Trying to diagnose AF issue. Is my test reasonable?

I'm having suspicions that my 400D is still front focusing (despite having been calibrated just a few weeks ago), and think it might be a good idea to do something about this before the warranty expires in 12 days! To that end, I have acquired a nice brand new EF 50mm f/1.4 USM lens. A typical result is shown in the figure below: front focusing by about 2cm. Note that the picture is a 100% crop to save space but preserve detail. I did the same with a Tamron 17-50mm f/2.8. At 50mm, it also shows a front-focus bias of about 2cm, which gets progressively worse as I get to 17mm (I think by as much as 30cm in this rig).

The test was set up as follows: camera was set to centre-point AF only, aperture f/1.4, mounted horizontally on a tripod about 1m from the subject. Subject was a lens box (sure you've figure that one out!) illuminated by a 20W halogen lamp about 30cm away. I trained the AF point on the big black letters, as the red square indicates. I put a piece of paper with a grid on it next to it; the line with the blue marks is in line with the front of the box. Picture was taken at 1/250s at ISO 200.

Now, before I go back to the shop for another two week calibration vacation, can anyone think why this test might be flawed? Surely the camera should be able to focus on something like this...

HOSTED PHOTO
please log in to view hosted photos in full size.

I'd say your test is a better one than the usual shot of a target tilted at 45 degrees, because the camera has a better target to focus on.

But 20 watts isn't very bright, even 30 cm away. I'd try it with 100 or 200 watts, or in daylight. Autofocus doesn't work at its best in low light.

One more thing: Canon is reported to test focus at 50 X focal length. That'd be 2.5 meters. Were you closer than that?


-js

I'm not sure a 50 f/1.4 wide open is the best test lens either. Softness from the lens can skew your results. I think you'd be better off stopping down to about 2.8.

Occasionally it hits the mark and then it's sharp as a knife... I'll try and get the camera as far from the rig as possible and see how it fares then, but I don't think focusing at 1m is unreasonable. (number six: it's a 20W halogen bulb --- I wouldn't want to have a 200W halogen 30cm away from anything!)

On the face of it, your test looks valid to me, but I have some somewhat random, disconnected points/suggestions/com​ments/questions.

Is this reproducible - does it miss by the same amount every time, time after time?

Have you tried having the system find focus starting from both too far, and too near?

What happens if you are further from the target? Canon calibrates based on 50 x focal length distance to subject. I'm still uncertain whether errors are constant or proportional to distance, or independent of it and variable. To echo the test Canon would do, you should be shooting at 2.5m.

There is a perennial problem with this situation, and that is what to use as a known good reference. You've tried to address this by getting the 50/1.4, but there is actually no guarantee that both it and the Tamron aren't both off by the same amount, and the body is actually fine!

When I was faced with this (a lens issue actually) I got my 30D calibrated and that became my reference -- by that logic it is your lenses that are off.

You can also send the body and the 50/1.4 to Canon together and get them calibrated to work with each other.

xarqi wrote in post #4711323
Is this reproducible - does it miss by the same amount every time, time after time?

Pretty much. Furthermore, the amount of error seems to depend on the camera's orientation in some cases.

xarqi wrote in post #4711323
Have you tried having the system find focus starting from both too far, and too near?

Yes, I when it mis-focuses, it tends to do so by the same amount each time.

xarqi wrote in post #4711323
What happens if you are further from the target? Canon calibrates based on 50 x focal length distance to subject. I'm still uncertain whether errors are constant or proportional to distance, or independent of it and variable. To echo the test Canon would do, you should be shooting at 2.5m.

I've moved the camera to just over 2m away from the rig. Of course, now the DoF has opened up and it's harder to tell, but I can still see a FF bias.

xarqi wrote in post #4711323
You've tried to address this by getting the 50/1.4, but there is actually no guarantee that both it and the Tamron aren't both off by the same amount, and the body is actually fine!

I've just tried it with a 50/1.8 at f/2.2. Unfortunately, this thing front-focuses from the maximum focal distance, and back-focuses from the minimum so I can't conclude anything from that.

ettlz wrote in post #4711460
Pretty much. Furthermore, the amount of error seems to depend on the camera's orientation in some cases.

Hmmm. cross-type AF sensor sensitivity? More light might help???

Well, that's helpful (sort-of). Consistent, reproducible problems are a lot easier to track down.

That could be part of the issue. The system is considered by Canon to be in spec if the target falls within the DOF at 50xFL. It just could be that everything is fine according to Canon, but that your requirements exceed that. Not much to do in that case except send both the camera and lens and ask nicely.

I've just tried it with a 50/1.8 at f/2.2. Unfortunately, this thing front-focuses from the maximum focal distance, and back-focuses from the minimum so I can't conclude anything from that.

Another can of worms opens.

ettlz wrote in post #4711321
(number six: it's a 20W halogen bulb --- I wouldn't want to have a 200W halogen 30cm away from anything!)

