p.19 #1 · Zeiss "Batis" AF lens line to be released for FF E-mount
DavidBM wrote:
Someone said that they have a massive psychological barrier against lenses designed to have their distortion software corrected.
i think I'm developing a psychological barrier against lenses which have their distortion optically corrected!
Here's why: optical distortion correction costs resolution. And if it is optically corrected, that's locked in: you can't get it back! At least if you took the same lens design and didn't optically correct (I know it's hard to compare) you will start of with higher resolution. Of course, after software correction, you lose it again. But there are scenes in which you don't need to correct - and then you can the full goodness of uncorrected resolution.
Of course exactly what the resolution costs of optical correction versus software is unclear. Modern lenses which are well corrected for distortion tend to be expensive too. So it's hard to compare apples with apples. Here's my guess: with low MP sensors, software correction will be worse than with high MP sensors, because the higher the sensor resolution, the more you can forget the limit of sharpness imposed by the sensor resolution, and the more you can think of the image as approximating an analogue representation of the sharpness of the lens....Show more →
I would agree because generally, resolution these days of both the lenses and sensors is so far beyond what 80% of users need.
but even a downsized jpg on screen from a forum is going to show distortion of the magnitude of even a few %
i'll have to admit, at first, I was a bit thrown off by the uncorrected distortion graph of the 85/1.8 - but if that was done in exchange for improving size, weight, price, and (far more importantly) all the other aberrations while providing beautiful rendering, i'll gladly accept some lens profiled distortion correction - which is pretty advanced these days.
p.19 #4 · Zeiss "Batis" AF lens line to be released for FF E-mount
Accepting major optical flaws which require a digital band-aid at the expense of ultimate IQ is as much about higher profits for the companies than anything else. I can see the consumer benefit for inexpensive lenses like the FE 28/2...I can't really see it on 1K+ lenses myself...but that's just me.
Except, as mentioned by theSuede (if I interpreted his post correctly), in actual use it would be like shooting one of those lenses (the Batis) on a lower MP camera vs the Otus even if the same camera is used. There is no free lunch. That could mean size, weight...or company profits (by selling lenses that are much cheaper to make at higher prices).
p.19 #6 · Zeiss "Batis" AF lens line to be released for FF E-mount
Tariq Gibran wrote:
Except, as mentioned by theSuede (if I interpreted his post correctly), in actual use it would be like shooting one of those lenses (the Batis) on a lower MP camera vs the Otus even if the same camera is used. There is no free lunch. That could mean size, weight...or company profits (by selling lenses that are much cheaper to make at higher prices).
I'm the least smart (dumbest) guy in the room - but wouldn't that only be true (noticeable) if the "crippled" resolution of the batis was below that of the sensor? these lenses are getting pretty sharp these days and many (i.e. the ones we're discussing) far outresolve even 36MP.
p.19 #7 · Zeiss "Batis" AF lens line to be released for FF E-mount
edwardkaraa wrote:
Well, I really hate to disagree with some of the most distinguished members of this board, but to me, a high quality lens should be well corrected for all aberrations, physically I consider the need for software correction as a kind of cheating Don't crucify me for saying that, just my opinion.
Don't get me wrong. The best cure is prevention and I would certainly prefer a well-corrected lens to a less well corrected lens plus post-processing. However, if this means Otus weight and cost, I am willing to reconsider. I am willing to deal with a few percent of distortion, but I am more reluctant to accept other aberrations. I don't do much architecture and there is no need for distortion correction in the majority of my shots anyway.
What I don't like is the misleading practice of reporting lens performance as if there were no distortion. The Touit lenses, for instance, have a few percent of distortion, whereas the lens sheets suggest that they are virtually free of distortion. The question arises whether the reported MTF curves include the in-camera correction of distortion as well.
p.19 #8 · Zeiss "Batis" AF lens line to be released for FF E-mount
ecarlino wrote:
I'm the least smart (dumbest) guy in the room - but wouldn't that only be true (noticeable) if the "crippled" resolution of the batis was below that of the sensor? these lenses are getting pretty sharp these days and many (i.e. the ones we're discussing) far outresolve even 36MP.
