7D Mark II - Focus Discussions

Pondrader wrote in post #17580131
Sure you can refocus ,...all you have to do is lift your thumb !!

LOL. I Use AI Servo a for any movement. You always capture the nice light.

I got to tell you I work really hard at it lol

Ok what about case settings for dogs that typically come at you in a hurry?? Still use case 5 with sensitivity down by one? Or case one with sensitivity down by one? This is my 165 pound Great Dane coming at me..lol it was case one no changes in settings. And second one is my Golden . But keeper rate is not high at all.

Re: long post on focus hit rates and the firmware update

Good post. Thanks.

The trip that delayed my sending my camera to canon is over. Didn't use the 7D2 as much as the T3i during the trip. Due to the results with the 7D2 I did the firmware update this evening. An impromptu test in the back yard in AI servo with center point expanded shooting only stationary objects in bursts of three or four shots all gave the same results - no chatter from the focus motor. One set of four was taken with the camera sitting on the deck and pointed at a wheel on the barbecue. Not a sound. Locked on and that was it. So it looks like the firmware did something. Perhaps the new firmware does tone down the aggressiveness of the focus system. Shots were only viewed on the lcd and not transferred to a computer.

Note: camera settings are after reset 'out of box' except for slow frame rate pushed to five and default case 1. Lens was 70-300L

The big thing now is my camera is about to be pushed to second place by my fly rod. Fishing season is here!!!

Rod

Thats a good question, I would shoot that with the centre group and case 5 or 6

Pondrader wrote in post #17579823
I shoot nothing but AI Servo, and slowing sensitivity is not new,..I posted the same months ago for all to see, I was asked over and over for all kinds of test's and files. I use case 5 and case 6 with the 100-400L II mostly and yup I shoot-um sitting in shade with Ai servo..... I worked this Kit up last night.,....If you ask me it works just fine??

Jeff,
You say you use case 5 and 6 adjusted for sensitivity at - 1. The difference between the two for single point of Accel decel at +1 for case 6. Can you elaborate when you switch between the two and in what way you. Otic the difference? I ask because many of us can't get a practical head around the accel decel setting. . Also, 90+% of your awesome posts here appear to be of stationary animals, have you had success with moving animals coming mostly toward or away from you faster than a walk? Which setting worked best?

gqllc007 wrote in post #17580313
Ok what about case settings for dogs that typically come at you in a hurry?? Still use case 5 with sensitivity down by one? Or case one with sensitivity down by one? This is my 165 pound Great Dane coming at me..lol it was case one no changes in settings. And second one is my Golden . But keeper rate is not high at all.

I have had about 66% keeper rate (2 of 3 shot burst) for either soccer or basketball player running at me. Now, like mentioned earlier, I too traditionally use a no-one mode, usually single point with a lower sensitivity for the times I slip off the player and don't want it to jump to the background.
I am actually trying to improve my keeper rate for this situation and am experimenting. Unfortunately I have had to tweak my mfa at the same, E time which needed priority for the soccer.

I will say that higher sensitivity will fix the 2of3 to higher keeper, but I believe there has to be another way without giving up the negative value sensitivity I just mentioned, . On my 7dc, I go 3of3 90%of my bursts and used negative sensitivity. So how do I emulate it on the 7dii?

Next I am going to try Accel at +2 instead of +1 in case 4, and sensitivity at - 1.

gschlact wrote in post #17579530

Michael132 wrote in post #17579199
Hi guys, long time reader-first time poster here
...

Now, I don't claim this to be solution to MY or anyone else's AF issue just jet as I haven't properly test it in the field and I remain skeptical until I do. For example; I can't explain how some of the pic didn't look like they miss the focus as much as they looked like there is no part of the image that is in tack sharp focus at all. That is another reason I have to wait for more in-the-field tests...to confirm if the camera can take tack sharp pic (50mm above f4 should be able to do at least as well as my XSi) Also
ow
...

...Specific yo the part I quote above.
Something moving in the system between and including the sensor and subject explains an image that is soft all over even though some point is sharper than others. Often causes by technique/shake, shutter speed, lack of IS settling (fighting mode 1). Other causes could include actively zooming or focusing the lens upon shutter release...

Perhaps I could have phrased it differently. What I meant is - given all the testing I went trough before and after 104 firmware I couldn't explain as to why some of the pics were soft-focused before and not after 104 IF I am to assume say my hypothesis of sloppy 50mm AF is to be right.

