Jump to content

Wow - read this re: Film versus Digital debate!

Recommended Posts

I have always wondered about what it is that people are always thinking about when they shoot a basic camera. It's

not that the film advance lever is that hard to find or you need a book to rotate a f/stop ring but then it finally dawned

on me..'It's where you can dump the thing. Once that popped into my head I piled it all in a box and junked it"

Stopped thinking about manual focus camera's after..Leaves plenty of time to trim my toenails now that I don't have

to think up a storm all the time.


I keep reading about Ken Rockwell and how people do not like his web site..Maybe if you toss out your manual focus

camera's it will clear up some cobwebs in the head and you can remember to not visit his web sight. Problem solved.


As far as film vs digital goes it's no problem. Just shoot what you want to. No brainer on that one.


Any other problems to be solved, just let out a holler. I am just sitting here eating some corn on the cob. It's yummy.

Link to comment
Share on other sites

  • Replies 997
  • Created
  • Last Reply

Top Posters In This Topic

Let me preface what I'm about to write by saying: I find most of Ken Rockwell's stuff very thorough & extremely valuable. I also still shoot & love film. 35mm RVP 50, to be exact. I scan using an IT8 & Hutch Color Target profiled Imacon Flextight 848 (I did the profiling myself using LPROF).


That being said, I have some serious problems with Mr. Rockwell's article. I am only going to discuss some very technical issues, because whether or not my girlfriend likes me to use a digital or film camera is not of concern to me when I'm trying to discern which gives me the best <i>quality</i>. Now, Mr. Rockwell makes valid points, but there are a few statements he makes that need obvious qualification:


First of all, re: film scanning, Mr. Rockwell writes: "The reason film became less popular was because there was such a large barrier between film and getting it into the digital domain via scanning. Now that it's trivial, by having my film all scanned at NCPS at high resolution at the same time it's developed, I can shoot film as transparently as a digital camera."


Film scanning <b>trivial</b>? My friends, please take a look at the 'Digital Darkroom' forum and illuminate yourself with the multitude of problems when it comes to film scanning. For example:



<li>How do you solve the film flatness problem?</li>

<li>Color accuracy when it comes to scanning slide film? Have you profiled your scanner?</li>

<li>Color accuracy when it comes to negative film? What's the best way to remove the cast from the orange mask? Erik Krause's advanced workflow method? Fuji negative film with an IT8 target printed on it for building a color profile?</li>

<li>Fuji pepper grain -- how to get rid of it, unless you only scan with one of the newer Imacons or the Nikon LS-9000 which has an optical diffuser in the light path?</li>



I could go on and on, but you get the idea. I would be seriously skeptical re: whether or not NCPS has figured out all these issues. On their website, I can't even see their scanning method or technique.


So I have to disagree that getting your film scanned is 'trivial', unless you're not a pixel peeper or a color nazi, both of which I am & I suspect lots of other discerning photographers are.


Now, re: his talk of 'dynamic range'. Mr. Rockwell says:


"The dynamic range is so great that the hellacious sunbursts you see are just what's naturally coming off the diaphragm blade at f/8, as if 1,000 suns were shining in the lens in the two-minute exposure."


That's largely because our eye-brain visual system is interpreting it as so because of the stark contrast between the wood of the barn and the light leaking through. Why the stark contrast? Because slide film has extremely high contrast, which is exactly what makes it so hard to shoot & why we use graduated neutral density filters so as to retain color information in the highlights while still capturing shadow detail.


Yes, I agree that on a light table, the difference between the darkest dark (black on Velvia can be incredibly dense) and the brightest bright (depending on the light source of your lightbox) can be significantly greater than that exhibited by a typical LCD screen.


But if we're talking about the dynamic range that can be recorded on slide film, it is at best equivalent to CMOS capture, and at worse, well, worse than CMOS capture. In fact, when I last checked the response curves of Velvia in a Fuji technical guide, the dye response to light was, not linear, but exponential. That means shadows stay dark and highlights get blown out easily. This is why we bracket. But let's give Velvia the benefit of the doubt & say its latitude of exposure is on par with digital... I back this up by the following comment made by Canon themselves. Canon, in a white paper about their CMOS technology, write: ""Canon’s full-frame sensors have reached another image quality milestone as well. Their gradations and dynamic range are now the equal of the best positive films, and their resolution and lack of grain are superior. No smaller sensor has achieved this level of performance." But with the newer 'Highlight Tone Priority' combined with increased signal:noise ratios in the shadows, I'm quite certain that digital capture's latitude will surpass that of positive film (but I admit I am not qualifying my statement). If Mr. Rockwell is talking about the dynamic range that negative film can capture, which I assume he is since he refers to it later on in his article, then, yes I agree with him. Negative film has a logarithmic response to light, since, as photons expose silver halide particles and more and more of these clusters are reduced to silver metal (which is black), less and less light can make it through the film. Beautiful principle, you must agree :)


Finally, I have some serious doubts with that overlay that he has at the beginning of the article, where he shows a 118% crop of a Nikon D3 image and a 100% crop of Velvia 50 scanned at approx. 4000 ppi (if my calculations are correct). Assuming my math is right, his film scan yields a 16.8MP image. He then took his 12MP image from the D3, then scaled up both the width & the height by 118%. So, for example, if the original image were 4000x3000 pixels (12MP), scaling each dimension by 118% you would get 4720x3540, or, 16.7MP. So far so good; the two crops should be comparable.


