Ricoh-Pentax recommendation for K1 specific AF-C tracking settings?

I dont have K1, only K3, thus I cannot check, but, IMO, this post by Awaldram deserves a special thread, with no arguing about K1 benchmarking performance, just trying to understand if some key settings have been overlooked to enable the AF-C tracking:

http://www.dpreview.com/forums/post/58044338

Quote from Awaldrampost:

PENTAX Real-time Scene Analysis System
  • This RICOH-original technology is available during viewfinder shooting when the exposure mode is set to Scene Analyze Auto and the Custom Image mode is set to Auto Select.
Supported by the PENTAX Real-time Scene Analysis System, the K-1’s Auto Tracking function accurately detects the subject’s motion based on various factors including color, then keeps pinpoint focus on the subject throughout the imaging process by automatically shifting the in-focus point.

In further posts in the same thread he shows the back LCD displaying what these settings (buried in submenus) look like.

IMO, it is worth that those of us who have a K1 try some shooting with these settings and report if there is a noticeable difference in AF-tracking performance.

If K1 AF-C tracking needs this kind of special settings, this would mean:
  • that a complete Ricoh-Pentax AF-C documentation must be developed and made available on Ricoh-Pentax website (maybe it is buried somewhere in the japanese site? ;-) )
  • that the camera settings was most probably wrong when passing the DPR bicycle test, and it should be redone;
 
I dont have K1, only K3, thus I cannot check, but, IMO, this post by Awaldram deserves a special thread, with no arguing about K1 benchmarking performance, just trying to understand if some key settings have been overlooked to enable the AF-C tracking:

http://www.dpreview.com/forums/post/58044338

Quote from Awaldrampost:

PENTAX Real-time Scene Analysis System
  • This RICOH-original technology is available during viewfinder shooting when the exposure mode is set to Scene Analyze Auto and the Custom Image mode is set to Auto Select.
Supported by the PENTAX Real-time Scene Analysis System, the K-1’s Auto Tracking function accurately detects the subject’s motion based on various factors including color, then keeps pinpoint focus on the subject throughout the imaging process by automatically shifting the in-focus point.

In further posts in the same thread he shows the back LCD displaying what these settings (buried in submenus) look like.

IMO, it is worth that those of us who have a K1 try some shooting with these settings and report if there is a noticeable difference in AF-tracking performance.

If K1 AF-C tracking needs this kind of special settings, this would mean:
  • that a complete Ricoh-Pentax AF-C documentation must be developed and made available on Ricoh-Pentax website (maybe it is buried somewhere in the japanese site? ;-) )
  • that the camera settings was most probably wrong when passing the DPR bicycle test, and it should be redone;
 
.... However, we won't go anywhere if we chaotically jump from one theory to another. I'm with Simon on this, there really doesn't make any sense to have subject tracking only on auto.

And I don't think Ricoh Imaging actually claimed that.
I guess this has already been posted, but I haven't spotted it:

"Supported by the approximately 86,000-pixel RGB light-metering sensor and the new PRIME IV imaging engine, the PENTAX Real-time Scene Analysis System analyzes such factors as brightness distribution in the image field, the subject’s primary color and its motion with great accuracy and efficiency. In addition, by adopting a breakthrough artificial intelligence technology called deep learning to its image detection algorithm *, this system assesses each individual scene more accurately, while optimizing the accuracy and performance of the light-metering, exposure-control and autofocus systems, and selecting the most appropriate finishing touch for a given scene or composition."

* "This RICOH-original technology is available during viewfinder shooting when the exposure mode is set to Scene Analyze Auto and the Custom Image mode is set to Auto Select."

http://www.ricoh-imaging.co.jp/english/products/k-1/feature/05.html

This says "deep learning" is available in Auto. I guess it doesn't say it isn't available otherwise, but that would be a strange way of saying it.
.... Yet people are able to take even action shots with it... is it "broken", or just "not up to par"?
As I've said many times, I primarily use "Spot" - center AF point only. So what I do can't be used to judge other options, one way or the other.
 
