More on focusing

divamum · November 16, 2010

If this should be somewhere else, by all means move it, but since it's a continuation of discussion in this thread I thought I'd start here :thumb

I've gone back through a host of shots from the last few days in Zoom Browser to check the focus points and in many cases that's a clear indicator of why something was or wasn't true; I can absolutely see where the issue is, and don't consider it a problem.

This shot, however, perplexes me. The AF point is clearly marked as roughly the tip of her nose, but her entire face is soft, and the point of sharp focus is indicated by the white circle, on the ribbing on the sleeve of her jacket. That's the direction the flash was coming from (yes, wrong modifier for the job so it's way too harsh); not sure if that's significant or not. Perhaps I might have inadvertently moved after I focused?

Any suggestions what might be going on here based on this shot? I'm still trying to diagnose if there's a hardware problem, or if it's unwitting user error. Yes, tripod tests still to come

(uncropped, sooc raw shot - not even converted since it was a screencapture)

divamum · November 16, 2010

Yes, single shot. What's curious is that, checking dof master, I should have had about 6-13" dof (I'm estimating I was 15-20ft away). It's entirely possible I swayed forward, however - we were on a hill, she was being less than cooperative (despite the look on her face here, it was a battle!) and the lightstand kept threatening to go over.......

Even so, tis a puzzlement....

divamum · November 16, 2010

I was above, shooting down towards her.

These are, again, SOOC raws that I've saved as screencapturs, so no processing of any kind.

ETA: arm with bracelet

zoomer · November 16, 2010

Sometimes focus just doesn't hit it right, it happens.
If you take the exact same shot ten times refocusing each time 8 might be sharp and 2 could be less than sharp, your camera may tell you it is in focus but it really isn't.
The thinner your depth of field the more likely this is to happen.

My guess is on this one is that something was moving just a bit to much or your focus just missed, I am leaning toward camera missed focus since all parts are a smidge soft.
Give it some extra sharpening and tonal contrast and this should sharpen up enough.

Shutter speed of 200 with 135mm does not give much leeway for movement, just an fyi.

At those settings focusing on her nose vs her eye would not make a difference.

kdog · November 16, 2010

That second shot of her sleeve and forearm looks sharp to me. So I guess judging from the pose, this was back focused. One thing to keep in mind is that the focus indicator really only tells you which focus point was active, and not what you focused on. For example, if you'd inadvertently focused on her sleeve with the active focus point and recomposed it to her face before you activated the shutter, you would end up with exactly what you are showing here.

Matthew Saville · November 16, 2010

divamum wrote: »

Yes, single shot. What's curious is that, checking dof master, I should have had about 6-13" dof (I'm estimating I was 15-20ft away). It's entirely possible I swayed forward, however - we were on a hill, she was being less than cooperative (despite the look on her face here, it was a battle!) and the lightstand kept threatening to go over.......

Even so, tis a puzzlement....

This is exactly the type of situation that I focus on a LOT in my "CameraTalk" workshop. Distance, aperture, and movement is EVERYTHING. In simple terms, if they're close and you're shooting telephoto, and if you're both standing up, AND if you're in one-shot AF, then you just gotta count for out of focus frames. Even on the 7D. There is just too much opportunity for movement for you to trust an important pose to just ONE click.

Having said that, I'm betting you do have a back focus issue. Go into the AF fine tuning menu, and pull focus forward by +10 or +20. Just play around and see which images look the sharpest.

And of course as I mentioned, it would be best to do this from a tripod, photographing a still-life with clear detail and a clear transition from foreground to background. (No need to buy a $200 calibration gizmo, a tree trunk and some grass make a GREAT focal plane test environment!)

ziggy53 · November 16, 2010

I suggest that it is time to test the lens and camera combination for focus accuracy.

1) A brick wall, shot straight on and squared with, using a tripod. This is a pretty good test for front-focus, back-focus, field curvature and vignetting issues.

2) A fence line or similar, shot at an angle to the subject. Put a singular strong-contrast target on the top of the middle post and use a single focus point in the camera to focus against the target. This shows focus accuracy and/or how easily distracting for/aft objects influence AF accuracy (compared to the above wall shot). It also shows bokeh tendencies at different aperture settings.

3) A focus target/chart like in the following:

http://regex.info/blog/photo-tech/focus-chart

I do recommend testing these charts at twice minimum focus distance or greater. Most lenses, especially most zoom lenses, do not do their best at MFD. True macro lenses are the major exception.

insanefred · November 16, 2010

Your problem: You're using a Canon
Your solution: get a Nikon

Just kidding.

