R10 "Easy Predictions" from Erwin Putt

[Irek]

Advertisement (gone after registration)

There were a lot of manufacturers of mechanical typing machines. Is there any in the wild now? PC killed them all. Why? Because of simple use and versality of computer systems.

Lens has 80% contribution to the picture quality. I agree, in the film era. Now lens has the same level but software can put quality to the highs. And lens will have ca. 50% contribution only. Why you use Photoshop or any RAW converter. to improve picture quality above those obtained from lens.

Ad rem.

Hasselblad is rather perceived as supllier of high professional camera systems. And they choose synergic approach to new camera design. They only work on RAW files. So why not to put to work computer power for more, and allow to make corrections of optical distortions just before the next part of picture workflow.

Each lens has its own flaws. Designers knew that and theirs work is basically allocated to optimize lens characteristics. You can always design superb lens whith 20 or so, but bulky and with some using restrictions. But it's not engineering approach. You have to optimize weight, size and costs to obtain very good lens for reasonable price. You can sell a few superb lens and thousands of very good, with satisfactory characteristics.

So which way to choose.

Regards

Irek

[Advertisement]

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

[giordano]

If one start thinking that it's possible to downgrade quality at one stage, then catch it up at another, it's a risky bet.

 

I don't think it's a matter of downgrading quality - and don't imagine Leica are looking for opportunities to do so. Rather, it's an opportunity to improve quality at a given cost, or reduce cost without losing quality.

 

All lens design - even modern Leica lenses - is a matter of compromise. Bringing digital processing into the equation just gives the designers more degrees of freedom. For instance, if distortion and vignetting can be to some extent corrected digitally, the designers can relax the tolerance on these in the optical design - and thereby increase their ability to reduce (say) curvature of field and focus shift.

 

Just my two penn'orth.

[AlanG]

With wide angle lenses on view cameras, I often used expensive center filters to minimize vignetting. This was a mechanical solution that was necessary due to the the difficulty or impossibility of designing such lenses with even illumination. So would you rather lose up to 2 stops and have a filter in front of the lens, or have firmware in the camera that corrects for vignetting? (I understand the M8 does this.)

 

So now it is a simple matter to correct for vignetting, optical distortion, c/a and probably some other defects in software. So even if a manufacturer uses every resource to make as good a lens as is technically possible, there will always be aspects that can be improved in software.

 

I think we are just seeing the tip of the iceburg as to what technology will be able to do for lens and camera design.

[Angora]

I agree, digital process should be considered as an « ally », provided that it adds more flexibility as John says (welcome, BTW ).

What I wanted to point out is that the software correction question is problematized on its usage : when is it considered a « normal feature » or conversely a « rescue feature » ?

 

I think special algorithms and firmwares, etc. should improve the lens ; not support it.

In clear, if a lens shows substantial performance gain on a specific body or using specific softwares, well... some questions come to mind, no ? Like : « Has the lens designing work been finished ? », « Why do they rely so much on post-processing ? »...

(of course this example is exaggerated)

[AlanG]

I think special algorithms and firmwares, etc. should improve the lens ; not support it.

In clear, if a lens shows substantial performance gain on a specific body or using specific softwares, well... some questions come to mind, no ? Like : « Has the lens designing work been finished ? », « Why do they rely so much on post-processing ? »...

(of course this example is exaggerated)

 

I'm not sure where you are going with this. What makes a lens design finished? They are all filled with compromises. Take any lens that is supported by DxO software and that software will improve it. Sometimes it is a big improvement, and sometimes it is a minor improvement.

 

Why not design a lens with software optimization in mind. Is there some kind of rule that says this isn't allowed? Let's say it is a wide range hi speed zoom lens that in order to be reasonably small will have a lot of barrel disortion and vignetting. Now if these issues can be corrected in the camera firmware or automatically with raw conversion software, what's wrong with that? It is just another tool to give us better equipment. (Or less expensive or smaller equipment.)

[Angora]

Yes I know that, since the beginnings of photography, lenses are made up of compromises.

