[bladerunner6]

It’s not dead, it’s pining for the fjords..

Lovely plumage.

[beholder3]

Nice explanation! Thanks! That also explains why these preciser sensors can only be in the center of the frame, eccentricity would cut them off on one side.

Look at the sensor hardware of the K-3 here: https://www.imaging-resource.com/PRODS/pentax-k3-ii/ZTECH_AFSENSOR.jpg
It is easy to see the pair of three "F2.8" pixel-lines on the extreme left and right sides. You can see that they are twice as far away from the center as the "F5.6" ones.
All these pixel rows "look through" the extra AF optics at the mount opening.

On the above photo of the hardware you can easily imagine what image areas are covered by which types of AF sensors when in your brain you move/shift all the line pairs (top/bottom and left/right) together in the center of the lens/hardware. Then you see which area is covered by line and cross type sensors.

It also tells the story of marketing the "numbers" of AF sensors. --> The number actually is a software decision to about 50%. The K-3 sensor only has 5 (F5.6) horizontal line sensor pairs. The question into how many marketing lines you split them is completely open. The only limitation is via the vertical line sensors because obviously you can have only as many cross type sensors as there are virtual "intersections" between horizontal and vertical. But they could have advertised "300" horizontal line sensors on the K-3 already.

[kwb]

Excellent illustration, beholder3!

OK. I'll see if I can find more information about how cross focus sensors work, because I'm sorry, but I couldn't follow the information which you posted. Yes, I know what a split image focus prism is like because the first SLR I used had one (the Chinon CE4s). But also I didn't see anything in your reply about flux luminance.

Oh, that's a translation error.

Usually it's written as "F/2.8光束対応AFポイント" or AF point that can use F/2.8 ray bundle. 光 is light, 束 is a bundle. 光束 in this case is a ray bundle, i.e. a group of rays (each ray being a straight line), which is a concept often used in ray tracing and geometrical optics.

光束 could also mean luminous flux, which is a measure of perceived power of light (unit: lumen).

As far as AF is concerned, 光束 has everything to do with how far the sensors are apart. F/2.8 sensors are farther apart than F/5.6.

So, which "ray bundle" are we talking about? I'll show you a cartoon I made some time ago for a different purpose, hopefully the combination of this and beholder3's illustration will help visualization.

Let's look at the top where the sensors are placed to look at two ray bundles close to the outer edge of conical ray bundle formed by a F/2.8 lens. That pair of sensors is "F/2.8光束対応AFポイント".

The bottom shows you a pair of sensors much closer together than F/2.8. The lens is still F/2.8 in this case, but the sensors are looking at ray bundles close to the outer edge of conical light bundle which would be formed by F/5.6 lens. This pair is "F/5.6光束対応AFポイント".

As you can hopefully see, the aperture number designation for AF sensors doesn't have anything to do with brightness of things.

[beholder3]

And on your graph you can nicely see the difference in precision between the top graph for F2.8 and the bottom one for F5.6 in that the spread between the orange and blue defocused postions is much larger on the F2.8 sensor.

So while a F5.6 sensor might think that both incoming sensor images (a contrast edge captured by the pixel line) do match and therefore the optimal focus is achieved, the F2.8 sensor still sees a significant deviation between the relative position of the contrast edge captured by both of its sensors.


Basically using a F2.8 sensor instead of F5.6 means using a magnifying glass for focussing.

[angerdan]

Look at the sensor hardware of the K-3 here: imaging-resource.com/PRODS/pentax-k3-ii/ZTECH_AFSENSOR.jpg

Interesting, i didn't know about the old images of the K-3 II. All images of the series from 2015 are published there: Pentax K-3 II Review - Technical Info

[gatorguy]

Great find, all images of the series are published there: Pentax K-3 II Review - Technical Info

For clarity's sake that is not the K3III sensor, correct?

If that's right someone might make this a bit more clearly relatable to the thread topic, K3III video.

[angerdan]

For clarity's sake that is not the K3III sensor, correct?
If that's right someone might make this a bit more clearly relatable to the thread topic, K3III video.

Thats right, it shows the AF Sensor of the K-3 II. So the image is 6 years old.

[gatorguy]

Thats right, it shows the AF Sensor of the K-3 II. So the image is 6 years old.

So not relevant to the K3III, or if so how?

[Serkevan]

So not relevant to the K3III, or if so how?

The principle should be the same... the schematics of the AF system show the f/2.8 sensors (now 9 instead of 3, IIRC).

[MMVIII]

Just for the record - I combined an old section drawing of the K-3III with the section drawing of the viewfinder. Maybe it is of interest someday.
It is taken from the video, so thought it might fit in here.

[RobG]

Usually it's written as "F/2.8光束対応AFポイント" or AF point that can use F/2.8 ray bundle. 光 is light, 束 is a bundle. 光束 in this case is a ray bundle, i.e. a group of rays (each ray being a straight line), which is a concept often used in ray tracing and geometrical optics.

Just for my own curiosity:

光束 = kousoku = a beam of light. Yes, it can mean a bundle like a sheaf of rice straw.

対応 = taioo = correspondence, as in measuring two things to see if they are the same

ポイント = pointo = point

Thanks for those diagrams. They make it much clearer how pairs of sensors relate to unfocussed light coming through a wide-open lens, and how it relates to the wide-open f-stop of the lens. The f2.8 sensors will presumably be most critical for f2.8 macro lenses, but I guess even a portrait lens wide open needs to be very carefully adjusted due to narrow DOF.

[RobA_Oz]

Just for the record - I combined an old section drawing of the K-3III with the section drawing of the viewfinder. Maybe it is of interest someday.
It is taken from the video, so thought it might fit in here.

I imagine many of us are interested in the difference the new pentaprism material might make to the K-1ii successor, but this comparison is also interesting for the work that seems to have been done to reduce the overall depth of the SR mechanism in the K-3iii. The individual component thicknesses (including the sensor stack) appear to be similar, but the gaps between them are less in the K-3iii than in the K-1. To me, that would be the result of some good engineering work to keep the body as compact as possible, maybe using different materials.

If this is true, then similar work could reduce the thickness of the K-1 by a couple of millimetres, or thereabouts, while still retaining the rear screen tilt-swivel mechanism. All very encouraging.

I also wonder if the new pentaprism material will allow for some further development of the hybrid viewfinder, but that’s a subject for another discussion.

[Sandy Hancock]

Just for the record - I combined an old section drawing of the K-3III with the section drawing of the viewfinder.

That has me very excited; it's a full centimetre thinner!
Do we have official specs to confirm this yet?

[leodp]

That has me very excited; it's a full centimetre thinner!
Do we have official specs to confirm this yet?

Yes, a full cm. No more room for improvement in the future, maybe 0.5cm max.


Also, what I find as something that stands out: the pentaprism & OVF make up a higher and higher percentage of the camera cross-section.

Probably the increased OVF magnification contributed to this, beside the thinning of the camera body.

This tells me that reducing the flange-to-sensor distance will have a reduced return in terms of camera size.

--> We are at a forking point between mirrorless APS or FF cameras and DSLR.
--> We can reasonably expect the flange-to-sensor distance to be kept constant (no new mount), with maintained backward lens compatibility for the future models.

[Kunzite]

That is a full cm, but compared to the K-1 - which has the moon lander articulated LCD.
Same optimizations applied to a K-1 III - keeping the moon lander - would get us a bit less. Still, it would be nice, as the K-1 feels just a tiny bit too thick.