BUT, here's my problem. His 100% crop of his 118% upsided image from the D3 just looks like <i>crap</i>. There is no reason why a good digital camera should give you a RAW file that when upscaled just 118%, you get that crap. Which leads me to ask these questions:



<li> What ISO did he use?</li>

<li> Did he shoot JPEG or RAW?</li>

<li> If the former, what sort of noise reduction did the camera use?</li>

<li> If RAW, which RAW converter did he use?</li>



Because when I shoot with a sub-par camera (Canon Rebel XTi - 10MP), and a subpar lens (EF 28-80mm II) at f/14 which should actually show diffraction-induced softness on an APS-C sized sensor, here's what <b>I</b> get when I upscale the RAW capture 118%:


<img src="http://staff.washington.edu/rjsanyal/Photography/FilmVsDigital/118PercentCrop_RebelXTi.jpg" width=800>

<a href="http://staff.washington.edu/rjsanyal/Photography/FilmVsDigital/118PercentCrop_RebelXTi.jpg">Link to full-size image</a>


Please click on the link and view the image at 100% to really gauge resolution. Compare that to Mr. Rockwell's 118% crop:


<img src="http://staff.washington.edu/rjsanyal/Photography/FilmVsDigital/D3R_2830.jpg">


So, in my analysis, yeah, my 118% crop is not that great. It's a bit soft. But not as soft as that crop that Mr. Rockwell shows from his D3, which also looks like it's had heavy noise reduction applied. And that's a 12MP camera, whereas this XTi is a 10MP camera using an inferior lens. No sharpening applied.


Interestingly enough, it does look a bit muddied in the shadows, especially in the region of the trees. Maybe the interpolation from the Bayer pattern is rocky in these areas of higher detail/low signal:noise. I'm not qualified to comment. But as those luminous landscape articles point out, sharpness is not all about resolution. Much of it has to do with contrast, edges, smoothness... these latter factors are some of the huge reasons why digital capture tends to look sharper to the human eye-brain system.


But this all got me pretty curious. What I really want to do is a side by side comparison of my film scans to digital capture. Unfortunately, I don't have access to a digital camera right now, but I have some old shots from before. And I have my film which I scan at 8000ppi on an Imacon 848. These scans give me 80MP files, which I then downsize to 20MP? Why? Because I can't deal with 500MB files, and because from articles I've read, there probably isn't more information than ~20-25MP worth in a 35mm frame of Velvia 50. Why scan at 80MP then downsize to 20MP? Because you then minimize the risk of softness induced by interpolation during the analog to digital conversion process of scanning.


Anyway, I decided to throw up these <b>extremely unobjective</b> comparisons (you've been forewarned, so please don't yell at me telling me that my comparisons are completely unfounded). Basically, I'm showing a 1:1 crop from a 10MP RAW capture (Rebel XTi), and a 1:1 crop from 35mm Velvia 50 scanned at 8000ppi and then downsized to a 10MP image (downsized using a bicubic sharper algorithm). The problem is, the images are entirely different. So, really, you can't come to any conclusion :) But, I try to discern sharpness anyway to see if they are at least in the same ballpark. In this extremely subjective comparison, I'd have to say that the 35mm film seems to have about as much, or perhaps a bit less, amount of resolution as the 10MP digital capture. If I were comparing apples to apples, I'd judge this by how 'sharp' or 'diffuse' an edge looks... the lower the resolution of the capture, the more 'diffuse' an edge would look (or, the lower contrast it'd have). Clearly, I can't do that objectively here, but it feels like they are on par. This seems to corroborate previous predictions of well scanned 35mm film having near the resolution of an equivalent 8-12MP digital SLR. Well, take a look at some of the comparisons yourself:


<img src="http://staff.washington.edu/rjsanyal/Photography/FilmVsDigital/FilmVsDigital_Resolution_1.jpg" width=800>

<a href="http://staff.washington.edu/rjsanyal/Photography/FilmVsDigital/FilmVsDigital_Resolution_1.jpg">Link to full-size image</a>




<img src="http://staff.washington.edu/rjsanyal/Photography/FilmVsDigital/FilmVsDigital_Resolution_2.jpg" width=800>

