APO-Vario-Elmarit-SL 24-90 zoom

[jaapv]

Advertisement (gone after registration)

I certainly hope Leica will leave out the software correction in the primes. I don't really mind what they do with the zooms, they're already a compromise by definition.

I think Leica will go for maximum excellence in the primes. With the newest lens design principles that means software corrections. I doubt that they will settle for less.

[edwardkaraa]

I think Leica will go for maximum excellence in the primes. With the newest lens design principles that means software corrections. I doubt that they will settle for less.

Considering the size of the 50 lux, I somehow believe there will be no software corrections. However if it's truly excellent right to the extreme corners it would be difficult to criticize the software intervention.

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[scott kirkpatrick]

To correct firststream's impression that the Q's 28mm lens is really a 24mm and the SL's Vario at its 24mm setting is really a 21 -- the "Adobe way" of doing the correction gives you a resulting image that is a true 28mm for the Q and a 24mm for the wide extreme of the SL.  If you play the simple trick I described, you can get at most effectively a 27mm Q or a 23mm SL.  But you don't have to do that. 

 

I agree with Jaap.  "Designed for digital" means using the cleanest optical solution plus software correction when the tradeoff makes sense.  I expect Leica's tradeoffs to be heavily weighted towards the maximum quality in the final image.  

 

Classic film lenses (except for the symmetric Zeiss 38mm found in the Hassy SWC) fight distortion by adding more distortion.  The distortion curves that you find in older M lens technical data sheets show barrel distortion of up to 2 % out almost to the top and bottom of the frame, but reducing to a much smaller correction at the corners and sides.  This competing distortion is hard to see if left alone, but it can now be corrected as well.  It's a choice.  I suspect less tricky approaches to handling distortion, with software to control it afterwards, allow improvements in other lens properties.

 

scott

[sandymc]

I agree with Jaap.  "Designed for digital" means using the cleanest optical solution plus software correction when the tradeoff makes sense.  I expect Leica's tradeoffs to be heavily weighted towards the maximum quality in the final image.

 

In terms of "the tradeoff makes sense", the question that most designers of sensing systems would tend to ask is what errors can easily be corrected in software, vs. what can't. So, in the film days, the target for the lens was to reduce overall distortion as much as possible within the constraints of size, weight, targeted cost, etc. In the new world, the target for the lens becomes to reduce distortion that is difficult to correct digitally as much as possible. So, e.g., relatively high levels of "plain" barrel distortion that only varies with focal length is fine. Distortion that e.g, varies with focus and/or aperture in complex ways, not so much.

 

This is no different to any other sensor system, btw - e.g., temperature sensors, etc

 

Sandy

[LocalHero1953]

If Leica can make a lens that achieves, by software correction, the same image as a lens that, without software correction, would have to be larger, heavier, more expensive etc, I'm all for it. 

OTOH, it makes you wonder how big this zoom would be w/o software correction.

[edwardkaraa]

If Leica can make a lens that achieves, by software correction, the same image as a lens that, without software correction, would have to be larger, heavier, more expensive etc, I'm all for it.

OTOH, it makes you wonder how big this zoom would be w/o software correction.

+1

[ramarren]

Advertisement (gone after registration)

I see no point to speculating about the 24-90's performance by looking at reviews; you folks have done that to exhaustion and there's nothing new to add.

 

I'll wait until mine arrives and let you know how I think it performs after I compare it to photos made with the corresponding R and M lenses at the paired focal length settings. I have 24, 35, 50, 75, and 90mm M lenses and 24, 35, 50, 60, and 90mm R lenses to compare against the SL 24-90, never mind 28, 35, 50, 55, and 85mm Nikkors. 

 

It will be an extravaganza of boring lens test images to celebrate... 

Edited November 9, 2015 by ramarren

[Leicaiste]

I have a question for those who used the SL with the SL 24-90.

 

Does it work at full aperture until the picture is taken or does it always work at the selected aperture ?

[lx1713]

I think that in action, at 24mp handheld, I won't notice the distortion nor the fractional difference in quality if the moment captured is memorable.

 

I would rather the lens be my muse than my nanny.

 

The shot discipline to display 24mp truly resolved is very high. Zooms are for action or convenience, for a non repeatable moment. Not careful compositions optimising every pixel or every mm of focal length.

 

It must not jigger its focus when snapping into focus but do it smoothly. As your shots come in, it should flow with the action. Life is too short. Use the right tool for the right job.

