PDA

View Full Version : LF versus smaller format optics, which is better pixel for pixel?



chris jordan
11-Apr-2015, 12:58
Hi guys, question for the optical experts here. I have wondered for awhile now what is the finest resolution that can be obtained photographically, per square inch of sensor/film. Imagine this test: you shoot an image with 3 cameras all side by side: an 8x10 with a 200mm lens (with a high resolution digital back mounted behind it), a medium format with 200mm lens, and a 35mm DSLR with 200mm lens. All three rigs are the best quality, highest-resolution available, and all the photos are taken with the same aperture, shutter speed, lighting, etc. Then you crop all three images down to the size of the 35mm frame. Which of them would show the finest detail? I have always thought that the optics get worse the bigger you go format-wise, but a friend suggested that good quality LF lenses may out-resolve the smaller formats pixel for pixel. Thoughts?

dmeckert
11-Apr-2015, 13:10
I'm interested to see where this goes. I've always operated on the same premise you have.

Mark Woods
11-Apr-2015, 13:37
For an empirical test, shoot a standard cinema lens chart and look at the film with a microscope. I've seen cinema lenses (standard 35mm) resolve to 200 lines/mm. I don't really care about resolution so much as the "personality" of the lens. I use to use a lot of diffusion when I was working as a cinematographer -- with my Ziess lenses.

Armin Seeholzer
11-Apr-2015, 13:53
35mm lens will win this test, the GF lens wins if you use the whole 8x10 format and you enlarge the 35mm to 8x10!

Cheers Armin

mdarnton
11-Apr-2015, 14:05
Check these numbers:
http://www.hevanet.com/cperez/testing.html
Then remember that 35mm lenses regularly clock in MUCH higher. But as Armin said, your test isn't really how one would use a camera, with different film sizes and the same FL lens, and at the same field of view from the same place. . . .

Eric Leppanen
11-Apr-2015, 14:14
Have you see these tests?

https://www.onlandscape.co.uk/2011/12/big-camera-comparison/
https://www.onlandscape.co.uk/2011/12/big-camera-comparison-comments/
https://www.onlandscape.co.uk/2011/12/camera-test-editors-commentary/

Nodda Duma
11-Apr-2015, 14:47
If you keep the focal length the same, the smaller the image plane the higher the resolution possible. There's no contest...it's not even a fair comparison.



Now, a fairer comparison would be to compare setups with the same field of view (say, 50mm focal length for 35mm format and the medium/large format equivalents), same lens designs, etc.

What you'll find is that everything scales, including aberrations. So your absolute spot size (blur size) is smallest for 35mm format, and it still beats out the larger sizes in absolute resolution.

*However* -- and this is important -- if you take your different sized negatives and make prints all of the same size (ie all 8x10s), you will find that their image quality is all equivalent: the minimum resolvable detail will all be the same size for all the prints. You've basically reversed the scaling of the geometric spot sizes for the larger formats.

This is what optical design theory states and what experience has shown me.

ic-racer
11-Apr-2015, 14:47
What "high resolution 8x10 back" is part of your test?

fishbulb
11-Apr-2015, 18:32
To me, this is one of those theoretical questions that doesn't really matter in practice. The more practical question to ask is 'how much resolution do I need for this, and what formats will deliver it?'. From there, other practical questions of size, weight, cost, ease of use, etc. are far more important than theoretical resolving power.

For what its worth though, my money would be on an ultramodern 35mm optic like the Zeiss Otus 85mm f/1.4 at about f/4.

Jac@stafford.net
11-Apr-2015, 19:31
[...] *However* -- and this is important -- if you take your different sized negatives and make prints all of the same size (ie all 8x10s), you will find that their image quality is all equivalent: the minimum resolvable detail will all be the same size for all the prints. You've basically reversed the scaling of the geometric spot sizes for the larger formats.

Emphasis is mine.

Resolution might be equivalent, but 'quality'? Quality includes tonal rendition. In that case, the miniature 35mm would be stressed and have lower tonal quality. No?
.

Nodda Duma
11-Apr-2015, 19:44
Sorry, I get -2 points for using a not-well-defined term :)

By quality I mean ability to resolve detail. If you understand optical engineering jargon, the spot sizes (blur size) would be equivalent for my 8x10 print example.