.... However, we won't go anywhere if we chaotically jump from one theory to another. I'm with Simon on this, there really doesn't make any sense to have subject tracking only on auto.

And I don't think Ricoh Imaging actually claimed that.
I guess this has already been posted, but I haven't spotted it:

"Supported by the approximately 86,000-pixel RGB light-metering sensor and the new PRIME IV imaging engine, the PENTAX Real-time Scene Analysis System analyzes such factors as brightness distribution in the image field, the subject’s primary color and its motion with great accuracy and efficiency. In addition, by adopting a breakthrough artificial intelligence technology called deep learning to its image detection algorithm *, this system assesses each individual scene more accurately, while optimizing the accuracy and performance of the light-metering, exposure-control and autofocus systems, and selecting the most appropriate finishing touch for a given scene or composition."

* "This RICOH-original technology is available during viewfinder shooting when the exposure mode is set to Scene Analyze Auto and the Custom Image mode is set to Auto Select."

http://www.ricoh-imaging.co.jp/english/products/k-1/feature/05.html

This says "deep learning" is available in Auto. I guess it doesn't say it isn't available otherwise, but that would be a strange way of saying it.
Yes, they clearly said that. However! That's only for the part after "In addition...".

The first part is not related to the deep learning technology.
.... Yet people are able to take even action shots with it... is it "broken", or just "not up to par"?
As I've said many times, I primarily use "Spot" - center AF point only. So what I do can't be used to judge other options, one way or the other.
Barry, you were able to prove with our pictures that the K-1's AF is not "broken". Of course, you do have the skills to use the camera... which a tripod doesn't ;-)
 
....

The following text is not referring to the Deep Learning/Auto algorithms: "Supported by the approximately 86,000-pixel RGB light-metering sensor and the new PRIME IV imaging engine, the PENTAX Real-time Scene Analysis System analyzes such factors as brightness distribution in the image field, the subject’s primary color and its motion with great accuracy and efficiency."
But it goes on:

"In addition, by adopting a breakthrough artificial intelligence technology called deep learning to its image detection algorithm *, this system assesses each individual scene more accurately, while optimizing the accuracy and performance of the light-metering, exposure-control and autofocus systems ...."

* "This RICOH-original technology is available during viewfinder shooting when the exposure mode is set to Scene Analyze Auto and the Custom Image mode is set to Auto Select."

It links "deep learning" to "image detection algorithm" to "optimizing the accuracy and performance of the ... autofocus systems" to "when the exposure mode is set to Scene Analyze Auto and the Custom Image mode is set to Auto Select".
 
....

The following text is not referring to the Deep Learning/Auto algorithms: "Supported by the approximately 86,000-pixel RGB light-metering sensor and the new PRIME IV imaging engine, the PENTAX Real-time Scene Analysis System analyzes such factors as brightness distribution in the image field, the subject’s primary color and its motion with great accuracy and efficiency."
But it goes on:

"In addition, by adopting a breakthrough artificial intelligence technology called deep learning to its image detection algorithm *, this system assesses each individual scene more accurately, while optimizing the accuracy and performance of the light-metering, exposure-control and autofocus systems ...."

* "This RICOH-original technology is available during viewfinder shooting when the exposure mode is set to Scene Analyze Auto and the Custom Image mode is set to Auto Select."

It links "deep learning" to "image detection algorithm" to "optimizing the accuracy and performance of the ... autofocus systems" to "when the exposure mode is set to Scene Analyze Auto and the Custom Image mode is set to Auto Select".

--
Yes, but it doesn't say it's about tracking. Autofocus is just one point - and it has to be about AF-S, since that's the only mode available. It might be just about selecting a suitable focus point.

Alex
 
I agree Simon, thank you.

I tried it as well, and it didn't work for me either.

However, what does seem to help is to:
  • shoot jpgs with no lens corrections on
  • shoot in aps-c mode
  • I'm having most success in SEL-9 or SEL-25 AF modes when i pan with the subject
I seems to me that the K-1 processor is getting overloaded with tasks when shooting RAW, with corrections on, saving RAW+Jpg and doing complex AF calculations at the same time. This could account for many of the discrepancies we're seeing regarding AF performance.

