DSLR Lens Design

[David]

Hi all, I have been puzzling over the differences between older 35mm film lenses and C Class sensor DSLR specific lenses. For example Pentax state that their DA lenses are specifically designed for flat sensors as opposed to non flat film cameras. I don't feel comfortable with this but can't justify my doubts.

I found this site Best Photo Quality System – Full Frame, APS-C or 3/4 ? which seems to confirm my suspicion that FF lenses are preferable to the DAs.

I accept the reflectivity differences between the two types of lenses and the need for different coating on the rear element but don't like the flat sensor bit

Any comments on this?

David

[Class A]

Originally Posted by David

For example Pentax state that their DA lenses are specifically designed for flat sensors as opposed to non flat film cameras.

I guess this is a "nice" way of stating that sensors are more iffy about incident angles than film. Strategically shifted microlenses can be used to address this problem that is sometimes referred to as "sensor vignetting". In particular the light emanating from FF wide angle lenses will not hit the sensor at 90 degrees towards the edges of the sensor. This makes it harder for the respective sensor photosites to register the incoming light.

Lenses optimised for use with digital sensors are often claimed to improve on the incident angles they challenge the sensor with.

Another aspect of optimisation is resolution. APS-C sensors are sometimes referred to as "resolution hungry" because they throw a lot of MP at the APS-C sized image circle of a FF lens. On a FF sensor, that APS-C sized image circle only hits a subset of the FF sensor pixels, say 9.3 MP (of a total of 21MP), whereas on, say the K-7, 14.6MP are used to resolve the same image circle. The latter amount of MP makes it easier to detect the limits of a lens.

Hope that helps.

[JonPB]

I'll agree with that article's general, theoretical premise: given a lens that casts a circle of light, the "best" light is usually found toward the center of the circle. But not all lenses cast the same size circle of light. And, more importantly, what a lens does with the light is far more critical for image quality than how large of a circle it casts.

Some of the DA lenses cast circles large enough to cover a full frame of 35mm film. What they do with the APS-C coverage can be fantastic. Still, I'd rather shoot with a DA Ltd that covers a just-large-enough APS-C circle than a consumer zoom from the 70's which covers a 35mm film frame. When it comes to my money, I buy a lens, not an optical theory. As it happens, almost all of my lenses were designed for film--but that's because they're cheap.

That said, my opinion of that article sank when it stated that the 3:4 ratio is ideal. Given a circle size, the largest amount of rectangular area bounded by that circle will always be a square. 3:4 might be better than 2:3, but 1:1 is better still. 3:4 might be ideal due to common print sizes, common monitor sizes, etc; but it does not capture the most amount of light a rectangle can within a circle.

I can't find a mention by Pentax that the DSLR sensors are flatter than the film plane. That is, indeed, not creditable. Can you link?

[David]

Originally Posted by JonPB

I'll agree with that article's general, theoretical premise: given a lens that casts a circle of light, the "best" light is usually found toward the center of the circle. But not all lenses cast the same size circle of light. And, more importantly, what a lens does with the light is far more critical for image quality than how large of a circle it casts.

Some of the DA lenses cast circles large enough to cover a full frame of 35mm film. What they do with the APS-C coverage can be fantastic. Still, I'd rather shoot with a DA Ltd that covers a just-large-enough APS-C circle than a consumer zoom from the 70's which covers a 35mm film frame. When it comes to my money, I buy a lens, not an optical theory. As it happens, almost all of my lenses were designed for film--but that's because they're cheap.

That said, my opinion of that article sank when it stated that the 3:4 ratio is ideal. Given a circle size, the largest amount of rectangular area bounded by that circle will always be a square. 3:4 might be better than 2:3, but 1:1 is better still. 3:4 might be ideal due to common print sizes, common monitor sizes, etc; but it does not capture the most amount of light a rectangle can within a circle.

I can't find a mention by Pentax that the DSLR sensors are flatter than the film plane. That is, indeed, not creditable. Can you link?

Thanks Jon here is quote from page 23 of K7 brochure:

"Lenses
Unlike film, digital sensors are perfectly
flat, requiring a higher level of precision in
PENTAX lens design. As a result, PENTAX digital
lenses have optimal lens curvature, particularly
in the rear-most lens element located
just before the sensor."

