[lol101]

All Pentax AF points since 1981 are F:2.8 sensitive (the ME-F had users selection between F:2.8 and F:5.6). Thats why Pentax AF is perceived as slower under certain condition as ir demands higher precision levels.

If all AF points were f2.8 sensitive, how would Pentax cameras be able to AF with f4 or f5.6 maximum aperture lenses?

Are you sure about this info? Do you have any source?

My understanding is that Pentax AF points are sensitive to f5.6 (at best) on the latest SAFOX systems.

---

Last edited by lol101; 01-23-2009 at 10:02 AM.

[m8o]

On one hand it seems pretty contrary to logic, as if you happen to have an eBay search saved that returns all classic long lenses, and monitor it daily for years, you'll conclude that Pentax had been the king of f/3.5 or f/4 lenses (and plenty slower) a few decades back.

On the other hand then I realized, those are mostly manual lenses! Uh duh to me. It does appear just working from my head and memory of the bdmitrov (sic) site, the lenses have on average become faster since AF became the norm.

[Pål Jensen]

If all AF points were f2.8 sensitive, how would Pentax cameras be able to AF with f4 or f5.6 maximum aperture lenses?
latest SAFOX systems.

It has nothing to do with ability to focus but precision level. Any manufacturer stating that they have optimized (a few) sensors to F.2.8, are in reality saying that the rest of the sensors has too low precision to be trusted with precise focusing at aperture as wide as F2.8.

The AF systems works by having sensors in front and behind the plane of focus. These sensors will provide two curves that plots contrast vs. distance. The plane of focus is where those two curves cross. The shape of the curves is dependent on DOF/Aperture. A smaller aperture (like F:5.6) gives a wide curve with a flatter crest; a larger aperture give a steep curve with a prononced top point. Finding the top point on "flat" crest does not demand much precision as deviation doesn't make much difference; if the curve is too wide, the system cannot reliably find any top pont at all (hence; no AF at very small apertures). However, for a steep curve finding the exact top point is of paramonut importance. A small deviation will give fuzzy images.

Theres a misunderstanding that Pentax AF do an extra check for AF; it does not - it is just that Pentax AF system is opimized for exactness over speed. It is about a ballpark figure (good enough for F:5.6) vs an "exact" figure.

[falconeye]

If all AF points were f2.8 sensitive, how would Pentax cameras be able to AF with f4 or f5.6 maximum aperture lenses?

Are you sure about this info? Do you have any source?

My understanding is that Pentax AF points are sensitive to f5.6 (at best) on the latest SAFOX systems.

I think what Pål Jensen meant is that the Pentax AF is calibrated to render sharp images at f/2.8 (while still working at f/5.6, right).

However, I don't see why this would be required when the camera knows at which aperture it will shoot.

Some brand (Sony?) has a dual implementation of the central AF point, one only working with wide apertures for improved speed/accuracy. I hope that Pentax will adopt this trivial idea.

[falconeye]

The AF systems works by having sensors in front and behind the plane of focus.

This is not correct. It works exactly like the split screen focus indicator by comparing two (1-dimensional) images captured from two distant ray beams. The distance is such that the AF "blacks out" for apertures smaller than f/5.6.

Accuracy/speed could be improved by taking rays from more distant ray beams at the prize of blacking out sooner, e.g., at f/2.8. Hence the dual AF sensor to overcome this.

With a single AF sensor, your narrow/wide curve is always the same -- it is just the question of how accuratly you try to find its maximum.

[deejjjaaaa]

The AF systems works by having sensors in front and behind the plane of focus.

well known text :: http://doug.kerr.home.att.net/pumpkin/Split_Prism.pdf

[Pål Jensen]

This is not correct. It works exactly like the split screen focus indicator by comparing two (1-dimensional) images captured from two distant ray beams. The distance is such that the AF "blacks out" for apertures smaller than f/5.6.

Accuracy/speed could be improved by taking rays from more distant ray beams at the prize of blacking out sooner, e.g., at f/2.8. Hence the dual AF sensor to overcome this.

With a single AF sensor, your narrow/wide curve is always the same -- it is just the question of how accuratly you try to find its maximum.

Huh? Pentax own words:

"...placing two sensors, one minutely in front, the other just to the rear, of the in-focus position. When both sensors detect the same degree of contrast, the in-focus position between the two is said to have been reached."

More:

"this is accomplished by using a beam-splitter to divide the beam in two so that it hits both the "a" and "b" (sensors). These two arrays are placed at points equidistant, before and after, the in-focus position at the bottom of the camera, equal in distance to the actual film plane."

About the 2.8 vs 5.6 aperture:

"This sets the contrast width of both "in focus" zone and the outer limit where the twin arrow lights (AF OK light) will light so that so that they match the lens in use, assuring a higher degree of focusing accuracy"

Pretty much what I have line out in the reply above....

[Pål Jensen]

Oops. Seems like I've gotten it slightly wrong!
After fine reading the technical paper it seems like it is the other way around: F5.6 needs higher precision than F:2.8. Everything else I've said is correct.
This is due to the fact that more relative depth of field make the contrast fluctuations against the light delivery amount decrease.


