Fuji X100 Lens design regarding F2 stop and digital sensor

[wudai_e]

Advertisement (gone after registration)

I was reading on the Fuji X100 from their official site and come across this paragraph:

 

2Road to the F2 aperture value.

 

 

• Designing an F1.6 or F1.8 lens is not so difficult; however, in the case of a digital camera, even if an aperture larger than F2 is used, the light receiving elements on the sensor cannot effectively use the brighter portion of the incoming light because of low incident light gathering efficiency.
  

So if the same principal apply to Kodak sensors, I assume it does, it's interesting to wonder why all the luxes are all sold out like hungry man's hot cakes. I don't own any luxes, have the new 35lux on order, now reading this makes me wonder and look harder at the cron, however, I'm still not canceling my order, interesting what I do with my wallet when the GAS hits.

[hoppyman]

The answer of course is that Summilux lenses work fine wide open with Leica M sensors whatever Fuji has to say about their quite different design and not yet released camera

[Advertisement]

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

[earleygallery]

I suspect that this is to do with the special sensor design of the M8 and M9, which is very different to other sensors from what I understand.

 

Also, the bokeh and wide open DOF of the faster lenses will apply regardless.

[01af]

I was reading on the Fuji X100 from their official site and come across this paragraph:

 

Road to the f/2 aperture value.

 

 

• Designing an f71.6 or f/1.8 lens is not so difficult; however, in the case of a digital camera, even if an aperture larger than f/2 is used, the light receiving elements on the sensor cannot effectively use the brighter portion of the incoming light because of low incident light gathering efficiency.
  

So if the same principal applies to Kodak sensors, I assume it does ...

There is some truth it that but it's not the whole truth. It's true that the individual photo sites on a sensor cannot see light that's coming from too low an incident angle. But what is too low? This angle generally will become lower with (a) shorter backfocus distances, ( wider angles of view, and © wider f-stops. We all remember the times when Leica insisted, contrary to popular demand, that making a digital Leica M camera wasn't possible mainly due to reason (a) above. To eventually overcome this problem, Leica now uses a custom-made Kodak CCD sensor which has particularly flat photo sites where the low-incident-angle problem is ... well, not entirely eliminated but greatly reduced. And this also reduces the wide-f-stop problem, so f-stops of f/1.4 or even f/0.95 are no major problem for this sensor. The maximum f-stop supported by a sensor depends on the sensor technology, lens design, and camera design. So it may be f/2 for the Fuji X100 but it's wider for the Leica M system. By the way, that's the reason why Leica cannot use modern CMOS sensors which would offer faster read-out speeds, less high-ISO noise, and live-view ability—those CMOS sensors have their photo sites at the bottoms of comparatively deep wells so while they're okay for DSLR cameras, they would not work particularly well in a digital rangefinder camera with a short flange distance and high-speed lenses.

 

Of course, technological progress won't stop here, so maybe in the future we will see a Leica M with a CMOS sensor. The new back-illumination technique might reduce the depth of the wells of the CMOS photo sites to almost zero. With the current manufacturing methods, however, it's not possible to produce large-size back-illuminated sensors ... yet. That's why we currently find back-illuminated CMOS sensors only in small digital point-and-shoot cameras and camcorders—but just wait another year or two ...

[rosuna]

That is my idea of the problem 10af, I agree.

 

Leica will go for new technologies when they are available. The good news at this moment is that they have a compelling offer right now, so they have resources and time.

 

I don't know if there is much incentive for back-illuminated large sensors for reflex cameras, but the new EVIL cameras will bring that incentive to the development of such type of sensors.

Edited December 22, 2010 by rosuna

[jaapv]

Olaf's post is the most comprehensive summary to answer the "outdated technology" blasts I've seen yet. Thanks

[lct]

Advertisement (gone after registration)

Great to have that small an equiv. 35/2 lens for €1K including a compact body.

IQ remains to see though and the .5x VF is not encouraging i'm afraid.

[wudai_e]

Great to have that small an equiv. 35/2 lens for €1K including a compact body.

IQ remains to see though and the .5x VF is not encouraging i'm afraid.

 

And most interesting of all, it has a internal retractable 3 stop ND filter for using F2 in daylight! And thx for the explanation 01af.

[lct]

Yes good idea indeed.

[hoppyman]

I'm very surprised at this part of the description. I have no idea regarding the physical depths of the 'wells' with photosites at the bottom for various sensors. However surely the physical feature that makes M lenses, especially wider lenses, work with the custom Kodak sensor is the progressively more offset microlenses above the photosites. Those of course are specifically to ideally bend the "light rays" to enter the wells vertically. After that you just have photons filling up the bucket whether or not they are ricocheting around in the well first:)

 

At least I've never heard of this 'shallow wells' theory. Happy to be proven wrong by informed technical explanation. In any event the retrofocus designs tend to have worse incidence angles, hence the issues arising with some (non-Leica especially) lenses.

