Nikon Introduces Mirrorless Z System

Discussion in 'Nikon' started by ShunCheung, Aug 23, 2018.

  1. If it has pixel level repeatable control. And if Nikon paid any patent fees that might be associated with the idea. (Pentax's use of the sensor shake to emulate an OLPF is something Nikon probably could have implemented by now, from a pure hardware perspective.)

    Despite the cost in waiting for vibration to die down, i'd quite like the feature for rare uses - though I've been trying the hand held superresolution trick (which is likely less effective) a bit.
     
  2. Pixel-shifting doubles resolution, but has little tolerance for motion. Sony EdgePro software tends to produce bizarre artifacts with blowing leaves, for example. Third party produces better results, and adds batch recognition and rendering. Hand-held shooting is unlikely to produce acceptable results. It's hard enough with a tripod and 2 second initial delay (or a remote release). I have not seen any Moire patterns, but to produce Moire, a lens must have significantly more resolution (2x or more) than the sensor. Moire would be harder to see in the absence of color bands, which pixel-shifting would eliminate.

    It's another tool in the box, but I wouldn't base a buying decision on it unless your are big into landscapes or architecture.
     
    Last edited: Sep 3, 2018
  3. Absolutely - I'd like the shifted raw files (potentially by half pixel offsets as well) for offline processing. Then you can blend layers and pick one when motion gets in the way. Still, I'd expect it to be useful even without this for macro, art reproduction (including architecture as you say), other still life, etc. I've seen (colour) moire from by D810 and D850 from buildings, although it wasn't exactly catastrophic. I'd be interested to see a two-shot approach to capture both options - you could probably automate detecting whether there's an issue. If I did more fashion photography I might hit more of it.

    The hand-held thing just involves shooting several (20 was suggested) frames at a high shutter speed to avoid blur, upscaling them by 2, then asking PhotoShop to align the layers - because its algorithm effectively works to sub-pixel alignment on the original pixel. Merge the result, then sharpen. It might help very slightly, and you have the same problems with moving subjects - and you can't use it to bypass the Bayer reconstruction as with proper shifting. But it's free and fairly convenient. It relies on being hand-held, since this is what's giving you different sub-pixel views in the first place. I've not given it a proper go yet - the nearest was when I was trying with Saturn and Mars a few Wednesdays back, but the images were small enough there that I had to use third-party dedicated stacking software to get anything.

    The inconvenience of moving subjects means it wouldn't be my primary consideration in camera shopping, but it could tip the balance for me - and I do quite a few landscapes. It's one of the main things I have mild A7RIII envy about.
     
  4. I posted and example of pixel-shifting in another thread. The inset is a 100% (original resolution) portion of the image. This is Mt. St. Helens in Washington, taken with a Sony A7Riii + Loxia 50/2 @ f/5.6 and ISO 100.

    1520537_7b01d91d82ed78bc342a92b3edb3f131.jpg

    This is the same level of detail, taken within a few minutes of the previous example, same lens, etc., but without pixel-shifting.

    _7R30843 Detail.jpg
     
    Last edited: Sep 3, 2018
    Andrew Garrard likes this.
  5. There is no substantive difference between taking three images through separate RGB filters and four pixel-shifted images from from a 2x2 pixel (GGRB) Bayer array, and combining the images algebraically. Every time you listen to a stereo broadcast, you are listening to main (sum) and side (difference) channels, which are combined, algebraically, into left and right channels, M+S and M-S. Color video is recorded as luminance (Y), blue (Pb) and red (Pr), then processed algebraically to RGB channels.

    The reasoning is similar. The first is fully compatible with single shot images, de-matrixed and interpolated from a Bayer array. The second is fully compatible with mono reception and playback. Video is encoded YPbPr to facilitate compression, reducing bandwidth for transmission and storage size.
     
  6. Agreed, Ed (although the fourth shifted exposure gives you some more green values to denoise with). Although pixel shifting can also, in some implementations, test sub-pixel offsets and gain extra resolution that way. My main argument against caring more about it is that - while I like having the detail of a 45MP image - it's not all that common that I need more than this, and you do need moire detail like the branches in your sample in order to see much of a difference. It's certainly a feature I'd like to have available. I wonder if there's a market for an automated shift lens that can be tuned to sensor resolution? (Hmm...)

    The difference from Y'CbCr and related encodings is that debayering assumes a correlation between luma data for the channels, and has to use a lower-frequency approximation of the chroma data in order to deduce the relation between luma and chroma at boundaries. This usually, helpfully, matches the eye's insensitivity to high frequency chroma changes, but obviously that's a heuristic that goes wrong with moire and other high-frequency colour data. Y'CbCr encoding typically assumes that the luma value is known and correct before chroma is downsampled (although not relying on this assumption is a reason it's not completely unreasonable to use 4:2:0 downsampling on bayer images). But the chroma reconstruction of broadcast formats is mathematically gibberish anyway (it relies on linear interpolation of nonlinear numbers), so I don't tend to think of it as a goal for image quality. :)
     
  7. My point was that matrix encoding is all around us, and largely transparent. I quickly found, pixel peeping in the camera, that pressing the shutter release manually required at least a 2 second delay to allow vibrations to decay. A cable release or smart phone would do as well.

