Nikon D850: Announcement of an Announcement of Development

[shadowblade]

Actually... the Nikon 24mm T/S is sharper than the Canon 24mm in the center and with the typical amount of modrate tilt used for landscape work. Where it falls short is with large amounts of shift.

It is an excellent landscape lens but not a very good architecture one.

The lens is old and is likely to be replaced soon. Based on the 19mm level of performance I am not worried.

The other Nikon T/S (45mm and 85mm) are excellent and totally comparable to the current generation of Canon T/S but also due for a replacement. I would expect the rumored Canon new T/S to be excellent indeed.

Cheers,
Bernard

Actually, wide-angle tilt-shifts are used in landscape photography more for the shift than the tilt, particularly in panoramic shots. At 24mm, it's not hard to get everything you want in focus without any tilt; however, stitching a rotational panorama with the angle of view of a 14mm lens (what you get when you shift-stitch a 24mm tilt-shift) requires a lot of distortion and subsequent loss of detail. So shift-stitching becomes a good option, provided you have a tilt-shift that's sharp enough when shifted.

At longer focal lengths, tilt becomes a lot more useful for getting everything in focus; at the same time, shift is far less needed for landscapes, since stitching a rotational panorama with a narrower angle of view introduces far less distortion, while the centre of a lens is going to be much sharper than the shifted edges of a tilt-shift.

Then there are cityscapes, which contain lots of strong verticals needing to be made parallel...

The PC-E 19 has a leg up on the TS-E 17 most likely because it's 7 years newer and 2mm longer (so a less extreme design). 7 years is a very long time in lens design.

The current Canon TS-E 45 and 90 lenses are literally antiques from the film era, not designed with 50MP sensors in mind and long overdue for a replacement. The Nikons were released in 1998, 17 years later, when the A900 and D3x were around.

[shadowblade]

I will not speculate about the dynamic range of the D850 this way... for starters the D500 sensor is an ASP-C sensor... just for being smaller has less dynamic range, and then, it is a sensor optimized for speed. If Nikon is still making D850 base ISO 64 it is more optimized for DR.


Michael, my theory it is that the D850 will use at least the same technology as de A7r II an BSI CMOS, that collects more light and has lower noise than the previous technology used in the D810. A7r II has a close dynamic range as the D810 (in different analysis... ), I will be very surprise if the D850, if ISO 64 is true, does not have more DR than the D810.

Bear in mind, though, that the D810 has the same DR at ISO 64 as the D800e and A7r at ISO 100. This continues all the way up the curve - at any given ISO, the D810 has about half a stop less DR than the other two, while, for any given DR, the D810 needs to be set at an ISO around half a stop lower than the other two. It only makes up for this at low ISO by having a base ISO around half a stop lower than the others.

In other words, the lower ISO of the D810 does not equate to more DR. I'm not sure if it's used to gain more colour depth or something else, but not DR.

The D850 also has a base ISO of 64. Who knows if that's a real ISO or a pulled ISO 100 shot? If it's a real one, does it sacrifice high-ISO DR for this (which would take away from its utility as an action camera), or does it at least match the D5/A9/A7r2 at ISO 3200-12800?

[BernardLanguillier]

Bear in mind, though, that the D810 has the same DR at ISO 64 as the D800e and A7r at ISO 100. This continues all the way up the curve - at any given ISO, the D810 has about half a stop less DR than the other two, while, for any given DR, the D810 needs to be set at an ISO around half a stop lower than the other two. It only makes up for this at low ISO by having a base ISO around half a stop lower than the others.

In other words, the lower ISO of the D810 does not equate to more DR. I'm not sure if it's used to gain more colour depth or something else, but not DR.

This is simply not correct.

Cheers,
Bernard

[BernardLanguillier]

Actually, wide-angle tilt-shifts are used in landscape photography more for the shift than the tilt, particularly in panoramic shots. At 24mm, it's not hard to get everything you want in focus without any tilt; however, stitching a rotational panorama with the angle of view of a 14mm lens (what you get when you shift-stitch a 24mm tilt-shift) requires a lot of distortion and subsequent loss of detail. So shift-stitching becomes a good option, provided you have a tilt-shift that's sharp enough when shifted.

