That's a test of the Canon 50/1,2 focus shift (backfocus) and what to do. And it's pretty accurate IMO of that lens.
Most lenses with focus shift that I have owned only show it at very close distance also. So most people including myself will not really notice it in normal shooting. Manual focus instead of AF will often help also.
p.5 #2 · Risk of focus shift issues in new 1.4/35?
Makten wrote:
Interesting that your opinion is the right one. Focus shift is an issue to me and to many other photographers, regardless of what you think about it. It's an issue even for people knowing exactly how it works too, because you can't use live view in all situations. My own camera is useless with LV because it doesn't tell what aperture is used during focusing.
And please, if you suggest that someone is wrong, tell us who you are talking about and what is wrong.
+1
As for the whole discussion back and forth, it has been most useful to me at least. I think I've gained a bit of understanding of the effect of the focusing screen . I've never had the issue in practice (focus shift has always equaled back focus at a couple of steps above wide open), but it's good to know what to expect when one uses a camera with a focusing screen that provides an equivalent of f/2.8 or so.
The AF discussion was also interesting, although I consider the question still open.
Focus shift can indeed be a significant thing especially at close distances. For instance you may be shooting a portrait, focus on the eye and get instead the nose in focus. For planar subjects it can be even worse where you end up with the whole image more or less out of focus.
p.5 #3 · Risk of focus shift issues in new 1.4/35?
First Wickerprint, then AhamB have already mentioned the correct answer to the "relative focusing screen DoF" question.
The focusing screen can be seen as a "choke", not in the true aperture sense, but as in a system relay sense. The matte screen's dispersion angle (the amount of "matte") determines the system throughput of the three-part (simplified) combined VF system.
Part one is the lens' aperture. Part two is the matte screen scatter angle profile. Part three is the combined viewfinder>eye aperture stop.
The screen determines how much light from the lens that gets through to the VF aperture in quite an elaborate way. The VF brightness (and the effective F-stop of the VF + lens combination) is found by getting the integral sum of ray solid angles:
( Lens ray cone [modified by] matte screen scatter function )
>[that is member of]
( viewfinder/eye combination ray cone angles )
-This is the amount of light that "gets through" from the lens to the eye in the viewfinder. If we remove the mirrors from the drawing to simplify the layout it looks like this - reality is the same, but with the mirror to redirect the light cone from the lens up into the matte screen, then we also have the pentaprism between the matte screen and the VF optics:
If you use a bi-brite screen like in the sample above, smaller angles (F5.6 and smaller apertures) have an almost 100% transfer. But as you can see, a large-aperture lens will get all the rays coming in from steeper angles (like the edge of an F1.4 aperture) cut away from the "path of possible transmission".
You can get around SOME amount of this problem by increasing the scatter angle spread (using a more matte screen) - but in this case the F2.8 - F8 lenses get dimmer in the viewfinder since the cut-off angle is reached earlier. But a lot more of the light from the F1.4 aperture rim now gets through to the eye.
But the effect is rather small, not at all as large as it would seem from my quick sketches above. You have to mind that the drawings are two-dimensional, the loss is almost comparable to what you see in the sketch squared.
Even though the -S screen is noticably better (I've modified and used them myself) it only adds about half an Ev of light from F3.2 down to F1.4. That's an 80% light-loss.
What you see in the viewfinder could be modeled by shooting two shots from a tripod, one at F1.4 and one at F4. Load them both in photoshop, the F4 shot as bottom layer, the F1.4 shot as top layer. Then lower opacity of the F1.4 layer to ~25%. This is the "F1.4" you see in the vievfinder.
p.5 #5 · Risk of focus shift issues in new 1.4/35?
I really don't understand why you are listing other problems when this thread is about focus shift. You are talking like most people here are idiots and that they have no idea of how their gear works. Why do you even bother?
p.5 #9 · Risk of focus shift issues in new 1.4/35?
Steve Spencer wrote:
The reason that autofocus can't solve this problem is because focus shift varies with shooting distance. If the lookup table changed with the aperture it would work for some distances and not others. For AF to try to overcome focus shift it would need to adjust focus for both aperture and shooting distance. This would be a much more complicated system and would require that the system could precisely measure shooting distance, which I don't think it can.
Possible? Yes. Complicated? Not really.
