Film and digital resolution compared

[Photo_Smith]

Or just choose a medium format or 4x5. How big do you print? What is the 'advantage' worth in real terms?
You obviously only need a drum scanner if you show your work digitally. What happens if I need a print? Does my inkjet or paper surface increase or decrease resolution?

Does this matter outside measurebation

 

[goamules]

Nice work in the analysis. Of course, you only touched on large format. To me, film has fantastic resolution, as do sensors. If you don't enlarge either, great! But we enlarge both many, many times from where they started. Not with LF contact prints or direct postitives. If I want the sharpest print, say, at 8x10, I use an 8x10 negative and contact print.

 

[Roger Hicks]

. . . ● @ Roger Hicks. I feel you are being too dismissive, despite using the phrase "more rigorous". I have been careful to base my initial calculations on sound theory and measurements from reliable sources (tests by Kodak and DPreview using the standard resolution charts). The later calculations do use estimates, but are most certainly not "wishful thinking", and although not precise are reasonable "ball park" values. My working is clear, so if you disagree with these estimates, feel free to substitute your own values. . . .

Sorry. That was the exact opposite of my intention, and I certainly did not wish to accuse you of wishful thinking. I just thought it rather interesting that your carefully stated assumptions and calculations came out at about the middle of others' calculations (many of which are, as you correctly say, much less rigorous than yours) and of empirical tests -- which rather suggests to me that (a) you and they are about right and (b) the advantages of further testing are limited. I am grateful that you confirmed what I had long suspected was the case, but as I say, it's all pretty much converged on the sort of figures you give, give or take a factor of two. EDIT: Well +/-50%, more likely.

Cheers,

R.

 

[bohdan]

FWIW, here's a link to Norman Koren's thoughts along the same lines from back when the Canon 5D was state of the art.

http://www.normankoren.com/Tutorials/MTF7.html

For myself, while my film scanner captures 4000dpi, I'm under no delusion that 4000 dpi are actually there.

So, we have to be specific here.

You can scan all day at 4000 dpi and in the scanned image you will get the appropriate number of dpi^2 in your image from the scan. There will be the 4000dpi there guaranteed.

Now, would this be an interpolated value because the native resolution of the scanner is lower? That's a different story.

Would there be a one to one correlation of scanned pixels to the grain? That's another story.

--bohdan

 

[bohdan]

What none of these web comparisons addresses is the need to consider bottlenecks in equipment that impact the theoretical resolution of film – i.e. lens resolution and scanner resolution (assuming images will be scanned). Consider a typical scan of 35 mm film at a scanner resolution of 2700 ppi: the resulting file has dimensions of about 3800 × 2400 pixels = 9 MP – but this is not equivalent to a 9 MP image from a digital camera because both are theoretical values based solely on the number of pixels, and fail to account for differences in resolving power between the two mediums. Comparing, say, a 20 MP film scan with a 20 MP digital camera image is thus comparing apples with oranges. What we actually need to compare is the resolving power of frame of film and a digital sensor.

Just comparing two different 20MP digital cameras is a flawed concept in and of itself. You have a lot of factors there...lens, size of photoreceptors, pixel density, method of color interpretation (Bayer, etc) as well as others I am too lazy to look up. Either way, you really need to stick with lines per measurement of the sensor in order to be correct here.

In your case, we are comparing 20MP scanner vs a 20MP camera image. That's complicating matters in that a certain scanner will interpret the scene differently than the camera. You are greatly

In order to do an experiment correctly:

+ Scene must be photographed the exact same way.
+ Sensor must interpret light the same way.
+ Film scanner must interpret that film the same was as the digital camera interprets the light from the scene.
+ Film must be developed to get the most perfect result as close to the theoretical limit of the film as possible
+ Same lens with no post processing by the internals

Food for thought....how can you reduce the variability?

 

[bohdan]

...<SNIP> Also you seem to forget (or be unaware) that the megapixel count of a camera is not a measure of it's resolution, that would be it's pixel density and size.
Further to that a 12mp sensor can't resolve 12mp of detail with a Bayer sensor at best you get 6mp green, 3mp Red and 3mp Blue the output being mathematically calculated from a grid sample.

Film doesn't work like that it has RGB records but has them stacked and randomised, no digital camera has that (foveon is stacked but gridded).

Awesome...and further, how does this compare to how the scanner sees the film and how the film records the scene?

Starting from a digital image is one thing, but starting from an analog image of the scene with a bunch of quantization errors and non-linearity thrown in is another.

