Microadjustment

[aldvan]

I just got back from Canon Service my 1Ds Mk3 for a focal plane adjustment.
After that, as I use to do, I made a microadjustment of all my lenses.
During the classical 'moiré' procedure, I needed adjustment in a +3/+8 range.
For the first time, after the MA, I verified pedantically in the real world the results and I found that the microadjusted focus is ok in case of Automatic AF selection, but, if I select the AF central point, that is my usual modus operandi, the microadjusted focus is slightly worse than the plain one.
Any interpretation?
TIA
Aldo

 

[Loswr]

Did you use the central AF point for the microadjustment calibration? How did you align your moiré-producing pattern to the camera (orthogonal alignment is important for an accurate calibration)?

 

[Mt Spokane Photography]

The moire pattern has not been reliable for me, so I've stopped using it. Its difficult to intrepret. I use the center point to micro adjust. If the other points are then significantly off, then the body may need some repair.

There will always be some inaccuracy in both the body and the lenses, nothing is perfect. Its just when it becomes a significant issue that service is needed.

 

[Loswr]

FWIW, I don't find the classical 'moiré' procedure to be all that accurate. I find that I can move the camera back and forth a fair distance without seeing any observable difference in the moiré pattern on the LCD. Personally, I use a LensAlign Pro for my calibrations.

 

[Viggo]

I also used to use that. But for my newly bought (second copy, missed it too much) 85 L II, I didn't do the moiré thing. I just shot at a distance quite far from mfd, something within 15-30 meters, and just tried several times, and adjusted (shot wide open). This method of real life examples straight away worked much better then I would have thought, but it makes sense. Just find something that's easy to focus on and take shots at 0, +10 and -10 to see where it get's way worse, then half your way through it.

 

[Mt Spokane Photography]

I also used to use that. But for my newly bought (second copy, missed it too much) 85 L II, I didn't do the moiré thing. I just shot at a distance quite far from mfd, something within 15-30 meters, and just tried several times, and adjusted (shot wide open). This method of real life examples straight away worked much better then I would have thought, but it makes sense. Just find something that's easy to focus on and take shots at 0, +10 and -10 to see where it get's way worse, then half your way through it.

The problem with shooting at a distance is that the depth of field increases, so you can't tell if the lens is front or back focusing. I keep a reasonably short distance to the target, depending on focal length so that I have a shallow depth of field. A shallow depth of field lets one set MA very accurately, I can usually discern a difference of +/- 1.

 

[aldvan]

@neuroanatomist: yes, I used the central AF point for MA, since this is my usual focusing system.
@viggo: after using the moiré system, I started just with the method you proposed, finding the best result at 0 adjustment...

 

[Loswr]

Normally, you should test at 25-50x the focal length. So the 85L, for example, should be tested at 2.125-4.25 m, not 15-30 m, where as Mt. Spokane pointed out, the greater distance makes DoF more than deep enough to mask focus errors (again with the 85L as example, 85mm f/1.2 at 30 m means a DoF of nearly 10 m deep!!).

 

[Mt Spokane Photography]

@neuroanatomist: yes, I used the central AF point for MA, since this is my usual focusing system.
@viggo: after using the moiré system, I started just with the method you proposed, finding the best result at 0 adjustment...

At 30 meters, almost everything will be in focus due to the large depth of field. Be sure to check at a close distance, say 2 meters wide open and see if its still accurate.

 

[Gothmoth]

Be sure to check at a close distance, say 2 meters wide open and see if its still accurate

as usual such general statements are wrong.
i would not test my 300mm or 400mm lenses at 2m.

i don´t even care if i have a back or front focus at such close distance with a 400mm tele.

 

[Loswr]

Be sure to check at a close distance, say 2 meters wide open and see if its still accurate

as usual such general statements are wrong.
i would not test my 300mm or 400mm lenses at 2m.

I would assume that Mt. Spokane was referring to the example raised by Viggo, namely the 85L. For that lens, a distance of 2 m is ok (85mm x 25 = 2.125 m).

Your 400mm tele should be tested at ~10 m. The distance used is really a compromise - for some lenses which exhibit a mild focus shift, you might get a different result testing at the minimum focus distance (MFD) vs. testing at a long distance (assuming you had a method that could accurately test with the deep DoF at a long distance). But, AF errors will be most evident at the MFD, where the DoF is shallowest.

 

[Mt Spokane Photography]

Be sure to check at a close distance, say 2 meters wide open and see if its still accurate

as usual such general statements are wrong.
i would not test my 300mm or 400mm lenses at 2m.

i don´t even care if i have a back or front focus at such close distance with a 400mm tele.

We are talking his 85mm f/1.2 lens, not a 400mm lens, and advice given for a 85mm lens is not applicable to a 400mm lens.