It seems like theres no such thing as a F:2.8 AF sensor; it is just software and the error levels used in processing the data. Hence, I believe that F:2.8 sensors are just markeing talk for something embedded in all AF systems.

[lol101]

Oops. Seems like I've gotten it slightly wrong!
After fine reading the technical paper it seems like it is the other way around: F5.6 needs higher precision than F:2.8. Everything else I've said is correct.
This is due to the fact that more relative depth of field make the contrast fluctuations against the light delivery amount decrease.


It seems like theres no such thing as a F:2.8 AF sensor; it is just software and the error levels used in processing the data. Hence, I believe that F:2.8 sensors are just markeing talk for something embedded in all AF systems.

Quote :

If the two “virtual AF apertures” are spaced further apart than the full diameter of
the exit pupil (as a result of our mounting a lens having a rather small maximum
aperture), there are no blue or red rays that can pass through to the detectors, and
focus detection will fail.
To avert this, the focus detectors are usually designed to have virtual AF apertures
near the edges of the exit pupil for a modest camera aperture—an aperture likely to
be available on most lenses to be used on the camera. Of course, this compromise
(a “reduced baseline”) results in a reduction in the precision of the focus detection
system.
In some cameras (such as the Canon EOS 20D), at one or more particularlyimportant
AF point locations across the frame (typically the one at the center), two
autofocus detectors are provided, one workable for modest aperture lenses and one
(with a “larger baseline”) workable only for larger aperture lenses. The appropriate
one is put into action based on the maximum aperture reported by the lens in place.
This arrangement exploits the greater precision in focus determination available
with a larger aperture lens in place.

End quote.

This is what I was talking about. Double sensors able to use the additional baseline width available (and necessary) to accurately focus f2.8 and larger aperture lenses.

This is not a software trick, nor marketing talk, believe me, it works.

[deejjjaaaa]

http://www.robgalbraith.com/public_files/Canon_Rebel_XTi_White_Paper.pdf

The Digital Rebel XTi has the same 9-point AF sensor and
AF unit as the EOS 30D, including the sophisticated dualprecision
cross-type sensor in the center AF position.
This represents a very substantial upgrade from the
Digital Rebel XT. The XTi is the first entry-level EOS Digital
camera to have an enhanced precision cross-type, center
AF point compatible with f/2.8, a feature unique in its
class. The XTi has some significant AF performance
improvements over the Digital Rebel XT. For example, the
XTi has the same wide-area focusing as the EOS 30D.
With an f/2.8 or brighter lens, the center AF point detects
focus with the f/2.8 light flux. The base line of the center
AF point's vertical-line sensitive sensor is twice as long
as that of an AF point compatible with f/5.6. This makes
focusing detection more precise.
Additionally, the center AF point has a vertical-line sensitive sensor compatible with
f/5.6. Therefore, cross-type focusing is possible with any EF lens. The center AF point's
horizontal-line sensitive sensor for f/5.6 has a two-line, zigzag pattern. This redundancy
of focusing methods makes focus detection more consistent. As in the case of the EOS
30D, fine-tuning has been incorporated to improve the AF precision. Focusing performance
under low light has been improved by 1 stop and is now possible from EV -0.5 to EV 18. Finally, the wider focusing area with nine AF points and the diagonal configuration
of AF points in and around the center afford improved subject focus tracking.

[Pål Jensen]

Quote :
This is what I was talking about. Double sensors able to use the additional baseline width available (and necessary) to accurately focus f2.8 and larger aperture lenses. .

All AF sensors are double. This is how the system knowns which direction focus lies by comparing the two. What precision (aperture) is decided by software. Which AF sensor to do what is also software controlled.
All AF points on Pentax cameras are 2.8 sensitive; there are really no such thing as we are just talking level of inaccuracy. It seems like Canon have increased the baseline on central AF point in order to reduce the inaccuracies. However we do not know the relative size of Pentax sensors anyway.
The value of this, except for marketing babble, is unclear to me....
What Canon is doing is adding an AF point good enough for F:2.8 and saying that the rest of the sensor cannot be trusted with such a task. Whether this is good or bad compared to how Pentax is doing it is debatable....

[vinzer]

vinzer, I totally agree. In body SR is one of the features that are a deal breaker for me as well. Not that I would but if I had to choose another body, I'd take a hard look at Sony for that alone.

A *HUGE* reason of why I upgraded to the K10D from a 300D back then instead of the 30D is in-body shake-reduction. Especially more so since I was in college back then, and Canon IS lens prices really put me off.

I hope no one gets started about the cost of Canon's 70-200 f/4 IS and the MSRP of Pentax's upcoming(?) 60-250 f/4 lens. What I'm talking about is the 75-150 f/4 I got for $55. Good focal length, terribly good performance, and with IS to boot. I can do without AF and auto-aperture, I shoot static subjects anyway.

A Sony body might be a good addition, if ever, coupled with Minolta lenses, of which most lenses are reasonable in price still. They really must have made a lot of those.