 

It's not a matter whether designers should make compromises or if they should take advantage of the computer in the « image processing mix ».

In digital age, people have to define the « rules » of making lenses in relation with software processes. Definitely. Question however : how far can/should one rely on postprocessing ?

 

Example : I got two cameras, A and B. Same type of body, same kind and generation of sensor, same whatever. Actually, the only difference would be that B officially has a « boosted » firmware.

Now I use the latest lens from the same manufacturer as cameras A and B.

Results : on A, lots of color fringing, unacceptable distortion, image like blurred. On B, acceptable distortion, reduced color fringing, sharp results.

What does it mean ?

It means that the lens has been poorely designed. That with a combination of a very efficient lens and, sideways, an optimised firmware, it'd have been possible to get very good distortion, no color fringing and razor sharp pictures.

 

Once again, exaggerated example.

[AlanG]

Advertisement (gone after registration)

Yes I know that, since the beginnings of photography, lenses are made up of compromises.

 

It's not a matter whether designers should make compromises or if they should take advantage of the computer in the « image processing mix ».

In digital age, people have to define the « rules » of making lenses in relation with software processes. Definitely. Question however : how far can/should one rely on postprocessing ?

 

Example : I got two cameras, A and B. Same type of body, same kind and generation of sensor, same whatever. Actually, the only difference would be that B officially has a « boosted » firmware.

Now I use the latest lens from the same manufacturer as cameras A and B.

Results : on A, lots of color fringing, unacceptable distortion, image like blurred. On B, acceptable distortion, reduced color fringing, sharp results.

What does it mean ?

It means that the lens has been poorely designed. That with a combination of a very efficient lens and, sideways, an optimised firmware, it'd have been possible to get very good distortion, no color fringing and razor sharp pictures.

 

Once again, exaggerated example.

 

 

I think manufacturers should take it as far as they want to. I have no desire to place any kind of restriction on their creativity. That's probably the only way any company will be able to design an 18mm or 20mm shift lens for full frame 35mm. If a specific lens will require special firmware or software for good results that's fine with me. (That is no different from the way cameras are supported in raw processing software today.)

 

If I shoot a photo with my fisheye lens and open it in DxO I can choose whether I want to keep it as a fisheye image or as a rectilinear image. So I'm used to it already.

[FastFashnReloaded]

i think Puts and others like MR over at luminous landscape said 4/3rds would be gone too

yet somehow everyone missed the fall of Konika/Minolta and Pentax

 

so much for predictions...

 

Leica could, and probably should do more though

4/3rds could use some wide primes

and another fixed lens 4/3 or APS C rangefinder would be a very good idea about now

and of course R10 needs to happen

 

Why do we need a wide prime for the 4/3 market? We already have a weather sealed constant F4 7-14 (14-28) mm zoom, and it is only $1600. Compare that to the price the fixed prime F8 Hologon for the G2 cost!

[photoarne]

With wide angle lenses on view cameras, I often used expensive center filters to minimize vignetting. This was a mechanical solution that was necessary due to the the difficulty or impossibility of designing such lenses with even illumination. So would you rather lose up to 2 stops and have a filter in front of the lens, or have firmware in the camera that corrects for vignetting? (I understand the M8 does this.)

 

So now it is a simple matter to correct for vignetting, optical distortion, c/a and probably some other defects in software. So even if a manufacturer uses every resource to make as good a lens as is technically possible, there will always be aspects that can be improved in software.

 

I think we are just seeing the tip of the iceburg as to what technology will be able to do for lens and camera design.

 

The only way to correct for vignetting AFAIK is to boost the ISO progressively towards the corners. In your example one would end with an image with ,say, ISO 160 in the center and ISO 640 in the corners. So the real question is would you rather use a center filter to even out illumination and maintain the same noise across the image, or accept increasing edge/corner noise. There is unfortunately no free lunch.

[giordano]

The only way to correct for vignetting AFAIK is to boost the ISO progressively towards the corners. In your example one would end with an image with ,say, ISO 160 in the center and ISO 640 in the corners. So the real question is would you rather use a center filter to even out illumination and maintain the same noise across the image, or accept increasing edge/corner noise. There is unfortunately no free lunch.

