Digital Capture & Standard LF Lenses

[neil poulsen]

I heard at a photo studio that standard LF current lenses don't work as well as the new Digitar lenses on digital backs. According to Schneider's LF technician, it's because of color fringing. He said that the Digitar lenses are much better.

I checked Schneider's webpage, and it indicates that the Digitar lenses are apochromatic, that digital sensors are especially sensitive to non-apochromatic lenses. So, what about the APO-Symmar lenses? Are these lenses not APOchromatic?

By the way, I'm not forgetting about Rodenstock.

Does anyone have personal experience with this? Practically speaking, how bad of a problem is this?

To add to my question, does Canon or Nikon talk about this regarding their digital cameras? I've not heard it, nor do I see mention of this on Canon's website. From my perspective, they regard their lenses as interchangeable between digital and 35mm film.

[Scott Fleming]

As one who has listened in on a lot of conversations over on robgalbraith.com I can tell you that the pro users of the 1Ds and the 1Ds Mk II are less and less happy with many of their canon lenses. Especially wide angle lenses. It is the opinion of many that the present day WA lenses from Canon are indeed inadequate and cromatic aberation is their main complaint. The extreme angle of the light strinking the photo-sites is what causes the purple fringing. It can be toned down with better glass.

The other complaint about existing lenses is that the new sensor is too good for them. It out resolves all but the best prime lenses. It has always been said that 35mm lenses have better MTF curves than MF or LF lenses. The new digital backs have the same resolution regardless of what camera platform you put them on. If resolving fine detail is your game I would think sharper LF lenses would be necessary.

[Ralph Barker]

I think there are several things working here, Neil. First is the physical characteristics of the digital sensors, where (I've heard, but don't know for a fact) the angle of incidence of the image-forming light with respect to the sensor surface is (may) be an issue. Second is the bucket of marketing issues at play. Canon and Nikon woul have a harder time selling expensive digital SLRs to their existing customer base if they told us we needed to buy all new lenses, as well. Also, there is the issue of focal length vs. format size - the angle of view or "perspective" of longer lenses designed for larger formats.

Among the miniature-format (35mm) manufacturers, it is my understanding that Olympus has designed new lenses for their digital cameras. On the flip side of the coin, Leica supposedly has hesitated making a digital M due, in part, to the angle of incidence issue that exists with the current line of M lenses, the design of which often extends farther into the body than SLR lenses, for example. My guess is that this issue may be related to the thickness of the digital sensor, where if the angle of the incoming light is too oblique, it may cause problems with adjacent pixels - the "color fringing" problem, or what might be called ghosting.

But, Canon and Nikon digital SLR users (including me) seem to be getting nice, sharp (even if dinky) digital images with conventional SLR lenses. So, I'm confused, too, while I wait for a full-sized digital back that I can afford for my 4x5, and which runs for a week or so on a couple of AA batteries. ;-)

[Glenn Kroeger]

Neil:

The other issue with the "digital" lenses is higher MTF values at higher spatial frequencies, ie, better resolving power. The trade off is that this better resolving power is realized over a smaller image circle. In many cases, the Digitars have large circles of illumination, but the image circles quoted by Schneider are based on higher resolution/MTF criteria.

In corresponding with Rodenstock, I have learned that some Apo-Sironar-Digitals such as the 90mm are unique designs. Others, such as the 55mm Apo-Sironar-Digital is a tweak of the Ap0-Grandagon. Element spacing is changed slightly, and some of the glasses are selected from more expensive blanks with tighter dispersion/index of refraction tolerances.

Apparently, the main result of all of this is "more" Apo-ness, or smaller lateral CA.

[Emmanuel BIGLER]

Speaking only about view camera lenses, the information is scarce and feedback by professionals is hard to find. The last Rodenstock brochure presenting the apo-sironar digital lenses, however, has two interesting technical features briefly discussed.

The first point deals with vignetting due to the recessed position of digital sensors. To address this problem, opticists have an old trick, using asymmetric lenses with a pupillar magnification greater than one, the rule in all regular retrofocus wide-angle lenses like Zeiss distagons. The information brochure of the new Olympus digital SLR mentions an oversized bayonet diametre in view of accommodating extremely asymmetic lenses in the limit of a quasi-telecentric lens. So there is some serious ground to modify the classical quasi-symmetric lens design used for standard and wide angle view camera lenses.

In the last Rodenstock-Linos 'digital' brochure, there is a comparison of lens designs between the classical 35 mm 'film' apo grandagon and the new 35 mm 'digital' lens. Clearly extreme rays are less slanted in the new design, but Rodenstock does not say anything about less extreme wide-angle lenses, and if you look at pupillar magnification ratios in 'digital' lenses, except extreme wide angle lenses, you find a ratio close to one like in standard film lenses. So the question of slanted rays might not be so important for 70 degree lens.

