What distance for lens test?

[Arjen van Andel]

I had my equipment calibrated twice in the last 6 months. The last time the technician made a lot more modifications than the first and he described how they test for focus. He claims an angle of 15 degrees and a distance of 80 times the focal length (for 1.6x crop, 50 times for full frame) and a 3D target.

In the picture below (200%) I've mimicked their test with the 17-55 f/2.8. But this leads to such vague images that, for the life of me, I cannot distinguish anything useful from these shots.

Canon 17-55 f/2.8 Focus test example, 15 degrees down at approx. 440 cm. Top: manual focus with Anglefinder-C and Af from Tele, Bottom: AF from Wide and last and AF from previous AF from Wide test.

For a 300mm lens the testing distance would be 24 meters. (!) I haven't tested the 70-300 in this way yet, but real-live shots are several meters off at 50 meters. That's useless, in my book.

Any thoughts?

Arjen.

 

[jerryk]

That is what you should test at IMHO.

Also, the camera should be on a tripod, the lighting conditions should be good, MLU used, and subjects stationary to ensure you eliminate as many variables as possible.

--
jerryk.smugmug.com

 

[Arjen van Andel]

That is what you should test at IMHO.

Also, the camera should be on a tripod, the lighting conditions
should be good, MLU used, and subjects stationary to ensure you
eliminate as many variables as possible.

Hi jerry,

I know the drill, tripod, MLU, light, etc., but what wonders me is that it's about impossible to see if a lens in in focus or not at these test distances (80x focal length for 1.6x crop). I sofar tested three lenses on 80x distances and none had readable results.

What test chart is Canon using that they can see what's what? I've tried an improvised ruler with 1 cm wide markings but those were still not big enough; they each wound up as a blurry line a few pixels wide, no clear difference in blurryness over a 40 or 80 cm track.

If I use a Circle of Confusion of 1/100th mm than the DOF is about 36 cm at the 80x test distance of 440 cm (focal length 55mm). The DOF behind the focus point is 17 cm in this case. I just cannot make out if the focus point is placed correctly or ends correctly or whatever.

My initial tests were done with Tim Jacksons test chart and from much nearer and I can actually see the difference in performance. DOF is much steeper there, of course.

Perhaps I need a longer test track. But that means still smaller detail at the far end.

Arjen.

 

[GoldenSpark]

I think to do it this way you need to have very crisp clear edges, but also large.

In other words you need a high contrast white/black edge where both the white and black are quite a bit wider than many pixels.

As it happens I have been playing around trying to understand MTF charts, and trying to derive MTF from test shots.

What I did was look at shots of a very dark black square about three inches on a side.

If you zoom in on a vertical edge, you can see the edge spread over a few pixels. The luminance value changes from bright to dark according to how well the edge is defined. This is blurred as a combination of focus and inherent sharpness effects.

You should be able to tell by looking at an edge that is in focus compared to one OOF by how many pixels it spreads across.

I think the problem with your test is that the things you are looking at (e.g small text) have low contrast anyway, even if in focus the black / white bands are not wide enough to discern.

 

[minordetail]

Some quick math in my head tells me that using the suggested calibration distance you would be over ten feet away with a 50mm lens ???

The DOF at that distance would cover up any focus issues that the lens may have .. I test just beyond the minimum focus distance and pretty much wide open .. tha reveals any focus problems for sure.

--

They call me 'minor detail' .. I keep the record straight. I have huge man hands and can comfortably crush a Rebel XT between my thumb & forefinger.

 

[JJMacks]

http://www.canon-dslr.com/Canon_Jan05/Canon_SLR_Focus_Test.htm
http://www.focustestchart.com/focus21.pdf

--
JJMack

 

[tmr131464]

40 to 50 times focal length is commonly recommended. This distance is just about as far as you can get with a 50 f/2.8 lens and even then you'll need a good long scale (DOF is about 20 cm at 40x). I've read that doing the test too close can mask some problems. I always do my tests at a variety of distances and especially, as another posted, always at about the range you would commonly use the lens.

 

[Arjen van Andel]

Some quick math in my head tells me that using the suggested
calibration distance you would be over ten feet away with a 50mm
lens ???

440cm or approx. 14 feet.

