Nikon Df and non-AI lens overexposing

As far as I know the non-CPU lens data affects only the EXIF information. If you forget to select the correct lens, or use an unlisted lens, the exposure does not vary. Since the listed aperture for a preset lens will be only what you set, and not what you use, it's really only useful if you need to be reminded later what lens you were using.

 

=

I don't pretend to know the actual "magic" that makes it work, but matrix meter DOES in fact need to know the maximum aperture of the lens when metering at full aperture. All Nikon lenses factory built to AI specifications(including AF and AF-D lenses, but not G lenses) have a lug that gives information on the maximum aperture of the lens. There are only a handful of cameras that read this, and only two that I know of-the FA and F4-that offer matrix metering based on the lug. In fact, when a lens without the lug(specifically AI converted lenses), the camera reverts to CWA metering.

CWA with AI lenses does work on the principal of "stops from wide open", and in fact most manual focus AI cameras have no way of knowing what the actual maximum aperture of the lens is(non-AI TTL metering is a bit different since the coupling shoe is always at f/5.6-this is why the non-AI cameras that most folks are familiar with require turning the lens to the minimum aperture after mounting-some really early coupled TTL cameras/prisms require that the maximum aperture be manually set).

As far as I know the non-CPU lens data affects only the EXIF information. If you forget to select the correct lens, or use an unlisted lens, the exposure does not vary. Since the listed aperture for a preset lens will be only what you set, and not what you use, it's really only useful if you need to be reminded later what lens you were using.

 

=

I have to do a test to show you but I did have the wrong data (or selecting the wrong lens) and the exposure was off.

Without meaning to sound uncaring, just use the Histogram of a test shot.

PC lenses are rarely used handheld in one-off, speed dependent situations.

...and yes I know it's based on the JPEG data even if your shooting RAW, but the amount 'out' it can be is no more than traditional metering errors.

EDIT. Does the Df not have Live Histogram?

I was told by a Nikon rep years ago that aside from EXIF, the "Non-CPU Lens" menu also tells the meter how to interpret movements of the AI follower for proper exposure with each manual lens. As usual with most direct Nikon communications I took for granted, this has proven wrong. Just tested mine, and changing the menu from f/1.4 to f/4.0 does not influence the meter reading at all with an f/1.4 lens mounted (it merely offsets the aperture selection reported in the finder display and EXIF by three stops). Interesting.

Put it on matrix and take it out to the sun where it's bright. You should see the difference.

Well, we learn a new thing every day. I never tested it exhaustively, but it does seem to differ, though the amount of variation is odd and not entirely predictable.

With a 55/3.5 AI-converted lens set at F8, with matrix metering, and using the different non-cpu settings I have, there was variation visible, though not enormous. Enough to make a little difference, though. Using spot metering, the difference was less, and very little difference occurred between one setting and those near it. Varying just the aperture resulted in a small but fairly consistent raising of exposure as the max aperture went up, while varying the focal length did less.

Switching to a 50/F2 AI lens at F2, it was different. Here, in spot metering, changing aperture alone resulted in a very small and consistent increase in exposure. But using matrix metering, F2 and 2.8 settings were identical. F4 had identical shutter speed but different white balance and slightly different histogram ( with set, not auto WB ?!?!?), and F5.6 jumped over a stop in shutter speed. F8 was identical.

Then I tried a converted 85/1.8 lens. It was the same at 1.8 and 2.8, a bit faster at 4, back down at 5.6, and then it jumped at F8.

My take on this is, first that I was wrong and BeBu right, but also that if you forget to set the correct lens you probably won't notice a difference unless the max aperture is more than a couple of stops off, and you may never be able to predict exactly how it will behave. If you switch from a slow telephoto to a fast normal lens it will probably show. If you switch between lenses that are fairly close it probably won't. But I also wouldn't be surprised if different cameras behave differently.

To an extent, I'd love Nikon to describe what exactly their matrix meter is doing. Unfortunately, at least in recent times, I suspect they can't - that it's a trained neural net, and as with my general concerns about neural nets, it's very hard to analyse what it's actually doing. Nikon have long claimed it was trained by a large number of sample images, but I don't know whether this was always automatic or sometimes a case of interns tweaking variables. I assume the absolute aperture information is a combination of approximating a drop-off from vignetting and wanting to know the absolute image brightness for subject identification - although I'd have thought filters would mess with the latter.

I'm even keener to know what highlight priority metering thinks it's doing - the histogram certainly claims it's still blowing highlights, but I've not done enough experiments to determine how often the raw data is intact and the histogram is misleading. I'm planning on giving the "uni-white balance" thing a go to see whether it helps, but I suspect the meter will still have an exposure designed to allow highlight roll-off rather than actually preserving the brightest channels. ETTR really shouldn't be so hard.