 

Leica Camera was telling the truth when they said that it was difficult to make a 24x36 sensor to work with M lenses. That dates back to genesis of the M8 and the custom sensor design with off-set micro lenses did not exist at that time AFAIK. Leica Camera was not resisting popular demand, they just had not been able to figure out (with Kodak of course) how to achieve what was needed at that time. See also what I already posted from technical staff on just how much correction is going on in firmware to make M digitals work at all.

I'm not personally convinced that the choice to use custom CCD sensors over CMOS was driven primarily by any difference in photosite architecture, at least as Leica Camera tells it.

 

............., Leica now uses a custom-made Kodak CCD sensor which has particularly flat photo sites where the low-incident-angle problem is ... well, not entirely eliminated but greatly reduced. ......

Edited December 22, 2010 by hoppyman

[mjh]

I'm very surprised at this part of the description. I have no idea regarding the physical depths of the 'wells' with photosites at the bottom for various sensors. However surely the physical feature that makes M lenses, especially wider lenses, work with the custom Kodak sensor is the progressively more offset microlenses above the photosites.

That’s a solution for a different problem. For each photosite there is a cone of light with the exit pupil as its base and the photosite as its top. Usually this picture is simplified in that one only cares about the principal ray that goes through the middle of the cone. When one talks about incident angles, one usually refers to the incident angle of the principal ray. This incident angle grows towards the edges of the sensor and the placement of microlenses is offset to account for that.

 

But there’s another issue. There isn’t just the principal ray and its incident angle, but a bundle of rays with different incident angles, all hitting the same photosite. If the exit pupil is small, the incident angles don’t differ that much, but with a large exit pupil they do. Since the photo diode sits at the bottom of a well, a large exit pupil might not be visible in its entirety when the well is deep, so an extremely fast lens might spread the light over a large exit pupil that isn’t fully visible from the point of view of the photo diode, rendering the supposed advantage of the fast lens moot.

 

At least I've never heard of this 'shallow wells' theory.

This is actually a well-known phenomenon; I have known about this for years. Shallow wells were touted as one of the main advantages of BSI CMOS sensors, for example.

Edited December 22, 2010 by mjh

[hoppyman]

Thank you Michael. That is very informative. Something new to learn every day.

 

I'll stick to my original response that Summilux lenses work fine with Leica M's whatever Fuji has to say about their upcoming camera

Of course Fuji's comment there does not address all of the other factors (correction) in producing a fast lens design either.

Edited December 23, 2010 by hoppyman

[giordano]

In any event the retrofocus designs tend to have worse incidence angles

 

I guess you meant to write "symmetrical" or "non-retrofocus".

 

It's worth noting that the X100 diagrams actually show an extremely short distance (5.6mm) between the rear element of the lens and the sensor. The rear element is huge (diameter almost equal to the diagonal of the sensor) and Fuji say here that "the positioning of the microlens on the sensor of the X100 has been customised to allow the capture of light rays with up to a 20° angle of incidence". I think this has to mean it has offset microlenses.

[hoppyman]

No John that is what I SHOULD have written if I had not confused the terms and got it wrong . Thanks for the correction. I presume that we agree on the part that Summiluxes work fine on Leica M's

 

There certainly seems to be a lot of anticipation and interest regarding that Fuji camera. No doubt we shall hear all about it when it actually becomes available.

[01af]

I'm very surprised at this part of the description. I have no idea regarding the physical depths of the 'wells' with photosites at the bottom for various sensors.

It's a general property of CCD sensors to have less depth than (regular) CMOS sensors. However the upcoming BSI CMOS technology most likely will change this aspect in the near future. [by the way, I think I can already hear the next question that's going to come up on this ...]

 

 

However surely the physical feature that makes M lenses, especially wider lenses, work with the custom Kodak sensor is the progressively more offset microlenses above the photosites.

That's another feature that adds to the shallow depth. After all, shallow depth doesn't mean no depth. The progressive offsets alone cannot solve the problem, only mitigate it, as different lenses have their exit pupils at different locations so we would need a different progression of offsets for each lens. And with really fast lenses, part of the problem would remain even with perfect offsets, as Michael already has explained above.