 

Isn't it equally possible to reduce the gain (or otherwise scale down) the signal in the centre of the image. If so, the TANSTAAFL effect hits you with wider apertures or longer exposures - just as with a centre filter. Either way, doing it digitally seems to mean a loss of dynamic range which one wouldn't get with a centre filter.

[AlanG]

The only way to correct for vignetting AFAIK is to boost the ISO progressively towards the corners. In your example one would end with an image with ,say, ISO 160 in the center and ISO 640 in the corners. So the real question is would you rather use a center filter to even out illumination and maintain the same noise across the image, or accept increasing edge/corner noise. There is unfortunately no free lunch.

 

I do it all of the time. Most lenses have way less than a stop of vignetting. But even 2 stops would work out fine. (A 35mm Apo Grandagon for the 6x9cm format is the only lens I own that has 2+ stops of vignetting - depending on the image circle used.) The Sigma 12-24 at 12mm has the most vignetting of any of my 35mm format lenses and its images show no disturbing noise after compensation.

 

Consider that the Leica M8 has this type of vignetting compensation as part of its firmware so it seems obvious that Leica is already embracing this technology. It's just a question of how far they'll take it in future firmware upgrades and new camera models and lenses.

 

But if you want to see how well a good raw processor can deal with extracting detail from under-exposure, look at these examples. DxO also does a bit of noise reduction to compensate. The noise isn't much of a factor and in this regard, cameras will only get better in the future.

 

The first example shows almost a 2 stop (+1.95) exposure compensation using Capture One on a photo shot at 400 iso. So this is almost a 1600 ISO equivalent after correction.

The second one illustrates DXO and how I can shoot underexposed to hold the highlights and then use "DxO Lighting" to simulate fill lighting in the shadows. I think this is about a 3 stop correction but DxO also adds contrast and enhances detail otherwise the shadows would look flat. The resultant noise isn't too bad and I have inserted a sample from a lighter image for noise comparison.

 

http://goldsteinphoto.com/agpexposure.jpg

http://goldsteinphoto.com/agpexposure2.jpg

 

In any case, vignetting is just one of the issues to be dealt with. I regularly eliminate c/a, and lens distortion too. Now images taken on the 35mm format with zoom lenses are better corrected than anything I could do with my specialized view camera lenses. (Except for maximum resolution.)

 

Here's another one:

 

http://goldsteinphoto.com/dxo-c1.jpg

 

This is just one example of what can be done with software tools that are available today. Wouldn't you like to have this as part of your camera firmware so all of your images come directly out of the camera with very little evidence of any lens "defects"? I would.

 

I wouldn't have been such a big proponent of this technology except for the proof I've seen in my work every day. Software compensation and correction is a geat tool for me. I am not aware of any nearly perfect lenses that I can use in its place. (Especially a better 24mm shift lens.) As a matter of fact, the lack of support for Leica cameras and lenses in DxO (and tethered support in C-1) are two main factors keeping me from considering an M8.

 

The software is every bit as important a tool to me as the cameras and lenses.

[AlanG]

So the real question is would you rather use a center filter to even out illumination...

 

Other than on view cameras, and specialized panoramic cameras, where photographers are critical, I don't think anyone uses a center filter. They just accept the vignetting.

 

You should see how hard it is to use an extreme w/a on a view camera indoors with a center filter on the front and a polarizer on the rear at an aperture of f16. And that doesn't allow for any cc filters if needed. I'm so glad I don't have to work that way any more.

[photoarne]

AlanG wrote: Other than on view cameras, and specialized panoramic cameras, where photographers are critical, I don't think anyone uses a center filter. They just accept the vignetting.

 

I agree. But when it comes to 2 stops (as in your post above) or more, a center filter becomes a serious option. Try using the 120mm SW Nikkor on 8x10. Unusable with reversal film without a graduated center filter, just usable in B&W.