Still, 20 watts isn't really bright light. Here's what I use for this kind of thing:

HOSTED PHOTO
please log in to view hosted photos in full size.

HOSTED PHOTO
please log in to view hosted photos in full size.

The results of your test depend on how precisely you can get the target parallel to the film plane. Also, it would be much better to have more and smaller text on your grid in order to better judge exactly where the focus is falling. Frankly, I think the standard 45 degree test is just fine -- the author claims he did extensive tests using a complicated setup like yours and a simple chart at 45 degrees and found no differences in the results. The results are easier to interpret too.

I'd recommend shooting with a wide open aperture and with as many different lenses as you can (ignoring the kit lens).

As for Canon's camera recalibrations, I know for a fact that they don't always get it right the first time. After the first recalibration of my 350D, it auto-focused correctly but ONLY in VERY bright light (e.g., the flash AF assist beam was NOT bright enough). Additionally, the repair/recalibration is covered by a new, separate warranty so you don't have to worry about your camera's warranty expiring. I was in the same position with my camera's warranty expiring and was told that this was not a problem.

Thing is, I've had another copy of the Tamron lens before this one which also front-focused and that prompted me to send the body in for repair. I'm not sure what the probability is of three lenses from two manufacturers all front-focusing by roughly the same amount at the same focal length with a particular body and it being the lenses at fault, but I bet it's pretty small. I did shoot all the pictures with the lenses as wide open as I could get them. The Nifty's AF accuracy is a joke at best, so I didn't include that, the Tamron 55-200 didn't seem appropriate, and nor did I bother with the 18-55mm kit lens as its DoF is too wide to show up any problem. I don't think that the light was too dim in my test (1/250s at f/1.4, ISO 100) and I would think in this case I'd see both back and front focus; the results show a consistent FF bias.

Anyway, I'm now in something of a quandary as how best to get this stuff to Canon UK. I bought the body at a department store and the 50/1.4 lens elsewhere. Might have to send them separately with some bleedin' obvious notes attached to the effect that the two are meant to be treated together.

ettlz wrote in post #4713563
Thing is, I've had another copy of the Tamron lens before this one which also front-focused and that prompted me to send the body in for repair. I'm not sure what the probability is of three lenses from two manufacturers all front-focusing by roughly the same amount at the same focal length with a particular body and it being the lenses at fault, but I bet it's pretty small.

Oddly enough, that Tamron lens is the same one that started me down the path you're following right now. Three copies across three different Camera recalibrations and I had to give up on the Tamron completely. After all was said and done, the camera turned out to still be miscalibrated, however, the Tamron seemed to be particularly sensitive to the the camera's calibration state (more so than the 50mm f1.8).

I'd still run the 50mm and 55-200 (at 55mm) just to double check your results. In my case, 11 of 12 lenses of varying quality showed some degree of front-focusing. Only the kit lens never showed a problem.

You might call Canon UK to verify this but the usual procedure in the US is to send camera and lens together asking them to recheck the camera's focus calibration. They will actually check both and run some tests with the lens+camera together.

Best of luck. I also highly recommend having a cheap EOS film backup camera handy to help keep soul and sanity in one piece during the process.

JackProton wrote in post #4713673
I'd still run the 50mm and 55-200 (at 55mm) just to double check your results. In my case, 11 of 12 lenses of varying quality showed some degree of front-focusing. Only the kit lens never showed a problem.

I think I'll try that tonight, but just with the 55mm end of the Tamron. My Nifty's focusing ring has more in common with a roulette wheel than an AF mechanism...

[I've been considering a chemical EOS... do you have any recommendations for a cheap model?]

OK, I've just done the test with the Tamron 55-200mm at 55mm, f/4. Again, front focus by roughly 1.5cm (see below). This has to be a body-related issue (AF sensitivity vs. available light?). Back to the shop on Saturday!

HOSTED PHOTO
please log in to view hosted photos in full size.

ettlz wrote in post #4723595
OK, I've just done the test with the Tamron 55-200mm at 55mm, f/4. Again, front focus by roughly 1.5cm (see below). This has to be a body-related issue (AF sensitivity vs. available light?). Back to the shop on Saturday!

Maybe its just that I'm used to it but I think the chart at
http://www.focustestch​art.com/chart.html provides a clearer idea of how bad a focusing issue you have.

But, yeah, sounds like a body issue. Consider sending Canon a lens to use with the body during their tests. You might also include a CD with the images from your focusing tests as well as some real-world shots that illustrate the problem. (And, for god's sake, NEVER EVER mention to them that you OWN a Tamron!)

JackProton wrote in post #4724002
(And, for god's sake, NEVER EVER mention to them that you OWN a Tamron!)

No, I just borrowed it... from someone who looks like me... has the same name... lives in the same house... Thanks, I'll have to remember that when writing a note to include with the camera detailing the problem...