Wow - read this re: Film versus Digital debate!

Happy Thanksgiving, Rich. It was entertaining, wasn't it? And if I'm in NJ, I'll take you up on that beer. Also, happy thanksgiving to all.

Happy Thanksgiving to all!

 

It's been a pleasure (mostly) :)

 

And who says it's over? I picked up a copy of Baines (1969) & Mees (1966) and have been doing some holiday reading... can't believe how many of our initial questions are all well answered & spell out there!

 

I will be reporting back, either to your dismay or not!

Rishi

Looking forward to it, Rishi.

 

And Bernie - nobody has come forward to answer my challenge - so now what?

Hello,

 

very interesting thread. Please let me add some test results to the theoretical knowledge presented here.

 

I am currently involved in an intensive test programme about resolution of modern films and sensors. We have not

finished all tests at the moment. A lot of work has still to be done. Some films and sensors have yet to be tested.

Probably we can publish all of our results next year. But so far we have found that a lot of modern films have

significantly higher resolution than the best digital sensors. The physical resolution limit of a 22 MP 24x36mm

sensor is 80 Lp/mm, the resolution limit of a 24x36mm 24 MP sensor is 85 Lp/mm. We reached it (as well as

dpreview for example, look at the Canon Eos 1 Ds Mk III test there). With many films we've got resolution values

above 100 Lp/mm for system resolution (film + lens). With most of the modern ISO 100/21° films such resolution is

possible (with very good prime lenses even higher resolution is possible). We are testing at a very moderate object

contrast of 1:20.

 

With Fuji Provia 100F, Velvia 100F, Velvia 100, Sensia 100, Astia 100F, Kodak E100 G(X), Elitechrome 100, Kodak

TMX, Fuji Neopan Acros and Agfa APX 100 we have achieved 100 Lp/mm. The 100 Lp/mm pattern is so clear with all

these films that we can conclude, that the resolution limit is significantly higher than this value. Therefore we will

continue our test trying to measure where the resolution limit really is (by the way: you need a 35 MP 24x36mm

sensor to resolve 100 Lp/mm).

 

Even with Fuji Provia 400X, Fuji Superia Xtra 400 and Kodak TMY-2 we achieved 100 Lp/mm. With the new Rollei

Advanced Technical Pan (ATP), developed in the dedicated ATP DC developer and exposed with ISO 40/17° we

achieved so far a system resolution of 140 Lp/mm. The 140 Lp/mm pattern is so clear that this film as well is not at

its resolution limit, further tests for higher resolution are necessary. This film has such extremely high resolution,

sharpness and fine grain that you can get results with 35mm film which are on a similar level with medium format.

And now this film is obtainable as 120 as well.

 

We have achieved all these resolution values not only under the microscope, but also on paper in the traditional wet

darkroom. We are using the Rodenstock Apo-Rodagon N 2,8/50. With a slide projector using a very good projection

lenses (Docter, Zeiss, Leica) we have got so far 100 lp/mm on screen with the color slide films and 140 Lp/mm with

Rollei ATP (much much more than every beamer on the market can deliver). Our tests ar not finished so far, because

we have not reached the resolution limits of some films and lenses (both projection and printing). Therefore we are

testing some films for even higher resolution values in the printing and projection. Then we can publish all of our

results.

 

But I want to add some results of other scientific resolution tests made by Carl Zeiss. Zeiss is clearly emphazising

the qualities of film, especially it's superior resolution. Zeiss is one of the world's most important optic companies

(probably the most important). About 99% of their camera optics are made for digital cameras. But nevertheless they

say clearly that film has higher resolution in many cases (of course dependant on the film you are using). Here are

some of their test results (using normal Zeiss 35mm and med format lenses): Agfa APX 25: 200 Lp/mm; Kodak

Ektar 25: 200 Lp/mm; Kodak TMX: 180 Lp/mm, Fuji Neopan Acros 100: 160 Lp/mm; Fuji Velvia: 160 Lp/mm; Kodak

BW 400 CN: 150 Lp/mm, Kodak Porta 160 VC: 160 Lp/mm; Kodak TMY (old): 120 Lp/mm, Fuji NPZ 800: 110

Lp/mm. And the absolute resolution record with Spur Orthopan UR with 400 Lp/mm (all published in their

magazine "camera lens news" No. 17, 19, 20, 22, 24 on their home page). Best regards, Henning.

