[biz-engineer]

They are two entirely different auto-focus systems.

Yes, absolutely.

Liveview will show the image just as the sensor sees it so it's the ultimate in obtaining correct focus.

Yes, and.... no. Theoretically, CDAF should give a perfect focus, especially when the camera stops down the lens aperture while doing its CDAF routine. Practically, there are a few things that contribute to a focus error in live view mode that does not occur with PDAF. 1) There is no second focus tweak with CDAF as it's done with PDAF (focus confirm step with a last tweak of lens AF), CDAF approximate where the focus should be by doing calculation of focus position based on the amount of blur calculated from a few frames taken randomly while spinning the lens focus motor, there is a always a residual error due to interpolation, CDAF can only guess where best focus is via interpolating a sharpness indicator from image captures. 2) Depending on camera drive mode and exposure settings, CDAF is done with the lens aperture stopped down, e.g f11, f16 for a f2.8 lens, this eliminate focus shift from aperture but this also increase the depth of field making CDAF frames interpolation less precise compared to interpolation from a lens aperture wide open. You may want to re-focus one or two times in live view mode to reduce the residual focus error to a lower value than its first focus routine. 3) CDAF interpolation is not as good near infinity or at infinity because it can't interpolate from two values in front of subject and behind subject to reach a minimum error goal. Near infinity , interpolation is from front of subject frames only, so it would tend to be less precise than PDAF for subjects near infinity.

It is unlikely that live view is giving you an incorrect focusing information.

Yes and no, every servo loop has a residual error based on how much time is given for the servo to reduce the residual focus error. The longer the time given for the CDAF optimization the better but only to a point where residual error become random and the subject distance may have changed (like noise, the focus error measured by the system become small and inconsistent with the actual focus error, that the maximum resolution of the system). So, there is a tradeoff between speed and residual focus error for both PDAF and CDAF, and none is able to give perfect focus. On my K1 CDAF, in low light, with macro lens, if CDAF isn't able to find a focus plane, it does a second routine going much slower... different speeds for difference situations...

I turn off Live View and focus through the view finder and the focus point is slightly different. Any ideas?

It is entirely possible and I could also experience this. You may refocus more than one time on the subject with CDAF to achieve better focus, and ultimately the very best focus is achieved by manually tweaking the lens while looking a x16 zoom in liveview mode + while closing aperture with DoF preview function to also eliminate focus shift from aperture, i.e tweak the lens focus at the same aperture as the aperture that will be used for the exposure.

---

Last edited by biz-engineer; 09-17-2021 at 10:52 AM.

[Bob 256]

Yes, and.... no. Theoretically, CDAF should give a perfect focus, especially when the camera stops down the lens aperture while doing its CDAF routine. Practically, there are a few things that contribute to a focus error in live view mode that does not occur with PDAF. 1) There is no second focus tweak with CDAF as it's done with PDAF (focus confirm step with a last tweak of lens AF), CDAF approximate where the focus should be by doing calculation of focus position based on the amount of blur calculated from a few frames taken randomly while spinning the lens focus motor, there is a always a residual error due to interpolation, CDAF can only guess where best focus is via interpolating a sharpness indicator from image captures. 2) Depending on camera drive mode and exposure settings, CDAF is done with the lens aperture stopped down, e.g f11, f16 for a f2.8 lens, this eliminate focus shift from aperture but this also increase the depth of field making CDAF frames interpolation less precise compared to interpolation from a lens aperture wide open. You may want to re-focus one or two times in live view mode to reduce the residual focus error to a lower value than its first focus routine. 3) CDAF interpolation is not as good near infinity or at infinity because it can't interpolate from two values in front of subject and behind subject to reach a minimum error goal. Near infinity , interpolation is from front of subject frames only, so it would tend to be less precise than PDAF for subjects near infinity.


.

When I stated that Liveview is the best way to focus, I was referring to manual focus and not AF. Liveview AF does have its issues, however with manual Liveview focus using magnification, there are no focusing servo problems. This too has issues in that it's a subjective visual judgement just where the best focus is, and it can be hard to set the best point using a focusing ring on a lens this way.

[claytoncramer]

Viewfinder focus produces better images. This picture was very sharp in LiveView. I find myself wondering if the mirror moving may alter the focal point. It would not have to be much.

[biz-engineer]

12 second delay. There is no motion on the screen. This is not a tripod. This is a 17.5" diameter reflector telescope.

Ah Ok, no camera autofocus then, focus by eye, highly subjective.

---------- Post added 18-09-21 at 05:47 ----------

I find myself wondering if the mirror moving may alter the focal point.

The optical VF has diopter correction that LV doesn't have. That's the difference that makes the manual focusing work better using the OVF. If you focus using live view, just wear correction glasses, if you focus with OVF remove your glasses and adjust the OVF diopter.

