This topic is in the 2010 archive. You can read it and like its posts here. It is read only.
Archive 2010Focussing accuracy with fast lenses
p.1 #1Focussing accuracy with fast lenses
I spotted this little snippet from The British Journal of Photography regarding focussing errors with fast aperture lenses;
Back and front focus errors are common with very fast lenses (and others). The AF sensors in all DSLRs operate at effective apertures between f/2.8 and f/7.1, and 'see' the focus point as if the lens was stopped down to the exact virtual f-stop the sensor in use imposes. When spherical aberration causes a shift in focus on stop-down, both wider and smaller apertures will no longer focus on the targeted plane. This is corrected in camera calibration, general or lens-specific.
I would be very interested to hear FM members thoughts on this particularly considering so many people seem to have problems with fast aperture lenses (especially the 50mm f1.2L and Sigma 50mm f1.4).
Back and front focus errors are common with very fast lenses (and others). The AF sensors in all DSLRs operate at effective apertures between f/2.8 and f/7.1, and 'see' the focus point as if the lens was stopped down to the exact virtual f-stop the sensor in use imposes. When spherical aberration causes a shift in focus on stop-down, both wider and smaller apertures will no longer focus on the targeted plane. This is corrected in camera calibration, general or lens-specific.
I would be very interested to hear FM members thoughts on this particularly considering so many people seem to have problems with fast aperture lenses (especially the 50mm f1.2L and Sigma 50mm f1.4).
Jan 03, 2010 at 12:09 PM
p.1 #2Focussing accuracy with fast lenses
They are common, yes, but is this due to user error, (AF) sensor error or lens design? That is a better question. From my experience certain lenses perform better in terms of AF accuracy than others. My 85/1.8 is quicker and more accurate than my 50/1.4 at f/1.8 and my 35L is quicker and more accurate than my 50/1.4 at f/1.4. The Sigma lens(es) that I have been testing over the past few days seem to exhibit a serious design flaw, as the issues I'm experiencing appear to be relatively wide spread. The same can be said of the 50L with its known focus shift issue. DOF at these apertures is, at times, paper thin and AF sensors are large enough to cause even the slightest focus error. For example when shooting a portrait at f/1.4 with an 85mm lens within three or four feet with the focus point directly over the subjects eye a lot of times the actual focus point will grab the tip of an eyelash rather than the eyeball itself (essentially front focusing or appearing as so)...in some cases the eyeball may appear sharp but it may not be critically sharp. However the examples posted in my Sigma 50/1.4 thread clearly portray a design flaw, there is no other excuse for it.
Jan 03, 2010 at 12:42 PM
p.1 #3Focussing accuracy with fast lenses
M Vers wrote:
They are common, yes, but is this due to user error, (AF) sensor error or lens design? That is a better question. From my experience certain lenses perform better in terms of AF accuracy than others. My 85/1.8 is quicker and more accurate than my 50/1.4 at f/1.8 and my 35L is quicker and more accurate than my 50/1.4 at f/1.4. The Sigma lens(es) that I have been testing over the past few days seem to exhibit a serious design flaw, as the issues I'm experiencing appear to be relatively wide spread. The same can be said of the...Show more →
Thanks for your reply MVers (I don't know your name so forgive me for using your username). I've been looking at your Sigma thread with interest and it's bummed me a bit as I'm (still) seriously looking at that lens instead of the 50mm f1.2L as I personally can't justify the extra cost of the Canon relative to it's apparent performance. Decisions decisions!
They are common, yes, but is this due to user error, (AF) sensor error or lens design? That is a better question. From my experience certain lenses perform better in terms of AF accuracy than others. My 85/1.8 is quicker and more accurate than my 50/1.4 at f/1.8 and my 35L is quicker and more accurate than my 50/1.4 at f/1.4. The Sigma lens(es) that I have been testing over the past few days seem to exhibit a serious design flaw, as the issues I'm experiencing appear to be relatively wide spread. The same can be said of the...Show more →
Thanks for your reply MVers (I don't know your name so forgive me for using your username). I've been looking at your Sigma thread with interest and it's bummed me a bit as I'm (still) seriously looking at that lens instead of the 50mm f1.2L as I personally can't justify the extra cost of the Canon relative to it's apparent performance. Decisions decisions!