- technique/shake --> I shot with and w/o tripod (perfect leaving room conditions with natural sun light trough the window)
- shutter speed --> around 1/1250
- lack of IS settling --> nifty-fifty doesn't have IS
- actively zooming --> nifty-fifty doesn't have zoom
- sensor itself --> only happens in AI SERVO burst
- subject itself --> perfectly still, high contrast black/white and decently illuminated by natural light.
- active focus upon shutter release --> AI Servo 1st and 2nd priority is set to FOCUS

gschlact wrote in post #17579530

I have suspected from the beginning that just as Canon let's the image get taken when IS isn't settled, that they also were letting the shutter release prior to the lens completing the focus command and still moving. Both exhaling the lack of sharpness.

Your post somewhat supports this theory, and 1.0.4 might have had a firmware tweak to halt focus movement at the time of shutter release. Close to completing move, eat will be much sharper than taking while focus is in movement. This approach would also let it maintain the different behavior of - 2 to +2 of the Sensitivity setting. Faster focusing lens would. OT have been impacted as much in the original firmware as you discovered.

I could understand how Canon would be forced of making such compromises on bodies which don't have the AF AI Servo image priority for 1st and following image. But that option being available to 7D2 body should make camera prioritize focus over shutter release unless one chooses to select shutter priority or the middle ground in between those two...which I did not do in my test. I don't see why Canon would do this by-design rather then some sort of faulty hardware patch - damage control.

That being said, I think you might be on to something and if we assume that you are indeed right, then we would have to also assume that 7D2, for whatever reason, does not respect the AI servo focus priority in the way I have assumed it should. Maybe they have min focus time limit that makes it possible to get 10FPS and so L lenses AF can keep up while nifty-fifty type ones cant and, as I believe you are suggesting, they are merely controlling the damage by making the AF holt the focus search just before shutter release for slightly better result. But why not honor the focus priority and instead make the lens shoot at slower FPS? It would not be the only limiting factor for full 10 FPS. It could be since Canon already have lens/f-depended active AF points type/amount selection they didn't want to have a lens depended FPS camera on top of that?
I can only speculate but would rather not since I do not have adequate knowledge of canon AF system and how the feed-back loop works...if present at all.

Your hypothesis (if I understand it correctly) is good in a way that it is easily disprovable if not true: Do people with good L lenses have the same AF issue as I do with nifty-fifty and are they seeing the same improvement with 104

You gave me food for thought, thank you. Definitely hypothesis worth exploring further.

MakisM1 wrote in post #17579649
Great post and not only for the 7DII. I just replicated your findings in the 5DIII. AIServo with the EF 50 f1.8 missed focus 6 out of 9 times on a stationary target. I grant you not by much, just being soft. AI focus was sharp 8/8 and so was One Shot.

Using a more capable (in AF) lens, like the EF24-70 MkII, AIServo improved to 7/8 and the one 'missed' was barely OOF.

So the issue exists, in varying degrees (although normally I don't spend my day shooting mechanical pencil tips-my 'stationary target'...).

Interestingly enough, the EF 50 f1.8 on the 60D aced ALL modes! 9/9, 8/8, 9/9 for AIS, AIF, ONEShot...

Highly interesting findings, thank you for doing those tests. Granted, the nifty-fifty leaves lot to be desired but I had no problem with it on my XSi. Then again, my XSi isn't shooting 10FPS and has far less demanding AF.

sarcasm //
Oh the things I had to do with Xsi+kit lens to get a half decent bird shot. I think I start with ritual dance and by the time the fire is lit most other big lens bird photographer start franticly packing their gear and racing away <-- that helps with scoring the right spot I would otherwise never have + ones the near-by fire starts spreading it tends to solve my lightning situation do to limited ISO capability and not to mention beautiful golden bokeh backdrop. The few brave photographers still present by the time I emerge completely camouflaged like a one-horned swamp monster from grass in-between them and the birds (<-- which solves my limited 250mm reach ) usually ran away screaming into the twilight scaring the birds into flight and thus I have a perfect opportunity at scoring that BIF pic.
Now if people could only appreciate that while looking at my underexposed, high noise, out of focus pic of some dark winged flying ...errr...UFB? in the night sky I would not have to worry about saving for "more capable" lens..pffffff
I do however have to worry about some of those people scouting the grounds in hopes of shooting me while I shoot the birds...some don't even bother bringing the camera <--- you would be worried too if you lived in a "stand you ground" state
sarcasm//

Thanks again, wish I had better lens to test with.

CyberDyneSystems wrote in post #17579709
Well Michael, that was a great first post with a some good information! (and well worth the time invested in getting through your War and Peace style prose!

Welcome to posting at POTN. In time you may even post small replies limited to a mere two or three paragraphs at under 1000 words.

Seriously great contribution!

