Real world focus shift

[Lindolfi]

Advertisement (gone after registration)

...if the lens is a single spherical surface?

 

If you look at s.a. curves for real lenses ("focus shift in the focal plane" vs aperture) they are all over the place, and it doesn't look like things work this way. But who knows...

 

The same principle holds for a complex design with spherical surfaces, like a double gauss. Here I simulated it with object distances of 0.3 meter and 5 meter, to show that the shift in the image plane grows for close focussing (even faster than image distance)

 

close focus:

 

far focus:

 

(Ray tracings were done with a package I wrote for educational purposes. The recipe of the lens (glass types, locations and radii) is available if required)

[Paul J]

Mine simply wasn't as 'sharp' as the 50 'lux, focus shift or otherwise. The 85/1.2 on the other hand was a fabulously sharp lens, but very difficult to focus on precisely the point at which focus was required. I have noticed no focus shift on the 50 'lux asph for what that is worth.

 

Yes these findings are same as mine. The 50mm 1.2 while sharp is not amazingly sharp. I still really like the look this lens has though and use it occasionally when I want that look. I have a noctilux on the M9 which is leaps and bounds ahead of it but I miss the close focussing distance of the Canon which is really very good. The level of detail in Canons though is really quite ordinary.

 

The 85mm L is also a nice lens. Yes it's sharper wide open but I had issues with focus too. All my concerns about moving to a manual focus Leica were blown out of the water and I really much prefer to manual focus now. With good eyes and a calibrated lens and body then it's so much more accurate and I don't think much more slower. It's more confidence and mindset that set it's apart.

 

Interestingly, I've not found any focus shift yet in my wide open 35mm summilux or noctilux 0.95 either.

[Lindolfi]

Interestingly, I've not found any focus shift yet in my wide open 35mm summilux or noctilux 0.95 either.

 

For the Noctilux 50/0.95 that is not surprising: the spherical aberration is well corrected, given that out of focus highlights further away than focus distance have even light distribution without the bright edges visible in those of the Noctilux 50/1.0 and Nokton 50/1.1, both suffering from spherical aberration and focus shift. The Noctilux 50/0.95 has more optical design parameters to play with than the other two.

[Guest #12]

A primer in the effect of stopping-down on focus-shift and MTF curves, see pp24-27 of Nasse, H.H. (2008) 'How to read MTF curves', published by Zeiss as no. 30 ('Measuring lenses objectively - Part 1') in their Camera Lens News series and available to download from the Zeiss website.

 

Thank you for the good reference. The example here is one lens wide open and then three stops down, so you will not see the behavior at all apertures, or the idiosyncrasies of a variety of lens designs; nor in the example of Bert's (thank you Bert for the interesting post).

 

Nor is there any discussion of the effect of subject distance on focus shift.

[ohnri]

I'll cut to chase first: For lenses known to focus shift (35lux asph v1, Zeiss Sonnar-C, Noctilux, etc), will you see focus shift at distances greater than five to ten feet, or only if shooting closer than five feet?

 

I ask for clarification, because this is the one detail I don't see spoken of in all the focus shift conversations. I've read Sean Reid's reviews and am surprised how many M-mount lenses show the shift, but all of Sean's focus targets are pretty close up. No one mentions farther distances.

 

I use and love the Canon 50mm 1.2 L lens, an optical design which displays focus shift from 2.0 to 5.6, only at distances of 5 feet or less. Beyond 5 feet, focus shift does not appear in real world shooting. Does the same apply to the Leica lenses?

 

My favorite M9 lens is my Noct f/1.

 

Not far behind is my 35 'lux v.1.

 

I routinely use these lenses for portrait work and I have yet to have any issue with focus shift. Certainly never with any landscapes either.

 

The way these lenses render is, however, something I would not sacrifice in any event. Remember, it is a trade off. If you get measurably (even if unnoticeable in prints) better focus shift you may well lose something far more important and harder to measure in a lab.

 

Best,

 

Bill

[SJP]

I'll reiterate...if you want the focus shift, you can get it on one side of the lens or the other...in one way by tracing rays of different heights, or another way by measuring the shift with a ruler like Lindolfi did. But this doesn't have anything to do with depth of field or the lens equation.