 

"Consider that the Leica M8 has this type of vignetting compensation as part of its firmware so it seems obvious that Leica is already embracing this technology. It's just a question of how far they'll take it in future firmware upgrades and new camera models and lenses."

 

Yes, and I hope they will allow the vignetting compensation to remain an option and to be separated from the (highly desirable) colour corrections.

 

"In any case, vignetting is just one of the issues to be dealt with. I regularly eliminate c/a, and lens distortion too. Now images taken on the 35mm format with zoom lenses are better corrected than anything I could do with my specialized view camera lenses. (Except for maximum resolution.)"

 

My experience with view camera lenses indicate that current lenses have extremely low distortion and CA, for practical purposes it is zero. So it is hard to imagine how a DSLR zoom lens after extensive electronic PP can be *better* than that. Just as good, possibly.

However one should not forget the manner in which SW corrects lens distortion; basically it disassembles the image and then puts it together again. This inevitably results in a loss of resolution. It might be slight and insignificant depending on the subject and intended usage of the image, but there is *always* a certain price to be paid.

 

Having said that I find the examples you have posted using DxO intriguing. I take it you are using this SW on a regular basis. Do you routinely run all your captures through this program or just selected images that need specific corretions? How do you find its highlight/shadow (fill light) performance compared to PS? I'm using a good deal of Canon equipment besides Leica, so I am wondering if this SW would be worthwhile.

[AlanG]

Having said that I find the examples you have posted using DxO intriguing. I take it you are using this SW on a regular basis. Do you routinely run all your captures through this program or just selected images that need specific corretions? How do you find its highlight/shadow (fill light) performance compared to PS? I'm using a good deal of Canon equipment besides Leica, so I am wondering if this SW would be worthwhile.

 

Our discussion is pretty much the practical vs. the theoretical. I just want resulst as easily as I can get them. I hated using center filters on interiors because of the light loss. Combined with a polarizer, f16 becomes equal to about f64.

 

The reason I say my 35mm images are better corrected thatn my view camera images is that DxO maps each lens for faults. Whereas with extreme w/a view camera lenses there is still some distortion, c/a, and vignetting (on my 35 Grandagon, the center filter leaves some.) That being said, those lenses work well.

 

Since I bought DxO last November, I have run every image though it except photos of people. I get very good skin tones with Capture One and don't usually need any of the optical corrections for that purpose. I haven't spent the time to make custom "Presets" in DxO that do as well on skin tones. (DxO probably can produce great skin tones, but most of the time it tries to pull out too much detail and shows every imperfection.)

 

I don't use PS but it is my understanding that those controls only work on the converted files in PS. DxO works on the raw file. The "DxO lighting" control is truely remarkable. (Expose to a point just before the highlights clip on the raw file and boost the shadow detail with it.) This has been a great solution for me on contrasty exteriors.

 

You can download a DxO trial and use it for a month. I warn you, the interface takes some time to learn and there are an awful lot of possible settings. Study the tutorials on their site. I couldn't work without it for architecture and interiors.

[EH21]

Alan,

I'm all for a combination of software and lenses but this is limiting for several reasons. Firstly when you use software to correct for things such as barrel distortion, you loose detail and information. No doubt about that. Same for chromatic aberrations, and host of other lens problems. Secondly if Leica or any other manufacturer designs a lens with known compromises to be fixed with software, they can't anticipate all the uses and its economically unfeasible for them to make software fixes for all uses. For example, using a lens with an extension tube or using a leica lens on a canon body. You brought up DxO software.... well I bought that and its never going to work with any of my leica lenses or my DMR and in fact it only works for about 25% of my mainstream canon stuff. Well that was about $300 wasted. Sure its great, but it shows you that at least for now software correction is difficult and expensive and it adds another step in an already complicated workflow.

