24-90mm Focus Shift (Diglloyd)

[Jared]

Advertisement (gone after registration)

I think you were just seeing the difference between "actual" and "effective" pixel counts. Not likely that there is any interpolation or up-scaling. Probably just cropping down from "actual" to "effective". Most camera manufacturers don't incorporate a few of the border pixels into the final image--they are just used for dark current, bias calculations, etc. They are still there physically, though, so can be read by GMaster. I assume the total pixel count is actually higher with the GMaster image than with Lightroom?

 

The more interesting question for me would be how dependent the lens performance is on the software for things like distortion correction and light falloff. As I'm sure you know, many lenses now utterly depend on the software for good image quality. This allows for simpler, smaller lenses, but actual results would be quite poor without the automatic adjustments. Leica, like everyone else, incorporates this into many of their AF and coded lenses. That's how they take care of distortion, falloff, and even the "Italian Flag" issues that crop up with non-tele-centric lenses.

 

- Jared

[Advertisement]

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

[CheshireCat]

As I'm sure you know, many lenses now utterly depend on the software for good image quality. This allows for simpler, smaller lenses, but actual results would be quite poor without the automatic adjustments. Leica, like everyone else, incorporates this into many of their AF and coded lenses. That's how they take care of distortion, falloff, and even the "Italian Flag" issues that crop up with non-tele-centric lenses.

 

 

One thing is correcting for sensor issues (like M cameras do), and another thing is correcting for lens issues.

Great lenses like most M lenses do not need any software correction (the sensor does).

The fact the 24-90 is so big, expensive, and still needs software assistance is not what I would expect from Leica's optical engineers.

[k-hawinkler]

Onew thing is correcting for sensor issues (like M cameras do), and another thing is correcting for lens issues.

Great lenses like most M lenses do not need any software correction (the sensor does).

The fact the 24-90 is so big, expensive, and still needs software assistance is not what I would expect from Leica's optical engineers.

Thanks. Great point.

The same holds even more so for Leica R lenses.

They perform excellently on my Sony A7r2 with Novoflex adapters.

I only started buying those R lenses a few years after I got my M9.

First I used the R lenses on an NEX-5N and NEX-7, having APS-C sensors.

When using them on an FF A7r I encountered shutter shock.

But with the A7r2 I didn't encounter any issues, to the contrary IBIS being a major benefit.

Will see whether future Leica cameras will be attractive enough for me and my R lenses.

 

I have pre-ordered the Techart M-E adapter that adds AF for M lenses on A72, A7s2, A7r2.

Hopefully it will arrive within the next month or two.

With an additional R-M adapter Leica R lenses should then also benefit from AF.

Will see how that goes! :-)

[CheshireCat]

With an additional R-M adapter Leica R lenses should then also benefit from AF.

Will see how that goes! :-)

 

Thanks, keep us posted !

[talt03]

I believe that in-camera processing happens with any lens attached to SL. The other lenses I use, R and M are all within the setup parameters. When using a different profile, it behaves differently. Seen this in LR.

[Peter H]

.....................

The fact the 24-90 is so big, expensive, and still needs software assistance is not what I would expect from Leica's optical engineers.

I'd prefer it if Leica's optical engineers didn't spend more of their expensive time designing features that are no longer needed and which confer no photographic benefit.

[CheshireCat]

Advertisement (gone after registration)

I'd prefer it if Leica's optical engineers didn't spend more of their expensive time designing features that are no longer needed and which confer no photographic benefit.

 

 

You mean features like compact size, low weight and optical excellence ?

Have fun using your expensive 24-90 on a third-party body when the SL system will go end of life, without the "optical makeup" software.

[cpclee]

Modern lenses that rely on electronic rather than mechanical linkages with the camera generally cannot be made to work when mounted on bodies of a different brand (eg. the aperture will not close down), let alone the fact that mirrorless cameras like the SL have such short sensor to mount distances that mounting a lens on a different brand body via an adapter (assuming an adapter with all proper electronic linkages can be made) means you will lose infinity focus.  So the issue that so troubles you is a nonissue.  Modern lenses are designed specifically for use with their matching brand bodies and that's all that you can expect.

 

Fussing over optical vs digital correction is a pointless exercise when all digital cameras rely on heavy digital processing to interpret colors, manage noise, reduce vignetting, color fringing etc.  With digital imaging, the lens, sensor and software need to be viewed as one integrated and inseparable system.  Engineers know to let each subsystem tackle what it's best at.

 

 

You mean features like compact size, low weight and optical excellence ?

Have fun using your expensive 24-90 on a third-party body when the SL system will go end of life, without the "optical makeup" software.