Tone, contrast, etc are functions of coating and the lens design type. Assume those parameters are the same to avoid muddying the waters.

Oren Grad
11-Apr-2015, 21:16
Tone, contrast, etc are functions of coating and the lens design type. Assume those parameters are the same to avoid muddying the waters.

It's the recording medium that blows up the equivalence. Even the sharpest/finest grained general-purpose 35mm films can't record all of the information that a really good 35-format lens at optimal aperture sends their way.

Jody_S
11-Apr-2015, 22:05
What you'll find is that everything scales, including aberrations.

The resolving power of the film does not scale. The size of grain on FP4 in 35mm is the same as the size of grain on FP4 in 8x10. That sets a very hard limit on what information can be recorded, if everything else is optimized. Of course you won't see that on an 8x10 print.

Taija71A
11-Apr-2015, 22:27
... if you take your different sized negatives and make prints all of the same size (ie all 8x10s), you will find that their image quality is all equivalent...

Ignoring for a moment your definition of 'Image Quality'...

This of course is simply not true. If it were... There never would have been a need for Larger Film Formats and Professional Photographers around the world -- Would all still be shooting Kodak Disc or 110 Film. :(

Nodda Duma
12-Apr-2015, 03:54
The resolving power of the film does not scale. The size of grain on FP4 in 35mm is the same as the size of grain on FP4 in 8x10. That sets a very hard limit on what information can be recorded, if everything else is optimized. Of course you won't see that on an 8x10 print.

The original thought exercise ignores the limiting resolution of the film, focusing instead on what the optics are doing.

And like you said, with certain films and print sizes you wouldn't see differences due to grain. If it's a concern we could instead say super-duper high resolution digital back ;)

Nodda Duma
12-Apr-2015, 04:07
Ignoring for a moment your definition of 'Image Quality'...

This of course is simply not true. If it were... There never would have been a need for Larger Film Formats and Professional Photographers around the world -- Would all still be shooting Kodak Disc or 110 Film. :(

It absolutely is true! All you're doing is scaling the aberrations and then reversing that scaling with the prints. Refer to any text on aberration theory.

What you are missing is that the development of the imaging media (grain size of plates and then film) drove the size of the camera.

Extrapolating to smaller sizes (and ignoring 110/disc which I think goes too far with film), modern digital focal plane arrays follow the same rule.

Emmanuel BIGLER
12-Apr-2015, 04:37
From Oren Grad
It's the recording medium that blows up the equivalence. Even the sharpest/finest grained general-purpose 35mm films can't record all of the information that a really good 35-format lens at optimal aperture sends their way.

Agreed with Oren so far, but the controversy between us will start when we try to define what "general purpose film" means, be it 35 mm or any other kind. So I can't resist pushing the exchange to a more provocative flavor, to a more theoretical approach ;) (anything theoretical, by definition, is provocative here for our real-life, down-to-Earth photographic activities)

Now that film has been pushed off the mainstream of photography, and that, on a daily basis, we can get new films from niche markets (yes, Kodak, Fuji and Harman-Ilford are still there, but look at the offer in B&W films from Agfa-Mortsel-Maco and ADOX) I'm wondering whether restricting the discussion to "general purpose" films is relevant or not.

The original question addresses the limits of LF film recording vs. small format recording on Silicon sensors.

Taking into account that anybody can freely purchase microfilm-type of B&W silver halide films, with resolving powers exceeding 400 cy/mm, and can easily use exotic chemistry to process microfilms in good gray levels, I friendly disagree with Oren.
And we can add that using the cumbersome, but actually tractable tri-color separation process, (http://trichromie.free.fr/trichromie/index.php?category/One-shot) B&W or color images is not an issue as well if we are ready for ... a cumbersome process.

If we push to the limits, the sensor is not the limiting factor, but the optics (to state clearly why I disagree with Oren ;) ).

In the past, comparison between images shot on 35 mm film and images shot on 8x10" films were irrelevant because each format and equipment, for professional use, was aimed at different jobs and different customers.
But it was no mystery that 35 mm images could never compete with 8x10" images.

