[hjb981]

Ya i agree I wasnt questioning the validity of the thread, I was really wondering out loud whether I should care about this particular bit of information - or not - because I am going on information overload these days . It seems to me that there are members of the board who are legitimate engineers, and I can try and join the conversation, but I will be left behind pretty quickly, with no improvement to my photography, and maybe a little disappointment at why I dont understand how a specific number of blades is important to my lens, or something like that .

This particular information seems to be a little more practical in nature, and from what I get from the discussion, it seems to me that while the autofocus of K52 may be similar to K5 on say an EV0 and a 18-55 3.5-6 lens, if I put a 2.8 lens in front of the K52, I may get a better autofocus because it will switch the AF system to the 2.8 ..err...arrays.

Right?

I think that you got that right. Put a 2.8 or faster lens on, and the AF will have better precision (be more consistent), in other words, if you focus on the same things many times, there will be less variation between shots.

It also seems that the AF would work with less light, both with a 3.5-5.6 lens and with at 2.8 lens. That is what my original question was about, and does not have to do with how precise the AF work, only how low light levels it can handle.

If I ever write anything that is too technical, do not hesitate to ask. I will try to explain. There are indeed many threads that get very technical, and also lots of incorrect information -- in many threads, someone says "it is this way" and someone else says "no, it is that way" -- obviously one of them has to be wrong. I do not think that anyone here is out to mislead people -- I think most people have the best intentions, but sometimes have misunderstood something. Because of this, you have to be able to evaluate the information you read, and this is of course nearly impossible if you don't have some background knowledge.

In the end, I think that the technical part and picture-taking part are two different (although connected) things. I am interested in both, but if you are interested mostly in the taking-pictures part, don't worry too much about all the super-technical forum talk. As an example, you do not really need to know how many aperture blades your lenses have to take good pictures. If you still want to know to be able to choose future lenses, a basic understanding should suffice: the best situation would be a round aperture. The more blades, the closer to round, so the better. At the "intersections" of the blades (the "corners" of the aperture), the light gets scattered. That means that it "gets spread out". It mostly gets spread out inwards and outwards (towards and away from the center of the aperture). This is the reason for the star patterns that can appear around bright lamps etc in pictures. If you have six aperture blades, the star will have six "spikes". If you have 7 aperture blades, you will have 14 "spikes" on the star. Why? Because remember that each corner causes diffraction in two directions. With on even number of blades, the two diffraction directions of one corner will overlap with the two from the corner opposite. With an odd number, there will be no overlap, and hence twice as many "spikes" as corners, but on the other hand the spikes will be less intense.

OK, the above was very off-topic... I hope that it was understandable, though. The point that I was trying to make is that even the technical stuff does not have to be too complicated. I hope that I was making sense above.

Not being knowledgeable about the technical stuff should not prevent you or anyone else from taking pictures. Like I said, I know several people that do not know that much about the technical stuff, but still take very good pictures.

[Nicks]

Changing the subject slightly:

Has anyone heard whether the K5II has the DO optic in the AF module like the K30?

The reason I ask, is that I am more than a bit convinced that a significant portion of the issues with the K5 AF are related to the color sensor part of the AF system. It seems clear to me that the initial tungsten FF issue was related to this system as it was fixed with a firmware update. Given that the color sensor was introduced to deal with CA shifts in the physical optics of the PDAF module that cause FF in tungsten light, it seems reasonable to conclude that the firmware fix was made to correct a programming problem in that system.

I think the DO optic is a much better solution, and so far seems to be working.

I have two k5s, and do not have many AF issues to complain about, but I have seen some misses that seem to be related to color. If you think about how inaccurate AWB can be, you can see that determining actual overall scene lighting color temperature and then making a precise adjustment to the AF might be a hit or miss thing.

Otherwise, there are other sort of vague claims about K5II AF improvements, but I would probably not be tempted by the K5II just for the low light improvements without the DO optic. Almost all of my lenses are f2.8 or faster, so the improved accuracy if the f2.8 sensors might be a good reason to consider the K5II, but I am not having any significant AF accuracy issues, so that isn't looking like something to tempt me to look at the K5II.

I guess time will tell, and I am certainly tired of Pentax reading AF complaints, so if the K5II corrects most of the things users complain about, I'm all for it even of I do not buy one.

If nothing else, the changes to the K30, and the emphasis on AF improvements for the K5 clearly show us Pentax is listening and tackling the things that users complain about most (well, except SDM...)

