3
\$\begingroup\$

While I have been analyzing sharpness and other optical properties of rectilinear lenses for a while, I have received a fisheye lens covering 180° and am having trouble producing an adequate test. Center sharpness is obviously easy to measure, which is most lenses is fairly good these days and so the main difference tends to be on how much softer are edges and corners and how much the improve while stopping down.

With the fisheye lens, it is difficult to get a target covering 180° field-of-view what is at edges is really distorted making it hard to make a comparison using what appears from outside the target. Bonus points if the method also applies to measuring vignetting. I know another major website that reviews and they show sharpness but not vignetting measurements for fisheye lenses.

\$\endgroup\$
10
  • \$\begingroup\$ Could you provide a bit more detail about your application? Specifically, are you trying to test the lens with or without the camera body? Are finite focus distances OK or are you testing at infinite conjugates? Are you testing the sharpness in terms resolving power such as lp/mm or are you testing Optical properties such as spherichromatism or wavefront error? If you provide more information about your current setup I might be able to recommend simple changes. \$\endgroup\$ Feb 19, 2018 at 19:50
  • \$\begingroup\$ Sorry for the double post; Budget would help too. DxOMark has excellent solutions in the $10k-50k US price range. \$\endgroup\$ Feb 19, 2018 at 20:01
  • \$\begingroup\$ Testing the lens on camera body, although I understand it has an influence, this is an acceptable compromise for testing something within a budget. Looking to compare resolution and resolving power at finite focus distances since I can't place a resolution chart at infinity while covering the entire field-of-view of the lens. \$\endgroup\$
    – Itai
    Feb 19, 2018 at 21:04
  • \$\begingroup\$ I decided not to address the vignette concern here. If you'd like to ask it as a separate question I'd be happy to share a few techniques for measuring vignette with various accuracy \$\endgroup\$ Feb 20, 2018 at 14:54
  • \$\begingroup\$ what are you compare sharpness of that fish-eye to? Is it fish-eye versus 50/1.8 or another fisheye? \$\endgroup\$ Feb 20, 2018 at 16:42

3 Answers 3

3
\$\begingroup\$

Inverse panorama

Use a single resolution target and pan your camera.

Remember that the goal is to obtain an image formed through different ray geometries of glass at the edge of the lens. You don't necessarily need to image through the glass of the center of the lens while doing so. To that end, you could accomplish your task of measuring sharpness in the corner using a panoramic photography rig. The difference is that, this time, instead of making multiple pictures to capture more than the FOV you will use multiple pictures to capture less. Some of the most advanced optical meteorology equipment in the world utilizes a single target and extremely accurate lens positioning systems.

There is plenty of information on creating a DIY pano rig here and on the web but for this test even a carefully measured lazy susan or turntable would work. To set the rig up you will need to mount that camera in your panorama rig with rotation around the nodal point as normal but also such that it is not horizontal. Instead you need to mount it at the angle corresponding to a bisecting angle of your sensor (that is the angle between the long side of the rectangle and a line going from corner to corner. The formula for that angle is angle=ATAN(height/width) Where height and width are either pixel counts or aspect ratios.

Once your camera is mounted at an angle, panning your panorama rig will cause the target to move from corner to corner. You are now ready to capture. Start with the target in the center of the sensor as a baseline. Next, move halfway to the corner and measure again, finish with the corner then repeat on the other side. If your lens contains an aspheric element you may wish to make more than five measurements so that you can look for third or fifth order trends in resolving power.

I'm not sure what your current measure of resolving power is but I will point out that SFR is a powerful metric for understanding lens performance. SFRMat used to be freeware and you can probably find a copy floating around in the web. the current version free trial allows 40 measurements. SFR or MTF tools can also be gotten from MITRE but I cannot vouch for them.

Freelensing could enhance your measurement accuracy further. Assuming you can effectively mask any stray light between the camera and disconnected lens, panning the lens while keeping the image in the center of the sensor has the advantage of not experiencing resolution loss to the sensor. This is an especially big concern with wide angle lenses due to exceeding the critical angle of the microlenses

Focus point is a matter of personal preference. Spherical aberration will make the focal plane for the edge of the lens different in the corner than it is in the center. You must therefore decide if you want to focus for the center and include spherical defocus in your measurements or focus each measurement to isolate that variable.

There are a few other ways to accomplish this task but I don't want to complicate this answer.

As a side note, Imatest produces a "pre-warped" resolution target for use with fisheye lenses. If you do want a target that can measure an entire lens in one go, they may be able to help. Expect to spend $1-4k US to get started though.

\$\endgroup\$
4
  • \$\begingroup\$ Going to follow this information, seems like it has good potential. \$\endgroup\$
    – Itai
    Feb 20, 2018 at 17:21
  • \$\begingroup\$ This technique has the added benefit of reducing the possibility of any target variation to zero. You will be comparing images of identical targets. \$\endgroup\$
    – Stan
    Feb 20, 2018 at 21:10
  • \$\begingroup\$ Front or rear nodal point, monochromatic illumination—comment? \$\endgroup\$
    – Stan
    Feb 20, 2018 at 21:11
  • \$\begingroup\$ Without wishing to start the nodal point debate here, I would say that you should rotate around the entrance pupil. Failing that, testing for the point of no parallax is the best bet. Illumination would be dependent upon intent. Monochromatic resolution has the advantage of probably reducing the complexity of the distortion equation. On the other hand a wide angle lens usually has significant spherachromatism. Might be interesting to compare monochromatic to polychromatic in order to define the chromatic effect. \$\endgroup\$ Feb 20, 2018 at 22:09
1
\$\begingroup\$

How about a satellite dish reflector (spray paint white) and a printable resolution chart (available at numerous websites).

Satellite Dish Reflector

Use an image editing program to crop the sharpness testing portion of the chart and attach a half dozen strips around the circumference of the reflector (adjusting by holding with fridge magnets).

Sharpness

Use an extremely short extension tube to get close focusing, you can test how long an extension tube you can get away with by freelensing. It would probably cost over $100, so you'd need a few lenses to test or be very keen on doing it. This should give you a 180° target that's large enough to light evenly (to test vignetting).

8mm Extension Tube

\$\endgroup\$
2
  • \$\begingroup\$ Very interesting. How would I ensure that focus is achieved at the edges of the frame though? \$\endgroup\$
    – Itai
    Feb 20, 2018 at 17:20
  • \$\begingroup\$ @Itai - Aperture, bracket, or adjust distance to bring the edge closer. \$\endgroup\$
    – Rob
    Feb 20, 2018 at 17:30
0
\$\begingroup\$

Here's an article I did on measuring an 8mm fisheye lens with 180-degree field of view using MTFmapper. It's free software, by the way.

https://www.photoartfromscience.com/single-post/2017/07/27/A-Better-Way-To-Test-Fisheye-Lens-Resolution

The program doesn't evaluate vignetting, however.

The MTFmapper program essentially un-distorts the fisheye photo to make it rectilinear, and then performs the analysis.

\$\endgroup\$
1
  • \$\begingroup\$ You need to give an overview of the method here, not just a link - even if it answers the question - since one does not know if it will become invalid. It's OK to keep the link since you disclosed it is yours anyway and also because there's much more details there than what you would put in the answer. \$\endgroup\$
    – Itai
    Feb 24, 2018 at 15:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.