fredmiranda.com Login
This topic is in the 2007 archive. You can read it and like its posts here. It is read only.

Archive 2007Cross point AF vs. High Precision x point AF?

  
 
Alan321
Offline
Details
p.2 #1Cross point AF vs. High Precision x point AF?
simonit wrote:
The other difference Canon are claiming on the 40d is 'Additionally the center point is now twice as sensitive as any other point with lenses of F2.8 or faster and has cross-type sensors set at forty-five degrees.'

Have no idea what difference it makes having them in the diagonal plane and what it means in the real world!!


Same as before - the centre sensor is the only high precision one. The others are now cross type instead of linear, so they can see focus on any straight edge instead of just those that are not parallel with the sensor, but they are still not high precision. If they are at 45 degrees then they should better handle the horizontal and vertical lines but any cross-type sensor should do that ok.

- Alan
Aug 20, 2007 at 10:17 AM
Alan321
Offline
Details
p.2 #2Cross point AF vs. High Precision x point AF?
RDKirk wrote:
The depth of focus is the range at the sensor within which a point produces an out of focus blur of less than the circle of confusion (0.035mm for Canon). Depth of focus varies only by aperture, not by focal length or focused distance. An f2.8 200mm lens at 10 feet has exactly the same depth of focus as an f2.8 20mm lens at 50 feet.


In my opinion this has to be at least partly wrong and is hardly surprising given Canon's own ambiguous description of depth of focus. Feel free to prove me wrong if you can (I'd like a definitive answer) but consider this: Single points are hard to find so in practise the AF system is focusing on edges. Some edges must appear to be sharper than others at the sensor if they are closer to being in focus (just commonsense). If that is the case then focus distance must have an effect on the depth of focus too. And if that is the case then depth of field - or factors that influence depth of field - must also affect depth of focus.

The AF system is measuring to within predefined depth of focus limitations but there must be a relationship to practical depth of field, albeit that there can be no allowance for or knowledge of future print enlargement or viewing distance when AF is being assessed.

Now it could be that what Canon refers to as depth of focus is just an arbitrary number such as 0.035mm - an acceptable width of fuzziness around what should be a sharp edge. In that case it is independent of any camera settings unless the focus computer allows for more or less fuzziness at different apertures. However, the actual fuzziness of the edges must depend on aperture and focus distance and actual subject distance, as well as lens quality. Another problem with this definition of depth of focus is that Canon's description in Lens Work III is far more complex, which brings me back to something that relates to depth of field.

A bit worrying is that a fuzziness of 0.035mm at the sensor is about 3 to 5 pixels on a DSLR sensor, and yet if we saw our required critically sharp edges that fuzzy on say a 10"x8" print we'd think the camera had mis-focused. That's why many people regard DOF based on 0.035mm as being overly optimistic by a factor of up to 5.



The 1D-series cameras have a dedicated computer for focusing, which is why they can afford to push high precision with slower lenses (and normal precision with even slower lenses).

Something from Canon that I read recently about the 1Ds2 indicated that the 1D and 1Ds had a single computer for AF and focus, whereas the 1D2 and 1Ds2 had two separate computers, plus a rationalisation of the number of circuit boards and the locations of those boards to improve data transfer speeds. These computers are quite distinct from the DIGIC computer used to process image sensor data.

- Alan
Aug 20, 2007 at 11:08 AM
Alan321
Offline
Details
p.2 #3Cross point AF vs. High Precision x point AF?
LotsToLearn wrote:
Will this greater sensitivity also allow for AF at f/8 with the center point like on 1 series bodies (eg. using a 2x TC on an f/4)?

I know IQ is degraded, just wondering if it's possible.




Nope. Still limited to f/5.6 or brighter for the effective aperture reported to the camera, according to data posted at dpreview. You may or may not get away with it if you use a TC that does not report to the lens. Usually there is significant deterioration in AF speed with the reduced light available to the camera, and if the camera thinks there is no TC then it can make no allowance for it when doing the AF processing, so it is more likely to get it wrong.

- Alan
Aug 20, 2007 at 11:15 AM
Jeff
Offline
Details
p.2 #4Cross point AF vs. High Precision x point AF?
Alan321 wrote:
Wow, that sure was a raw nerve he touched there

With that unresolved AF problem in mind you'd be nuts to rush out for a 1Ds3. I'll put up with my little 1Ds2 for a while. Jeff, I really feel for you - it's usually me who gets all the lemons. Even my 1D2 was a dud until it was serviced by Canon.


Yeah, my sour grapes are mostly directed at Canon's silence on the matter, more than the fact that the camera has a problem. I should probably just cross my fingers and send it to Irvine, see what happens. Someone (can't recall who) did report that Canon cleaned dust out of the AF module in addition to the 'electronic adjustments' and it seemed to help (post-repair), but can't recall the ultimate outcome of that one. I keep hoping for a manufacturing defect to be revealed (rather than a design defect), so that they have a discrete part replacement to fix it once and for all. The longer this goes, the less it leans toward manufacturing and the more it tends toward design... And thus my indirect 'dig' at the newly-announced cameras touting the new AF technology, which in my mind remains unproven until my camera works as advertised.