To an extent, I'd love Nikon to describe what exactly their matrix meter is doing. Unfortunately, at least in recent times, I suspect they can't - that it's a trained neural net, and as with my general concerns about neural nets, it's very hard to analyse what it's actually doing. Nikon have long claimed it was trained by a large number of sample images, but I don't know whether this was always automatic or sometimes a case of interns tweaking variables. I assume the absolute aperture information is a combination of approximating a drop-off from vignetting and wanting to know the absolute image brightness for subject identification - although I'd have thought filters would mess with the latter.

 

I'm even keener to know what highlight priority metering thinks it's doing - the histogram certainly claims it's still blowing highlights, but I've not done enough experiments to determine how often the raw data is intact and the histogram is misleading. I'm planning on giving the "uni-white balance" thing a go to see whether it helps, but I suspect the meter will still have an exposure designed to allow highlight roll-off rather than actually preserving the brightest channels. ETTR really shouldn't be so hard.

Preserving the brightest part in the image is an easy thing to do but I don't think it's a desirable algorithm for a metering system designed as a one size fit all like matrix. A lot of time it's best to let the highlight goes because there are more important things in the deep shadows.

Preserving the brightest part in the image is an easy thing to do but I don't think it's a desirable algorithm for a metering system designed as a one size fit all like matrix. A lot of time it's best to let the highlight goes because there are more important things in the deep shadows.

Oh yes, I don't say I never want to use matrix metering. But a lot of the time I find myself capturing a scene whose dynamic range I know full well falls within the capabilities of my camera, and the one thing I don't want to do is lose highlights - and in fact I want the shadows elevated as high up the exposure graph as possible to maximise detail. You'd think highlight priority metering would avoid this; as far as I can tell it doesn't. (It would be easier to tell with raw histograms...)

Preserving the brightest part in the image is an easy thing to do but I don't think it's a desirable algorithm for a metering system designed as a one size fit all like matrix.

On the contrary, I think not blowing any highlights is a very good (and simple) algorithm.

Once the highlights are truly blown, in any of the 3 RGB channels, it's impossible to retrieve them, or their true colour. Whereas even with a crappy JPEG you can pull a lot back from the shadows, and if you shoot RAW at a sensible ISO speed and in a CT close to daylight, the amount of shadow detail you can claw back is pretty incredible.

That's the very reason that many people use ETTR as a matter of course. I don't know anyone that sets out to deliberately blow highlights, unless they're ridiculously bright, or you're aiming for a very high-key look.

Personally, I suspect the 'thousands of images' that Nikon supposedly trained its matrix metering with must all have had blown highlights in the first place!

Maybe Nikon's software and hardware engineers should familiarise themselves with the concept behind the acronym K.I.S.S?

When shooting RAW, about the only situation I can think of where I'd set the camera to avoid blowing the highlights is if the sun is actually in the photo-something I generally try to avoid.

Otherwise, at base ISO most of Nikon's current FX lineup(and stretching back a bit to some older cameras like the D800) has something like 14 stops of DR-there are few scenes where anything you can see with your unaided eye would fall outside of that.

On the contrary, I think not blowing any highlights is a very good (and simple) algorithm.

 

Once the highlights are truly blown, in any of the 3 RGB channels, it's impossible to retrieve them, or their true colour. Whereas even with a crappy JPEG you can pull a lot back from the shadows, and if you shoot RAW at a sensible ISO speed and in a CT close to daylight, the amount of shadow detail you can claw back is pretty incredible.

 

That's the very reason that many people use ETTR as a matter of course. I don't know anyone that sets out to deliberately blow highlights, unless they're ridiculously bright, or you're aiming for a very high-key look.

 

Personally, I suspect the 'thousands of images' that Nikon supposedly trained its matrix metering with must all have had blown highlights in the first place!

 

Maybe Nikon's software and hardware engineers should familiarise themselves with the concept behind the acronym K.I.S.S?

For back lighted subject I would generally let the highlight in the background goes. I much rather have good exposure in the shadows that is with low noises. For me many images that I have I don't care about the highlight. They can be pure white.

For me, it depends what the background is. I don't like blue skies turning white if I can avoid it, and I don't like completely (or certainly partly) blown-out clouds. The sun or a light is different, but I'd at least like to know it's happening. The dynamic range of cameras these days is good enough that you can usually still get good shadow detail - and at least being told whether to bracket by the metering system would be useful. E-TTR is supposed to be about retaining as much shadow detail as possible while still keeping the highlights.

Examples of my attempts to retain very bright detail in a very dark scene (Antelope Canyon, upper and lower) on this Nikon Wednesday thread and this one from a while back. YMMV.

So I take it the Df, doesn't have Live Histograms?

I'm always amazed how much sky and cloud detail can be brought back by using selective 'development' of RAW files in DxO. The grad tool for localised adjustment is nearly always enough to get it back to how I remember it looking...:-)

So I take it the Df, doesn't have Live Histograms?

 

I'm always amazed how much sky and cloud detail can be brought back by using selective 'development' of RAW files in DxO. The grad tool for localised adjustment is nearly always enough to get it back to how I remember it looking...:)

I would have to check the manual as I never use either live view nor histogram.