Samuli Vahonen
Offline
• • • •
Upload & Sell: Off
|
 p.40 #9 · Zeiss "Batis" AF lens line to be released for FF E-mount | |
theSuede wrote:
I still can't agree with your overall stance on the matter.
No problem, people disagree all the time, would be pretty damn boring if everyone would have same opinions. If the theory is proven in practice as well, I have no problem changing my mind.
theSuede wrote:
So, a cheaper and smaller lens, that gives better sharpness in a flatter focus field - BUT requires some distortion correction! - is IMO still a much better production choice than a larger, more expensive lens that gives no better sharpness (and more focus field curvature).
Theory is nice but I don't see it practice, at least not in Batis lenses:
- 2/25 has less distortion (1.5% max vs 2%, shape similar) and larger lens barrel than ZE/ZF version, and much better field curvature characteristics
- 1.8/85 has a lot of distortion (for 85), but usually 85/90 lenses are "easy" and field curvature issues are extreme rare. 85 is not small lens. So I'm not sure what was the gain from high distortion in 85 if the theory "was in use"?
Neither is particular cheap thou with Zeiss production amounts it's not comparable to Canikon. Prices are similar to ZE/ZF series and I would assume expected production quantities are much bigger than ZE/ZF series.
theSuede wrote:
The real detail, detail you can trust was actually in front of the lens/camera system when you pressed the shutter, starts at about sqrt(2) of 1/f - when you have downsampled the image to about 1/2 the MP count. Before that scale, the sampling theorem falls apart, which destroys the accuracy of ANY kind of geometric transform you try to apply. This is also one of the main reasons why you have the contrast loss on 1/f detail from the original no matter what your scaling ratio is.
Practical replacement of good old slides is showing photos on UHD-1 (3840x2160) 65+ inch television, which are slowly coming more popular. I don't yet have one, but I have had the privilege to present my photos from these few times. It's pain in the #ss to make it happen in real life, as one has to carry desktop computer - or atleast I don't have good enough GPU in any of my laptops, and my MacMini is one generation too old... then setup everything and calibrate-adjust-calibrate-adjust-calibrate... (TVs usually have really crappy colors, and if forcing via profile without finding best TV settings, colors may "break" and for example gradients like sky can look really crappy).
When using UHD-1 and cropping to 16:9 to avoid black bars with A7 we are:
- linear 1/1.575 (6048/3840=1.5757)
- MP count 1/2.48 (UHD-1 = 7.91Mpix, A7 16:9 = 19.62Mpix)
So just barely above your described 1/2 MP "limit". Photos look great, but are extreme fragile and it's very easy to spot all kind of processing done to the photos. A7r gives a little more freedom, but also it's images quite easily show the processing.
The question for me isn't that much "Do I mathematically have details worth of 3840px wide?" (I maybe won't) but "what has been done to this image? There is circle in center of image which looks different than rest of the image" if using method describe by Matt leaving part of the image non-interpolated and interpolating after some threshold. Maybe there is no real mathematically accurate detail on A7 image to show accurately 3840px wide, but depending on texture and demosaic algorithm used, I'm usually able to create artifacts (as you call them), which have appearance of texture, maybe not the original texture, but atleast not "plastic". Please keep in mind, that I shoot almost 100% nature, same artifacts shooting humanmade stuff most likely would never work on our world full of patterns (e.g. AHD's maze pattern is much more easily triggered by humanmade stuff, and you rarely ever see it in nature subject using similar parameters)
As I'm not in forensic business, I can't just think about theoretical real detail presentation, but I prefer to take into account as well how the viewer perceives the image. During presentation I would hate question "Doesn't leaves in that tree have that kind of small white/gray hair on them, why your's look just smooth green?" because I have lost microcontrast and apperance of texture due to way how I (or in worst case; my camera) have processed the photos and turned the leaves to "plastic" instead. Based on what you wrote earlier, we do have different preferences also here.
Any of this won't matter if one shows photos on web, for example A7 and 1280px wide web photo makes the ratios: linear 1/4.725, MP 1/22.28 (3:2), which will hide a lot of processing. Also I find prints to be more forgiving than UHD-1 display, perceptually, mathematically it's naturally the same.
I was long time thinking that I don't need many megapixels, but moving presentation from prints to UHD-1 (and in future whatever larger resolution format displays come available) I start to change my mind that it maybe doesn't harm to have some extra resolution. It's not that long time ago, back at 2001, I was really happy to 3Mpix resolution of my 1.6 crop DSLR, and was sure that I never need more 
Samuli
|
