When will canon fix the focus issues with the 70D?

[Photogirl55]

I was "T.H.I.S." close to buying the 70D.  I have read WAYYY too many posts about issues with the focusing on the 70D.  How is Canon handling the issue?  I would love to buy this camera, but not willing to gamble with that much money.

[Molybdo42]

@TTMartin : I'm not arguing about the need or not to do mfa, it's just that whatever the AF algorithm does, it can give some out of focus results.

I did further testing and it looks like my camera doesn't work as well as I though in sunlight.

I now think the AF performance will depend a lot on the general illumination of the scene. The picture I had trouble with was taken in a scene illuminated by sunlight. I had no focus issue when aiming at the brightest element in the scene (a car light by sunlight). But I had noticeable defocus when aiming at a gray wall in the shade (the car I previously aimed at was clearly out of the focus sensitive area of the central focus point, but it was still in the field of view).

Now the eyes have a very good dynamic range, so what appeared to be bright in the shade, can be in fact too dark to focus on for the sensor, or the AF sensor might be distracted by stray light.

Canon's focus algorithm must be doing something to prevent the bright spots to saturate the AF detector, so the signal of shaded areas must be quite low, too low for the AF algorithm to consider.

For the ones with AF issues, could you try pointing the camera at the brightest element in the scene (aim at distant objects) ? And avoid pointing the camera at shaded areas. And then compare

The point I'm trying to make is that there is a clear limitation on most prosumer DSLR, not well documented because most users never where in a similar situation (either they don't have fast lens, either they use full frame sensor). Micro focus adjust could be in part to blame , but what appears to be a mfa issue might in fact be a limitation of the AF system.

I think we are converging towards gears that keep getting more challenging for the AF :

- Current generations of lenses from Sigma and Canon are very sharp,

- the DSLR sensors keep getting more pixels,

- Also, take into account the gear : on a 24mm 1.4, for a full frame sensor the hyperfocal distance is only 14 m, on an APSC it's 22 m.

- Take into account the lens construction : on these wide angle lenses the focus ring turn between 10m and infinity is extremely small

Because of those facts any focus issue won't be visible on a full frame sensor (you won't see any difference if the camera focused at 15 or 20 meters), but that error will be visible on an APSC sensor.

Also are AF sensors build differently between full frame and APSC, are they bigger, do they get more light ?

[TTMartin]

---

@Molybdo42 wrote:

@TTMartin : I'm not arguing about the need or not to do mfa, it's just that whatever the AF algorithm does, it can give some out of focus results.

I did further testing and it looks like my camera doesn't work as well as I though in sunlight.

I now think the AF performance will depend a lot on the general illumination of the scene. The picture I had trouble with was taken in a scene illuminated by sunlight. I had no focus issue when aiming at the brightest element in the scene (a car light by sunlight). But I had noticeable defocus when aiming at a gray wall in the shade (the car I previously aimed at was clearly out of the focus sensitive area of the central focus point, but it was still in the field of view).

Now the eyes have a very good dynamic range, so what appeared to be bright in the shade, can be in fact too dark to focus on for the sensor, or the AF sensor might be distracted by stray light.

. . .

---

AF systems require adequate contrast. What you describe is covered and explained in these 3 videos.

A Look at The Canon Autofocus System Part 1

A Look at The Canon Autofocus System Part 2

A Look at The Canon Autofocus System Part 3

Understanding what the capabilites of an AF system are and are not is key to getting the optimum results from your camera.

[Molybdo42]

I agree, contrasted objects help achieve better focus. But what I was aiming at in the shade appeared (to the naked eye) well contrasted and detailled enougth, see below (24mm 1.4 sigma lens) :

Apparently it wasn't enougth for the AF sensor. There might be a link between the metering and the AF sensitivity threshold.

I'll watch again the videos in case I missed some information.

[TTMartin]

---

@Molybdo42 wrote:

. . .

I think we are converging towards gears that keep getting more challenging for the AF :

- Current generations of lenses from Sigma and Canon are very sharp,

- the DSLR sensors keep getting more pixels,

- Also, take into account the gear : on a 24mm 1.4, for a full frame sensor the hyperfocal distance is only 14 m, on an APSC it's 22 m.

- Take into account the lens construction : on these wide angle lenses the focus ring turn between 10m and infinity is extremely small

Because of those facts any focus issue won't be visible on a full frame sensor (you won't see any difference if the camera focused at 15 or 20 meters), but that error will be visible on an APSC sensor.