Wow, very interesting. For me, a good picture is not only megapixels or lines per millimiter, but reading those things, well, put a BIG smile on my face ...

Vijay... see my last post for your answer (well, it was actually 3 posts by the time I got the wording right).

 

Rishi... I can't wait to see your interpretation of the process. I had a go at explaining quite a bit of the process, but without a (recent) strong chemistry background, I couldn't delve into all the specifics.

 

Henning... Just a query on the resolution numbers you quote. Are these of black and white resolution charts? If so, this won't help us elucidate the binary or non-binary nature of b&w film grains. If the charts contain grayscales, then that is a much different story, and much more interesting results.

Bernie - you said that if no one came forward, you'd believe in "gray grain" (partially transmissive silver specks) - and after about a week, nobody came forward. Just asking if you believe in "gray grains" now, or you'd still like to wait. Just asking.

 

Henning already said that he did the tests for the following B&W films - Rollei ATP, TMX, Neopan Acros, TMY2 and APX 100. He also stated that they did the test at an object contrast of 20:1, which is essentially two gray tones, or grayscale.

 

So the story is indeed, as you say, much different, and the results are indeed, as you say, much more interesting. What say you?

"Henning... Just a query on the resolution numbers you quote. Are these of black and white resolution charts?"

 

Bernie, we have used resolution charts with black and white lines (linepairs) and "Siemenssterne". The object contrast of the black and white lines was 1:20, a bit more than 4 stops (both on matte paper, not a "deep" black and not a "brillant" white). We used this moderate contrast (and not 1:100 or 1:1000) because in real life photography an object contrast of 1:20 is existant in details of most scenes. It is relevant for normal daily photography (and not only in a laboratory surrounding).

The "Siemensstern" show you that film resolution is equal in horizontal, vertical and diagonal direction. But with digital sensors you get max. resolution only in vertical and horizontal direction. In diagonal direction the resolution is significantly lower, due to the square pixel shape and the Bayer-Matrix.

 

With my posting I didn't want to refer to the discussion of the binary or non-binary nature of b&w film grains, but to the question of the resolving power of film/sensor/lens combinations, which showed up in this very long thread as well.

 

Vijay, we have tested a lot of more films than mentionend. I gave only some examples in my posting.

Best regards, Henning.

I can't wait to see this paper Henning et al. will be putting out!

 

Also, I'm so tired of conflicting reports saying resolution of film is limited by diffraction limits of lenses, and other reports saying no this is not the case, that the resolution of 35mm film is too low for lenses to be the limiting factor. Which is it?!

 

Also, why do the curves in film MTF graphs stop at 20-30% response? I don't even see the data that shows Fuji Velvia can resolve down to 160 lp/mm, because the curve cuts off at like 80 lp/mm. Please someone point me to the data?

 

Thanks,

Rishi

Rishi - in electronics, we consider the bandwidth of a system to be defined generally by the 3 dB points. 3 dB refers to 70.7% response (50% power, which is the square of the signal). So actually, film manufacturers, if they accurately wanted to show bandwidth they should cutoff much sooner, like at 70% MTF. This would show the bandwidth for signal not being attenuated significantly.

 

Unfortunately this would result in an artificially low resolution number (like 10-20 cpmm) which is not truly representative of the system because our brain "reconstructs" detail even if it is at lower contrast. It is a matter of judgment about where you "stop" the curve, but a 20-30% response is being conservative from their side. If they stopped too soon, you'd get artificially low resolution numbers, and if you stopped too late, you'd get resolution numbers that were so high that people wouldn't hit them except in controlled conditions, and some people could accuse film manufacturers for misrepresenting their products.

