortunately most peoples' faces have both vertical and horizontal detail (nose and lips) so it's not that common a problem. And if you want to focus on the eyes, well they are round....

--
Erik

 

[carlk]

The center high precision AF point can do it when I use AI servo but I have to put the object right in the middle of the frame with no other framing possibilities. I think I know how to use the equipment better than you do. From what you said it will be a disaster to put the 85/1.2 in your hand. I hope you really don't mean AF with live view. So much for theory and speculation.

Don't care what (matter of fact I do know what) but I know it won't
do for shooting super fast aperture and super thin DOF.

No current phase AF in an affordable full-frame system will do what
you ask. You will just have to learn how to use your equipment and
depend less on the automation. Or use live view AF where you can put
a high-precision point anywhere independent of the lens aperture.

--
Erik

 

[Rcihardc]

doesn't make sense for them to be, because they are only active in
AI-servo, if they are too sensitive they will just lock on BG.

They are only getting a longer baseline look at the same subject.
They are neither more or less likely to lock onto the background.

well the size is less compeling because the center f/5.6 horizontal also has a long base line. so do the peripheral ones.

I did and did not find any argument contradicting what I say.

Why do you think Mr. Kerr's paper combined the descriptions of
split-prism and phase AF?

I have no problem with the article.

I don't like the idea of rotating the camera, there is not enough
time if you want to capture a critical moment.

Fortunately most peoples' faces have both vertical and horizontal
detail (nose and lips) so it's not that common a problem. And if you
want to focus on the eyes, well they are round....

In general I have found that cross sensors of 40D work better than single axis sensors of 20D, for a variaty of subjects

--
Erik

 

[thx1138]

You missed the point and are being picky.

It is NOT going to take another 9 years for and AF upgrade and it is
NOT unreasonable to have expected more from Canon.

I wrote elsewhere recently that people expected the 5D to grow into a
3D over 3 years. People are expecting the sun and moon with every
new model and that simply can't happen.

Here's what people want:

d300 body
5D sensor
d3 AF
1Ds3 viewfinder

All for $2000.

The customer is always right (unless they are wrong)

No here's what people want 5D II + better AF (not 1 series) for drumroll please .... $2799

Would it have hurt them to spend a bit of time on the AF, would it have hurt them to make the centre point at least x-type to f/5.6 and improve the AI servo.