Jan 03, 2010 at 12:48 PM
p.1 #4Focussing accuracy with fast lenses
While this is getting a bit OT, based on what I've seen so far I think the Canon 50/1.4 is still a better performing lens than the Sigma variant. However, like many others have found, ymmv.
Jan 03, 2010 at 12:53 PM
p.1 #5Focussing accuracy with fast lenses
... The AF sensors in all DSLRs operate at effective apertures between f/2.8 and f/7.1, ...
Can anyone explain why it is so ? I'm not at all challengeing it but I just don't understand.
The AF sensor is seeing the image through the same lens wide open. Where does the "effective" aperture come ?
Can anyone explain why it is so ? I'm not at all challengeing it but I just don't understand.
The AF sensor is seeing the image through the same lens wide open. Where does the "effective" aperture come ?
Jan 03, 2010 at 12:59 PM
p.1 #6Focussing accuracy with fast lenses
Back OT, I have the lowly 28mm f1.8 and that isn't too bad at all on the focussing accuracy wide open, however, I put that down to a slightly deeper plane of focus because it's a WA, but I'm no expert.
Jan 03, 2010 at 01:02 PM
p.1 #7Focussing accuracy with fast lenses
Pondria wrote:
Can anyone explain why it is so ? I'm not at all challengeing it but I just don't understand.
The AF sensor is seeing the image through the same lens wide open. Where does the "effective" aperture come ?
Join the club, that's one of the reasons I posted this to get a bit more info on it.
Earlier quote
... The AF sensors in all DSLRs operate at effective apertures between f/2.8 and f/7.1, ...The AF sensor is seeing the image through the same lens wide open. Where does the "effective" aperture come ?
Join the club, that's one of the reasons I posted this to get a bit more info on it.
Jan 03, 2010 at 01:12 PM
p.1 #8Focussing accuracy with fast lenses
Interesting, could you post a link to the paper please?
Jan 03, 2010 at 01:50 PM
p.1 #9Focussing accuracy with fast lenses
Andrew Burnett wrote:
Interesting, could you post a link to the paper please?
Interesting indeed as it's from their review of the Sigma 50mm f1.4. Draw your own conclusions considering so many have issues with that lens.
http://www.bjp-online.com/public/showPage.html?page=837772
But I'd still like to know if anyone with knowledge of optical physics could tell me (us) a bit more?
Interesting, could you post a link to the paper please?
Interesting indeed as it's from their review of the Sigma 50mm f1.4. Draw your own conclusions considering so many have issues with that lens.
http://www.bjp-online.com/public/showPage.html?page=837772
But I'd still like to know if anyone with knowledge of optical physics could tell me (us) a bit more?
Jan 03, 2010 at 01:54 PM
p.1 #10Focussing accuracy with fast lenses
This article tries to explains virtual AF aperture. I don't understand it. It's near the end in the autofocus section.
http://doug.kerr.home.att.net/pumpkin/Split_Prism.pdf
http://doug.kerr.home.att.net/pumpkin/Split_Prism.pdf
Jan 03, 2010 at 02:13 PM
p.1 #11Focussing accuracy with fast lenses
There are two issues being discussed here. One is effective aperture AF sensitivity, and the other is focus shift. These are two different but related phenomena.
Focus shift in a lens with under-corrected spherical aberration exists independently of AF, and causes focus error even when manually focusing. It is an intrinsic property of the optical design. One way to avoid the problem is to manually focus while stopped down to the shooting aperture.