<a href="http://staff.washington.edu/rjsanyal/Photography/FilmVsDigital/FilmVsDigital_Resolution_2.jpg">Link to full-size image</a>


Again, please click on the links and view the images at 100%. Interestingly enough, in the first example, the digital capture seems a bit sharper if you look at distinct, smaller features. But in the second example, the film scan looks sharper. Is it coincidence that the digital capture looks sharper in the case where there are less fine details mostly consisting of sharp, linear, high contrast edges? Probably not... but someone who knows more will have to comment on this. The individual smaller branches certainly look sharper in the film scan in the bottom example in comparison to those branches below I-5 in the digital capture. Of course, this could largely be a contrast issue, with Velvia getting the upper hand due to its high contrast & exponential response to light. Why didn't it win in the top example? Well, the top example is a bit unfair because the digital capture shows a crop of a portion with high contrast in the subject matter itself wheras the Velvia scan shows a crop of a portion with pretty low contrast overall.


Anyway, this is really not an objective comparison, as I've said 100 times, and I'd really like to do a side-by-side comparison when I get the chance just to convince myself (I mostly believe, but have a problem, with Michael's shootout over at Luminous Landscape, but that's probably for another thread. Here's a link to his article on film vs. digital: http://www.luminous-landscape.com/reviews/shootout.shtml).


But I do feel that Mr. Rockwell's article needs some objective qualification. I find it hard to believe those results, as my own results, as well as many other's (on Luminous Landscape, for example) who've, apparently, objectively qualified their methods, disagree.


Yet I do love film & continue to shoot it :) I admit that on a light-table my slides look overwhelmingly better than after I scan them. After scanning them, even with a profiled scanner, they just appear a bit 'lack-luster', and I always find myself trying to apply a tone curve to increase contrast... probably because I want highlights to pop off the image, and darks to remain deep black, much as it does on the light table.



Link to comment
Share on other sites

>You made Michael Reichmann's point for him. Your 50,000x view of film grain is all the proof needed. And your

descriptions of film grain structure are essentially descriptions of a "binary" system. I'm quite frankly baffled that you

don't realize this.


What rubbish if grain is filamentary and passes varying amounts of light depending how many photons hit it how is it


To be binary would mean it only has two states solid black or clear.

hear is a 50K times image.



look at the image Daniel. What about the partially exposed grain on the left? surely it should be clear or black, instead it

has been exposed round the edge near the sensitivity ears 'irradiation' an id most certainly Not binary.

I know it, Kodak experts that helped me write the article know it

Film is NOT binary.

The Riechman article is pure FUD

Link to comment
Share on other sites



<P><i>What rubbish if grain is filamentary and passes varying amounts of light depending how many photons hit it

how is it binary?</i></P>


<P>At the level Reichmann is talking about, it doesn't pass varying amounts of light. It's there are not, black

or clear. There's no gray in the 50,000x image you posted on your blog. I've seen other images like it and I've

looked at my own B&W film under a microscope and at high enough magnification, there is no gray, only black

specks and clear areas. (I forget the magnification I went to in order to observe this, but a standard high

school or college microscope can magnify enough that you see this at play, at least with B&W film.)</P>


<P>Backing off, looking at film at lower magnification, varying density passes varying levels of light. At that

level you can call it "continuously variable" or "analog". But Reichmann was talking about the actual image

forming elements in each system because people tend to make 1-to-1 comparisons between grain and pixels. His

point was that you cannot do that because a grain filament is not a pixel. A pixel represents much more

information. A pixel can precisely represent any one of millions of tones. A single grain filament cannot.</P>


<P><i>What about the partially exposed grain on the left?</i></P>


<P>Sorry, that image also confirms Reichmann's statement. There's either silver or there isn't. There's no light

gray silver, mid tone silver, dark gray silver, etc.</P>


<p>No offense, but I think you don't want to see this. It seems to me that calling film "binary" strikes a nerve

with you. Just an observation. Use whatever term you want, a grain filament is either there or not. There aren't

any 35% gray grains or hot pink grains. A pixel can be any one of millions of colors. That's why you can't

compare grains to pixels on a 1-to-1 basis.</p>

Link to comment
Share on other sites

"At the level Reichmann is talking about, it doesn't pass varying amounts of light. It's there are not, black or clear.

There's no gray in the 50,000x image you posted on your blog. I've seen other images like it and I've looked at my own

B&W film under a microscope and at high enough magnification, there is no gray, only black specks and clear areas. (I

forget the magnification I went to in order to observe this, but a standard high school or college microscope can magnify

enough that you see this at play, at least with B&W film.)"


Thats false.

I'm talking about what happens at a MOLECULAR level MR asserts that grains are either black or white and clump

together to form tone-that just doesn't happen. Grains are filamentary like a wire wool pad and the denser the pad the

more photons have hit it and the less light it passes.