[nulla]

Originally Posted by David

Thanks Jon here is quote from page 23 of K7 brochure:

"Lenses
Unlike film, digital sensors are perfectly
flat, requiring a higher level of precision in
PENTAX lens design. As a result, PENTAX digital
lenses have optimal lens curvature, particularly
in the rear-most lens element located
just before the sensor."

Thanks David for that post

Neil

[Class A]

Originally Posted by JonPB

I can't find a mention by Pentax that the DSLR sensors are flatter than the film plane. That is, indeed, not creditable. Can you link?

I think it is credible in the sense that film that isn't a 100% flat will add imprecision to the image and thus dominate any such imprecisions introduced by a lens. A completely flat sensor raises the bar for lenses.

[bdery]

I'll chime in, since my job title is "optical designer"...

The "problem" with digital sensors is not that they're flatter, this actually helps in comparison with film (you know where the image plane will be). The problem is that their angles of acceptance are smaller. On film, light arriving at an angle would still react with the silver particles and be recorded. With digital sensors, if the light arrives at an angle that's too important, it will not be detected.

Microlenses mitigate this a lot, but it's one key difference.

The size of a lens is ONE of the elements that have to be taken into account when designing. The size impacts mainly vignetting. That is, it can happen that only a small part of the lens will be able to direct light in the corners of the frame, and the corners will get darker.

Other things to be taken into consideration are aperture, glass quality, chromatic dispersion, resolution (linked to all the former elements), focus capabilities, etc.

Now, what is different between a lens made for digital and a lens made for film? Two things:

1-coatings : since the digital sensor is different than film, its sensitivity to light is different, and coatings optimized for digital will be better matched for the sensor. However, coatings adapted for the visible spectrum in general will still perform well.

2-Telecentricity : because of the smaller light acceptance angle of the sensors, lenses must be designed as telecentric as much as possible. That is, lenses should be designed such that light will hit the sensor near to a perpendicular angle. Again, not REQUIÈRED, but all things being equal, it's better.

I hope I was able to help to some extend. People interested in learning about this (and who are not scared by maths) could read the OSLO software's manual. It's actually an excellent lens design textbook, available for free.

[Class A]

Originally Posted by bdery

The problem is that their angles of acceptance are smaller.

This is what I meant to express in the first part of my first post.

Originally Posted by bdery

The size of a lens is ONE of the elements that have to be taken into account when designing. The size impacts mainly vignetting. That is, it can happen that only a small part of the lens will be able to direct light in the corners of the frame, and the corners will get darker.

This is why FF lenses typically fare much better in terms of vignetting on an APS-C sensor compared to their performance on an FF sensor.

Originally Posted by bdery

1-coatings : since the digital sensor is different than film, its sensitivity to light is different, and coatings optimized for digital will be better matched for the sensor.

As far as the visible spectrum is concerned, the different sensitivity of the sensor shouldn't be of relevance, should it? Isn't one objective of the "digital" coatings to reduce reflections from the sensor?

Originally Posted by bdery

That is, lenses should be designed such that light will hit the sensor near to a perpendicular angle.

I'm assuming this problem only needs to be addressed for wide angle lenses, right? Is there a focal length below which the incident angles start to become critical?

Originally Posted by bdery

People interested in learning about this (and who are not scared by maths) could read the OSLO software's manual. It's actually an excellent lens design textbook, available for free.

I tried to find it but didn't find a place to download it from. Does it come with the free evaluation version or do you know a place where to get it from?

[JonPB]

Originally Posted by Class A

I think it is credible in the sense that film that isn't a 100% flat will add imprecision to the image and thus dominate any such imprecisions introduced by a lens. A completely flat sensor raises the bar for lenses.

I see no credible* reason to make DA lenses sharper than their FA counterparts, which Pentax appears to be claiming. If the light is more precisely cast, the resulting image should be sharper. True, with film, the imprecision makes unpredictable where the image will be sharper; but, on average, sharper it would be. I can't believe they're saying that consumer DA lenses are more precisely made than professional FA lenses.

*"Creditable" is legal-speak. They/we like slipping in extra syllables. My apologies.