Now that amateurs who dare to spend $4000 in digital photographic equipment (without any idea of any future return-on-investment) can get a 36 Mpix silicon sensor, with sensor pixel pitch @5 microns, hence capable of detection 100 cy/mm, the old comparison between film formats, to say the least, has to be re-examined from scratch.

We should remember that in the XX-st century, the trend has been to smaller and smaller film formats and camera sizes. The trend actually stopped with Kodak 110-size film and APS; 110-size images were found unacceptable by the discerning amateur, and APS films came too late to prove anything but the imminent demise of amateur film-based photography as a mass market.

Smaller film size & camera sizes were made possible thanks to the improvement of films, not to the improvement of lenses; but actually the effort in R&D for photographic lenses was certainly the most important (in terms of man.months) in 35 mm photography, after, say, ~ 1955.

Fine grain in B&W silver halide, we already had this with Lippmann plates the end of the XIX-st century. But with such a low ISO rating that ... well no further comment is necessary.

Finer grain at a given ISO rating and better color rendition, actually if you look at the improvement of color slide film, from Kodachrome 25 to Ektachrome E-100G and Fuji Frovia 100 F, same fine grain but 4x ISO rating. Same applies in pure b&W films for the comparison of conventional silver halide grains with "modern" flat grains. It means that the photon detection efficiency of amateur films has been dramatically improved in the last decades of the last century.
But even so, photography is not restricted to photojournalism and the "decisive moment", with moving subjects in poor light and hand-held cameras. For us who love to carefully set-up the camera in front of something eternal like red rocks in the West, or the Rouen cathedral like Claude Monet, the "decisive moment" and high ISO-ratings, are secondary.

Here enters Silicon on the scene.

If we look at a classical high-performance B&W film like Kodak Tri-X, the equivalent photon detection efficiency (or equivalent quantum detection efficiency, this is exceedingly difficult to define for film) is about 0.5%, whereas any amateur-grade Silicon sensor features a quantum efficiency above 30%.
It means that virtually no-noise images can be detected if you allow enough photons to be detected.
So, set you DSLR ISO rating to the minimum and allow a generous exposure time with your small-format camera on a tripod, the 35 mm format on SIlicon, provided that good lenses are used, competes with 8x10" on film in terms of granularity and noise, or at least the question of comparing 35-mm size Silicon sensors and 8x10' film is not irrelevant.
The question of resolution is important but now only secondary, since we can have access to more than 20 Mpix sensors on a 24x36 mm size (and up to 1 Gbyte on a bigger image size, using recent scanning backs).

Now regarding really small size Silicon detectors, downsizing digital cameras and sensors, the only limitation I see is diffraction.
The reason I see to explain why there is a limit to downsizing camera and sensor formats is that the wavelength of light is fixed and not scalable.

The minimum angular feature detectable in object space in a diffraction-limited lens is (λ / a), where "a" is the diameter of the entrance pupil of the lens and "λ" the wavelength of visible light.
Let's consider a worst-case approach with λ = 0.7 microns, this is the actual limit of the human eye in terms of sensitivity in the red-end of the visible spectrum.
0.7, this makes calculations easier : 1/0.7 ~ 1.4 we fit within a well-know series of f-numbers.
It means that for an entrance pupil of diameter 7 mm, the smallest angular feature detectable in object space, whichever the focal length might be, is about 0.7/7000 = 1/10000-th of a radian.
This is already extremely good if we remember that the human eye can detect something like 1 minute of arc, 1 / 3440 radian.

7 mm of diameter for the entrance pupil corresponds to the following f-numbers and focal lengths,
going up in focal length from the 35 mm world:

f/5.6 - 40 mm or f/8 - 56 mm in this range we have some state of the art lenses covering the 24x36 mm format used at their best f-stop
f/11 - 80 mm no problem with a standard lens for the medium format, we could even open by one or two f-stops
f/16 - 110 mm easy for a good MF of LF lens
f/22 - 150 mm! here fit our standard lenses for the 4x5" format, f/22 is the classical recommended f-stop for maximum homogeneity of image quality / coverage with movements
f/45 - 300 mm no problem with the 8x10" format, we even do not have to stop down so much!