[tashley]

I think you were just seeing the difference between "actual" and "effective" pixel counts. Not likely that there is any interpolation or up-scaling. Probably just cropping down from "actual" to "effective". Most camera manufacturers don't incorporate a few of the border pixels into the final image--they are just used for dark current, bias calculations, etc. They are still there physically, though, so can be read by GMaster. I assume the total pixel count is actually higher with the GMaster image than with Lightroom?

 

The more interesting question for me would be how dependent the lens performance is on the software for things like distortion correction and light falloff. As I'm sure you know, many lenses now utterly depend on the software for good image quality. This allows for simpler, smaller lenses, but actual results would be quite poor without the automatic adjustments. Leica, like everyone else, incorporates this into many of their AF and coded lenses. That's how they take care of distortion, falloff, and even the "Italian Flag" issues that crop up with non-tele-centric lenses.

 

- Jared

 

OK, here goes - and it seems worse than I suspected.

 

The first file is a JPEG export at full size of the DNG as shown in Lightroom having been imported with a sharpening profile I generally use but with no changes to image dimensions. It is 6000 x 4000 pixels (i.e. 24mp) -  as you would expect. Lightroom has clearly applied the lens corrections proposed to it by Leica and there is no option to un-apply them.

 

The second file is from a TIFF export from RAwDigger, showing the file with no lens corrections applied whatsoever. It is, as far as I can tell, RAW RAW. It measures 6016 x 4016 - which is 24.16mp.

 

The third file is processed from the same source as the second file so as to match, as closely as possible, the first file. The involves distortion corrections and then cropping. As you will see it is fairly close but not identical to the DNG. I can't get it exact because the lens correction profile provided by Leica is more complex than that available using manual controls in LR but I hope you will agree that it is at least a reasonable approximation. It measures 5314 x 3548 pixels which is equivalent to a touch under 18.9 mp.

 

Please do suggest alternative methodologies if this exercise seems to be incorrectly conducted or if you believe I have drawn the wrong conclusions. But if I am approximately right, the so-called 24mp DNG is in fact an extrapolated <19mp file. That's a 26% difference - approximate, of course, due to the slightly different lens corrections, but it seems ballpark correct unless I have all this wrong....

 

http://tashley1.zenfolio.com/img/s7/v154/p1682167176.jpg

 

http://tashley1.zenfolio.com/img/s6/v141/p1630997552.jpg

 

http://tashley1.zenfolio.com/img/s4/v9/p1778833252.jpg

[Peter H]

You mean features like compact size, low weight and optical excellence ?

Have fun using your expensive 24-90 on a third-party body when the SL system will go end of life, without the "optical makeup" software.

 

 

Don't you agree that optical excellence only matters insofar as it translates visibly into photographs?  Otherwise it's just a competitive game of no interest to me.

 

There's no reason to suppose that the sort of optical excellence you refer to would result in smaller, lighter and less expensive lenses, and I rather think the opposite would apply. If I buy a lens like the 24-90 it would be purely for use on the camera body that it was specifically designed for, or for which the camera body was specifically designed. 

 

This "no digital correction" idea feels to me very wasteful of resources for no real photographic benefit, and I'm not comfortable with that at all.

[tashley]

Don't you agree that optical excellence only matters insofar as it translates visibly into photographs?  Otherwise it's just a competitive game of no interest to me.

 

There's no reason to suppose that the sort of optical excellence you refer to would result in smaller, lighter and less expensive lenses, and I rather think the opposite would apply. If I buy a lens like the 24-90 it would be purely for use on the camera body that it was specifically designed for, or for which the camera body was specifically designed. 

 

This "no digital correction" idea feels to me very wasteful of resources for no real photographic benefit, and I'm not comfortable with that at all.

 

I think that 'all depends'. For example, with the 24-90, if I am correct and the wide end of the zoom requires such extreme corrections that you are effectively getting an interpolated <19mp file, then that is into very significant territory AFAIAC.  One might argue that if the DNG looks great why worry and if you are not intending to print to largish sizes then that is a fair point but if you are intending to print large, then you can't upses the file as much you might need because it has already been up-resed. In effect the lens is turning a full frame sensor into a cropped sensor when shot wide.

[Peter H]

I think that 'all depends'. For example, with the 24-90, if I am correct and the wide end of the zoom requires such extreme corrections that you are effectively getting an interpolated <19mp file, then that is into very significant territory AFAIAC.  One might argue that if the DNG looks great why worry and if you are not intending to print to largish sizes then that is a fair point but if you are intending to print large, then you can't upses the file as much you might need because it has already been up-resed. In effect the lens is turning a full frame sensor into a cropped sensor when shot wide.

