I also can demonstrate the front focusing issue at MFD.
I also believe this is due to a limitation of dual pixel autofocus when objects of focus are too large in the frame.
AF_1.jpg and AF_2.jpg are focused on the smallest box in the AF chart near MFD.
MF_1.jpg clearly has the most accurate focus and that was done in MF at 10x magnification.
AF_1_Pink.jpg and AF_2_Pink.jpg were just to see if changing the target would result in a perfect critical focus, but it didn't work.
AF_1_Dog.jpg and AF_2_Dog.jpg were taken of my dog, who so kindly helped out. Both of these are in perfect critical focus on the eye (where the spot focus box was placed).
AF_2_Dog reports exif distances -
Focus Distance Upper : 0.76 m
Focus Distance Lower : 0.71 m
Which is around the MFD of 0.70m. Notice that there are very small specular highlights in the eye, I beleive these were probably what was focused on, and thus the camera was able to be precise and focus in the correct spot.
I don't know whether this lens has a front focusing issue, but my hypothesis is still that the EOS R is not capable of resolving perfect focus at high magnifications when the objects are large in the frame and there is nothing very small to focus on.
Also, sorry for choosing a lackluster naming scheme as it has caused some less than perfect ordering in the folder.
stanj wrote:
Got it. It's still about the wording - to say that "it doesn't have it" is incorrect, misleading, and can lead to confusion later on.
It's absolutely accurate. AFMA is ONLY available when using the LA-EA2 or 4, which have MIRRORs inside them. Basically downgrading Sony mirrorless cameras into DSLRs (SLTs for Sony). So you have to have the feature that will reduce the impact of the problem mirrors have always created. If you use the LA-EA1, 3, or EVERY other adapter made for Sony mirrorless (Metabones, MC-11, Techart Pro, cheap Chinese adapters, etc.), along with every single native E-mount lens, AFMA will be grayed out an inaccessible, as it has no purpose when using on-sensor AF (which cannot go out of alignment). Think of it as legacy support for owners of older Sony A-mount cameras and lenses.
Thanks everyone for taking the time to look at this issue, and I greatly appreciate it. I also have to give a quick thanks to Jim! He is a pleasure to work with, and has gone above and beyond to help me sort out this issue.
Jesse Evans wrote:
Jim, could you please take some using the EF 70-200 with EF-RF adapter and post those here as well?
I suspect the problem is actually that the EOS R is not focusing properly on the squares of the focusing chart.
Also, how far are you from the focusing chart? Is this near MFD?
I did and I mentioned that in my post. The EF 70-200 did not exhibit the same level of front focusing, if any. I also tried a couple RF 70-200mm lenses on the RP to make sure it wasn't just the R. The front focusing showed up on both the R and RP.
I won't be able to access the test files until Sunday, but if I can, I will post the results of the EF version.
robert_in_ca wrote:
There’s no micro adjusting on mirrorless cameras.
Because micro-adjustment is the calibration of three different focusing points (Imaging sensor, groundglass, and focusing sensors) when in an mirrorless camera there is only one sensor.
This must be a problem with communication between the camera and the lens. I suspect the lens is lying to the camera when it reports it has followed the focusing command. Although I've been disputed about this on DPReview, my earlier information is that after sending the focus command to the lens, the camera does not do a followup check to make sure the command was effectively completed.
JimPatterson wrote:
I did and I mentioned that in my post. The EF 70-200 did not exhibit the same level of front focusing, if any. I also tried a couple RF 70-200mm lenses on the RP to make sure it wasn't just the R. The front focusing showed up on both the R and RP.
I won't be able to access the test files until Sunday, but if I can, I will post the results of the EF version.
Thanks, I know you mentioned it in your post but was hoping you could take some for pasting. I appreciate your participation in this!
RDKirk wrote:
my earlier information is that after sending the focus command to the lens, the camera does not do a followup check to make sure the command was effectively completed.
Isn't this what Servo mode would solve / bypass? It doesn't take the camera's word for it, it keeps driving the lens. Again I don't know, I can't explain how this would happen in the first place.
