Focus difference between live view & viewfinder

rockfall

rockfall

Suspended / Banned
Messages
247
Name
Tony
Edit My Images
Yes
Hi,

I seem to have a problem with my Canon 550d where if i use live view to focus it does so perfectly but when using the viewfinder and the middle focus point it is always slightly soft.

I've tried 3 different lenses and the all have the same issue, has anyone else encountered this? Surely the focus should be just as good with both the viewfinder and live view?

I have done tests outdoors in good light.

Thanks

Tony
 
The camera's phase detect AF system is out. It works independently of live view. It'll have to go back to be recalibrated (easy job).

It sounds like you have done all the right tests :(
 
oh right, well at least it's still under guarantee. how long do you reckon canon will take to fix it?
 
Are the images soft or is it just the viewfinder (diopter adjustment)?
 
The images are definitely soft, if I take one with and one without live view and then zoom in 100%, one is sharp and one is soft.
 
Have a look at a thread I created some time ago, back when I was more of a novice at this kind of stuff:

http://www.talkphotography.co.uk/forums/showthread.php?t=216110

I think part of the problem turned out to be camera shake in my images (particularly of the keyboard) but also that I was just expecting too much from an entry level camera body, in less than perfect light, or challenging conditions. When I stopped expecting miracles from the body every time, and just started enjoying the images it produced, I suddenly stopped being aware of any slight differences, which essentially become null and void once you've sharpened in post processing, anyway.

You could try posting some examples taking on a tripod, with live view and with contrast detect, to see what we can come up with. The problems I had, and what eventually lead me to believe I was just being picky, was that there was no pattern at all as to my 'focus issues'. Sometimes I would get slight back focus, sometimes slight front focus, other times it would be perfect. Regardless, unless I was pixel peeping 100%, mere mortals wouldn't have noticed anyway. If your issues are consistent every time though, you could indeed have a problem. let's see some pictures.
 
Last edited:
Are you using 'quick' or 'live' focus modes in liveview?

I've found that the live focus, although miles slower is extremely accurate, much more so than the usual viewfinder points. I always assumed this was because the viewfinder uses a different method and sacrifices a bit of accuracy for vastly improved speed. So is this actually a fault with the camera or just down to the different focus methods?
 
Im not sure if live view is on quick or live, i will check later.

I've just printed out a focus test chart so i will do some tests later and also take some real world samples with both live view and with the view finder.
 
HoppyUK said:
Haha! Good question!
Click to expand...

Sorry, crossed post with this comment that you have now answered.

If the problem has been there since new, then I would suspect a problem.

But if it has only just appeared, then it's unlikely the camera has decided to misbehave. Unless you've dropped it or something.

The thing is, you have checked Live View AF (and yes, not the quick AF which actually uses the same system as the viewfinder AF) and it is better than normal AF, with a range of different lenses, and at distance where other focusing errors are much less likely. Is that so?

I'm assuming you are comparing pic A in live view AF with an identical pic B taken using regular AF. If not, you must do that.

The camera's normal phase-detect AF is capable of extreme accuracy, but it must be calibrated correctly.
 
Do you have examples?

The two AF systems used are different - so it may be limitations based on the subject matter?

If you do send it for calibration you can generally expect the camera to be away for about a week.
 
Hi, thanks for your replies.

Yep I'm comparing pic A in live view AF with an identical pic B taken using regular AF.

I think the problem must have been there from new, I didn't think anything of it, just presumed it was what it was capable of until a friend tried out the camera and told me it was soft and recommended doing a comparison with live view, the live view image was noticeably sharper.

I will take some samples when i get home and post them here but last night i took a shot of a jar of coffee and the text on it was soft with normal AF but perfect with live view. I spoke to a chap about returning it and he said to do a test to see if it is front or back focusing.
 
Simple test. Get a cerial packet and line it up on the work top, square to the camera. Get a rule and lay it across the front of the box so it half in front and half behind the front surface.

Use a tripod if you can - not essential, but make damn sure you don't move a fraction of a mm after focusing! Shoot from 5-6ft and take a pic at lowest f/number. Raise the ISO as high as you need to get a shake-free shutter speed. Engage IS if you have it.

