Anybody tried this? AF / Focus checking

[ajohnw]

I came across this while looking around for something else. I haven't the things I would need to try it out properly with me but hand held plus a lap top screen it does look promising and much simpler than other methods


http://www.northlight-images.co.uk/a...djustment.html

The actual article is some way down the page.

Having a positive attitude - it will be ok - I would AF onto the target and then see how good it looks in live view. Then maybe see how much the AF needs to be shifted manually to correct.

John
-

[DanK]

yes, I tried it with my old 50D and three lenses. I never got past the initial checks because none seemed to need adjustment.

[Black Pearl]

Saw that a few years ago and had a play - might have helped that I worked in a camera store at the time, had the space, the screens, the time (we weren't always busy) and more gear to fiddle with than is healthy.
I think from memory my AF-S 70-300mm VR was a bit out....though I may have went through a few copies until I found a perfect one....instead of tweaking the system I owned until my original one was tuned in.

[Downrigger]

I have done this with my Nikon D600 for each of my lenses, but used a standard target (Siemens star focus chart) and looked at final images in a sequence of with graduated corrections - multiple shots at each correction, as the autofocus is not perfectly consistent). In each case, the optimum microadjustment was several points off "0".

I doubt it makes much difference in the field, but gives me peace of mind that any softness is my fault, or autofocus mechanism error, not the result of imperfect mating of the lens with camera.

[NorthernFocus]

I've seen this method but never actually done it. In general I kind of poo pooed focus tuning until I bought a brand new 500mm prime lens and couldn't get a sharp image with it. I went along about six months blaming myself for poor technique. I then did some preliminary testing that showed a problem and followed that up with whole hog, Focustune software plus LensAlign pro version. I sent the lens to Nikon on warranty repair along with some of my test images. They re-calibrated the lens and life is now good.

What that experience taught me was to come up with a quick and dirty method of doing a quick check. Anything with a regular pattern will do. I simply shoot at an oblique angle to a bit of wood beadboard paneling and visually verify that the DOF is more or less centered on the focal point. If the quick check fails, then I can break out the LensAlign etc.

Ironically the only lens I've had issues with to-date has been the 500mm, the most expensive one by a factor of four or so

[PhotomanJohn]

I check and adjust the micro focus on all of my lenses. It is most important when shooting with fast lenses wide open. The easiest method is to lay a ruler on the floor with a predominate mark to focus on then adjust the micro focus value until actual focus is at the desired point. I recently purchased the FoCal program and rechecked my lenses with it. The adjustment values came out to be nearly what I had determined with the ruler method. It is easy to do and worthwhile.

One should note that the AF accuracy can vary with distance, zoom level and aperture so make your checks accordingly.

John

[ajohnw]

I think this is a bit more sensitive than a star type focus test. It uses Moire interference patterns, There is a bit about these here. Also a large image of the same bulls eye type target down the page. Try looking at different places on it suddenly and scrolling it up and down on the screen. I suspect the instructions here get round pixel density and screen resolution problems as well - find a distance with live view where the bulls eyes are distinct as possible. I would suspect as many as possible could be added to that. The distance should still be long in relationship to the focal length of the lens.

http://www.komar.org/faq/camera/auto-focus-test/

I have just played with a 16mm F2.8 180 degree semi fish eye and it looks like even that could be checked to some 10's of mm.

John
-

[chauncey]

Ditto with everything that John said...made a huge difference when AF is used.

[Stagecoach]

John,

As I'm stuck inside awaiting a possible cyclone I gave it a go today just out of interest.

Trying with just one lens my general walkabout 18-200mm mid zoom which is where I would roughly want it to be most accurate I undertook the following;

a) two shots manually focussed using remote release
b) two shots using mirror up using remote release with each being pre 'de focused' near then far
c) two shots (no mirror up) with remote release taken quickly, no deliberate half press again being pre 'de focused' near then far.

The purpose of taking c) was to see if there was any noticeable difference not allowing AF time to react fully.

I really can not see any difference in the results that would suggest anything is out with this lens at this setting.

As with all these test methods an appreciation of repeatability, deadband and hysteresis helps when making decisions and determining test calibration protocol.

Grahame

[ajohnw]

Grahame makes a good point about making sure any "errors" really are errors. Mentioned in the 1st link but not in detail.

Having bought a D7000 I have wondered about this aspect but photo's seem ok so haven't bothered to build a target plus ruler type set up. This method is easier to do so can be used to reassure.

Curious that I found it when I need it though. More cheap Nikon glass. Full frame semi fish eye, 16mm F2.8 and totally manual. It covers 120 degrees on crop. Nikon's AF confirm doesn't like it one bit but I need to know if the distance scale on the lens is correct just in case. Not easy as the focus scale from 0.6 to 1.2m is bigger than 1.2m to infinity. I don't really feel the need for AF on a lens like this. Set F11 for instance and it covers a bit over 0.5m to infinity on FF.

John
-

[Downrigger]

One should note that the AF accuracy can vary with distance, zoom level and aperture so make your checks accordingly.