Now let's downsize, trying to keep the same 1/10000-th of angular resolution in object space
f/2.8 - 20 mm: probably a good lens of this kind can be found, but do not stop it down beyond f/2.8 if you want to compete even with a modest medium format camera!!
f/1.4 - 10 mm: well a diagonal of 10 mm is not uncommon in amateur-grade silicon sensors, but f/1.4 is less common!
f/1.0 - 7 mm focal length: many, many amateur cameras or cell-phones use 7mm sensor, but @f/1.0, we clearly enter into Science Fiction.

Why? Simply because we actually demand a minimum of image quality and angular coverage to our photographic lenses, and that wide apertures in the f/1.0 - 1/1.4 are detrimental to image coverage and quality, even in an hypothetic diffraction-free optical word governed only by the laws of geometrical optics and aberration correction.

My conclusions are simple, as of 2015 - 04 - 12

1/ We are now back to the old hierarchy between amateur cameras and professional cameras.
A cell-phone is in direct competition with amateur-grade digital cameras with no wireless communication capability, but not in competition with 35 mm (sensor size 24x36 mm) digital cameras, be they reflex or not, with a mirror or mirrorless.
Exactly like in the last century the photojournalist's 35 mm film camera had no ambition to compete with a 8x10" camera.

Million of people are happy with 7 mm focal length covering a 7 mm Silicon sensor, they do not even know what a f-stop is, most probably they use f/4, hence their angular diffraction-limited resolution is probably 10 times worse that any standard medium-format or 4x5" equipment that we use here routinely. But who cares ?

2/ probably we come to a more stable market, downsizing sensor sizes cannot be continued without degrading image quality, like 110 film was the very end of film-size downsizing; but for 24x36 mm sensor sizes I would not be foolish enough to state that with 36 or 40 Mpix as of 2015-04, we have reached the ultimate number of pixels that can actually be manufactured and marketed in the $2000-$3000 range of street prices.

Nodda Duma
12-Apr-2015, 05:10
Good analysis Emmanuel... I will point out that if, as an imaging system designer, I need better angular resolution than 100 microradians, I simply choose a more appropriate focal length, f/#, and focal plane array. Usually the trade off is in field of view.

Ken Lee
12-Apr-2015, 05:13
Now regarding really small size Silicon detectors, downsizing digital cameras and sensors, the only limitation I see is diffraction.
The reason I see to explain why there is a limit to downsizing camera and sensor formats is that the wavelength of light is fixed and not scalable.

Clarification please:


MAF (minimum angular feature) = (λ / a), where "a" is the diameter of the entrance pupil of the lens and "λ" the wavelength of visible light
λ = 0.7 microns (wavelength of red light)
a = 7mm
Then MAF = 1/10,000 radian
The eye can detect 1/3440 radian

How do we determine that 7mm is the diameter of the entrance pupil ? Why is 1/10,000 radians the desired limit ? How does that relate to what the eye can detect ? What happens when we enlarge the image ?

rfesk
12-Apr-2015, 05:21
Off on a somewhat different tangent - years ago there was an article on Photo-do (still around?) that showed with
star tests that 6X6 was the most practical and efficient format (with lenses to match) for practically all photography.
Anyone remember that? Haven't seen the article in 10 years of so - so maybe the results were disproven.

I believe it was because the combination of format, limits of sharpness of lenses, and film sharpness gave the sharpest rendition at
normal viewing distances of enlarged prints.

chris jordan
12-Apr-2015, 09:04
Hi guys, really interesting thread, thanks for all the thoughts. In the original question I tried to eliminate the film grain/format size issue by specifying that the 8x10 camera has a high resolution digital back mounted on it. I am trying to equalize all other factors and compare only the optics. What I'm hearing is that 35mm optics outperform 8x10 by a lot, which is what I suspected. I am curious if 35mm optics outperform medium format as well? So let's say the LF camera has been eliminated from the playoffs, and now we run the same test with a top-quality 35mm and MF rig side by side, each with a 200mm lens, same aperture and exposure, focused on a distant target. We import both raw images into Photoshop and crop the MF photo down to the size of the 35mm and compare the resolving power of the two systems. Would 35mm still win, and by how much?

