[GMounk]

I am wondering if they can come up with a software fix. Perhaps the software is written in such a way that it is slow. For example, I have a friend who is a computer programmer. A mutual friend told me that he uses one line of programming where other programmers use 11 lines of programming. So maybe there's room for software improvement, maybe it can be written differently, or are we dealing with a slow processor? The other thing is how much stuff is going on in the background that really isn't necessary.

[UncleVanya]

I am wondering if they can come up with a software fix. Perhaps the software is written in such a way that it is slow. For example, I have a friend who is a computer programmer. A mutual friend told me that he uses one line of programming where other programmers use 11 lines of programming. So maybe there's room for software improvement, maybe it can be written differently, or are we dealing with a slow processor? The other thing is how much stuff is going on in the background that really isn't necessary.

I'm sure there are possibilities for code improvement - but that costs money. Not just for coding but for testing and validation and so forth. Good luck getting the coffers opened for this.

[Jeffvan]

I am wondering if they can come up with a software fix. Perhaps the software is written in such a way that it is slow. For example, I have a friend who is a computer programmer. A mutual friend told me that he uses one line of programming where other programmers use 11 lines of programming. So maybe there's room for software improvement, maybe it can be written differently, or are we dealing with a slow processor? The other thing is how much stuff is going on in the background that really isn't necessary.

There is efficient code and then there is code that is just written in a compressed manner. 11 lines of code to 1 line of code doesn't always equate to efficient code. It depends on how the compiler translates your code into machine code. I have seen 1 line of code end up being massive on the machine code side. Now working out the logic of the code, so that it needs less processing power, that is where magic happens, and it requires understanding your CPU platform. On the up side, someone that can compress 11 lines of code to 1 line, most likely can write code that is efficient for the CPU, but would be loathed by other programmers for making there code hard to read. I have been blamed for writing black-box code before.

There are several causes of slow code:
1) Improper implementation of algorithms for the hardware platform.
2) Extra conditional evaluations to avoid incorrect data entering an algorithm, such as bad user input. Or in the case of a K1 a hot pixel value.
3) A Lack of processing power for the tasks at hand. (The task is overloading the processor capability)

In the case of number 1 there is an issue with the K1 and JPEG values with RAW files. I noticed the significant delay between auto-focus acquisition and shutter release with the wrong JPEG settings while shooting RAW. The proper way would be to have the CPU dedicated towards auto-focus and shutter release and dump all the information into the buffer, and then work on JPEG processing. But things are not always this simple. There could be a dedicated JPEG pipe in the K-1 CPU. This is common in a lot of micro processors, to have dedicated pipes for specific algorithms such as encryption. This JPEG pipe might be fed from the Image Sensor input side of the CPU. If this is the case, it is a CPU design problem and nothing on the coding side is going to fix this.

Regarding issue 2: This is not always such an easy fix, since any device that has several inputs need to have those inputs evaluated for errors.

On the issue of number 3: This is where things get interesting. If the K-1 could have a high performing FPGA inside, it would be one of the best solutions out there. Since you could literally create a new CPU architecture when ever you need it. You could potentially have a different CPU for each of the different modes. BUT, this comes at a great cost of actual power consumption. It is one of the toughest battles going on right now in the DSLR world. Look at the Nikon D850, which can do 7 frames a second in shooting, but with the grip that gives more power it goes up to 9 frames a second, allowing the CPU to overclock. This comes at the cost of having a larger battery, which costs more. So if you are hoping for the 500+ shots from a single battery charge, then you need a very efficient CPU. Just look up the number of shots the Sony A7R II got (~100) per battery versus the need for the larger battery in the A7R III. So a more powerful CPU in the K-1 will either require new CPU technology or more battery power. Yet again a design issue that we cannot alter at this stage.

So at this point I doubt there are many areas of improvement on the K-1's efficiency side. They could make it that when you shoot only RAW, that the JPEG settings are forced to low, and effectively give you better results.
Now in many implementations there isn't just one CPU that does everything. Quite often there are dedicated processes being done by smaller computing units. These units' firmware is updated by the CPU, and one such unit might be the phase detection sensor unit. I am not positive on this but I could envision how this would be implemented as a modular system. If this is the case, there might be some algorithm improvements there, if that unit isn't already over tasked. But this wouldn't be related to the shutter release delay, it would be related to improving the tracking capabilities of the auto-focus.