That's interesting, John, but a bit discouraging. If this is so, I may misunderstand the reason and purpose for setting microfocus adjustments for my lenses, and may need some help with this concept and procedures for it.
I thought the point for microfocus setting adjustment was that for a particular, specific lens and camera mating, the right setting corrected for minute differences between the length of the light path to the interferometer and to the sensor. If that is correct then I would think the test is 1. relevant solely to "mirror down" shooting when using autofocus and 2. applicable to any focal length on a zoom, since that structural distance discrepancy remains the same regardless, and regardless of aperture too. Moreover, the adjustment cannot address autofocus repeatability, which always shows small variations, and means that in finding the right setting, one should evaluate multiple exposures at each tested microadjustment setting.
I test my lenses wide open, as I would think this most sensitive to the problem for a particular lens camera pair (wouldn't all adjustment settings look good at f22?), and, in any case, the adjustment is probably only important for satisfactory focus when using the lens at wide apertures.
I'm working up to purchasing a new, longer lens, so could use enlightenment if I'm off the mark.

[DanK]

1. relevant solely to "mirror down" shooting when using autofocus and 2. applicable to any focal length on a zoom, since that structural distance discrepancy remains the same regardless, and regardless of aperture too.

I believe the answer to the first is yes, and I am pretty certain the answer to the second is no, because of imprecisions in the construction of lenses. What I have read--and this was borne out by my own experience--is that back- or frontfocus can vary considerably from one end of a zoom to another. I have tested 4 or 5 lenses over the years using both the method John wrote about and Focal. All but one were essentially spot on. One zoom that I no longer have was not. It had slight front focus at one end but slight back focus at the other. I worried about this until I realized that the amount of misfocus was too small to worry about anyway.

I believe the 5D Mark III allows separate adjustment for both ends of a zoom, which would be useful if the direction is consistent.

[PhotomanJohn]

Mark - The amount error between the AF focus point and the actual focus point on the sensor CAN vary with focus distance, zoom position and aperture like I stated earlier. The amount of the variation varies significantly between lenses based on their design and assembly. One of the reasons I didn't buy the current version Sigma 50mm f/1.4 is that the focus varies significantly as the lens is opened up the first few stops. On most lenses this is not noticeable. I believe that DanK is correct in that Canon allows for two micro focus values to be stored for zoom lenses which is a nice feature for lenses that have a significant variation over their zoom length. Nikon only allows for one so I use an average biased toward the zoom position I deem to be the most important. Several folks talked about repeatability which is important and one thing that the FoCal program checks.

I recommend that the focus accuracy be checked when the lens is new. I do a quick check in the store before I purchase the lens or when a mail order lens first arrives. Like Alaska Dan, I have only found one lens so far off that it needed to go back to the manufacturer for calibration because it exceeded what could be corrected for in the camera. It was also the most expensive (pro with the gold band) lens that I had purchased up to that time. It's a simple test - make the correction (big or small) and then forget it. You have eliminated one variable to being happy with your lens purchase.

John

[Downrigger]

Thanks, Dan and John. Helpful, and - I'm convinced, but don't quite get why what you say is so.

[Stagecoach]

Mark, Dan

Regarding the 'Mirror Up' mode test I undertook (b) it's function was to verify performance of AF with response to speed of operation not actual accuracy of the AF. I should have made that clearer in the post.

You often hear it mentioned especially in landscape photography to "select and place your 'AF focus point' over an object 1/3 of the way into the scene, use mirror up and take shot". Normal procedure of course would be to 1/2 press to allow AF to work then full press to lock mirror up, then another press to take shot.

I just wanted to use what would be considered a bad procedure (no deliberate half press) to see the result. Whilst I have done various checks on AF before using picket fences, brick walls and steel tapes this screen image seems to give a more distinguishable difference for a very small amount.

Grahame

[Glenn NK]

Is this method similar to this one: http://www.fredmiranda.com/forum/topic/1187247/0

[Stagecoach]

Is this method similar to this one: http://www.fredmiranda.com/forum/topic/1187247/0

It's a different method Glen that simply adjusts the 'in focus' indicator to have a band that is equal either side of what you decide is correct focus. Again with this test it does not stress the importance of checking repeatability accuracy before any conclusions are made that determine if adjustment is necessary.

Grahame

[ajohnw]

I just checked the focus marking on my Zenith 16mm/F2.8 at 0.6m. It's a very sensitive test even in a lens like that. Set smack on the 0.6m mark no additional bulls eyes. The optimal position was slightly to the right of the centre of the letter. I adjusted for no red or green but the shot came out showing red. I set the screen virticle with a spirit level and lined up by eye and then measured the 0.6m from the sensor plane. Biggest problem was using live view. My screen is at 100 cd/m^2 so had to turn the lights out. The screen resolution is about 100dpi.

This is the sort of thing that should be seen. 50% reduced crop with contrast boosted a little. The crop took in most of the target.



John
-

[Downrigger]

I just wanted to use what would be considered a bad procedure (no deliberate half press) to see the result. Whilst I have done various checks on AF before using picket fences, brick walls and steel tapes this screen image seems to give a more distinguishable difference for a very small amount.

Thanks Grahame for helping me understand your point. BTW I just sprang for the Nikkor 80-400 and had the shop use lens align with my camera before I took it home. This produces nice curves with what looks like high precision and clearly locates the adjustment with optimum "focus score" graphically. Twas +8 at 80, and +11 at 400, so maybe not trivial for shooting wide open and perhaps adding crop. Proves the point made about zooms, above, and I think validates the concept of lens calibration, in general.