Disagree. The DoF range does not change appreciably but it moves with the point of focus. DoF is governed by the aperture, the focal length and the focus distance and can be calculated easily. However, as the name implies the point of focus shifts - which means that the focal length is slightly different for different apertures. This is caused by more or less of the lens(es) surface being used and thus can indeed be calculated via ray tracing & similar software. So determine focus shift experimentally, or via ray tracing, and then you can back calculate that to the real focal length versus aperture. This then allow to predict the focus shift versus aperture for any focus distance. You could also calculate the real DoF, but focal length change should be so small that that is negligible.

 

If someone can post some focus shift data at 1m and 5m I can show how the calculation works (35/1.4 summilux previous version?) or is there place where I can find that sort of numbers?

[01af]

Advertisement (gone after registration)

... the focal length is slightly different for different apertures.

That's right. Usually, the actual focal length is slightly shorter at wider, slightly longer at narrower apertures. So the plane of focus will shift towards greater distance when stopping down.

 

 

This then allows to predict the focus shift versus aperture for any focus distance.

No, it doesn't. Focus shift depends on many factors, including focusing distance. So if you know the amount of shift for one distance, you cannot simply extrapolate from there to other focusing distances.

 

Typically, camera lenses are optimised for infinity focus, so they will become worse at shorter focusing distances. Worse means, more spherical aberrations and hence, more focus shift. This does not mean focus shift will always be zero at infinity ... but it does mean focus shift usually will be less (as measured at the image plane) at infinity than at shorter distances.

[Lindolfi]

This does not mean focus shift will always be zero at infinity ... but it does mean focus shift usually will be less (as measured at the image plane) at infinity than at shorter distances.

 

That appears to be the case in four other designs I tested with raytracing with spherical surfaces. Unfortunately I don't have the design prescriptions (glass types, locations and radii of the surfaces) of Leica, Voigtlander and Zeiss lenses for M mount, else I could get the story for those with ray tracing.

 

Does anybody have design data of those lenses? ( Probably not, given that it is part of the crown jewels of those companies...)

[IWC Doppel]

From conversations I have had with Leica people actual focus shift must be less than 20-30 microns otherwise there would have been a noticeable issue on film

 

I guess if you really get to know your lens and have access to testing to know how and when you need to adjust a little manual adjust ment ( moving post focus ) could be acceptable

 

Are Nikon/canon and other lenses as 'susceptible' I wonder, does physical lens side play a part ?

[jaapv]

It might be a little bit more - film has no instant feedback.

[Lindolfi]

Here's a little experiment I did to show the relation between focus shift and spherical aberration. A Nokton 50/1.1 was focussed at 1 meter to a focus test chart at 45 degrees angle. Three images were made:

 

[1] At f/1.1

[2] At f/2.8

[3] At f/1.1 with a central circular obstruction, the size of the opening at f/1.4

 

These three images are mounted in the below figure from left to right at 100% with arrows indicating the position of sharpest focus. As you can see the focus shift from f/1.1 to f/2.8 is backward by about 20 mm, while the focus shift is forward with the central obstruction by about 8 mm relative to the image at f/1.1

 

 

Obviously the focus from the outer ring of the entering light is closest by, while at the center it is furthest away. At f/1.1 without obstruction, you get a mixture of all these contributions, resulting in an image with some drop in contrast and finest detail.

 

The shift in the image plane of an object at 1 meter shifting by 28 mm in object space is 0.085 mm for a focal length of 50 mm.

[SJP]

Nice, my trusted spreadsheet gives 0.987 - 1.01 m for the DoF (50mm, f/1.1, 1m), your chart seems to be quite close to that although I guess everyone will put the DoF boundaries at a different place.

 

If we assume that the focal length is 50 mm at f/1.1 then the observed shift of about 18 mm corresponds to a focal length change from 50 to 50.0445 mm.

 

For 5m focus that would boil down to a focus shift from 5m to about 5.5m when going from f/1.1 to f/2.8 which seems an unreasonably large shift. Do you have any data on that?

 

The central blocking disk is really fun as it shows that by using such an obstacle you can pretty much keep the DoF while inceasing contrast. You probably do not even lose too much light capturing ability. The disadvantage is that the bokeh is suddenly like a mirror (schmidt-cassegrain) objective.

[01af]

For 5 m focus that would boil down to a focus shift from 5 m to about 5.5 m when going from f/1.1 to f/2.8 which seems an unreasonably large shift. Do you have any data on that?

That's what I was talking about.