Or maybe I'm not?

Can you tell us if your AF is any better on continuous?

divamum · November 16, 2010

Tx guys - pouring rain and drak here right now, so can't nip outside to shoot tests, but they will be forthcoming in due course.... Brick wall (well, pillar) and fence readily available outside in brighter, drier conditions!

@Matt - I've seen you trailing your classes on Twitter - I'd totally be signing up if I didn't live on the wrong coast!

@Ziggy, this is from the shoot that started all this, btw, on Saturday - it's not a new set of photos, I just went back through the shoot systematically in Zoom Browser expressly to look for things like this.

@Insane Fred - I've never used continuous focus, actually - I've always used single point, single shot. Is there an advantage to continuous for portraits, which is 75% of what I do?

Matthew Saville · November 16, 2010

divamum wrote: »

Tx guys - pouring rain and drak here right now, so can't nip outside to shoot tests, but they will be forthcoming in due course.... Brick wall (well, pillar) and fence readily available outside in brighter, drier conditions!

@Matt - I've seen you trailing your classes on Twitter - I'd totally be signing up if I didn't live on the wrong coast!

@Ziggy, this is from the shoot that started all this, btw, on Saturday - it's not a new set of photos, I just went back through the shoot systematically in Zoom Browser expressly to look for things like this.

@Insane Fred - I've never used continuous focus, actually - I've always used single point, single shot. Is there an advantage to continuous for portraits, which is 75% of what I do?

I do lots of calibrations indoors too, actually, You can just put a cereal box on your carpet. Put the focus point on the flat surface of the cereal box, but compose the shot so you can see the carpet at the base of the box. Use your tripod with the legs not extended and voila, you have a lab test! Just try it from 6-10 feet away, and play around with the AF fine tuning.

Yes, for test situations single focus is the way to go. But continuous focus, aka AI-Servo, is actually really useful for portraits, it is not just a setting for sports photographers to shoot runners etc. with... Like I said I could spend a whole afternoon discussing it, (which is what I do hehe) ...but the bottom line is that if your subject is moderately close, and especially if your aperture is fast, ...then single focus (one-shot) is just not going to be accurate enough when considering the faint movements that occur with two people standing upright. For medium length and close up portraits, I almost always use continuous focus. As long as you can move the focus point around to properly frame your shot, just fire off 3-5 images and you'll get 2-3 in perfect focus. Assuming your camera is calibrated properly!

=Matt=

NeilL · November 16, 2010

dm, if you were getting repeat images with the same gear with consistent oof issues, then I think you would have to investigate. However, this doesn't seem to be the story you are telling, and my understanding is that AF doesn't come with any guarantees anyway, even on the 7D!

As well, it appears to me that there are other bits of the image that are likely in the same focal plane as the in-focus bit, such as parts of the subject's hair, which are not in focus. So there is a possibility that there is subject movement here, especially if to your coaching the subject "perked up", changing the posture of her upper body slightly while putting on a smile, just as you were depressing the shutter button. 1/200 would not have stopped this.

Neil

divamum · November 16, 2010

Tx Neil. If it had just been a shot here and there I wouldn't have given it a second thought, but there were WAY more oof on this shoot than I consider acceptable, hence why I've been trying to figure out if it was me, the situation, the camera or the lens. I'm working my way down the list.....

divamum · November 16, 2010

@richy (btw, literally laughing out loud at your shark+moving car post in the other thread): the 135l is just as fine at 2.0 as stopped down and I use it wide open all the time in the theatre. For portraits I sometimes forget to keep both eyes square to camera (whoops) which isn't so good with that limited a dof, so I try to consider ~3.2+ as the sweet spot for it when shooting at headshot distances, although sometimes I'll grab a few at 2.0 just because I can - it's 100% useable at maximum ap, and the bokeh is to die for.

rwells · November 16, 2010

My guess is that which I've already mentioned in a different thread, about the one design flaw on the 135L. For me, the focus ring is too big. I have to change my camera/lens grip to keep my hand off of the focus ring, and inadvertently changing the focus.

Simple Test:
Just grab your camera with 135L, and quickly put it up to your eye like you normally would to take a shot.

Now, don't move a muscle...

Take specific note of EXACTLY where your hand, or anything else, might be touching the focus ring.

If nothing is even "close" to touching the focus ring, (in both landscape and portrait orientation), then your likely OK with this particular error.

Hope that helps...

chrisjohnson · November 17, 2010

Assuming your DOF calculation is correct and why would it not be, this is much tighter than I would ever try to get all of her looking sharp given her side-on pose. But then I am an amateur.

