Strange problem with images from 180-600

Leonard Shepherd said:

The focus accuracy part of the thread IMO developed from the question of whether AF was capable of consistent accuracy as distinct from whether the AF settings were the best for the subject (perhaps part of your point) or if with a very few of the images in the thread where the subject had enough contrast for reliable AF detection.

Click to expand...

The question here was whether or not the system stayed focused on a scene in the field, not whether it was accurate on a static lab situation. Those are two different problems. And based upon my tests, I'd say that the answer is different.

Leonard Shepherd said:

I am satisfied the best way to test focus accuracy with Nikon ML is a flat target parallel to the sensor using pin-point AF with a good contrast subject parallel to the short dimension of the screen.

Click to expand...

I would argue that the only thing the flat target testing might tell you is whether or not a component is out of tolerance. This is how we discovered the D800 left focus problem, for instance.

Leonard Shepherd said:

Nikon to the best of my knowledge has not specified if there are photo detection sites on Z bodies in addition to the selectable viewfinder AF points.

Click to expand...

Not sure what you mean here. We've pretty much reverse engineered what Nikon/Sony are doing, and it agrees with the patents. The viewfinder "AF points" are not exactly the detection points. First, there's a slight accuracy issue with the overlay positions. Second, the focus information is coming from hundreds to millions of individual collection points (depends of AF-area mode). Third, subject detection alters the focus points looked at based upon learned patterns.

Leonard Shepherd said:

I expect there are many more extra PD sites than selectable AF points.

Click to expand...

Absolutely. Every blue photosite on every twelfth row, basically (there's a small margin at the edges). This makes the system highly dependent upon long axis detail.

Leonard Shepherd said:

The D700 seemed to have a few with outer vertical AF detection. The D850 and the more advanced D6 AF systems according to Nikon have them.

Click to expand...

The D6 is unique in that it has crosshatch detection at every detection point. The D500 and D850 don't, and moreover, have a bigger gap between some of the detection points, introducing blind spots.

Leonard Shepherd said:

There are likely more PD sites in the Z8/9 than Z6/7 - to make possible the wide range of AI subject options and AF D6 AF area "shapes" possible with the later generation Z9 & Z8.

Click to expand...

Again, given that it's every twelfth row and every blue photosite, yes, the Z7/Z8/Z9 have more (and smaller) focus data points than a Zf/Z5/Z6. The smaller bits of the 45mp cameras create a small problem when not all the photosites are reporting useful data, as can happen with low contrast in low light.

Thom Hogan said:

The question here was whether or not the system stayed focused on a scene in the field, not whether it was accurate on a static lab situation. Those are two different problems. And based upon my tests, I'd say that the answer is different.

Click to expand...

IMO some yes - often with some no.

If the subject or the camera/lens are not stable then what the AF is aimed at changes - and focus may not be consistent - independent of the AF system accuracy.

Similarly if an AF subject is borderline sometimes AF may lock on or off - due to subject issues rather than AF accuracy.

Thom Hogan said:

I would argue that the only thing the flat target testing might tell you is whether or not a component is out of tolerance.

Click to expand...

If a flat target incorporates fine resolution detail it will record, ideally with repeat testing, if there is the highest resolution with nil or some other fine tune setting.

Most photographers usually aim to achieve the highest resolution.

Thom Hogan said:

This is how we discovered the D800 left focus problem, for instance.

Click to expand...

I am not generally part of "we"!

It is easy to demonstrate the outer AF columns of the D800 have AF detection limited to central and inner AF sensitivity, unlike the more central AF columns that generally have both left and right as well as central sensitivity.

Some reported they found the D800 did not AF to the left of centre of an outer left AF point, and many more reported the D800 did not AF to the right of centre of an outer right AF point.

How many checked both left and right outer columns seems not recorded.

I did

While there seems to have been the occasional defective D800, my methodology indicated the underlying issue was highly likely due to limited AF detection ability; either outside (left) of a left outer column AF point, or outside (right) of a right AF detection column.

For those interested the AF detection width of a D800/D850 AF point is easily established.

First find a small subject that a central AF point can just detect. Then, with first pressure remaining on the shutter, carefully move, the camera left or right until focus is lost.

I found AF generally stayed locked using AF-S to just beyond the outer edge of the viewfinder AF rectangle, and interesting to just inside the inner edge of the next AF point in AF-C mode - though there could be some gaps.

AF sensitivity is clearly not restricted to the centre of a D800/D850 AF point, and also covers a distinctly wider segment of the viewfinder in AF-C than in AF-S.,

Back to the D800, I used this technique to confirm (on numerous other photographers D800 bodies) that there was narrower AF point detection width with either left or right outer vertical column AF points than with the inner AF points.