Not sure how to put it, but frankly you don't seem to have much first hand experience with spherical stitching... nor with the value of tilt for landscape images, be it with a 24mm lens.

I'll keep it a that.

Cheers,
Bernard

[shadowblade]

Not sure how to put it, but frankly you don't seem to have much first hand experience with spherical stitching... nor with the value of tilt for landscape images, be it with a 24mm lens.

I'll keep it a that.

Cheers,
Bernard

'Not much first-hand experience'?

Each one of those probably accounts for a third of all the shots I've ever taken. I also print huge sizes (1x3m is common), so it's not like I don't care about sharpness/resolution or focus either. I know what spherical stitching can and can't do. When stretching it into a UWA rectilinear shot, I've seen the stretching and blurring of corners (you don't get that if you use a cylindrical or spherical projection, but those don't keep straight lines straight). It's simple geometry.

As for tilt at 24mm for landscapes, you probably lose less resolution stopping down to f/10 or so if you really need the DOF vs tilting it. You need it at longer focal lengths, since even stopping down to f/14-f/16 won't give you enough DOF, but what are you trying to shoot at 24mm that calls for that, even when you're very close to the ground?

Some rotational panoramas attached.

[davidgp]

'Not much first-hand experience'?

Each one of those probably accounts for a third of all the shots I've ever taken. I also print huge sizes (1x3m is common), so it's not like I don't care about sharpness/resolution or focus either. I know what spherical stitching can and can't do. When stretching it into a UWA rectilinear shot, I've seen the stretching and blurring of corners (you don't get that if you use a cylindrical or spherical projection, but those don't keep straight lines straight). It's simple geometry.

As for tilt at 24mm for landscapes, you probably lose less resolution stopping down to f/10 or so if you really need the DOF vs tilting it. You need it at longer focal lengths, since even stopping down to f/14-f/16 won't give you enough DOF, but what are you trying to shoot at 24mm that calls for that, even when you're very close to the ground?

Some rotational panoramas attached.

I have already told you before... but I really like your work


http://dgpfotografia.com

[shadowblade]

This is simply not correct.

Cheers,
Bernard

Sure. Don't believe me. See for yourself: http://www.photonstophotos.net/Charts/PDR.htm#Nikon%20D800E,Nikon%20D810,Sony%20ILCE-7R

A7r at ISO 100 - 11.71 stops
D800e at ISO 100 - 11.45 stops
D810 at ISO 100 - 11.06 stops

Dialling the D810 down to ISO 64 gives you 11.6 stops - in between the values for the other two at ISO 100.

Are you saying these PDR measurements are wrong?

[BernardLanguillier]

Sure. Don't believe me. See for yourself: http://www.photonstophotos.net/Charts/PDR.htm#Nikon%20D800E,Nikon%20D810,Sony%20ILCE-7R

A7r at ISO 100 - 11.71 stops
D800e at ISO 100 - 11.45 stops
D810 at ISO 100 - 11.06 stops

Dialling the D810 down to ISO 64 gives you 11.6 stops - in between the values for the other two at ISO 100.

Are you saying these PDR measurements are wrong?

I am looking at this: https://www.dxomark.com/Cameras/Compare/Side-by-side/Canon-EOS-5D-Mark-IV-versus-Sony-A7R-II-versus-Nikon-D810___1106_1035_963

Cheers,
Bernard

[BernardLanguillier]

'Not much first-hand experience'?

Each one of those probably accounts for a third of all the shots I've ever taken. I also print huge sizes (1x3m is common), so it's not like I don't care about sharpness/resolution or focus either. I know what spherical stitching can and can't do. When stretching it into a UWA rectilinear shot, I've seen the stretching and blurring of corners (you don't get that if you use a cylindrical or spherical projection, but those don't keep straight lines straight). It's simple geometry.

Nice images, thanks for sharing.

Yes, there is of course some stretching happening when using flat projection, but the image quality you can reach in corners is still significantly superior compared to what you get with a T/S shifted all the way, be it with the Canon 24mm T/S.

This does of course depend on the focal length used to so the spherical stitching. I am typically using focal lengths longer than 55mm, often 100 or more (up to my 400mm f2.8 in fact).

A recent example shot with the H6D-100c and a 100mm, which would correspond to a 60mm in 35mm terms.