Perhaps they're saying that they are implementing professional-grade precision into this aspect of all DA lenses. But then, why not say that?

On the other hand, the way I think of it is that digital sensors are deeper than film: the light needs to penetrate through a lens/filter and into an optical-digital sensor rather than merely striking a chemical somewhere on a film. Something about "angles of acceptance," as bdery mentioned. But this has little to do with the flatness of the sensor's surface, unless perhaps that very flatness is what causes reflections away from the sensor at angles that would work fine on film. Still, in that case, being "flat" isn't the problem, but rather the problem here is that the optical sensor isn't the first part of the sensor-unit which light strikes.

My guess is that Pentax just wanted to say that digital sensors behave differently than film, and that DA lenses are optimized for this difference. Some explanation was necessary--and I don't doubt that "flatness," or what was translated into that term, plays some role in it--but that analyzing their language won't teach us much about lens design.

Originally Posted by Class A

I tried to find it but didn't find a place to download it from. Does it come with the free evaluation version or do you know a place where to get it from?

This looks to be it:
Lambda Research's OSLO Optics Reference Manual [pdf]

Thanks for the insight, bdery. I'm curious to look through this, perhaps brushing up my maths along the way.

And thanks to everyone else--I find lens design to be fascinating, and like reading other opinions to help inform my own.

[bdery]

As far as the visible spectrum is concerned, the different sensitivity of the sensor shouldn't be of relevance, should it? Isn't one objective of the "digital" coatings to reduce reflections from the sensor?

Not really, no. A coating's job is to let light pass through. Without coatings, a regular optical surface will reflect about 4% of the incident light. On a loupe, you have two optical surfaces, so that's a bit less than 8%. On a simple 4 elements 50 mm lens, you get 8 surfaces, so roughly 32% (I'm rounding things up here). so coatings are necessary.

Now the problem is that a simple coating is usually optimized for one single wavelength. Here is an example pulled from a filter maker's website:

Semrock : LL01-808-12.5

So to allow multiple wavelengths to get through, you need multi-coated filters, and that gets more complex.

Nowwwwwww. The coatings' job is to let light get to the detector, be it film or a sensor. But film and CCDs/CMOS don't respond to light exactly the same way, so to optimize everything with regards to your sensor, the coatings for film and digital will be slightly different. That is the reason why digital and film sensors might differ.

Light reflecting from the sensor will exit the lens the same way (and with the same efficiency) as light entering the lens coming from your scene. That part is no different between film and digital.

I'm assuming this problem only needs to be addressed for wide angle lenses, right? Is there a focal length below which the incident angles start to become critical?

Not really, no. It's related to the clear aperture, or the f-number as we photographers say (two different but related things). Vignetting can happen with any focal length.

Thanks for the insight, bdery. I'm curious to look through this, perhaps brushing up my maths along the way.

A pleasure. The textbook MIGHT be too steep, depending on your level. Do not be insulted if that is so, I never touched this before my Masters'.

I find lens design to be fascinating, and like reading other opinions to help inform my own.

It's an art as much as a science

[Class A]

Originally Posted by bdery

A coating's job is to let light pass through.

Yes, and the light which doesn't pass through is reflected (there won't be much absorption with lens and sensor surfaces).

A digital sensor reflects more light (back to the lens*) than film which is why the the rear element of a lens gets more light to potentially reflect back. (* Apparently sensors reflect less light in total, but the "little reflection they have is almost entirely "specular reflection", bouncing light directly in the direction it came from, right back into the back of the lens.")

That's why a lens optimised for digital cameras should reflect less light back from its rear element than a "filml lens". Reducing reflections is achieved by letting the light pass through, hence the coatings.

Or not?

Originally Posted by bdery

so to optimize everything with regards to your sensor, the coatings for film and digital will be slightly different. That is the reason why digital and film sensors might differ.

The job of both film and sensors is to record the visible spectrum. Hence coatings need to be optimised for the latter and not for specific recording media. There are some idiosyncrasies with current sensor technology but it would be foolish to optimise lenses for these, given that technology progresses and tomorrow's sensors will have different idiosyncrasies. No?

Originally Posted by bdery

Vignetting can happen with any focal length.