I don't think that's the case - over the whole frame at least (after correction). Presumably, the market for more expensive Zeiss lenses consists of those who want to squeeze out all of the potential capability (which is one of the reason's the user bought a 36MP based camera to begin with). All this said, it's 3% distortion with the Batis 85 so I don't think it's that huge of a deal in practice. I mostly have a "moral" issue with the price versus compromise - and I am one of those that likes to record what the lens "sees", flaws and all. Digital correction at this level doesn't sit well with me.
p.19 #10 · Zeiss "Batis" AF lens line to be released for FF E-mount
The rationale is that if you can sacrifice the most easily correctable (in post) aberration - distortion - you can make the lens smaller, cheaper and and still increase performance for the aberrations that are more inaccessible for software postprocessing.
The complexity of a lens increases - OR you accept a manyfold increase in surface and material (glass property) tolerance susceptibility - if you try to correct many aberrations at once.
You can make a one-lens system that corrects for spherical aberration perfectly. That's simple, just grind a perfect parameter aspheric. Unfortunately, it will only be perfectly corrected for the center point, for one wavelength of light - and only at one certain object distance...
**Make that aspheric a compound lens, like an achromatic doublet, and you've corrected (better) the longitudinal CA.
**Add another element, a positive lens to split off almost half of the strain on the positive flint part of the compound, and then you can start playing around with field accuracy parameters - i.e how far out from the center axis you have an acceptable image formation.
**Now make two of that three-part lens, and mount the second one reversed in front of the first one - and you have a double-Gauss lens.
**Split one of the doublets, and add another positive lens at either front or rear to spread strain even further, and you have an Ultron (base for most normal 50's and 85's)
Keep on adding either surface complexity (more aspherics) or more elements to keep chipping away at the aberration amounts and increase the usable field size... Making distortion a have a smaller weight in the simulation optimization feedback WILL make the lens smaller and more simple (cheaper!) while still retaining the same performance in all other aberrations.
The longer I work in this field, the more I notice the very large and quite obvious gains in "image workability" through the image processing chain by actually trying to optimize for what you might call "object accuracy".
Object accuracy as I talk about it here really means conformance to the sampling theorem. That each pixel in the image has a mathematical meaning that supports the entire IDEA of representing an almost infinite resolution ideal image by a number of squares on a grid (pixels)...
Correcting distortion in post allows for better object accuracy and lower amounts of other aberrations without increasing lens manufacturing cost. IF you do the resampling right, and IF you realize that all resampling necessitates a sharpening after the fact to "look" sharp even though the data/image is accurately presented. What is accurate and what looks good isn't the same thing.
p.19 #11 · Zeiss "Batis" AF lens line to be released for FF E-mount
Tariq Gibran wrote:
Except, as mentioned by theSuede (if I interpreted his post correctly), in actual use it would be like shooting one of those lenses (the Batis) on a lower MP camera vs the Otus even if the same camera is used. There is no free lunch. That could mean size, weight...or company profits (by selling lenses that are much cheaper to make at higher prices).
Yes, that is quite correct.
I'd say that you have to take the SCALE of the "most compressed" part of the image, and make that the actual scale of the output. Otherwise, you have to upsample the other parts of the image. Upsampling an already Bayer-sampled image isn't a good idea.... Downsampling OTOH is most of the time a very good idea
So if the most "compressed" part of the image is at say "95% scale" compared to the most "expanded" part of the image, you should conform all of the image to the 95% scale - and then crop. This would leave you with even less "MP".
But what my real intention was was to emphasize that you do NOT really affect the "resolution" of the real image. That is still the same. It's your ability to translate that into "pixels" that gets lowered.
The necessity to sharpen after the resampling is the same as when you downsample normally, what we normally call "output sharpening".
p.19 #12 · Zeiss "Batis" AF lens line to be released for FF E-mount
edwardkaraa wrote:
I would argue that a lens like the Otus is designed for SLR, while for a mirrorless cam, there are less design restrictions. I mentioned the ZM earlier. They are small, light, and extremely well corrected with very high image quality.
ZMs are good lenses, but they have their own compromises: in the case of the excellent 35 f2 biogon and its loxia spinoff, (the only zm I have personal experience of) that trade off is performance at f2 and to a degree f2.8. Of course to improve that would likely require a retrofocal and considerably larger design
You can see it is resizing the image in the centre, the dark rink shows where the images are the same. Here's the Lightroom correction laid on top of the version I corrected by hand in Photoshop:
My version with no resampling in the centre is larger all round, not just in the corners, in order to preserve the field of view and produce a file with the "correct" dimensions Lightroom shrinks the corrected image.
