Focus accuracy

[NZDavid]

Advertisement (gone after registration)

Which is most accurate? Rangefinder, autofocus (which kind?), microprisms? I would appreciate some technical answers because I can only go by experience. Does one system produce quantifiably sharper pictures than another?

 

I have always found RF highly accurate, even with wide lenses. But focusing the M9 is more exacting than the M6 or M3. Why? I have read that digital sensors are less forgiving than film. I always used to think focusing with an old SLR (OM series) was pretty easy, but looked at closely, old images were not always bang on, especially using wide angle lenses. I struggle with autofocus, but my only experience has been the X1 and small sensor cameras like the D-Lux 5, not superfast DSLRs. For 90 per cent of the time images are fine, especially scenics and some portraits. I think wide DOF and smaller apertures can create an impression of acceptable sharpness across the whole frame, but I still like to choose one exact point of focus. I find it very hard -- actually impossible -- to focus on someone's eyes with AF. I just aim the central spot and hope for the best. I haven't tried manual focusing with compacts; it's just too fiddly compared with a rangefinder. Another complicating factor is my changing eyesight as I near my first half century.

 

So, which focusing system do you find most accurate and works best for you?

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[k-hawinkler]

That depends ...

 

K-H.

[Sp12]

Live view x10, then contrast detect, then microprism/split prism, then AF, then the standard screen in most pro DSLRs tied with RF for wide lenses, followed by poor DSLR finders, followed by RF for long lenses.

 

For speed I go AF, RF, contrast detect/split prism, and it goes downhill from there.

 

Of course, contrast detect is getting better and better and better as processors get faster.

[darylgo]

Without the use of electronics microprism is highly accurate and for my eyes consistently better than split image. The rangefinder on the M9 is accurate but needs help for our eyes to squeeze every last bit of it's accuracy. To improve accuracy I am currently using a 1.4x magnifier along with a diopter correction. This has all but eliminated any focus issues. I find that focus accuracy might be more variable in low light. Lastly, my accuracy improves with use.

[giordano]

IMHO the time it takes to achieve best focus with whatever device is usually at least as important as its potential accuracy. For accuracy, it's hard or impossible to beat a modern "ground glass" screen and a powerful magnifier (or the live view equivalent) - but it takes an age to set up and use.

[payasam]

The Leica type range-finder, being an indirect method (that is, not through the taking lens) depends on correct adjustment for accuracy.

[lars_bergquist]

Advertisement (gone after registration)

If we assume conditions close to those of 'general photography', i.e. a medium lens (50mmm) and a subject distance of c. 10 meters, the ranking would probably be, in descending degrees of accuracy and speed:

 

1. Leica rangefinder

2. SLR with split-image screen (given a first-class finder, not current run-of-the-mill)

3. SLR, ditto with microprisms

4. SLR all matte screen

5. Autofocus with phase detection (= SLR)

6. Autofocus with contrast detection, or 'focus peaking' which is the manual version

 

Focusing on a magnified cropped finder image is not worth consideration. The 1936 Kine Exakta used that method – a flap that covered the screen, with a small magnifier inset, so that you saw only the small center part. That was an egregious mistake, and even Exakta understood that when after the war, they changed to a standard Rolleiflex style flip-up magnifier. The method is slow enough to be useless, and 'electronifying' it is no improvement.

 

During my SLR days, I came to prefer a plain matte screen with a finely matted circle in the center. But that was a large, bright Olympus OM finder, not the abysmal screens you find on many current DSLR cameras, especially in the APS-C format.

 

In low light, the superiority of the rangefinder increases perceptibly, because the finder image is not dependent on the lens. The RF is also superior for wide angle lenses, which are increasingly difficult to focus precisely as depth of field increases, but RF focusing is independent on that. Accuracy is about equal for RF and SLR versions 2, 3 and 4 with a 135mm lens, where AF is still behind.

