PDA

View Full Version : AF Area question?


David Morison
8th November 2013, 09:43 PM
I am sure that there is a simple answer to this but if the zoom function is used to select the size of the single point AF (not "small") does this change the number of on sensor AF sites used? The size of the single point can be changed from x5 to X14 by zooming, and the smaller size appears equivalent to one of the 81 AF sites whereas the larger size appears equivalent to perhaps four (estimated by reference to the centre 9). I am really struggling to get my head around this and I am only asking because the larger single point appears to have some advantage over the centre 9 for BIF.

If I'm talking rubbish please ignore me but I hope someone will be able to shed some light on this for me.

Thanks

David

bredman
8th November 2013, 10:43 PM
x5 to X14 by zooming

I can't help you David.

Now. There seems to be 2 slider scales that appear with the M1. The 'x5 to x14' and the 'x2.5 to x7'. If anyone knows how to toggle between these two scales i'd love to know. The x2.5 to x7 has appeared twice for no apparent reason, then disappeared again for no apparent reason.

David Morison
8th November 2013, 11:07 PM
The 2.5 to 7x is when the 2x DTC is in use.

David

bredman
9th November 2013, 12:46 AM
Thanks David. I should have realised that as it almost explains another issue i had with the DTC :). My brain is fried.

photo_owl
9th November 2013, 11:17 AM
I'm not convinced that the use of the magnification functions changes the active size of the AF sensor in use at all.

the change to (s) does

the use of different lenses, and working distances, will change the relationship between the subject elements and the AF sensor but the use of the magnification function only (but usefully in many cases) enables the placement of the AF sensor area to be made more accurately on the target.

whilst it was relatively easy to map the relationship between normal and small points on the PD E's I'm not so convinced about the M1 ones - I may try at some point!

this was the E3

http://farm4.staticflickr.com/3466/3780246145_24d33b4232.jpg

from which it can be seen that at normal setting the points had no effective gaps between them, but as (s) they did.

bilbo
9th November 2013, 11:24 AM
from which it can be seen that at normal setting the points had no effective gaps between them, but as (s) they did.Sorry, I don't follow you at all... can you explain a little more please?

photo_owl
9th November 2013, 12:05 PM
Sorry, I don't follow you at all... can you explain a little more please?

blue is normal on the E3, red is (s), overlay the blue (only shown on the centre point here) on all the red bits and you won't find many 'gaps'.

but the M1 AF points aren't crosses in the first place...

bilbo
9th November 2013, 12:41 PM
blue is normal on the E3, red is (s), overlay the blue (only shown on the centre point here) on all the red bits and you won't find many 'gaps'.

but the M1 AF points aren't crosses in the first place...Gotcha. Thanks. *chr

David Morison
9th November 2013, 12:43 PM
So if I use a 5x zoom this only covers the same area for AF as 14x zoom, even though it is 2.8 times larger?

pvasc
9th November 2013, 02:58 PM
Check page 54 of the manual. It shows the focus points and the zoom frame. The only question is, is it accurate or is it for illustration?

David Morison
9th November 2013, 03:49 PM
Check page 54 of the manual. It shows the focus points and the zoom frame. The only question is, is it accurate or is it for illustration?

Yes I've studied that and having also owned the E-M5, which is the same, the mechanics of selecting the size of the single target AF point I fully understand. But my question relates more to the science - does changing the size of the single AF target affect the number of AF sites used? My feeling is that it does otherwise why use the function for anything other than magnifying the image for MF?

David

photo_owl
9th November 2013, 04:11 PM
So if I use a 5x zoom this only covers the same area for AF as 14x zoom, even though it is 2.8 times larger?

from a manual perspective, as pvasc references it down to page 54

unfortunately it's open to interpretation both in terms of the wording of AF zoom "You can zoom in on a portion of the frame when adjusting focus. Choosing a high zoom ratio allows you to use autofocus to focus on a smaller area than is normally covered by the AF target. You can also position the focus target more precisely." This could be interpreted both ways, although I agree it has a strong implication that the relevant AF point becomes smaller relative to the subject.

the example 'grid' shown doesn't help because it starts with all 81 points, and even at 14x it doesn't go smaller than single point (s) might be expected to cover.

it should be a relatively simple thing to test (using a 60macro would be the easiest way) but I'll have a go with the 12-50 tomorrow.

David Morison
9th November 2013, 05:55 PM
it should be a relatively simple thing to test (using a 60macro would be the easiest way) but I'll have a go with the 12-50 tomorrow.