 

[RichC]

I think most of your logic is flawed. Some glaring mistakes you might want to look into is the role of diffraction, and calculating the differences in diffraction over different formats and even different sensors (pixel size) and how they change data.
For instance a 6x7 film with standard lens won't become diffraction limited until after ƒ22 given a 12" wide print.

I disagree... This is only true if aiming to make prints of the same size from all formats. As I've stated, I'm not interested in print size here, just the amount of information recorded on a negative or by a sensor.

Also you seem to forget (or be unaware) that the megapixel count of a camera is not a measure of it's resolution, that would be it's pixel density and size.

Which is precisely what I said. If you go to, say, Nikon's website, they talk about "a resolution of 36 MP" for the Nikon D800 - so, for better or worse, "resolution" is now used to mean "number of pixels". To reiterate, I am not using it in this sense.

Further to that a 12mp sensor can't resolve 12mp of detail with a Bayer sensor.

Of course it can't - I state that too! This is precisely why I use measured resolving powers (in lp/mm) from actual real-world tests. These values tell us how exactly how much detail is resolved (assuming the test results are trustworthy).

I don't know where you get the figure of 80% degradation with drum scans, or the idea that film needs to be scanned.

A scanner is an optical system with a lens, sensor, etc., and as another link in the imaging system it will inevitably lead to loss of data. How did I get 80%? The resolving power of a drum scanner is about 90 lp/mm (measured - Google for confirmation). To work out the loss in resolving power, the resolutions of film and the scanner are combined. The following formula is empirical (i.e. it closely matches experimental results but is not based on theory):

1/R = 1/√[(1/f)² + (1/s)²]

where R is the resolution of the scan (not the scanner), f is that of film (70 lp/mm) and s is that of the scanner (90 lp/mm). Plugging in these values gives

1/R = 1/0.0173
R ~ 55 lp/mm

So, scan resolution = 55/70 = 80% that of film.

I don't know where you get the idea that film needs to be scanned.

Of course you don't have to scan colour film - if you don't, ignore that step. Most people do scan today, and I certainly do. In fact, I can't imagine where I can get large (> 20 inch) colour darkroom prints today, and dread to think of the price!

If we output to print what effect does mathematical dithering have on inkjets?

Which is why I barely touched on printing. As the post stated, I primarily wanted to know how much detail film captures compared with digital - ending up with a digital image file, not a print. I'm not dismissing printing - it's important to me, and I always print my photographs, aiming to get prints on gallery walls. But printing is irrelevant to my calculations .

So many basic errors, mis-calculations and false assumptions in your OP that really I don't know where to start.

By reading my post more carefully...? 😉

 

[CameraQuest]

I moved this thread into optical theory and placed a sticky on it because I think this issue and discussion is important.

To all the naysaysers which disagree, fine. This is a discussion forum.

However, taking partial potshots of his methods without offering a better overall view is not really all useful or helpful. Neither does it make your own arguments convincing since they are not assembled in a complete form as the OP did.

So, for those of you who think you can do a better job on this subject, let's see you post something better covering all the points which Rich does (and more) that works better !

Stephen

 

[Photo_Smith]

I disagree... This is only true if aiming to make prints of the same size from all formats. As I've stated, I'm not interested in print size here, just the amount of information recorded on a negative or by a sensor.

If you're not interested in printing then fine but that still doesn't change the fact that you can't approximate the diffraction of 4x5 by adding a 3x multiplier because the focal length is 3x.
Diffraction is one of the min causes of degradation of the image in other words the spread function of the light.
It is dependant on the size of your output though so I'm not sure how you are going to duck out of that one.
Simple fact diffraction will act like a blurring filter over your sensor/film and will greatly lower your 'theoretical' lp/mm figure.

Which is precisely what I said. If you go to, say, Nikon's website, they talk about "a resolution of 36 MP" for the Nikon D800 - so, for better or worse, "resolution" is now used to mean "number of pixels". To reiterate, I am not using it in this sense.

If we are talking science then what Nikon say in their literature is not worthy of discussion.

Of course it can't - I state that too! This is precisely why I use measured resolving powers (in lp/mm) from actual real-world tests. These values tell us how exactly how much detail is resolved (assuming the test results are trustworthy).

Sure, I'm more pointing out when you say the D800E has xx resolution it doesn't-just clarification.