[GMounk]

There are several causes of slow code:
1) Improper implementation of algorithms for the hardware platform.
2) Extra conditional evaluations to avoid incorrect data entering an algorithm, such as bad user input. Or in the case of a K1 a hot pixel value.
3) A Lack of processing power for the tasks at hand. (The task is overloading the processor capability)

Thanks for the insight I am starting to get some understanding to the issue.
I have always been the type of person who likes to do things the most efficient way. I know next to nothing about computers but oddly enough I was able to figure out a simple fix for an issue that our company and at least one other company was having with a certain computer program. It was so simple the programmers completely overlooked it. I think sometimes smart and knowledgeable people seem to have either complicated or expensive solutions for simple problems. I certainly would never understand how the K-1 programs work, but on the other hand I think, they should have a a few minimalist type of settings that can be activated from the menu screen. You go to the screen and it would have setting for landscape, action etc. Then after a person has a feel for the camera they can get into the more complex settings a pro photographer would use. It has been a very significant jump to go from a prosumer camera that is ten years old to a professional camera, with tons of capabilities.

[zapp]

Lots of people say the C-AF performance in the K-1 is weak. I think I have two specific performance issues I can bring up from my own experience shooting this year's Walk-A-Thon at my kid's elementary school.

1) Lag - slow shutter activation. The K-1 with a D-FA 24-70mm can acquire focus and track subjects decently well. It's the speed of the shutter firing that makes images soft. The kids who ran toward me leaned forward. Their chests were ahead of their feet. In all the images I captured their feet were very sharp but their chests and their faces were soft. Yes, I used SEL9. Yes, I set the focus position point correctly. Yes, I used a very high shutter speed in TAv mode (1/2000). We have to remember that the camera stops focusing when the mirror goes up and the shutter starts to cycle. It's during that time the subject moves forward and is now slightly out of focus. Focus is perfect-perfect for still subjects and even when panning.

Wish : I wish the mirror could go up faster. I wish the shutter could cycle faster.

2) SEL9 seems to pick the furthest subject in its selection area. Given two people side by side with one person standing 3 feet back the K-1 seems to focus on the further back person. It's like the camera wants to focus closer to infinity instead of minimum. I thought Nikon had an option in their menus to let the photographer decide the priority. Where is this in Pextax? Am I missing something very fundamental somewhere?

Wish : Give priority to focusing to closer subjects instead of more distant subjects.

Now, with all that said, I was still able to produce lovely 4"x6" prints from every image captured. The softness we see when pixel peeping at 100% on the screen disappears at that size. Some were good enough for 6"x8". My stills are sharp enough for 9"x12" and beyond to poster sizes if not banners.

We are looking for better tracking and better prediction. Phase detect AF used to be the reason to stick with DSLR vs mirrorless. Look at current mirrorless cameras, they are ahead of Pentax with this regard.
You want to lock on a face and track it while moving towards you. 3D AF is a Nikon Feature, it is nice and works well. For FF cameras you also want to see better focus point coverage of the whole viewfinder. I am not saying Pentax AF-C is completely useless, you can get godd tracking if you know how to control it, but all others are a couple generations/years ahead.

[clackers]

You want to lock on a face and track it while moving towards you. 3D AF is a Nikon Feature, it is nice and works well. .

?

It's a mess.

Youtu.be

Youtu.be

[Jeffvan]

?

It's a mess.

NikonD810: 3D subject tracking at close subject distances - YouTube

That looks awesome!

[clackers]

That looks awesome!

Look how long it takes to find the eye, Jeff!

Eg, if the subject keeps moving, it keeps hunting ... notice he basically has to come to a stop before it hits the target.

[Jeffvan]

Look how long it takes to find the eye, Jeff!

Eg, if the subject keeps moving, it keeps hunting ... notice he basically has to come to a stop before it hits the target.

You have to keep in mind that the spot sensors are not where they indicate on the screen. You can measure your own cameras spot sensors, so any hits that land close to the eye is actually spot on. It might be the sensor that has the greatest contrast on the eye or what is indicating the selected spot might be an area that covers the eye even if the display doesn't exactly land on it. In any case the tracking is very close to the area he selects. I would love to see how this looks on a D500, which has 3 times the spot sensors compared to the D810.

Any mishaps I saw were instantly corrected, like the one at 5 seconds and the one at 22 seconds. The hunting is perfectly expected, since he is constantly moving around, not just latterly but in the Z-Axis. You can also tweak these behaviors in the settings of the Nikon cameras. Like how erratic you expect your target's movements to be.