I have 3 AF lenses, the 3 FA limiteds. Obviously none is a sports lens, but the 43 focuses quickly, definitely good enough for kids running around in the garden if you're OK with the focal length.

--
-----------------------------------------------
Miles Green
Pentaxian with chronic LBA
Corfu, Greece
N.B. All my images are protected by Copyright
 
Last edited:
Is it possible, that in their desire to build a small FF dslr and pack so much into it without the experience that Canikon have ..... they have overcooked the processors ability to keep up ?

Does this theory stack up ?

I used jpeg only for my sequence and it worked great.

Do we have to delve into various menu's to turn things on or off to activate the best out of the machine? thereby leaving only the tech savvy or those with time to read their manual at the side of a football pitch before the game ??

or with a list each time you choose to shoot some action .... hang on, i'll just get my K1 flip card menu .... !!!

I activated AF-C, in jpeg used 33 auto points and that's enough for me.

Certainly looking forward to the next firmware update !!!
 
Is it possible, that in their desire to build a small FF dslr and pack so much into it without the experience that Canikon have ..... they have overcooked the processors ability to keep up ?

Does this theory stack up ?
yes, i think that's the problem as well
I used jpeg only for my sequence and it worked great.
me too.
Do we have to delve into various menu's to turn things on or off to activate the best out of the machine? thereby leaving only the tech savvy or those with time to read their manual at the side of a football pitch before the game ??

or with a list each time you choose to shoot some action .... hang on, i'll just get my K1 flip card menu .... !!!
Save your ACTION settings to a USER mode. :)

Save your usual settings to another USER mode, so that you don't fumble around when you're done shooting action.
I activated AF-C, in jpeg used 33 auto points and that's enough for me.
I'll deactivate all the lens corrections, they are heavy on processing power.
Certainly looking forward to the next firmware update !!!
Me too.
 
What about, in for example TAV-mode (or AV):

- AF-C

- 33 AF-points

- Matrix metering

- AF-hold set to off (by default it is set to medium... What was the case in the DPR test?)

- Lens correction to off

- RAW

This setup works best for me so far with the the DFA 150-450mm and BIF.

Greetings
 
Last edited:
What about, in for example TAV-mode (or AV):

- AF-C

- 33 AF-points
There are two of these: "Auto" (no pre-selected AF point), and "Expanded Area AF (Large)" (with a pre-selected point).

Which are you referring to?
- Matrix metering

- AF-hold set to off (by default it is set to medium... What was the case in the DPR test?)

- Lens correction to off

- RAW

This setup works best for me so far with the the DFA 150-450mm and BIF.
 
What about, in for example TAV-mode (or AV):

- AF-C

- 33 AF-points
There are two of these: "Auto" (no pre-selected AF point), and "Expanded Area AF (Large)" (with a pre-selected point).

Which are you referring to?
Auto!
- Matrix metering

- AF-hold set to off (by default it is set to low... What was the case in the DPR test?)

- Lens correction to off

- RAW

This setup works best for me so far with the the DFA 150-450mm and BIF.
--
http://www.barrypearson.co.uk/
http://en.wikipedia.org/wiki/User:Barry_Pearson
 
What about, in for example TAV-mode (or AV):

- AF-C

- 33 AF-points
There are two of these: "Auto" (no pre-selected AF point), and "Expanded Area AF (Large)" (with a pre-selected point).

Which are you referring to?
Auto!
Thanks!

I'm going to have to try this. Perhaps I'll experiment a bit on planes on Saturday.
- Matrix metering

- AF-hold set to off (by default it is set to low... What was the case in the DPR test?)

- Lens correction to off

- RAW

This setup works best for me so far with the the DFA 150-450mm and BIF.
 
