[UserAccessDenied]

I am having difficulty locating the answer to my question...

What is the duration between each of the 4 exposures when using Pixel Shift? Particularly with K-3ii.

There has to be some lag between the 1st shot and the sensor moving to take the 2nd shot, and so forth.
What is this duration?
1/6000s?
1/1000s?

I am not sure whether it could play a roll into the artifacts we see when the subject is moving, regardless I just want to know.
If there is some correlation, then perhaps a firmware update could lessen the time lag between each of the 4 shots to reduce the "artifacts" we see in PS.

Someone told me there is no lag, but that is physically impossible as the sensor is moving a pixel in each direction between shots. That movement takes time. Even the actuation of the shutter takes time. So what is it?!

Thanks to anyone who has the answer.
If you do come up with something, please post a source!

If there is no information on the subject; first, why not? And second, is there a way to test the duration between the 4 exposures when using PS?

-Logan

[stevebrot]

What is the duration between each of the 4 exposures when using Pixel Shift? Particularly with K-3ii.

Very fast, but also very real. The post exposure processing time is greater. The rub is that the total time the mechanical shutter is open is about the same as the X-sync speed (1/180s) and can be no faster. That is plenty of time to introduce blur to foliage courtesy of the wind.

If there is no information on the subject; first, why not?

There is plenty of information on the subject. A rule of thumb would be to take the nominal exposure value x 4 + mechanical shutter curtain dwell + a few milliseconds for the shift itself. Any higher precision is sort of silly.


Steve

[UserAccessDenied]

Very fast, but also very real. The post exposure processing time is greater. The rub is that the total time the mechanical shutter is open is about the same as the X-sync speed (1/180s) and can be no faster. That is plenty of time to introduce blur to foliage courtesy of the wind.



There is plenty of information on the subject. A rule of thumb would be to take the nominal exposure value x 4 + mechanical shutter curtain dwell + a few milliseconds for the shift itself. Any higher precision is sort of silly.


Steve

How can the total time not be any faster than 1/180s?
Can you link a source to this?

Also what do you mean by, "nominal exposure value x 4 + mechanical shutter curtain dwell + a few milliseconds for the shift itself. Any higher precision is sort of silly. "

Can you please clarify?

I'm pretty sure Pixel Shift uses an electronic shutter and not the mechanical shutter, I could be misinterpreting the information I've read though...

Thanks!

[enoeske]

The rub is that the total time the mechanical shutter is open is about the same as the X-sync speed (1/180s) and can be no faster...

Real Resolution (pixel shift) uses an electronic shutter for the 4 exposures and doesn't use the mechanical one.

[Adam]

How can the total time not be any faster than 1/180s?
Can you link a source to this?

Also what do you mean by, "nominal exposure value x 4 + mechanical shutter curtain dwell + a few milliseconds for the shift itself. Any higher precision is sort of silly. "

Can you please clarify?

I'm pretty sure Pixel Shift uses an electronic shutter and not the mechanical shutter, I could be misinterpreting the information I've read though...

Thanks!

The shutter is definitely electronic in between the exposures. I would say that the gap between the photos is short as the hardware will allow, but it's not really relevant to be honest, since any motion at all in the scene will still be observable at the pixel level.

On the K-1, you can see each frame as it is captured, and things happen really quickly:

You Tube

You Tube

The K-1 does have a new feature that detects motion and reverts to using a single frame for affected areas. With the K-3 II, you would need to do this manually in post with layers, etc.

You Tube

You Tube

[UserAccessDenied]

The shutter is definitely electronic in between the exposures. I would say that the gap between the photos is short as the hardware will allow, but it's not really relevant to be honest, since any motion at all in the scene will still be observable at the pixel level.

On the K-1, you can see each frame as it is captured, and things happen really quickly:

The K-1 does have a new feature that detects motion and reverts to using a single frame for affected areas. With the K-3 II, you would need to do this manually in post with layers, etc.

Thanks Adam!
wow, that's sweet how the K-1 handles PS...
I really wasn't thinking about upgrading to FF though, it's well above what I need from my camera and I actually like having a little crop factor for wildlife photography.
Just a personal preference^

So is this something that could be updated in the K-3ii firmware, or is it likely to never work the same way the K-1 is shown in that video?

I don't mind editing out layers in post, but that in camera function is cool. Also like how all 4 images are show in the screen while it's processing. It's overall just a slicker interface.

Thanks!

[nomadkng]

wow, that's sweet how the K-1 handles PS...

The problem is that's for JPG output only. RAW output requires a lot more user intervention, from what I can tell. For someone like me, who shoots 100% exclusively RAW, pixel shift on the surface doesn't seem as useable, yet.

I don't mind editing out layers in post, but that in camera function is cool. Also like how all 4 images are show in the screen while it's processing. It's overall just a slicker interface.

The PS images are NOT readable as layers.. again yet. Maybe the firmware 1.0 will allow that, but right now you have no control over the individual images in a PS composite. Adam showed in a demo how the sharpness? slider in Pentax PP software will allow you to access the composite PS image, but you can't access the individual ones. Same for Photoshop/Lightroom. I like the idea of super resolution, I really do. But until I can control layer masking of 4 separate RAW images in post myself, I'm having a tough time seeing this as game changing in my shooting process.