I just don't think an erratic target, like what is simulated, could be tracked by a human with a center spot only. You are going to try to select some area on the subject and hope to get good hits, or you are going to increase the DOF to include the whole target.

[clackers]

You have to keep in mind that the spot sensors are not where they indicate on the screen. You can measure your own cameras spot sensors, so any hits that land close to the eye is actually spot on. It might be the sensor that has the greatest contrast on the eye or what is indicating the selected spot might be an area that covers the eye even if the display doesn't exactly land on it. In any case the tracking is very close to the area he selects. I would love to see how this looks on a D500, which has 3 times the spot sensors compared to the D810.

Any mishaps I saw were instantly corrected, like the one at 5 seconds and the one at 22 seconds. The hunting is perfectly expected, since he is constantly moving around, not just latterly but in the Z-Axis. You can also tweak these behaviors in the settings of the Nikon cameras. Like how erratic you expect your target's movements to be.

I just don't think an erratic target, like what is simulated, could be tracked by a human with a center spot only. You are going to try to select some area on the subject and hope to get good hits, or you are going to increase the DOF to include the whole target.

Company propaganda's one thing, real life performance, another.

I saw a video of Rich Corey, possibly *the* action photographer in the world, choosing Manual Focus on his D5 for a skateboarder scene.

Here's that same AF system, but put in the APS-C D500. Watch the focus point get placed out into the grass again, and again, and again.

Honestly, Jeff, if you couldn't do better tracking yourself, you'd give the hobby away!

Youtu.be

Youtu.be

[Jeffvan]

I saw a video of Rich Corey, possibly *the* action photographer in the world, choosing Manual Focus on his D5 for a skateboarder scene.

Great! This doesn't mean everyone can do this, that is why he is *the* action photographer. AND how many takes could he have? How much planning went into deciding where the skateboarder was going to be. Did he practice his focus ranges and establish his focus limits. How much experience does he have taking these kinds of shots? Such shoots become the very definition of non-erratic, because he knows his subject, his gear, and that makes him a pro.

Company propaganda's one thing, real life performance, another.

Here's that same AF system, but put in the APS-C D500. Watch the focus point get placed out into the grass again, and again, and again.

3D AF Nikon D500 - YouTube

Out of the 229 shots shown, there were 37 miss placed tracking shots, that I counted. That is an 83% accuracy for tracking. In some of the sequences it appears as if the target was never established to begin with. For example: the sequence started at frame 60, the dog was never picked as the target, but I included that in the sequence count. Some of the targets that were in front of the dog (on the grass) was actually on the dog, because the displayed bracket on the camera is not necessarily where the actual sensor is. Look at frame 20 and 21, the dog is in focus, but the sensor bracket is on the grass just in front of the dog.

At frame 200 the system completely looses the target, and through frame 208 still doesn't re-establish focus. The lens has focused to the distance and cannot find the target. This is a known issue with any 3D tracking. If your lens isn't close enough in the focus distance of the target; then it will keep searching for the target in the current focus distance. Even with these flaws the dog was selected at 83% of the times, including the sequences when the dog was never picked as the target.

Honestly, Jeff, if you couldn't do better tracking yourself, you'd give the hobby away!

3D AF Nikon D500 - YouTube

I by no means think that Nikon 3D Tracking is the solution to everything. A good example is dogs. Their features don't lend well to tracking specific body parts. Their nose, eyes and ears or features on their coat can all be taken as the same object, If you wanted the eyes in focus, you might have to go to spot focus for small DOF. Or in the case of birds in the sky where dynamic group focus allows you to use the featureless blue sky to pickup birds for focus. But in the case of a dog having a white coat and dark eyes, you could then use a good 3D tracking system for body parts. It comes down to options. It isn't something to use all the time, but when there is a case for it, it does make a difference. It can reduce the workload dramatically when you are attempting to get that unique shot.

So when it comes to good options with the Auto-Focus systems, the K-1 just doesn't have as many as the Nikon does for action photography. The Nikon D3000 that I have, does about as well as the K-1 does right now, when it comes to autofocus. That might mean we have to dig deep into the Rich Corey strategy, and operate a lot more of the tracking (and other autofocus features) on a manual basis and get to learn our subject a lot better.

