An autofocus oddity

[simpo two]

About a month ago I got a Nikkor 17-55mm f2.8, specifically for low-light interior work.

After my experience with the myopic Sigma 20-40mm, I kept a close eye on the focusing accuracy and got the impression that it was variable at best - some shots were OK, others defo seemed slightly front-focused.

Last night, as there was nothing to watch on TV, I set the camera up on a tripod and did some AF tests. Not in the usual way, by taking photos and examining them afterwards, but by just triggering the AF by half-depressing the shutter release. Normally you do this once, the camera goes 'beep' and the focus is set. Pressing the button again doesn't change the focus as it's already set. Invariably you're looking through the viewfinder as you do this.

Last night I did it differently and immediately noticed something odd. With the camera fixed and pointing at a suitably contrasty subject, I triggered the AF as above but this time watched the distance ring on the lens. Each time I triggered the AF, it would move a little, to left or right. In other words, the AF is not consistent - it focuses in different planes each time. I did the test again this morning with better light and the result was the same.

The effect is neglible at 55mm but very noticeable at 17mm and is surely the answer why some of my shots look out of focus at wide angle and f2.8.

Could anyone else with a 17-55mm do the same and see what they get please?

[speedychrissie]

i wont be able to help as i dont have that camera or lens, but one thing that might be useful to know:
what kind of autofocus is it set to? ie on the D70 you can tell it to focus directly in the middle, or focus on the closest thing it can find or choose what area of the view to focus on.

it might not make a difference, but you never know.

[chris.mapey]

If you want to try a different D70 body with your lens, give me a call. I can't help you with the lens (as you know!!)

Chris

[fergusd]

Camera on tripod, Single servo mode, center focus spot only enabled, pointing at a static scene with good contrast on the focus spot.

At 55mm the lens never moves, at 17mm the lens moves very slightly perhaps 5% of the time - when the lens does not move the 'focussed' spot just lights up, when it moves one of the wee arrows lights up briefly then the focussed spot.

Interesting for me as I basically never use the camera on single servo mode, I use constant mode almost all the time and change focus sensors rather than frame/focus and move.

I've also not really noticed any focus issues, and it's said that front or back focus issues are a body problem not a lens problem (which makes sense to me as the body is what actually determines what is and is not in focus).

Fd

[te51cle]

I've just tried it on my EOS1 with variety of Canon lenses and one old Sigma telephoto lens. The only time I get any autofocus hunting is at f5.6 on poorly lit low contrast subjects - which is what you'd normally expect. Sounds odd, as long as the light on the subject is sufficient the camera should lock on and stay locked on and you're perfectly capable of setting that up. I did wonder if the camera was in some kind of predictive autofocus mode but as you mounted it on a tripod it should have eliminated that particular problem.

[_dobbo_]

Without a 17-55 to try I whacked the 18-70 on. at 70mm the distance marker only moves about 10% of the time, and only a fraction of a millimeter. At 18mm it moves every time, between 1/2 and 2mm. Perhaps goes some way to explain the extremely inconsistent performance of my 18-70, and why I don't trust it.

[LongQ]

S'funny. I get the same sort of results from all my lenses and not one of them has autofocus adjustment(well, not intentionally ...)

[elderly]

fergusd said:

the body is what actually determines what is and is not in focus).

If there is a back focus problem, it shows up more with wider angle lenses;
the opposite of what you would expect.

[simpo two]

speedychrissie said:

what kind of autofocus is it set to? ie on the D70 you can tell it to focus directly in the middle, or focus on the closest thing it can find or choose what area of the view to focus on.

This is my standard set-up, ie single servo focus and centre sensor.

chris.mapey said:

If you want to try a different D70 body with your lens, give me a call.

You're on; it will help me define the problem, if that's what it is. Time you took a day off and called into Simpo HQ

fergusd said:

Camera on tripod, Single servo mode, center focus spot only enabled, pointing at a static scene with good contrast on the focus spot.

That's it!

fergusd said:

At 55mm the lens never moves, at 17mm the lens moves very slightly perhaps 5% of the time - when the lens does not move the 'focussed' spot just lights up, when it moves one of the wee arrows lights up briefly then the focussed spot.

Thanks for the test; mine changes its mind about 80% of the time when at 17mm.

fergusd said:

I've also not really noticed any focus issues, and it's said that front or back focus issues are a body problem not a lens problem (which makes sense to me as the body is what actually determines what is and is not in focus).

But all my other lenses work fine on this body...

elderly said:

If there is a back focus problem, it shows up more with wider angle lenses; the opposite of what you would expect.

Odd. However this is slight front focus.

[elderly]

simpo two said:

elderly said: If there is a back focus problem, it shows up more with wider angle lenses; the opposite of what you would expect. Odd. However this is slight front focus.

Maybe we are at cross purposes about the term 'back focus' ?

Back focus is where the plane of (lens) focus lies behind the image sensor
(maybe because of an incorrectly aligned sensor in manufacture , like in the days of film when not having the film perfectly flat).

The depth of focus inside a camera body is in any case very small but is at its smallest with short lenses
and AFAIK with an SLR the sensor does not receive the image until the mirror is out of the way,
by which time the lens has already been focused.

[simpo two]

elderly said:

Maybe we are at cross purposes about the term 'back focus' ?

Good point. How do I know which side of the film plane the image is focusing? I used the term 'front focus' to mean the lens is focusing in front of the sbject.

[elderly]

simpo two said:

elderly said: Maybe we are at cross purposes about the term 'back focus' ? Good point. How do I know which side of the film plane the image is focusing? I used the term 'front focus' to mean the lens is focusing in front of the sbject.

