David Langworthy
Well-known member
Does moving down to an f/4 lens completely disable the cross type sensors or is there some drop off, e.g. it only works in bright light? The documentation states f/2.8 as an absolute, but I don't understand why.
-
Welcome to the new forums! Please read this first. For known issues we are working to resolve, click here.
5D Autofocus at F/4?
- Thread starter David Langworthy
- Start date
M
misha
Guest
It works fine, just presumably with less accuracy (1 DOF vs 1/3 DOF?)
--
Misha
jerryk
Veteran Member
Perhaps Canon has determined at f/4 there is not enough contrast to use these sensors. They can detect the aperture of the lens, so they could turn off the sensors.
--
jerryk.smugmug.com
--
jerryk.smugmug.com
David Langworthy
Well-known member
I tried an experiment with 4 parallel lines drawn with a sharpie. Basically, what I determined is that faster lenses focus better than slower lenses and vertical lines resolve with less light than horizontal lines. The vertical sensors (that resolve horizontal lines) do not seem to shut off but they do appear to need an extra stop to work.
I’d love results from a more rigorous test.
My first test with a single line drawn with a fine uniball did not resolve well either horizontally or vertically at f/4, so I switched up to the 4 parallel lines drawn with a sharpie. These resolved almost instantly in either orientation, so I turned the paper over where they were fainter. In this case, the vertical lines resolved but the horizontal lines did not. I tried this same test with an f/1.2 and both orientations resolved almost instantly. (The f/1.2 did a pretty good job of fixing on a blank section of the page.) After I turned a light off, I got the same behavior as the f/4. The vertical lines resolved but the horizontal lines did not.
The f/4 was the 24-105 IS and the f/1.2 was the 85. I focused on something far away in between each trial so the starting points would be fair.
I’d love results from a more rigorous test.
My first test with a single line drawn with a fine uniball did not resolve well either horizontally or vertically at f/4, so I switched up to the 4 parallel lines drawn with a sharpie. These resolved almost instantly in either orientation, so I turned the paper over where they were fainter. In this case, the vertical lines resolved but the horizontal lines did not. I tried this same test with an f/1.2 and both orientations resolved almost instantly. (The f/1.2 did a pretty good job of fixing on a blank section of the page.) After I turned a light off, I got the same behavior as the f/4. The vertical lines resolved but the horizontal lines did not.
The f/4 was the 24-105 IS and the f/1.2 was the 85. I focused on something far away in between each trial so the starting points would be fair.
Rusty1
Senior Member
Misha you'd know more about this than me. I have always wondered if there was much of a real world advantage to this. I cannot say I have ever seen the greater accuracy demonstrated. I've seen F2.8 lens focus faster than F4 but not better to a degree I could measure. I've just reserved judgment on this presumed advantage. Your thoughts very welcome
--
Good luck and good shooting.
Regards Rusty
Rusty1
Senior Member
Very interesting, thanks for posting this.
--
Good luck and good shooting.
Regards Rusty
--
Good luck and good shooting.
Regards Rusty
When the lens is mounted, the camera "interrogates" the lens for its operational parameters, including maximum aperture. If the lens reports a maximum aperture of f2.8 or greater, the camera switches the autofocus system to its "high precision" focusing algorithm.
The "high precision" algorithm uses 0.01mm as the circle of confusion in its calculations instead of 0.03mm.
This is NOT an accuracy bonus. This is more of a "cost of living adjustment," because the faster lens has an inherently more shallow depth of focus and needs an algorithm of greater precision.
By definition, everything within the depth of focus is rendered "sharp enough" in the print (when viewed at the enlargement size the circle of confusion was designated to reach).
It's like shooting at a baloon with a rifle or a shotgun. The rifle requires more precision to hit the baloon at all than the shotgun does, but it doesn't destroy the baloon to any greater extent than any single one of the shotgun pellets striking it with less precision.
--
RDKirk
'TANSTAAFL: The only unbreakable rule in photography.'
The "high precision" algorithm uses 0.01mm as the circle of confusion in its calculations instead of 0.03mm.
This is NOT an accuracy bonus. This is more of a "cost of living adjustment," because the faster lens has an inherently more shallow depth of focus and needs an algorithm of greater precision.