Lets take the person in this video and lets have a little thought experiment: Review what this D500 video would have looked like if the person had a K-1 in their hands. On several occasions the dog is on the edge of the frame for several shots in a sequence, so the person isn't the best at tracking their target to begin with. The K-1 3D tracking is really not a viable option, so that would be out. It would be some spot focus that would work the best. It would take them quite a bit of time to improve their tracking. And sure in the end they might be a better photographer for it. But that is just scratching the surface. The hunting of the K-1 auto-focus system will show itself as soon as they go for those small DOF. And any dim light situation or some situation that reduces the contrast a bit, like fog, and the miss focusing of the K-1 would only get worse. How do you tell this person that the K-1 is a better action camera than the D500?

The K-1 accel at many things but action photography isn't one of them. With practice, one can become proficient, but you will be limited by the lack of tools in the toolbox.

[Dan Rentea]

The K-1 accel at many things but action photography isn't one of them. With practice, one can become proficient, but you will be limited by the lack of tools in the toolbox.

I do agree with you. With practice some occasional action/wildlife can be taken with K1 and we have the proof of it in the 300mm+ lens thread. But K1 is far away from being an action orientated camera and it was never advertised like one. No one with experience in shooting with other systems will say that K1 can compete with Canon/Nikon when it comes to action because it lacks:
- speed (with 4.5 fps you can miss a lot of important moments)
- af tracking performance
- buffer and precessing power; it takes 20s or more for the buffer to clear 10-12 shots

But then again, with practice anyone can have some good action images taken with K1. If you're shooting action for a living, then Pentax is not the right system for the job. If you're an amateour with high expectations for action, then again, Pentax may not be the best option. But if you go out shooting some action for fun and you don't care about the numbers (70-80-90% of images in focus), then K1 may deliver better results than expected. Knowing the camera's limitations, knowing how to deal with those limitations and also with a lot of practice, K1 can do a decent job in this department (action/wildlife). Also, changing a little the mentality can help. Almost everybody who has only 10-15-20% of images in focus blames the camera. I also used to do that. But then I've started to blame me instead my cameras and this pushed me to practice. Better results started to apear soon after I stoped blaming the camera and after I started working on my technique.

So yes, if you want to become proficient in action photography, Pentax lacks tele lenses and tracking performance. Let's see what K3 Mark III will bring to the table in this regard.

[Digitalis]

I wish the mirror could go up faster.

The greater the acceleration the higher the inertia. Pentax went through a lot of trouble to dampen mirror shock with the K-1. As it is i'm happy with the camera, it's no D5 AF speed demon, but as i'm sure I don't have to remind you: people in the past did fine catching photos of kids without AF.

[shardulm]

Do we know what lens was used in these videos.. I have a feeling that the wider zooms will behave better for AF-C in general because of much less AF travel it needs to do in general.

[Jeffvan]

The greater the acceleration the higher the inertia. Pentax went through a lot of trouble to dampen mirror shock with the K-1. As it is i'm happy with the camera, it's no D5 AF speed demon, but as i'm sure I don't have to remind you: people in the past did fine catching photos of kids without AF.

I think they are referring to the latency between auto-focus acquisition and shutter release. It is a problem with the K-1 (maybe other Pentax) cameras that don't have the correct settings.
These settings mostly resolves that issue:
Menu 1
--Phase Detection AF
----AF Mode: AF.C
----AF Active Area: SEL
----AF.S Setting: Focus Priority
----1st Frame Action in AF.C: Focus-priority
----Action in AF.C Cont.: Focus-priority
----Hold AF Status: Off
Menu 2
--Crop: FF
--Image Capture Settings
----File Format: RAW
----JPEG Recorded Pixels: XS
----JPEG Quality: 1-star
----RAW File Format: PEF
----Color Space: AdobeRGB
--Dynamic Range Settings: All Off
--Noise Recution: All Off

Just setup one of your User modes for action shots with these settings. As mentioned earlier, I didn't proceed to test DNG file types so I don't know if they latency returns with those file types.

Do we know what lens was used in these videos.. I have a feeling that the wider zooms will behave better for AF-C in general because of much less AF travel it needs to do in general.

It is better to completely skip the concept of getting into what the focal length was. It is simpler just to refer to the Depth of Field (DOF) size, since that is what matters with perceiving what is in focus. It effectively gets to the same concept you have. Increasing your DOF will hide any flaws of the autofocus system. With the "Autofocus system" I include the ability of the lens into the system. The response speed of the lens can make a big difference!