I have the K1 in my hand and I can tell you that changing the Custom Image mode doesn't affect X-Y tracking at all. Setting the exposure mode to Scene Analyze Auto
Multi system user here (And I had to spend about a week reading the Canon forum to get a 7DII set up correctly once). What is Scene Analyze Auto? I saw only Auto on the exposure dial.
does affect the tracking - it disables it (and you cannot use AF-C).
Perhaps they mean set Program Line - on menu 1 - to Auto instead? Or is that Custom Image mode referred to earlier? I almost always set this to neutral to set up for post.
I think this is a translation error. I would suggest that it's trying to say that the AF-tracking uses the RTSA system, but that doesn't mean you need to 'activate' it.
Or it is only active in certain shooting modes which weren't clearly translated. Or it isn't what we think it is. Maybe other settings, such as, turning off lens corrections & shooting in jpeg are obfuscating peoples perceptions by improving things to an acceptable level.
It would be very odd if Pentax only offered this feature in a mode where you cannot select AF-C, or if you select one specific JPEG rendering mode.
It would be an implementation error they could fix though.
SJ

--
Simon Joinson, Editor-in-chief & GM
dpreview.com
 
Perhaps they mean set Program Line - on menu 1
The program line settings are only active when P or Sv mode is selected. or when you customize the E dials in TAv or M mode, see page 46 of the manual.

My earlier explanation and pic (in another thread) of my user setting for AF.C show this is how I've set the K1 up for action shooting.

--
Cheers Brian
http://500px.com/bjlawson
Smoking is a mugs game, nearly got me!!!
 
Last edited:
I dont have K1, only K3, thus I cannot check, but, IMO, this post by Awaldram deserves a special thread, with no arguing about K1 benchmarking performance, just trying to understand if some key settings have been overlooked to enable the AF-C tracking:

http://www.dpreview.com/forums/post/58044338

Quote from Awaldrampost:

PENTAX Real-time Scene Analysis System
  • This RICOH-original technology is available during viewfinder shooting when the exposure mode is set to Scene Analyze Auto and the Custom Image mode is set to Auto Select.
Supported by the PENTAX Real-time Scene Analysis System, the K-1’s Auto Tracking function accurately detects the subject’s motion based on various factors including color, then keeps pinpoint focus on the subject throughout the imaging process by automatically shifting the in-focus point.

In further posts in the same thread he shows the back LCD displaying what these settings (buried in submenus) look like.

IMO, it is worth that those of us who have a K1 try some shooting with these settings and report if there is a noticeable difference in AF-tracking performance.

If K1 AF-C tracking needs this kind of special settings, this would mean:
  • that a complete Ricoh-Pentax AF-C documentation must be developed and made available on Ricoh-Pentax website (maybe it is buried somewhere in the japanese site? ;-) )
  • that the camera settings was most probably wrong when passing the DPR bicycle test, and it should be redone;
 
Last edited:
Is it possible, that in their desire to build a small FF dslr and pack so much into it without the experience that Canikon have ..... they have overcooked the processors ability to keep up ?

Does this theory stack up ?
As a software engineer I think it is more software than hardware at this point. And because its software it is actually a more difficult problem as they can't just buy the best AF ASIC chip from some hardware vendor. It is a software intellectual property advantage. People keep harping on 'Scene' analysis or whatever. Nikon doesn't brag about this like Canon does with iTR, their AF just works. Notice how Canon/Pentax are more fiddly about settings? How Canon's iTR can actually *reduce* keeper rate? Software problem.

The reason I think it is software is because Nikon AF is very good *across* models and from mirrorless to DSLR. It is good on a Nikon 1 J5 that can shoot 20 FPS...

Canon just *recently* released the EOS M10. An 18MP mirrorless with crappy autofocus *again*. If they had excellent software AF IP, AF should be better across the board by now from top to bottom. Only Canon's top bodies are good at AF and they require significantly more hardware to do so. Tech organizations with poorer software IP always try to compensate by throwing better hardware at the problem but are always left behind.

Panasonic has *excelllent* software AF IP with DFD...now any recent releases tend to be *excellent* at AF since *any* of their cameras from high end MFT to a bridge camera, now incorporates excellent software AF IP.