You don't need to tell that a 18-250/3.5-6.3 owner. . However, I was talking about "sensor vignetting" not optical lens vignetting. I'd be surprised if the incident angles from a long lens would be anything but non-problematic for a sensor. I guess, however, that the crucial point w.r.t. incident angles is the location of the exit pupil of a lens and that telephoto and retrofocus designs can decouple incident angles from focal lengths by deliberately changing the location of the exit pupil.

[Lowell Goudge]

In another thread I made a posting outlining the differences between digital and film lenses

http://www.pentaxforums.com/forums/689082-post5.html

[Marc Sabatella]

Not knowing anything about the OP's level of technical knowledge or how much of a pixel peeper he might be or how he demanding he might be regarding any other aspects of image quality, I'm going to at least *mention* a couple of things:

- While the might indeed be very good reason that, all else equal, a lens designed for digital might perform better than one not so designed, all else is rarely equal. Meaning, it does not follow that *all* digital lenses beat *all* film lenses.

- Overall, the differences discussed here may have be smaller than any given photographers inclination to or ability to actually see.

- It may well be a given film lens also provides some characteristic simply not available in any digital lens (any 77/1.8 lenses designed recently? 135/2.8? how many digital lenses even cover a 35mm film frame?). These concerns might easily trumps any of the IQ concerns.

As a theoretical discussion, I'm as interest in this as anyone, but as a practical guide to buying lenses, I woudn't put *too* much stock in it, at least not to the exclusion of other concerns. For instance, I know from empirical testing my M28/2.8 outperforms my DA18-55II in most of the ways I care about. Doesn't matter if being designed for digital gives the 18-55 some advantage over a non-digital version of that same lens; it doesn't magically make it better for my purposes than the M28/2.8. So I use the 28 when I am primarily concerned about the things I know it gives me; I use the 18-15 when I am more concerned about the things I know it gives me (basically, autofocus and a range of focal lengths).

I say this only because I wouldn't want the casual observer to get the idea that this discussion measn that no film lenses are worth using on digital.

[bdery]

As a theoretical discussion, I'm as interest in this as anyone, but as a practical guide to buying lenses, I woudn't put *too* much stock in it, at least not to the exclusion of other concerns.

I agree. I own 8 lenses, only one of which was designed for digital. I think it's important to give accurate information, but it's important not to panick with this stuff

The job of both film and sensors is to record the visible spectrum. Hence coatings need to be optimised for the latter and not for specific recording media. There are some idiosyncrasies with current sensor technology but it would be foolish to optimise lenses for these, given that technology progresses and tomorrow's sensors will have different idiosyncrasies. No?

The visible spectrum, as you put it, has no clear-cut boundaries. I once was able to see light from a "IR" laser reflected on a wall (a pretty powerful laser, granted).

Film lenses were not designed for digital sensors, since it was impossible to guess what they would be at the time. So yes, they have their own idiosyncracies.

A CCD records light intensity, the same for a CMOS. The filters in front of the pixels define the colour channels, nothing else. So if I had to design a lens right now, I would optimize it for the three wavelengths ranges that those filters let go through. That would make me cover roughly the whole visible spectrum, but with varying efficiencies depending on wavelength.

A filter is never a flat top letting 100% light go through on a range. Even the best filters are flawed, and have efficiencies (% transmissions) that vary with wavelength.

the global efficiency (over the visible spectrum) of the bayesian array in front f your sensors is probably different from the efficiency of film. Not VERY different, but not the same. that's why there CAN be a difference in the design of coatings for digital and film.

That's why a lens optimised for digital cameras should reflect less light back from its rear element than a "filml lens". Reducing reflections is achieved by letting the light pass through, hence the coatings.

Or not?

It's the hope of every coating maker to make perfect coatings. Of course. The lower the reflections, the better the transmission. But even though it's true that digital sensors reflect a bit more light, reducing reflections was a concern with film too.

[David]

Wow, thank you all, I sure have learnt a heap about lens design in the last few days. I am happy that I can't go too wrong by sticking to a policy of buying mostly good FF lenses for a future FF body. If that dose'nt happen then I still have good glass to use on C or H class sensor bodies.

Thanks to all on this thread

David