The DOF at that distance would cover up any focus issues that the
lens may have .. I test just beyond the minimum focus distance and
pretty much wide open .. tha reveals any focus problems for sure.

That's pretty much what I did and wrote to Canon, but Canon's tech wrote me a note stating they use an angle of 15 degrees at 80 times the focal length (for 1.6x, or 50 times for FF).

My own tests were done from 80 cm with the 17-55mm at tele and at an angle of 45 degrees on a modified Tim Jackson test chart.

Arjen.

 

[Arjen van Andel]

http://www.canon-dslr.com/Canon_Jan05/Canon_SLR_Focus_Test.htm

This test uses 50 times the focal length for the test distance and I can clearly see any errors. Canon claims 80 times focal length for their 1.6x crop bodies, that's a 60% increase in distance, perhaps enough to make this test distance useless (too little pixels covering any target detail). I'll try 50x when I have time, and perhaps I should try 80x with a white paper target instead of a glossy tape measure.

http://www.focustestchart.com/focus21.pdf

This is indeed the chart I used to show Canon the performance of 6 of my lenses. I modified the target in an attempt to minimize error and limit any number of confused viewers, and it mimics the test above, but at approx. 1/5th the distance. I used various patterns as the target but found no discernable difference in performance to the plain line. I also did the tests outside, not the best choice as the light isn't very reliable. But I didn't want to introduce any focus errors (if any) by the fluorescent lighting in my apartment.

Arjen.

 

[The Wanderer]

My jaundiced take on Canon's recommended distances is that quality control is now so bad, and it is so much easier with a digital camera to check for front/back focus than with a film camera, that they recommend you are in the next field and can scarcely see the test target anyway!

Wanderer

 

[abqnmusa]

Forget rulers, brick walls, $20 bills, etc.

Just take photos of real world subjects and see if they are in focus.

--

Canon EOS 5D / XTI -- If you are in Albuquerque NM area & want to take pics contact me

 

[BryanP]

I agree.

I have taken focus test shots on the Tim Jackson Focus Test Chart with three different lenses, from their wide open aperture down to f/5.6, and at the major increments of the lens' focal lengths. Some shots of one lens are back focused, then dead on, then front focused. I got alot of mixed results within one lens. In most cases, I took about 12 different shots with one lens, with the different focal lengths and f-stops.

So instead, I tried a real life test. I took my kids' rectangular shaped juice boxes and stacked them side by side in a stair stepped fashion. I chose this object because I needed something that would show up wider than the center focus box in the EVF. First I tried crayons side by side standing vertically, but then I found that two or three crayons fit inside the center focus box. So that was no good. The first juice box was set 30" from the camera, the next one beside it was 31" from the camera, the next was 32", the next 33" and the 5th one was 34" away. I focused on the center juice box (32" away) and shot all major focal length imcrements labeled on the lens and this times 3 different aperture settings. I got more consistent results with this real world test than shooting the test chart.

The problem I see with the test chart is that you set it on table with the camera aimed at 45 degrees. Who takes pictures of objects in a horizontal plane at 45 degrees? Not me. 99% of my pictures are of vertical objects (people, buildings, landscape, vehicles) where my camera is usually on a 0 degree horizontal plane or within about 15 degrees up or down.

From now on, all of my future lens focus tests will be real world type compositions, not a test chart on a horizontal surface.

Someone please tell me if my testing method is flawed and if my reasons against the test chart are also flawed. Thanks. )
--
Cheers,

Bryan P.

http://www.flickr.com/photos/29386469@N00/

 

[John Rain]

I have also been getting extremely variable results shooting $20 bills. Within a series of images taking right after each other with identical conditions, some were in focus and others weren't.

I'll try some stacked 3D objects one of these days. The only comment I'd have is that the distance at which these tests are performed probably matters too. It's one thing at the minimum focus distance but probably different at real-life distances, e.g. 5, 10, 20 or 30 meters. Question is of course to find a good test subject with sufficient detail for the latter.