AF sensitivity is different--I don't know the details but I do know that when it is stated that an AF sensor is sensitive to f/2.8, or f/5.6, this means that additional light does not assist in that sensor's ability to acquire focus. For example, an f/1.2 lens does not AF more accurately than an f/2.8 lens if both are using an f/2.8 AF sensor. My understanding is that this has to do with the width of the baseline for the sensor (which is proportional to the f-number rating)--since opening up the diaphragm wider than the baseline doesn't confer any additional advantage, the sensor works the same at all lower-rated f-numbers, and fails to work at all for higher f-numbers (so an f/5.6 sensor works with the same precision at all apertures f/5.6 and wider, and fails to work at all for apertures smaller than f/5.6).
Therefore, when using AF on a very fast lens, the sensor is really only seeing at best f/2.8. It is theoretically possible to add another sensor with a longer baseline to allow AF at, say, f/1.4, but then only those lenses f/1.4 and faster would be able to use it, and in all other cases, it would never be used. Thus it would not be cost-effective.
This effective f-number for AF sensitivity then contributes to focus shift as well, but only because the AF sensor sees only f/2.8 light and is unable to assess focus from the marginal rays outside the f/2.8 effective aperture. So for lenses that exhibit focus shift, this can cause misfocus at all apertures wider than f/2.8, including the maximum aperture, whereas in the MF case, focus shift should not be present at the native aperture (wide open) because you are focusing at the same aperture as you are shooting. Consequently, the limited AF sensitivity may be mistaken for miscalibrated AF for fast lenses, and the user may use MA to compensate.
Note that if a lens is perfectly corrected for spherical aberration, focus shift should be entirely absent, and AF misfocus should not occur due to focus shift.
Focus shift in a lens with under-corrected spherical aberration exists independently of AF, and causes focus error even when manually focusing. It is an intrinsic property of the optical design. One way to avoid the problem is to manually focus while stopped down to the shooting aperture.
AF sensitivity is different--I don't know the details but I do know that when it is stated that an AF sensor is sensitive to f/2.8, or f/5.6, this means that additional light does not assist in that sensor's ability to acquire focus. For example, an f/1.2 lens does not AF more accurately than an f/2.8 lens if both are using an f/2.8 AF sensor. My understanding is that this has to do with the width of the baseline for the sensor (which is proportional to the f-number rating)--since opening up the diaphragm wider than the baseline doesn't confer any additional advantage, the sensor works the same at all lower-rated f-numbers, and fails to work at all for higher f-numbers (so an f/5.6 sensor works with the same precision at all apertures f/5.6 and wider, and fails to work at all for apertures smaller than f/5.6).
Therefore, when using AF on a very fast lens, the sensor is really only seeing at best f/2.8. It is theoretically possible to add another sensor with a longer baseline to allow AF at, say, f/1.4, but then only those lenses f/1.4 and faster would be able to use it, and in all other cases, it would never be used. Thus it would not be cost-effective.
This effective f-number for AF sensitivity then contributes to focus shift as well, but only because the AF sensor sees only f/2.8 light and is unable to assess focus from the marginal rays outside the f/2.8 effective aperture. So for lenses that exhibit focus shift, this can cause misfocus at all apertures wider than f/2.8, including the maximum aperture, whereas in the MF case, focus shift should not be present at the native aperture (wide open) because you are focusing at the same aperture as you are shooting. Consequently, the limited AF sensitivity may be mistaken for miscalibrated AF for fast lenses, and the user may use MA to compensate.
Note that if a lens is perfectly corrected for spherical aberration, focus shift should be entirely absent, and AF misfocus should not occur due to focus shift.
Jan 03, 2010 at 02:16 PM
p.1 #12Focussing accuracy with fast lenses
Thank you! I can't speak for anyone else, but that made sense to me! 
Jan 03, 2010 at 02:20 PM
p.1 #13Focussing accuracy with fast lenses
The broader baseline of sensors that only work with wide aperture lenses does allow for higher accuracy, but I don't think that necessarily means that the accuracy of an f/2.8 sensor (i.e. one that requires an f/2.8 or faster lens) cannot be sufficient for the shallow DOF of an f/1.4 lens shot wide open. There's more involved than just the baseline (just as with rangefinder focus accuracy).