 

Since 80 cpmm is considered high enough for most applications and 20-30% MTF means you can still extract some more resolution if you really tried, I guess that is why manufacturers use this number.

<i>Bernie - you said that if no one came forward, you'd believe in "gray grain" (partially transmissive silver specks) - and after about a week, nobody came forward. Just asking if you believe in "gray grains" now, or you'd still like to wait. Just asking</i><p>

 

Yes OK Vijay, you annoying mosquitoe ;). I'm leaning towards the gray grain hypothesis. But I will reserve my final opinion until AFTER we hear Rishi's interpretation of how the process works.

>Diego Buono , Nov 17, 2008; 11:52 a.m.

To Dave Wyman who wrote: You've been living in a cave with Bin Laden? No, even he has a digital camera.

 

Dave I have'nt gone digital, actualy I leave in a cave as you can see..... may be this is the reason I havent gone digital :-

)

 

Anyway, I don't live with Bin Laden! :-)<

 

 

Diego, I wish I lived in your cave. I'm in more of a hovel. ;-)

I looked at a bunch of film under a microscope tonight.

 

I actually ended up placing the Scanhancer in the optical path of the microscope (between light source and film) because the air bubbles in the Fuji emulsions were so distracting that I could hardly focus on the image itself! Funny, it worked! The Scanhancer reduced the number of bubbles I saw quite a bit (or, reduced the overall intensity of bubbles... whatever, end effect: made the bubbles much less noticeable, but certainly didn't get rid of all of them). ARGH, pepper grain. I hate it.

 

Anyway, whoever posted the image earlier showing the dye clouds... yup, that's pretty much what I saw (though I was at much lower magnification). It's amazing... in the end all you see is magenta, cyan, and yellow dyes... perhaps some other darkening dye in dense portions... I couldn't really tell.

 

A patch of the green sign I posted above on I-5: you can actually see, if you look very hard, the cyan and yellow dyes making up the green of the sign. Interestingly enough, where the writing "Express Lanes" should have been entirely white, there is some cyan leakage into the white. What does this mean? That where white light should have been hitting the film, possibly some red(ish) photons did not productively cause enough exposure in the red-sensitive layer (not enough grains extended to that particular region... and remember, this is a long exposure with a very small aperture), and so significant cyan dye formed there.

 

The red regions: you can see dye clouds of magenta and yellow. Pretty amazing.

 

Anyway, to get to how this relates to the discussion: both an Imacon at 8000ppi and a Nikon LS-9000 at 4000ppi resolve most of the information that's on the film. Except for shadows of Velvia... regions near Dmax -- the Imacon could not pick up some of the tree branches that I could distinctly see under the microscope. Haven't run the same comparison on a LS-9000 yet...

 

In my subjective opinion, so far, I don't think that the Imacon at 8000ppi really resolves any useful detail that the LS-9000 can't resolve at 4000ppi. Other than the pepper grain. Which is undesirable.

 

Rishi

Rishi... stick some b&w film under the scope and see if you can see any grey grains. Re the colour trannie: what's your conclusions about resolution and the dye clouds?

Don't have any B&W film... yet. Will shoot some.

 

My initial conclusion is that there is no more than 20MP of useful information on a 35mm Velvia frame. I picked one of the smallest features visible (a leaf of a far away tree), and compared the 4000ppi & 8000ppi scans to what I saw under the microscope. Both scans sufficiently resolved the tiny distant leaf, and the film under the microscope may have had a tiny bit more 'resolution' in the sense of blobs of dye... but when is enough enough? In that case, I could say that the 'resolution' of film is much higher and won't be met by a scanner until the scanner resolves some of those dye clumps I was barely able to see under the microscope!

 

Which is meaningless.

 

There's no point in discerning those dye blobs because it's many of those dye blobs that, from a distance, give the illusion of a particular tone.