Ray

Ray, this is exactly what I was thinking. As far as I know, the new 2.8 sensors will improve accuracy in low light but it will not fix the real problem K-5 is struggling with and that is the light wavelength issues. I own both K-5 and K30 and it's obvious to me the DO in K30 works really well. I too shoot most of the time in low light with fast lenses and it's been a quite frustrating experience with K-5. One simple thing that I think would've helped K-5's accuracy is for the AF assist lamp to light up sooner than it does. On the other hand I know that when K30 locked focus the accuracy is there. The AF is not fast in low light but at least I know it's accurate. I think it would be a big mistake for Pentax if they don't implement the DO in K-5II as well. Time will tell.

[Ray Pulley]

Ray, this is exactly what I was thinking. As far as I know, the new 2.8 sensors will improve accuracy in low light but it will not fix the real problem K-5 is struggling with and that is the light wavelength issues. I own both K-5 and K30 and it's obvious to me the DO in K30 works really well. I too shoot most of the time in low light with fast lenses and it's been a quite frustrating experience with K-5. One simple thing that I think would've helped K-5's accuracy is for the AF assist lamp to light up sooner than it does. On the other hand I know that when K30 locked focus the accuracy is there. The AF is not fast in low light but at least I know it's accurate. I think it would be a big mistake for Pentax if they don't implement the DO in K-5II as well. Time will tell.

It is good to hear from someone with the K30 and the K5. The AF assist on the body is all but useless, IMO. here is why:

Many lenses block all or part of the light.

It is inconsistent about when it comes on.

It is not very bright.

It has no stripe or pattern to create contrast.

I find that with wide lenses like the DA 15 that have such a large FOV, the dim body light creates so little additional illumination on targets beyond about 5-6 feet that it makes no difference in the focus lock.

The assist light on the 540 is much better, however. It seems more consistent, is centered and clears all lenses, is brighter and has a contrast stripe. Accurate lock in low light is much better with the 540 on the camera.

Back to the K30/K5:

I believe that the lenses are a big part of the problem with AF speed, especially SDM. This would be equal between the two bodies.

While I have not seen significant or consistent AF issues with my K5s, I did shoot a few party pics at work this week with one of my K5s, mostly indoors with flash, but a few outdoors in daylight, and quite a few shots were just not sharp. Oddly, I shot similar shots at my grandson's birthday a week or so ago with the same exact setup, indoor, outdoor, flash, mid-day, and out of 54 shots, maybe 3 or 4 were not in focus.

The only semi-consistent focus issue I have seen until this week, is slight misses when pink'ish young children or baby faces fill most of the screen. They are not all misses, but more than average.

Although all of my earlier Pentax bodies had consistent FF in tungsten light, the subject was usually just in the DOF at f4 or f5.6, so this was not a big problem. Pentax was slow, but otherwise accurate. Of course, those cameras just would not focus at all in light levels well above where the K5 will lock, and tracking anything moving was pretty much impossible.

Anyway, I am starting to lose a bit of confidence in getting the shot with the k5s. I changed to the S type focus screen so that I can better judge the focus point, but there often isn't time to carefully judge the focus point, and the APS sized viewfinders do not help.

Lastly, if the K5II has no DO AF module, I would be more tempted to pick up a K30 than a new K5II.-3EV is truly shooting in the dark, but -1EV is fine with me if it nails the focus consistently.

Ray

[Anvh]

I think that you got that right. Put a 2.8 or faster lens on, and the AF will have better precision (be more consistent), in other words, if you focus on the same things many times, there will be less variation between shots.

It also seems that the AF would work with less light, both with a 3.5-5.6 lens and with at 2.8 lens. That is what my original question was about, and does not have to do with how precise the AF work, only how low light levels it can handle.

You're mixing things here, you've the sensitivity of the AF sensor and the maximum aperture threshold so to say.

So for the AF to work better on f/2.8 lenses they need to be spaced out more but that means that it doesnt work any more with for example f/5.6 lenses because they are spaced out too much to see that "slow".
It's not like f/2.8 sensor has a bigger opening then f/5.6 sensor, that's what most think but the only difference is the spacing nothing more so they are both as sensitive.

---

Last edited by Anvh; 09-19-2012 at 09:22 AM.

[ENicolas]

Err...will this knowledge make me a better photographer, or did you guys happen to be photographic equipment engineers? If this knowledge will make me a better photographer - I wanna get in on it. Soon as I understand how knowing what aperture my AF sensor is at will help me with my pics....until I stumbled into this thread i didnt even know there is an aperture value for AF?

All knowledge makes you a better photographer - in this case better understanding of your equipment makes you better able to choose equipment, operate it, and understand it's capabilities and limitations.

