Q 28mm lens design

[chrismuc]

Advertisement (gone after registration)

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

http://www.fredmiranda.com/forum/topic/1370077/3 Leica Q: The first camera with fixed 28mm fisheye lens with an image circle that does not cover the sensor size. How to possibly achieve sharp edges after digital correction with such a lens design? Edited June 13, 2015 by chrismuc

[bernie.lcf]

I think at 28mm only the Summilux is better and that thing costs the same as the Q just without the camera

 

The design is not so bad. When you eliminate distortion (which is done in Adobe via DNG codes, but of course not in any other RAW converter), those dark corners are removed. Similar things can be seen, for example, with all Canon L lenses where the level of distortion is even worse (but no automatic correction is applied to the RAW file)

 

So, what you are posting and what the thread is discussing is all true and valid but not Leica specific and not even unusual. 

 

If you need absolute perfection with no post processing, you will need to buy a Leica M with M lenses. All others, pro level cameras by Canon and Nikon included, will require some compromise....

[Advertisement]

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

[viramati]

Well my Summicron 28 asph (which I note cost £30  more than the Q) and the Summilux 28 costs another £1075) was not sharp in the corners wide open and even at f5.6 was not 100% if you really want to pixel peep. From what I have seen so far the lens is more than good in nearly all respects but if you want it for architectural photography and need huge prints then it may not be the camera for you. As documntary camera i feel that this camera will really excel and that is why I am buying it

Edited June 13, 2015 by viramati

[-ph-]

When Olympus introduced their most recent µ4/3rds lenses, there was an outcry by some who detected that the camera performed similar corrections of the raw file. Indeed the 12-40mm/2.8 zoom shows strong distortion natively. But as a user who opens the raw file in Lightroom I never get to see this. What I do get to see is, that the images from this lens are stunningly sharp and distortion-free. So in fact, this is one of the best zoom lenses I have used.

Lens design is mostly about correcting the multitude of optical errors a naive lens would have. So the final lens is a compromise of trading off these corrections, while maintaining a reasonable size and price. If size or price don't play a role you get lenses like the Zeiss Otus, or the 50/2 APO. Lenses which are supposed to be used with film cannot rely on post processing corrections.

But now we have a generation of lenses which are to be used on digital cameras only. And this allows for taking in-camera corrections into respect for the whole design. Doing e.g. geometrical correction in the camera allows the lens designer to focus on other aspects of the design, like sharpness and contrast. As a result, the lens is cheaper, or better or both. Correcting a 5% optical distortion (which is huge by old standards) digitally would mean that 1 in 20 pixels gets interpolated - I dare to say that few lenses have such good resolution at the borders to make that visible.

So we get to see how Leica manages to get such a stunningly good 28mm lens at the price of the Q. I think that is good news!

 

Peter

[jaapv]

Lens design is about shifting aberrations somewhere where they are least seen. I think it is a very good idea to shift them into an area that can be corrected in software, making the lens as aberration-free as possible in all other respects.

 

The results speak for themselves. Images from the RX-1, XV, now this camera (and others) are significantly  better than they would  have been from the same cameras, had they had lenses that relied on optics  for correction solely.

 

Trying to present digital lens correction as an easy cop-out by camera makers and being purist about this matter is cutting off your nose to spite your face.

[chrismuc]

From the picture of the FM guy I would say that that lens has at least 10% geometrical distortion. If it is stretched by the in-camera digital correction, you loose up to 10% resolution towards the image corners. For CA correction only about 1-3 pixels are interpolated (and mainly only in 1 or 2 color channels) while here maybe up to several hundred pixels are interpolated. Hopefully someone can upload a detailed corner crop to see the achievable sharpness.

This is actually a 24mm f1.7 lens with tremendous geometrical distortion. IMO it would have been better to design a 28mm f2 lens with least possible distortion for the Q.

[Daedalus2000]

Advertisement (gone after registration)

In this review (it is in French it needs to be translated) they show that the lens is good (but not exceptional) and that it is relatively consistent between center and edges (see the comparison with the Sony 28/f2). How do they achieve this consistency with such big distortion?

 

http://www.focus-numerique.com/test-2372/compact-leica-q-typ-116-precision-colorimetrie-11.html

Edited June 13, 2015 by Daedalus2000

[chrismuc]

Hmm, Chinese firewall, cannot log into the French site :-(

Anyhow, the current 28 Summicron IS a 28mm lens with good corner sharpness (acc lenstip), has only 0.8 % distortion and is tiny and light.

