This is a continuation of a series of posts on the Nikon D850. The series starts here. You should be able to find all the posts about that camera in the Category List on the right sidebar, below the Articles widget. There’s a drop-down menu there that you can use to get to all the posts in this series; just look for “D850”.
A while back I wrote this post about the accuracy of manual focusing with the D850 using the “focus confirmation” feature, aka the green dot (it’s not green in the D850), aka “the meatball”.
A two word summary of the accuracy and repeatability of this method: not very.
In researching the presence or absence of D850 PDAF compensation for focus shift, several people introduced me to an AF tuning procedure that uses the meatball. I tried it, and initially, it didn’t work for me; I focused manually and saw the meatball all lit up with the AF focus trim set from -20 to +20. I turns out the Nikon engineers have a (insert pejorative adjective here) trick up their collective sleeves: in manual focus mode, the meatball tolerance is deliberately widened so that it’s easier to get it to light up.
For people who care about manual focusing accuracy, this is a disaster, and it explains part of the rotten results I got when I tried to use it under controlled conditions.
But there’s a workaround. It has been known to the Dot Tune inventor for quite some time but is new to me. It may be new to you, too, and it has implications that extend far beyond Dot-Tune users. It affects everyone trying to us the meatball for accurate manual focus.
Here’s the trick:
- Set the camera in AF mode.
- Set the lens in AF mode.
- Set the camera up for back-button focusing only; defeat the option that allows the camera to focus with a half-press of the shutter release.
- Half-press the shutter release to see the arrows and the meatball.
- Focus until the arrows go out and only the meatball remains.
- Press the shutter release all the way down.
With the 105 mm f/1.4 lens, this makes the meatball tolerance — the range where the meatball will stay on — a bit less than 10 AF trim units. That’s not very low, but it’s a heck of a lot lower than the more than 40 AF trim units when the camera is in MF mode. You’ll still have to deal with the vagaries of the phase detection mechanism, and there seems to be no correction for focus shift, but you’ll be a lot better off than using the meatball in MF mode.
It’s hard to say for sure what the camera does when you affix a lens that doesn’t support autofocus, but it appears to me that the camera switches to the loose-tolerance mode for the meatball in that case.
I still say, with the D850, stick with live view if you want accurate manual focusing.
As a manual lens user predominantly this is of interest but is it not the case that the “meatball” “focus indication is derived using only rays within the f/5.6 (or f/8) aperture circle of the lens.”
Marianne
https://www.dpreview.com/forums/post/60264302
And given the DOF at those apertures the tolerance change you note when focussing from 10 to 40 AF units does seem remarkably large, but if focussing a 50mm f1.2 manual to shoot wide open a split screen aid would seem to be a better choice (availability of such screens on DSLRs noted, the F6 does have one) especially if the lower precision mode is only enabled, could you spare the time to verify that, manual lens shooters would be obliged, well this one would.
Yes. Marianne has demonstrated that well. However, I believe that split-image ground glass focusing aids are similarly handicapped. Not sure; just vaguely remember that. I don’t have a camera that has such a feature. I don’t believe I’ve seen it on a digital camera. Do any digital cameras still have interchangeable ground glass?
Some third parties will fit/supply screens for film bodies (still available as parts) that are a direct switch or sometimes “trimmed” or mount swopped but it’s not popular as it throws away any warranty and results seem variable, I suspect shimming is often required and rarely performed. (as the screen is not intended to be changed the manufacture may include a shim set up step as the body is not intended to be accurate with a different screen unlike film SLRs where changing screens required the body to be in spec for any screen) Katz eye did them, the D800 I know was swappable, the screen being held in place by a spring clip, the Df screen is exchangeable with F6 screens I understand, experience is mixed: https://www.dpreview.com/forums/post/51149976 for example.
On split image as I understand them the lower limitation is only that of the screen brightness as they are central and not “fed” light other than from the main mirror so light loss is minimal and blackout occurs at around f5.6/f8 depending on the angle of view of the eye and the split prism construction. (The F6 split is much smaller than in in say the F2/F3 screens and rather hard to use being overlaid with autofocus markers).
If manual focussing I prefer to ignore the meatball personally as the wide tolerance with faster lenses is clearly visible even on a DSLR screen, focussing using the screen only without aids, bear in mind my current DSLrs
Take a look at what Doug Kerr has to say about the split image device and tell me how it’s essentially different from the phase detection mechanism in PDAF.
http://dougkerr.net/Pumpkin/articles/Split_Prism.pdf
I’m no optical expert, but it seems the principle is the same, and it should have an “effective aperture” like PDAF.
Am I wrong in this thinking?
Jim
Hello Chris
Im having issues with my D850 manual focusing… Im back focusing, where my D800 this happened with one lens 80-200 f2.8 but not with my 24-70 f2.8.
Do you have any advice to setting up the focus for the D850, or any adjustments and tests I should do. Have done the flat wall test, its when I shoot in flat light or very bright front light I notice the problem most. Yes it could be me, but Im not so sure, as the D800 gave me very good results. Any advice is welcome. Thank you
I would always hesitate to call you wrong.
I have seen that paper elsewhere as well as at that link and think this statement is relevant:
Pg10 When it doesn’t work
“The previous figures showed the lens set at essentially its maximum aperture, and
in fact that is the normal situation when in the viewing/focusing mode. The
proportions also suggest that the maximum aperture of this particular lens is fairly
large (fairly small f/number).”
To me that implies the system works at “essentially the maximum aperture” why essentially is added is of course relevant as a qualifier and I don’t see that explained there, but see later.
