Sam Bennett wrote:
Read the last dozen or so pages of the thread. The issue people are reporting is that at f/1.2 focus is perfect and it gets worse as you stop down. Specifically, past f/2 at close distances the lens backfocuses pretty badly.
I've read the thread on and off since it started.
What button are you guys focusing with? Has that been discussed? The 'stopped down' problem almost sounds like a software issue in the camera.
Sam Bennett wrote:
For what it's worth, I think I've successfully established this is not the issue. I purchase an Ec-B split-prism focusing screen and tested focusing with the lens wide-open, and then focusing with the lens stopped down (by using DoF preview) - the result was the same. Focusing in either case with the split-screen results in a perfectly focused image. If focus shift was an issue, the wide-open shot would be out, while the stopped-down shot would be in focus.
Sam, with all due respect, I interpet your results to reach the opposite conclusion. I would ignore the wide open shot in this evaluation. If you use a split prism screen to focus stopped down and get perfectly focussed results, to me, that indicates there is a possibility that focus shift when stopping down is causing the problem. If I understand, if focus shift is the problem, then focusing wide open and then stopping down for the shot would sahift the focus plane and cause misfocus. The way you tested, focusing stopped down with a split prism, would prevent focus shift from throwing off the shot, so your test proves that focus shift might indeed be the problem. Am I missing something?
I'd love to see just how sharp you have gotten the lens to perform at 2.8 by focusing that way... is it as good as the 1.2 shot aaronlam posted? If so, that would be very promising for a firmware fix.
RoyPertchi wrote:
If I understand, if focus shift is the problem, then focusing wide open and then stopping down for the shot would sahift the focus plane and cause misfocus.
Yes, this is what should happen. Needless to say, all focusing by default using all the camera based focus options - auto-focus or MF, assisted by the camera through the MF indicator lights - will occur with the lens wide open. So if focus shift is occurring the camera is unable to take that into account.
RoyPertchi wrote:
The way you tested, focusing stopped down with a split prism, would prevent focus shift from throwing off the shot, so your test proves that focus shift might indeed be the problem. Am I missing something?
Yes, and I'm not sure where the confusion is arising. When using the split-screen to focus, both the f/1.2 and f/2.8 shot is perfectly in focus. The expected behavior would be that the shot focused at f/1.2 but then exposed at f/2.8 would be OOF. But as I've shown, they're identical. So, no focus shift.
RoyPertchi wrote:
I'd love to see just how sharp you have gotten the lens to perform at 2.8 by focusing that way... is it as good as the 1.2 shot aaronlam posted? If so, that would be very promising for a firmware fix.
Yes, the lens is quite sharp at f/2.8, as one would expect. I'm not going to get into posting proof of this since its so incredibly subjective and highly dependent on what someone is used to in terms of post-processing. Since I'm a high ISO guy, I virtually never sharpen my photos before the noise reduction step, so my standard for "sharp" is much different than most.
Sam Bennett wrote:
.. The expected behavior would be that the shot focused at f/1.2 but then exposed at f/2.8 would be OOF. But as I've shown, they're identical. So, no focus shift.
Oh, the first was focussed at 1.2, then shot at 2.8... I thought the first was focussed at 1.2 and was shot at 1.2 and the second was focussed at 2.8 and shot at 2.8. The way you did it, focusing at 1.2 and shooting at 2.8, then focusing at 2.8 and shooting at 2.8, yielding identically sharp results, does indeed eliminate focus shift as the culprit. It also seems to say that it's a firmware problem, I think. Thanks for clearing that up
For the fun of it, with your caveats accepted, I'd still love to see how sharp you got your manually focussed 2.8 shots to be, if you have a moment to post a sample.
Now that Sam's methodology for ruling out focus shift caused by optical design flaw has been completely clarified, I agree with Roy that something in the firmware seems to be a likely cause of the problem.
This might be a technique for a cross-check of the logic involved:
Locked down on a tripod, do a series of your 3/4 view portraits with focusing always done by AF, but with the camera always set to f/1.2. Set your stopped down apertures after AF is completed. Naturally this requires assigning AF to the thumb buttons.
Yesterday I shot a series of about two dozen 3/4 view head and shoulders shots from wide open to f/2.8. The focus was consistently on the near eye. In the shots at 2.8, the DOF clearly covers a little more of the closer cheek and extends to the closer eyelashes of the distant eye, but focus is definitely peak at the near eye, where I wanted it.
I don't know if there could be something different in the firmware of my lens or body, or what might cause you (Sam and others) to have a very real problem, but I don't seem to have it. It makes me wonder if Canon has intentionally designed the AF algorithms to shift focus slightly when stopping down a certain moderate amount. It could be useful, I suppose, to focus on the closest part of a subject and automatically get maximized DOF behind that plane. I doubt many of us want that, though, and some of us don't seem to get any shift (or too little to notice) in the plane of focus when we stop down.
Watch, the next Canon will have a new custom function: "shift plane of focus to maximize useful depth of field"! Should I add this to the PMA rumor thread? Remember, you read it here first!
Here's an example of the AF'd shot at f/2.8 compared to the split-screen focused shot at f/2.8. These were prepared with Lightroom's standard sharpness/NR settings - 25, 0, 25.
AF Misfocusing (this is the best I was able to achieve with AF)
Sam, you've got a great chart and test setup. Have you tried doing the AF with camera set at 1.2, then stopped down after AF? It seems to me your earlier tests showed accurate AF with lens wide open. Yes, of course I know that AF is always done with the lens physically wide open, but the setting on the camera seems to affect the results that you get.
Photon wrote:
Sam, you've got a great chart and test setup. Have you tried doing the AF with camera set at 1.2, then stopped down after AF? It seems to me your earlier tests showed accurate AF with lens wide open. Yes, of course I know that AF is always done with the lens physically wide open, but the setting on the camera seems to affect the results that you get.