Also are AF sensors build differently between full frame and APSC, are they bigger, do they get more light ?

---

We are subjecting cameras to higher degrees of scrutiny. In the days of film cameras and early digital cameras the AF standard was that the focus plane fell within the depth of field for a consumer camera, and within the 1/3 center for pro cameras.

The thing to keep in mind was that that depth of field was based on either an 8X10 or 11X14 print at normal viewing distances. When we look at a 20 megapixel photo at 100% on a computer monitor it is equivalent to staring at a 3'X5' poster size print from 28 inches away. Use those parameters in a depth of field calculation and you find that it becomes a minuscule fraction of what cameras were designed for. This exaggerates even the smallest focus error. Take that same photo with that small focus error and make an 8X10 print and it would look fine. View the whole photo on an electronic display at normal viewing distances and it would look fine.

So it's not so much about focal lengths and sensor sizes as it is about pixel peeping on computer monitors. 

I can provide links to technical documentation for all of the above. So if there is a specific point you'd like to learn more about just ask.

[Molybdo42]

Of course that reducing the file size will help but I still want to be able to keep the full res picture to reframe in post if needed. I won't be trying to use autofocus with a wide angle lens anymore (at least on an APSC camera).

If you have any documentation detailing the elements taken into account by the camera during AF post them, I'm still trying to figure it out. But I would like informations about the 19 AF point sensor.

For example the lightning condition (stray light, shadows) seem to have an impact on my AF performance. Is there a way to configure it in the options ? Any information or tips will be welcome.

My dream camera would look like a full frame sensor (for lowlight : no noise at mini ISO 6400), as lightweight as possible, with touchscreen, swivel screen, 15-20 Mpx max (for noise issue), dual pixel AF and integrated flash (never use it but in a pinch...)... In short : a 70D with full frame sensor.

Last thing : I hope you are ready for the "5Ds focus issue" post

[TTMartin]

You don't feel it was covered in the videos?

They certain some of the most comprehensive information on Canon's AF, the AF system in the 70D is a direct descendant of the AF system in the 1D Mk III so even that part of the 3rd video applies to some extent.. 

[Molybdo42]

I already use back button AF ; know the difference between one shot, Ai servo, and AI focus ; know when to use micro focus adjust, etc...

For the moment I'm more interested in knowing what the 19 point AF sensor sees when aimed at different objects (the raw data). The focus sensitive areas extend beyond the marks in the viewfinder and sometimes (like in my picture above) the camera focuses where I don't want to. I want to know what caused that.

One thing I still don't understand is when I look at pictures of the AF sensor, compared to the focus point distribution it looks completely different. What do those sensors see ?

I would also like to know what other sensors Canon might use. For example can they use the data from the exposure metering when doing AF ? Is there a different focus algorithm when using different focal length.

How does the raw signal measured by the AF sensor changes between a wide angle lens and a telephoto lens ?

I'm doing this mainly to know the limitations of my gear. I missed some nice shots because I wasn't aware of these limitations. This knowledge will also assist me in getting the most suited gear for my shooting style.

The only references I found were these :

http://fr.wikipedia.org/wiki/Autofocus_%C3%A0_d%C3%A9tection_de_phase

http://www.clarkvision.com/articles/understanding.autofocus/

Higher end cameras from Canon offer more customization options but they are so heavy and lack some features that make the 70D attractive : touch screen, swivel screen, dual pixel AF, weight...

[TTMartin]

---

@Molybdo42 wrote:

I already use back button AF ; know the difference between one shot, Ai servo, and AI focus ; know when to use micro focus adjust, etc...

For the moment I'm more interested in knowing what the 19 point AF sensor sees when aimed at different objects (the raw data). The focus sensitive areas extend beyond the marks in the viewfinder and sometimes (like in my picture above) the camera focuses where I don't want to. I want to know what caused that.

One thing I still don't understand is when I look at pictures of the AF sensor, compared to the focus point distribution it looks completely different. What do those sensors see ?

I would also like to know what other sensors Canon might use. For example can they use the data from the exposure metering when doing AF ? Is there a different focus algorithm when using different focal length.

How does the raw signal measured by the AF sensor changes between a wide angle lens and a telephoto lens ?

I'm doing this mainly to know the limitations of my gear. I missed some nice shots because I wasn't aware of these limitations. This knowledge will also assist me in getting the most suited gear for my shooting style.

