olyacme wrote:
cogitech wrote:
Do we need two different targets for the two different types of CA?


Except for macro and closeup lenses, and in the absence of a good "artificial" star, real stars at focus remain the best way to test for lateral and longitudinal CA, among other aberrations.

For "defocus CA", a star can be placed slightly out of focus and the resulting disk observed for non uniformity (also good for checking SA and astigmatism). Otherwise, a tilted focus chart works fine.

/Acme

+999999999

Spot on! Everyone gets one guess which star I suggest...

If you poked a very tiny hole in a piece of aluminum foil wrapped over the head of a normal speedlight, would that be "small" enough?

Well, we are starting to go in many directions here. Re: the softbox idea; not everyone has a softbox (I speak mainly for myself here, but I can't be the only one). Re: stars; anyone who lives in a large city is out of luck (again, I speak for myself, but I can't be the only one).

Most everybody has a flash. Flash outputs a pretty good sunlight equivalent.

Kaa

theSuede wrote:
If you poked a very tiny hole in a piece of aluminum foil wrapped over the head of a normal speedlight, would that be "small" enough?

Pin pricks in aluminum foil can be used, but it's best to make a bunch and then chose a good one to minimize diffraction. Projecting light through a batch of prospective holes onto a wall is one way to do this. A laser cut 300µm (or so) hole, such as for a pinhole camera, would be a nicer option and fine for testing wide angle lenses at reasonable distances.

A flash's instantaneous output is high, but without continuous output or an effective concentrator its intensity per surface area over time is quite low compared to other light sources. A common 3-watt LED flashlight with the pinhole plate mounted on it might be a better choice. Exposures of defocused disks can be quite long, so mount the flashlight and the camera on stable footings.

A fancier option would be a bare LED with a concentrator, such as a reversed telescope eyepiece, mounted some distance in front of it. This could improve the quality of the "star", making it suitable for use with longer focal lengths at closer distances (with shorter exposures).

/Acme

Tree branches against a bright sky is always the best test for me.

How about wrapping them tree trunks with tin foil?
That should do it.

kosmoskatten wrote:
How about wrapping them tree trunks with tin foil?
That should do it.

Not far wrong. Silver glass Christmas bulbs hung from trees and hit by full sunlight make decent targets.

any light sourced behind an aperture refracts on the way through; i.e. the aperture itself introduces CA which is added to lens CA, thus light behind holes punched in foil (even by laser-cutting the hole) will not work for measuring lens CA.

star brightness is affected by atmosphere; we want a specific light-value. stars emit specular light; we want to use full-spectrum white light.

inexpensive, commonly available equipment, and a specific setup, is required to give commonly reproduceable results. here's my list; the "specifics" still need to be worked out:

a specific target with specific reflectance is necessary, i.e. the specific target is printed on a specific brand and version and size of photo paper at a specific size using a specific brand/version ink, at a specific dpi.

lens and target are oriented exactly perpendicular to line-of-sight, set a specific distance(s?) apart; specific distance(s?) is dependent on lens focal length.

in an otherwise dark room with non-reflective walls, floor, and ceiling, front-light the target from a standard position (specific x, y, and z) and distance with a specific brand/version of halogen bulb mounted in a specific socket in a pecific orientation; adjust brightness using rheostat to produce a standard EV as measured at the camera. (all this is perhaps the most difficult to set up)

photos are taken at various aperture/shutter speed according to measured EV using a camera with properly calibrated shutter speed, and set to a specific iso.

specific portions of images are examined for CA at a specific magnification, chosen to permit usage of most folks cameras, thereby eliminating differences in pixel density and sensor size.

I'm not sure we need to take this to such an extreme as wrapping trees in tin foil or buy HMI lamps and turn our living rooms into studios. Is it OK to take it back to a more realistic level?

Here is an example of a target which simply is a laser print on ordinary office paper, halogen lamps:


To the left an EF85/1.8 at f/5.6, to the right the Samyang 85/1.4 at f/5.6, left border (not really the border this example is from the border if the image is cropped to an aspect of ratio of 1:1.33 (or 4/3). I used the method mentioned in my first reply and the sample is taken from the series of images comparing the EF85/1.8 to the 85mm Samyang over here.

The sample shows 100% crops, the difference is clearly visible. At 200% any faults can easily be detected.

siriusdogstar wrote:
any light sourced behind an aperture refracts on the way through; i.e. the aperture itself introduces CA which is added to lens CA, thus light behind holes punched in foil (even by laser-cutting the hole) will not work for measuring lens CA.

It diffracts on its way through, but does not disperse. A smooth aperture is desirable to keep diffraction uniform, but not even a rough pinhole will suffer from CA.

/Acme

Jonas B wrote:
The sample shows 100% crops, the difference is clearly visible. At 200% any faults can easily be detected.

But remain extremely difficult to classify. Star testing allows much more sophisticated analysis:

olyacme wrote:
But remain extremely difficult to classify. Star testing allows much more sophisticated analysis:

Indeed. I guess there is a balance between what results "we" need and the setup. Thinking of Paul's original post:

cogitech wrote:
(...)
- Must be a readily available object, so that we can all use it.


... is it easy for anyone to setup and process the images to get the results you demonstrated? To my eyes it looks so advanced. Maybe it isn't?

Jonas B wrote:
... is it easy for anyone to setup and process the images to get the results you demonstrated? To my eyes it looks so advanced. Maybe it isn't?


