Resolution - Digital versus 35mm Film

[FoodLover]

Advertisement (gone after registration)

General question; there's no specific camera involved.

 

Can anyone tell me how many pixels must there be in a digital image so its resolution is equal to that of 35mm film?

 

Thanks!!

[TimF]

There's probably as many answers to that as there are photographers!

 

PS. Why the massive text size?

[Advertisement]

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

[lct]

Depends on the film. From 10 to 18 MP for Velvia 50, or from 4 to 8 MP for Sensia 200 according to Clarkvision: Film versus Digital

[lct]

...Why the massive text size?

I don't know.

[stunsworth]

Compared to scanned film we're there already, and have been for a while IMHO. So I'd say 6mp upwards.

 

With the usual proviso that there's more to photography than resolution. There was a program about street photography on UK TV last night. Some of the images were pretty ropey from a definition point of view, but were outstanding images none the less.

[batmobile]

prob about 8MP for fine grained 35mm trannie film. But film looks totally different, so take you pick based on aesthetic wishes! I prefer film personally. Digital is too clean!

[TimF]

Advertisement (gone after registration)

prob about 8MP for fine grained 35mm trannie film. But film looks totally different, so take you pick based on aesthetic wishes! I prefer film personally. Digital is too clean!

Except when it's noisy, when it can often be intensely ugly compared to film grain.

[Guest joewehry]

Remember that pixels don't tell the whole story. You can fit 10 MP on a small sensor or a large sensor. The images will be different.

[SJP]

If you believe the CoC concept and the Nyquist theorem then you need about 2 pixels per CoC. On the M8 the CoC is about 21 micron and the pixel size is about 7 micron (ie 3x smaller). So to have "the same" resolution as film you would be able to live with a sensor of (2/3)^2x10.5Mpix => about 5 Mpixels.

 

But if you want a film like grain then 2 Mpixels should be enough:D

 

EDIT: the same no. of pixels (> 5 Mpix) should be appropriate for a FF sensor as well.

Edited March 15, 2009 by SJP

[adan]

The deal with film, especially 35mm film, is that it has a certain native resolution. But there is no way to access that resolution directly. You ALWAYS have to use a secondary optical device (enlarger lens, scanner lens, projector lens, loupe, microscope, etc.) in order to see what it has resolved.

 

And since no optical device is "perfect" you will always lose a bit.

 

Direct digital capture does not require a secondary lens to view the image at a size where the full native resolution can be seen.

 

(Technically, I suppose one could say that the human eye is also an optical device in the chain, which will add its own errors)

 

Fuji lists Velvia resolution as 125 lpm under ideal conditions, and 80 under average conditions (80 is what that Clarkvision site uses for Velvia)

 

To equal 125 lpm with no aliasing (Nyquist frequency), you need at least 250 pixels per mm - or 4-micron pixels. That will get you equal resolution to film so long as the sensor and the film are the same size. (Micron = 1/1000 mm)

 

Tiny P&S digicams have pixels in the 2-micron range, so AREA FOR AREA, they resolve twice as much as film. But we are talking an image area about like an 8mm movie frame.

 

For a 24 x 36mm image area and 4-micron pixels, you get 6000 x 9000 pixels, or 54 Mpixels, as the theoretical or mathematical sensor needed to always do as well or better than Velvia under ideal conditions.

 

Under less than ideal conditions (i.e. 99% of the time), Velvia reverts to 80 lpm, or 160 pixels per mm, or pixels 6.25 microns each dimension. On a 24 x 36 sensor that comes out 3840 x 5760, or 22.1 Mpixels.

 

However, bear in mind that those numbers are necessary to totally avoid moire patterns and other digital artifacts by oversampling by a factor of two (160 pixels per mm to capture 80 lpm cleanly with never a risk of moire).

 

In MOST pictures, that oversampling is overkill, and an oversample of 1.4x in each dimension will be enough to cleanly reproduce an image from a very good (80 lpmm resolution) lens. So, 80 lpmm x 1.4 = 112 pixels per mm = 9-micron pixels = 2688 x 4032 image (24x36 sensor) = 10.8 Mpixels.

 

Go back and pick up the transmission losses from secondary optical systems (enlargement) that I mentioned to start with - plus the fact that most sensors are natively about 160-200 ISO and thus should be compared to films of that speed - plus the fact that most 35mm images are NOT full 24 x 36 but often 23.5 x 35.5, cropped further by slide mounts, negative holders, etc.....

 

...and you see why, in the huge majority of cases, any double-digit sensor will usually do at least as well, and usually better, than a 35mm film image at equivalent ISOs - in terms of resolution.