On top of that, in a field environment, my waterfalls/seascapes etc just aren't good fits for pixel shift. Standard half second to 5 second exposure times seem to render the idea moot even. Unless someone can further enlighten me? I honestly want to be excited about pixel shift, but I'm not seeing how it will help me at all.

[stevebrot]

How can the total time not be any faster than 1/180s?

1/180s (5.6 msec) is the fastest shutter speed at which the full image frame is exposed. It is that action that sets the minimum time to do a PS exposure. Any shorter and one or both of the mechanical shutter curtains will impinge on the frame.

Edit: While the struck out text above is true, the opposite is what happens in practice. The curtain dwell (start of leading curtain travel to end of trailing curtain travel) will be at least ~5.6 ms but that time span is incidental and not pertinent to the PS exposures themselves. I was fuzzy. See sequence discussion below.


Steve

[Oakland Rob]

The problem is that's for JPG output only. RAW output requires a lot more user intervention, from what I can tell. For someone like me, who shoots 100% exclusively RAW, pixel shift on the surface doesn't seem as useable, yet.



The PS images are NOT readable as layers.. again yet. Maybe the firmware 1.0 will allow that, but right now you have no control over the individual images in a PS composite. Adam showed in a demo how the sharpness? slider in Pentax PP software will allow you to access the composite PS image, but you can't access the individual ones. Same for Photoshop/Lightroom. I like the idea of super resolution, I really do. But until I can control layer masking of 4 separate RAW images in post myself, I'm having a tough time seeing this as game changing in my shooting process.

On top of that, in a field environment, my waterfalls/seascapes etc just aren't good fits for pixel shift. Standard half second to 5 second exposure times seem to render the idea moot even. Unless someone can further enlighten me? I honestly want to be excited about pixel shift, but I'm not seeing how it will help me at all.

You can separate the layers (more like frames I guess) with dcraw PS. You can then do whatever you'd do with four shots, like a bracket. We've had a couple of other threads here with details and examples. Eg: https://www.pentaxforums.com/forums/172-pentax-k-3/311865-pixel-shift-finicky-3.html

[stevebrot]

Real Resolution (pixel shift) uses an electronic shutter for the 4 exposures and doesn't use the mechanical one.

The exposure sequence begins with the mechanical shutter closed. Therefore the total time is gated by the curtain speed to open/close the opening. This is basic focal plane shutter stuff. I chose the 1/180s figure because it is the nominal X-sync speed which in turn is based on the speed of the curtains. The real sequence is a little more complicated, but distills to the same similar logic:

1. Shutter release pressed
2. Mirror up
3. Electronic shutter closed
4. Leading mechanical shutter curtain begins travel
5. Leading mechanical shutter curtain reaches end-of-travel (EOT, frame is now fully open)
6. Leading mechanical shutter curtain EOT confirmed
7. Pixel Shift exposure #1 *
8. Shift #1 **
9. Pixel Shift exposure #2
10. Shift #2
11. Pixel Shift exposure #3
12. Shift #3
13. Pixel Shift exposure #4
14. Trailing mechanical shutter begins travel ***
15. Trailing mechanical shutter EOT (frame is now fully obscured)
16. Mirror drops and mechanical shutter resets to ready state

* All pixel shift exposures are done using the electronic shutter at the shutter speed indicated by the exposure system. E.g. If 1/1000s, the time for each will be 1 ms or 4ms total for the actual exposures.
** I am assuming that the electronic shutter will reset/restage during the shift operation along with sensor reset and data flush. The time required for the combined operation is unknown, AFAIK, but is assumed to be speedy.
*** Image processing likely begins as the trailing curtain begins its travel. The camera will not be able to process another frame until this is accomplished.

As I was typing this last little bit, it occurred to me that I have been fuzzy in my thinking and explanation. The total cycle time is per the above sequence. The pertinent time window for subject motion of the PS system is simply the time it takes to do the four exposures + the time to shift/restage/flush. As such the mechanical shutter component is sort of a red herring unless the concern is total time per cycle.

Edit: In case it is not obvious, @enoeske is quite correct. I have bolded the time-critical steps for subject motion for emphasis.


Steve

[stevebrot]

On top of that, in a field environment, my waterfalls/seascapes etc just aren't good fits for pixel shift. Standard half second to 5 second exposure times seem to render the idea moot even. Unless someone can further enlighten me?

I too have been pondering this issue. I typically shoot moving water at 1/10s - 1s but think it would be really incredible to get PS detail and quality for the stationary portions of the frame. I don't currently own a PS-enabled body, but I would suspect that at least for the K-3II implementation, extraction of the first image to represent the blurred motion with later reintegration in PP might be the best solution (four images for the PS merge + one more for the blur. Yes, it would be very involved (horrible masking task), but probably also very worth it.

I am anxious to see how well the K-1 deals with this.