I own a couple Samsung K Mount 6mp dslrs too. My first two dslrs purchases. Neither has shake reduction and I really wanted a shake reduced body to see what I could do with my Pentax glass and made for Pentax mount lenses. Having K20D inhand I have to say its way more feature rich than my 5D but the D700 is pretty feature rich too, however its 3 times the money, soon to be 4 times the money. and of course both are full frame with bigger viewfinders which are a great benefit to me and my aging eyes.

That said, the nikon lens I've wanted since I handled one some 15 months ago:

14-24mm 2.8

is a STELLAR piece of glass. I've always been drawn to ultra wide photography and this nikon lens is The King of UltraWide. In all reviews "Sharp as Primes" at every focal length of 14-24mm. I can confirm if ultra wide capture is your thing this is a Must Have lens.

Had it not been for the "heads up" on 18% nikon price increase I would have waited for further price drops. I think 2009 will go down as the year all the prices we are use to seeing globally Will Go Up instead of Down.

Which means I am now done buying brand new equiptment for the next year, maybe longer. Hopefully in the meantime Pentax will all Wow Us with fast frame per second, more af points, as opposed to just the same K20D specs with more megapixels. I don't need video and I have a feeling thats another direction camera makers are heading in. So for me the next year or two, I'll happily deploy the 5D 12mp, D700 12mp and the 15mp K20D.

I'd happily try out that 14-24mm, but I know no one who has it here, and shops here are different in the US and elsewhere. Here, it's "try only if you're sure to buy". I'm pretty sure I'd be wow-ed by its performance on a FF DSLR (I've never experienced that wide of a FOV before), but it'll be just that - a test run, unless I hit the lottery.

I would have to agree that a FF viewfinder would be a wonderful thing to have. I've been shooting some film SLRs lately, and going back to the VF on the K200D, it leaves me wanting. What I'd like to see in the K30D would be the same K20D sensor, better algorithms/electronics to be able to shoot from ISO 25-12600 with relatively clean noise, 5 FPS, bump in AF performance, and 100% VF. I'm already a happy camper by then. Anything else would be gravy.

[falconeye]

Huh? Pentax own words:

"...placing two sensors, one minutely in front, the other just to the rear, of the in-focus position.

Ok, well, this explains the confusion. This quote, even if from holy Pentax, is a false statement. The original author had a misconception of how a phase AF sensor works, including that of Pentax. There are images of the Pentax AF sensors (on the bottom of the mirror cage) which show that Pentax uses the same AF sensors as everybody else.

A phase AF sensor uses two (in the case of a cross type, four) 1-dimensional CMOS arrays and the phase shift is the shift between the two images where coincidence (maximum correlation) is reached. The function to be maximized is Norm(correlation(shift))^2. Not contrast, why this type of AF system is called phase, not contrast, AF. It works up to minimum aperture (such like f/5.6) only.

What we (we = everybody else in this thread) call a "double sensor" is a full double sensor. One working up to f/5.6. And another, second, one made to only work up to f/2.8, using more shift (combining rays more far away) and providing more accuracy. A full double cross sensor would have 8 CMOS arrays.

If in doubt, always listen to engineers, not marketing people.

[Pål Jensen]

Ok, well, this explains the confusion. This quote, even if from holy Pentax, is a false statement. The original author had a misconception of how a phase AF sensor works, including that of Pentax. There are images of the Pentax AF sensors (on the bottom of the mirror cage) which show that Pentax uses the same AF sensors as everybody else.

A phase AF sensor uses two (in the case of a cross type, four) 1-dimensional CMOS arrays and the phase shift is the shift between the two images where coincidence (maximum correlation) is reached. The function to be maximized is Norm(correlation(shift))^2. Not contrast, why this type of AF system is called phase, not contrast, AF. It works up to minimum aperture (such like f/5.6) only.

What we (we = everybody else in this thread) call a "double sensor" is a full double sensor. One working up to f/5.6. And another, second, one made to only work up to f/2.8, using more shift (combining rays more far away) and providing more accuracy. A full double cross sensor would have 8 CMOS arrays.

If in doubt, always listen to engineers, not marketing people.

It is written by engineers with diagrams and illustration.

You are confusing AF point with sensors. Each AF point have two sensors; that is what the beam splitter is used for; splitting the beam to each sensor.

It is not correct that AF do not work at aperture slower than F:5.6. It has nothing to with blacking out as stated earlier. It works, but that depends on brightmess and contrast of the subject. It is just that it doens't work consistently enough. This according to Pentax technical paper. It has also been collaborated by user who has managed to get F:8 lenses to AF.

To me Canon F:2.8 AF looks like marketing speak as Pentax had this feature in the first AF SLR, the Pentax ME-F of 1981. It was manually engaged via a switch on the top of the camera body. Canon seem to do this automatically, I wont be surprised if Pentax does that to now that all lens info is transmitted to the camera....

[Pål Jensen]

This figure show the two sensors (a, b) and curves derived from them. It also show that they are in front and back of the film plane and that the (AF) film plane is calculated to be where they cross. It also show what the 2.8 vs. 5.6 is all about; precision in determing top point of the curves.