 

 

 

Yes, I understand that, but you have to balance the costs, as always. Lens design and manufacture is an exercise in compromise as we all know.

 

Do we really believe that an optically perfect or close to perfect 24-90 lens built to the physical standard of the current lens, incorporating AF and Image Stabilisation would be the same size and weight as the current lens? And the same price?

 

I've speculated before that if Leica were to produce the optically perfect lens and keep ithe same dimensions and the same features as the current lens it would, like the M50 Apo Summicron, cost multiples of its run-of-the-mill (!) Leica equivalents. Maybe they should make one to run alongside the current 24-90, but not instead of it unless you want to kill the SL stone dead before it's left the ground.

[cpclee]

Look no further than the S and Cine lenses for how much the next quality grade lenses will cost...   And I'd bet you in those systems, especially the S, Leica deploys digital correction just the same.  

 

People have to stop thinking about digital correction as somehow suboptimal compared to optical correction.  There are certain types of distortion that have very predictable mathematical properties and can be calculated and hence undone to an arbitrarily high level of precision and excellence.   In those in instances using optical correction is silly because it comes at a physical cost and can only do a worse job.

 

 

Yes, I understand that, but you have to balance the costs, as always. Lens design and manufacture is an exercise in compromise as we all know.

 

Do we really believe that an optically perfect or close to perfect 24-90 lens built to the physical standard of the current lens, incorporating AF and Image Stabilisation would be the same size and weight as the current lens? And the same price?

 

I've speculated before that if Leica were to produce the optically perfect lens and keep ithe same dimensions and the same features as the current lens it would, like the M50 Apo Summicron, cost multiples of its run-of-the-mill (!) Leica equivalents. Maybe they should make one to run alongside the current 24-90, but not instead of it unless you want to kill the SL stone dead before it's left the ground.

[LocalHero1953]

I think that 'all depends'. For example, with the 24-90, if I am correct and the wide end of the zoom requires such extreme corrections that you are effectively getting an interpolated <19mp file, then that is into very significant territory AFAIAC.  One might argue that if the DNG looks great why worry and if you are not intending to print to largish sizes then that is a fair point but if you are intending to print large, then you can't upses the file as much you might need because it has already been up-resed. In effect the lens is turning a full frame sensor into a cropped sensor when shot wide.

How does the SL lens compare to similar size zooms from of the brands in this respect?

[cpclee]

The interpolation is going to be mostly in the edges of the picture especially if the distortion correction is radially symmetric.  So it's not a uniform resolution loss across the image.

 

But, looking at that default DNG, do you actually think the 24-90 has worse corner performance than any film-era 24mm/f2.8 lens mounted on a digital camera?  Or produces a less enlargeable image?  I don't think so.   On such a lens you'd have significant optical resolution loss into the corners whatever is its measured distortion. So who is to say modern camera manufacturers' approach of "preserving optical resolution into the corners as much as possible (as evident in the MTFs), but let distortion take its course because we can fix that rather perfectly in the camera later" isn't far more effective?

 

 

I think that 'all depends'. For example, with the 24-90, if I am correct and the wide end of the zoom requires such extreme corrections that you are effectively getting an interpolated <19mp file, then that is into very significant territory AFAIAC.  One might argue that if the DNG looks great why worry and if you are not intending to print to largish sizes then that is a fair point but if you are intending to print large, then you can't upses the file as much you might need because it has already been up-resed. In effect the lens is turning a full frame sensor into a cropped sensor when shot wide.

[tashley]

How does the SL lens compare to similar size zooms from of the brands in this respect?

 

From memory the Sony A7RII and 24-70 F4 (which is a good but not great lens and is fairly reasonably priced) the figure is about 11 or 12%

[tashley]

The interpolation is going to be mostly in the edges of the picture especially if the distortion correction is radially symmetric.  So it's not a uniform resolution loss across the image.

 

But, looking at that default DNG, do you actually think the 24-90 has worse corner performance than any film-era 24mm/f2.8 lens mounted on a digital camera?  Or produces a less enlargeable image?  I'm don't think so.   On such a lens you'd have significant optical resolution loss into the corners whatever is its measured distortions. So who is to say modern camera manufacturers' approach of "preserving optical resolution into the corners as much as possible (as evident in the MTFs), but let distortion take its course because we can fix that rather perfectly in the camera" isn't far more effective?

 

I think engineers at Leica, Zeiss, Sony, etc. know better than you or I.

Yes the distortion is stretching and extrapolating some areas of the image more than others. But load those three images into a viewer and flip between them and you will see that a lot of cropping is going on regardless of the pixel stretching. Of course, if one area of the image is stretched on average less than others, then by logic, other areas will be being stretched by more. That means that some areas are in effect up-resed by more than 26% assuming that I am correct. In any event, when shot at the wide end the SL appears to be an ~19mp crop-sensored camera with some areas of the frame bearing the brunt of up-resing to 24mp output more than others. 