 

But obviously, the range would be 54 Mpixels to 10 Mpixels (or less) - depending on what assumptions you make about the conditions, enlargement, etc. FOR A 35mm-sized original capture area!! Move up to medium-format film or sensors, and the numbers go up as the square of the dimensions.

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

 

Put another way - a chip of Velvia the size of an M8 sensor would resolve 80 lpmm most of the time outside the lab. Over the image area of 27mm x 18mm, we get total resolution of 2160 lines x 1440 lines. To perfectly register that resolution at the Nyquist frequency (2x oversampling) we'd need 4320 x 2880 pixels or 12.4 Mpixels. If we go with 1.4x oversampling (with risk of occasional moire), we need 3024 x 2016, or about 6.1 Mpixels. The M8's 10.1 Mpixels is in the high end of that range - and one can see why, since most users do NOT shoot films with the resolution or ISO of Velvia, 6 Mpixels was the tipping point (Nikon D100, D70, Canon 10D) that pushed digital out in front.

Edited March 15, 2009 by adan

[lars_bergquist]

Very good, Adan. For decades I used medium format in parallel with and mostly in preference to 35mm, because that gave me more grain than I liked. I went as far as doing long range backpacking up above the Arctic Circle with a Mamiya 6 and two lenses. Finally I used some acane tricks and got a visually grainless 18x24cm print out of a 35mm T-Max 100 neg, but that was not easy! (Note that by 'grainless' I do not mean just that I found the grain acceptable, which is the common usage. I mean that I did not see it.)

 

My own impression is that the M8 gives me definition (resolution + useable contrast) roughly comparable to what I could get out of a 6x9cm rollfilm negative on medium speed BW film. And I have looked at many such negs projected through a 90mm El-Nikkor. Like most people I do not pine for a FF M because I want (let alone need) more pixels---it's mainly because I want my wide angle lenses back.

 

The old man from the Age of Plus-X

[k_g_wolf ✝]

>> Lars

 

That´s a beautiful post No. 1,500.

 

Congratulations and thanks for many contributions on this site,

regularly raising the quality here.

 

 

Best

[andit]

Hi Guys,

 

Forgive me for turning this thread a bit on it's head. I was always told and been under the impression that with current sensor technology you can not equal film resolution. This is also why Hollywood studios still use 16/35mm film for making movies. WHY?

 

From what I have been led to believe, this is due to the deep well construction involved in making sensors (a sensor is actually a matrix with x amount of rows and x amount of columns. There are literally walls between adjacent pixels - thus the problem in making a full frame M8. These walls cast shadows onto the pixel which affects the light gathering ability - mainly at the edge of the sensor where light enters at extreme angles.

 

Film does not have rows and columns, thus does not have these walls between elements and can capture light at virtually any angle. This makes film a far superior capturing method to digital - that is with current technology.

 

There are many very technical people who work with these things in their daily lives here on the forum. As I said, this is what I have been led to believe up to now - this may be the biggest amount of rubbish and I would love to hear from these experts as to how much of this theory is correct.

 

Andreas

 

AFTER THOUGHT: From what I recall hearing sometime ago, Fuji has started to address this deep well issue by developing sensors with octagonal pixels. This reduces the shadows and increases the light gathering ability of the sensor.

Edited March 16, 2009 by andit

[jaapv]

I think it is a bit the other way around, Andreas. Although there are sums to be made for objective resolution figures for both film and sensors the end result is the impression on the viewer. With sensors the sharp transition between the pixels makes for high edge microcontrast, giving an impression of higher resolution, with film there will be diffusion and refraction within the emulsion, lowering this effect, giving the impression that it resolves less. On the other hand, film grain will give an impression of higher resolution. The net result is that the two can only be compared subjectively, not in numbers.

[pgk]

Last October I carried out a project using digital cameras (2 x 5Ds, a D200 and a D50) in order to see whether it was possible to identify small fishes (gobies) in situ in Loch Fyne, Scotland. The result was that it was very possible to identify the gobies. However, what is pertinent to this thread is that whilst some of the nuances (sensory papillae on the face and head of one goby species) can be be seen in the 5D images (12.8 MPixels), they cannot be made on Velvia 50 transparencies taken of the same species (identified from other features) in the same place several years earlier.

 

Although this was a qualitative rather than quantitative test, it has convinced me that 12.8 MPixels exceeds 35mm transparency in terms of usable information yield, however the capacity to store usable information is not necessarily the same as resolution and IMHO trying to compare film and digital numerically is not a particlarly useful exercise as to do so caveats must be applied.

