HEADS UP: AF FINE TUNE

cheers - I'll probably pick one up through sheer laziness (you can print them and construct)
 
jacob12_1993 said:
Amateur Photographer are giving away a free lens calibration chart with their magazine in Junes issue, On sale 31st May.

http://www.amateurphotographer.co.uk/features/FREE_AF_calibration_test_chart_features_307653.html

Got to be worth the couple of quid for the magazine :thumbs:
Click to expand...

That chart is so flawed it borders on the irresponsible. It will cause more problems than it solves :nono: This is what you need, and how to it properly, with something like the LensCal (though I would move back a bit from the distance shown in the video, to make sure) http://spyder.datacolor.com/learn_videos_spyderlenscal1.php

The LenCal is $60, but this set up is identical. It is far better than that horror from AP - more accurate, far less prone to error, and free. Though you can send me a couple of quid if you like.

IMG_3381-1.jpg
 
HoppyUK said:
That chart is so flawed it borders on the irresponsible. It will cause more problems than it solves :nono: This is what you need, and how to it properly, with something like the LensCal (though I would move back a bit from the distance shown in the video, to make sure) http://spyder.datacolor.com/learn_videos_spyderlenscal1.php

The LenCal is $60, but this set up is identical. It is far better than that horror from AP - more accurate, far less prone to error, and free. Though you can send me a couple of quid if you like.

IMG_3381-1.jpg
Click to expand...

What is so flawed about the chart, if you set it up in the place of the ruler in your diagram I can see it working better in fact. I for one will be picking one up :thumbs:
 
jacob12_1993 said:
What is so flawed about the chart, if you set it up in the place of the ruler in your diagram I can see it working better in fact. I for one will be picking one up :thumbs:
Click to expand...

You don't focus on the ruler. You focus on the flat face of the box - as in the LensCal video. Then read off any error from the angled rule.

The problem with the AP chart is that it's at an angle so you can never be sure which part of the target the sensor has actually picked up on (the sensor area is quite a bit larger than shown in the viewfinder).

You must have a flat target or you don't know where you are. You can guarantee people will have problems with that. There are several similar targets you can download from the internet and they are all the bane of camera service departments.

The other common problem is shooting at too close distance, which exagerates errors or even creates some where none exists in normal use. No lens is ever perfect and zooms are prone to being out by fractionally different amounts at various zoom and focusing distances. If you correct for one extreme setting, you will throw it out for normal working range.

If Geoffrey Crawley was still with us, he would never have allowed that thing to go out with AP. He died a few months ago and was the only person left on AP with half a clue :( It's no different to laying the a newpaper on the floor and shooting that at an angle. It's a stupid idea.
 
Last edited:
HoppyUK said:
You don't focus on the ruler. You focus on the flat face of the box - as in the LensCal video. Then read off any error from the angled rule.

The problem with the AP chart is that it's at an angle so you can never be sure which part of the target the sensor has actually picked up on (the sensor area is quite a bit larger than shown in the viewfinder).

You must have a flat target or you don't know where you are. You can guarantee people will have problems with that. There are several similar targets you can download from the internet and they are all the bane of camera service departments.

The other common problem is shooting at too close distance, which exagerates errors or even creates some where none exists in normal use. No lens is ever perfect and zooms are prone to being out by fractionally different amounts at various zoom and focusing distances. If you correct for one extreme setting, you will throw it out for normal working range.

EDIT: Just curious really as you clearly know more about it than me, not trying to cause any arguments
If Geoffrey Crawley was still with us, he would never have allowed that thing to go out with AP. He died a few months ago and was the only person left on AP with half a clue :(
Click to expand...

If a cereal box was placed in line with the 'middle' of the test chart and the chart was shot from 50x the focal length of the lens, how does it differ from a ruler?
 
It probably wouldn't, but it would have cost the price of a copy of AP ;)

I was one of those people who gets a new lens, takes a few pics of a diagonal ruler on a desk and then wonders why I doesn't appear to focus properly. So I think Hoppy's warning is very helpful.

Not sure of his choice of cereal though! ;)

Toby
 
jacob12_1993 said:
If a cereal box was placed in line with the 'middle' of the test chart and the chart was shot from 50x the focal length of the lens, how does it differ from a ruler?
Click to expand...

So what do you need the chart for? The important thing is that there is absolutely no room for error on the target - it must be flat (check that the box isn't bulging with cerial goodness - AF systems can be that accurate) and that it's square to the camera, and with a random bold pattern for the AF to lock on to.

Do that at a sensible distance and that's the focus sorted. You just need an adjacent angled subject so you can see where the focus has actually landed, either in front or behind or, chances are, smack in the middle.

Check the video above :thumbs:
 
Have to say the best way I've found to do this is photo real world objects and vary microadjust and take a set (i.e. microadjust at steps of 2 or 3 and photo at the focal lengths marked on the lens. To test close things, focus on some text at an angle and see where the fuzziness is - you'll see the position of focus change as you vary the microadjust. For further objects, flat on with some detail and see which microadjust position is best at each focal length. I do this for close (<2m), mid distance (~5m) and long distance (~50m).

