17mm TS-E and focal plane curvature?

I have read that the 17mm TSE (and other TSE’s) have some focal plane curvature, necessary to accommodate the tilt-shift. However, I never appreciated the potential consequences for landscape photography and have not seen anything explicitly describing the phenomenon I’m going to show. So I decided to share some observations using the 17mm TS-E that I hope may be useful to some people buying this lens, or maybe even some who already have this lens. I’m putting this out there basically for both comments and as something to watch out for if using this lens for some landscape applications with close foreground elements.

I would also say that although there is a possibility that my copy of this lens is faulty, I have not been able to produce any examples where there is something obviously wrong. More comments on this at the end.

So what I usually do to get landscape shots with both near and far elements in best possible focus, is:

1. guess an aperture based on experience and how close the nearest image element is
2. focus on infinity close to the center (or top center) of the image
3. Move the focus to the ‘near’ until infinity is just acceptably sharp
4. Check nearest part of the image
5. If the nearest part is not in focus, I will stop down up to a point (usually f16), if I am at f16 and it looks like I could open up the aperture I will try the same procedure at f11 and go from there.

So without any tilt I expected this method to work with the TS-E. Unfortunately I was proved wrong, fortunately for me this did not occur on a shoot where I got anything worth printing.

So what happens is that if you are focused relatively near, say 2-4ft away and pushing the limit of the DOF with the aperture you have, the center-infinity part of the image may be very sharp, but the sides can be OOF.

The attached images show the phenomenon. Shot at 1/80s f11 and focus set near the foreground (the bolder is about 6-8 ft away). The main image just shows the overall shot with boxes indicating the crops (in next post). I don’t know exactly what % magnification the crops are, but this effect is easily visible on a computer monitor, so would definitely be very obvious in a print. In this example the near and far in the center of the image are relatively sharp, but right on the edge of the DOF, whereas both left and right portions of the image are quite soft. If I move the focus towards infinity, the sides of the image become in focus and as sharp as the center infinity (which also sharpens up a bit as expected).

As I mentioned earlier, I don’t think this is an issue with my copy of the lens. If I’m focusing at 10+ ft at f8, everything from 5ft to infinity is sharp. I haven’t really tested this using hyperfocal distances, since I don’t normally use them, but that and properly comparing with say the 17-40 focused at the same distance would be interesting. If I use tilt, then since I’m focusing near infinity, everything I would expect to be sharp is sharp. Also the left and right sides are always approximately the same sharpness and the bottom corners are relatively sharp (although distorted), which suggests it is not due to a misaligned element.

Any comments/suggestions of other info on using the TS-E that are not commonly out there would be appreciated.

I have more questions, thoughts and comments on the 17mm TS-E, but since this is already a long post, I’m going to leave it at that for now. Definitely an interesting lens.

HOSTED PHOTO
please log in to view hosted photos in full size.

First two crops - soft (OOF). These parts do became in focus and very sharp is I more the focus towards infinity.

HOSTED PHOTO
please log in to view hosted photos in full size.

HOSTED PHOTO
please log in to view hosted photos in full size.

next two crops (sharp)

HOSTED PHOTO
please log in to view hosted photos in full size.

HOSTED PHOTO
please log in to view hosted photos in full size.

ejenner wrote in post #13663656
I have read that the 17mm TSE (and other TSE’s) have some focal plane curvature, necessary to accommodate the tilt-shift.

That doesn't sound right to me. Why on earth would a lens need focal plane curvature to accomodate for tilt-shift?

What does happen is that when positioning the camera/lens closer to the ground, and generally getting closer, the focal plane curves in, towards the lens plane and hence appears to have a curve.

Generally speaking, if a lens shows field curvature, that is not because it is designed in, but because it couldn't be designed out. This is especially difficult with asymmetrical (U)WA lenses and zooms, IOW, retrofocus lenses.

This lens may well have some field curvature, and from what you say quite a typical one, i.e., in a wide arc around the lens, going through infinity so to speak and closer to the left and right.

I must say I have never noticed this yet, however. I tend to try and get the things in focus I want in focus, and that rarely is infinity - not enough pixels to resolve infinity anyway.

I have compared my own TS-E 17 with a ZE 21, a Nikkor 14-24, and a Sigma 12-24 EX (old model), and nothing comes close to it in the (extreme) corners, let alone a 17-40L. I was looking for something better than a 17-40L at 17-20 mm in the corners when I tried and/or owned all of the above, the TS-E 17 is the only one remaining.