 

As to the comparison with a film era lens mounted on a digital body, this isn't apples to apples - which I suppose is your point. The issue isn't a huge one but it is clear that all things being equal, an 19mp image enlarges less well than a 24mp image. It might well be the case that the net result is sharper corners than a 24mp file with a film era lens but then if I were shooting a film era lens, I would expect issues. I merely, here, we responding to Jared's opinion that I was mistaking two different kinds of crop by presenting the evidence. And the evidence surprised me when I attempted to quantify it.

 

So I totally agree that I would rather have a certain amount of digital correction and that it can be very effective - I am merely surprised at the pixel cost of the corrections in this case.

 

I don't like to print at less than 180DPI (in fact I prefer higher resolution but that's what I feel I can get away with at worst) and that, in this case, is a difference between a 33.33" print and a 29.52" print. Not huge, just more than I would have thought.

[cpclee]

It will be (square root of 1.26) - 1, so about 12-13% in linear resolution.

 

But you are missing my point.  What I'm saying is the optical resolution loss you normally get from a traditional 24/2.8 lens (eg. designed for the film era, where distortion has to be dealt with optically) would far exceed 50% into the corners.  If Leica tells you they can design a lens with virtually no optical resolution loss into the corners, but the image will be distorted and fixed in the camera where you will lose about 13% of resolution from cropping, that's fantastic, right?

 

Yet if you look at the charts of the 24-90 you'll see that it's exactly the kind of lens that Leica has designed.  Virtually no optical resolution loss into the edges whatsoever and in that regard completely unmatched by any Leica zoom from the film era.  It can do that because Leica freed up degrees of freedom from having to deal with things like distortion optically and redeployed those degrees of freedom towards things like corner resolution which can't be recovered digitally.  

 

Bottomline is, I think Leica knows what tradeoff they are making and have settled on something that in their opinion strikes the best balance.

 

 

Yes the distortion is stretching and extrapolating some areas of the image more than others. But load those three images into a viewer and flip between them and you will see that a lot of cropping is going on regardless of the pixel stretching. Of course, if one area of the image is stretched on average less than others, then by logic, other areas will be being stretched by more. That means that some areas are in effect up-resed by more than 26% assuming that I am correct. In any event, when shot at the wide end the SL appears to be a 19mp camera with some areas of the frame giving and some taking. 

 

As to the comparison with a film era lens mounted on a digital body, I disagree that this is apples to apples I'm afraid. The issue isn't a huge one but it is clear that all things being equal, a 19mp image enlarges less well than a 24mp image. It might well be the case that the net result is sharper corners than a 24mp file with a film era lens but then if I were shooting a film era lens, I would expect issues. I merely, here, we responding to Jared's opinion that I was mistaking two different kinds of crop by presenting the evidence. And the evidence surprised me when I attempted to quantify it.

[Advertisement]

Hallo,

Look here: Leica L-Mount

[tashley]

It will be (square root of 1.26)- 1, so about 12-13% in linear resolution.

 

But you are missing my point.  What I'm saying is the optical resolution loss you normally get from a traditional 24/2.8 lens (eg. designed for the film era, where distortion has to be dealt with optically) would far exceed 50%.  If Leica tells you they can design a lens with virtually no optical resolution loss into the corners, but the image will be distorted and fixed in the camera where you will lose about 13% of resolution from cropping.  That's fantastic, no?

 

I don't honestly think I'm missing your point: I'm talking about what is versus what appears to be, rather than what might otherwise have been. We all make our own choices on these matters as to where we want the compromises to lie and from memory the Q, which I like, has a compromise which I am more easily able to accept at about effective 21 point something mp resolution. 

 

I totally get where you're coming from but I do think that 26% is quite a lot and that 19mp is quite a lot less than 24mp, and that one has to do quite a bit of work to understand exactly what compromise one is accepting...

[cpclee]

24mp of mushiness is less resolving than 19mp of pristine quality upscaled to 24mp.  So I wouldn't equate mp count with resolution.  

 

I don't honestly think I'm missing your point: I'm talking about what is versus what appears to be, rather than what might otherwise have been. We all make our own choices on these matters as to where we want the compromises to lie and from memory the Q, which I like, has a compromise which I am more easily able to accept at about effective 21 point something mp resolution. 

 

I totally get where you're coming from but I do think that 26% is quite a lot and that 19mp is quite a lot less than 24mp, and that one has to do quite a bit of work to understand exactly what compromise one is accepting...