[Lindolfi]

 

For 5m focus that would boil down to a focus shift from 5m to about 5.5m when going from f/1.1 to f/2.8 which seems an unreasonably large shift. Do you have any data on that?

.

 

Yes, please see my posting #9 of this thread: http://www.l-camera-forum.com/leica-forum/customer-forum/225595-real-world-focus-shift.html#post1992799

It appears to be 5.4 meter. That seems large but it isn't, just try it yourself. You can use a textured carpet on top of which 5 cm. marks on paper.

[SJP]

That's what I was talking about.

Talking is useless, all we need is numbers.Worth noting that this is still well within the 50/2.8 DoF range so maybe no-one is complaining. I really don't know, all we need is data of focus shift at 1, 2 ,5, 10 m nominal focus versus aperture.

[01af]

... all we need is data of focus shift at 1, 2 ,5, 10 m nominal focus versus aperture.

We? I don't need them. If you feel you need them then I'd suggest you take your lens and camera, a tripod, and a yardstick and collect those data. You already got them for 1 m and 5 m, and your results basically are in accordance with theory. I think that's all the data 'we' need. If you want more then ... more power to you.

[SJP]

We? I don't need them. If you feel you need them then I'd suggest you take your lens and camera, a tripod, and a yardstick and collect those data. You already got them for 1 m and 5 m, and your results basically are in accordance with theory. I think that's all the data 'we' need. If you want more then ... more power to you.

Indeed, I think we have all the data we need. I hadn't seen all the data available earlier on in this thread or indeed Bert's posting @ 22.39 yesterday.

 

Anyway, it appears that focus shift at any focus distance can be predicted reasonably well using the effective focal length and the simple lens formula. I expect the errors in visual/experimental determination are well within the calculated shifts & DoF errors of margin. Still it is nice to see that theory actually works, pretty much straight out of the box high school physics.

[Guest #12]

...

 

If we assume that the focal length is 50 mm at f/1.1 then the observed shift of about 18 mm corresponds to a focal length change from 50 to 50.0445 mm.

 

For 5m focus that would boil down to a focus shift from 5m to about 5.5m when going from f/1.1 to f/2.8 ...

 

Anyway, it appears that focus shift at any focus distance can be predicted reasonably well using the effective focal length and the simple lens formula.

...

 

I don't understand what you're doing...you seem to be using a focus shift of "18mm" (not sure where you got that) at 1m with a 50mm lens, and the usual lens equation, to get a focal length after stopping down of 50.044mm. Then you assume the focal length after stopping down will be the same at any distance. Then you apparently used the same 50mm lens at 5m and the focal length after stopping down of 50.044mm, and the lens equation, to get a new distance of "about 5.5m" and a focus shift of about 0.5m. Was this supposed to be close enough to the 0.4m focus shift actually measured? Am I missing a post, or are you comparing your calculations to only two measurements that someone else made? If you use one measurement instead of the other to start with, you will get something completely different for the "effective focal length."

 

I think as someone suggested you have the wrong premise, and focus shift depends on other factors.

[SJP]

Nope, see Lindolfi #31, the posted images give the 18 mm at 1m shift and #9 also not measured by me gives the 0.4m shift at 5m. This is both in agreement with the minimal change in focal length from 50 to 50something.

 

So all of this agrees with the lens formula which I find surprising, or at least somewhat surprising. Apparently what my brain tells me is not the same as what simple lens equations tells me. I trust the latter rather than the former.

 

Edit: and yes 0.4 m or 0.5 m focus shift at nominal 5m is close enough to the mark by any standards. That is why I requested more data, not only 1m and 5m, just to firm things up. But considering the good agreement without any preconceptions I am happy enough with 0.5 vs. 0.4. In fact I am still surprised.

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[Lindolfi]

Perhaps I can help here:

 

1] if you assume that focus shift in terms of image shift (due to spherical aberration) is independent of object distance, then you can do what SJP suggests, simply calculate the shift in the image plane back to object space with the lens formula. I agree with him that the large shifts you get at large object distances are counter intuitive, but do agree roughly with measurements.

 

2] both according to my ray tracing simulations and measurements, spherical aberration and its focus shift in terms of image shift does depend on object distance in many designs: it is worse at short object distances than further away. In that case the lens formula is insufficient, you need complex ray tracing simulations and data on the lens design (glass types, radii and locations of surfaces)