The second point does make sense to those who have worked with biological microscopes. Rodenstock mentions that since silicon sensors are recessed behind a certain thickness of glass (silica ?) an optical correction is required. It is well-known for biological microscopes that you are supposed to look trough a glass of .18 mm thickness, and that the lens is corrected for the supplement of spherical aberration induced by this thickness of glass, but again the effect is visible for slanted rays mostly. a 'biological' microscope lens of high power is supposed to deliver bad quality images if you do not look throught the standard .18 mm flat glass.

A third point seems obvious by reading the specs is that 'digital' lenses being optimized for small sensors, the engineers have changed the traditional trade-off between performance at the centre and homogeneous image qiality over a wide image circle in favor of more sharpness at the centre and less image circle.

The real puzzling point is the interaction between residual aberrations and a bayer pattern of colour pixels. If the image is not good on silicon on certain circumstances behind a 'film' lens, I'd like to know, conversely, what is magic in film, that makes the same lens acceptable on colour film !!

[Doug Dolde]

I have deliberately avoided Canon for this reason as well as the fact that I am not a big fan of the 2:3 aspect ratio. I recently purchased a Contax 645 system with the Kodak 645C Digital Back. Not only do I get Zeiss lenses but they only use the sweet spot as the sensor is about 37mm square. Some images from it are on my website (http://www.painted-with-light.com/" TARGET="_blank) now.

[Struan Gray]

Emmanuel, my guess is that the difference lies in aliasing, particularly with Bayer sensors. There may be a bit of 'noise pumping' going on too although that's less likely with modern high-bit backs. Film has built in spectral and spatial dithering, but digital doesn't, and current cutting-edge digital is undersampled at the 'resolution' limits claimed in advertising and reviews.

Take a look at the high magnification images of the York hotel towards the bottom of Luminous Landscape's review of the Canon 1DS


http://www.luminous-landscape.com/reviews/cameras/Canon-1ds-mkii-p1.shtml (http://www.luminous-landscape.com/reviews/cameras/Canon-1ds-mkii-p1.shtml)

See the bands of yellow and turqoise in the scrolled railing at the bottom of the DCS proback image - I'd lay money that is caused by aliasing accentuating residual colour abberations. It's a very subtle thing in the overall image for this case, but when you remember that one large target market for these tools is people who shoot catalogues full of textiles I can see why 'digital' lenses with low lateral CA become important.

[Bob Salomon]

"By the way, I'm not forgetting about Rodenstock. But, Schneider has a webpage! (Hint, Hint.)"

So does Linos - that is Rodenstock's owner.


http://www.linos.com/en/prod/index.html (http://www.linos.com/en/prod/index.html)

Been there for quite some time.

And yes we also have lenses for digital applications and yes they do produce better images then traditional optics and yes there is more then one type of digital lens for backs on LF cameras.

We also have brochures that spill out the differences and the reasons.

[Glenn Kroeger]

Neil, Struan, Emmanuel

I have just picked up an Apo-Sironar-Digital 90mm. I will shoot some tests and report back. Hopefully I won't need to cement a thin glass sheet to my rollfilm backs.

[Bob Salomon]

" Hopefully I won't need to cement a thin glass sheet to my rollfilm backs."

Only the Apo Sironar Digital HR series require a glass corrector plate when shooting on film.

The reason for this is that these lenses were designed with the glass cover plate on the scanning backs they were designed for as being the last element in the lens formula. Without the corrector plate you would have a focus shift.

With the Apo Sironar Digital no corrector plate is needed.

[Chad Jarvis]

>>And yes we also have lenses for digital applications and yes they do produce better images then traditional optics...<<

I love reading DRIVEL like this, which is tantamount to an admission of making inferior lenses for other purposes. Now THAT'S marketing.

[Glenn Kroeger]

Chad:

Drivel is an unfair characterization of both the issue and Bob's words. Traditional LF lenses aren't inferior, but they do tradeoff some resolution and other corrections in the center of the field for a larger usable image circle demanded by LF photography. All product designs are balances of features and performance. It helps to know what the tradeoffs are in each.

[Scott Fleming]

Chad,

That's not really fair. We all accept that the small format 35mm lenses of necessity must be and are a lot sharper than MF and LF. It's the same with digital sensors. They are tiny compared to 4 x 5 film and need sharper lenses. There are economics involved here. We have not wanted to pay for ultimately sharp lenses for LF work because they were not necessary. Have you checked the prices of the 'digital' lenses?

[Bob Salomon]

Chad,

Prior to commenting you might want to study the plentiful literature on the how and whys of digital lenses. We, Rodenstock, have several technical brochures explaining the uses and need for these lenses and Schneider must have some also.