The face of the box should be sharp and, if you zoom in to the rule on the LCD, you should see the sharpest point coinciding exactly with the box. If it's out, you will see the sharp zone somewhere along the length of the rule.

Cross check with live view AF, and double check with another lens. Repeat until you are sure; if the AF calibration is out, it will be out every time, with all lenses. And the live view pic obviously will not be.
 
I've just done a quick test with this and thought i would post the results:

(on tripod, mirror lock up, 2sec timer, shot RAW with no PP sharpening, focused on the ES of cookies)
F/2.8 iso400 1/200
img8531c.jpg

100%
8531100precent2.jpg


F/8 iso800 1/50
img8533.jpg

100%
8535100precent2.jpg


I think there is a slight difference between the left and right photos (great at 2.8 than 8), but dont really think it is big enough to cause any real concern. Oh and just for the record, left is 'viewfinder' phase-detect and right is liveview's 'live' mode.

p.s, if these photos look like they've been warmed up in an oven for 2 hours then i blame my netbook's screen.
 
Last edited:
Blue Eagle said:
I've just done a quick test with this and thought i would post the results:

<snip>
Click to expand...

That doesn't look great BE, but I'm not happy with that test.

It is too close. Phase-detect AF is a compromise and it may be slightly better at some distances, and focal lengths, than others. Being very close emphasises any errors, but you are also pushing the system right to the worst end of its performance. It may be fine at a greater distance. 50x focal is the recommendation, or at least reasonably close to it.

You are not square to the target. The AF sensor is quite a lot bigger than the spot on the viefwinder. Only when you are fully square can that be ruled out.

Shoot square and put a rule across the front and behnind, so you can see where is it actually focusing.
 
Oops, i forgot about being square as in not looking down and yea i deliberately made it quite close to push both systems and have a shallow depth of field. I also reasoned that with a larger depth of field i would not be able to notice, or care, about slight inaccuracies.
Maybe you are right though as it might matter with my longer lens, i think i'll wait for rockfall's test and then maybe have another go :)
 
Sorry for the delay, here are the samples

Canon 50mm mk2 taken at f1.8

4950521691_c8e3a598af_b.jpg

Viewfinder, phase detect 50% crop

4951113780_567b26d900_b.jpg

Live view 50% crop


Tamron 17-50mm at 21mm and f2.8

4951092610_0eaaffd36e_b.jpg

viewfinder, phase detect 100% crop

4950496489_75451d9648_b.jpg

live view 100% crop


Looks to me like the camera is front focusing when using phase detect.
 
rockfall said:
Sorry for the delay, here are the samples

Canon 50mm mk2 taken at f1.8

Tamron 17-50mm at 21mm and f2.8

<snip>

Looks to me like the camera is front focusing when using phase detect.
Click to expand...

It does look like it from those shots, but...

Why are you shooting so close? As stated above, that will exagerate any issues to the point where you think you have a problem that in reality is insignificant, or quite possibly non-existant at normal range. And the 50 1.8 has a very crude focusing mech at the best of times.

I would do it again, from a distance of four or five feet at least, and put a mark on the rule next to the subject plane, and raise the ruler at an angle so you can see what's happening more clearly.
 
I took it from around the same distance as i would be if i was shooting a portrait so if it doesnt work in that situation then it isnt much use.

here is the original phase detect shot from the tamron at 21mm

4951093808_25273fd99d_b.jpg



I decided to send the camera back to the supplier to be fixed, better safe than sorry whilst it's under guarantee.

Tony
 
rockfall said:
I took it from around the same distance as i would be if i was shooting a portrait so if it doesnt work in that situation then it isnt much use.

here is the original phase detect shot from the tamron at 21mm

<snip>

I decided to send the camera back to the supplier to be fixed, better safe than sorry whilst it's under guarantee.

Tony
Click to expand...

Ah right. I didn't realise quite how much cropping was going on there. But seeing the full image, portrait distance is a pretty good range to test at. And there is clearly an error.

I wouldn't be happy with that.
 
Interesting. That's a nice sharp lens too.
 
Yea, that looks a lot worse than mine which were only a tiny bit off, i'd be sending it back too.
 
You must log in or register to reply here.
Back
Top