 

There are calculations to support this argument, at least the non-AF part of it, in e.g. Günter Osterloh, Leica M, Frankfurt a. Main 2002, p.47

 

The old man from the Age of Scale Focusing

[giordano]

In my experience, current DSLR AF is significantly faster in ordinary circumstances than any focusing screen I ever used. But my eyesight was never up to sniper standard.

[lct]

1. Split-image rangefinders within their accuracy range (no fast teles)

2. Split-image SLR screens at faster apertures than f/5.6

3. Microprisms SLR screens at faster apertures than f/5.6

4. Phase detection autofocus

5. Coincident-image rangefinders within their accuracy range

6. Plain matte SLR screens for anything else than macro, micro or astrophoto

7. Contrast detection autofocus

? Focus peaking, no experience at all but being contrast detection based i don't believe in miracles.

[Redridge]

at f16.... they are all fast.

[NZDavid]

I agree practicality can affect accuracy, for example using split screen focusing in contrasty light. Rangefinders, too -- the bigger and brighter the better. And using LCDs and the latest touchscreen focusing would be nigh on impossible in bright light -- I find it's impossible to even frame with an LCD in our extremely bright and glary sunlight.

 

I also used to have Olympus OMs and finders were big and clear. I also once tried out the original Leicaflex with its central microprism focusing (the rest of the finder was always sharp) and thought this was a very easy way to focus.

 

I really like rangefinder focusing, but of course it is restricted to 135 lenses and less. More and more we are likely to see mirrorless cameras taking over, but to date the AF system used in these camera has been slower than DSLRs. The latest Olympus OM-D claims to be a lot faster.

 

I still don't get how you can perfectly focus on someone's eyes using AF compared with RF.

[jamesk8752]

(Snip)

I still don't get how you can perfectly focus on someone's eyes using AF compared with RF.

 

Surely, you simply cannot. That's why we still have rangefinder cameras.

 

Regards, Jim

[lct]

Focussing on someone's eyes is not difficult with a dSLR if light is good enough for the AF to work. Nothing beats a good rangefinder in low light though.

[Sp12]

If we assume conditions close to those of 'general photography', i.e. a medium lens (50mmm) and a subject distance of c. 10 meters, the ranking would probably be, in descending degrees of accuracy and speed:

 

1. Leica rangefinder

2. SLR with split-image screen (given a first-class finder, not current run-of-the-mill)

3. SLR, ditto with microprisms

4. SLR all matte screen

5. Autofocus with phase detection (= SLR)

6. Autofocus with contrast detection, or 'focus peaking' which is the manual version

 

Interestingly, this is wrong. At least, by your statement of speed and accuracy. The single most accurate system is contrast detect -- it's immune to calibration errors and precise to the pixel level. The Leica rangefinder base is actually too short (assuming 20/20 vision and perfect calibration, which is not always the case) to offer greater than 70% accuracy for the 135/3.4 (at any range), and has a non-100% hit rate on anything 50/1.4 or faster/longer at some focus distances. SLRs and the RF are limited by the precision of their adjustment and the viewer's eyesight, and Canon had a whitepaper out comparing AF accuracy in the 1V (old tech) to pro-level sports photographers on MF. Canon's old pro-level AF was consistently more accurate and faster than professionals making a living off being able to MF.

 

I still give RFs the major nod for fast focusing of wides, especially in low-light, but for longer lenses, moderate light, and shots requiring critical accuracy the RF is not it.

[lars_bergquist]

I don't follow you there. Contrast AF is a mechanised version of eyeball focusing on an all matte screen. You turn the focus and see if the image gets any sharper, if it gets fuzzyer you turn the other way, continue until the image gets fuzzier again, change direction, turn until it gets fuzzier ... repeat until you are satisfied. Hopefully, every 'stroke' will be shorter than the previous one.

 

Sometimes it isn't. What then happens is called "AF hunting". It can go on for some time. We have all seen cases when it never stopped.