The density of that pad TRANSMITS different amounts of light that is what makes it ANALOGUE.

Here is a photo (which I'll explain)



You'll notice that the grains aren't either black or clear as MR suggests but filamentary far from validating his claim that

the grains themselves are either black or not there it shows that grains can be partially exposed to light.

In that state the amount of light the individual transmit VARIES.

Look Reichmans argument is a binary one, the grains are either solid black (binary) or they are semi transparent as


Those semi transparent grains ARE the grey tones!!!!


"No offense, but I think you don't want to see this. It seems to me that calling film "binary" strikes a nerve with you. Just

an observation. Use whatever term you want, a grain filament is either there or not. There aren't any 35% gray grains or

hot pink grains."



Its not me that doesn't want to see....

The reason I got together with a former Kodak engineer to write this is that MR is wrong, bad information is poor

information and coming from someone who sets himself up to be an educator is poor science.


Grain transmits a varying amount of light proportional to the amount of photons that strike it and the developer that

magnifies that process.


Link to comment
Share on other sites

BTW Daniel:

"His point was that you cannot do that because a grain filament is not a pixel"


A grain filament is 3 single atoms, a bit smaller than a pixel in reality!!! Those grains you see at a macro level are made

up of billions of Ag atoms, your school microscope cannot resolve individual grains for that you need an electron

microscope, the picture above was supplied to me by Kodak- I doubt you school lab has such a microscope.


So to recap grain is filamentary (note: not a filament!) like a wire wool pad, transmits varying amounts of light, and

therefore is analogue.


It is not a black speck or clear, as those black specks have holes!! The grains do not move around within the emulsion

to 'clump' together to form tone (contary to MRs claim) but are layered (sometimes 10 layers thick)

That is what you see when you look though your school microscope many layers of varying density depending on how

dense the grain 'wire wool pads' are.

Link to comment
Share on other sites

Eh, Daniel, you go small enough, the whole universe is "digital" - the correct term is of course quantum. Of

course, a single "unit" is either a molecule of AgBr or an atom of Ag, but that is far too small to be a grain.

If a single grain is some microns across, it would contain what, something to the order of microns/nanometers

cubed - or perhaps as Mark says, billions of atoms or molecules. Unless you insist that these billions of

molecules must all change state from AgBr to Ag simultaneously, grain must have all the states possible - 2 to

the power of that billion or so.


By comparison, a pixel has say 2 to the power of 48 states. (16 bits per color times 3 colors). Sorry, but if 48

bit digital is what you call analog, then a billion bit digital can't be errr... binary. For all purposes, it is

continuous - in fact, the quantization of tones is probably so tiny that you can't even pick it up on the most

sensitive instrument.


Thus, this Reichmann chap's stuff about grain being binary is bunk. However, there may be merit to the argument

that film can't resolve details much finer than an individual grain, but that is a different issue and quite

unrelated to the point that an individual grain can have continuously variable tones.

Link to comment
Share on other sites

I view this a little bit differently - which is what does a final print look like? I really don't care whether a workflow is binary or digital at a molecular/pixel/theory level - this is the "how many angels can you fit on the head of a pin" type discussion - it's the end result that matters.


What I know is that with a good digital camera (I use an M8) - I can make 18x28 inch prints that can be shown along side prints of the same vertical dimension (18-inches) made with a 6x7 film camera (either E100G or Provia film) - and the prints from the digital camera can stand the side-by-side scrutiny.


Now, I'm NOT saying that the digital camera image has the same resolution, but it does not look fuzzy and grainy - which, in a print from 35mm film gives the appearance of lower resolution. What I am saying is that displayed side-by-side, you don't have the grain "fuzz factor" of film in the digital image, you see more details, and you don't have the very apparent resolution disparity that you would have with a 35mm film print along side of a 6x7 film print.


Having printed Ilfochromes professionally for a number of different photographers; from all different types and sizes of transparency film - I can tell you that no 35mm transparency film printed at 18-inches in height would stand the side-by-side comparison to a print made from 6x7 film.


One other comment - if Mr. Rockwell thinks that Ilfochrome does, in any way, make accurate color - he's either color blind, totally unfamliar with Ilfochrome, or delusional - take your choice.

Link to comment
Share on other sites

Just stepping aside from the issue of which is capable of "the best" image quality, let's just say both are capable of very good IQ. I view both as tools that can have different advantages for different applications.


I have digital cameras and enjoy their benefits greatly. Can erase and retake shots on the spot. Great for their immediate on-screen viewing, electronic storage and at-home processing, printing or CD-making, as well as easy sharing through e-mail. Quick, easy access for image adjustments in software like photoshop.


But I still use my 35mm film bodies much of the time, in fact more often than digital.