The expansion of that section is of the blackout that occurs as the lens is stopped down ( and kills the split image) which my experience is around f5.6/f8 which is precisely where the meatball does work, so to my reading the split image must work at apertures wider than the meatball.
Also as I mentioned previously the split image gets the maximum light available from the lens, the AF (meatball) relying on that proportion “stolen” from the viewing screen to feed the AF detector array which must leave that system at a disadvantage. This is where I think Marianne’s effective f5.6/f8 figure comes from, that being the effective aperture the AF array sees in the light it does get. So IF I am correct with say an f2 lens the split image “sees” f2 (less glass transmission and flare light loss and loss to the AF array) which may = “essentially” full aperture) and the AF array/meatball sees f5.6/f8 which is all the secondary mirror system passes.
I see that there is confusion in using f stops here in two ways: as we understand them optically in changing depth of field which must affect focus perception and as indicators of purely the amount of light in the system as we use them in adjusting for exposure. So my take is Marianne refers to the light (photons) reaching the AF system being as if the lens is at f5.6/f8 and not the depth of field the AF system sees as being that of a lens system stopped down to that aperture.
Of course I am open to being completely wrong in my interpretation.
Chris, I interpret Doug’s article differently. I think there is are off-axis bands of ray bundles that do the work in both systems. The f-stop where the split image begins to get dark is an indication of how far off-axis. I could be wrong, too.
My interpretation of Marianne’s statement is also different from yours, and has to do with ray angels, not photons.
I can certainly see the other POV and “at the end of the day” we both know the “meatball” sits just this side of useless (perhaps not even that good on higher MP) ,the split image is for all practical purposes dead on a DSLR (or MFD) and on chip is the future so it’s academic. Thank you for the discussion , the continued great work and have a fantastic year and I trust better health this time around.
I’d be interested to find out if you tried the ‘dot tune’ method on your D850 with the 105mm at f/1.4 and what result did you get in comparison with the LensAlign FocusTune system you initially used?
FocusTune says +6 at f/1.4 and -4 at f/5.6.
I am getting upper limits of +10 and lower limits of 0 for solid meatball, at both f/5.6 and f/1.4 with the D850, with Dot-Tune. Averaging, DotTune says +5 at both f-stops.
If there is some correction for focus shift, it is coming later in the chain than the point that the decision is made whether to light the meatball or not.
So just to confirm.
There is absolutely no way to get good accuracy with say a modem MF lens such as Milvus?
Depends on your definition of “good”. The only way to get the best out of such a lens is to use live view and manual focusing.
That’s unfortunate. A lot of users had reported they found the focusing improved on the D850 over its predecessors after tuning for MF lenses.
But you are saying that is not the case. Basically one could not get by with using just focus confirmation expecting to have a good hit rate. That’s s shame. Would love to have the Milvus 35mm as a walk around.
I bought three D850 bodies within the first two weeks it was released. One of them from a NPS member who ended up with a spare body. All three of them showed identical behavior with the “dot”, and remarkably better than my D810, so I thought everything was ok.
Until at a trip the week after, my daughter got 3 pics of a distant boat with the D810, I immediately attached the same lens on my D850, and got 3 less sharp pics than hers (70-200 f4 VR). Her 3 pics were identical to each other in sharpness, and my 3 pics were identical to each other in sharpness (compared to hers, mine were all slightly out of focus).
I never fine-tuned the D810 with any lens (nor the D850). Considering also that I was shooting raw, while her D810 was accidentally set on jpeg (!!!). Similar behaviors were apparent throughout our trip, with the most disturbing being the very choppy sea surface with the D850, which was rendered more clear and smooth with the D810. Other than that though, the default WB settings of the D850 proved to be more brilliant in the SOC images. A manually focused Zeiss 25/2 produced stunning results with the D850 and out of this world resolution, so I know the camera itself (the sensor etc), besides the AF system, was ok.
I’d still keep the D850 because most of my lenses are manual so I get to blame myself, but I will have to fine tune the AF lenses, at least around the aperture values I prefer, due to focus shifting, and get used to multiple shots of critical subjects. Carrying around 2+ lbs Milvus lenses and large loops is not an option: I’m not going on the street looking like a cinematographer, just to take about the same pics a Leica Q would have taken without all this mess and weight.
Nikon must do something to compensate for this focusing accuracy, at least for its native expensive lenses, and tighten the tolerance of the meatball dot via firmware. It’s embarrassing to go to a middle school volleyball shoot with the 70-200E and get less sharp pics than the guy with the old Canon or a Sony 6500 with a kit lens or (even worse) a lucky iPhone 7 plus. Because the performance of the D850 in the prosumer market will determine Nikon’s mirrorless sales next year. Actually, all Nikon’s sales next year…
I’m horrified. I had no idea this was going on. Not only that but it seems to me that when I am manually focusing with my 70-200 E FL, brand new, that just by eye it looks great in the finder but then is not the point in focus in the image. I had a shoot in a church last night and shot manual focus at 2.8 to begin with then switched to AF. Almost every manual focus image is not in focus. I used visual, as well as at times the meatball. Luckily I overshot. But I lost a ton of good work.
I replaced my D610 focus screen with a split screen made in Taiwan. I shoot with only Zeiss manual and Nikon classic tele lenses. Just bought the classic Nikon 300mm F2.8 EDIF , shot a few quick shots and am totally blown away with the images. I’ve been holding off purchasing the D850 but am planning on purchasing one. My buddy shoots with a Cannon and I’m envious of Cannon’s Live View focusing – blows away Nikon’s Live view. But he is more envious of Nikon’s ability to use classic manual focusing lenses with the modern digital bodies.