Eric

[AlanG]

Alan,

I'm all for a combination of software and lenses but this is limiting for several reasons. Firstly when you use software to correct for things such as barrel distortion, you loose detail and information. No doubt about that. Same for chromatic aberrations, and host of other lens problems. Secondly if Leica or any other manufacturer designs a lens with known compromises to be fixed with software, they can't anticipate all the uses and its economically unfeasible for them to make software fixes for all uses. For example, using a lens with an extension tube or using a leica lens on a canon body. You brought up DxO software.... well I bought that and its never going to work with any of my leica lenses or my DMR and in fact it only works for about 25% of my mainstream canon stuff. Well that was about $300 wasted. Sure its great, but it shows you that at least for now software correction is difficult and expensive and it adds another step in an already complicated workflow.

Eric

 

 

You bring up good points. Software correction is is a tool like any other and is but one part of the process. I get excellent results so I don't worry about how DxO is achieving those results. Your other point only illustrates the need for manufacturers to put this type of correction into the camera firmware. (Which is exactly what Leica is doing, although they are still limiting this to some extent.) DxO currently only supports some of the most popular cameras and lenses (or lenses that benefit the most from it.) They are always adding more modules and maybe some day that will include Leica.

 

As for adding another step in the workflow, not for me. It is part of the raw processing now. Where as in the past, I'd have to use special software to correct for c/a, barrel disotortion and even correct for tilt and converging lines wiht other programs after the images have been converted. Now I do it in one step.

 

Look, I have no stake in trying to convince you or making life easier for other photographers. I definitely get superior results efficiently through this process.

[AlanG]

Alan,

I'm all for a combination of software and lenses but this is limiting for several reasons. Firstly when you use software to correct for things such as barrel distortion, you loose detail and information. No doubt about that. Same for chromatic aberations...

 

Actually, there is doubt about that. My experience using DxO shows that images gain detail. I'd be glad to post as many examples as you'd like to illustrate this. When using the software you can press a button to toggle back and forth between an overlay of the corrected and uncorrected versions so it is very easy to see the improvement in detail.

[EH21]

Alan,

I've got DxO and I have had it since it first came out. I know the pace at which they are adding modules....its slower than canon and nikon are adding new cameras. There is no way they will ever get to the Leica. Plus leica is doing itself. I gave up on DxO because other methods are giving me equal or better results for RAW conversion. DxO is good but useless if they don't have a module for the camera lens combination you use. If it doesn't work for most then it becomes an inconvenient workflow tool because you have to use one thing for some shots and another thing for the rest which slows down batch processing. DxO does not play well with image cataloging software so that's another problem. Well anyhow glad it works for you but don't hold your breath for DxO to support anything Leica.

 

The biggest reason I went to Leica from Canon is because I get better images without having to do another step in my work flow. Simple as that. That's one good reason to still make good lenses and not rely on some software to fix it, and why I'll still buy leica lenses.

 

Now as far as your argument that fixing distortion in software doesn't result in less detail - well you'll never convince me that it doesn't. Try any software that fixes barrel distortion for one. It crops a portion of the image unless you want a resulting image that has conic sections on the border. If the capture has distortion, you are never getting as much detail as an image that didn't have it to start with. A better lens will always give you a better detailed image no matter what software you use. Like I said in my first post, I'm all for new technology to improve imaging. But it may very well be new software tools to design optic formulae or new manufacturing technologies that make imaging better than image correction software. I don't want it because I use alternative lens combinations a lot and don't want to be locked into some standard because of software linked to certain optics.

I'd rather have good optics from the start.

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[AlanG]

Alan,

Now as far as your argument that fixing distortion in software doesn't result in less detail - well you'll never convince me that it doesn't. Try any software that fixes barrel distortion for one.

 

I never made an argument that fixing distortion doesn't result in less detail. You said,

"I'm all for a combination of software and lenses but this is limiting for several reasons. Firstly when you use software to correct for things such as barrel distortion, you loose detail and information. No doubt about that. Same for chromatic aberrations, and host of other lens problems." I did not answer specifically to only one correction.