I tend to shoot in AI Servo nearly 100% of the time. But I use [AF ON] button and do not have the shutter button linked to AF. For stationary subjects I get focus lock and let go. I wonder if any of this practice has reduced the number of these oof images?

Much to consider with your findings, and thanks not only for doing all the testing to develop your theorem, but for also for taking the time to post them.

Thanks CyberDyneSystems, appreciate the welcome.

Yes, in time I shell indeed learn to express my self in 1000 words...I just don't think my pictures are that good jet

I am really new to the blazingly fast FPS capability of the 7D2 and as to my test shots; they were taken in full burst servo mode for testing purpose to further understand OOF scenario I experienced while shooting my little nephew sliding and swinging few days before. The stationary pics of him while in full servo burst were the ones that captured him for example while he would come to full stop at the bottom of the slide.
I usually never use servo for stationary subjects since I do lot of recomposing and it works great...I am new to 7D2 and do like the idea of BBF but have jet to test it out as well as many other neat and time-saving things one can customize...so looking forward to it.


And to think I almost scratched the post twice during the write up

gschlact wrote in post #17575355
Dan,
Regarding your comparison, were these shot at the same distance and therefor you cropped further with bare lens for the comparison's same FOV? If not, then I don't hint this represents the comparison we have been discussing.

Same spot, otherwise no need
to test!

Michael132 wrote in post #17580625
Perhaps I could have phrased it differently. What I meant is - given all the testing I went trough before and after 104 firmware I couldn't explain as to why some of the pics were soft-focused before and not after 104 IF I am to assume say my hypothesis of sloppy 50mm AF is to be right.

- technique/shake --> I shot with and w/o tripod (perfect leaving room conditions with natural sun light trough the window)
- shutter speed --> around 1/1250
- lack of IS settling --> nifty-fifty doesn't have IS
- actively zooming --> nifty-fifty doesn't have zoom
- sensor itself --> only happens in AI SERVO burst
- subject itself --> perfectly still, high contrast black/white and decently illuminated by natural light.
- active focus upon shutter release --> AI Servo 1st and 2nd priority is set to FOCUS

I have suspected from the beginning that just as Canon let's the image get taken when IS isn't settled, that they also were letting the shutter release prior to the lens completing the focus command and still moving. Both exhaling the lack of sharpness.

Your post somewhat supports this theory, and 1.0.4 might have had a firmware tweak to halt focus movement at the time of shutter release. Close to completing move, eat will be much sharper than taking while focus is in movement. This approach would also let it maintain the different behavior of - 2 to +2 of the Sensitivity setting. Faster focusing lens would. OT have been impacted as much in the original firmware as you discovered.



I could understand how Canon would be forced of making such compromises on bodies which don't have the AF AI Servo image priority for 1st and following image. But that option being available to 7D2 body should make camera prioritize focus over shutter release unless one chooses to select shutter priority or the middle ground in between those two...which I did not do in my test. I don't see why Canon would do this by-design rather then some sort of faulty hardware patch - damage control.

That being said, I think you might be on to something and if we assume that you are indeed right, then we would have to also assume that 7D2, for whatever reason, does not respect the AI servo focus priority in the way I have assumed it should. Maybe they have min focus time limit that makes it possible to get 10FPS and so L lenses AF can keep up while nifty-fifty type ones cant and, as I believe you are suggesting, they are merely controlling the damage by making the AF holt the focus search just before shutter release for slightly better result. But why not honor the focus priority and instead make the lens shoot at slower FPS? It would not be the only limiting factor for full 10 FPS. It could be since Canon already have lens/f-depended active AF points type/amount selection they didn't want to have a lens depended FPS camera on top of that?
I can only speculate but would rather not since I do not have adequate knowledge of canon AF system and how the feed-back loop works...if present at all.

Your hypothesis (if I understand it correctly) is good in a way that it is easily disprovable if not true: Do people with good L lenses have the same AF issue as I do with nifty-fifty and are they seeing the same improvement with 104

You gave me food for thought, thank you. Definitely hypothesis worth exploring further.

You mention the halting focus vs delaying shutter.
Is it possible that in the eos system, the lens doesn't issue a completion signal¿ assuming it does, I agree with you that canon should autimatically delay/slow the fps if shutter oriority is Focus. This begs the question and another test for you. Now that the nifty fifty is sharp, what are the fps when you request 10fps on moving object? Can you try a 3second burst using a stop watch and then you how many shots it took? Lastly, not all L glass is equal in AF speed.some had issues like the500.