voigtf64
12-Apr-2015, 10:05
great thread, but I have just discovered a game changer , the Sigma DP Merrill Foveon series of cameras , these are producing {and this is my observation) large format quality images in both detail and tonal gradation, OK so they are
dreadful on battery usage and asa speed, but for an old film person like myself they are what I have been missing in current technology, not perfect there is a potential to burn out highlights, but really give them a look ,the DP1, DP2 and DP3, mind blowing capability.

chris jordan
12-Apr-2015, 10:34
voigt, agreed, I had a DP1 for awhile and the images were amazing, incredibly rich and lush feeling, I loved that camera. For me the low megapixels limited it to being my personal snapshot camera, but rumor has it that there's a much higher resolution version of that technology in the works, possibly from several different companies at once. Fingers crossed.

Corran
12-Apr-2015, 12:13
If you equalize all sensor technology (same sensors, just with more area for different formats) the answer is obvious. The best lens wins.

paulr
12-Apr-2015, 16:56
Back to the OP's somewhat theoretical question, I think the very best small format lenses of mid- to long focal lengths would win this today.

Quality of lenses in the small digital camera world is all over the place, but the best models are insanely good.

The very best wide angle lenses are probably the Schneider and Rodenstock lenses for medium format technical cameras. If you compare the MTF curves of these lenses to the same companies' LF lenses, the differences are almost obscene.

paulr
12-Apr-2015, 17:33
Now that amateurs who dare to spend $4000 in digital photographic equipment (without any idea of any future return-on-investment) can get a 36 Mpix silicon sensor, with sensor pixel pitch @5 microns, hence capable of detection 100 cy/mm, the old comparison between film formats, to say the least, has to be re-examined from scratch.

Here's a comparison from actual prints. I've been making 60" prints from 36 megapixel files from one of these 5 micron pixel pitch cameras. They look significantly better than 50" darkroom prints from 4x5 (TMX 100 negatives). They look roughly comparable to 50" digital prints from desktop scanned 4x5 negatives. I'm willing to bet that they don't look as good as digital prints from high quality drum scans of 4x5, although I haven't tried.

The film here really isn't the limiting factor. TMX has good modulation (high signal / noise ratio) at least up to 150 lp/mm and possibly well beyond. Getting the detail off the film is the hard part. Drum scans are required if you're printing big and want the best possible quality. And here's the catch: the cost of 20 drum scans equals the price of my camera. Never mind the price of film and processing.

As far as I'm concerned, that dslr was an incredible bargain. It's paid for itself a few times over in the 3 years I've had it.

Jim Andrada
12-Apr-2015, 21:28
And now Canon is offering a version of the 5D with a 50 megapixel sensor.

Regardless, I like film better!!!!! Don't bother me with facts, my mind's made up! Then again a 50Mpx camera paired with a Zeiss OTUS sounds like something that would be worth a try. I read a few reviews of the OTUS and people were complaining about the price, but compared to a $40k video zoom or even a $5k video prime, the Otus isn't bad.

mmerig
12-Apr-2015, 22:16
Here's a comparison from actual prints. I've been making 60" prints from 36 megapixel files from one of these 5 micron pixel pitch cameras. They look significantly better than 50" darkroom prints from 4x5 (TMX 100 negatives).

Hmm, 5 micron pitch at 42.3 X enlargement would be 120 dots per inch of pixels or "grain".

60 inches*25.4mm/inch/36 mm = 42.33
42.33 * 5 microns = 211.66 microns
211.66 microns /1000 microns/millimeter = 0.21166 millimeters
25.4 millimeters/1 inch/0.21166 millimeters = 120 pixels (dots) per inch

Of course these calculations don't take into account Bayer filter arrays (which could make it worse).

I think 120 dpi would not look that good to me on a print. If I optically (darkroom) enlarge an FP4 4 by 5 negative to about 50 inches, the grain is not noticeable without magnification. The grain is quite noticeable without magnification if I do the same enlargement with a 135 format FP4 negative taken of the same subject at the same time.

I can post examples if anyone is interested, although my flatbed scanner subdues the detail on both prints somewhat.