A scanner is an optical system with a lens, sensor, etc., and as another link in the imaging system it will inevitably lead to loss of data. How did I get 80%? The resolving power of a drum scanner is about 90 lp/mm (measured - Google for confirmation). To work out the loss in resolving power, the resolutions of film and the scanner are combined. The following formula is empirical (i.e. it closely matches experimental results but is not based on theory):

1/R = 1/√[(1/f)² + (1/s)²]

where R is the resolution of the scan (not the scanner), f is that of film (70 lp/mm) and s is that of the scanner (90 lp/mm). Plugging in these values gives

1/R = 1/0.0173
R ~ 55 lp/mm

So, scan resolution = 55/70 = 80% that of film.

You might like to look at your figures again if you think a drum scanner tops out at 80 lpm I think and I'll have to find out it depends on the aperture of the scan it can resolve grain and further to that I'm not sure of your maths either.

Of course you don't have to scan colour film - if you don't, ignore that step. Most people do scan today, and I certainly do. In fact, I can't imagine where I can get large (> 20 inch) colour darkroom prints today, and dread to think of the price!

When arguing the ultimate difference in resolution of two systems you need to think about those systems as a whole, price shouldn't come into it!
Since we are talking about resolution I presumed (falsely) that that resolution would have a purpose-large display.

Which is why I barely touched on printing. As the post stated, I primarily wanted to know how much detail film captures compared with digital - ending up with a digital image file, not a print. I'm not dismissing printing - it's important to me, and I always print my photographs, aiming to get prints on gallery walls. But printing is irrelevant to my calculations .

By reading my post more carefully...? 😉

I read your post, possibly I didn't understand it certainly I don't understand why you would need resolution if it's not for large display?

 

[RichC]

If you're not interested in printing then fine but that still doesn't change the fact that you can't approximate the diffraction of 4x5 by adding a 3x multiplier ...

I read your post ... I don't understand why you would need resolution if it's not for large display?

I'll look into diffraction in lenses more carefully.

My original post arose from a recent project. As I've mentioned, I shoot with a Nikon D800E and a Mamiya 645. I regularly make 30-36 inch prints with the Nikon, and that's about the maximum without the print looking slightly soft seen very close up. The Mamiya is new to me - more convenient and cheaper to shoot than my 4x5 camera - and I wanted to compare the resolution of 645 with the 36 MP Nikon, to have some idea how large I can print 645 from scans. So, I sort of had a print size in mind, but it was more useful for me to directly compare cameras - the 645 with my Nikon D800E - in terms of megapixels than find out the maximum "sharp" print size of 645.

I couldn't find an answer on the web - you might find my approach flawed but some of those on the web are beyond hope! Hence my attempt at more rigour.

Film is fairly new to me - I'm used to digital cameras and megapixels; my first "proper" camera was digital, not film. (Hey, at least I'm using film now!) So, to get my head round film, I need to think in terms of digital - for example, I have a very good idea of how large I can print a file from a digital camera if I know its sensor size and "resolution" in megapixels, but am still a bit lost with maximum film print sizes. But not now I've calculated the megapixel equivalents of film formats, assuming my values are about right.

My calculation that 645 colour film and a 36 MP full-frame sensor have similar real-world resolving power has since been born out in practice: my test scans on an Imacon Flextight and Nikon files have similar detail seen on-screen. The Nikon images are a little sharper with a bit more detail, but the Imacon isn't as good as a drum scanner, and film also has grain, which makes the Nikon images look cleaner, but not necessarily better - if I'm printing digital camera images large, I always add a little noise, otherwise I feel the prints look a bit "clinical". Ironically, I use a Photoshop plugin that replicates film grain to add this noise, which is barely visible in the print but still has a tangible effect on its perception (at least to my eye)!

 

[Sejanus.Aelianus]

It may appear tangential to this discussion but an immense amount of research into the techniques of microfilm was carried out in the 'thirties, 'forties and 'fifties. The people working in that area were very interested in resolution, for obvious reasons, and did a lot of work on the variables that affected real world micro-recording.

It might be worth getting one or two of the technical books on the subject. I used to have one, dating from the late 'forties, that contained a lot of information on diffraction, high contrast development and similar subjects, which you may find useful in this context.

Unfortunately, I've mislaid my copy and can't even remember the correct title. 🙁

 

[thegman]

Not sure if I understand this post correctly...

There is a theoretical advantage in resolution of ISO100 35mm film over a 21MP full frame digital sensor. This will turn into a real world advantage only if the film camera is perfectly calibrated, shot from a tripod with a good lens stopped down, only with access to a high end drum scanner in post processing.