---------- Post added 18-09-21 at 05:58 ----------

I’m pretty sure the OP is using the camera on a telescope. The keyword: prime focusThe Prime Focus Method (Astrophotography)

Interesting to me is that the article about prime focus talks about everything except focusing. I think we have an issue of communication here, basically incomplete information, at the beginning of the threat wasn't mentioned the use of a telescope, assuming we know it already, which makes diagnose the problem impossible.

---

Last edited by biz-engineer; 09-17-2021 at 09:02 PM.

[UncleVanya]

Ah Ok, no camera autofocus then, focus by eye, highly subjective.

---------- Post added 18-09-21 at 05:47 ----------


The optical VF has diopter correction that LV doesn't have. That's the difference that makes the manual focusing work better using the OVF. If you focus using live view, just wear correction glasses, if you focus with OVF remove your glasses and adjust the OVF diopter.

---------- Post added 18-09-21 at 05:58 ----------


Interesting to me is that the article about prime focus talks about everything except focusing. I think we have an issue of communication here, basically incomplete information, at the beginning of the threat wasn't mentioned the use of a telescope, assuming we know it already, which makes diagnose the problem impossible.

And yet diagnose it we did.

Two clues: 17.5” and prime focus. 17.5” isn’t a typical way lenses are expressed, some older lenses have focal lengths in inches, but 17.5”/444mm is an odd focal length. Prime focusing is literally attaching a camera to a telescope using the telescope focuser. The camera can’t focus except by operating the telescope focuser. It’s as though you have a giant manual focus lens.

[biz-engineer]

Two clues: 17.5” and prime focus. 17.5” isn’t a typical way lenses are expressed, some older lenses have focal lengths in inches, but 17.5”/444mm is an odd focal length. Prime focusing is literally attaching a camera to a telescope using the telescope focuser. The camera can’t focus except by operating the telescope focuser. It’s as though you have a giant manual focus lens.

Thanks for bringing some clarification. Why is it called prime focusing, not simply focusing? (I was expecting a special technique, but I've found nothing more than what I'd do focusing a lens manually). I guess what I'd do if I had problems with focus, I'd do a series of exposures shifting focus slightly prior to every exposure, then keep the sharpest image. I remember having done some focus interpolation manually based on graduated focus and ranking JPEG files by size in file explorer (keeping ISO constant), the JPEG having the largest size is the one that recorded the most detail (considering JPEG compression record pixel variation primarily, less sharp images product smaller JPEG file sizes). I pixel peeped the 4 best files and indeed, the sharpest image had the largest JPEG file size.

[MossyRocks]

Thanks for bringing some clarification. Why is it called prime focusing, not simply focusing?

My understanding is that people using telescopes have their own terminology when talke about things. Prime focus basically means no additional eye pieces in the optical formula other than the scope or any field flattners/coma correctors.

Scopes are also measured in odd ways basically they are measured as open aperture in inches and f-stop. So a 17.5" f/4.5 is a scope that has a 17.5" open aperture so working back to focal length that puts it at right about 2000mm. Using open aperture as a measure is actually useful as that determines what you can actually image since things are so dim and there are known minimum absolute aperture size to see them.

In looking at the image there may be some shake but it is also just out of focus. My suggestion is find a bahtinov mask that will fit a 17.5" scope and focus using that instead. Finding masks for scopes is easier than finding them for camera lenses. Once you have one find a bright star (at 17.5" aperture lots of stars a bright) turn focus peaking off and use magnified live view to get the diffraction spikes centered on the star. Do bother playing around with focusing in the view finder or playing with trying to guess focus.

[UncleVanya]

My understanding is that people using telescopes have their own terminology when talke about things. Prime focus basically means no additional eye pieces in the optical formula other than the scope or any field flattners/coma correctors.

Scopes are also measured in odd ways basically they are measured as open aperture in inches and f-stop. So a 17.5" f/4.5 is a scope that has a 17.5" open aperture so working back to focal length that puts it at right about 2000mm. Using open aperture as a measure is actually useful as that determines what you can actually image since things are so dim and there are known minimum absolute aperture size to see them.

In looking at the image there may be some shake but it is also just out of focus. My suggestion is find a bahtinov mask that will fit a 17.5" scope and focus using that instead. Finding masks for scopes is easier than finding them for camera lenses. Once you have one find a bright star (at 17.5" aperture lots of stars a bright) turn focus peaking off and use magnified live view to get the diffraction spikes centered on the star. Do bother playing around with focusing in the view finder or playing with trying to guess focus.

Is the moon going to be at “infinity” like a star or is there a difference in light years vs the moon?

[claytoncramer]

My understanding is that people using telescopes have their own terminology when talke about things. Prime focus basically means no additional eye pieces in the optical formula other than the scope or any field flattners/coma correctors.