I have to say that I shoot mostly at F/6.3, and only occasinally I may stop down to F/8, but never beyond that. Also, I don't see these results in the TS-E 17 sample thread, so I am just wondering whether you possibly have a less good specimen?

As to hyperfocal distances, don't bother, they don't really work . Best is to make sure that the main subject is sharp, whatever the eye is led towards. Anything else rarely matters. That's the compact version of Merklingers blog and article on the subject, with a slight twist .

HTTP response: 401 | MIME changed to 'text/html'

HOSTED PHOTO
please log in to view hosted photos in full size.

wimg wrote in post #13664150
That doesn't sound right to me. Why on earth would a lens need focal plane curvature to accomodate for tilt-shift?

I don't actually know, I've been trying to figure it out in my head, but to be honest I still haven't completely sorted out exactly what happens with tilt with regards to ray paths and focal planes. I'm content for now to keep learning what to do in practice.

http://www.luminous-landscape.com/tutorial​s/focusing-ts.shtml

It is mentioned here at the bottom, with a bit of an explanation and saying that you should stop down one stop, but saying it is more of a problem at WIDE apertures. Some of his wording seems confusing to me though. I don't take pictures of brick walls and don't care if my lenses aren't designed to take super-sharp images of flat subjects at wide apertures (OK, that's a different rant also involving test chart examples posted on the net which totally distort some lens' true capabilities.).

wimg wrote in post #13664150
What does happen is that when positioning the camera/lens closer to the ground, and generally getting closer, the focal plane curves in, towards the lens plane and hence appears to have a curve.

I'll have to think about this, but I haven't noticed same thing with the 17-40 although as you point out, it would be much harder since it is no were near as sharp to begin with, and a bit of loss of sharpness on the edges would be expected. I agree that anything slightly OOF or off with the TS-E is so much more obvious becasue it is sooo sharp when focused exactly. Actually in re-looking at my examples, the 'poor' areas aren't much worse than one might expect with the 17-40 anyway. It's only when you are looking at the image as a whole that it really becomes obvious becasue of the variation across the frame. And actually those edges are somewhat worse than a properly focused 17-40 at f16 - I did compare while I was shooting.

Just as a point, in the above situation I was able to get very nice overall sharpness at f16 with some tilt - this example wasn't where I first noticed this and I was specifically trying to get a good example of the phenomenon.

wimg wrote in post #13664150
I have to say that I shoot mostly at F/6.3, and only occasinally I may stop down to F/8, but never beyond that. Also, I don't see these results in the TS-E 17 sample thread, so I am just wondering whether you possibly have a less good specimen?

I thought and was hoping you might comment in this thread, but I remember you mentioning the apertures you usually use, and I have to say I have not seen this in situations when I have used/needed less than f11 - landscape or otherwise. For instance I just took a shot of the snow coming down in the back yard, hand held and standing at f6.3 - everything is tack sharp.

In your examples, chances are you were focused close to infinity (like 6ft or more) and in this case I never notice it. At f4, if I focus near infinity, the edges are just as sharp as the center. It only happens when I'm focusing close, definitely less than 6ft. I also plan on doing more tests at f4, but so far everything looks as expected (I haven't tried a brick wall yet), including with tilt on a relatively flat area.

I understand there are a lot of uses for this lens and even on flikr, looking through the examples most are not where I think you would see this. I have not seen a lot, if any examples with the foreground at less than 2-3ft away. That's one reason I'm not convinced I have a 'bad' copy, but I do intend to keep testing. I figure if I do have a bad copy, I prefer to send it to Canon (under warranty) than just return it anyway, and it is clearly super sharp (I know that's all relative), so if there is an element alignment issue, I'd rather just get that fixed than try another copy.

I also don't know, for instance, what the edges would look like compared to the center with the 17-40 focused at exactly the same distance with the same aperture. It might be that infinity is all in focus or that infinity is all slightly OOF.

ATM I am considering it as just something to be aware of and watch for in certain situations, rather than being overly concerned about having a 'bad copy'. But I do like to take shots with my foreground a foot or so from the lens with this wide angle.

Thanks for your comments, I appreciate it,

Ed.

wimg wrote in post #13664150
T
As to hyperfocal distances, don't bother, they don't really work . Best is to make sure that the main subject is sharp, whatever the eye is led towards. Anything else rarely matters. That's the compact version of Merklingers blog and article on the subject, with a slight twist .