By definition, everything within the depth of focus is rendered "sharp enough" in the print (when viewed at the enlargement size the circle of confusion was designated to reach).
It's like shooting at a baloon with a rifle or a shotgun. The rifle requires more precision to hit the baloon at all than the shotgun does, but it doesn't destroy the baloon to any greater extent than any single one of the shotgun pellets striking it with less precision.
--
RDKirk
'TANSTAAFL: The only unbreakable rule in photography.'
By definition, everything within the depth of focus at the sensor plane will be rendered sharp in the enlargement, provided we're looking at the enlargement for which the circle of confusion was intended.
This is an 8x10 enlargement viewed at 19 inches, for Canon (actually, the Canon-Europa site gives a standard of a 6x9 print viewed at 10 inches).
--
RDKirk
'TANSTAAFL: The only unbreakable rule in photography.'
This is an 8x10 enlargement viewed at 19 inches, for Canon (actually, the Canon-Europa site gives a standard of a 6x9 print viewed at 10 inches).
--
RDKirk
'TANSTAAFL: The only unbreakable rule in photography.'
I'd say less sensitivity. The cross type AF point in the center when shooting at f/2.8 and faster, has two horizontally oriented sensors and one vertically oriented. When shooting at slower f-stops, one of the two horizontal sensors, is deactivated, leaving you still with a cross-type, only supposedly less sensitive.
Dave Largent
Senior Member
So, basically what you're saying is that the lenses slower
than 2.8 just don't need the same degree of accuracy
to make a sharp picture?
than 2.8 just don't need the same degree of accuracy
to make a sharp picture?
Ben_Egbert
Forum Pro
A 500 f4 also has razor thin DOF and needs high precision focusing. They should not allow people to buy a 500 f4 unless they have a camera with the focus system to support it. Or maybe they should make them pass an eye test.
--
http://www.pbase.com/roserus
Ben
It doesn't for the same subject magnification. It has about the same that other focal lengths when not focusing near infinity.
David Langworthy
Well-known member
This makes sense. Just to make sure I understand. For a good image shot at f/8 and with a large print (24x30) viewed up close (say 18") there will be no noticable difference auto focus between the 24-70 f/2.8 and the 24-105 f/4, is this correct?
Ben_Egbert
Forum Pro
A robin at 20 feet is not frame filling with a 500f4. the DOF is .03 feet. If it is off this much, the robin will be soft. This holds for larger birds as well, you need to stop down some anyway, just to get enough DOF, but you hate tossing in another DOF just for focus accuracy.
http://www.pbase.com/roserus
Ben
Canon's autofocus system is based on depth of focus, not depth of field, and the depth of focus of a 500mm f4 lens at any focused distance is exactly the same as the depth of focus of a 50mm f4 lens at any focused distance. The focusing algorithm need change only for aperture--no other factor.
--
RDKirk
'TANSTAAFL: The only unbreakable rule in photography.'
Ben_Egbert
Forum Pro
So my 500f4 ought to be sharp at 20 feet, right? So maybe I need to sell it.
--
http://www.pbase.com/roserus
Ben
Yes, there will in a 24x30 print viewed at 18 inches. But Canon doesn't care about that.
There won't be any difference in a 5x7 print viewed at 10 inches (which is the system design specification standard Canon quotes in their "Lens Work III").
The system is designed around depth of focus--which occurs at the sensor plane. Depth of focus varies by aperture--no other factor.
A large aperture creates a shallow depth of focus at the sensor plane. A small aperture produces a deeper depth of focus at the sensor plane. We're talking about depths of focus of only a few millimeters in the case of a small aperture lens and one millimeter or less with very fast lenses.
When you're talking about getting the focus point correct by fractions of a millimeter, you need a more precise algorithm.
Why not use a more precise algorithm all the time? Because a higher precision algorithm would take more time to calculate. A lower light level takes more time to calculate. Less subject contrast takes more time to calculate.
Canon's engineers probably have figures they've designated as the maximum amounts of time they can allow for focus calcuations, depending on their performance targets for each model of camera.
If they expect the 20D to maintain satisfactory Servo AI focusing at 5fps, for instance, they have to limit the factors affecting focusing speed somewhere. So they don't allow the system to do high precision calculations at less than f2.8.