    Theoretically, shifting the lens could accomplish the same thing. However only IS-enabled lenses would qualify, and these are generally longer focal lengths. The longer the lens, the more precision would be required. Finally IS in the lens mainly addresses tilt and yaw, whereas the pixel-shifting solution is translational and independent of the lens or focal length. Lens-shifting is possible, but not practicable.

    The vacation in Washington state was the first time i experimented with pixel-shifting, and this post was the first time I made a direct comparison of the results. I was surprised at the difference it makes. That said, I can't make a print large enough to see the difference (13" wide at most). For some subjects, at least, it delves deeply into the sandbox reserved for high-end medium format.
     
    Andrew Garrard likes this.
  8. Well that feels like a challenge. :) There are several third parties doing tilt-shift lenses, although here we're talking about very tiny shifts that wouldn't need wide coverage (and which may save us since we're moving the lens and not the camera). You couldn't (reasonably, in the F-mount) do it for existing Nikon glass, but it's the kind of thing that someone like Laowa or Samyang might like to think about. Even if the shifting were manual and driven by a micrometer-like worm gear arrangement (we're talking less than 1/200 of a mm shift for a D850) it would be an interesting selling point, although something using solenoids or similar would be more convenient. Keep it under $1000 and ship it with some conversion software, and I think you might get takers; I might buy one, although I certainly wouldn't pay the enormous premium to have one custom made. Or you could just get your sensor monochromised and just use a filter wheel, as you suggested.

    I'd much prefer it all integrated in the camera body, of course. Some medium format bodies can do the sensor shift thing, which is one reason they can get the quality!
     
  9. The pixel-shifting wheel has already been invented. It's. up to everyone else to catch up, or simply ignore it. Medium format has been an option for a decade. If you have $50K you can spare, and paying customers, go for it.

    Lens shifting isn't going to cut it. The demand for precision and reproducibility is too high. 1/200 mm is about the limit for a precision micrometer, well beyond that of the relatively crude mechanism on a tilt-shift lens.
     
  10. Oh yes, you'd need something very precise (although with a very small movement range) to do it properly - potentially something like the mechanism used for sensor shift, but it would have to go less far and move much more weight. I'm talking about doing this in a dedicated lens (presumably a re-housed version of one of the lenses these companies already sell); arguably the difficult bit is allowing it to shift appropriate amounts for different sensors, and a simple spring-loaded mechanism with a precision-ground mask for movement ought to suffice if models were sold differently for different pixel densities. Personally I think it's a small matter of engineering, but then I'm a software guy, so all mechanics seem like they should be trivial to me. :)

    I'd absolutely like Nikon to put the tech in their cameras (especially if they're now at the stage where they're less worried about sensor shifting weakening the sensor for cleaning purposes, which was apparently an excuse before); it's clearly the right way to do it. I've been mildly advocating for rear movements in general for a while. But there are enough people out there with older cameras (by which I mean "all the current Nikons") that it felt like a moderately-priced niche product might find a home. As you may be able to tell, I'm no product manager!
     
  11. I would like to mount perhaps a Z6 on the back of a 4x5 due to the very short flange distance. Have a special back that shift the camera to take several shots to cover the entire 4x5 area and then stitch them together in post. It would be cool.
     
  12. Voila! there are adapters for practically any camera.

    Cambo ACTUS-XCD View Camera Body with Hasselblad XCD 99010883
     
  13. So, as rumoured, Canon have announced the "EOS R" full-frame mirrorless camera with a new mount. 30MP, 8fps (but only 5 with AF), priced between the Z6 and Z7. They have an interesting 28-70 f/2 to go with it.

    Comparison time?
     
    Last edited by a moderator: Sep 5, 2018
  14. ShunCheung

    ShunCheung Administrator

    Just like Nikon, Canon is making a ridiculously fast lens to showcase the new mirrorless camera. The 28-70mm/f2 weights over 3 pounds and uses 95mm filters, just like Nikon’s 200-500, and costs $2999. I don’t want to hold that lens all day to shoot a wedding. Personally, I think a 24-70mm f2.8 is far more practical. I also wonder how good this lens is at f2, 28mm and 70mm.

    Canon’s 50mm f1.2 costs $2299. I am afraid the f0.95 Nikon Noct will eventually be in the $4000 to $5000 range and will more be a collector’s item, rather than a practical lens.