[firststream]

I think your question boils down to whether or not straight lines in the 3D object space, which get curved when they are mapped onto the image plane by a lens with some radial distortion, will be straightened out by the corrections applied in 2D at the image plane, even if they don't lie in a plane parallel to the chip.  I think they do, but I'm doing some experiments to see if anything funny happens.

 

scott 

 

Sorry Scott, I hadn't seen this reply until now.

 

Your description of mapping 3D space onto a two dimensional plane is elegant, and really is the crux of my worries with in-camera corrections for the wide angle focal lengths. Again, it's intuitive; when the Q first appeared I was struck by the exaggerated keystoning. I even went back to the street images made by Winogrand and Gilden, both of whom favour both the 28mm focal length and the tilted image. No where in their photographs did I see keystoning to the degree I did with the Q.

 

The traditional bench test of test-charts don't measure a lens' foreshortening, and I don't know if there is any variance from one lens to another, for any given focal length. Is there even a way to mathematically calculate the foreshortening of a focal length in say, the same way its depth-of-field can be? (Are the two measurements related?) 

 

Maybe the way to see if there is any difference between the foreshortening of the focal length of lenses corrected in-camera, versus those that are not, can only be done in-field. It's possible that a simple series of skyscraper shots, shot off one-axis, could show if there is any difference resulting from in-camera corrections to the foreshortening/keystoning properties of the given focal length.

 

We seem to disagree on the expected outcome, and I'd be very happy to have my intuition proven wrong...as it often is!  If it isn't however, then it brings up a philosophical  question as to whether or not foreshortening variance in a lens is a measurement that changes the definition of what that lens is. That is to say by example, if a 24mm lens shows the same foreshortening/keystoning of a wider focal length lens, then can that lens be considered a true 24mm focal length? Ditto for the 28mm lens of the Q.

 

I look forward to your tests. Thanks also very much for answering my other queries in this missed reply.

 

Kind regards,

Jeff

[ramarren]

I have a question for those who used the SL with the SL 24-90.

 

Does it work at full aperture until the picture is taken or does it always work at the selected aperture ?

 

In the brief moments I had handling and using it, it felt to me like using a similar lens on the Olympus E-M1: in normal viewing mode, the camera strives to produce the best image possible for focusing and framing. I suspect this means it holds it close to maximum aperture but that's not guaranteed, it can adjust aperture as well as EVF/LCD amplification to achieve the "best" goal. 

 

The SL does have an aperture simulation (aka preview) mode that holds the lens at the selected aperture so you can assess DoF and other aspects of the image to be captured. It also has a full simulation mode that allows viewing the lens aperture and the shutter effects at the current settings. Neither of those guarantee that the lens is always at maximum aperture in normal viewing mode, but it likely holds it there most of the time. 

[scott kirkpatrick]

Your description of mapping 3D space onto a two dimensional plane is elegant, and really is the crux of my worries with in-camera corrections for the wide angle focal lengths. Again, it's intuitive; when the Q first appeared I was struck by the exaggerated keystoning. I even went back to the street images made by Winogrand and Gilden, both of whom favour both the 28mm focal length and the tilted image. No where in their photographs did I see keystoning to the degree I did with the Q.

 

The traditional bench test of test-charts don't measure a lens' foreshortening, and I don't know if there is any variance from one lens to another, for any given focal length. Is there even a way to mathematically calculate the foreshortening of a focal length in say, the same way its depth-of-field can be? (Are the two measurements related?) 

 

 

Keystoning is just geometry.  When you shoot with a wide angle lens, stuff in the near part of the frame is closer than you would think, based on looking at the distant part of the image.  Shoot close to a table top, for example, and the edges of the table top diverge faster than you expect.  But this is because the camera is probably very close to the table's edge, although the person at the other edge looks to be a more comfortable distance away.  Correcting for radial distortion just makes slightly curved lines get straight, it doesn't do a keystone correction.  Winogrand generally shot at eye level, holding the camera level but tilting it around the horizontal axis to let his horizon reflect objects in the image -- catching interesting bits of faces at the corners (he is famous for the number of people facing the camera with eyes open that he could include in a single shot, like his park bench shot at the '63 World's Fair).  Gilden I think invades his subjects space more, and shoots down at them, so I bet if you look more you will find vertical keystoning of chair arms and such.  Incidentally, a little barrel distortion is a good thing when your frame is filled with faces, as it pulls in details at the corners, reducing the stretch that comes with wide angle perspective.