Downsampling the otherwise unaltered image centre ought to make the image sharper and thus produce better MTF figures. This is easy to prove, even after cropping the manually corrected image to a rectangle it's still larger than the original. Yet it should produce identical Imatest figures in the centre, since the centre is untouched by the correction (see the difference image I posted here). So if you get the same line pairs/mm figure, but a larger image you have a greater number of line pairs / picture height, i.e. a sharper image (in the centre, corners will of course be worse as they have been stretched).
The fact that Imatest gave worse results in the centre when Roger tested the 24-70 just proves how sensitive it is to resampling, as stated by theSuede, thus it's a false result in my opinion.
Given this I'm warming to the concept of the fix-it-in-software approach to distortion. At the end of the day I'm not a purist, I'm results oriented, all things being equal I'd prefer it was done optically, but things are never equal. If it means fewer aberrations that can't easily be fixed in post then I would accept correcting distortion (where necessary) digitally.
Losses through resampling itself will be smaller the higher the sensor resolution, and in order to deliver good corner results with high resolution sensors without breaking the bank manufacturers may have to resort to this approach for all lenses, this could be the start of a trend. Again if it produces sharper images for the same price I don't really care.
p.19 #14 · Zeiss "Batis" AF lens line to be released for FF E-mount
Matt Grum wrote:
Bit harsh maybe (I can't claim I fully grasp all the stereotypes from each US state), I hope jsv_20 has a sense of humour, but yes, the Zeiss Batis samples and their sometimes unusual dimensions were discussed several pages back.
I've just got back from the pub and think I've finally grasped the intricacies of distortion correcting transformations but insights will have to wait until the morning...
It serves me right for not reading every page before posting...
p.19 #15 · Zeiss "Batis" AF lens line to be released for FF E-mount
jsv_20 wrote:
It serves me right for not reading every page before posting...
hopefully your American skin is a bit thicker than our British friend's :-)
{I've never used an emoticon in my life. I do so now to make obvious the levity of my comments}
p.19 #17 · Zeiss "Batis" AF lens line to be released for FF E-mount
If I have never felt bad for adjusting a horizon, I will never feel bad using software correction for distortion. So what if the end resolution drops from 24mp to 22? Canon's make pretty images, too.
p.19 #18 · Zeiss "Batis" AF lens line to be released for FF E-mount
The theory is fine and informative, but in the end software correction of distortion ends up with an image requiring more sharpening, somewhat negating the high mtf.
p.19 #19 · Zeiss "Batis" AF lens line to be released for FF E-mount
alundeb wrote:
The theory is fine and informative, but in the end software correction of distortion ends up with an image requiring more sharpening, somewhat negating the high mtf.
Lets say it perfectly negates the higher MTF achieved by reducing other aberrations instead of distortion, that means for scenes that don't require correction you have a net win.
I don't apply lens corrections to every image, I only do it when something is clearly off and it's detracting from the image. Sometimes I'll flick it on and off to see what it does to the composition, but rarely. I don't envisage correcting the 85mm Batis much at all unless I'm shooting images of buildings.
p.19 #20 · Zeiss "Batis" AF lens line to be released for FF E-mount
charles.K wrote:
Hi Derrick. I agree I really don't like the focus by wire with the FE 55/1.8 and FE 35/2.8. The FE 35/1.4 is very different IMO. The dampening and movement has now been optimized to the point where I now freely use AF and MF, particularly when I am shooting close @ f/1.4. I also like the Loxia's too for this reason
that's good to hear charles, the 35/1.4 still changes the throw depending on how fast you turn it though doesn't it?
bjornthun wrote:
I understand that. Indeed I have used several manual lenses with exactly the right focus throw, Leica R, Nikon Ai are good. Some MF lenses on the other hand have a too short throw, particularly close to infinity, the Zeiss 100/2 MP is a example of this. Most AF lenses that are not focus by wire have a too short throw in my opinion. So I actually like FBW lenses. The sensitivity to speed of the focus throw allows me to quickly get the lens from near to infinty, so I think it's good. Particularly compared to AF lenses with clutch mechanism, I prefer FBW. Still for a good MF feel I think Leica R and Nikon Ai are among the best, though I prefer FBW lenses, since they give me the option of AF as well. To me they are the best compromise. ...Show more →
i agree the FBW can be better than the manual focusing of most AF lenses when taking your time shooting non moving subjects, i haven't found it to be any better than most good manual focus lenses though. i agree that some MF lenses have nicer throw and damping, though i don't really like nikkor AI lenses (better than the AIS ones though). i actually like the rokkor MCs almost as much as the leica Rs with takumars following close behind.
and think I've finally grasped the intricacies of distortion correcting transformations but insights will have to wait until the morning...