 

For example, in the image of I-5 from the bridge above, you can barely make out the blue/red I-5 interstate sign. Under high magnification, that familiar 'shield' shape blue is rather ugly... the analog-to-digital conversion process actually made it more 'recognizable' or 'readable'... much as zooming out on the microscope made it more the familiar I-5 sign I am used to. Close up, though, it appeared as interspersed cyan blobs... take a step back, and it's a nice continuous cyan color.

 

If the resolution of the film really were much higher than what the 4000ppi scanner is picking up, then the cyan/blue of the I-5 sign would have been a continuous cyan/blue, right? But remember, as Vijay so eloquently pointed out above, that the resolution is limited by the largest grains in the emulsion (remember, silver ions are free to move around within a give grain, thereby making subgranular resolution rather impossible). So basically, in this scenario, if some grains in the blue sensitive layer weren't exposed enough, some yellow dye may have formed there, when, really, only cyan & some magenta dye should've formed. Conversely, perhaps grains in the red and/or green sensitive layers within this area got exposed, thereby not forming *enough* cyan and/or magenta dye.

 

That was a rather rough description; of course take it with a grain of salt. What I'm trying to say is that with all the randomness in the process, I don't think there's any more useful detail beyond 20MP. Personally, I'd probably say less than 20MP.

 

But I will be shooting some resolution test charts and doing some scans to better determine this.

 

I think someone already did this for the Nikons... right? Does anyone have that link?

 

Rishi

Oh, and also, as I mentioned previously, there's DEFINITELY detail in shadows of Velvia that the Imacon 848 DOES NOT pick up.

<i>Bernie: Rishi... stick some b&w film under the scope and see if you can see any grey grains.</i>

<p>

Make sure you underdevelop first - i.e., pull the film by a stop or more.

<i>My initial conclusion is that there is no more than 20MP of useful information on a 35mm Velvia frame.</i>

<p>

Careful - are you sure you were film limited and not lens limited? Your lens itself may not have the ability to

resolve more. You have to get the lens limitations out of the way by choosing a really good lens and using the

center part etc. You have to get vibrations out of the way by using flash, and so on and so on if you want to

isolate the

resolution of film alone.

<p>

So if 20 MP useful info comes from a "casual" test, dare I say that a controlled test will provide say 25-28 MP?

And for a fine grained B&W - 35 MP is not so far fetched? I mean as an upper limit of course, under controlled

conditions and all.

<p>

And you are quite right - you saw for yourself that the largest crystals (and eventually the largest dye blobs)

limit resolution. No surprise here.

<i>"And you are quite right - you saw for yourself that the largest crystals (and eventually the largest dye

blobs) limit resolution. No surprise here."</i>

<p>

I'm sure you realize this, but larger crystals don't particularly have anything to do with larger dye blobs.

They're resolution limiting in the sense that a larger crystal may barge into an area of the film that, for

example, is not supposed to get any exposure (say a black line in a resolution chart), but the other 2/3 of it is

in an area receiving white light (and therefore exposure) from one of the gaps in between resolution chart

lines... end result? A small portion of the film that was supposed to be black (corresponding to a black line on

the test chart) is partially 'exposed', or partially white, thereby limiting contrast, thereby limiting perceived

resolution.

<p>

Anyway, you're right, that the lens could be limiting. Anything's possible really. Which is why this'll be much

more worthwhile to shoot resolution test charts side-by-side with digital vs film, which is what Mauro is trying

to do. Anyone else performed these rigorous tests? Or can you provide me to links where they've done these tests?

<p>

Also, Vijay, understand that I said there may be 20MP worth of information, but how much of that is actually

useful vs. how much of that could be recreated to similar perceived 1:1 sharpness by a lower resolution digital

sensor, I don't know. If at 20MP you're resolving soft dye clouds trying to represent a distant leaf, could you

reconstruct said leaf with fewer pixels? Do you see what I'm getting at?