The factors that determine the size of your DOF is the Focal-Length, f-stop, Pixel Pitch, and distance of your subject. I will list the effects one by one for those that don't know:
Pixel Pitch: the larger the pixel pitch the larger your DOF will appear to be.
f-stop: The larger your aperture (Small F-Stop values) the smaller your DOF will be.
Focal-Length: The larger your focal length the smaller your DOF.
Distance of your subject: The further away your subject, the larger your DOF will be.

So you can play with any of these 4 factors to manipulate your DOF. The smaller your DOF the more precise your focus of your camera needs to be and the more important the accuracy of your focus ring will have to be. Less throw in your focus ring and your system will have to be that much more precise to be in focus.

Almost everybody who has only 10-15-20% of images in focus blames the camera. I also used to do that. But then I've started to blame me instead my cameras and this pushed me to practice. Better results started to apear soon after I stoped blaming the camera and after I started working on my technique.

This is a great way to improve your own technique! I agree! But at some point we do have to be honest, which you were! I just like to express that honesty with hard numbers. My procedure was to measure the area of the center spot focus sensor. Next, I went and shot several action shot sequences (~100 shots per test run) in great conditions, with an extremely small DOF. Then I analyzed how many of the shots had the center spot area in focus. My numbers turned out to be 41% at max. Next I had conditions where the situation was worse (less contrast on the subject) and the number fell to 10%. In total my tests had 2000 shots. For some people these numbers mean nothing and does nothing for them. But such information allows me to plan and develop better techniques for my self. I know when it is a low contrast subject (such as a subject in a wooded area), I need to pre-prefocus on where my subject will be, and take the shot as they get to that spot. Or plan to know my subject and take a shot when they have the smallest speed in my direction. In good light conditions I can increase my DOF and get more in focus shots that way.

One additional note, most action shots where the subject is moving towards the camera, can become extremely boring real quick. Action shots usually means something when the subject interacts with what ever the action is (such as the environment), unless there is something interesting about the subject alone. So getting more in focus (larger DOF) will usually make the shot more interesting. But in the rare occasion where the subject's face reveals their struggle, and you want to isolate that (small DOF), those percentages I listed above will become important. Depending on the light conditions you might have to take 10 shots for the focus to land where you would like it to be.

Being able to reduce the size of the DOF has benefits in another way. Sometimes we do have low light conditions with a moving subject. If you plan to freeze your subject you might need a high shutter speed. With the combinations available between ISO, shutter speed and f-stop, it is great to have the option available to open the aperture to keep the ISO lower and the shutter speed high. Yet again it comes back to options, or tools in your toolbox. Understanding the numbers listed above can help you yet again, in planning. If you are going to be in these situations often, then look for an camera that accel at action.

One tip, if you don't need large resolution files you can "expand" the size of your DOF by reducing the resolution of your image. Reducing the size of your image will effectively increase the circle of confusion.

P.S.
My tests had the following setup, if you want to have a reference for the DOF. I was using the Pentax 70-200 lens with the f-stop at 2.8 and the Pentax K-1 camera. The subject was between 15 and 5 meters for my tests. That is a 0.9 meter DOF at 15 meters and a 0.098 meter DOF at 5 meters. (Out of the 2000 shots I had, finding the usable settings, I had about 500 shots with the good settings. Out of those shots I got these https://www.flickr.com/gp/153388317@N08/A5w17a. You can look at the EXIF data and pixel peep large versions.)

With the Nikon D3000 I was using a 300mm lens at f/4. At 15 meters the DOF was 0.371 meter and at 7 meters it was 0.079 meter. (On a 27 shot test with 3D tracking I got a 41% accuracy. One of the shots was this one: https://www.flickr.com/gp/153388317@N08/87Mi22)

If anyone is interested in how to measure your center spot phase detection sensor and create a photoshop action to show it on your picture, let me know; since I can make a video of how to do this. I am prepare to do this if there is interest. You can get this kind of layer, that you can turn on and off, on your picture to see where the sensor was https://www.flickr.com/gp/153388317@N08/5mx5bg. If you look at the original size (click the download icon and select view all sizes and select Original) you can see that the grass was selected as the focus and not the dog. This allows you to learn how the focus system works. What I learned from this is; that if I wanted the dog in focus, I should have placed a focus spot on the grass next to him, since the grass offered good contrasting lines.

Then again, most of my shots are not action shots:
https://www.flickr.com/gp/153388317@N08/UnJ037
https://www.flickr.com/gp/153388317@N08/jae51G
https://www.flickr.com/gp/153388317@N08/qY91W8