I just don't get it why Leica didn't use it.

Let's assume that the corner sharpess (MTF at 40 l/mm) of a "good" wide angle lens is about 20 % below the center sharpness. If you have to correct 10% geometrical distortion, the center sharpness isn't much effected but the corner sharpness will be decreased by this 10% to now nearly 30% below the center MTF. So digital post (in the camera or at raw conversion) does not replace the optical quality of lens designs or can compensate all its flaws.

Edited June 13, 2015 by chrismuc

[-ph-]

Hmm, Chinese firewall, cannot log into the French site :-(

Anyhow, the current 28 Summicron IS a 28mm lens with good corner sharpness (acc lenstip), has only 0.8 % distortion and is tiny and light.

I just don't get it why Leica didn't use it.

Let's assume that the corner sharpess (MTF at 40 l/mm) of a "good" wide angle lens is about 20 % below the center sharpness. If you have to correct 10% geometrical distortion, the center sharpness isn't much effected but the corner sharpness will be decreased by this 10% to now nearly 30% below the center MTF. So digital post (in the camera or at raw conversion) does not replace the optical quality of lens designs or can compensate all its flaws.

The thing is, a lens which is not corrected for distortion optically, might not have as much or any of the sharpness degradation in the corner an optically corrected lens has. So even after the digital correction it might be equal or even better than a purely optically corrected lens. Also, this lens is brighter and obviously far cheaper than the M Summicron.

In any case, before we can have a final verdict, the lens performance needs to be measured in comparison of the M lenses.

 

Peter

[edwardkaraa]

Does/will Leica publish the MTF of the lens?

[wattsy]

This is actually a 24mm f1.7 lens with tremendous geometrical distortion.

 

Does this lens differ greatly (or at all) from the lens used on the aps-c X typ 113?

[Daedalus2000]

Hmm, Chinese firewall, cannot log into the French site :-(

Anyhow, the current 28 Summicron IS a 28mm lens with good corner sharpness (acc lenstip), has only 0.8 % distortion and is tiny and light.

I just don't get it why Leica didn't use it.

Let's assume that the corner sharpess (MTF at 40 l/mm) of a "good" wide angle lens is about 20 % below the center sharpness. If you have to correct 10% geometrical distortion, the center sharpness isn't much effected but the corner sharpness will be decreased by this 10% to now nearly 30% below the center MTF. So digital post (in the camera or at raw conversion) does not replace the optical quality of lens designs or can compensate all its flaws.

 

Ok I do not know if that helps you.

 

At f1.7 the Leica gets 0.23/0.22/0.16 cycles per pixel (whatever that means! :-) ) at center/side/edge

At f2 it gets  0.25/0.23/0.17

At f2 the Sony gets 0.32/0.19/0.10, so much better in the center, much worse on the edges

Edited June 13, 2015 by Daedalus2000

[jaapv]

From the picture of the FM guy I would say that that lens has at least 10% geometrical distortion. If it is stretched by the in-camera digital correction, you loose up to 10% resolution towards the image corners. For CA correction only about 1-3 pixels are interpolated (and mainly only in 1 or 2 color channels) while here maybe up to several hundred pixels are interpolated. Hopefully someone can upload a detailed corner crop to see the achievable sharpness.

This is actually a 24mm f1.7 lens with tremendous geometrical distortion. IMO it would have been better to design a 28mm f2 lens with least possible distortion for the Q.

let's be thankful that Leica knows a thing or two about designing modern optical instruments. Such a design would have resulted in a more expensive camera with less IQ with a heavier and larger lens. The trick is to "collect" as many aberrations as possible in one aberration that can be corrected digitally i.e. distortion ( and vignetting).

This camera is smaller, weighs less, costs half and gives better image quality than an M240 plus Summicron 28. Regressing to purely optical lens corrections would have negated these advantages.

[rosuna]

let's be thankful that Leica knows a thing or two about designing modern optical instruments. Such a design would have resulted in a more expensive camera with less IQ with a heavier and larger lens. The trick is to "collect" as many aberrations as possible in one aberration that can be corrected digitally i.e. distortion ( and vignetting).

This camera is smaller, weighs less, costs half and gives better image quality than an M240 plus Summicron 28. Regressing to purely optical lens corrections would have negated these advantages.