 

My own opinion, for what it is worth, based on my own experience is that 10MPixels equates to 35mm film in many situations. Beyond this there will be some applications which will benefit from film and others which will benefit from digital. at what point digital become the clearly better system (assuming a sensor the same size as 35mm film) is a tricky one - the latest 20 MPixel + sensors are very lens sensitive for example. I am using a 5D2 now and so far my two 'best' lenses on this are the Canon 100 macro and the Leica 80/1.4R. Using these two the 5D2 is clearly producing results beyond the capabilities of 35mm film IMO.

[tgray]

Sounds like most people agree here. For all intents and purposes, 8-10 mp usually beats film for resolution. 6mp looks pretty good. Even if they technically give up a hair of resolution on a given system, film's going to have more grain and digital will be smoother, so it will probably take enlarging better. For a nice comparison between Ektar 100, Velvia, and the Sony 25 mp camera, look here. Just to give you and idea of what those numbers are actually capable of.

 

That being said, resolution is not everything. A lot of of us here already shoot things like TMZ and Delta 3200. Heck, actually anyone shooting Tri-X, or FP4+, or anything other than TMAX 100 or finer grained films has already decided to trade resolution for other qualities, so resolution isn't everything. But if that and grain free images are what you're looking for, digital or larger formats is the quickest way to get there.

[adan]

Andit: In fact, if you've ever seen stills from motion pictures, the resolution (and graininess) is actually pretty mediocre, except for some of the really big formats like cinerama/vistavision/panavision. Resolution is not a high priority in cinema because any individual image is only visible for 1/24 or a second and because shutter speeds are often not much faster.

 

Which is why studios hire "stills" photographers to shoot the promotional pictures, rather than just grab frames from the movies themselves.

 

(Same for video - 1080p HDTV is only 2 Mpixel images, but still looks awesome even blown up to wall size where a 2Mpixel still photograph would crash completely - because it is the motion, sound, color and drama that carry the experience - not the resolution.)

 

On the flip side, there is a huge installed technology for using film in movies, a huge "knowledge base" of HOW to use film, whether lighting, exposing or editing (tens of thousands of cinematographers, lighting technicians, camera operators, directors) - and other factors beyond resolution (as tgray says) that still benefit film: dynamic range, flexibility, "the look", etc.

 

Not to mention the huge installed base of film-projection theaters.

 

A fast conversion to digital movie-making would be incredibly expensive and disruptive, without a whole lot of immediate pay-back in terms of "better" product. It will happen slowly, as digital-savvy professionals replace the retiring film pros.

 

None of the above applies to animation or special effects techniques, which went happily digital over a decade ago without a second thought (even the claymation guys). Who really wants to spend three weeks shooting a 10-second stop-action sequence (and then another three weeks reshooting because something needs to change) when in the same 3-6 weeks they can do 50 variations and renderings out of a computer?

[tgray]

Another fact in the digital/film issue in motion picture is that that it's one thing to make a 35mm sensor that works at 1,5,10 fps, it's another when one needs to work at 24 fps, 60 fps, or even 120-150 fps. For example, I thought I remember reading about Apocalypto being filmed on a digital cinema camera, but it ran at a max rate of 50 fps, which isn't fast enough for convincing cinematic slow motion shots. As a result, some scenes were filmed at 60 fps, and the actors had to act in slow motion at the same time. I think other scenes were shot on film for the same reason.

 

Not that this is an insurmountable problem. Yet another reason film is still widely used in cinema is as adan says, it's actually cheaper right now to shoot on film. When your budget is $150 million, what's $200k in film costs? The infrastructure is there for film for both production and distribution, whereas a full digital set has other (expensive) requirements. There usually needs to be a compelling reason to go digital. FX heavy films like the new Star Wars or the new Superman are good candidates since it's all going to be digitized anyway and heavily worked over, but even that isn't always a good reason for digital origination.

 

This has all the workings of a troll. Apologies to the original poster if he actually plans on sticking around.

[georg]

Cinematography is a different beast, but this document explains many things that also count for still photography:

 

http://www.arri.de/fileadmin/media/arri.com/downloads/Camera/Tutorials/SystemsTechnologyBrochure.pdf

 

It's worth reading and also states about 80lp/mm max. real resolution.

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[jaapv]

This has all the workings of a troll. Apologies to the original poster if he actually plans on sticking around.

 

Well, if this is a troll, he must be sorely disappointed as this is developing into quite a harmonious thread, instead of the bloody film-digital fight he was hoping for....