Pondrader wrote in post #17580225
I got to tell you I work really hard at it lol

It shows.

digital paradise wrote in post #17579086
I wish things had gone smoothly for me as I really wanted to work with Canada. Glad it is working for you. Perhaps the basic CPS is just not good enough for them to take that extra step to work with you.

A few years ago I dropped my 300L F4 IS and the IS went all wonky. I double checked everything, fresh battery and it would not slap or do the normal grind sound and was shaky. The women who answers the calls said they could not reproduce the problem. At one point she told me maybe it fixed itself during shipment. I kid you not. I begged to have their techs take another look to avoid re-shipping. It was like pulling teeth but she agreed. I called a week later and asked if they had a chance to look at it and she told me they replaced the IS assembly for the tune of $450 and it was on it's way back. No estimate or anything. It still was not great so I shipped it to NJ not telling them about what TO did and they told me the IS had to be replaced. TO had to refund me because Canadian law states any repair over $100 requires a written estimate. If it had worked correctly I would have let it go.

These things happen but I went through a few more frustrating events so I gave up. Maybe I should try again at a different level. I might have to change my name.

Wow..what a nightmare...

gschlact wrote in post #17580976
You mention the halting focus vs delaying shutter.
Is it possible that in the eos system, the lens doesn't issue a completion signal¿ assuming it does, I agree with you that canon should autimatically delay/slow the fps if shutter oriority is Focus. This begs the question and another test for you. Now that the nifty fifty is sharp, what are the fps when you request 10fps on moving object? Can you try a 3second burst using a stop watch and then you how many shots it took? Lastly, not all L glass is equal in AF speed.some had issues like the500.

Indeed, that was the first thing I thought of but, as mentioned before, I do not have adequate knowledge of canon AF system and how it exactly works...nor am I aware of canon fully disclosing it. Indeed, if the AF has full feed-back loop then this should not be an issue but I suspect that it does not, which is why contrast AF is arguably more accurate then phase detection for stationary subjects.
We have to keep in mind however that the AI SERVO's main job is to keep focus on "moving" subjects. Using strict feed back loop to confirm and lock focus of high speed subject would be of limited usefulness there. It's reactive nature would often allow enough time to pass between initial AI focus seek and focus confirmation check that you would seldom get your focus lock confirmed, leaving you with endless loop where AI is chasing the focus and never allowing shutter release.
Thus, you have to be able to "predict" where the subject is going to be by the time shutter has opened and allow shutter action to happen without knowing if you are going to be "in focus". The longer into the future it has to predict the less accurate it will be.
Canon can get closer to it by using more powerful algorithms (making it smarter at predicting) and faster Lens AF drive etc (reducing the time camera spends executing AI focus command thus reducing the length of time into the future the focus has to be predicted for) etc. which requires more powerful processors and hardware off curse. Another way is to limit the possible predictions needing to be made by means of "case scenarios" for the AI AF to improve it's prediction success...and since no one knows better then the photographer, who is taking the pic, as to what type of movement it is to be expected, we were given the case scenarios and their micro adjustments as to help the AI AF improve our hit rate ever so more.

It could be that, excluding hardware fault and unintended behavior of the AF, the main culprit of OOF images in AI SERVO comes down to AF prediction algorithm which could be aggravated by specific lens ability to follow it. However I would rather pursue a more "unified" theory of sorts that could explain how some people needed hardware fix while others don't.

Also, while it is clear that canon has custom AF behavior modifier for individual lenses we are not sure to what extend are those implemented in what lens and camera combos. Do low end lenses/cameras get some generic algorithms with none or very few custom tweaks compered to high end? I feel we need someone who actually knows more about canon AF to chime in since I would rather just help by contributing hard data then soft speculation about areas I do not fully understand. My hope is that we can, trough deduction, narrow down the possible solution...say in case A person should spend more time tweaking AF and technique versus case B where a person should send in a camera do to high possibility if hardware problem. Anyhow...

Speaking of hard data, here are some as per your suggestion. Well not quite. I didn't have moving subject (I know, I know... but its been long day and dark by the time I did it. I might try that this weekend if its sunny) so I did a few 20 shot bursts aiming at a artificially illuminated stationary target with a digital stopwatch in frame. I handheld it with intention to repeat the test in case of OOF pic but it was not necessary

Test: 50mm Av@ f1.8; 1/1250; default Case 1; AI SERVO Full speed; RAW; 1st/2nd focus priority; handhold; ISO 3200
Result: 20 pictures full buffer burst all taken 0.1 second apart as indicated by digital stopwatch within each frame and all pic are adequately, and most importantly, equally sharp.

Please keep in mind that by "sharp" I do not mean "tack sharp" since that is subjective and I have jet to test that out.