 

I stated that the whole process of using DxO generally gains detail. It might lose a very slight amount in the distortion correction but gain some in the c/a correction and sharpening. I see a net gain. First of all the distortion in most lenses (except for a fisheye) is slight so the amount of corrective pixels generated are few and don't result in noticeable detail loss. But when the program corrects for c/a that brings out detail. DxO also has its own sharpening that is specific to each lens. I'll post an example when I get to my office. I just looked at a case on Friday on an image shot with my 24 tse. The DxO conversion corrected for c/a and other issues, in this unsupported lens, and brought out additional detail in tree branches and leaves above a building. I did a comparison image with C-1.

 

In any case, DxO overlays the corrected image on top of the uncorrected one allowing you to instantly switch back and forth. So it would be pretty easy to study it closely and see if choosing distortion correction actually loses noticable detail somewhere. You could always turn that correction off if you didn't need it.

 

Over the years I have processed many hundered of images in DxO and other programs where I have straightened out converging lines where the software regenerated some pixels and everything looks fine to me. That's just my experience in practice not dealing with theoretical concerns about losing detail. Whatever detail was lost was slight compared to what I gained. Even images taken with "perfect" lenses will often need geometric correction for my type of photography.

 

As for lenses supported by DxO. It supports my 12-24, 15mm, 16-35, 24-105, 28-135, 70-200 2.8 IS and 100-400 IS. (And other lenses I don't own.) But you know what? While the 24 TSE is not supported, DxO automatically corrects for the c/a and adds sharpening when I use it. And my "unsupported" 45 TSE, Nikkor 28 PC, and Nikkor 35 PC work the same. Some of the controls for c/a, etc, can be adjusted manually for unsupported lenses as long as the camera's raw file is supported.

 

DxO also allows for manual geometric correction for converging lines or to make an object level or square. And it can compensate for other types of wide angle distortion that is a problem with even the most perfectly made lens. Such as the way they stretch out a person's head at the corner of a frame. (Volume anamorphosis.)

 

It really isn't my fault that DxO doesn't support Leica cameras or some other camera/lens combos However, I wrote to DxO suggesting that they allow the program to work in manual mode even on unsupported cameras and lenses by just giving DxO the raw file support for those cameras.

 

Yes ideally we'd all have great lenses that don't need any kind of correction, but I don't see that ever happening. In the case of the M8, even with good lenses, firmware/software does correct for some and could correct for a number more of its imaging issues.

 

I stand by my original point that I feel that manufacturers should look at the whole process - camera - lens - firmware/software to come up with total photographic solutions. I think this is where the new optical designs and improved imaging performance is likely to come from. I only wrote about DxO to show that this approach is viable today. Wasn't that what this thread is about?

 

By the way, how do you know that a specific lens type that is designed with a "defect", let's say with a certain amount of barrel distortion which is corrected in firmware, can't be made to outperform a similar lens that doesn't have this distortion? Isn't this up to the manufacturers and their own creative designers to try? Or do you know for sure it isn't worth trying?

[zapp]

Yes I know that, since the beginnings of photography, lenses are made up of compromises.

 

It's not a matter whether designers should make compromises or if they should take advantage of the computer in the « image processing mix ».

In digital age, people have to define the « rules » of making lenses in relation with software processes. Definitely. Question however : how far can/should one rely on postprocessing ?

 

Example : I got two cameras, A and B. Same type of body, same kind and generation of sensor, same whatever. Actually, the only difference would be that B officially has a « boosted » firmware.

Now I use the latest lens from the same manufacturer as cameras A and B.

Results : on A, lots of color fringing, unacceptable distortion, image like blurred. On B, acceptable distortion, reduced color fringing, sharp results.

What does it mean ?

It means that the lens has been poorely designed. That with a combination of a very efficient lens and, sideways, an optimised firmware, it'd have been possible to get very good distortion, no color fringing and razor sharp pictures.

 

Once again, exaggerated example.

 

We have to look at the complete sensor-lens system not at lens or sensor only. If firmware can reduce color fringing, it probably fixes lens trouble, but at the same time it alters the raw image by alot. Instaed of complaining about lens design I would rather complain about heavy recalculation of images inside the camera.

Color fringing was a little probelm back in the film days, today most wide angle lenses look like rainbow lenses. We still need to wait a long time before all lenses are really digital ready.