Wouldn´t it be easier to just chose a D800E or Sony A7R if your photography requires max resolution? (Not even talking about medium format digital...)

By the way, what happens at ISO 200? 🙂

I think that's a fair summary of the thread.

However (!)

The thing with film is that you can increase it's size without any great increase in cost, when it comes to cameras. My Leica M3 (sold now), Rolleiflex, and Fotoman 45SPS all cost roughly the same. Cost per shot of course changes, but film is not bound by cost of camera in the way digital is.

So I can shoot 4x5 on a (brand new) camera which costs less than new FF digital, let alone medium format digital.

If you want resolution at any cost, by all means use a Phase One IQ180 (or better it with 8x10 film). But most of us are bound by cost to a point, and for me, medium format film makes a lot more sense than a pricey digital equivalent.

 

[Photo_Smith]

I have most of the books and papers and can publish some info. The studies you mention are the ones by Jones and Zwick where they used microfilm to resolve text down to 2µm high
This is from the paper–

I don't want to publish too much...

If the OP wants know the resolution each system is capable of he'll need to work out the losses from each part of the image chain.
If he does factor in the camera/lens/format/film type/scanner/output/viewing distance etc he'll get different results.

The only way to measure the capabilities of film itself is to count line pairs by microscope, which would be more valid for slide film that could be looked at under a loupe, but then that ignores the viewing medium projector/paper/monitor.

Taking others figures then extrapolating results without verification is fuzzy at best and not going to allow you to arrive at anything more than an VERY rough estimate.

It isn;t possible to get an absolute from a mess of estimated data.

 

[Photo_Smith]

I'll look into diffraction in lenses more carefully.

My original post arose from a recent project. As I've mentioned, I shoot with a Nikon D800E and a Mamiya 645. I regularly make 30-36 inch prints with the Nikon, and that's about the maximum without the print looking slightly soft seen very close up. The Mamiya is new to me - more convenient and cheaper to shoot than my 4x5 camera - and I wanted to compare the resolution of 645 with the 36 MP Nikon, to have some idea how large I can print 645 from scans. So, I sort of had a print size in mind, but it was more useful for me to directly compare cameras - the 645 with my Nikon D800E - in terms of megapixels than find out the maximum "sharp" print size of 645.

I couldn't find an answer on the web - you might find my approach flawed but some of those on the web are beyond hope! Hence my attempt at more rigour.

Film is fairly new to me - I'm used to digital cameras and megapixels; my first "proper" camera was digital, not film. (Hey, at least I'm using film now!) So, to get my head round film, I need to think in terms of digital - for example, I have a very good idea of how large I can print a file from a digital camera if I know its sensor size and "resolution" in megapixels, but am still a bit lost with maximum film print sizes. But not now I've calculated the megapixel equivalents of film formats, assuming my values are about right.

My calculation that 645 colour film and a 36 MP full-frame sensor have similar real-world resolving power has since been born out in practice: my test scans on an Imacon Flextight and Nikon files have similar detail seen on-screen. The Nikon images are a little sharper with a bit more detail, but the Imacon isn't as good as a drum scanner, and film also has grain, which makes the Nikon images look cleaner, but not necessarily better - if I'm printing digital camera images large, I always add a little noise, otherwise I feel the prints look a bit "clinical". Ironically, I use a Photoshop plugin that replicates film grain to add this noise, which is barely visible in the print but still has a tangible effect on its perception (at least to my eye)!

Ok this would have made a much better OP because it lays out your requirements with real world issues and things you've personally noted.

The resolution of any system is a chain, of which no part can be ignored.

The issue has no real definitive answer, certainly not on the internet, what you have noticed though is the differences in character between film and digital film looks sharper even with more grain, digital smooths and has better edge detail in some cases.
Viewing distance matters too, most people don't have 20/20 vision (text on a eye chart 20 feet away 480 lux lighting) Kodak have their nice PGI with a figure of 25 being no detectable grain.

The resolving power of any given film varies with contrast and colour, you may not be aware that magenta subjects will have less resolution than blue and the red/cyan layer being at the bottom is less able to record fine detail due to light spreading as it travels down.
The magenta layer has the most noise and the red the least acutance so that make our estimate of resolution somewhat subject dependent.
It's similar with digital where it has more green pixels.

I pick a print size, I top out at 20x16 so 6x7 with 100 ISO film like Ektar is fine, but Ektar isn't the sharpest 100 ISO that would probably be Velvia.