Then make your comments.

[Frank Petronio]

Didn't Rodenstock write them for Schneider?

[Jonnie Payback]

"my guess is that the difference lies in aliasing, particularly with Bayer sensors."

I am waiting to change over til they get Tylenol sensors. Then the headaches will be worked out.

[Henry Ambrose]

Seriously, there is a lot to the aliasing factor and the angle of acceptance of the sensor - I think those are the two big factors. Feeding each tiny receptor on the sensor is a difficult task and software and tricks can't correct fully for the mechanical shortfall. And aliasing is voodoo as far as I can tell. Its not entirely predictable and very difficult to eliminate without loss of image quality.

I think a lot of the surprise by DSLR owners is a result of having never seen what their photos looked like under high magnification. If you've been turning in 35mm tranparencies picked with a loupe as finished work, then viewing your digital camera files at high magnifications must be very disconcerting. I suspect that on the printed page, the files that have so many wailing and gnashing their teeth, look just fine. That little bit of CA in the corners just dissappears at 150 line screen 4X6 inch repro. ; >)

I doubt that we'll find the new "digital" LF lenses to be necessarily and consistently better on film than the standard models. And its a matching of lens performance to sensor performance. The digital lenses will be corrected and adjusted for a chip instead of film. Just as a process lens is purpose made for close range reproduction and a standard lens for photographing more distant objects. Tuned for the job so to speak.

It'll be interesting to see what Glenn finds about his new APO Sironar Digital. Glenn, I'd really like to hear back about your test results and impressions.

[neil poulsen]

From Bob Solamon: "So does Linos - that is Rodenstock's owner."

Terrific! I hadn't yet seen this website. For a long time, Rodenstock went without a webpage. I was curious about what Rodenstock offers and now have the opportunity to find out. Thanks for the heads up.

[Emmanuel BIGLER]

It is interesting to see the origins of the different German companies merged into the Linos group. Among them you find Spindler & Hoyer, a highly-regarded brand of laboratory optics. So it does make sense if Rodenstock concentrates on ophtalmic glasses, that Rodenstock-photo-lenses becomes the companion of a scientific optical company. Good for us who insist on top-class instruments, be they lab intruments or photographic or scanning instruments.



Since I use a 6x9 view camera I have always regretted the limited number of view camera lenses specifically designed for 6x7-6x8-6x9 on film. I'm not speaking about dedicated brands of 6x7-6x8-6x9 lenses, but view camera lenses adaptable to all kinds of view cameras. To me a 90mm would be a good general puropose lens for 6x9 like the 135 is preferred by some 4"x5" users.



For 6x7 and 6x9, 90mm was the focal length on the Linhof 220, the rangefinder Fujis 6x7 and 6x9, etc.. 90mm lenses designed for 4"x5" have their optimum f-stop at f/22 and offer too much image circle. So I thought by reading the specs on paper that the 90 apo sironar digital could be an excellent lens for me on rollfilm, only the high premium to pay vs. a conventional 90, 100 or 120 and the lack of used ones has prevented me to seriously consider this 90mm 'digital' lens.



So I'm happy if Glenn shares his experience with this 90mm lens, be it on silicon or on film.



Going off-topic for one minute : 6x9 aficionados are really happy to see the apo-sironar-S introduced in 100mm. And the only regret that we could have is that all the effort by Linos-Rodenstock on new digital lenses might have deprived the company from the required workforce and money to maintain or (even introduce new) long focal lenses on catalog. Probably the market was not there... but I know at least 3 competitors who keep attractive long focal lenses ;-);-) Even if we acknowledge, reading Rodenstock's recent litterature, that the 210 apo-sironar-N is an excellent, not too heavy and affordable long focal length for 6x9 and 4"x5", even if one might dream of a 300 apo-sironar-S, as a long focal length for my 6x9 (and even for 4"x5") is concerned (if I had the budget) for a 300 I would still prefer something lighter and with less image circle like the apo-ronar... so here for the moment our shopping basket will aim at used or non-Rodenstock lenses ;-);-)



Back to the subject ; to Struan : I understand the point of refining the correction of chromatic aberrations for 'digital' lenses. It would be interesting to see if multilayer sensors like the foveon® is subject or not to the same problems as one-layer "bayer-headache" ;-) sensors.