Yes, why didn't I think of that, thank you. I did a quick test using the 75-300mm @ 100mm (to limit DoF) and positioned a small subject 9 inches in front of a larger one. I this case it was the top of a cough mixture bottle (yes, I have a cold) in front of a cereal packet with writing on it. I positioned myself so that the top of the bottle filled the 14x target area. I then took a photo of the bottle top with the zoom frame at 14x and then at 5x with the subject dead centre of the enlarged target area, and which overlapped the subject. The image with the 14x zoom frame had the bottle top in perfect focus whereas the one with the 5x zoom frame had the top OoF with cereal packet in perfect focus. To my simple mind this would indicate that the 5x is employing more detector pixels than the 14x. The whole purpose of my original question is that using the 5x single target is a much faster and more accurate method of locking focus on a BIF than the centre 9 AF area, which makes sense to me as the detector pixels in the centre 9 don't talk to each other (as they do in the Canon 7D) whereas whatever the number of detector pixels in the single 5x target area must do.

Am I barking up the wrong tree?

David

photo_owl
9th November 2013, 08:18 PM
Yes, why didn't I think of that, thank you. I did a quick test using the 75-300mm @ 100mm (to limit DoF) and positioned a small subject 9 inches in front of a larger one. I this case it was the top of a cough mixture bottle (yes, I have a cold) in front of a cereal packet with writing on it. I positioned myself so that the top of the bottle filled the 14x target area. I then took a photo of the bottle top with the zoom frame at 14x and then at 5x with the subject dead centre of the enlarged target area, and which overlapped the subject. The image with the 14x zoom frame had the bottle top in perfect focus whereas the one with the 5x zoom frame had the top OoF with cereal packet in perfect focus. To my simple mind this would indicate that the 5x is employing more detector pixels than the 14x. The whole purpose of my original question is that using the 5x single target is a much faster and more accurate method of locking focus on a BIF than the centre 9 AF area, which makes sense to me as the detector pixels in the centre 9 don't talk to each other (as they do in the Canon 7D) whereas whatever the number of detector pixels in the single 5x target area must do.

Am I barking up the wrong tree?

David

David, don't know about trees, but I think I follow what you were doing. We agree about the key difference in Group function but not the size of active AF through the magnify process.

I've just run my testing as follows -

1. establish a background that the lens will snap to - as it's CDAF I used the back of a cork table mat; it's both flat and contrasty at the same time :)
2. put a 12-50 in macro and set up the plane of focus to match the slope of the mat at about 300mm distance.
3. set the AF to single point normal, and grabbed a pipe cleaner (string or wool will both work pretty well but the pipe cleaner is a one hand operation!)
4. holding the pipe cleaner in vertical and about 25mm in front of the background I moved it around chimping to establish the necessary overlap to the indicated AF point on the screen/vf - was about 1/2 the diameter. Less and it stayed with the background, more and it stuck on the pipe cleaner. At 1/2 it wasn't happy - but then again the edge of the cleaner is fuzzy anyway.
5. Having established how to estimate the actual AF point from positioning the target I then magnified to 5x and repeated. All good in that the marked on screen AF point acted in the same way.
6. Now the important bit - I marked the indicated AF point on the cork mat before going back to non-magnify. It was smaller than the AF point on the screen! Switching to single point (s) I got an exact match - bingo.

So, my initial conclusion based on a quick and dirty test, is that when you switch to AF magnify mode the single AF point reverts to (s), and that this is retained as you magnify further. As such the AF point gets smaller, and smaller, relative to the target size as it gets magnified but that physically it's the same size from 5x through 14x ie (s)

Incidentially, I'm not convinced that the group points aren't talking to each other - if you repeat the above set up with group centre, once they have locked to the mat they ignore the pipe cleaner even when it's straight across the middle 3. What get's indicated is an AF lock from whichever point has un-interpupted access to the mat, and this will change as the position of the target is moved when the AF is triggered by a half press. Now, I have AF lock HIGH and this may be a factor - again more to 'test'. However, the indication is that the group are interacting in some way - acting as one large AF point as a minimum but it's more given that they don't shift to a new target even if it covers the centre point completely. Hoever, I completely agree that this is a totally different behaviour to the PDAF group function on the E3/30/5 7D D3x etc which not only act with a centre priority but work together to predict changing position in 3D.

