[Guest]

Hi Klaus & others,

 

Would it be possible to introduce a metric for field curvature in your tests, so that one can estimate how strong the effect actually is?

 

There is a concrete background to my question; I currently use the EF-S 15-85 on APS-C and the field curvature almost rules out this lens for scenery where there is relevant detail in the corners.

 

Now I am contemplating the EF 24-105 3.5-5.6 STM for a FF setup, but in your review I read about field curvature being "significant and noticeable". Now I am of course wondering how bad it really is. If it is as bad as my 15-85, I would definitely not buy such a lens.

 

Could you think of any way to express field curvature as a number, or quantify it otherwise?

 

If not, perhaps one could include sample pictures in future reviews that are taken with the following method: focus on a far-away subject with the center AF point; take a shot; then disable AF and recompose; take four more shots, recomposed, showing the focused subject in one of the four corners of the frames each time. This would also give an impression of the centering quality of that particular lens sample. While the centering may not be representative for other samples, this procedure would give a good impression of field curvature as well; something like the average over all 4 corners.

 

Then of course for a zoom, one would want these samples at different focal lengths, so it is getting a bit tedious. ;-)

 

But perhaps someone with better background in optics could think of a more scientific metric for curvature of field?

 

Thanks & best regards

 

Tom

 

 

 

[Klaus]

The problem is that the EXIF data is not accurate enough to produce "hard" data. If it was we could just take the delta between the center focus and the corner focus.

 

If things are really extreme we tend to publish a 2nd MTF chart e.g.

http://www.opticallimits.com/samsungnx/6...24?start=1

 

Sample images are a difficult topic. It would have to be a constant scene in order to allow a comparison across different lenses.

[Scythels]

Field curvature can mathematically be described (really, is) in terms of mm of defocus at the image plane.  The issue is that this is scaled by the focal length.  There is no way to model the field curvature in terms of the object distance in a way which is independent to FL.  Short FLs will also obfuscate that with large DoF.

[youpii]

Lensrental has a "toy" to measure field curvature but that's probably beyond Photozone's budget

 

http://www.lensrentals.com/blog/2015/03/...onstration

[Scythels]

That's a trioptics machine - it costs about $100k.  It's doing a model of the field curvature in mm of defocus at the image plane. 

[Guest]

The problem is that the EXIF data is not accurate enough to produce "hard" data. If it was we could just take the delta between the center focus and the corner focus.

 

If things are really extreme we tend to publish a 2nd MTF chart e.g.

http://www.opticallimits.com/samsungnx/6...24?start=1

 

Sample images are a difficult topic. It would have to be a constant scene in order to allow a comparison across different lenses.

What would EXIF data be used for? FL and f-stop?

Personaly, I'd prefer slightly inaccurate data (let's call it "an indication") over just words in the text that are very subjective.

 

I also think that curvature information is sorely missing because the test procedure you are using masks it completely, if I understand correctly. In real life however, there is no practical difference between a lens that has blurry corners due to curved field and one that displays blurry corners for other reasons. So checking charts for corner sharpness of a lens is almost meaningless without knowing how strongly the image field is curved.  Again, I am talking from a practical, not scientific perspective, but still... One lens may have good test results, like the Canon 15-85, but in reality, it's corner sharpness may not be better than that of another lens which tested badly, such as the Canon 17-85.

[Brightcolours]

What would EXIF data be used for? FL and f-stop?

Personaly, I'd prefer slightly inaccurate data (let's call it "an indication") over just words in the text that are very subjective.

 

I also think that curvature information is sorely missing because the test procedure you are using masks it completely, if I understand correctly. In real life however, there is no practical difference between a lens that has blurry corners due to curved field and one that displays blurry corners for other reasons. So checking charts for corner sharpness of a lens is almost meaningless without knowing how strongly the image field is curved.  Again, I am talking from a practical, not scientific perspective, but still... One lens may have good test results, like the Canon 15-85, but in reality, it's corner sharpness may not be better than that of another lens which tested badly, such as the Canon 17-85.

In real life there is a big difference: if corners are blurry, they are blurry. Like the extreme corners of my(the) EF 35mm f2. If there is some field curvature, chances are there is something sharp in corners, because a scene usually is deep (3D).

Can you give examples where the field curvature from that lens gives you problems? Which focal length?

When I look at image samples taken with that lens, I only notice the CA in the corners, CA which is easily corrected.

https://pixelpeeper.com/lenses/?lens=13192

[Klaus]

What would EXIF data be used for? FL and f-stop?

Personaly, I'd prefer slightly inaccurate data (let's call it "an indication") over just words in the text that are very subjective.

 

I also think that curvature information is sorely missing because the test procedure you are using masks it completely, if I understand correctly. In real life however, there is no practical difference between a lens that has blurry corners due to curved field and one that displays blurry corners for other reasons. So checking charts for corner sharpness of a lens is almost meaningless without knowing how strongly the image field is curved.  Again, I am talking from a practical, not scientific perspective, but still... One lens may have good test results, like the Canon 15-85, but in reality, it's corner sharpness may not be better than that of another lens which tested badly, such as the Canon 17-85.

 

It is not that simple I'm afraid.

If we would take the readings on the flat plane they would be very misleading simply because you won't really notice a mild field curvature but it would spoil the numbers. Thus there's a risk of overshooting in either direction. If it helps we could provide the 'flat view' results more often - I have no problems with this per se. However, I am very sure that there'll be more questions than without - also because the centering quality will play a much higher role in those charts.

[popo]

How "easy" is it to get the camera aligned to the test chart?

 

Just thinking out loud here so feel free to shoot down... how about if you configure to maximise the MTF for two opposite corners at the same time, and without changing anything measure the centre. Then only changing the focus, maximise the centre resolution. Would the ratio of the two centre measurements be a useful indicator of the difference in focus from centre to corner? My mind hurts thinking how any other defects might impact this...

[Scythels]

You can't use MTF to measure defocus if you vary the position within the frame, as the aberration balance will change dramatically and mask the defocus (field curvature).  A plot of defocus (X) vs image height (Y) vs MTF (color/other indicator) is perhaps the best possible way to demonstrate FC.