After 40 years photography I accept it as an analogue medium - now trying to go digital. Most obvious is the weight of glass that has to be shifted by micro-motors, less obvious is the amount of DSP going on in-camera (even on manual mode) to try to give you the shot the Canon folks think you wanted to take. We are moving, the subject is moving, the background is moving. The DOF tables are for rough indication only. In this case the difference between the back of her and the front of her is greater than the DOF you have allowed yourself, even assuming the absolute accuracy of the component elements which is doubtful. This is a shot that was always likely to go wrong.

Practically speaking I would go for a bigger DOF tolerance - reduce the aperture. Maybe the bokeh is not optimally blurred but at least my subject would be sharper. At 1/200 and ISO 400 you have some room to manoevre, even hand-held.

I think you are pushing your system to the limits. This is quite legitimate to improve your understanding of how far it will go and in this case it does not go. The micro-adjustment argument looks sound and you can doubtless spend several hours photographing brick walls and fiddling with the micro-adjustment to get a zillionth of a percent more accuracy. Whatever turns you on. You could even ship your combo off to Canon for a calibration. Personally I would not bother: "different strokes for different folks".

These days, when I want something to be sharp (which is increasingly less often), I stretch the DOF so I don't have to worry. Nice question though, and I look forward to more opinions from the real experts.

billythek · November 17, 2010

I may be fooled by the resolution of the images posted on the web, but I don't see anything in sharp focus in that picture. The fabric has more high-contrast edges, which why it might seem more in focus, but at close inspection it doesn't look that sharp to me. I'm thinking camera shake. Did you use a tripod, or was this hand held?

divamum · November 17, 2010

Great suggestions and discussion, guys - tx for all the ideas!

@Randy - I will absolutely check that - never occurred to me!

@Goldenballs - Your point is a good one. In this instance, I needed the extra light gathering - I didn't want to go up to iso 800 (although I could have, I didn't want to - probably a misjudgement). Also, artistically, I hugely prefer shallow depth of field shots - it's pretty much what fires me up, so it's what I aspire to achieve. It has a much higher failure rate because the margin of error is much smaller, but when it works....

I usually expect OOF shots when I shoot like this (and make sure to shoot lots of frames to ensure that at least some are really sharp), but usually I can tell what the error was and where it went wrong; what's curious about this particular batch (and all of them in this position were duds, sadly) is that the entire string of shots in this position were all like this.

@BillytheK - entirely possible since it was handheld, although motion blur from this lens usually looks a little different than that (you can usually see it on eyelashes), and also it wouldn't be likely to vary between face and hand. It may be that there is ALSO a small amount of motion blur, but there is for sure a discrepency between the indicated focus point and where the image is sharpest. Also, these were not even converted to jpgs and as such are entirely unprocessed (I used screencapture to make these jpgs, thus the files are 100% SOOC without even any conversion); raw files from the 7d often look a tad flat and fuzzy before they've been run through a raw processor, so nothing unusual there.

Ok, time to go check out some bricks, fenceposts and rulers

NeilL · November 17, 2010

richy wrote: »

I'm thinking that one big half of the issue is that her head due to the slope of the hill is a couple of inches closer to the camera and therefore out of the area in focus. <...> I took some last week of our 2 yr old at 2.0 (i had wanted to use 1.4-1.6 but changed my mind) and missed the focus on several sets of shots. I feel your pain! When they eventually play the game and pose you dont want to get the shots wrong.

The selected focus point was undoubtedly on the subject's face. Even if the angle of the camera was not orthogonal to the subject (and I don't know why that would cause the active focus point to fail to give focus?), the tilt of the camera does not appear to have been extreme.

The point about dof is not that it affects the active focus point, but that a narrower dof produces more oof area in the image excluding the area (focal plane) selected by the focus point-AF. Dof does not explain why that has not occurred here.

Neil

NeilL · November 17, 2010

divamum wrote: »

Great suggestions and discussion, guys - tx for all the ideas!

@Randy - I will absolutely check that - never occurred to me!

@Goldenballs - Your point is a good one. In this instance, I needed the extra light gathering - I didn't want to go up to iso 800 (although I could have, I didn't want to - probably a misjudgement). Also, artistically, I hugely prefer shallow depth of field shots - it's pretty much what fires me up, so it's what I aspire to achieve. It has a much higher failure rate because the margin of error is much smaller, but when it works.... I usually expect OOF shots when I shoot like this (and make sure to shoot lots of frames to ensure that at least some are really sharp), but usually I can tell what the error was and where it went wrong; what's curious about this particular batch (and all of them in this position were duds, sadly) is that the entire string of shots in this position were all like this.