The photographers I helped were generally satisfied that a left as well as a right detection limitation was a camera design issue and not an AF fault.

I appreciate your view is different to mine.

(snipped)

Thom Hogan said:

Again, given that it's every twelfth row and every blue photosite, yes, the Z7/Z8/Z9 have more (and smaller) focus data points than a Zf/Z5/Z6.

Click to expand...

What is not easy to establish is how many Z AF sites are needed to satisfactorily detect say an eye or a bird.

What might or might not be relevant is that, when an eye is detected the very small green square over the eye is smaller than an AF rectangle displayed in Auto Area AF.

I find good technique combined with an easy for AF subject and an appropriate AF selection for the subject matter dramatically reduce the percentage of AF issues.

Very occasionally when there is good quality evidence of consistent front or back focus then many would agree a camera or occasionally a lens is likely to benefit from a service.

Leonard Shepherd said:

Thom Hogan said:

The question here was whether or not the system stayed focused on a scene in the field, not whether it was accurate on a static lab situation. Those are two different problems. And based upon my tests, I'd say that the answer is different.

Click to expand...

IMO some yes - often with some no.

If the subject or the camera/lens are not stable then what the AF is aimed at changes - and focus may not be consistent - independent of the AF system accuracy.

Similarly if an AF subject is borderline sometimes AF may lock on or off - due to subject issues rather than AF accuracy.

Leonard Shepherd said:

I would argue that the only thing the flat target testing might tell you is whether or not a component is out of tolerance.

Click to expand...

If a flat target incorporates fine resolution detail it will record, ideally with repeat testing, if there is the highest resolution with nil or some other fine tune setting.

Most photographers usually aim to achieve the highest resolution.

Click to expand...

Perhaps a thought experiment: say that some company - Sony, or Nikon, or perhaps Fuji, it doesn't matter - were able to develop a system which could focus on any flat target with 100% accuracy out to 500 meters and record the image with razor sharpness sharpness resolvable to 1000 megapixels at that 500 meters.

Now imagine that the system was also incapable of focusing on three dimensional objects.

For all of its incredible, almost unimaginable technological power, it would be altogether useless.

Focusing on flat targets in a controlled environment might provide a great deal of insight into how well a system focuses on flat targets in a controlled environment, but without good reason to think that this necessarily applies to 3D targets in the real world the value of these insights is limited.

An update - of sorts, anyways - to the original topic.

After a lot of thought and further experimentation I decided to pick up a 500pf.

Now it goes without saying that this is an altogether different class of lens and so comparing them is not exactly apples and oranges.

Still, my initial impressions are not so much about sharpness of the 500 pf vs my other lenses, but something relevant here, which is that at least on one day with pretty poor weather conditions and subjects that didn't make themselves especially easy targets for an AF system, I was been able to acquire focus and get some shots with reasonable sharpness, even perhaps impressive for the situations.

I would say that this is relevant to the discussion as it stands because while comparing the image quality of the 500pf to the 180-600 may be unreasonable, if we're concerned with whether one handles focusing better especially in more challenging conditions or of more challenging targets, at least one day's worth of experimenting suggests that it has been - and that's with the same user applying the same methods and techniques.

Leonard Shepherd said:

Leonard Shepherd said:

This is how we discovered the D800 left focus problem, for instance.

Click to expand...

I am not generally part of "we"!

It is easy to demonstrate the outer AF columns of the D800 have AF detection limited to central and inner AF sensitivity, unlike the more central AF columns that generally have both left and right as well as central sensitivity.

Click to expand...

Let's not rewrite history. The D800 problem was a manufacturing issue, where the rig that programmed about one-third of the early cameras put the wrong position table in for the left focus cluster. This is different than the expected issue of central on-axis accuracy being higher than the left/right clusters.

Leonard Shepherd said:

While there seems to have been the occasional defective D800

Click to expand...

Again, let's not rewrite history. I had pretty substantial data on this at the time, and was able to figure out that it was a single machine (Nikon had three) that was programming the table incorrectly. This is in a calibration step to lock in what the actual position of the focus clusters is relative to the film plane, so if you put a wrong value in that table, you get a misfocus because it thinks the film plane is in the wrong spot.

Leonard Shepherd said:

my methodology indicated the underlying issue was highly likely due to limited AF detection ability; either outside (left) of a left outer column AF point, or outside (right) of a right AF detection column.

Click to expand...

Again, in the DSLR phase detect system, the outer areas are less reliable than the central one for a number of reasons. This is particularly true for fast lenses with spherical aberration.