The 1D cameras, however, have a cpu dedicated for focus calculations that (unlike the lesser cameras) don't have to share processing cycles with other functions.
--
RDKirk
'TANSTAAFL: The only unbreakable rule in photography.'
Depth of field is variable by the final enlargement. Depth of focus is based on the permissible circle of confusion, and that is based on a standard enlargement designated by the designers of the system.
Canon bases all its system designs on a 5x7 print viewed at 10 inches (according to "Lens Work III," page 204).
Based on that enlargement factor, they have set their permissible circle of confusion at 0.035mm (ibid, page 204)
The circle of confusion is a section of the cone of light focusing on the plane of focus with the apex of the cone at the plane of focus. The section varies directly in diameter, of course, with its distance from the apex. Canon considers the image "sharp" as long as the section diameter is no greater than 0.035mm.
A cone with a very large apex angle arrives at the permissible section diameter at a shorter distance from the apex than a cone with a small apex angle.
The only factor that changes the apex angle of the cone of light focusing on the image plane is aperture--because "aperture" is a ratio of the diameter of the base of the cone to the height of the cone, which keeps the apex angle constant when the ratio is constant.
So a 500mm f4 lens has the same depth of focus as a 50mm f4 lens. The fact that both lenses are "f4" means their respective cones of focus have the same apex angle, and thus the dimeter of the cone sections vary by the same amount as the distance from the focused plane varies.
--
RDKirk
'TANSTAAFL: The only unbreakable rule in photography.'
Canon bases all its system designs on a 5x7 print viewed at 10 inches (according to "Lens Work III," page 204).
Based on that enlargement factor, they have set their permissible circle of confusion at 0.035mm (ibid, page 204)
The circle of confusion is a section of the cone of light focusing on the plane of focus with the apex of the cone at the plane of focus. The section varies directly in diameter, of course, with its distance from the apex. Canon considers the image "sharp" as long as the section diameter is no greater than 0.035mm.
A cone with a very large apex angle arrives at the permissible section diameter at a shorter distance from the apex than a cone with a small apex angle.
The only factor that changes the apex angle of the cone of light focusing on the image plane is aperture--because "aperture" is a ratio of the diameter of the base of the cone to the height of the cone, which keeps the apex angle constant when the ratio is constant.
So a 500mm f4 lens has the same depth of focus as a 50mm f4 lens. The fact that both lenses are "f4" means their respective cones of focus have the same apex angle, and thus the dimeter of the cone sections vary by the same amount as the distance from the focused plane varies.
--
RDKirk
'TANSTAAFL: The only unbreakable rule in photography.'
Whilst your description is correct, the loss of the f/2.8 sensor directly equates to less accuracy.
The AF works just like a rangefinder with a baseline which extends over the aperture of the lens. When you have an f/2.8 lens or faster, one of the horizontally aligned AF sensors is comparing light coming in from the vertical peripheries of the f/2.8 aperture, while the other is comparing light coming in from the vertical peripheries of the f/5.6 aperture. The diameter of the f/2.8 aperture is twice that of the f/5.6 aperture, so the f/2.8 AF sensor is therefore working with twice the rangfefinder baseline. Hence the horizontal f/2.8 AF sensors have twice the accuracy of either the vertical or horizontal f/5.6 AF sensors.
If the lens isn't as wide as f/2.8, then that sensor switches off, this reduces sensitivity as you say, but the f/5.6 sensor has only half the accuracy, so you lose both.
If your taking aperture is the same with both lenses (ie. both lenses are stopped down below f/2.8) then that reduced accuracy can be significant.
For a pretty good description of how dSLR AF sensors work, and how they derive from the old split image viewfinders, take a look at this document:
http://doug.kerr.home.att.net/pumpkin/Split_Prism.pdf
Once you understand that, you will see that the AF sensor sensitivity isn't related to the f/# it works at (you don't even get more light on the AF sensor from the faster lens, but its accuracy is. It should also be apparent why the AF stops working above certain f/#s and that this again has nothing to do with the AF sensitivity.
Similar threads
Keyboard shortcuts
- f
- Forum