Scopes are also measured in odd ways basically they are measured as open aperture in inches and f-stop. So a 17.5" f/4.5 is a scope that has a 17.5" open aperture so working back to focal length that puts it at right about 2000mm. Using open aperture as a measure is actually useful as that determines what you can actually image since things are so dim and there are known minimum absolute aperture size to see them.

In looking at the image there may be some shake but it is also just out of focus. My suggestion is find a bahtinov mask that will fit a 17.5" scope and focus using that instead. Finding masks for scopes is easier than finding them for camera lenses. Once you have one find a bright star (at 17.5" aperture lots of stars a bright) turn focus peaking off and use magnified live view to get the diffraction spikes centered on the star. Do bother playing around with focusing in the view finder or playing with trying to guess focus.

Yes. Prime focus means the mirror's focal point is at the sensor. Eyepiece projection puts an eyepiece between focal point and sensor. There is no camera lens. Attached is a prime focus image (cleaned up a lot with GIMP) and an eyepiece projection with a 25mm eyepiece. Bahtinov mask is a good idea but for some thing 20" diameter (tube diameter)? I have made a Hartmann mask instead. I was just hoping Live View would be an alternative.

---------- Post added 09-19-21 at 08:25 AM ----------

Is the moon going to be at “infinity” like a star or is there a difference in light years vs the moon?

Mountains 30 miles away and 60 miles away are theoretically going to need differences in focus, but I defy you to measure it. The Moon is effectively infinity. Everything that needs a telescope to image can be considered infinity.

[claytoncramer]

Thanks for bringing some clarification. Why is it called prime focusing, not simply focusing? (I was expecting a special technique, but I've found nothing more than what I'd do focusing a lens manually). I guess what I'd do if I had problems with focus, I'd do a series of exposures shifting focus slightly prior to every exposure, then keep the sharpest image. I remember having done some focus interpolation manually based on graduated focus and ranking JPEG files by size in file explorer (keeping ISO constant), the JPEG having the largest size is the one that recorded the most detail (considering JPEG compression record pixel variation primarily, less sharp images product smaller JPEG file sizes). I pixel peeped the 4 best files and indeed, the sharpest image had the largest JPEG file size.

Astrophotographers also use a technique called stacking where we use software such as Registax to gather similar pitcures (usually from video) and stack them to minimize noise. I have not quite mastered this yet. My first attempt reminds me of Blow-Up (1966) where the photographer discovers his fashion shoot catches a murder in the background. The attached result captured the alien building program instead.

[MossyRocks]

Is the moon going to be at “infinity” like a star or is there a difference in light years vs the moon?

Yes. Depending on what measure one uses for infinity from the lens's perspective they both would be a infinity. However picking a lamp post or tree down the street and focusing on it won't work because a telescope is so large. If we use even the most liberal definition of infinity from the lens's perspective of 500x focal length a 2m scope would be properly focused at infinity at 1KM this generally would be close but likely not perfect. Instead if we use something that would be the most conservative estimate for where infinity from the lens's perspective would be at 5000x that would put it at 10KM. Going even more conservative and saying 10,000x would put it at 20KM. Considering the moon at its closest is 356,400KM it would be at infinity for probably all scopes ever made even using the ultra conservative estimate of 10,000x.

Astrophotographers also use a technique called stacking where we use software such as Registax to gather similar pictures (usually from video) and stack them to minimize noise. I have not quite mastered this yet. My first attempt reminds me of Blow-Up (1966) where the photographer discovers his fashion shoot catches a murder in the background. The attached result captured the alien building program instead.

I have found that using Hugin to align and remap the images works great. I don't know if you are on an equatorial or not but if you aren't then something like Hugin would likely work better for alignment. From there I have been playing with using the Enfuse tool with contrast blending to combine the best parts of the aligned shots into one. Here is what I managed with my 400mm lens and 2x converter on friday:

Here as a thread I started on it over in the astro group with some more details:
Because I was working on technique - Astrophotography - PentaxForums.comFor that my basic work flows was:
Initial develop in RawTherapee, nothing much other than some basic crop of all shots (to try to minimize the file size), 2x upscaling using nearest neighbor, set the develop mode RCD+VNG4, and set the white balance and develop to 16bit Tiffs.
Feed the hundreds of shots into PIPP and only keep the 100 best and do a proper auto crop of each image
Feed those crops into Hugin to do a better alignment and exposure correction (get the brightness basically the same) and output remapped images
Use Enfuse on the command line to do a contrast blend of the Hugin remapped images.
Do some final cleanup in RawTherapee.

I do the 2x upscaling as that gives even more space to get a better alignment and with a large number of shots you can do a super resolution shot, or in astro image processing software terms 2x drizzle.