Emmanuel BIGLER
13-Apr-2015, 02:33
From Nodda Duma

... as an imaging system designer, I need better angular resolution than 100 microradians, I simply choose a more appropriate focal length, f/#, and focal plane array. Usually the trade off is in field of view.

I agree with Nodda Duma: with our photographic lenses, usually we love to get a one-shot image (no stitching) and we like to have a minimum of field of view!

If I had to design a military surveillance camera, without disclosing here anything top-secret, I can tell you that 7 mm for the entrance pupil is far too small: if I had an unlimited budget and unlimited storage space at home & no problems for transportation of the huge instrument, I would vote for a telescope lens in the genre of the Hubble space telescope, coupled with some kind of scanning system. For landscape images this would be THE ultimate, as they often say in commercial ads we love to read ;)

----------------------------------------

From Ken:
How do we determine that 7mm is the diameter of the entrance pupil ? Why is 1/10,000 radians the desired limit ? How does that relate to what the eye can detect ? What happens when we enlarge the image ?

Hi Ken.

I did not mean that 7 mm was anything optimum nor ultimate. I just wanted to put some realistic numbers to play with the basic diffraction limit and see where we are. I forgot to mention, but most of us know this, that the resolution limit for the human eye is close to the diffraction limit. If we take a size of 2 mm for our entrance pupil in daylight, we get, @.7 microns of wavelength, something like .7/2000 ~= 1/2850 ; one minute of arc is 1/3440.

Regarding the size of the entrance pupil vs. the diffraction limit, I have already mentioned here that if we make a compilation of the best f-numbers for state of the art standard photographic lenses for film use, lenses available in the 1980's, we get a trend as follows:

best f-number = (focal length in mm) / (8 mm)

This corresponds to a constant entrance pupil of 8 mm for all formats!

However, in a recent discussion as Ken knows,
http://www.largeformatphotography.info/forum/showthread.php?121548-CZJ-Tessar-quot-critical-f-stop-quot (]http://www.largeformatphotography.info/forum/showthread.php?121548-CZJ-Tessar-quot-critical-f-stop-quot)
one of our readers quotes a Zeiss source, the critical "f-stop" for a CZJ F/4.5 - 180 mm Tessar being f/12.
But we need to stop down a little more in rdre to get a better homogeneity in the field, we sacrifice resolution at the center to slightly increase resolution in the edge. 180/8 yields about 22 which is a significantly smaller aperture than f/12. The situation is what we are facing all the time in LF photography, if we want to somewhat shift or tilt, we need more image circle and at f/12 probably the 180 mm tessar will be somewhat limited.
Hence if we do not take into account film grain & resolution, all those lenses deliver the same angular resolution in object space, with the same useable field of about 50 to 60 (can go up to 80 for some standard view camera lenses). So there is no advantage of any kind using a bigger camera, except of course film grain that becomes, in relative values, negligible with bigger film formats. However one cannot extrapolate this "8 mm" trend to smaller formats, however for sure you will hit the "f/1.0 wall" quickly when focal lengths and formats get smaller and smaller.

May be the most recent standard 150 mm view camera lenses for the 4x5" format can be stopped down to f/16 and not f/22, for those family of lenses the trend would be something like (f in mm) / (11 mm) i.e. a bigger entrance pupil and a better angular resolution.


-----------------------

From PaulR:

Here's a comparison from actual prints....the cost of 20 drum scans equals the price of my camera.

Paul, many thanks for sharing your experience with us, yes this is were we are now in the real world as far as good prints are concerned. And it is better to face this reality: if we love our film camera it is for many good reasons, but not to outperform recent digital cameras.

This a problem when a friend comes to you and asks you to help him choosing a classical medium format film camera or a large format camera. Before entering the discussion, we should explain him that if he tries to scan his images with an amateur-grade scanner, he will never outperform what he actually gets with a recent digital camera, at least with film in medium format.

A possible answer for B&W could be: use the traditional wet darkroom, do not try to scan. But even with an amateur-grade flatbed scanner, for a 8x10" final print, I have several good friends faithful to their film cameras who will never go back to the wet darkroom, the digital darkroom being so flexible!