JR

--

 

[Erez_Katz]

In Sigma's case, the technician's book says to test a lens at 20x the focal length, so a 30mm lens should be test at 60cm. The target is a standard lens test ruler at 45 degrees, and the difference of the achieved focus point from "0" is used to offset/calibrate the lens using software.
The camera should be at the exact height of the '0' target.

the one time i had to "cook up" a focus test without the latter setup, I placed a memory card at 45 degress (tilted against a battery) on asfalt, and tried to shoot it by pointing the camera straight at it (meaning camera was pointed at 45deg to the ground). I used the blur of the asphalt to determine where the actuall focus point was, and I was right, the tested lens front focused while a lense I knew for sure was ok was spot on... not the most scientific thing though...

HTH

Erez

--
'I hope this letter finds you'
Captian Spaulding
http://www.pbase.com/katzer

 

[BryanP]

Thanks for the compliments.

Very true about the distance you test for. I see no point in testing at the minimum shooting distance unless that's how you shoot that particular lens. I test lenses with 17mm to 55mm focal lengths at + - 4 feet and also around 7-8 feet. Tele zooms are a different story.
--
Cheers,

Bryan P.

http://www.flickr.com/photos/29386469@N00/

 

[Arjen van Andel]

Forget rulers, brick walls, $20 bills, etc.

Just take photos of real world subjects and see if they are in focus.

That's what I did, but I had to throw out way too many shots. So I went for the less-than-real-world-but-not-moving-around-so-much focus charts to see if the error is consistent (or perhaps my doing), and give the beasts a (millimetre) name so I have something to say to a technician.

All lenses showed reasonably consistent offsets on the chart. Real world images differ much more and in both directions and are hard to quantify.

I still think that the body has something wrong with it because the errors are quite severe and erratic (front and backfocus). I own a Sigma 30 f/1.4 and focus on near subjects can be off by a meter and more in front or behind the subject. Sigma tested this lens on their 30D and 400D bodies and found no such behaviour.

Arjen.

 

[GoldenSpark]

I have no doubt that the test chart (Tim Jackson's) works fine. Some lenses are a bit erratic on AF anyway, but even these show whether they are front or back focussing.

On comparing to real-world subjects, I have done loads of testing both ways, but more importantly have tweaked my 350D and observed the effect on both test chart and real-world. There is very good correlation.

Testing close to a test chart with a setup that shows front focus also shows front focus at distance and with real-world subjects, even to the point that with tripod or not, MLU or not, the results are the same.

The big problem with normal real-world shots is that often both the focus point and the focus indicators (parts in the subject tat show what is in focus and what is not) are so variable. Sometimes the focus point is not a high contrats plane surface that covers the whole of the focus sensor, leading to errors. Then it is unusual to have the equivalent of the 45 degree graduated scale to see where the real focus is.

 

[JJMacks]

Forget rulers, brick walls, $20 bills, etc.

Just take photos of real world subjects and see if they are in focus.

I will take test pictures to check out a lens's focus and sharpness. These will be of real world subjects that have dimensional depth and not be too far away I want detail resolution not distant objects. I take these in good light for a good fast shutter speed. Mostly shoot wide open and take both Center point AF and auto AF point shots . I like to use Canons EOS Viewer Utility to check the focus for it can display which AF points the Camera thought should be in focus with it's AF point overlay option. So far the four Canon lenses I have purchased seem to work well on both my 1D2 and 20D. My two f2.8 zooms lenses focus quit well though-out their focal lengths ranges and produce reasonable sharp images wide open within the in-focus range. My macro lens is very sharp and it's focus is right on with good light macro manual focus is doable with its f3.5 speed. However these lenses fall far short on my best lens which continues to amaze me. The Canon 300mm F2.8 IS lens is simple stunning it is sharper wide open at F2.8 then my other lenses are stopped down some. You would think a lens that is as big and heavy as the 300mm F2.8 would not have the fastest Auto-focus speed. It the fastest Focusing lens I own and it has a focus memory you can set to focus the lens to a set spot with the touch of a special focus ring. Additionally this lens a buttons on it when any one of the is press will perform the function programed in with the custom function. I have it set to switch the AF mode I using. If I have set the Camera to AI servo focusing mode Pressing and holding one of these buttons puts me in One shot mode. Releasing it I mack in AI Servo. This is very handy when you using a tripod with a gimbal head when your subject lites down and stop you can switch to one shot mode and get you Auto focus conformation without removing you eye from viewer in an instant.

--
JJMack