 

[Viggo]

Mt Spokane:

I assume you have never tried a 85 L II? I shot across the yard outside, and aimed at a cars license plate a,d it was only sharp focus on the plate, and widway up the hood and maybe 0,5 meters infront of the car, that is much easier to determine wheter or not if the focus is hitting straight than at MFD. Because at mfd the veeery tiny back or front focus won't show, but it makes a HUGE difference when you put in some distance, trust me. I had a 300mm f2,8 L IS that was bang on at mfd and 40 meters off at 200 meters. And that is very common with lenses.

The trick is to shoot at such a distance that let's you see where the focus really hits, and not so long you go past hyperfocal and everything is in focus as you say.

Besides, focusing errors at 90cm at 1,2 is MUCH more common, and it becomes very difficult to see if the lens misses or if the focus misses in camera. I've done this with perfect results for as long as microadjustment have existed, so please try before you bluntly assume something.

 

[Loswr]

Mt Spokane:

I assume you have never tried a 85 L II?

Sounds like a blunt assumption to me...

I shot across the yard outside, and aimed at a cars license plate a,d it was only sharp focus on the plate, and widway up the hood and maybe 0,5 meters infront of the car, that is much easier to determine wheter or not if the focus is hitting straight than at MFD. Because at mfd the veeery tiny back or front focus won't show, but it makes a HUGE difference when you put in some distance, trust me. I had a 300mm f2,8 L IS that was bang on at mfd and 40 meters off at 200 meters. And that is very common with lenses.

The trick is to shoot at such a distance that let's you see where the focus really hits, and not so long you go past hyperfocal and everything is in focus as you say.

Ok, so your subject is sized proportionally to your DoF, so that can work. It all depends on how you test. With a calibration tool such as the LensAlign, resolving that 'veeery tiny back or front focus' is quite easy to do, because unlike a car parked across your yard, that tool is designed to do just that.

Here's what it looks like in use, and you can see that any front or back focus would be immediately obvious - the thinner the DoF, the better.

In fact, 2 m on an 85L is no where near the MFD (which is actually slightly less than 1 m). IMO, using a tool designed for the job is better, because it eliminates some of the variables that could otherwise cause problems. Are you orthogonal to the license plate? Is the plate itself providing adequate contrast? Is the AF system really even focusing on the plate? That last question is important, because at the distances I believe you're talking about, a standard sized license plate will cover only a small fraction of the AF point on the sensor (which is a larger area than suggested by the little box in the viewfinder) - so, you may intend to focus on the license plate, but your camera's AF system doesn't know that, and will just grab the highest contrast feature at the proper orientation within the sensitive area, which might be the bumper...or the shadow of the bumper on the ground several cm in front of the car.

Probably most importantly, when you shoot with the 85L, are most of your shots taken from a distance similar to 'across the yard'? Personally, most of my shots with my 85L are portraits taken at a distance of 2-3 m. So, having spot-on AF at a distance of 20 m, even if that is achieved, is not very helpful if it's not applicable at the distances where I most often use the lens...

 

[Viggo]

Wow, this is beginning to become a discussion I'm not going to bother bickering about. First off, When you use 1dmkIV the focusing point is so large that it covers at least too much of those fine lines on a chart to be even remotley accurate. Second of all, it must be at a larger distance to see what happens. I have countless times been really upset when I have fiddled around with charts and moiré patterins at 1m or 2m and getting it PERFECT, then turn around and shot a black and white super-high contrast road sign outside, perfectly level and everything (I'm not a noob at this) and it's waaay off, I keep shooting that sign and also the bottom of the sign as that is much easier to see where the dof hit the asphalt, and clearly see the dof moving back and forth, then I just leave it where it hits sharp (which is usually a whole other MA setting then before) and go back to the moiré pattern, or test chart and find that it hits EXACTLY the same spot as the first time, only now it's a very different MA adjustement, and what do you know, now it works at a distance too. Adjusting at mfd will never be accurate enough. At Northernlight they suggest you use at LEAST 50 times the focal to adjust.

So whatever works for you, fine! But do not tell me that my method doesn't work, and for me it works much better, because aiming in between very long and very short distance to subject I get sharp POF on ALL distances. I did so with these lenses: ef 14L II, 24 1,4LII, 50L, 85L (twice) 70-200 II, 300 f2,8 L IS. That's the way it was and always will be.

 

[aldvan]

Please forgive me for starting a discussion that is becoming quite harsh...
By the way, although the idea to microadjust by a simple tool intrigued me at first, I'm convincing my self that the procedure suggested by Canon itself should be closer to the needs of the real world.
Rather, there are some points that I overlooked and that are very important.
The first one, as others already pointed in their post, is the perfect parallelism between camera and target (and this could be a point in favor of Lensalign).
The second one, that I forgot to do until now, is TO TURN OFF the image stabilizer...
I'll repeat the procedure next week end...