I like slide film for its straightforward accuracy, that is, in terms of your exposure settings, etc- what you shot is what you got. I can also more predictably expect the kind of color palette I will have by selecting a particular film.


I often shoot print film for its exposure latitude/dynamic range, and its convenience and low cost for prints. For pics of special occasions with friends or family, for the sports events I sometimes shoot, the convenience and deal I get using print film can't be equalled by digital use. I often supply copies of prints for others, for their photo albums. Through our local K-Mart, and another local chain store I use, I get a roll developed with an order for three copies of each print of 5x7 size for about $16. I usually get 25 shots on a 24 roll. That comes to 75 5x7 prints, which is about 21c each! Try getting 5x7's from digital at 21c- no way. Quality from their Kodak Perfect Touch service is very good.


A number of photographers have said to me that they go to film for very long exposures, because with film, the grain/noise does not increase with exposure time.


So each has its own advantages. I enjoy the benefits of both. But others may only need those benefits of one or the other, and may just shoot film or just shoot digital.

Link to comment
Share on other sites

Well Phil, you know what they say about opinions.......Here's another one for ya'! Although his writing style is

a bit .....novel, it's my opinion that Ken's message (and opinion!) is basically sound. I too have used film for

40 years, bought into digital (NIKONBIG!) 3 years ago, and have found the instant gratification to be less and

less rewarding. I've tried everything to get greats shots from digital (and HATE the computer work!) and for me

there seems to be "no place like chrome". In fact I recently have been driven over the edge to a Chinese built

4x5 view camera.....madness takes it toll! I think you've done well grasshopper. -john

Link to comment
Share on other sites

People have been getting great shots from 2mp cellphone digicams to the mammoth polaroid 20x24 camera. I don't think the question is so much about resolution, as in what works best for you and your approach to photography. I find that photography is not so much about the final product as it is about the process of getting to the final product. Using the camera, visualizing the scene, anticipating the result, trying to improve constantly, just enjoying yourself and appreciating; this is what makes photography so fun and special. The killer image is just the bonus at the end of the road.

The only people who should be concerned with resolution, dpi, and end product are the professionals who make a living off of it, who know that if it ain't perfect they won't get paid.

There you go. Out of my system now.

Link to comment
Share on other sites

philip, shooting with a Pentax 67 has great potential for large, finely detailed prints. The question is, what are you going to use it for? Scanning medium format takes a rather expensive scanner unless you have a pro lab that prints from negs or trans. 12 shots per roll of 120 is expensive if you shoot a lot. If you intend on just using chromes to send to an agency, maybe that will be the way to go for you. Check out Walter Tatulinski's folders here on pnet: http://www.photo.net/photodb/user?user_id=605046 to see what you can to with a Pentax 67. On the other hand, if you shoot a lot of images that you intend on scanning and printing, it can get expensive and tedious. I myself moved away from medium format and even large format and love the Nikon digitals. I generally don't print over 11x14 and a 10 or 12 mp camera can make prints at that size that rival medium format (but not large format IMO). I enjoy the ease of use, flexibility in post processing, high iso, low grain, and low cost of making lots of images. I've shot film since the 1960's using all formats up to 4x5. I made all my own B&W prints and developed my own film. I prefer the flexibility of shooting in RAW and processing in ps over working with the chemicals of the darkroom. What works for you depends on your style of shooting, how big you want your prints, what type of post processing you enjoy, etc. Don't obsess about film vs digital a la Ken R. As others have pointed out, comparing is like apples to oranges and not worth agonizing over. It boils down to what are you having the most fun with!
Link to comment
Share on other sites

First of all - Michael, thanks for pulling the subject back down to a level most of us plebs can understand - I was getting a bit lost in the ''binary' 'pixels' and 'molecular' stuff. Furthermore - like my archery mates mentioned above, it looked as if things were about to end up in a figurative stoush, so well done Michael.


Secondly - I think Lucas has pretty well summed up the conclusion that I have reached.


I phoned the chief editor at the Stock Library, and as I suspected things have really changed in the ten years since I was actively submitting. So much so it sounds all a bit daunting now.


So like Lucas suggested I am just going to start taking pic's the way I like it, enjoy the process and if the final results are worth something to somebody - then so be it. If not, then I will still be happy.


In fact, since I penned my submission up top, I received my F5 back from Nikon in Perth. I sent it in for a service and check-up. They did such an awewome job. It only cost me 260AUD. They put on a new grip, pulled it all apart, checked seals, replaced several mechanical parts and recalobrated everything. It is really awesome, it looks and sounds like a brand new camera.


I am about to order either the 80-200 2.8 professional zoom, or the 180 IFED. Either way I will just enjoy my F5 like I used to and will persevere with the Pentax when it arrives. Hopefully I can use both cameras to compliment each other.