You'll find hundreds of other examples here: https://www.flickr.com/photos/bernardlanguillier/albums/72157600916381270

Cheers,
Bernard

[bclaff]

Sure. Don't believe me. See for yourself: http://www.photonstophotos.net/Charts/PDR.htm#Nikon%20D800E,Nikon%20D810,Sony%20ILCE-7R

A7r at ISO 100 - 11.71 stops
D800e at ISO 100 - 11.45 stops
D810 at ISO 100 - 11.06 stops

Dialling the D810 down to ISO 64 gives you 11.6 stops - in between the values for the other two at ISO 100.

Are you saying these PDR measurements are wrong?

I am looking at this: https://www.dxomark.com/Cameras/Compare/Side-by-side/Canon-EOS-5D-Mark-IV-versus-Sony-A7R-II-versus-Nikon-D810___1106_1035_963

Cheers,
Bernard

You both may find this interactive Photographic Dynamic Range (PDR) versus DxOMark Landscape Dynamic Range scatter chart interesting. (static image attached).
PDR at PhotonsToPhotos is a different measure than the Landscape score at DxOMark.
I have always contended that PDR is more meaningful.
For example, DxOMark places the D7200 above the D600, D610, D750, D800, and D800E whereas PDR does not; do you really think that excellent APS-C sensor outperforms all those FF ones?

[BernardLanguillier]

Yes, I do having used a D7200.

Incredibly clean shadows.

Cheers,
Bernard

[shadowblade]

Nice images, thanks for sharing.

Yes, there is of course some stretching happening when using flat projection, but the image quality you can reach in corners is still significantly superior compared to what you get with a T/S shifted all the way, be it with the Canon 24mm T/S.

This does of course depend on the focal length used to so the spherical stitching. I am typically using focal lengths longer than 55mm, often 100 or more (up to my 400mm f2.8 in fact).

A recent example shot with the H6D-100c and a 100mm, which would correspond to a 60mm in 35mm.

You'll find hundreds of other examples here: https://www.flickr.com/photos/bernardlanguillier/albums/72157600916381270

Cheers,
Bernard

It depends on the total angle of view rather than the angle of view/focal length of the individual frame. A panorama with a narrow angle of view (e.g. the monastery in my post, consisting of two stitched 420mm shots) requires almost no distortion. Something covering a very wide angle of view requires a lot of distortion, like the edges of a rectilinear UWA lens.

A shift-stitched panorama from a 24mm tilt-shift is going to have sharper corners than a rotational panorama of the same angle of view taken using a 24mm lens, or even a 35mm lens. You can probably get as much corner detail  (and more central detail) using a 50mm lens, but that requires many more frames and much larger file sizes.

Also, rotational techniques complicate the use of filters  (especiall GNDs and panoramas).

With a 90mm lens, you wouldn't shift-stitch, unless you specifically needed to do so due to filters or to deal with parallax (nodal offset systems being clumsy, heavy and usually unnecessary, except when dealing with very close foreground elements that are going to show up in more than one frame of the panorama). At the angle of view subtended by a 90mm lens shifted fully in each direction, there would be little rectilinear distortion required to rotationally stitch the same image, and you'd probably get sharper results just keeping the lens in the centre position and rotationally stitching the images.

[BernardLanguillier]

Ok, we agree.

For me stitching isn't just about aspect ratio, it is about added resolution also, which means that nearly never stitch with lenses wider than 55mm.

And 95% of the thousands of stitches I have made in the past 10+ years were with a dedicated pano head. It decreases significantly the issues with parallax and philosophically increases the "correctness" of the image. But I understand most would consider this attention to details overdone and pointless.

Cheers,
Bernard

[shadowblade]

Ok, we agree.

For me stitching isn't just about aspect ratio, it is about added resolution also.

And 95% of the thousands of stitches I have made in the past 10+ years were with a dedicated pano head.

Cheers,
Bernard

I look at it in terms of effective sensor area and pixels per inch at final print size.

Shift-stitching using a lens that can go 12mm in each direction gives me an effective sensor area of 24x60mm horizontally, or 36x48mm vertically. This gives an effective sensor width equal to or greater than Phase One digital backs (since most of my shots are between 1:2 and 1:3 aspect ratio, the height limitation doesn't come into play much). A 1.4x TC increases this to 24x70mm horizontally or 36x58mm vertically, although obviously at a different focal length. A single frame from a modern MFDB contains a lot of detail - as much as a large-format film shot - and is suitable for printing at a large size.