Precise measurement of shooting distance is the fundamental on which autofocus actually works in the first place. Wanted focus distance and aperture is sent to the lens electronically. A 2-dimensional LU table is fully feasible. The distance or actually the DoF limits near and far can even be found as tags in a RAW file today.
Mar 14, 2011 at 07:18 AM
Steve Spencer Offline • • • • • • • Upload & Sell: On
p.5 #10 · Risk of focus shift issues in new 1.4/35?
I think the key to what you are saying is that, "focus distance and aperture is sent to the lens electronically," in order to make the system work to overcome focus shift you would need the LU table to vary by the lens that is attached. Right now there is a micro focus adjust that varies by the lens attached but this is a pretty primitive adjustment that does not take distance into account. Granted a more complicated system could be developed so that the LU table is adjusted for lens attached, aperture, and focus distance, but we are still a long way off from that happening and it would be a substantially more complicated system than the current one.
p.5 #11 · Risk of focus shift issues in new 1.4/35?
I agree that we are a long way from that happening, but I was thinking that the focus shift is largely a design issue and not a camera calibration issue, so the LUT could be implemented in the lens microcontroller on a permanent basis.
p.5 #12 · Risk of focus shift issues in new 1.4/35?
theSuede wrote:
Even though the -S screen is noticably better (I've modified and used them myself) it only adds about half an Ev of light from F3.2 down to F1.4. That's an 80% light-loss.
What you see in the viewfinder could be modeled by shooting two shots from a tripod, one at F1.4 and one at F4. Load them both in photoshop, the F4 shot as bottom layer, the F1.4 shot as top layer. Then lower opacity of the F1.4 layer to ~25%. This is the "F1.4" you see in the vievfinder.
thanks for the wonderfully informative post! out of curiosity, how do split prisms fit into this equation?
p.5 #13 · Risk of focus shift issues in new 1.4/35?
sebboh wrote:
out of curiosity, how do split prisms fit into this equation?
From what I remember from previous discussions, split prisms need a steep angle to be accurate for wide apertures, but this causes them to black out quickly when slower lenses are used or when doing stop-down metering.
p.5 #14 · Risk of focus shift issues in new 1.4/35?
Split screens works just like the AF-system. It redirects rays from two different parts of the aperture to the same direction (towards the eye or the AF-line sensor pairs). Where they differ is in systematical approach; the AF system makes two pictures (that the system tries to align) by aiming two sets of prisms with an aperture of ~F22-32 at two opposing points on the lens aperture. The matte screen prism has a relative aperture as large as the surface of the prisms allow.
The higher the angle between the prisms, the faster the picture halves "pass each other" when you focus through the target. The same goes for the AF optical system, and here it's called the base line distance and determines the size of the aperture needed for the system to work. It's all about angles, every time.
Trying to make a F2.8 base pair work with a F5.6 lens will make the line sensors look at the back of the aperture mechanism. Darkness. The same goes for the split prism optics in the matte screen. If you trace a ray from your eye in the VF in through the camera, too large an angle will redirect the ray so that it hits the back end of the aperture mechanism in stead of "fitting" through the aperture itself. Darkness.
BTW;
I think they implemented the focus shift control in the latest HB4 series of medium format cameras. it needs a fully electrical coupling to the lens, and it makes the system slightly slower. The order in which you normally want to do this is: Focus measurement, AE metering and focus actuation at the same time, shutter. To implement focus shift control you need to know the stop position of the focusing actuation, AND the aperture you're going to stop down to. There are ways to solve this quickly, but none that's efficient enough for an all-round camera built for speed.
p.5 #16 · Risk of focus shift issues in new 1.4/35?
I didn't quite get in what respect a Nikon camera is working differently with regards to focus shift than a Canon camera. Can someone please clear that matter for me?
I use a D700, and I own the ZF 85/1,4. Anything I can do to make fokusing easier and more reliable?
There are no different screens you can buy from Nikon, and the screens from older cameras seem not to fit, or do they? Katz-Eye sells some, but is anyone using them on a D700 and are they really making things better?
I recently got a magnifying eypiece DK 17-M (1,2 x). Works alright, but doesn't help all that much and you see less of the viewfinder. Getting the focus right with the 85 is hard work, and I'd like to improve the system if this is possible.
As to focus shift: is it correct that Nikon cameras stop down in live view to the apperture set, so if focused with live view focus shift should not be an issue? Or is the stopping down of live view irregular?
I understand that the "focus uncertainty" is hard to overcome at f 1,4.