Fortunately I'm not hankering for a 1Ds MkIII; I'm waiting more for the successor to the 5D, which better be more like the D200 than its current incarnation, which is essentially built to the 30D's standard. I'm not sure why Canon thinks that a $2500+ camera shouldn't be designed to approach the build of the 1-Series bodies, but I guess that's for another thread...

Regards,

Jeff
Aug 20, 2007 at 11:56 AM
RDKirk
Offline
Details
p.2 #5Cross point AF vs. High Precision x point AF?
Now it could be that what Canon refers to as depth of focus is just an arbitrary number such as 0.035mm -

Depth of focus is based on that specific number, however, it's not totally "arbitrary," rather it's based on a historical norm.

The fact that depth of focus is dependent only on aperture is a physical fact. It's easier to draw out than to explain, but the thing to know is that depth of focus--how quickly the size of the out of focus circle increases in size as you move from the image plane--a matter of the acuteness of the apex angle of the focusing cone of light from the optical center of the lens to the image plane.

A 100mm lens at f2.8 will have an optical center twice as far from the image plane as a 50mm lens at f2.8...but will also have twice the optical diameter. In other words, the focusing cone of the 100mm f2.8 lens has twice the hight and twice the diameter, so the angles of the apexes of the focusing cones will be identical, hence the depth of focus will be identical.

Canon uses a circle of confusion of 0.035mm. Their lens formulations are based on that as well as their focusing systems. Back in the manual focus days, the focusing scales on the lenses were based on the same number.

That number isn't arbitrary. Ultimately it's based on one normality: The average human eye has a resolving capability of about 300 lpm at about 20 inches (average comfortable reading distance).

That's their target. Using a circle of confusion of 0.035mm, a 24x36mm image can be enlarged about ten times (8x12 inch print) and the blur circle will be just under the resolving capability of the average human eye at normal reading distances. Actually, the Canon-Europe website quotes their target as a 6x9 print at 9 inches, but I think they're only counting college students in their consideration of "average reading distance."

The AF system is measuring to within predefined depth of focus limitations but there must be a relationship to practical depth of field, albeit that there can be no allowance for or knowledge of future print enlargement or viewing distance when AF is being assessed.

Mechanically speaking--as far as a fairly simple computer can think it through--there is no such thing as depth of field. The depth of focus at a particular f-stop is a single number for any lens; that's one factor. The circle of confusion is a specific number that is constant for the computer. The lens reports the maximum aperture and the computer then has the two hard, distinct numbers it needs to determine it's accuracy tolerance...which will be in movements of fractions of a millimeter at the sensor plane.

Remember the the computer knows nothing about "subject," it only knows about matching out-of-phase light rays to a specific tolerance.

Something from Canon that I read recently about the 1Ds2 indicated that the 1D and 1Ds had a single computer for AF and focus, whereas the 1D2 and 1Ds2 had two separate computers, plus a rationalisation of the number of circuit boards and the locations of those boards to improve data transfer speeds. These computers are quite distinct from the DIGIC computer used to process image sensor data.

I don't know about the orignal 1D and 1Ds, but the Mk II models have separate computers (not the DIGIC) for AF and exposure, where the lesser cameras use one computer for both functions.
Aug 21, 2007 at 05:33 PM
Alan321
Offline
Details
p.2 #6Cross point AF vs. High Precision x point AF?
I still don't see how focus distance or subject distance can be ignored in working out depth of focus. I understand your description of the cone of light from the lens to the focus point being dependant on aperture, and that the diameter of that cone at any distance from the focus plane is also dependent on aperture. However, as that lens is moved back or forward to adjust focus the cone diameter at the sensor changes a little according to the focus distance and so is related to the focus distance. Furthermore, if the subject moves without refocusing the lens then the cone of light representing the subject image is more or less blurred. The extent of blur has to be affected by subject distance as well as aperture.

Thus it seems to me that the depth of focus seen at the sensor (where the light cone diameter is within tolerances) is related to all three of aperture, focus distance and subject distance rather than just aperture.

- Alan
Aug 22, 2007 at 12:25 PM
Canon 10D
Offline
Details
p.2 #7Cross point AF vs. High Precision x point AF?
OSev wrote:
I thought that cross point AF sensors were high precision cross point sensors, but apparently there is a difference between the two. Can anyone describe what the difference is?

Thanks


For the 1D Mark II, 1D Mark II N, and 1D Mark III (not sure about other models), cross point AF sensor and high precision cross point sensor are the same thing, where cross point is AF point with both vertical and horizontal sensitive, also know as high precision. The number of high precision (cross point) AF points, however, in 1D Mark III is different (more) compared with the 1D Mark II and 1D Mark II N.
Aug 22, 2007 at 01:04 PM
FM ForumsCanon ForumJoin Upload & Sell
  
 

You are not logged inLog inRegister