Leonard Shepherd said:

What is not easy to establish is how many Z AF sites are needed to satisfactorily detect say an eye or a bird.

Click to expand...

That's because we're not dealing with rectangles any more.

Leonard Shepherd said:

What might or might not be relevant is that, when an eye is detected the very small green square over the eye is smaller than an AF rectangle displayed in Auto Area AF.

Click to expand...

This is not the actual focus area being used. The reason the rectangles change size and color is for feedback as to what is happening. For instance, with a person with subject detect active, there can be four different sized rectangles that tell you something: outer rectangle is the AF-area one, and would indicate "human not seen"; next largest would be "human body"; next largest is "human head"; and smallest is "human eye." As far as I can tell, these boxes are centered on the detected area, but don't necessary say exactly where the focus is placed.

As bandwidth increases inside our cameras, we'll get better information, I'm sure. As it stands now, Nikon and Sony are prioritizing differently while using the same basic underlying phase detect system. Sony is better at showing you the absolutely current focus area (but the subject could have moved slightly) while Nikon makes sure it follows the subject and may have left a slightly misleading focus area in the display. To the user, the Sony looks like it is nailing the focus but isn't, the Nikon doesn't look like it's nailing the focus, but is. This is a very small nuance that a lot of folk don't see, but I've dealt with enough A1 and Z9 continuous sequences now to be able to see it.

Leonard Shepherd said:

I am not generally part of "we"!

It is easy to demonstrate the outer AF columns of the D800 have AF detection limited to central and inner AF sensitivity, unlike the more central AF columns that generally have both left and right as well as central sensitivity.

Some reported they found the D800 did not AF to the left of centre of an outer left AF point, and many more reported the D800 did not AF to the right of centre of an outer right AF point.

How many checked both left and right outer columns seems not recorded.

I did

While there seems to have been the occasional defective D800, my methodology indicated the underlying issue was highly likely due to limited AF detection ability; either outside (left) of a left outer column AF point, or outside (right) of a right AF detection column.

Click to expand...

That really wasn't the case. The issue was a "programming" error on the line for some bodies that left the factor

Leonard Shepherd said:

Back to the D800, I used this technique to confirm (on numerous other photographers D800 bodies) that there was narrower AF point detection width with either left or right outer vertical column AF points than with the inner AF points.

Click to expand...

Also not the case

Leonard Shepherd said:

Leonard Shepherd said:

Again, given that it's every twelfth row and every blue photosite, yes, the Z7/Z8/Z9 have more (and smaller) focus data points than a Zf/Z5/Z6.

Click to expand...

What might or might not be relevant is that, when an eye is detected the very small green square over the eye is smaller than an AF rectangle displayed in Auto Area AF.

Click to expand...

You're confusing the "focus boxes" (and what they are intended to convey) with the actual focus arrays. That can lead to a bit of confusion

If this is caused by VR + High shutter speed at times can they fix this with a lens firmware update?

Thom Hogan said:

Let's not rewrite history.

Click to expand...

Not re-writing history - I tested over 30 D800 of various ages for other photographers - and did not find one that performed better on outer left AF compared to outer right AF,.

Leonard Shepherd said:

Thom Hogan said:

Let's not rewrite history.

Click to expand...

Not re-writing history - I tested over 30 D800 of various ages for other photographers - and did not find one that performed better on outer left AF compared to outer right AF,.

Click to expand...

Again, we're talking about different things.

To this day I still encounter someone buying a used D800 and finding that it has the left focus sensor issue. Your "different ages" modifier basically changes the statistical analysis to one where you couldn't accurately predict from only 30 samples. The D800 had something well over 300k units made. It was only about a third of the first 25k units that had the issue.

If itcwas atmospheric, the issue would have appeared on all images. Same with decentered elements. It is not the copy of the lens as the bad results would have appeared on all images.

I am facing the same issue only with sport vr (in camera) with longer bursts. First or second image are super sharp while the next few (2-4) are blury then 1 super sharp then few blury and so on.

It is less likely to be AF issue as it does not happen when the vr is off or in a different mode. I am excluding the shuttercspeed as all the above is happening with enough light and high shutter speed.

My only coclusions are that the sport in camera VR is not working properly all the time, or that it cannot cope with the frame rate. Another thing might be an interference between AF and VR in high rate burst mode that not well thought of.

I agree with the topic derailing statement you mentioned earlier and it can be very annoying, especially when people do not know when to stop and keep on going in the same direction.

This was the only thread I was able to find that addressed my observation.

Quick fix: deactivate vr when you have enough light.