If I decide to persue the stock seriously again in the future, and have to get a DSLR, then my new Nikkor lenses can still be used.


Thanks everybody for contributing to this great thread. Even though the comments differed markedly, I enjoyed reading all of your thoughts and observations.


Happy shooting - Phil from Down Under!

Link to comment
Share on other sites

Digital, film, and chalk sticks all make images on paper the same way cheesecake, meatloaf, and garden salad are all food.


Make your own assignments about which item belongs to which course of the meal in your kitchen, but don't try to claim that the three are close enough that one can compare "which is better" than another.


If I rephrase the question to ask is digital better than chalk, then I can get a whole new crop of digital artists into the fight. Some of the stuff that the uber-photoshop folks produce with nothing more than a mouse amazes me, but it's *DIFFERENT* from digital image capture, not better or worse. The same for film. It's just different.



Link to comment
Share on other sites

well I am still using film and will for a long time yet ,digital is great and oh so easy but I just dont have the cash to fork out on digital gear that to my opinion cant give me what my bronica SQ with ISO 100 or 50 film can give me .


maybe the AU$55,000 39 Mp hasselblad will give it tho .well when I can get one for around $1000 I will go digital

Link to comment
Share on other sites

Does no one else here agree that Mr. Rockwell's 1:1 crop from his Nikon D3 image is unacceptably soft, and <i>not</i> the type of results typically seen with 1:1 crops from, say, a Canon 5D?


Also, here's a slightly more objective comparison of a 10MP file from a Canon Rebel XTi vs. a 10MP scan (Imacon 848) from a 35mm Velvia 50 frame (well, more accurately, an 80MP scan downsized to 10MP). Yes the time of day is different and the zoom setting is actually not the same (with the advantage going to the Rebel since it's zoomed in more, hence any feature in the frame is represented by more light-gather elements, or pixels and grain in this case), and the lenses are not the same (with the advantage going to film as a 24-70L lens was used as opposed to the EF 28-80mm on the Rebel).


<img src="http://staff.washington.edu/rjsanyal/Photography/FilmVsDigital/FilmVsDigital_Resolution_3.jpg" width=800>

<a href="http://staff.washington.edu/rjsanyal/Photography/FilmVsDigital/FilmVsDigital_Resolution_3.jpg">Link to Full-Size Image</a>


As always, please follow the link to view the image at 100% to make sharpness comparisons.


If you look at the 'Express Lanes' sign, there's a slight advantage to the digital image, though this is arguable especially in light of the fact that the sign is smaller on the film (shot is zoomed out more) so the fact that the feature is smaller on the capturing-media frame yet still fully legible is impressive. But if you look at some of the rails or the buildings, even though they're smaller on the film scan, they appear to be sharper than in the digital image. The digital image is, however, much cleaner in these areas (whereas you can see the grain on the film in the buildings).


So it's a bit of a toss-up. I'd say that what's capable of being resolved on both these frames is similar, going along with my previous observations that film scanned well (though I haven't tried drum scanning) yields resolution at or near a 10-12MP digital camera. Certainly the cleanliness of a digital image, though, makes images appear sharper to the eye.


Additionally, I find it interesting that straight edges seem to be better represented in the digital capture, whereas more complex edges (e.g. trees) seem to appear sharper in the film. Is this my imagination or is this because straight edges are better represented by the linear array of a CMOS/CCD sensor, whereas the random distribution of silver crystals better represents non-linear complex surfaces/paths?


Any thoughts?


Thanks, <br>Rishi

Link to comment
Share on other sites

Interesting point, Rishi. I have posted before that when testing film vs digital, that what might be descirbed as "complex edges" (or more loosely described as "random") is a better test subject than subjects with lots of straight or simple edges. I suspect that the interpolation algorithms applied to Bayer patterns may be performing some kind of edge enhancement which contributes to digital images looking so sharp. After all, to take the argument to absurdity, a black box which always spits out sharp lines regardless of the subject is indeed producing a very sharp image, but one that has nothing to do with an imaged subject.


Years ago, I studied information theory (rate distortion theory, etc.) I don't have the inclination or time nowadays, but it sure would be interesting if an information theorist would take up the issue of whether a particular digital sensor, say the new Nikon CMOS chip or the chips in the Hassy back, captures more or less information than a particular film and negative size, such as 35mm Kodak portra 160NC or the new Etar 100. The information content of a digital sensor is easy to compute. But for film, it will take some thinking. Any information theory professors out there up for it?

Link to comment
Share on other sites



Yeah, look closer at the buildings in the image I posted above at 100%, and I think you'll see that the absolute resolving power (subjectively) is higher in the film than the 10MP digital shot. Essentially meaning that there's more than 10MP worth of information on a 35mm frame of Velvia 50. Which I'd certainly hope there would be, given that it's the highest resolution film out there (right?).