Jan 03, 2010 at 03:58 PM
p.1 #14Focussing accuracy with fast lenses
Photon wrote:
The broader baseline of sensors that only work with wide aperture lenses does allow for higher accuracy, but I don't think that necessarily means that the accuracy of an f/2.8 sensor (i.e. one that requires an f/2.8 or faster lens) cannot be sufficient for the shallow DOF of an f/1.4 lens shot wide open. There's more involved than just the baseline (just as with rangefinder focus accuracy).
From what I understand that is correct. The baseline width of an AF sensor array only determines the relative accuracy, may other factors of the AF system design such as the AF sensor pixel density come into play in defining absolute accuracy. As a sensor with a (wider) f/2.8 baseline will be able to register smaller phase differences than a sensor with a (narrower) f/5.6 baseline, the f/2.8 array is inherently more accurate. But that does not say anything about the absolute accuracy of the whole AF *system* where many additional factors such as lens positioning tolerances etc come into play.
The baseline choice of f/2.8 or f/5.6 or f/8 etc is just a practical choice to enable the AF sensor array to work with the largest possible range of lenses. While a baseline wider than f/2.8 would increase accuracy even further for large aperture primes, it would not work reliably with any of the EOS zooms and of course the slower EOS primes. Canon specifies that an f/2.8 baseline is sufficient to drive positioning of the focus plane within 1/3rd Depth of Focus (not quite the same but similar to Depth of Field) of the lens wide open for any of their lenses - including their f/1.2 primes.
The broader baseline of sensors that only work with wide aperture lenses does allow for higher accuracy, but I don't think that necessarily means that the accuracy of an f/2.8 sensor (i.e. one that requires an f/2.8 or faster lens) cannot be sufficient for the shallow DOF of an f/1.4 lens shot wide open. There's more involved than just the baseline (just as with rangefinder focus accuracy).
From what I understand that is correct. The baseline width of an AF sensor array only determines the relative accuracy, may other factors of the AF system design such as the AF sensor pixel density come into play in defining absolute accuracy. As a sensor with a (wider) f/2.8 baseline will be able to register smaller phase differences than a sensor with a (narrower) f/5.6 baseline, the f/2.8 array is inherently more accurate. But that does not say anything about the absolute accuracy of the whole AF *system* where many additional factors such as lens positioning tolerances etc come into play.
The baseline choice of f/2.8 or f/5.6 or f/8 etc is just a practical choice to enable the AF sensor array to work with the largest possible range of lenses. While a baseline wider than f/2.8 would increase accuracy even further for large aperture primes, it would not work reliably with any of the EOS zooms and of course the slower EOS primes. Canon specifies that an f/2.8 baseline is sufficient to drive positioning of the focus plane within 1/3rd Depth of Focus (not quite the same but similar to Depth of Field) of the lens wide open for any of their lenses - including their f/1.2 primes.
Jan 12, 2010 at 12:47 PM
p.1 #15Focussing accuracy with fast lenses
Thanks for all your input everyone, it's much clearer to me now. 
Jan 12, 2010 at 02:18 PM
p.1 #16Focussing accuracy with fast lenses
The AF system looks at light paths coming in from two directions simultaneously and compares them to determine focus status. Both of those light paths first have to get through the partially transmissive section in the middle of the main mirror, before bouncing off the sub mirror that is behind it and then on down to the AF lens and mirror assembly. As it is only looking through a "window" in the main mirror then perhaps it cannot fully utilize the wider lens openings.
- Alan
- Alan
Jan 12, 2010 at 04:58 PM
p.1 #17Focussing accuracy with fast lenses
wickerprints:
very good write-up
very good write-up
Jan 12, 2010 at 05:15 PM