RDKirk wrote:
my earlier information is that after sending the focus command to the lens, the camera does not do a followup check to make sure the command was effectively completed.
stanj wrote:
Isn't this what Servo mode would solve / bypass? It doesn't take the camera's word for it, it keeps driving the lens. Again I don't know, I can't explain how this would happen in the first place.
This doesn't seem to depend on server vs. single-shot AF. I used servo mode in my tests and saw the same issue as others here who used single-shot.
I’m really at disbelief as to how Canon has allowed this issue to get by given the lens only works on 2 camera bodies. I’m not here to bash on Canon, and for the most part, I have had little issues with their lenses or bodies. With that said, I noticed the issue instantly even by reviewing the photos on the BOC. I can’t quite understand how they allowed this one to slip by. Once I receive the replacement lens I will be sending it in to Canon for review. Hopefully Canon will issue a fix if we continue to shine light on this issue. I would hope this is an easy fix with firmware since the lens can focus manually. I left a review for the lens on their website outlining my concerns. Not sure if they will post it.
This sounds like an encoding or programming error of the exit pupil distance supplied by the lens EEPROM @ 200mm to the camera over the RF mount. On-sensor PDAF mechanisms need the precise exit-pupil distance to correctly calculate the phase differentials seen at the on-sensor phase pixel pairs.
Same thing happened this year with their flagship supertele, EF 600mm f4 IS III.
It had a IS issue that translated into some vertical blur, basically destroying every image at shutter speeds between 1/50s and 1/250s. Sharp images at 1/10s but unusable at 1/200s. It took 7 months of long discussions between 600mm users and Canon, hundreds of images transferred to the service center until they finally admitted that problem and issued a third firmware for that lens. That 600mm III is a 13000$ lens.
jkochuni wrote:
There is no micro adjust for mirrorless. Thanks for running that test, and your picture looks like it hit focus. Could you do the same test at mfd? I have a new lens coming Friday, so I will report back if I’m having the same issue. The pictures I took were on a flat surface and a dimensional object.
I've come to the conclusion that Canon are either total idiots or really clever by making the cheapie R and Ra without critical features so that people will have to upgrade to better bodies later. Nikon has AF MFT obo Z7 and even the cheap 90D has the AF micro adjust.
EB-1 wrote:
I've come to the conclusion that Canon are either total idiots or really clever by making the cheapie R and Ra without critical features so that people will have to upgrade to better bodies later. Nikon has AF MFT obo Z7 and even the cheap 90D has the AF micro adjust.
EBH
Unless something goes terribly wrong, mirrorless is immune from AF misalignment. The entire fine tuing thing came about because of mirrored cameras, which have the AF sensor separate from the imaging sensor, and use mirrors to "mirror" the view from the lens to the AF sensor (where the imaging sensor gets a direct shot from the lens). So software came about that allowed slight misalignments of the sensor, AF, and lens to be somewhat compensated for. It's still not perfect, nor will it ever be, due to the complexity of the system and growing complexity of the AF system (try tuning all 153 points of the D5 at close, medium, far focus individually for wide, medium, telephoto, on EVERY SINGLE LENS you own).
Move to mirrorless, and the AF and imaging sensor are the same exact part, on the same plane, and cannot be misaligned unless the sensor gets bent. Every lens is 100% plug and play for 100% accurate focus, unless a FW issue arises. That Nikon included a fine tuning option is either for their legacy users to make them feel comfortable, or it means they're not intending to do much in the way of FW updates for lenses. Either way, it's a swing and a miss. Sony already had an issue with focus misalignment on their FE 24-105/4, and it was 100% solved by lens FW, because misalignment isn't caused by the body in mirrorless. Canon's problem should also be solved in FW, and probably affects a certain range of lenses that weren't QC'd properly.
BlueBomberTurbo wrote:
Unless something goes terribly wrong, mirrorless is immune from AF misalignment. The entire fine tuing thing came about because of mirrored cameras, which have the AF sensor separate from the imaging sensor, and use mirrors to "mirror" the view from the lens to the AF sensor (where the imaging sensor gets a direct shot from the lens). So software came about that allowed slight misalignments of the sensor, AF, and lens to be somewhat compensated for. It's still not perfect, nor will it ever be, due to the complexity of the system and growing complexity of the AF system (try tuning all 153 points of the D5 at close, medium, far focus individually for wide, medium, telephoto, on EVERY SINGLE LENS you own).