However, the straight edges of the buildings themselves don't look so impressive; in fact, they look 'fuzzy' whereas the straight edges of the buildings in the digital shot look, well, <b>straight</b>. I'd be willing to bet a nickel that that's because of the regular pattern of the CMOS along with, probably, what you suggested in the Bayer pattern interpolation method.


Yes I'd love to see such an analysis about the information captured by film vs. digital. However, what will render this analysis very difficult will be how to figure out how many silver crystals make up a 'clump' that represents the absolute tone of the smallest feature being recorded? Yes I'm of the Reichmann from Luminous Landscape camp, even though I really wanted to believe Mark Smith here. I will follow up with a post of why, including a number of quotes from Ansel Adams. Of course, I'd like to be challenged and proven otherwise because I <3 film :)



Link to comment
Share on other sites

I thought that this thread was over - apparently not. Benny, a simplified information theoretic analysis for you

follows. Rishi, your image shows some serious vibration or camera shake or scanner shake or <i>something</i>,

especially visible in the Velvia picture. Unfortunately, with camera shake present, any meaningful comparison

about resolution gets tossed out the window.


Look at the traffic streaks - instead of being smooth lines, they look "stepped" - either there were ground borne

vibrations, or it was windy and your tripod (or the entire structure you were on) "rang", or <i>something</i> -

but there is clear evidence of some kind of vibration and its there in both pictures. Naturally this is a

resolution killer; meaningful data is lost and comparisons can't be made or conclusions drawn other than the fact

that both systems are capable of <i>more</i> resolution.


Benny - here's a simple information theoretic analysis.


Lets assume that a very good lens is capable of 100 line pairs per mm, or 100 cpmm (cycles/mm). Nyquist's theorem

states that you need 2x the sampling frequency to capture this information, and reconstruct it without loss. The

simplification here is that the cycles are treated as sine-wave cycles, not as square-wave alternating light and

dark values. In theory a square wave contains infinite information (because there is a transition from light to

dark in time or space approaching zero). Well then, to sample and reconstruct the information, you need a capture

mechanism that can sample 200 cycles/mm.


What does this translate to in terms of pixels? 200 lines/mm x 24mm for the height, and 200 lines/mm x 36mm for

the width of the image. This is 34.56 Megapixels, say around 35 MP.


This is more or less the maximum information content of the input source, i.e., the lens. In other words, there

isn't more information to be had from the source to begin with. Now when you capture this with a 35 MP sensor or

high resolution film or whatever, you will probably do justice to whatever high end lens gave you 100 cpmm to

begin with. Any more pixels and you are <i>oversampling</i>, i.e., going into a subjective realm similar to where

audiophiles sample 20 kHz audio at 1 Msamples/sec, and claim that they can somehow "hear" this. Maybe the brain

invents information where there is none - however it is outside the current abilities of optical or audio

engineering to make the brain see or hear what doesn't exist. But I digress.


So we are back to this magic figure of 35 MP as a fairly accurate upper bound of information content in a

24mmx36mm system.


This renders moot (or will as soon as digital sensors hit 35 MP) all discussion that film outresolves digital or

vice versa; after all, the lens isn't providing any more than that much information to begin with.


Now lets talk a little about film. I already said why I think a single grain has a continuous range of tones.

Don't confuse this with a the ability of a single grain to resolve information content smaller than the grain

size. This would require that the alternating black and white line pattern be reproduced by a single grain, and I

haven't seen any claim to that effect. In other words, a grain is a picture element (or pixel), with a continuous

range of tones possible. Grain size is variously stated as being 1-10 microns (micrometers, 10^-6 m). 200 cpmm

means that each individual line is 1mm/200 or 5 microns across. if you want this to be properly resolved, you

need grain size less than 5 microns.


Now if film grain were binary as Reichmann says, then you would need somewhere in the neighborhood of 48 grains

to give you the equivalent of 16-bit color depth; or you would need an area occupied by about 7x7 grains, or

about 35x35 microns. With a film "pixel" occupying 35 microns linearly, you could only resolve 1000/2x35 cpmm or

14 cpmm. This number is absurdly low for film, based on all sorts of empirical and experimental evidence.


Now before somebody comes back and says that there may be some way for grains to be stacked along the third

dimension (i.e., that of emulsion thickness) that can give the effect of having a smaller area for 48 grains

(i.e., a film pixel), please realize that this can't be possible with <i>binary</i> grain since a single "black"

grain will completely occlude all lighter grains, essentially resulting in a binary AND operation, i.e., creating

another binary grain, third dimension or not. Meaning that there have to be 48 grains in the x-y axes to form a

grain. Meaning that film must have 14 cpmm resolution, which is false.