Steve

[Oakland Rob]

I too have been pondering this issue. I typically shoot moving water at 1/10s - 1s but think it would be really incredible to get PS detail and quality for the stationary portions of the frame. I don't currently own a PS-enabled body, but I would suspect that at least for the K-3II implementation, extraction of the first image to represent the blurred motion with later reintegration in PP might be the best solution (four images for the PS merge + one more for the blur. Yes, it would be very involved (horrible masking task), but probably also very worth it.

I am anxious to see how well the K-1 deals with this.


Steve

dcrawps can automate some of that.

You can set a threshold for moved pixel percentage (moved due to blur, not PS movement). 20% was seeming to be a default. It can pick up some tiny movement but that movement won't show in the image; sorta like very slight noise. There's a good example posted in the thread I cited.

And whether it's worth it depends on your output, even more so if you had the K-1 and a very sharp lens. It can exceed the ability of a poor lens or poor display to see the benefit.

Also, at slow exposures the blurring is OK if there is enough of it. Kinda like the difference between a 3 second shot of water flowing vs you just shot it at 1/60 handheld. The blur in each PS frame blends without the artifacts, which tend to show more on the edges.

Just get a copy of dcrawps and/or use PDCU and download some of the PS images and mess around with it to see if you find it worthwhile. Adobe and DxO also process the PS images.

[nomadkng]

You can separate the layers (more like frames I guess) with dcraw PS. You can then do whatever you'd do with four shots, like a bracket. We've had a couple of other threads here with details and examples. Eg: https://www.pentaxforums.com/forums/172-pentax-k-3/311865-pixel-shift-finicky-3.html

XPosted from that thread:

LOL... I have been using Photoshop since before the "CS" designation. I have a degree in software design, and NONE of this makes sense. I'm sure some of it is since I have NO clue what dcrawps even is. (sounds like a casino game actually...lol)

I also can't find a single mention in ANY adobe publication about being able to extract the 4 pixel shift images and import them as layers in Photoshop so I can do my masking there.

I consider myself pretty computer literate and an advanced Photoshop user, but nothing I have read anywhere calms my fears. If I (ME PERSONALLY) can't manually control layer masking using 4 images in post, pixel-shift is useless to me. I am willing to spend hour(s) in post, but I am NOT willing to use a jpg file, let a camera decide for me what pixels to keep, or even let a single slider do this. I may let SOME automation do some or a lot of the work, but at the end of the process I still want 4 layers that I can do manual quality control on. I want to be able to save my file with these separate layers in the event I want to re-edit the file, I don't want to have to start from scratch if I change my mind or realize I've missed something.

[stevebrot]

I have NO clue what dcrawps even is. (sounds like a casino game actually...lol)

dcrawps is a custom branch of the ubiquitous dcraw utility designed to support Pentax pixel shift files.

I also can't find a single mention in ANY adobe publication about being able to extract the 4 pixel shift images and import them as layers in Photoshop so I can do my masking there.

Not surprising. You use dcrawps to extract the four images from the PEF/DNG. FWIW, regular dcraw can do this as well. Edit: Wrong again Steve...regular dcraw cannot extract the images. Once you have the four images you can use your extensive PS experience or consult with other forum users as to the best way to leverage your harvest.

Dcraw home page: Decoding raw digital photos in Linux

Wikipedia entry on dcraw: https://en.wikipedia.org/wiki/Dcraw


Steve

[stevebrot]

I am willing to spend hour(s) in post, but I am NOT willing to use a jpg file, let a camera decide for me what pixels to keep, or even let a single slider do this. I may let SOME automation do some or a lot of the work, but at the end of the process I still want 4 layers that I can do manual quality control on.

It is unlikely that you will be able to duplicate the pixel shift merge process from the four layers. I have yet to try this out, but if I were to start shooting with a K-3II, I would likely use a flow something like this:

• Shoot in RAW (PEF or DNG, choose your poison)
• Use Lightroom for pixel shift RAW conversion
• Apply the usual LR non-destructive edits to fit the image intent

In other words, treat the file the same as any other capture. There is no need to edit the four images separately. To be honest, doing so would be contraindicated.

If there is motion artifact there are (at least) two possible paths:

• Process the original RAW in Lightroom as above. The original RAW is your master image.
• Use Lightroom's edit-in facility to work with the image in Photoshop (or similar)
• Extract the #1 image capture from the original RAW using dcraw and save as a flat curve* 16-bit TIFF and edit that TIFF as needed in either LR or Photoshop
• Use the extracted TIFF as the pixel source to mend the motion artifact in the master image.

In this scheme all edits except the final merge are non-destructive and can be modified at will using the usual Lightroom methods.

-- OR --

• Use dcrawps to mask the motion pixels with output to 16-bit flat curve TIFF
• Import TIFF into LR and process as usual. The flat curve TIFF becomes your master image.

Note that none of the above use JPEGs, in-camera processing, or PDCU. All leave the original RAW intact and the LR options allow for round-trip non-destructive edits.**

* I may have used the wrong term here. Dcraw has the ability to output the computed data as an image file with "flat" curves. The result is similar to an import in Lightroom with a linear contrast curve and all sliders zero'ed (i.e. no import profile).

** The "edit in" facility in LR is very powerful and allows real time simultaneous edits in both Photoshop and Lightroom on the same image.


Steve