Now if I understand correctly (and that's always suspect!), the primary difference in an f/2.8 AF system is the mask that overlays the AF sensor: the slits in the mask are farther apart, giving the system a broader "view" of the image projected through the lens. This will make AF more capable for any given image because more of the image is being considered (or at least the discrete points are more isolated, so there is a broader cross-section of contrast data).

So right-off this wider spacing will increase accuracy of AF for most scenes. If the contrast levels near the center of the frame are somewhat uniform then f/5.6 AF will struggle to find an accurate position for the lens elements. And this is why when shooting wide angle landscapes I sometimes have to frame the image on a contrasty part of the scene, use AF, then re-frame for the composition I want (similar to the way you might meter off the sky and then re-frame). Or I use manual focus to adjust the AF the camera finally settles on.

With a wider f/2.8 mask I probably would encounter that situation less often. The thing is, with my f/4 16-45mm zoom the aperture may not be wide enough for all of those f/2.8 mask points to "see" the entire image: some will be outside the circle of light the lens projects. So for a slower lens the f/2.8 mask might actually be less capable than the f/5.6 mask.

[JinDesu]

In phase detect AF, I thought the sensors try to align the red and green (or whatever) wavelengths that are split by the lens. For the 2.8 sensor, it would obviously have greater sensitivity (larger difference between the incoming wavelengths) than the 5.6 sensor. I assume that is where the improved fast lens accuracy comes from.

In addition, based on what I hear about Canon's AF systems, the 2.8 and 5.6 reside in the same sensor location - the camera chooses which to use based on the situation. I don't think you lose the 5.6 sensor just because you have a 2.8 sensor.

[Anvh]

This one is from 7D




As you can see f/2.8 are in different location.
So yes you can have 5.6 and 2.8 but not in the same spot.

[JinDesu]

Ah ok, thanks for that info

[Anvh]

In a way what you said can work, the only difference is the spacing so if you make the sensitive strips larger it might work.
Or you simply have strips all over the place so that you can vary the distance that way.


ps. just see i've read your comment wrong.
yes the f/2.8 and f/5.6 spots are on the same sensor.

[Ray Pulley]

In phase detect AF, I thought the sensors try to align the red and green (or whatever) wavelengths that are split by the lens. For the 2.8 sensor, it would obviously have greater sensitivity (larger difference between the incoming wavelengths) than the 5.6 sensor. I assume that is where the improved fast lens accuracy comes from.

In addition, based on what I hear about Canon's AF systems, the 2.8 and 5.6 reside in the same sensor location - the camera chooses which to use based on the situation. I don't think you lose the 5.6 sensor just because you have a 2.8 sensor.

The secret of Phase Detect is that the one ray of light is split into two rays by the beam splitter (this is usually a small part of the mirror that is semi-transparent so some light goes through instead of being reflected up to your eye) and these two rays are then directed to a pair of sensor points, one on each side of the array. The distance between these pairs of sensor points is the aperture diameter we have been discussing.

f5.6 sensors are simply closer together because the f5.6 circle of the lens is smaller than the f2.8 circle (assuming the lens even has a f2.8 aperture). Regardless, the PDAF system knows the distance between the two sensor points, and it reads the sensor point pixels and adjusts the focus until the readings are the same on each side. Since the two rays were originally one ray that was split, when the two match (in phase) the lens is in focus.

So, 2.8 and f5.6 sensors cannot be located at the same place. The 2.8 sensors have to be further apart than the f5.6 sensor points. Being further apart means that the calculation used to focus the lens can be more accurate as the two points are used as a baseline for a triangulation type of calculation. In that type of calculation, the further between two points on the baseline, the more accurate the calculation. Of course, being further apart makes the calculation take a little longer as well, and also the focus adjustment can be slower too because the camera is reading and calculating this while it is moving the lens.

Lastly, it is hoped that the design of the AF optics will focus all wavelengths at exactly the same point on the AF array. However, it is when the AF optics are not well corrected for different wavelengths that you might get focus shift in different wavelengths (like FF or BF in tungsten, for example). This is chromatic aberration (CA).The K5 introduced a color sensor in the AF system to try and determine the color temperature of the light for the shot and adjust the focus accordingly, and the K30 uses a diffractive optic element to correct the actual optics in the AF system so that all wavelengths focus correctly.

I think it is always preferred to get the optics right rather than trying to adjust some other way. I also think that this color sensing element is why the K5 does odd things while focusing sometimes. I know it is not exactly the same thing, but how often is AWB correct? AWB does nothing but try and guess the color temperature from the main sensor data. It seems to me that it is equally hard to sense the true color temperature of the scene for focus adjustment, especially when it is very dim.