The first image is idealized / simulated, the second from a bench test of a more demanding optic than we'll typically encounter. The second one does demonstrate some "defocus CA", though, so that's a little interesting.

Results produced by typical camera lenses won't look like the shown images, but we're lucky in that the focal lengths are shorter and the aberrations are mostly much greater. Basic tests can be performed with just a field, a flashlight, a tripod, and some attention to detail.

/Acme

olyacme wrote:
(...)
Basic tests can be performed with just a field, a flashlight, a tripod, and some attention to detail.

That sounds good. What is a suggested setup and what will we be able to learn from the results?

Jonas B wrote:
That sounds good. What is a suggested setup and what will we be able to learn from the results?

A quality pinhole over a decent LED flashlight will be good enough for wide angle lenses. Size its aperture and place it far enough away that it spans less than the diffraction limit of the lens under test. For infinity testing, this should also be far enough away that it amounts to infinity.

We can test (and to some degree quantify) astigmatism, spherical aberration, longitudinal CA, lateral CA, defocus colour, coma, field curvature, mechanical vignetting, and possibly flare. Some of these require examination of the "star" while defocused or off axis. Ideally the star should be placed at several points in the field and compared both inside and outside of focus, but focusing much beyond infinity is not generally an option for camera lenses. Of course a "star" that is big enough to be resolved as a disk by the lens in question will spoil the results, by muddying the data in exactly the same way non star tests do.

In a real sky and with narrow fields of view, the test is also a powerful way of gauging optical vignetting.

I think using flash is a good way to have standardised lighting, with specified intensity, distance to subject, and exposure setting.

Next we need an object that every one has access to. Paper/test chart is a good place to start but we don't have access to the exact same paper to ensure identical reflectance.

cogitech wrote:
I've been doing LoCA testing this way, thus far:

siriusdogstar wrote: ... the aperture itself introduces CA

olyacme wrote: ... not even a rough pinhole will suffer from CA.

scroll down to "Aberrations of a pinhole" here:

http://books.google.com/books?id=e5mC5TXlBw8C&lpg=PA218&ots=tve-lkHHEP&dq=pinhole%20chromatic%20aberration&pg=PA218#v=onepage&q=pinhole%20chromatic%20aberration&f=false

and search for "chromatic aberration" here:

http://www.wesjones.com/pinhole.htm

Paoletto wrote:

bet this is the apo lanthar, isnt it?

Yes.

I think Paul is just setting us up to buy his new lens testing rig.

gasrocks wrote:
I think Paul is just setting us up to buy his new lens testing rig.



Why let the cat out of the bag? I could have let you in on the deal. I have 200 pinholes on order from a Chinese manufacturer....!!! Already have a box of paperclips here, so I'm saving a bundle there. Got a source for cheap white rulers?

siriusdogstar wrote:
and search for "chromatic aberration" here:
http://www.wesjones.com/pinhole.htm

Learn something new every day, and of course this makes perfect sense. However, it's demonstrably moot for testing wide-to-normal camera lenses. This is a uninterpolated 400% crop:

Do you really need to be that far away from the pinhole? A really high resolution camera has got ~5µm pixel c-c distance, on a 50mm lens that gives that a 0.5mm hole (normal pinhead needle) only needs to be 0.5*50/0.005mm = 5m away before the pinhole is smaller than a single pixel in the camera...
A 0.5mm hole is easy to calculate minimum distance to, 28mm gives 2.8m, 35mm gives 3.5m, 85mm gives 8.5m and so on.

The comatization in the above example actually makes the Lateral CA much harder to judge, as Coma is often corrected very differently for the three different primary wavelengths. For lateral CA, the USAF target, or even any other high-contrast pure black/pure white edge will give more pictorially relevant information - without making things to hard on the more non-scientific target users who only want an overview of classic lenses' CA behaviour.

LoCA smudges the information available in a lateral CA pinhole target so much that you almost have to deconvolve the channels separately to get a good overview of what's going on.

theSuede wrote:
Do you really need to be that far away from the pinhole? A really high resolution camera has got ~5µm pixel c-c distance, on a 50mm lens that gives that a 0.5mm hole (normal pinhead needle) only needs to be 0.5*50/0.005mm = 5m away before the pinhole is smaller than a single pixel in the camera...

No, but the pinhole was bright enough, and I wanted to test at infinity, for best comparison with an actual star.

theSuede wrote:
The comatization in the above example actually makes the Lateral CA much harder to judge, as Coma is often corrected very differently for the three different primary wavelengths. For lateral CA, the USAF target, or even any other high-contrast pure black/pure white edge will give more pictorially relevant information - without making things to hard on the more non-scientific target users who only want an overview of classic lenses' CA behaviour.


Maybe, but I think most photographers can produce an image like that, and see how far the various aberrations extend, and how intense they are, and then make a good assessment on how they will affect their images. For example, I knew that this lens delivered good detail but low contrast wide open, and figured this was probably due to Coma. The star test confirms this speculation, and shows its extent.

theSuede wrote:
LoCA smudges the information available in a lateral CA pinhole target so much that you almost have to deconvolve the channels separately to get a good overview of what's going on.

The images shown were chosen to compare a pinhole star with a real star, rather than look at any particular aberration, so you're right to call them a smudge. They're greatly overexposed, for one thing, to bring up the Coma. I did take additional images at f/16 in order to look at Lateral CA more or less by itself, and out of focus shots to look at SA. On-axis shots could be used to inspect Longitudinal CA in better isolation.