@BillytheK - entirely possible since it was handheld, although motion blur from this lens usually looks a little different than that (you can usually see it on eyelashes), and also it wouldn't be likely to vary between face and hand. It may be that there is ALSO a small amount of motion blur, but there is for sure a discrepency between the indicated focus point and where the image is sharpest. Also, these were not even converted to jpgs and as such are entirely unprocessed (I used screencapture to make these jpgs, thus the files are 100% SOOC without even any conversion); raw files from the 7d often look a tad flat and fuzzy before they've been run through a raw processor, so nothing unusual there.

Ok, time to go check out some bricks, fenceposts and rulers

Yes, it doesn't look to me to be camera shake, mainly. Although the sleeve is not brilliantly sharp (the area of focus would look sharper even in an unprocessed RAW) it is certainly sharp enough to be usable, and certainly sharper than the face. 1/200 with a 135 hand held is however a risk for camera shake, and for motion blur completely suicidal!

With the extra information about the series of shots this image is a part of - that the focus has similarly failed in all of them - we are back to square one. Motion blur, which I suggested earlier, would appear to be ruled out, as a major cause. Repeated and consistent focus problems with the same gear in the same circumstances would be a reason for a calibration-function check. However, it would seem from what you say that outside of this particular series there is no such problem.

I am tending to think that the cause of this result is not singular, but possibly a mixture of things - a little motion blur, a little camera shake, but mainly misfocusing by the AF, where the selected focus point has been overidden in favour of a higher contrast area. All focus points have an area of sampling which might well overlap each other, and the AF (which is never infallible) might in some few circumstances (eg where artificial light is very strong on some parts of a subject, as here) opt for a higher contrast item in the overlap of a nearby focus point. DPP shows your selected focus point, but that might not have been the active focus point, in this particular image, and in the others in this series with very similar composition and lighting.

I don't think there is any good reason to reduce aperture, but I do think you could consider using a higher shutter speed, and checking that your lighting is not causing the AF to stray by creating tempting very high contrast areas in the periphery of the sampling area of your selected focus point.

Neil

billythek · November 17, 2010

If you are doing static poses like this, I think you will get best results with a good tripod, live view + 10X manual focus, + remote trigger. Little vibrations make a big difference.

If you are shooting something moving, then crank up the shutter speed, use AI focusing, and hope for the best.

divamum · November 17, 2010

I'm not going to post them since tape-measure shots are beyond boring (!), but I did manage some testing today. The 135 focus was dead on the money without any calibration adjustment; I actually tinkered with the AFMF to see if it made a difference, and even the smallest adjustments threw it off the mark. So we can rule that out, at this point. I was using a tripod and spot focus and it was 100% accurate as far as I could tell.

I'm beginning to think Saturday's issues were a perfect storm of backlighting, low-contrast light on her face, high-contrast on the shirt she was wearing, a shutter speed on the edge of fast enough, and my legendarily bad handholding skills (which could well have included focusing then moving as well as not-steady-enough hands). I'll continue to monitor this until I'm SURE it isn't a hardware issue, but so far it seems to have been some kind of perfect storm of bad AF glitches. Then again, the 7d usually surmounts potential AF impediments, so I'm still puzzled, but there we have it. Any further developments, I'll post 'em for anybody who's interested!

Allen · November 18, 2010

I think this has been totally over analyzed. Nice discussion though. Both
camera and subject were free floating. DOF was small and exact focus AF was
achieved at a plane in that DOF. Only that exact plane would be in perfect
focus and the rest in 'apparent' focus. Very minor leaning in or out of
camera and/or subject would shift that exact focus plane.

I almost always spot AF and then manual tweak to hopefully an eye glint. I
shoot a lot of small birds and even with spot focus, which covers most of the
bird, there's no way knowing where that perfect focus plane is without that
final manual tweak.

Matthew Saville · November 18, 2010

Allen wrote: »

I think this has been totally over analyzed. Nice discussion though. Both camera and subject were free floating. DOF was small and exact focus AF was achieved at a plane in that DOF. Only that exact plane would be in perfect focus and the rest in 'apparent' focus. Very minor leaning in or out of camera and/or subject would shift that exact focus plane.

...What he said!

NeilL · November 18, 2010

Allen wrote: »

I think this has been totally over analyzed. Nice discussion though. Both
camera and subject were free floating. DOF was small and exact focus AF was
achieved at a plane in that DOF. Only that exact plane would be in perfect
focus and the rest in 'apparent' focus. Very minor leaning in or out of
camera and/or subject would shift that exact focus plane.