 

When thinking more about it ... here's another way of explaining the same thing: We have rays of light that hit the sensor's center and rays that hit the sensor's corners. These rays arrive at different incident angles that progressive offsets can compensate for—albeit not perfectly for all lenses wide-angle, standard, and telephoto. But then we also have rays of light coming from the exit pupil's center and rays coming from the exit pupil's rim. These are hitting the sensor at different incident angles again—for each single photo site. So here progressive offsets don't help. Shallow wells do.

Edited December 23, 2010 by 01af

[rosuna]

Sony seems to have problems with the 16mm lens for NEX cameras. The final production model has a low resolution off axis. It may be an optics problem, but I think this depth of wells problem is involved as well.

 

Offset microlenses and back-illuminated CMOSes will help in designing EVIL cameras, and that development may help future solutions for Leica based on CMOS sensors.

[ho_co]

I don't want to re-open a can of worms. This is only a question:

 

Is it possible that this depth-of-well effect might be at the bottom of the DxOMark data which Mark Dubovoy "mined" for his LuLa Open Letter To The Major Camera Manufacturers? (Latest DxOMark data and conclusions on the topic are at DxOMark - F-stop blues.)

 

My intention is not to criticize DxOMark's methods nor to accept their data as valid; I'm just curious about how this supposed 'discovery' might be explained.

 

 

(Maybe FujiFilm's copywriters are spending too much time trying to figure out what the DxO graphs mean? )

 

 

Thanks.

Edited December 23, 2010 by ho_co

[adan]

Howard - that occurred to me, too. I'm sure there is a complex "dance" between: aperture size, AoI, sensor wells, microlenses, yada yada. Things like the DoX/Dubovoy experience and Fuji's statement (and Leica's statements over the years) tend to focus (ahem!) on different pieces of what is a more general optical situation.

 

Digital sensors prefer light to come from directly overhead. Otherwise the sides of the pixel wells cut off part of the light. Some sensors have deeper wells and more shade (just as one is more likely to find one's self in the shade on a NY street surrounded by skyscrapers than on a Texas country road where everything is flat - especially this time of year when the sun is low and has a shallow "angle of incidence")

 

The new "revelation" as described by Du/Fu/mjh/o1af is that the difference in AoI between the light from opposite sides of a wide aperture can be as significant as the AoI difference with short-focus lenses.

 

BTW, it should be noted that what matters as regards "short-focus" is the location of the exit pupil - NOT the actual location of the rear-most piece of glass. In the Digilux 2, the rear element practically touched the sensor - but it showed no problems because the exit pupil was much further forward than the physical back of the lens. In fact the rear elements were effectively corrector plates that redirected the light to be directly from in front (more telecentric).

 

DIgilux 2 lens X-section diagram in this .pdf from Overgaard's site - scroll down to "Summary of the features":

 

http://www.overgaard.dk/pdf/digilux_2.pdf

Edited December 23, 2010 by adan

[mjh]

Is it possible that this depth-of-well effect might be at the bottom of the DxOMark data which Mark Dubovoy "mined" for his LuLa Open Letter To The Major Camera Manufacturers?

Yes, that’s the reason for the findings reported by Mark Dubovoy.

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[ho_co]

Then it seems to me DxO Mark is on a wild goose chase.

 

Tell me where I’m wrong:

 

CASE 1: EDGE PIXEL

With the wide exit pupil of a fast lens, some edge rays will hit the pixel well at an angle that precludes their entering the photosite. Result: Vignetting at the image edge.

 

Well, that’s why we had our lenses coded, right, or why we’re using “CPU lenses” etc? We want to avoid sensor-originated vignetting, so we use lenses coded to tell the camera to do that.

 

 

CASE 2: CENTER PIXEL - (DxO Mark’s real point)

With the wide exit pupil of a fast lens, some edge rays may still strike the pixel well wall and not reach the photosite. Result: Vignetting at the image center.

 

Again, isn’t that why we had our lenses coded: To avoid sensor-induced vignetting?

 

 

 

 

FOLLOWUP & CONCLUSION

1. The effect is dependent on a number of issues including pixel well depth. Future sensor designs will likely reduce these effects. Thus there’s no disadvantage to buying high-speed lenses today, and possibly getting increased advantage from them later.
    
    
   
2. Fast lenses are used not only to milk available light, but also for artistic reasons such as narrow depth of field. This latter reason is valid despite sensor-induced vignetting, and DxO Mark’s “discovery” doesn’t mitigate against this use of wide aperture lenses.
    
    
   
3. According to DxOMark - F-stop blues, the amount of EV adjustment the cameras in question make ‘secretly’ is up to about a half stop! For this we should consider abandoning high-speed lenses due to increased noise potential? Give me a break.
   

 

 

 

Much ado about nothing, in my book.

 

 

 

And happy holidays to everyone to boot!

Edited December 25, 2010 by ho_co