The problem with the focus test charts is that it is a single snapshot at a single distance and focal length (if a zoom). I regularly take photos at << 25x focal length (that aren't macro ;)) and the static lens tests have proven both more long winded and less accurate than handholding and taking photos of real things ('cos that's what I do in real life...).
 
HoppyUK said:
If Geoffrey Crawley was still with us, he would never have allowed that thing to go out with AP. He died a few months ago and was the only person left on AP with half a clue :( It's no different to laying the a newpaper on the floor and shooting that at an angle. It's a stupid idea.
Click to expand...

Well said Richard! :thumbs:
 
I've been checking the general sharpness of my kit and have come to the same conclusion as stated here that this "free chart" will be, for all intents and purposes, useless.

One thing I observed when using these "angle at 45 degree" charts is that they produced wildly inconsistent results. Using the same AF fine tune settings one shot would front focus, the next would back focus, and one would be more or less right. Also if the the actual focus sensors don't correspond exactly to where the focus points are in the viewfinder, which I'm led to believe is often the case, then any results obtained using this form of focus chart won't apply to real world use.

Needless to say I became quite frustrated with this and finally found another chart based on the ISO 12233 standard. This one has finally allowed me to consistently reproduce any front/back focus issues and dial them out.

The original ISO 12233 chart is under copyright, however the author of this chart claims that the data used to create the chart appears not to be protected, and has used this data to create a very similar chart.

Anyway you can find it here in a fully scalable PDF format. Print it out as higher resolution as possible (at 600 dpi I'm getting moire patterns in the finer areas, but it's more than good enough to assess focus), and at 100%. 100% will crop on an A4 page but as long as you get the center area you'll be OK.

Best thing is it's even more free than buying a magazine with a dodgy test chart.
 
HoppyUK said:
That chart is so flawed it borders on the irresponsible. It will cause more problems than it solves :nono: This is what you need, and how to it properly, with something like the LensCal (though I would move back a bit from the distance shown in the video, to make sure) http://spyder.datacolor.com/learn_videos_spyderlenscal1.php

The LenCal is $60, but this set up is identical. It is far better than that horror from AP - more accurate, far less prone to error, and free. Though you can send me a couple of quid if you like.

IMG_3381-1.jpg
Click to expand...

Sorry, am I missing something here, you paid £40 for a ruler? :lol:
 
Save your money because these things could drive you mental especially if there is any hint of OCD in your personality, akin to giving hypochondriacs a guide to illnesses and their symptoms
 
Dear God! I've just seen that travesty in AP. Stay well clear. Listen to Hoppy. Nobody should be focusing on angled targets when trying to calibrate AF. The only thing that should be at an angle is the scale from which you determine the focus position. The target you aim at should be perpendicular to the camera, no exceptions.

Ignore the price, just look at the LensAlign product for an example of a well designed AF calibration setup - http://www.whibalhost.com/lensalign/how-lensalign-works.html - and then fabricate your own equivalent for free from household bits and bobs, just as Hoppy has done and just as I do.....

20110412_010137_000.jpg


Note also the garden setting and bright, sunny conditions for my example. This is because I was calibrating a 400mm lens and wanted to calibrate at a representative distance for real world use. I also wanted good light because I was testing focus accuracy, not focus sensitivity, and wanted the camera to have enough light to work with properly and to maintain a sufficiently high shutter speed to eliminate camera shake as a concern, even with a tripod. A lot of people make the mistake of calibrating at distances far too short to be representative of real world shooting and in light levels so low as to bring other potential problems into play, like slow shutter speeds and camera shake, or high ISOs which corrupt the image with noise. These problems will not help your AF calibration efforts.

If you're going to put in the effort to calibrate your equipment then you might as well put in the effort to do it properly, not with some half baked approach that will accomplish nothing useful other than to keep you occupied for a while.
 
Have to say that I've had inconsistent results using ruler/square on method. Had much better results (i.e. the microadjust figure was different) taking real world pictures at real world distances (which, for me, are invariably closer than 50x focal length).
 
Slaphead said:
I've been checking the general sharpness of my kit and have come to the same conclusion as stated here that this "free chart" will be, for all intents and purposes, useless.

One thing I observed when using these "angle at 45 degree" charts is that they produced wildly inconsistent results. Using the same AF fine tune settings one shot would front focus, the next would back focus, and one would be more or less right. Also if the the actual focus sensors don't correspond exactly to where the focus points are in the viewfinder, which I'm led to believe is often the case, then any results obtained using this form of focus chart won't apply to real world use.

Needless to say I became quite frustrated with this and finally found another chart based on the ISO 12233 standard. This one has finally allowed me to consistently reproduce any front/back focus issues and dial them out.

The original ISO 12233 chart is under copyright, however the author of this chart claims that the data used to create the chart appears not to be protected, and has used this data to create a very similar chart.