I almost always spot AF and then manual tweak to hopefully an eye glint. I
shoot a lot of small birds and even with spot focus, which covers most of the
bird, there's no way knowing where that perfect focus plane is without that
final manual tweak.

2 things -

1. remind your doctor not to "overanalyse", especially when your life depends on it! Advances in technology run on the tread of minutiae!

2. the active focal point determines that the focal plane intersects at that point. The selected focal point in dm's image was squarely on the face. So that is where the focal plane should have been.

Neil

kdog · November 18, 2010

NeilL wrote: »

The selected focal point in dm's image was squarely on the face. So that is where the focal plane should have been

Unless she inadvertently moved the camera after acquiring focus, but before activating the shutter (as I pointed out previously). In other words, a focus-recompose. In that case, one would not have a clue what the camera actually focused on from looking at the active focus point.

NeilL · November 19, 2010

kdog wrote: »

Unless she inadvertently moved the camera after acquiring focus, but before activating the shutter (as I pointed out previously). In other words, a focus-recompose. In that case, one would not have a clue what the camera actually focused on from looking at the active focus point.

Yep

However, there is one spanner in that idea's works, and that is that this problem did not happen in just one image, but in a series shot together. The problem was the same in each image in the series. A bit improbable that divamum did the leaning and swaying thing to the same degree consistently in all these shots, don't you think?

Neil

billythek · November 19, 2010

NeilL wrote: »

Yep However, there is one spanner in that idea's works, and that is that this problem did not happen in just one image, but in a series shot together. The problem was the same in each image in the series. A bit improbable that divamum did the leaning and swaying thing to the same degree consistently in all these shots, don't you think?

Neil

Well, one thing she did consistently was hand-hold without AI focus at somewhat marginal shutter speed for that lens. I would suggest repeating the test at 1/1000 using AI focus hand held, and then also compare to 1/250 non-AI, and also tripod using live view + 10X manual focus, remote release at various shutter speeds. You may have to use a different model, though.

Yes, it is a lot of work, but you may learn something about the value of good support.

kdog · November 19, 2010

divamum wrote: »

I've gone back through a host of shots from the last few days in Zoom Browser to check the focus points and in many cases that's a clear indicator of why something was or wasn't true; I can absolutely see where the issue is, and don't consider it a problem.

This shot, however, perplexes me.

NeilL wrote: »

Yep However, there is one spanner in that idea's works, and that is that this problem did not happen in just one image, but in a series shot together. The problem was the same in each image in the series.

divamum · November 19, 2010

billythek, while I don't disagree that a tripod can provide better support than I can, it just isn't the way I shoot and for many of the situations in which I shoot just wouldn't work. That does mean I discount its value, simply that, for me, it would be a very clinical "non real world" test... because I just don't use it in real life situations so much.

Also, one other thing to factor in: in the past shooting at similar setting I"ve had success with it! So, my goal is to figure out why this particular batch failed on me, and address any issues that I find. If it turns out to be the camera, I'll send it in for calibration; if it's more likely it was me, I'll figure out the exact element which let me down here and work on ensuring that isn't a problem in future shoots.

For the record, here's what the 7d and 135L CAN achieve... handheld:

f2.2 (cropped a bit, but not that much - I'd guess I was ~10-15ft away)

divamum · November 24, 2010

rwells wrote: »

My guess is that which I've already mentioned in a different thread, about the one design flaw on the 135L. For me, the focus ring is too big. I have to change my camera/lens grip to keep my hand off of the focus ring, and inadvertently changing the focus.

TINGTINGTINGTINGTINGTINGTING I think we may have a winner....

Just went to do some location scouting for a shoot on Sunday - used the 135 to see if I could get to the bottom of this.

Sure enough, Randy, I caught myself nudging that focus ring. I've never been aware of doing this before, but then it suddenly dawned on me that any "focus issues" I've had with the 135L have been since I got the 7d *where I'm not using a grip*. Particularly in portrait orientation, that means I'm holding it differently than I do when using the gripped xsi. This may well explain a LOT! Certainly, today, once I noticed that, my sharpness rate went back up to what I would consider normal, so I think we're definitely on to something...

While I still can't be 100% sure, I'm willing to bet that in the example that prompted this thread, it's a combination of mild camera shake+having moved that ring ever so slightly. I'll take that as one mystery solved for the time being!

Now, to get to grips with AI focus ... still haven't got that under my belt yet

Icebear · November 24, 2010

Hooray. Glad you figgered it out. Even more glad you have a "technique" solution.

More on focusing

Comments