How can the problem exist with manual focus? Is it being suggested that if one used an accurate means of manual focus (Live View or whatever) with a subject distance of 4 feet, at any aperature, that focus would be inaccurate?
Supposedly the aperture shutting down actually shifts the lens' projected focus by a small distance, enough to take the edge off. I don't know enough about optics to understand how aperture blades could do this. This means that if Live View or MF focusses correctly at f1.2 focus will go off as the lens stops down to 2.8. No problems wide open though. Can someone confirm this?
brainiac wrote:
Supposedly the aperture shutting down actually shifts the lens' projected focus by a small distance, enough to take the edge off. I don't know enough about optics to understand how aperture blades could do this. This means that if Live View or MF focusses correctly at f1.2 focus will go off as the lens stops down to 2.8. No problems wide open though. Can someone confirm this?
Actually yes, I did this once just to confirm that I was correct. And with liveview and 10x zoom you don't even need to take a picture to see it. Just press the DOF button and you can see the focus shift. Although if you stop down at some point the increased DOF starts to hide the shift. By 2.8 the DOF is increasing enough that your focus point will likely still be focused, but at f2 you can clearly see the focus is wrong.
focus shift has been around forever. it was pretty common in older lenses with 100% spherical elements. The problem is spherical aberration - near the center of the lens elements, the curvature is relatively low and the aberration between ideal and actual is pretty small. When you stop down, the lens only sees through the well-corrected center so you have "correct" focus. Near the edges (included in the wide open image), spherical aberration piles up and can cause a shift in the plane of best focus between the ideal (similar to where it would be at say f8) and actual. Since spherical aberration varies by color, there's usually also some smearing and loss of sharpness with non-monochrome subjects.
All of this can be fixed with aspherical and APO elements. Problem is, most often those corrections bring along nasty bokeh. For world class bokeh, it's a balancing act between leaving aberrations in the wide-open lens (for nice portrait bokeh, since a 50 1.2 is going to be used as a portrait lens on 1.6) and being fully-corrected stopped down. Apparently the possibility of focus shift either eluded Canon, or was deemed to be acceptable.
For more examples of focus shift, read up on the Leica 35mm Summilux-M ASPH and the 50mm 1.0 Noctilux. The newish Zeiss ZM 50mm 1.5 Sonnar is purposely designed with loads of spherical aberration, for a 1930s glamour portrait look. The price is loads of focus shift (Zeiss will calibrate the lens for use wide open or stopped down, your choice - but you can't have both). A google seach will reveal a lot of large format guys talking about it (LF guys commonly use 80-year-old lenses, which are very simple spherical symmetric lenses prone to focus shift). The Leica M8 has been exposing focus shift in classic lenses with a vengeance. One theory is that the thickness and waviness of film emulsions has been covering up this kind of focus shift since the dawn of time, and only the ruthlessly flat digital sensors reveal it.
For example, Zeiss had to apologize/clarify to customers over the focus shift in the ZM Sonnar-C:
C-Sonnar T* 1,5/50 ZM
Information about special features for dealers and users
The C-SONNAR T* 1.5/50 ZM is a very special lens; based on a classical lens design concept from the 1930´s. The additional letter “C” in the name of the lens expresses this designation.
This lens design helps to achieve pictures with a special artistic touch. This lens ‘draws’ your subject in a fine, flattering manner and is therefore ideally suited for portraiture. It renders a sharpness that is slightly rounded, being less aggressive than in contemporary lens designs, but at the same time not soft in its rendition.
Many famous portraits of glamorous and prominent people during the 1930´s used this technique to great effect. These images are characterized by portraying the person in a shining, nearly celestial way. This effect is very well balanced and not exaggerated; therefore many viewers see it in a subconscious way. The trained observer, however, understands the underlining technique and enjoys the results.
This lens design exhibits some additional effects, which should be understood to achieve the maximum benefit from the C-Sonnar T* 1.5/50 ZM:
Because of the above mentioned classical characteristic of the lens the best focus position in the object space can not be kept exactly constant for all f-stop settings.
The passionate photographer might notice a slightly closer best focus in his pictures than expected. When stopping down the lens to f/2.8 or smaller this effect is minimized, so the focus position will be as expected.
In order to balance the performance at full speed and other f-stop settings the lens is adjusted with above described characteristic.