<p>

Also, this should be obvious, but dye clouds certainly vary in density/intensity... so this is very different

from our discussion B&W silver deposits. I think.

<p>

Rishi

Forgot to mention that I'm making my comparisons based on an image of Mt. Hood & forests, taken from the side of a road, using a Gitzo 2-series tripod, 2 second shutter delay with mirror lock-up, using a 70-200mm f/4L lens at 75 mm shooting at f/8.

 

The comparison with the signs on I-5 falls apart because it was shot from a bridge. Plus I'm not convinced of how great the 17-40mm f/4L lens is at 17mm at f/18 with 30 second exposures...

<i>Rishi: I'm sure you realize this, but larger crystals don't particularly have anything to do with larger dye blobs.</i>

<p>

Well, the dye cloud size and/or density has to be proportional to the silver formed. Larger crystals would mean that larger amounts of silver will be deposited, which will create larger or denser dye clouds. So not directly as in B&W, but indirectly, there is a connection. Beyond that, I can't really say.

<p>

Useful information in film is a tricky thing; because it is subjective. At the extreme limits, the information may be there, but as you say, it may be difficult to extract and do much with. In the real world, we move to larger formats, such that we don't have to work on the hairy edge all the time, but that is a different discussion. For that reason, because I value quality so much, I mostly shoot 8x10. Now, at 8x10 lenses don't even resolve 40 cpmm with high MTF, and it is common to see 10-20 cpmm. But the immense gain with the size of the film means that you have something like say (40 cpmm * 2 * 25.4 * 8 * 40 cpmm * 2 * 25.4 * 10 =) 330 MP of information. Lets round that down (since I took 40 cpmm) to say 200-250 MP. That's huge, especially since you can extract the most out of your lenses etc. by clamping down on a huge tripod with an additional monopod for stability. You are not film limited any more, but massively lens limited.

<p>

Anyway, just rambling.

<i>"Larger crystals would mean that larger amounts of silver will be deposited, which will create larger or denser dye clouds."</i>

<p>

Well, technically, exposure and dye cloud presence are inversely related in reversal film. So the larger the silver deposit, the less the dye for that particular layer at that spot.

<p>

Anyway, my (our, I think) point being that the larger the grain at any given spot, the more it hurts local contrast for very fine detail (detail any finer than the size of the grain). Say you have grains 10 microns in size. If you're trying to resolve 100 lp/mm, that's 1 line pair per 10 microns... you'd run into trouble with grains 10 microns in size.

<p>

I need to now go and read more about lenses & diffraction limits. In the meantime, let me post this interesting article:

<p>

<a href="http://www.luminous-landscape.com/tutorials/resolution.shtml">Do Sensors Outresolve Lenses?</a>

<p>

Rishi

<p>

P.S. I think shooting 8x10 is crazy. I wish I could because the results must be astounding, but I already complain enough about lugging around a 35mm Canon EOS 3 with 3 lenses, ND filters, adapters, etc. and a Gitzo tripod with a Manfrotto head. Long hikes with anything more would just be a pain in the...

Actually 8x10 isn't that heavy. A wooden field camera runs you about 10 lbs, a few well chosen lenses about 3-4 lbs, film holders and film about 5 -7 lbs, a tripod about 5 lbs, so we are talking about say 20-25 lbs. Say 30 lbs with a darkcloth, loupe, lightmeter etc. There are people who get this down to maybe 20 lbs by being very picky about lenses etc.

 

A good high DSLR runs you about 2 lbs, some high end lenses (L series or the Nikon equivalent) around 8-10 lbs, a tripod around 4-5 lbs so you are already at almost 15-18 lbs.

 

So yes, there is a difference, but it is not that bad, maybe 10 lbs or so.

<center><img src=http://www.geocities.com/dainisjg/jimsn.jpg><P>

Nothing crazy about it.</center>

Oh...my....God...

 

Two 8x10s and one amateur aluminum tripod? Real photogs use wood tripods ;-)