Ray

[Class A]

Has anyone heard whether the K5II has the DO optic in the AF module like the K30?

I haven't but perhaps you'll find these K-30 & K-5 II cut out images helpful.

It seems clear to me that the initial tungsten FF issue was related to this system as it was fixed with a firmware update.

The initial FF issue was mitigated with the firmware update (for some situations) but not fixed. The underlying problem was CA in the AF module and a limited sensitivity of the colour sensor (metering chip). It is not unlikely that even the latest firmware contained a bug; it should have been possible to either better correct in low-light situation or at least stop pretending that a reliable lock has been achieved.

I think the DO optic is a much better solution, and so far seems to be working.

I agree. I furthermore have no doubts that the K-5 II's AF module is at least as good as that of the K-30. It wouldn't make sense for Pentax to let the K-30 outperform the K-5 II in terms of AF. I fully expect future tests to demonstrate the K-5 II to be superiour to the K-5 with low-light locks both in terms of accuracy and speed.

Of course, being further apart makes the calculation take a little longer as well, and also the focus adjustment can be slower too because the camera is reading and calculating this while it is moving the lens.

What makes you think that?
Why should there be any correlation between processing speed and the length of the AF baseline?

The only disadvantages I can see for f/2.8 AF areas is that

• they stop working with slow lenses, and
• they lose track of an out of focus subject earlier. Slower AF sensors can still match (and thus enable calculation of the appropriate focus action) out of focus subjects when faster AF sensors have already lost an overlap between the separate images.

[andymuns]

Snapsort is saying that k5II and the K5IIs only has contrast focus i have sent a post telling this is wrong but they have not done anything about it . I think this might mislead people.

[Ray Pulley]

I haven't but perhaps you'll find these K-30 & K-5 II cut out images helpful.

What makes you think that?
Why should there be any correlation between processing speed and the length of the AF

What makes you think it does not?

I probably should have been more clear and said focus process or cycle rather than calculation.

While we do not know exactly what Pentax is doing with the new AF module, we have been looking at what Canon says about their f2.8 sensors in this thread. From what I gather in some of that information and discussions about it, the speed of the focus cycle, the f2.8 sensors are more accurate but slower.

In fact, Canon seems to use the f5.6 sensors to get close because it is faster, and then complete the cycle using the f2.8 sensors, at least in some models.

The reason seems to be that the f2.8 sensor areas have more resolution or more pixels in the same area (which also means they are smaller). This makes sense as the baseline is not the only thing that affects AF accuracy, the resolution of the pixels measuring the split light beams is also part of the equation. If you only increased the baseline for a more accurate calculation, you might not gain the full accuracy advantage if your array resolution stayed the same. After all, the camera looks at the two split rays on a the pixels of the arrays and moves the lens to hit the right spot. The resolution of these arrays determines how accurately you can judge the difference between the split rays and which direction to move and also how far to move.

So, you will need to drive the lens in smaller steps as you get close, and/or perform an extra check and adjust cycles when you are close in order to fine tune the position and end up with a more accurate position. This takes more time.

Pentax is not Canon and might be doing something completely different than this, and I could be completely wrong too.

I just want Pentax to give us something else to talk about rather than Autofocus, so hopefully the improvements in the K5 and K30 will at least be a good start on this goal. At least they finally seem to be trying to address what has to be the number one complaint about the product overall over many models now.

Ray

[D4rknezz]

If you still want to know to be able to choose future lenses, a basic understanding should suffice: the best situation would be a round aperture. The more blades, the closer to round, so the better. At the "intersections" of the blades (the "corners" of the aperture), the light gets scattered. That means that it "gets spread out". It mostly gets spread out inwards and outwards (towards and away from the center of the aperture). This is the reason for the star patterns that can appear around bright lamps etc in pictures. If you have six aperture blades, the star will have six "spikes". If you have 7 aperture blades, you will have 14 "spikes" on the star. Why? Because remember that each corner causes diffraction in two directions. With on even number of blades, the two diffraction directions of one corner will overlap with the two from the corner opposite. With an odd number, there will be no overlap, and hence twice as many "spikes" as corners, but on the other hand the spikes will be less intense.

I kind of got that - and that helped me understand why the number of blades affects bokeh. Thanks!

[Class A]

What makes you think it does not?

The same reason that stops me from entertaining the thought that the moon is strung to the earth by an invisible rubber band.

It make sense to use the f/5.6 sensors first as they see more subject distance range. It also makes sense for the AF procedure to take a bit longer with the f/2.8 sensors as the focus movements will have to be more precise. However, I do not believe that camera internal processing adds further delays or that higher resolution sensors take noticeably more time to evaluate.