Move to mirrorless, and the AF and imaging sensor are the same exact part, on the same plane, and cannot be misaligned unless the sensor gets bent. Every lens is 100% plug and play for 100% accurate focus, unless a FW issue arises. That Nikon included a fine tuning option is either for their legacy users to make them feel comfortable, or it means they're not intending to do much in the way of FW updates for lenses. Either way, it's a swing and a miss. Sony already had an issue with focus misalignment on their FE 24-105/4, and it was 100% solved by lens FW, because misalignment isn't caused by the body in mirrorless. Canon's problem should also be solved in FW, and probably affects a certain range of lenses that weren't QC'd properly. ...Show more →
This is mostly correct except for the cases of bent lens mounts and decentered/tilted lenses, which can cause AF issues on MILCs the same as they do for DSLRs and for the same reason - they cause incorrect phase differentials that can lead to the AF system misjudging what is "in focus".
snapsy wrote:
This sounds like an encoding or programming error of the exit pupil distance supplied by the lens EEPROM @ 200mm to the camera over the RF mount. On-sensor PDAF mechanisms need the precise exit-pupil distance to correctly calculate the phase differentials seen at the on-sensor phase pixel pairs.
That's interesting. I wonder if the lenses delivered to photographers have slightly different exit pupil distances than those used by Canon in the EEPROM?
It might also explain what I experienced with the 90D in my post on the previous page.
If that's the case, and if it's difficult for Canon to fix, maybe there will still be a need for AFMA after all?
snapsy wrote:
This sounds like an encoding or programming error of the exit pupil distance supplied by the lens EEPROM @ 200mm to the camera over the RF mount. On-sensor PDAF mechanisms need the precise exit-pupil distance to correctly calculate the phase differentials seen at the on-sensor phase pixel pairs.
I thought this would only be necessary to calculate the amount of movement needed to bring the phase differential to 0. I can tell you that the camera continues to believe that it is in sharp focus. If you press the shutter button again, or have it in AF Servo, it still stays out of focus.
I am curious if anybody here has a 180mm macro and an EOS R. I’m really wondering whether this is a telephoto close focusing limitation of DPAF, but I do not have access to one of these lenses. I would like to eliminate this as a possibility.
snapsy wrote:
This is mostly correct except for the cases of bent lens mounts and decentered/tilted lenses, which can cause AF issues on MILCs the same as they do for DSLRs and for the same reason - they cause incorrect phase differentials that can lead to the AF system misjudging what is "in focus".
Hmmm, I had a lens that I dropped twice onto concrete, bent the mount, decenterd an element or two (by a lot), and dislodged an aperture blade. Despite all that, the lens still focused accurately, though the bokeh was terrible.
rscheffler wrote:
That's interesting. I wonder if the lenses delivered to photographers have slightly different exit pupil distances than those used by Canon in the EEPROM?
It might also explain what I experienced with the 90D in my post on the previous page.
If that's the case, and if it's difficult for Canon to fix, maybe there will still be a need for AFMA after all?
AFMA could correct it but the better solution would be to fix the root issue. Btw, there is also a remote possibility that Canon's' dual-pixel AF structure just can't accommodate the geometry/exit-pupil distance of the lens correctly. Unlike traditional PDAF sensors on DSLRs that can handle pretty much any exit-pupil distance, on-sensor PDAF have shallower geometries, which make them less suitable at longer focal lengths. Sony tries to work around this on their sensors by having a mix of PDAF pixel geometries on their sensors, each of which is designed to accommodate a certain range of exit-pupil distances. It's pretty interesting actually and is described in detail on their patent filings.
The EF 70-200 did not exhibit the same level of front focusing, if any. I also tried a couple RF 70-200mm lenses on the RP to make sure it wasn't just the R. The front focusing showed up on both the R and RP.