 

If the LUF forums software will let me embed some shots, I'll show you an extreme example, taken in our living room with a 7mm lens on an Olympus M5.2 (micro 4/3, so this is effectively 14 mm focal length.  The first shot is without radial distortion corrections, but I can put those in and all the bowed out lines get straight.   As you requested, I tilted the camera over at 45 degrees, applying the same corrections to make all the lines straight.  The resulting perpective looks strange, but it is exactly what your eye would see if you put it where the camera was, within about a foot of the coffee cup in the foreground.  So let's see:

 

PB030116 by scott kirkpatrick, on Flickr

 

That's pretty disturbing, but it is easily corrected.  I'll throw in the rotation next, and correct to make the lines come out straight, and we get

 

PB030119 2 by scott kirkpatrick, on Flickr

 

This is correct geometry, but a pretty good example of where not to place the camera for architectural work.  I think I'll go back to taking pictures of cats and kids.

 

scott

Edited November 15, 2015 by scott kirkpatrick

[firststream]

Thanks for that, Scott. What I've been trying to ask is: Is foreshortening a function of Field-of-View, or of focal length? In the corrected image you show, is this the foreshortening of a 14mm lens or a 7mm lens? When I use a 21mm lens on an M8 with a sensor factor of 1.33x, I have a 28mm FoV, with the foreshortening of a 21mm. This is why I've requested a comparison with either the SL and its zoom at 24mm or 28mm, and comparing the foreshortening to that of a prime at the same focal length.

 

I see you have ordered the SL. If the lens is also coming, it would be possible to test the zoom at 24 or 28mm against a prime, and see if the distortion corrections affect the foreshortening.

 

I agree images of dogs and cats are preferable to the aesthetics of the bottom shot...but it beats the proverbial brick wall, no?  At least it magnifies the foreshortening effect by the way you've shot off-axis on two planes. If, the next time, the camera could be pointed up or down, and shot otherwise the same way as you've done here, then the effect would be even further exaggerated.

 

Thanks again.

Edited November 16, 2015 by firststream

[scott kirkpatrick]

It's all POV.  Put your eye about 2-3 inches from the screen and things will look normal.  The picture I showed could have been taken with a 14 mm lens on a full frame camera.

 

scott

[dgktkr]

I hope this is a good thread for this question: what comes with the lens? Hood? Front and back caps? Soft case?

 

dgktkr

[Rolf42]

I hope this is a good thread for this question: what comes with the lens? Hood? Front and back caps? Soft case?

 

dgktkr

 

Yes. Yes. Yes. 

[ramarren]

Yes: the lens comes with caps, lens hood, soft case, instruction manual, warranty, etc. It's all straight-forward, just as you'd expect a premium lens from a premium manufacturer to be. :-)

 

I recall some thread where the question "is the 24-90 or 90-280 lens parfocal?" was raised. Obviously, we can't test the 90-280 yet. But my first test on the 24-90 was to test whether it was parfocal or not. I set aperture priority mode, AutoISO, and manual focus. Set the lens wide open. First I focused it (with peaking and magnification) at 24mm and made exposures 24/28/35/50/75/90. Then I refocused it at 90mm and made exposures 90/75/50/35/28/24. Results: 

 

• Marked focal lengths are accurate enough that EXIF reported within 1mm of the set focal length.
• The series focused at 24mm and shot through 90mm is slightly less sharp than the other series (as you would expect).
• When cropped to a central subject to display the same size, both series show the same degree of focus sharpness.
• Sharpness degradation by cropping and as ISO rose to accommodate the longer focal length, smaller lens opening is minimal.

 

So, yes: the Vario-Elmarit-SL 24-90mm is parfocal. Noting the difference in absolute quality of focus from bullet 2, this means that when using manual focus, the technique of zooming-in, focusing, then zooming out to frame as desired nets the sharpest results. Also, at least in the constrained range from ISO 640-1600, noise/DR/sharpness is nicely held to a finely comparable level. 

[ramarren]

Oh yes: I tried and found my simple Acratech tripod adapter plate for the M-P is a perfect fit on the SL. It does not obstruct the battery lever or battery, the anti-turn ridge connects with the back edge of the camera nicely, etc. A couple of my other plates might fit it too. 