For those familiar with how to prove things with logic, I just used reductio-ad-absurdum to prove Reichmann wrong.


I am quite sorry for being arcane, but this thread has raised some excellent points and counterpoints, and I

think that we are headed in the direction where the limiting factor is the lenses and format, as opposed to the

medium. This is excellent from my point of view, because these arguments about which medium has more resolution

will be finally laid to rest.


A few personal observations: All tests of film and digital I have seen so far (especially ones in with APS-C

digital outperforms medium format, these were popular a few years ago) - have one glaring shortcoming - they

<i>don't</i> have information content more than what the lens provides; and this is generally quite low, i.e.,

maybe in the neighborhood of 25-30 cpmm; that is relatively easily handled by both media, so the apparent

grainlessness of digital is taken as proof that it is superior. We somehow equate lack of grain to more

resolution, but this is quite untrue.


Second, using a scanner to compare film resolution to digital resolution is indicative of poor methodology. I

happen to

agree with Mauro Franic that examination under a loupe or microscope is the correct way to determine resolution

of film.


Finally, digital pixels have edges, so it is easy to create a completely sharp 200 cycles per mm pattern from

digital when the lines exactly line up with sensor pixels. It is also easy when they don't, since software can

detect straight lines and optimize them very well. It is when you have random curved lines that digital chokes

up, because you have to represent that curvature with square pixels. If the curvature is smaller than a pixel,

you lose that information.

Link to comment
Share on other sites

Good Post Vijay,


Just looking at this image will show that individual grains aren't binary (either totally black or clear)


I don't know how to embed so you'll need to follow the link.

The grain on the left is Black with a hole in it, so in effect transmits a small amount of light, in other words it's black(1)

and clear (0) at the same time which is not possible with binary states. The filamentary (Kodak term: Lattice) grains

increase in density depending on how many photons strike the individual grain and can be anything from 3 single atoms

right up to the entire grain, they are also stacked in the emulsion (sometimes 10 layers) with the larger grains ≈10µm

(faster) at the top and the smallest grains ≈ 0.2µm at the bottom of the stack.

The denser the lattice the less light that passes, the inverse also being true all states between the minimum (one

photon) 3 Ag atoms right up to the whole grain are possible- in other words they are analogue.

Reichmanns assertions are way out even if he is confusing grains with Ag atoms.




Your photo's although interesting seem to be taken at different times, do another side by side test with cameras on

tripods in the same lighting conditions, try also less linear subjects like trees.

In my side by side tests I have found that slower (25-200) film has a higher resolution and resolves more detail at the

expense of some noise in the mid tones- a trade off.


Link to comment
Share on other sites



I have to disagree with you on camera/scanner vibration. I see no evidence for it whatsoever.


The film shot was taken with a Canon EOS-3, with mirror-lock up (though really that's not even needed on a wide-angle shot such as this one taken at 17mm), on a Gitzo 2-series tripod. The wavy-ness you see in the streaks is because that section of I-5, which I drive across quite often, is notoriously bumpy. Those are simply the headlights of cars bumping up & down. There is no such evidence of this wavy-ness *anywhere* else in the scan. If you want, I can send you the full-size image.


The digital shot doesn't show any wavy-ness; I assume you're referring to the jagged-ness of the headlight streaks indicating vibration? Those are caused by interpolation patterns from Bayer pattern sensors... I see that on any diagonal lines even on shots taken at high shutter speeds in broad daylight.


The REAL reason my comparison is partially meaningless is because they weren't shot with the same lens, not shot at the same time of day, not shot at the same zoom, etc. When I get a chance, I'd like to do what Mark suggests: a side-by-side comparison of more complex & interesting subjects.


If you still find any other evidence that supports your claim of 'vibration' or 'shake', please do let me know.




Link to comment
Share on other sites

Actually, let me clarify:


In the digital image, the stair-stepping I'm referring to is in the highlights of the streaks... these are Bayer interpolation problems.


Looking more closely at the non-highlights, I do see the wavy-ness. Again, this is due to the road surface of I-5... those are cars bumping up and down.


Why is the wavy-ness not as 'sharp' & 'resolved' in the 10MP digital image? Because <b>the film resolved it better</b> :) Which goes along with my argument that the film does seem to have a higher resolving power than this 10MP digital image (barring lens differences, yes, I know).


This is not camera vibration because if it were, it would be evident elsewhere in the image, which it is not. Plus, any vibration of the camera would have had to have been continuous, because these are ~30 second exposures, and a short-lived vibration would not have led to those very well defined and resolved and <b>COHERENT</b> wave patterns! It would have led to random wavy-ness at different times, which would not have formed those coherent sine-like waves.


So, I think, by logic, I've proved the 'shake' or 'vibration' hypothesis wrong. Vijay, please let me know if you are convinced of my argument.



Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

  • Create New...