The only references I found were these :

http://fr.wikipedia.org/wiki/Autofocus_%C3%A0_d%C3%A9tection_de_phase

http://www.clarkvision.com/articles/understanding.autofocus/

Higher end cameras from Canon offer more customization options but they are so heavy and lack some features that make the 70D attractive : touch screen, swivel screen, dual pixel AF, weight...

---

FWIW, I think BBF is overrated. I reprogram the AF ON button, to be an AF OFF button and use that to stop focus when needed in AIServo.

This is from the classic 7D, but, since the AF sensors appear to be very similar it should answer your question. He also explains his methodology if you want to duplicate it with your 70D. By the way this was originated by someone who had a very similar complaint about the classic 7D's AF that you have with the 70D.

Cross type AF points in EOS 7D * all 19 points are shown by rolling over the list below the 4th photo

Again not a 70D reference, but, a 7D Mk II. On pages 32, 33, and 34 you can see a list of what lenses work with what AF points. Not all lenses can use all 65 of the 7D Mk IIs AF points. Like you surmised wider angle lenses like the EF-S 10-22mm f/3.5-4.5 USM and the EF-S 10-18mm f/4.5-5.6 IS STM are lenses that have a reduced number of usable AF points. While this doesn't directly address your issue with the 70D, it does show that your thinking is on the right track.

EOS 7D Mark II AF-Setting Guidebook * link to PDF is on the lower left portion of the page 

[TTMartin]

---

@Molybdo42 wrote:

I agree, contrasted objects help achieve better focus. But what I was aiming at in the shade appeared (to the naked eye) well contrasted and detailled enougth, see below (24mm 1.4 sigma lens) :

Apparently it wasn't enougth for the AF sensor. There might be a link between the metering and the AF sensitivity threshold.

I'll watch again the videos in case I missed some information.

---

In the left photo the camera very obviously AF'd on the plants in the foreground. This can be explained by the AF point size as shown in Andre's Blog above. 

In the right photo the camera appears to have focused exactly where I would have expected it to focus.

[jeffandellie]

I am left to wonder what your motives are for flooding the thread with all kinds of unrelated nonsense and redirection and subject changes.

I noted half a dozen lenses that I tested, all canon except for one sigma (art series), and your response is that 3rd party lenses don't work well with canon – no acknowledgment of the canon lenses.

Then you start going off about the old timers and their issues, and you talk about your problem with wide angle and shadows, as if this is related, and throw in a bunch of info on focus systems.

You are condescending towards the people raising issues, you poo-poo all the issues, and basically have come out and said it is user error, a condition caused by a bunch of clueless consumers using camera gear that is over their head - while effectively burying the real grievances a few pages back in the thread.

I can only assume there is a reason you are doing this. is it possible that any of you are compensated in some form, way, or manner to purposely muddy the waters, redirect the conversation and bury the complaints? Would be an easy assumption for some, but who knows – we never will know the truth.. But this thread was dormant for the most part until you two demonstrated the need to flood it with many, many posts, which as noted basically buried the issues.

you refused to acknowledge that I tested ALL of my lenses (center focus square in viewfinder plus same exact pic in live view exactly per canon’s directions), at several apertures (and focal length in some cases) each, in bright but not direct sun, with a tripod, changing focus in-between sots so that the camera had to refocus, and did all of this not at the widest angle available, but generally in the 40-50 mm range. (In fact, the only lens done at 24mm was the 24mm prime)

ALL of my lenses produced the exact same effect at every aperture - the center square was WAY off of focus. Canon told me to send it in, which I did.

Also worth noting is that canon fixed the issue, without touching any of my lenses, and now none of my lenses need micro adjusting - or at least are so close that I have not noticed any lenses being off (except for my own bad pictures from cameras shake, etc. - as we all get on occasion).

Maybe my problem isn’t exactly the same as some others, and of course I have no way to verify the thousands of users who have had issues – can you get that data from canon? I can’t. I couldn’t even get them to tell me what they did to fix my camera. “check firmware, clean sensor” is all they would tell me. And of course I see no comments about the issue regarding the two different versions of the same firmware 1.1.1 (as noted by magic lantern developers for the 70D)

Yes, it is true that there are generally complaints with every new model of camera that comes out, but to try and suggest that with all the issues, it has 99.99% been user error is just plain ignorant and insulting, and looks to be intentionally misdirecting.

I’ve spent somewhere between 7,500 and 10,000 already, and I consider this payment enough to have my right to raise complaints.