If I get what you're saying, I'm not really sure what it would show. You're saying, set the camera to f/1.2, focus, then set it to f/2.8 before taking the shot? I haven't done that particular test, no.
Sam Bennett wrote:
If I get what you're saying, I'm not really sure what it would show. You're saying, set the camera to f/1.2, focus, then set it to f/2.8 before taking the shot? I haven't done that particular test, no.
Yes, that's what I'm suggesting. What it would show is that the focus is being moved in response to the set aperture.
I see only two ways that AF could be accurate at 1.2 and off at 2.8:
1. The lens has an optical design that results in major focus shift with stopping down.
2. The AF system changes the focus depending upon the aperture setting.
Make sense?
Now, it would be possible for the AF system to adjust the focus after you change the aperture, but I don't think it could do that when the AF system isn't activated (which would be the case using the thumb buttons, if you release the button after focusing.)
Photon wrote:
Yes, that's what I'm suggesting. What it would show is that the focus is being moved in response to the set aperture.
Yeah, seems like a worthwhile test. Unfortunately my lens is already on its way back to Canon. If it comes back and it's still screwed up, I'll give that a shot.
Does anyone know wat they do when they recalibrate the focus of a lens? Do they do it via software or do they actually realign the elements? I was thinking about tat on the way to work today and also wondered if there is a computer chip in the lens itself. I seem to remember hearing about rechipping a 35L once.
I did some extensive talking to CPS during my 3- 50L trips....and from what I understood...
1. Basic focus adjustment is through changing an electronic setting in the lens.
2. However, on the 50L if the focus does not perform in a linear fashion....then they have to re-align the elements for "best focus", and then recalibrate the electronic setting as well.
So to answer your question..is it can be a combination of both. Many of their lenses are NOT done this way....but my guess is since in the new 50L it appears that ALL elements move as a group inside the barrel....it possibly creates a slightly different set of challenges in the way they align lenses.
Photon wrote:
Yes, that's what I'm suggesting. What it would show is that the focus is being moved in response to the set aperture.
I see only two ways that AF could be accurate at 1.2 and off at 2.8:
1. The lens has an optical design that results in major focus shift with stopping down.
2. The AF system changes the focus depending upon the aperture setting.
fwiw, i discovered that all three of my copies appeared to backfocus more when stopped down simply because it misfocused from the beginning using AF. let's say that at 1.5' (minimum focusing distance) it would backfocus by 1/2" at f/1.2. this was enough to throw off any results from f/1.2 through f/5.6 before the DOF limit range would catch up to the subject plane and put it within the near limit. using MF and the focus confirmation light provided the same results, telling me that the focus confirmation system in the lens (firmware perhaps?) was incorrect at close distances. using MF and trusting the focusing screen instead of the confirmation light produced tack-sharp results all day long at any aperture.
my standpoint is that it doesn't focus shift whatsoever. if it did then i would never have had sharp results when using MF and the focusing screen, stopped down or otherwise. perhaps there are copies out there that have this anomaly but in the three that i tried this wasn't the case. it was caused by incorrect AF operation every time.
while canon may claim that AF is the responsibility of the body and not the lens, i don't buy it. why would my 35L, 50/1.8 mkII, 135L, 300/2.8L IS and every other AF canon lens i have ever owned nail focus at minimum distance without a problem but the 50L comes along and in three copies can't do it? the lens seems to have more to do with AF than i originally thought, that's for sure.
Went home for lunch and took a few shots with the previous suggestion of focusing at 1.2 and then taking the shot at 2.8. This time I used a tripod and it made no difference. I took two shots at just 2.8 and two at 1.2-focus, shoot at 2.8. I did have the * at my Af. All shots were identiclal. focus was the left battery again.
ward1066 wrote:
Went home for lunch and took a few shots with the previous suggestion of focusing at 1.2 and then taking the shot at 2.8. This time I used a tripod and it made no difference. I took two shots at just 2.8 and two at 1.2-focus, shoot at 2.8. I did have the * at my Af. All shots were identiclal. focus was the left battery again.
2.8 pic
Clearly you have a problem that I'm not having.
Did you also shoot at wider apertures? Are you finding consistent back focusing, or like Sam are you finding that focus is good at 1.2, 1.4, 1.6...?
ward1066 wrote:
Its good at 1.2 but progresssively backfocuses as it is stopped down, my quick test is back on page 174 https://www.fredmiranda.com/forum/topic/467828/173
To me, it looks as if the best focus is on the second battery in all of the wide aperture shots, and tending toward the third battery at 2.5 and 2.8. In other words, based just on that test, it looks as if you're getting back focusing at all apertures, more so than any progression as you stop down.
Two different and independant french magazines ("Chasseur d'Images" and "Réponse Photo") tested the 50 f/1.2. It appears that the wider aperture gives a lot of trouble to the autofocus system leading to very poor focusing accuracy (even on a 1DsMarkII)...
The results show that the f/1.8 has the better focusing accuracy, the f/1.4 second and the f/1.2 the worse.
This would be linked to the wider apertures used during focusing that lead to reduced microcontrast, which is not compensated by the increased luminosity : the autofocus has trouble locking.
I suspect that for fast lenses, this is most troublesome around 50 mm because with longer focal lengths, the reduced depth of field helps with focusing accuracy; and with wider lenses, the increased depth of field renders the focusing error irrelevant in the final image.
Therefore, I think I will stick with my f/1.8 lens for now... I am really missing FTM :-(
Sam Bennett wrote:
Here's an example of the AF'd shot at f/2.8 compared to the split-screen focused shot at f/2.8. These were prepared with Lightroom's standard sharpness/NR settings - 25, 0, 25.