[scott kirkpatrick]

I recall some thread where the question "is the 24-90 or 90-280 lens parfocal?" was raised. Obviously, we can't test the 90-280 yet. But my first test on the 24-90 was to test whether it was parfocal or not. I set aperture priority mode, AutoISO, and manual focus. Set the lens wide open. First I focused it (with peaking and magnification) at 24mm and made exposures 24/28/35/50/75/90. Then I refocused it at 90mm and made exposures 90/75/50/35/28/24. Results: 

 

• Marked focal lengths are accurate enough that EXIF reported within 1mm of the set focal length.
• The series focused at 24mm and shot through 90mm is slightly less sharp than the other series (as you would expect).
• When cropped to a central subject to display the same size, both series show the same degree of focus sharpness.
• Sharpness degradation by cropping and as ISO rose to accommodate the longer focal length, smaller lens opening is minimal.

 

So, yes: the Vario-Elmarit-SL 24-90mm is parfocal. Noting the difference in absolute quality of focus from bullet 2, this means that when using manual focus, the technique of zooming-in, focusing, then zooming out to frame as desired nets the sharpest results. Also, at least in the constrained range from ISO 640-1600, noise/DR/sharpness is nicely held to a finely comparable level. 

 

I was reminded of this post when reading LensRental's recent tests on the actual precision and variabiiity of still photo zooms: https://www.lensrentals.com/blog/2016/03/mythbusting-parfocal-photo-zooms/ .  His conclusion is that only rare copies of 70-200 mm photo zooms are made with sufficient control that the distance to the image plane for sharpest focus (at infinity -- this was done on an optical bench) differs by as little as 0.1 mm.  Some in his sample of ten Canon lenses shifted by 0.8 mm from the infinity settings at the two ends of the range.  By contrast, he claims that a cinema grade lens, with fewer moving elements, can be held to 0,05 mm variation in all samples, but at 10 or more times the cost.

 

Godfrey's test used careful manual focus at full aperture, and compared with the results while zooming to the opposite extreme focal length.  He didn't say whether he focused at infinity or at a closer target.  An error of 0.1 mm with a 24mm lens brings the point of sharpest focus in from infinity to about 6 m, which I think is a change Godfrey would have noticed.  So the 24-90 Vario is really built to an exceptional standard.  Or perhaps firmware is helping to keep tolerances tight.  Has anyone performed the same test with the new 90-280?  Another interesting test for parfocal-ness, would be to focus the lens at 90 mm wide open, then stop down to f/5.6 or 8, and zoom out to 24 mm.  Has anyone tried that?

 

scott 

Edited April 13, 2016 by scott kirkpatrick

[Advertisement]

Hallo,

Look here: Leica L-Mount

[ramarren]

...

Godfrey's test used careful manual focus at full aperture, and compared with the results while zooming to the opposite extreme focal length.  He didn't say whether he focused at infinity or at a closer target.  An error of 0.1 mm with a 24mm lens brings the point of sharpest focus in from infinity to about 6 m, which I think is a change Godfrey would have noticed.  So the 24-90 Vario is really built to an exceptional standard.  Or perhaps firmware is helping to keep tolerances tight.  Has anyone performed the same test with the new 90-280?  Another interesting test for parfocal-ness, would be to focus the lens at 90 mm wide open, then stop down to f/5.6 or 8, and zoom out to 24 mm.  Has anyone tried that?

 

I set up my test target at 2.5m distance, a very and common focusing distance for a lot of my photography at the closer end. I specifically wanted to test at the closer end because this is where I'm not testing my focusing error, I'm testing the lens' potential tendency to focus shift when zooming. Focusing near and at infinity is always much harder to do precisely (despite what it might seem) because there are all sorts of other influences which can degrade focus quality, never mind that focusing a wide angle lens near infinity is a very very tetchy thing and you usually just go so far and let DoF carry you the rest of the way. 

 

I haven't noticed any difficulties with the SL24-90 focusing accurately and consistently at any focal length near infinity, but then I only rarely do that anyway so perhaps it's a lack of the experience on my part. 

 

Your "another interesting test..." is essentially what is happening all the time when you are using MF and using the zoom in, focus, zoom out, shoot. The SL, like SLR cameras, strives to keep the aperture fully open for focusing and framing, and stops down for exposure only with dedicated lenses. So if you're lens is actually parfocal and also doesn't have a great deal of focus shift, the fact that you've focused wide open at 90mm and then shoot at f/5.6 or f/8 when you rack it down to 24mm is a normal part of taking pictures. The DoF at the short focal length, stopped down, will mask minor focus inaccuracies well enough in most cases that you simply won't see them.