If film is new to you try some 100 ISO Fuji E6 and some Ektar scan them and see if you like it at the print size you require, move up a format to 6x7 if you need (I think 645 is too small for 30" prints)

The Imacon is a nice scanner and probably good enough for most but it isn't s drum scanner whatever Hasselblad tell you.

If not use digital and add in noise to your requirements...

 

[RichC]

Taking others figures then extrapolating results without verification is fuzzy at best and not going to allow you to arrive at anything more than an VERY rough estimate.

It isn;t possible to get an absolute from a mess of estimated data.

No argument from me!

The point was not to get precise values but ballpark figures of the correct order of magnitude. Simply to be able to say that a 645 camera with low-grain colour negative film approximates an 35-45 MP digital camera, for example, or that a 20 MP dSLR gives similar results to a 35 mm camera loaded with the aforementioned film.

My results are a starting point, not an end point. They give me an idea of which film format I need to use if I want to make 40 inch prints, say (645 OK if viewer is at a sensible viewing distance, but 6x7 will be perfect if the prints need to be sharp close up - if "massaged" in Photoshop (adjust contrast, sharpness, etc., to taste).

The Imacon is a nice scanner and probably good enough for most but it isn't s drum scanner whatever Hasselblad tell you...

No. But it's what the university has, and drum scans aren't cheap or always necessary - there's no point aiming for high resolution if it's not needed.

I prefer the look of film - Portra has subtle colour that I cannot replicate with a digital camera, with more delicate tones, and exposure is more forgiving. But digital is faster, and cheaper (once you've paid for the camera!), and has the appearance of being sharper. I think large film prints look better than large digital prints, but the smaller the print, the more similar. I'll continue to use both formats, depending on the needs of each photography project.

 

[noisycheese]

While we're at it, no one has yet conclusively answered Descartes' equally worthy query:

How many angels can dance on the head of a pin?

🙄

 

[sector99]

You may wish to consult the recently March 2014 updated film vs digital scientific analysis by an imminent imaging/computer expert at this link:

Last Updated: Mon Mar 3 17:53:33 UTC 2014
​

Some Observations on Digital vs Wet Film Photography

Dr Carlo Kopp, MIEEE, SMAIAA, PEng
Text, Images © 2010, Carlo Kopp

​

http://www.csse.monash.edu.au/~carlo/bayer-patterns.html

He lists EKTAR 100 at 154 lp/mm (High contrast).

He also shows normalized MTF curves for various films and discusses the results showing photographic comparisons to illustrate Bayer mosaic/scanner limitations.

His conclusion is that wet film resolution is in fact comparable to the highest digital sensor results—especially when medium format color is selected.

Oil immersion Wet Drum Scanning

A key omission from the extant discussions (Including the above example) is the positive effects of wet drum scanning in reducing scattered light from the negative/transparency original. Often the illuminating scanner's scatter degrades the final file by adding what appears to be veiling, white noise. This scatter is eliminated by fluid immersion strategies.

Though expensive, the wet drum scan process looks perhaps to be the best candidate for those hoping to extract all that their film can offer—especially with EKTAR.

Ideally two camera bodies, one with advanced transparency and the other with advanced color negative film (EKTAR) might be the very best strategy to capture varying dynamic range scenes.

A Photograph Isn't a File...

Finally, a photograph (IMO) isn't a file. It's rightfully an object, optimally with a gelatin media and archival substrate. It's features offer album storage and convenient display that eliminate electricity as a viewing requirement.

A true photograph can properly be with you at all times and its parent negative/transparency. Here, I discount inkjet printers as a poor facsimile to the real thing.

 

[antonynorthcutt]

Now this is an interesting post, some really food for thought here, thanks for taking the time to write and share with us. Much appreciated.

 

[Daniel Teoli Jr.]

OP...I dodn't know where you got your info.

35mm film is = to about 4 mp with a cheap P&S cam. And that is being generous to film.

http://photographycompared.tumblr.com/

 

[charjohncarter]

Impressive, That was a great essay. One thing that came to my mind was how the final print is made: digital screen, digital printer, analogue printer, or my favorite at this time digital file to laser exposure of photographic paper. A properly or even not so properly done Silver Gelatin print is wonderful, but for me I don't have a darkroom any more so I send my files to Costco and they print on photographic as does Mpix on true B&W paper. These two methods analogue prints and laser printing on photographic paper give the best results in my opinion. I know that digital printers have made great advances but still to me don't look as good as either of the above. So, even though the argument of digital vs film is still alive, my final print is what I look at and how I judge everything that came before it.