 

[Viggo]

You should also consider this one. I tried it and it made sense of the adjustments I already had done. (tried one we had in our camera-store were I work). It's ridicolously expensive so might want to pitch in with a few buddies ;oD

http://spyder.datacolor.com/product-cb-spyderlenscal.php

 

[Loswr]

So whatever works for you, fine! But do not tell me that my method doesn't work, and for me it works much better, because aiming in between very long and very short distance to subject I get sharp POF on ALL distances. I did so with these lenses: ef 14L II, 24 1,4LII, 50L, 85L (twice) 70-200 II, 300 f2,8 L IS. That's the way it was and always will be.

Agreed. I've gotten excellent results with the LensAlign Pro and a much longer list of lenses than yours, and it works for me. If your method works for you, keep doing it!

I won't quibble, but I will correct a few inaccuracies:

Adjusting at mfd will never be accurate enough. At Northernlight they suggest you use at LEAST 50 times the focal to adjust.

Once again, not MFD. The usual recommendation is 25-50x the focal length of the lens - frequently qualified by the advice to test at the most relevant/common subject distance for your typical subject(s). Granted, that's not always easy because most people shoot more than one type of subject with a given lens.

I have fiddled around with charts and moiré patterins at 1m or 2m

The problem there is not the distance, it's the other part of the statement - fiddling around with charts and moiré patterns. Printed charts don't cut it, and moiré patterns have too much play (little to no discernable difference with a pretty large degree of axial movement) and problems with alignment. You need to use the right tool for the job. If your auto mechanic 'fiddled around' with a crescent wrench when working on your car's powertrain, instead of properly tightening the bolts with a torque wrench, you'd have maintenance issues down the line.

Your subsequent statement about the SpyderLensCal, "I tried it and it made sense of the adjustments I already had done," is totally consistent with what I'm saying here - when you use the right tool, you get accurate results.

First off, When you use 1dmkIV the focusing point is so large that it covers at least too much of those fine lines on a chart to be even remotley accurate.

With this and other, similar tools, you're not focusing on the ruler with the DoF scale - the ruler is at an angle to the vertical so the scale has depth, and your focus target is a larger square with a contrast scheme optimized to activate any AF sensor point geometry (horizontal- or vertical-line, cross-type, or diagonal cross-type) that is aligned to be parallel to the image plane (and thus orthogonal to the AF sensor).

The SpyderLensCal you linked to is constructed in the same way (but it seems to lack a feature to facilitate parallel alignment of the camera with the focusing target.

Rather, there are some points that I overlooked and that are very important....I'll repeat the procedure next week end...

One point that has not come up, but which is quite important, is what autofocus microadjustment really does. What AFMA does is correct for systematic error in the AF system, specific to manufacturing tolerances in cameras and lenses. It corrects for bias in the AF system, or put another way, it improves the accuracy of the AF system. The terms accuracy and precision are sometimes (improperly) used interchangeably - accuracy is 'closeness to true' whereas precision is repeatability. Here's a diagrammatic example:

AFMA corrects for accuracy but does nothing for precision. Chuck Westfall has told me that, "The AF precision for the standard precision sensors is within the depth of focus for the maximum aperture of the lens, while the AF precision for the high precision sensors is within 1/2 or 1/3 the depth of focus for the maximum aperture of the lens, depending on the camera model under discussion." Note that depth of focus is measured at the image plane of the sensor, and is different from (but affected by the same parameters as) the depth of field. I interpret that to mean that although the precision is specified, for example with the f/2.8 sensor all the shots will be focused within a region that is 1/3 of the depth of focus deep for that lens' maximum aperture, focusing could still be inaccurate (i.e. it would look like case C above).

AFMA moves the center point - the 'average' focal plane of multiple measurements. But it's important to remember that precision plays a role, so if you test your AF system with just one or a couple of tries, random chance says your test won't be 'spot on' due to imprecision. So, when testing AF performance, you need to perform multiple tests for a given lens, and fewer tests would be needed for an f/2.8 or faster lens due to the higher precision. I suspect this is what's behind Canon's statement that AFMA can prevent correct focusing - if you base an adjustment on one or two test shots, and those shots were relatively far off the correct focal plane, then on average most of your subsequent shots will miss focus.

Bottom line - no camera will produce a perfectly focused image every time at every distance, period. No adjustment procedure will change that. What a proper AFMA can do is increase the percentage shots that are correctly focused.

 

[Loswr]

As a side note, some time back epsiloneri posted a very nice method for quantitatively determining the best AFMA setting. It's something I plan to try when I get a chunk of free time.

 

[Viggo]

Well, fair enough, it seems that we actually agree on most things, and the rest I'm not going to discuss further.

The image you posted of the chart didn't include the whole thing (similiar to the spydercal) which was misleading at best. It look liked a printed chart. These tools seem to be the best solution. But I have found that when going shooting real life subjects, I still need to make adjustments, which kind of is crazy considering how much these tools cost.

The Spydercal has a level on it so it's simple to keep that and the camera parallell.