 

Phase comparison AF on the other hand is technically similar to both a rangefinder, and to a split image aid on a screen, which is in effect a simulated mini-rangefinder, as it simultaneously compares two images. Like the split-prism, the AF system's base length is limited by the aperture of the lens which is hopefully wide open. But like a RF, the separate AF sensor and its associated electronics can decide both in what direction focus is to be changed, and also how much. So it can get it right in the first try. Even for a medium-long lens, this may well be faster and more precise than a RF, if the user is inexperienced.

 

The great advantage of the RF is that it is independent of the lens. It delivers one precise level of accuracy irrespective of what lens you mount on the camera, as long as both the RF and the lens are correctly adjusted, while short focal lengths and slow speeds result in lower accuaracy with 'TTL focusing'. With longer forcal lengths however the accuracy requirement increases, until at around 150mm it does no longer suffice. But TTL focusing, being lens dependent, delivers greater accuracy with increasing focal length.

 

This is slightly simplified, but essentially correct. I stand by my ranking.

 

L.B.

[pico]

The easiest cameras for me to ever focus was the Nikon F with the optical sports finder and the later F with high-point finder. But then I wear thick glasses. I have a miserable time with DSLR autofocus. I feel out of control with it, and the screens are almost transparent (to me).

 

The Leica M - well, I can no longer see full frame but experience has made up for that.

[kanzlr]

 

The great advantage of the RF is that it is independent of the lens. It delivers one precise level of accuracy irrespective of what lens you mount on the camera, as long as both the RF and the lens are correctly adjusted, while short focal lengths and slow speeds result in lower accuaracy with 'TTL focusing'. With longer forcal lengths however the accuracy requirement increases, until at around 150mm it does no longer suffice. But TTL focusing, being lens dependent, delivers greater accuracy with increasing focal length.

 

This is slightly simplified, but essentially correct. I stand by my ranking.

 

L.B.

 

 

Exactly. For example, my 28/2.8 is way easier to focus correctly on the M8 than on the GXR, because on the GXR you judge what is sharp, not if you have set the lens to the correct distance.

With a longer, faster lens and somewhat close focusing distance, the GXR works perfectly fine, and at 200/4 it is still spot on, whereas this gets tricky with the M (it works with longer lenses, but you have to be pretty careful and accurate).

[adan]

The Leica type range-finder, being an indirect method (that is, not through the taking lens) depends on correct adjustment for accuracy.

 

True, but except for actually focusing with the image off a sensor (live view or EVF or contrast-detect AF), all other means of focusing are also dependent on correct adjustment.

 

SLRs depend on the ground-glass and/or focus screen or AF sensor being correctly positioned relative to the lens mount and film/sensor plane. And also the various moving mirrors.

 

Even a large-format view-camera is dependent on the back and film holders being correctly adjusted, to ensure that the film is installed in exactly the same plane previously occupied by the ground-glass.

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[pgk]

Like many topics discussed here there are caveats and trying to discuss absolutes and come up with an order of accuracy table will only ever provide a generalised and not always correct one.

 

For those who think that current AF systems are spot on, try the Canon 85mm f/1.2 - its not easy to focus absolutely partly due to the difficulty of pointing it at the precise, exact point required in focus, when looking through the viewfinder. Uncluttered ground glass screens allow us to use more than a simple, single point and we can assess overall focus by our appreciation of the entire scene and in some circumstances this is better than any other method. Rangefinders allow for excellent wide-angle focus to be achieved, and can also use the technique whereby the lens is rotated back and forth to enable the precise point of focus to be obtained by multiple assessment of the coinciding image in otherwise difficult situations. Magnified live view is extremely accurate but awkward to use in bright light conditions or handheld.

 

So my view is that its horses for courses. No one system is best because all have advantages and disadvantages depending on the conditions faced when using them.

[algrove]

NZDavid

 

Have your eyes recently changed and do you wear eye glasses? Can't tell from your photo. May sound stupid, but this can type of change cause focusing issues with your RF.