David Morison
9th November 2013, 11:36 PM
Thank you for this, you appeared to have carried out a much more detailed and scientific study of the issue than me. I gather from your testing that you believe that only one detector site is involved whether you use 5x or 14x zoom. This would make sense if the 5x, if one could believe the diagrams on p.54, didn't overlap adjoining sires.

Apart from my even shorter and dirtier test than yours I still find it difficult to understand why pointing a 5x AF target point at a small bird in s large bush usually ends up with a lot of infocus twigs whereas the 14x has a much better chance of getting the bird.

David

Ross the fiddler
10th November 2013, 12:46 AM
I can see various points being discussed here on this subject & I will through my bit in too (maybe it will only add to confusion).

With the M4/3's lenses they are using the Contrast Detect -AF (except when using C-AF Tracking on the E-M1) & the PEN's & OM-D's generally operate the same way. The magnification function from 5 X to 14 X selects the area size (of the sensor) that the contrast is detected for focus (as I understand it on the E-M5). If I am using my 75-300 lens for example, & I am trying to focus on a bird in a shrubby tree, there is a fair bit of distracting branches & leaves that could be focussed on instead of the bird so I use the 14 X for the smallest focus area (I usually do it by using the touch screen to get the position I want straight away) & I can quite often get to focus on the subject I want that way. Conversely though, if I use too small an area (14 X) on a subject with little contrast detail in it, I then need to enlarge that area so the camera has something of contrast detail to focus on. Another thing with contrast focus is it will focus on the subject plain with the most contrast detail in it & with a bird hiding behind some thin branches, the camera will actually focus on the bird & ignore the branches in front. This can also work against me when a bird is closer to me with a tree, shrub or something else behind when the focus locks onto the subject behind, even with the smallest focus point (14 X) & it needs to be prefocussed at a similar distance to that of the subject before the camera can actually see the bird to focus on.
A far as DTC use is concerned, the focus area doesn't change, but just a 2 X crop view is seen instead (with the zoom showing the half zoom values). I hope that helps & doesn't confuse.

*chr

David Morison
10th November 2013, 05:34 AM
Ross, this is exactly my experience with the E-M5 and M1, and the one great improvement with the M1 is the Small AF target point. You mention that the use of the 5x to 14x function selects the size of the area of the sensor that is used for contrast detect AF which would indicate that different numbers of detector sites are employed, the subject of my original question. And what about the "Small"target point, how does this relate to the number of detector sites used?

The magnify and "Small" target point functions are also employable using PDAF on 4/3 lenses with equal results.

I guess I'll have to email Olympus technical to get a definitive answer, unless Ian has the "knowledge"

David

Ross the fiddler
10th November 2013, 07:19 AM
Ross, this is exactly my experience with the E-M5 and M1, and the one great improvement with the M1 is the Small AF target point. You mention that the use of the 5x to 14x function selects the size of the area of the sensor that is used for contrast detect AF which would indicate that different numbers of detector sites are employed, the subject of my original question. And what about the "Small"target point, how does this relate to the number of detector sites used?

The magnify and "Small" target point functions are also employable using PDAF on 4/3 lenses with equal results.

I guess I'll have to email Olympus technical to get a definitive answer, unless Ian has the "knowledge"

David

I would think it would have to apply to the photo sites otherwise it wouldn't have a useful function. The twin or single cross hairs option of the E3, 30 & 5 DSLR's were for finer focussing situations & the smaller focussing boxes or areas would be for the same thing, so the 'small target point' should use the smaller area of detector sites too. Again, if there isn't sufficient detail to focus on with the small targets then larger or grouped focus points may need to be used. My experience with using single cross hairs on the E30 was less reliable focussing but with the default twin cross hairs it locked onto focus more reliably. You know the saying, "To be sure, to be sure" rings true here too. ;)

PeterBirder
10th November 2013, 07:47 AM
I'll chip in my two pen'orth on this interesting discussion.

In CDAF there are no "focus points". The camera uses the normal photo sites not any extra "focus sites". The focus algorithm mathematically assesses the average contrast in the actual image within a focus "box". It will look for the highest average within the box as the focus distance. If the scene within the box includes subjects of differing inherent contrast it will be biased to those subjects which have the highest inherent contrast. If the subject you want has less inherent contrast than its surroundings it may miss your subject. By making the focus box smaller you isolate your wanted subject from it's surroundings improving the chance of focusing correctly.