    Finally, I just can’t believe Canon is introducing yet another incompatible RF mount for their new FX mirrorless cameras, incompatible with their APS-C EOS-M mount. Now Canon is split among the EF mount for DSLRs and two different mirrorless mounts. I doubt Canon will be able to support 3 lens lines.

    BTW, the new Canon FX R also has just one memory card slot: SD UHS-II. I am already not too happy with just one XQD card slot in the Z cameras, although XQD is very reliable. Personally I wouldn’t buy a camera with just one SD slot now in 2018; SD cards are too fragile and too easy to lose. At the Z launch event I attended, naturally the Nikon rep claimed that XQD cards are far more reliable than CF and SD, but he didn’t provide any data to back up that claim. While that maybe the case, any card can still fail. I think the Z7 is way too small a camera. Hopefully XQD and CFX will be much cheaper in a year or two, and we’ll see a larger Z body with two XQD slots in the near future.
     
    Last edited: Sep 5, 2018
    paul_b.|1 and Andrew Garrard like this.
  15. Forget about the EOS-M it's about as good as the Nikon 1 in my opinion. The EOS-R with an adapter is fully compatible with all EF and EF-S lenses. They even has an adapter which add a control ring. The control ring on the RF lenses is in addition to the focus ring not like the Z. You can now use EF-S lenses on a Canon full frame. Its performance level is about as good as the 5D MK IV and $1000 less. It is enough at least to keep any Canon DSLR users from buying the Sony.
    Let say if I were someone without any camera and wants an MILC (both of those conditions are not true for me) I would say I pick either the Canon or Nikon but not Sony.
     
  16. I'd say Eos-M is more in the micro 4/3 space (with many fewer lenses) - it's a small, competent system, helpful if you want compactness, but not challenging the high end. I see this launch more as a statement of intent and a placeholder with the Z series than a class leader.

    I do like some of the adaptor options Canon has - of course they don't have to fit in a mechanical aperture lever. The control ring is nice. I'm with Shun that a 28-70 f/2 is a bit extreme (and Nikon's f/4 is more suitable), although they do have a 24-105 IIRC - but, like the Nikon f/0.95, there's something to be said for a technology demonstrator, and persuading customers that there's something in it for them (not just cheaper for the manufacturer). On that note, interesting tactic by Canon, using reduced production costs to make a cheaper camera. I wonder if Nikon will try that at some point? (I'd not realised the 5DIV was still going for over $3K, but I guess the US availability of the D850 isn't helping compete.)

    I'd say it might be enough to stop Canon shooters who haven't already switched to Sony. I'm not sure how many it would pull back, but obviously we don't know much about the actual performance yet. I'm happy to have a D850 over an A7RIII, but features like the sub-pixel shift do give me mild envy of that system. Having been over 36MP for a few years now, dropping back to a 5DIV would lack appeal if I were to change system; I can see 5Ds users going the Sony route despite this announcement. And I'm assuming the new body's dynamic range is in the same class as the 5DIV: much better than Canon has been in the past, but still a stop or two behind Sonikon at minimum ISO. Of course, most people who care would have switched away from Canon in the 5DIII era!

    Not a world-changing announcement as such (although the new mount may turn out to be), but it doesn't seem horrendous either. I guess the same as for the Z series Nikons.
     
  17. I would welcome a really good, optically and mechanically, f/4 24-70 for mirrorless. Aside for the lack of a compatible tele-extender, the 70-200/4 suits my needs well at half the weight (and cost) of the f/2.8 version. Maybe one of the 24-105 lenses is "The One."
     
    Andrew Garrard likes this.
  18. I'm with you, Ed. If an advantage of mirrorless systems is size, Nikon's choice of smaller lenses makes more practical sense to me than Canon's (although the 24-105 is apparently smaller than the EF one); otherwise Canon seems to be trying to have a halo effect more than practical launch lenses. They've a track record of this, producing the f/1.0 50mm EF lens early on.

    Though, relatively small though the 24-70 Tamron is, I might also be interested should Nikon ever make an F-mount premium f/4 24-70 that's optimised for size. I've got to say I approve of putting a "storage size" position on the Z-mount 24-70.
     
  19. The advantage of mirrorless system is no longer size because you would need the same lenses as when you use DSLR and the smaller body wouldn't reduce the size by that much. Besides to solve the problem of battery life the battery may have to be bigger. The real advantage and disadvantage of the mirrorless system is the EVF.
     
    paul_b.|1 likes this.
  20. For many customers, lighter weight and smaller size are the main advantages, but there are other things such as a more seamless integration of video.

    I think the EOS M mount services customers who would only need a few lenses and want compactness and a fluid user experience. It may not be a problem for Canon to maintain the different mounts if not all of them need to be so ambitious in terms of lenses. There should be a lot of resources freed from not making all those compact cameras any more.
     

Share This Page