The special features of the C-SONNAR T* 1.5/50 ZM are best used in emotional, artistic, narrative images, portraits or atmospheric landscapes. For documentation or technical subjects CARL ZEISS recommends to stop down the lens at least to f/5.6 or to use the PLANAR T* 2/50 ZM lens.
(I find it endertaining how they avoid using the word 'glow' - a Leica trademark - choosing rounded, shining, celestial instead)
fourfa wrote:
focus shift has been around forever. it was pretty common in older lenses with 100% spherical elements. The problem is spherical aberration - near the center of the lens elements, the curvature is relatively low and the aberration between ideal and actual is pretty small. When you stop down, the lens only sees through the well-corrected center so you have "correct" focus. Near the edges (included in the wide open image), spherical aberration piles up and can cause a shift in the plane of best focus between the ideal (similar to where it would be at say f8) and actual. Since spherical aberration varies by color, there's usually also some smearing and loss of sharpness with non-monochrome subjects.
All of this can be fixed with aspherical and APO elements. Problem is, most often those corrections bring along nasty bokeh. For world class bokeh, it's a balancing act between leaving aberrations in the wide-open lens (for nice portrait bokeh, since a 50 1.2 is going to be used as a portrait lens on 1.6) and being fully-corrected stopped down. Apparently the possibility of focus shift either eluded Canon, or was deemed to be acceptable.
For more examples of focus shift, read up on the Leica 35mm Summilux-M ASPH and the 50mm 1.0 Noctilux. The newish Zeiss ZM 50mm 1.5 Sonnar is purposely designed with loads of spherical aberration, for a 1930s glamour portrait look. The price is loads of focus shift (Zeiss will calibrate the lens for use wide open or stopped down, your choice - but you can't have both). A google seach will reveal a lot of large format guys talking about it (LF guys commonly use 80-year-old lenses, which are very simple spherical symmetric lenses prone to focus shift). The Leica M8 has been exposing focus shift in classic lenses with a vengeance. One theory is that the thickness and waviness of film emulsions has been covering up this kind of focus shift since the dawn of time, and only the ruthlessly flat digital sensors reveal it. ...Show more →
menatwork wrote:
The focus problem is the same whether I use AF or MF on my 5D.
I was wondering if this issue is camera specific. May be it's only on the 5D and 1D Mark III. I have read some happy results from the 40D camp. Does any one know for sure?
I found the lens backfocuses on my 5D but not at all on my XT. I wondered if it was a problem with focusing on the f/2.8 focusing points that the 5D has and the XT does not have. When I switched to the non-f/2.8 focusing points on the 5D the lens did not backfocus....
I had nearly the same problem with my 24-70L. Drove me nutz. Sold it and got the 24-105... I do not own the 50F1.2 but a partner of mine does...He loves it.. However, he could be naive to the problem. I'll ask him about it.
I would've thought that the amount of back focus is calculatable (indeed, some people who have experienced this problem with the 50Lf1.2 claim they can compensate for it to their satisfaction)... and if the AF system can determine roughly the distance to subject (through the position of the focus ring), then surely this could be accurately negated in a firmware upgrade? Obviously a firmware upgrade in the lens would be ideal as it would improve the issue for both AF and MF shooters... I thought these things work on some sort of "look up table" where the values within could be merely recalculated to compensate for the effect?
The explanation above for spherical aberation is excellent - well researched *and* explained - thankyou. But I'm still not happy! ;-) Even if we can't change the physics of light, some rules can be bent by using the supporting electronics. This is no longer the 1930's - we have the technology these days to have our cake and eat it too.
Thnx fourfa - useful to know. I still find it surprising that the effect can be so accentuated. You would think that you would notice such strong spherical aberration in the image quality at f1.2.
I was disappointed with the low contrast of the F1.4 at big apertures and bought the 50L and sold the F1.4. The 50L has a fine IQ at F1.2 to F2.8 at which point you could trade it for a $85 F1.8 and still have comparable quality. The upshot is that all these UltraSpeed lenses are very specialized with quirks and limitations in order to achieve their fast speed.
My 50L remains in inventory for specialized conditions (extremely low light, fixed camera position with multiple shots to bracket focus-not exposure) but I would have thought a $1600, standard focal length 50mm prime could have been better realized.
Frankly I like the 85L better (but also keep the 85 F1.8 for fast focus and utility). My most trouble free and hi IQ UltraSpeed is the 135L. So, rather than buying the 50L first, I consider it as a "last to buy" seldom needed/used lens and for that reason enormously expensive.