PPI is another issue. Shift-stitched horizontally (without TC), an A7r2 gives me enough pixels for 150ppi at almost 90" print width. A 60MP 24x36mm sensor would give me enough for around 105", or 200ppi at greater than 72" width  (for a 24x72" print), while 70MP would give enough for 115".

For a shift-stitched panorama, the effective sensor area and PPI are evenly distributed - there is as much in the corner as thete is in the centre.

Effective sensor area and PPI for rotational panoramas are a bit more difficult to calculate, depending on the angle of view of the final image, the projection (rectilinear, cylindrical or spherical) as well as the focal length of the lens. Not only that, but it's unevenly distributed - there is more sensor area and more PPI contributing to the centre of the image than the corners. You can calculate what focal length you need to shoot at (and how many frames it will take) to get the same or more PPI and sensor area contributing to the corners on a rotational panorama as compared to a shifted panorama.  The sharpness aspect is somewhat mitigated by the fact that the rotational panorama uses the sharper part of the lens than the shifted edge of a tilt-shift, but the effective sensor area and its implications in terms of noise still apply. If you don't have enough sensor area contributing to the corners, if they're stretched from too small an area of sensor, you'll have noisy corners, regardless of how sharp the lens you shot it with is.

[Bart_van_der_Wolf]

Ok, we agree.

For me stitching isn't just about aspect ratio, it is about added resolution also, which means that nearly never stitch with lenses wider than 55mm.

I agree (with some reasoned exceptions) to the higher resolution aspect of stitching.

And 95% of the thousands of stitches I have made in the past 10+ years were with a dedicated pano head. It decreases significantly the issues with parallax and philosophically increases the "correctness" of the image. But I understand most would consider this attention to details overdone and pointless.

Indeed, but count me out on some (!) of the 'correctness'. It's the resulting image, and whether it relays my creative intent, that matters.

Attached, an up-close (for perspective) hugely reduced 2-row, 6-image, TS-E 45mm Pano-shot (using RRS multi-row pano gear) with my Canon camera+lens (tilt used for focus plane), to retain the wider angle-of-view look that's inherent from looking at the resulting image from too far a viewing distance (intended to add to the mystery/drama) of the scene, one of the several different ones that I shot in an urban exploration (Urbex) environment of an abandoned (now demolished) warehouse in Antwerp.

Likewise, I've shot a number of hyper-resolution Pano images and attached a small crop of a larger 'Autumn soil-texture' image (with a 135 mm lens on a 24x36mm sensor).

Cheers,
Bart

[BernardLanguillier]

Bart,

Nice image from my home country.

Sorry, not 100% sure I got your point nor how you captured the image though.

Is that a spherical pano or a flat stitch done by shifting your T/S lens?

I am also not too sure what all this has to do with the D850...

Cheers,
Bernard

[Bart_van_der_Wolf]

Bart,

Nice image from my home country.

Sorry, not 100% sure I got your point nor how you captured the image though.

Is that a spherical pano or a flat stitch done by shifting your T/S lens?

Rotational stitch, using the lens' tilt capability for improved focus plane resolution.
 

I am also not too sure what all this has to do with the D850...

None whatsoever, as far as the D850 is concerned (since it's not available yet ),  but everything as far as stitching for resolution is concerned.

My point being that stitching offers ultimate freedom in composition/resolution, if the subject's (lack of) motion allows.
 
Cheers,
Bart

[BernardLanguillier]

Rotational stitch, using the lens' tilt capability for improved focus plane resolution.

Ok, clear, thanks.

This is indeed exactly how I have been using my T/S lenses also.

Cheers,
Bernard

[BernardLanguillier]

https://nikonrumors.com/2017/08/12/nikon-d850-specifications-recap.aspx/

Cheers,
Bernard

[Jim Kasson]

A shift-stitched panorama from a 24mm tilt-shift is going to have sharper corners than a rotational panorama of the same angle of view taken using a 24mm lens, or even a 35mm lens.

Not my experience with the Nikon 24 T/S.

.http://blog.kasson.com/the-last-word/stitched-panos-slide-or-spin/

Jim