Anyway you can find it here in a fully scalable PDF format. Print it out as higher resolution as possible (at 600 dpi I'm getting moire patterns in the finer areas, but it's more than good enough to assess focus), and at 100%. 100% will crop on an A4 page but as long as you get the center area you'll be OK.

Best thing is it's even more free than buying a magazine with a dodgy test chart.
Click to expand...

That link is to a resolution test chart - not the same as an AF target. Different application, and also seriously prone to errors especially with wide angle lenses because it has to be shot at close distance and all lenses suffer from field curvature to some extent, wide and standard zooms especially. Which means the edges and corners of a flat 2D target often look poor when in fact they're not at all. So you think you've got a duff lens, when you haven't.

For AF testing, the target needs to be flat, and set square to the camera, in good light. Large enough to cover the AF point, and a bit more to be absolutely sure - like filling the centre circle in the viewfinder. Shoot at a sensible/typical/normal distance - it needs to be close-ish because otherwise depth of field covers any focusing errors and it becomes hard to see anything accurately. But if you go too close, to see changes more clearly, it's very easy to push the lens right to the edge of its operating envelope where any small errors are exagerated. If you then correct for those, the lens will be out at normal range. No lens or AF system is 100% perfect - they are mechanical and analogue devices and subject to tolerences.

Then the target needs a bold high contrast area of random graphics for the sensor to lock on to - the logos and lettering on boxes and packaging are perfect for that. In theory, AF systems can be upset by small regular/symetrical patterns like resolution targets.

As the cerial box shows, and Tim's garden target shows, this kind of test is super-simple and easy with a really basic target and set up. I've gone nearly blind using all manner of pseudo scientific-looking targets. I'll swap that for a bit of sound kitchen science any day.

On the other hand, I do rather like that little LensCal device linked above, and $60 is not that mad (some others cost $250 :eek:). It's a cool gadget, but actually doesn't do anything better or easier or faster than a box and a ruler.
 
The target should have high contrast, well defined edges to allow the AF system something to latch on to. There is no point having wishy washy swirls of pattern. Sharp edged text should be perfect.

As for bowing of a cardboard box, if you're calibrating an 85/1.2 at close range then every mm matters and you need good precision in your setup. Just look at the LensAlign site to see examples of how wafer thin the DOF is with that lens. If you are calibrating a slow (f/5.6) 400mm lens at 20m (50X focal length) then in my opinion an error of a mm or two is unimportant, especially when your DOF at that range, even viewed at 100% runs to a few cm fore and aft. The closer your focus target the more important the accuracy of your setup. An error of 2mm at 1,000mm (1m) distance is more significant than an error of 2mm at 20,000mm (20m).

Basically, the faster your glass the more important the attention to detail in setting up your test. The slower your glass the more relaxed you can be, because the AF system, combined with slow glass, just won't be mm perfect time after time after time. Even so, you may as well do what you can to maximise accuracy as best you can. After all, it's accuracy that you are chasing.
 
Last edited:
dave_bass5 said:
Ive never found cereal packets to be particularly sharp, nor perfectly flat as they tend to me made out of thin cardboard. Wouldn't this cause issues.
Click to expand...

It shouldn't, though I've mentioned that above. Cerial boxes can bulge a bit, just make sure they don't. And make sure the box sits square and doesn't lean (Weetabix is not good for this test ;)). Not sure what you mean by cerial packets not being sharp, you just need a bit of contrast in the AF target - text and graphics are perfect.

This is another area where a bit of distance helps to minimise errors. If you're too close, then if the box is bulging a bit or there is significant distance between the AF target point and the ruler, then basic geometry suggests that the two will be at fractionally different distances, and even slight field curavature might come into play. So make sure the box is flat, focus on an area adjacent to the ruler rather than the middle of the box.

If you want to eliminate all potential errors, then don't use the ruler. In fact, that's the best way to test wide angles which are frankly very difficult becaase they deliver so much inherant depth of field. The best method there is to set the target up as described, but taking very particular care to get everything absoluetly square and flat.

Then AF on the target dead centre, and make sure you have some fine text to look at (becuase you haven't got the ruler markings to reference) and shoot a sequence of images at different microadjust settings either side of zero, at +/- 5 increments. Zoom in on the LCD and click through the images and pick the sharpnest one. Then do the same again at increments of +/- 1 either side of that. You will find it very hard to distinguish the images with settings that fine, and it is a good camera/lens that can deliver that kind of accuracy every time. Check that by defocusing and refocusing a few times.

Don't expect perfection, and keep it real. I find my Canon L lenses on a 5D2 are actually amazingly accurate and consistent, but my cheapo 50 1.8 is sometimes all over the place in a test like this :gag: But having said that, in practise the weakest link in the chain is usually me. In reality, nailing the focus to within a few mm, which is the level of error we're looking at here, is rarely possible unless the subject is stone dead and you've also stopped breathing.

Edit: crossed post with Tim :)
 
Last edited:
You must log in or register to reply here.
Back
Top