Speaking about the above mentioned test on the new Canon 1DsMkII vs a 645 Contax plus a Kodak 16 Mpix sensor, the colored effects you mention are very subtle, whereas if you look at the "York" letters in the Canon image, they show an obvious "staircase effect" not present on the Contax-Kodak side. Well "obvious" only at such magnifications that this doesn not probably change anything in real-world printed images as seen with the naked eye



However I am not totally satisfied with the statement : silicon sensors are now better than film lenses. The mere fact that aliasing is invoked to explain some artefacts in digital images proves that fine periodic structures beyond the sampling limit of the sensor are actually imaged by the lens with a minimum, non-zero contrast. So on this issue, I'm tempted to argue : one cannot say at the same time "film lenses have a lesser resolution than the sensor" and : "aliasing generates artefacts". You have to choose. Either it is not aliasing, or the lens is better in terms of its ability to transfer fine periodic objects. You cannot get true aliasing (at least as far as I remember my course on signal processing) and a lesser lens resolution at the same time. Except if... what follows is valid, which is just a conjecture proposed to the group's criticism.

What is subtle is the fact that the absolute diffraction limit is very far away from the sensor sampling limit, except @f/22 or f/32 for view camera lenses. So the conjecture is : residual geometrical aberrations, like plain de-focusing, do not actually prohibit that images of periodic objects can actually be recorded with a certain contrast. Those of you who have imaged radial targets like the Siemens' star know what I mean : even de-focused, the star is still visible but with some parts exhibiting a low-contrast, sometimes inverted. The MTF for de-focusing and some geometrical aberrations is oscillating and you can get some reversed contrast for periodic objects beyond the first zero of the MTF. Only diffraction generates an absolute zero of the MTF. So may be the issue is here, residual aberrations do not prohibit that periodic objects generate periodic images beyond the first zero of the de-focused or aberration MTF, those images generating cahotic, "voodoo-style" ;-) aliasing, so improving aberration correction and not only chromatic will counter-fight certain aliasing effects (??) but could introduce others, I do not know.
This would not show up on film due to the random dithering film pattern. But modern slide film being credited of 150 lp/mm, I am still tempted to say : digital sensors can still be improved in terms of a finer pixel mesh, the finer the better... as long as noise does not show up... imagine the noise figure of a 1 micron square pixel ? if you have to average them ten by ten, you'd better stay with present state of the art...

[Glenn Kroeger]

Emmanuel:

Given the choice between a 10 micron image element, and averaging over a 10x10 group of 1 micron elements, I would much prefer the latter! Assuming the CPU power to do the computation, I would retain the same luminance sampling frequency, but would greatly increase the sampling frequency in chroma and eliminate much of the color aliasing problem. So in some sense, it is the size of the Bayer filter pattern that still needs to shrink.

[Struan Gray]

Given a choice between marketing a 10 megapixel back that does things correctly - i.e. oversamples by a factor of three or so and then digitally dithers, bandlimits and downsizes - or a 90 megapixel back that 'resolves' more, albeit at the expense of some aliasing, I can't think of a photographic supplier or user that won't take the second option. Astronomers, microscopists and the CIA, all of whom know what true resolution is, might prefer the former, but the photographic industry beyond the factory doors of the optics houses has simultaneously fetished 'sharpness' while failing to understand what it really is.

A bit of aliasing isn't all bad. In fact, a lot of people prefer the edge effects you get in non-periodic images. But when it bites, it bites hard. I have seen some truly hideous photographs of textiles with fine-scale structure such as Thai silk. There again, I'm a probe microscopist, and not getting fooled by aliased 'atoms' is part of the job description.

Glenn, I too would be interested in seeing your results, although my wallet tells me that in my case the information will remain purely theoretical :-)

[Glenn Kroeger]

Struan is of course correct... and the marketing folks would never miss a chance to use a larger number to characterize any product, since larger must be better. OTOH the poor buyer who has to feed the beast with CF cards would end up preferring somewhat smaller sized files!

[Chad Jarvis]

I stand corrected, but seeing how other companies and marketing efforts truly create inferior products to satisfy some segment of the market or another, I'm dubious of any claim that a certain product is "better" because it's engineered for a different task. Like I said, I stand corrected. Apparently Rodenstock's digital lenses really do "produce better images then traditional optics". Question, though...why not make traditional optics better?

[Bob Salomon]

Chad,

Yes the digital lenses do produce better images. However they also need to be used at larger openings then a large format lens and will be in diffraction at f16 and 22. Additionally they cover smaller circles. The GR series cover a 70mm circle. The Apo Digital cover a 105 to a 150mm circle (depending on the focal length).

If you want to use a smaller circle and a larger aperture we don't need to make LF lenses better you can just use the digital ones instead.

As a comparison we can send a CD if you are interested that has comparison product shots of a detailed original between a digital Rodenstock and a similar non digital one. I will send out one if someone wants to upload it. I can not send out several.

[Bob Salomon]

"GR"

Sorry, HR series not GR.

I have been writing 26 new press releases for PMA later this month and got caught in too much typing.

No, nothing new for Rodenstock but new stuff from Linhof and Wista at the show.