I doubt if one could tell;
you could give the body back to Nikon and ask them to check it -
As prices for DSLRs have dropped, I fear that quality control has dropped as well, for all but the 'high-end' stuff.
With a sensor it's no longer just a question of machining the film plane as it used to be in the good old days.

There could also be a problem with the relative positions of the AF sensor and the image sensor.
I would talk to the pro. dept. at Nikon or a reputable repairer (do Fixation still do Nikon?) and
see what they say.

[LongQ]

Interesting comments which provoked this thought ...

One of the things you pay for with a fully pro body is longevity. The Pro body will normally have been built with additional water and dust exclusion systems and a duty cycle of, from memory, 150,000 + images expected. Maybe 250,000? I guess I could look it up but that is not the point.

So what is the expected max duty cycle of consumer market bodies?

I Japan I guess it would be about 6 month's worth of shots for a camera used regularly - so about 2,000 shots for a pocket job, maybe 25,000 claimed for an SLR with about 5,000 expected in the 6 months if the user is a keen photographer and so on for cameras in the middle markets.

All these numbers are guesswork but for my purpose here the real numbers possible don't matter.

The fact is simpo, I conclude from my guesstimates, you have worn out your instrument (how many frames have you taken?) and are due to hand over some more readies to Mr. Nikon to refresh your kit as was their original plan.

You could of course check this by popping down to your nearest Nikon dealer to check out your lens on a new D70s/D200(?stock)/D2H/D2X . Just forget to take you cheque book/credit card/pen-that-you-always-use-when-signing-things if you do so.

I would be delighted to be wrong but suspect that elderly was right with comments about quality control or, rather, the quality products are built to in a market where usage can be very low. eBay is awash with kit which claims, often genuinely I suspect, to have 'hardly been used'. My guess is that in the last 10 years the manufacturers have relied upon little usage as part of their business plan.

>> Edited by LongQ on Sunday 12th March 19:38

[V6GTO]

LongQ said:

Lots of stuff that rings true

[fergusd]

The man said it didn't exhibit the same problem with other lenses . . . suggests it's not abody problem . . .

Fd

[simpo two]

elderly said:

you could give the body back to Nikon and ask them to check it etc

I've been perfectly happy with the body's performance with the 18-70mm lens - this is only 1mm away from 17mm; although the f2.8 aspect might be showing up tolerances.

LongQ said:

The fact is simpo, I conclude from my guesstimates, you have worn out your instrument (how many frames have you taken?) and are due to hand over some more readies to Mr. Nikon to refresh your kit as was their original plan.

I have been out-cynicked I got the camera in July 2004 and it's done about 12,500 frames.

As AF seems so fickle, especially with fast lenses, and we're probably talking microns in film-plane terms, you'd think they'd include an adjusting screw... I didn't imagine myself adding shims to my lens mounts!

Do you think they make a D2X lens mount any differently?

[fergusd]

Shimming your lens mount will make no difference . . . shimming the position of the AF sensing array on the other hand will . . . I also expect that the tolerances to which components like this must be mounted will be very small.

Fd

[LongQ]

fergusd said:

The man said it didn't exhibit the same problem with other lenses . . . suggests it's not abody problem . . . Fd

Yes indeed, it could simply be a general quality control problem on lenses - or a design issue.

But Simpo has had the problem with the Sigma lens and now with the similar Nikon lens selected to replace the problem Sigma.

Now it could be simply that at that end of the lens market/operating range these things happen.

Or it could be that elderly's comment

"The depth of focus inside a camera body is in any case very small but is at its smallest with short lenses"

highlights that a potential problem will be more evident with lenses at that end of the scale.

Now I have no idea how the lenses are designed but I would imagine that they are likely to be based on step focussing, albeit it fine resolution. So if the measurement is marginal the automatic adjustment might well cause a change of setting on a regular basis.

The thing is that wide angle lenses, on 35mm film or full size sensors at least, have a very short hyperfocal distance which should really mean that in most normal circumstances the subject matter will be in acceptable focus (to the human eye) almost independently of the focus set. The exceptions would be relatively close focus or highly critical depth of field issues for some subject types.

At those tolerances a small amount of wear on components somewhere in the chain MIGHT result in fractional mis-measurements which in turn MIGHT result in marginal, but sometimes visibly evident, focus decision anomalies.

My (limited) experience with wide angle lenses suggests that the results are either remarkable or stink. I have one ancient independent 28mm prime lens that produces really nice results but only at specific settings under good lighting conditions. It's replacement manufacturer lens is only marginally better.

By way of contrast I am in the middle of scanning some films I shot last weekend some of which were through a recently acquired used 600mm Canon L spec lens with a worn bayonet mount. Every body I have wobbles on it, by several degrees, suggesting that the distance to the film plane is not going to be very consistent. But the shots I have scanned so far, where the 'operator' managed to activate the shutter when the subject was in the narrow DoF, have been superbly sharp, which suggests that the longer focal lengths can tolerate the anomalies much better then the shorter ones.

Or am I very wide of the mark on this?

[simpo two]

I'm still at a loss to understand how my Sigma 70-200mm f2.8 can nail a fly's eyebrows at 40 yards, yet a almost £1,000 Nikkor 17-55mm can't manage it.

[fergusd]

;-) you're not focussing on the real issue ;-)

your badly fitting lens mount does NOT change the difference in distances between the focussing screen (or AF sensor) and film plane (or imaging sensor) and the lens itself . . . so if it looks in focus on the screen (and is in focus w.r.t the AF sensor) then it is in focus, regardless of whether the lens itself in wobbled in or out . . .

The difference in distances (as above) should be zero (else back or forward focus will be the case - although may not be evident as Simpo suggests - aperture and DOF will change how visible it is).

Fd