Yea, I think we've already semi-discussed this and I totally agree. In fact I was just looking at the DOF scale on the TS-E and noticed that is it quite different than the typical DOF or hyperfocal distances one might encounter.

But, in many situations I do like to at least start of with somewhat even sharpness across the image if possible, and then perhaps selectively sharpen from there.

In fact, also to re-iterate the point you are making is that I went to Utah recently with my 17-40 which had an element mis-alignment issue at the time that I hadn't fully appreciated (I tend not to shoot brick walls or test my lenses much unless I notice a problem). It somewhat ruined the large (14-20") printability of some shots, but on others it is unnoticeable or even improves the image depending on where the important elements were in the frame.

Hi Ed,

Just a quick reply for now, it is getting a little late over here. Have you had a look at or own Lester Wareham's site yet?
http://www.zen20934.ze​n.co.uk/photography/ti​ltshift.htm
He explains it quite well, both the tilting of the plane, and the influence of the pivot point height. I think the plots make things clear.

I find that one doesn't have to focus very close with the TS-E 17 to get everything really sharp. My castle tower picture in a stone door frame is an example. The plants in front, of which I can see the individual leaf nerves in the full picture, was less than a foot away. I had troubles not having my shoes or knees in that particular shot. That was shifted a little however, and focused about 2.5 m (say 8 ft) away if I am not mistaken, and shot at F/6.3. I'd have to check my notes.

That shot is easily printable at 60 cm X 90 cm (24"X 36"), which I will, next month or so.

Kind regards, Wim

Hi Ed
I've not seen anything remotely like the effect your lens is producing. Assuming the tilt is zeroed, mine is sharp, edge to edge even when I max out the shift. I seldom have to use it past f/8 and never past f/11.

In the picture with the four red squares, the top three are all of areas that are OOF. You've got bigger problems from trying to set hyperfocal distance than field curvature would cause. I agree that this just looks like the lens' scale being a bit confusing. (I oughta read the manual and see what it says about the use of the scale - I also would be interested to see what the DEP mode would do with this lens, on a DEP-able body.)

Ever tried using the tilt to use the Scheimpflug principle to get the ground in front of you sharp and the distance as well? It won't be as handy for good - light landscapes as setting a hyperfocal distance but it's useful to experiment with for marginal lighting (or for dramatic effect). Difficult to use would be putting it mildly, but breathtaking when done right.

Yes you are using the Tilt shift lens movements incorrectly. You are not trying to create curvature in the focal plane. Firstly, read this much simpler tutorial on focusing a TS lens
http://www.davidsummer​hayes.com/techniques.h​tml
and open the link to "Focusing the Tilt -Shift Lens" pdf

The very first figure is exaggerated, but the idea is that you are tilting the focus plane from parallel to the sensor plane to something much more angled. The thickness or depth of the focal plane (volume of blue area) will depend upon your f stop value, a thicker wedge for small apertures.
Far down in the article you will find a tilt table for the 17mm focal length as a function of height above the focal plane (usually height above the ground if you want to keep the ground in focus). Study it carefully!
If the center of your camera lens is at standing height (2m) you only need 0.5 degree of shift. The amount of shift hardly changes for kneeling height. However as you get within 1 foot of the ground you need about 3.5 degrees of shift, and as you get closer the shift angle increases rapidly with every inch.
Now read the article's description as to the technique for fine adjusting the focus and the tilt angle, as they interact. Finally choose a very simple example to learn your tilt technique. Notice Summerhayes use of the flat beach for this purpose. I used to use my driveway as I refined my shift technique for distances below 8" height.
Mike K

Mike K wrote in post #13665092
Yes you are using the Titl shift lens movements incorrectly. You are not trying to create curvature in the focal plane. Firstly, read this much simpler tutorial on focusing a TS lens
http://www.davidsummer​hayes.com/techniques.h​tml
and open the link to "Focusing the Tilt -Shift Lens" pdf