 

That is OK to me in fixed lens cameras... or interchangeable lens cameras if, and only if, the cost/size/performance advantages are translated to the user.

 

Anyway, I find it somewhat annoying the fact of a hardware "thing" like a lens needing a software support for a proper behavior. This happens with Olympus lenses, for example. The Zuiko 12mm lens is quite expensive, but you need a complementary software in order to get acceptable results with it. I know is the sign of the times, but... a bit annoying for me. 

[jaapv]

We have to move with the times IMHO if we see improvements, and it is clearly so in this case. If there is an advantage in correcting a lens in a hybrid way, it would be quite foolish if Leica persisted in the older way for Luddite reasons.

[edwardkaraa]

We have to move with the times IMHO if we see improvements, and it is clearly so in this case. If there is an advantage in correcting a lens in a hybrid way, it would be quite foolish if Leica persisted in the older way for Luddite reasons.

Isn't it one of the main attractions of Leica? I wouldn't call it Luddite.

[tgm]

Ok I do not know if that helps you.

 

At f1.7 the Leica gets 0.23/0.22/0.16 cycles per pixel (whatever that means! :-) ) at center/side/edge

At f2 it gets  0.25/0.23/0.17

At f2 the Sony gets 0.32/0.19/0.10, so much better in the center, much worse on the edges

dpreview has provided in the first impression report, jpg and raw files  for apertures from 1.7. to 16. The claim that the lens is impressive already wide open in center. In the whole range from 1.7 to 8 a lot of moire is visible, whch starts to dissapear at 11 and is gone at 16.  In another section in a comparison with the Sony RX1 it becomes obvious that the summilux is manry areas better than the Zeiss 2.0/ 35 execpt the center were they are on par. Download the raw field and see yourself, I think you have to look very carful seeing any differences between 1. 7 and 4 or 5.6 which give optimum already optimum sharpness.

[AlanJW]

There is much talk about the brilliance

 

We have to move with the times IMHO if we see improvements, and it is clearly so in this case. If there is an advantage in correcting a lens in a hybrid way, it would be quite foolish if Leica persisted in the older way for Luddite reasons.

Yes! In today's times the capture is not just a matter of pressing the shutter.  Image quality is a function of many factors, starting with the lens and the sensor of course, but also the software,  including in the camera and continuing to post processing. They are all now part of the process.  This is a good thing.  The technologies work together to produce results.  Leica lenses are just a good starting point. 

[cpclee]

This is all about degrees of freedom: if you allow some geometric distortion (which is easily addressable in software with no detectable residual artifacts), then you gain degrees of freedom that can be used towards addressing other forms of aberration that have to be addressed optically (e.g. cannot be dealt with easily in software without leaving behind artifacts).

 

This is now standard practice in lens design and is deployed by every manufacturer under the sun.  I don't see what is there to contest.  It's a way to achieve outstanding image quality while keeping the lens compact and the cost reasonable. 

 

More generally, every time you allow some kind of tradeoff in one place you can gain performance somewhere else, all else held constant.  Just look at the Zeiss Ortus lenses for example.  They accept much larger lens volume in order to gain optical performance.  That's another kind of tradeoff lens manufacturer may pursue.  

 

It is very likely that the lens and sensor are one enclosed module in the Q, which is common for fixed lens cameras of all manufacturers.  It reduces the cost of manufacturing.  They did this with the X1.  The downside is if any part of this module breaks, they might have to replace the whole module which can get expensive.  I was quoted a repair cost of about US$1000 when the lens motor on my X1 broke.  
 

[Advertisement]

Hallo,

Look here: Leica Q

[cpclee]

Because geometric distortion can be dealt with rather easily in software without leaving behind artifacts.  

 

To my eye the Leica Q blows the Sony 28/2 out of the water.  It is not noticeably worse in the center (if you can even call it that), but is way better by country miles in the edges.  And it's got a faster starting aperture to boot.  A 1.7 lens is much harder to design than a f2 lens because aberrations grow faster than linearly in the size of the aperture. 

 

In this review (it is in French it needs to be translated) they show that the lens is good (but not exceptional) and that it is relatively consistent between center and edges (see the comparison with the Sony 28/f2). How do they achieve this consistency with such big distortion?

 

http://www.focus-numerique.com/test-2372/compact-leica-q-typ-116-precision-colorimetrie-11.html