When you focus manually via the viewfinder using your eyes/ brain you are doing much the same. Your Mk 1 human brain may actually be better at this as it has the intelligence to know from pattern recognition which part of the image is the part you want. CDAF still works pretty well nevertheless.:)

Just my thoughts.*chr

David Morison
10th November 2013, 09:19 AM
Thank you Peter, that certainly helps to give me a clearer understanding of what's going on and fits with my experience. My reference to "detector sites" is a result of reading about "detector pixels" in a couple of well regarded reviews. However it's done I now have good reason to do more tests using the largest (x5) single box as an aid to BIF, I have found just recently that this appears to be a much faster and more accurate method (together with S-AF) than using the centre 9.

David

photo_owl
10th November 2013, 10:57 AM
I'll chip in my two pen'orth on this interesting discussion.

In CDAF there are no "focus points". The camera uses the normal photo sites not any extra "focus sites". The focus algorithm mathematically assesses the average contrast in the actual image within a focus "box".

thank you for highlighting this Pete - I was somewhat careless in my use of language earlier when refering to 'focus point' but my meaning related to that area of the sensor being used for CDAF at the time :)

my conclusion relating to the use of single point focusing is that when using focus magnification the camera switches to the actual sensor area associated with single point (s) for all magnifications.

today I will move on to what's happening if you start with group and add 5X etc!

OK here's what I found -

1. if you select a group focus area, which is normally an area associated with 9x the normal size active AF area, and then increase the magnification to 5x etc, the camera switches to the AF area associated with single point (s). ie once you switch to 5X it doesn't matter what AF area size you were originally using it move to the centre of that area and single (s) size. I want to check more precisely the relative size for 10 and 14 times as my current testing 'bed' isn't precise enough to be anything but indicative at that scale.
2. and interestingly it always centres the selected AF area in the viewfinder, which can be very useful with closeup and macro work but is slightly dis-orientating! there's an obvious logic to this action because you would otherwise have the AF area outwith the viewfinder, esp at higher magnifications.

as to the diagram of 'AF and zoom frames', I remain of the view that this is giving a representation of the effective AF area relative to the target but I need to come up with a new test bed to be comfortable with that - at the moment it's more 2+2...OK; theory was good but I knew it didn't match the facts exactly! BY replicating the total AF area grid as a target and matching it to the AF grid on screen, then cycling the magnifications it became crystal clear :) What they are trying to illustrate is the proportion of the target, in terms of the full AF grid overlay, that will be visible in the viewfinder when magnification view is used. ie at 10x what you will see in the full viewfinder (of your image) is pretty much what is covered by a normal single point AF area.



the really usefull thing for me is that I'm happy that the E-M1 is happy using the AC-1 for long periods - and I've also established that it doesn't charge the battery when connected :( (although I've only established this for a grip mounted flat battery at this point and it may differ for a body mounted one...)

David Morison
10th November 2013, 12:07 PM
if you select a group focus area, which is normally an area associated with 9x the normal size active AF area, and then increase the magnification to 5x etc, the camera switches to the AF area associated with single point (s). ie once you switch to 5X it doesn't matter what AF area size you were originally using it move to the centre of that area and single (s) size.


This is exactly what I found out when having trouble locking focus on a BIF using the centre 9, I pressed the button I had assigned to magnify and the 9 were replaced by 1, and because I had already set that to 5x I started getting more consistent results. This lead to me starting this thread and I think we have unearthed some information and experience that will serve to help everybody, if they're not confused. Thank you all for your contributions.

David

Ross the fiddler
10th November 2013, 12:17 PM
Thank you Peter, that certainly helps to give me a clearer understanding of what's going on and fits with my experience. My reference to "detector sites" is a result of reading about "detector pixels" in a couple of well regarded reviews. However it's done I now have good reason to do more tests using the largest (x5) single box as an aid to BIF, I have found just recently that this appears to be a much faster and more accurate method (together with S-AF) than using the centre 9.

David

My understanding is the review would be talking about Phase Detect-AF that is used for 4/3's lenses & for C-AF Tracking in a Hybrid mode for all (including M4/3's) lenses & Pekka Potka talks about it in his preview (http://www.pekkapotka.com/journal/2013/9/9/olympus-om-d-e-m1-hands-on-preview.html).

Gwyver
10th November 2013, 04:09 PM
...This lead to me starting this thread and I think we have unearthed some information and experience that will serve to help everybody, if they're not confused. ...
David

David,
Once you and your fellow experimenters have reached some generally applicable conclusions it would be really helpful if they could be summarised and added to Ian's Wiki on the FTU site. Thanks in anticipation.