The very first figure is exaggerated, but the idea is that you are tilting the focus plane from parallel to the sensor plane to something much more angled. The thickness or depth of the focal plane (volume of blue area) will depend upon your f stop value, a thicker wedge for small apertures.
Far down in the article you will find a tilt table for the 17mm focal length as a function of height above the focal plane (usually height above the ground if you want to keep the ground in focus). Study it carefully!
If the center of your camera lens is at standing height (2m) you only need 0.5 degree of shift. The amount of shift hardly changes for kneeling height. However as you get within 1 foot of the ground you need about 3.5 degrees of shift, and as you get closer the shift angle increases rapidly with every inch.
Now read the article's description as to the technique for fine adjusting the focus and the tilt angle, as they interact. Finally choose a very simple example to learn your tilt technique. Notice Summerhayes use of the flat beach for this purpose. I used to use my driveway as I refined my shift technique for distances below 8" height.
Mike K

Thanks for that, but I think you mis-understood me. I am not trying to create focal plane curvature - I just think that is what I'm observing. The above example is without any tilt, but there was a little shift. I have tested without shift too though.

I do understand the basics of tilt, and was just making the point that I can use it in the way I would expect - in other words I don't see anything fundamentally or obviously 'wrong' with the lens when I use tilt that may suggest something catastrophically wrong.

Personally I used a frozen lake for my tests - but they were essentially the same and I indeed I could get everything in focus, from 2ft to infinity at f4 at various tripod heights. When using tilt I DO NOT see the above effect (I think becasue I am focused near infinity). In the above shot the only issue I had when using tilt was the rock in the very lower left corner which was just under the focal plane.

Since we seem to be collecting articles, I will just add this one, which for basics I found really good for visualizing what is going on with the animations.

http://www.northlight-images.co.uk/article_p​ages/using_tilt.html

I carefully reread your OP and realized you were not talking about using tilt at all. I am sorry about the confusion. I was trying to find the word "tilt" in your steps 1-5 and realized you were not describing a process of lens tilt, but trying to detect field curvature as described by Summerhayes in LL.
"It is necessary for the lens not to be a flat field design to accommodate ‘shift’. Imagine what would happen to your focus when you shifted left or right if the edges of the lens focused further away from the centre! When shifted the centre focus would jump forward, magnification would change ever so slightly making stitching difficult."

The accuracy of Summerhayes claim as to T/S necessity for field curvature is not obvious to me, but In any case the field curvature referred to is with strong shift, at the extremes of the image circle, while the example you show is unshifted. There should be no visible field curvature unshifted. The center 1/2 of the image circle should be not have field curvature; if there, it would be at the edges of the image circle, only visible at extreme shifts.

I recently bought a ts-e24mm ii and at the risk of getting boxed around the ears a bit again (like the tse-24mm ii thread):

When I shoot the lens close and square to a flat surface (wall), close enough to have a newspaper mostly fill the viewfinder (maybe 3ft?),with the center sharp the edges are visibly softer than say my 24-105L both at f/4 (swapped in changing nothing but focus).

Even at f/6.3 a close flat surface is not all in focus. Esp the left side.

Multiple tries yield the same results.

When I focus further it does not seem to be nearly as true if at all (not positive on that as harder to control)....

Anyone with a tse17 or tse 24II willing to shoot a flat target up close at f /3.5 or f/4 and report? I am curious....

I am talking no shift no tilt here.

All wide angle lenses on FF are softer in the corners and extreme edges. Some are more dramatically so than others. The 24 TSE II f3.5 is not unusual in this regard. Show me a WA FF MTF graph that has the corners as sharp as the center. This corner softness is a pretty universal performance characteristic. If all WA FF lenses are softer in the corners does that mean that all WA FF lenses have field curvature? Your test of flat field sharpness is a necessary but insufficient criteria of field curvature. Perhaps a better test would be if the corners sharpen up if the target were concave? How much concavity is required for the label of field curvature?
I believe enlarging lenses and most macro lenses are of flat field design but the vast majority of camera lenses often have a bit of field curvature, especially WA designs. Macro lenses are also much sharper in the center than the corners when wide open.

Is there some specific concern of yours?
Mike K

Mike K wrote in post #13670607
A Your test of flat field sharpness is a necessary but insufficient criteria of field curvature. Perhaps a better test would be if the corners sharpen up if the target were concave?

Or if they sharpen up a lot wen changing focus?

I think this has been talked about when either shifting a lot, or on flat targets, and I haven't tried that yet, but I was surprised to see it near infinity.

I'm going to investigate more before deciding what to do with the lens or whether to try a different example.

I have also done a lot more reading since posting this and it does look like something others have observed, mostly with different lenses though (TS-E 24 (I/II) and possibly ZE21).