Fuji is open about 're-writing' the firmware of much older releases because they are essentially the *same* hardware, but the *software* has improved.

The best solution to this problem is to poach Nikon/Panasonic software engineers, but there are obviously regional IP protection laws to deal with ;p. Or swallow pride and engage in discussions with Samsung who showed in a short engineering cycle how it could be done.

AF advantages might not be something that can be overhauled overnight. Yes Samsung showed they can do it, but for an existing organization it may mean existing company politics with outdated senior software engineers and depts with egos and reputation. Sensor DR advantages is something that is more easily overhauled as you can just shop around for decent sensors many of which have good DR.

WIth DSLR's it is even more tricky because it is a software problem with tight integration with hardware (dedicated PDAF sensors of different natures). Regardless if you have the *best* software AF IP, you can always couple it with good hardware, and if you have poor software AF IP, even with the best hardware it will still lag.
 
Last edited:
Is it possible, that in their desire to build a small FF dslr and pack so much into it without the experience that Canikon have ..... they have overcooked the processors ability to keep up ?
You've got a lot of concurrent processes the camera is capable of. The more you ask the camera to do at once, the more it's going to have an issue. Now, some of it can be code - if the code is not doing it's job efficiently. But only to a point. That's why Canon's flagships have dedicated processors for AF. That gets expensive - not just in money but also real estate within the camera.

If I want to shoot action with my 5dIII I turn off RAW+JPEG and just shoot jpeg. My buffer can't do it well. I didn't try shooting RAW+JPEG for action long enough to see how adversely AF was affected.

I mean, think about it - If you're using your PC and you've got a couple heavy-duty apps running you start getting delays. Those delays can go away when you upgrade memory or CPU. These cameras are no different. If you want the machine to be able to do everything concurrently you need better CPU or you start shutting things down.

Do you honestly expect a manufacturer to tell you that if you use all the bells and whistles we gave you at the same time you'll see a performance hit? They're not going to do that. At the same time, they have to weigh the cost of going with better or multiple chips against that performance loss and how many people are going to experience it.
Does this theory stack up ?

I used jpeg only for my sequence and it worked great.

Do we have to delve into various menu's to turn things on or off to activate the best out of the machine? thereby leaving only the tech savvy or those with time to read their manual at the side of a football pitch before the game ??

or with a list each time you choose to shoot some action .... hang on, i'll just get my K1 flip card menu .... !!!

I activated AF-C, in jpeg used 33 auto points and that's enough for me.

Certainly looking forward to the next firmware update !!!
 
You've got a lot of concurrent processes the camera is capable of. The more you ask the camera to do at once, the more it's going to have an issue. Now, some of it can be code - if the code is not doing it's job efficiently. But only to a point. That's why Canon's flagships have dedicated processors for AF. That gets expensive - not just in money but also real estate within the camera.

If I want to shoot action with my 5dIII I turn off RAW+JPEG and just shoot jpeg. My buffer can't do it well. I didn't try shooting RAW+JPEG for action long enough to see how adversely AF was affected.

I mean, think about it - If you're using your PC and you've got a couple heavy-duty apps running you start getting delays. Those delays can go away when you upgrade memory or CPU. These cameras are no different. If you want the machine to be able to do everything concurrently you need better CPU or you start shutting things down.
Yet I would rather browse the web using a recent version of Chrome on an outdated CPU, than load up a brand new PC with an early version of Netscape.

The fact that Canon needs to start adding 2x the hardware to compete on AF is not good. Maybe that is why even more entry level Nikon's can track well yet *still* give leading class battery life. Because the *software* AF is good.
 
Pentax previously got into bed with Samsung to solve their software ip with a jiont camera development contract

Samaung stitched them up like a kipper releasing their Milcs and dropping their k10d copy leaving all their customer in the lurch and Pentax with egg on their face.

So I don't see Pentax going that route again and are probably shy of 'joint' cycles as they've been badly bitten .
 

Keyboard shortcuts

Back
Top