35mm Film vs DSLR: Gradation, Resolution, and Dynamic Range

Rik Littlefield
Draft 1.0, 5/28/2007
Minor edit, 2/11/2009


This article compares images obtained from the electronic sensor of a digital SLR (Canon 300D, 6.3M pixels in 22.7x15.1 mm) with those from 35mm film (Kodak Gold 200, full frame 36x24 mm).

Overall, this 6.3M pixels DSLR is approximately equal to the tested film, when imaging the same subject area.  The DSLR has less resolution than film for high contrast features of the subject, but it captures more low-contrast features.

In more detail:
Note that the DSLR used for this test was an inexpensive prosumer unit purchased in early 2004.  Presumably its performance is at or below the bottom end for current models.


In April 2004, after 40 years of shooting 35mm film, I suddenly switched almost completely to digital SLR. 

What prompted me to consider a DSLR in the first place was a side-by-side comparison of 8"x10" prints from film and DSLR, shown to me by a colleague.  The subject was boring -- some buildings in a desert -- but the comparison grabbed my attention.  At first glance, the images seemed very similar.  On closer study, I realized that I could see more high contrast detail in the film print, but more low contrast detail in the DSLR print. That made no sense to me at first, but after some thought, I realized it was an easily explained although unexpected consequence of noise -- relatively large levels in the film image due to grain, versus low levels of pixel noise from the DSLR.  In any event, the major point at the time was that the DSLR image did look good enough to keep me happy.

The other advantages of DSLR seemed obvious, so I decided it was time to buy one and see how I liked it.

As it happened, I liked the DSLR very much.  In fact, I liked it so much that I just kept using it and using it.  Periodically, the thought occurred to me that I really ought to objectively compare the DSLR's images with film, but somehow I always had better things to do.  For three years and 50,000 DSLR exposures, I never got around to doing the comparison.

In May 2007, several discussions on a variety of topics prompted me to (finally!) run the test.  DSLR versus film -- same time, same subject, what happens?

The major results are listed above, in the section labeled "Abstract".  Details, of course, are recounted in the sections that follow.

Test Setup

Here's how the tests were set up. 

The target consists of a weathered wooden fence plus reproductions of a USAF resolution chart in high, medium, and low contrast. 

Optics are an Olympus 135mm f/4.5 bellows lens with adaptors to fit a Canon 300D digital SLR and a Pentax ZX-30 film SLR. The entire setup is exactly the same, except for which camera is mounted on the back of the bellows.

DSLR images were shot as raw (.CRW) and converted to 16-bit RGB using Adobe Photoshop CS with default conversion parameters.

Film was processed at a local one-hour photo service, then scanned at 13,500 pixels per inch (534 pixels/mm) using a 4:1 digital photomacrography setup, as shown in this photo.

For comparison purposes, what I'm showing here are actual pixel crops from the DSLR images, matched against the same subject areas from the film.

All image pairs from here down are shown in grayscale, to avoid unnecessary confusion from color differences.

First, here is an image pair to illustrate the conclusion that: "Overall, this 6.3M pixels DSLR is approximately equal to the tested film, when imaging the same subject area.  The DSLR has less resolution than film for high contrast features of the subject, but it captures more low-contrast features."  Of course, this had to be shot using a different lens on the DSLR, so as to image the same area as the 135mm Olympus on 35mm film.  (The lens is a Canon EF 55-200mm USM zoom, at f/11.)

(Note:   These two frames were shot long enough apart that there is visible difference in illumination of the middle board, due to different angles of light reflecting off the back side of the outer boards. This doesn't affect the appearance of the resolution charts or the two outer boards, so it's not worth yet another photo session to fix it.)

Now let's get into the bulk of the data.  This consist of two long sequences of images, one for DSLR and one for film, varying exposure in 1 f-stop increments from grossly under- to grossly over-exposed.  In addition to comparing DSLR and film at corresponding exposures, it will also be interesting to look at how the gradation and effective resolution varies depending on exposure.

Following are thumbnails that summarize the sequences.  For "dynamic range", I'm using a criterion of being able to recover the medium-contrast target at 20 line pairs per mm.

Canon 300D Raw at ISO 200. Optimum exposure is at approximately 1/30 second (00p03330sec.jpg).  Dynamic range is 11 f-stops, ranging from 1/8 second at f/11 down to 1/4000 second at f/22.

Kodak Gold 200 film.  Optimum exposure is also approximately 1/30 second (00p0333sec.jpg).  Dynamic range is15 f-stops, ranging from 15 seconds at f/11 down to 1/2000 second at f/11. 

Here are the gory details.  Shown as actual-pixels crops from the DSLR (on left), correspondingly resized scans from the film (on right).  All images have been level-adjusted by hand to produce comparable appearance to the extent possible.

Note that your browser window must be set to 1280 pixels wide for this to display side-by-side -- otherwise it will be DSLR actual pixels on top, film scan below.

1/4000 second at f/45.

1/4000 second at f/32.

1/4000 second at f/22.  I'll give credit here for resolving group -2, element 5 (2nd from the left, top row), although it's so noisy I'm not really sure.

1/4000 second at f/16.  Group -2, element 5, is clearly resolved here.

1/4000 second at f11.

1/2000 second at f/11.  This is about the minimum usable density for film.  Group -2, element 5 is clearly resolved.

1/1000 second at f/11.

1/500 second at f/11.

1/250 second at f/11.

1/125 second at f/11.

1/60 second at f/11.

1/30 second at f/11.  This is arguably an optimal exposure for both film and DSLR.  Group -2, element 5, corresponds to 20 line pairs/mm at the film/sensor. 

1/15 second at f/11.  The DSLR image is saturating at this point

1/8 second at f/11.  Group -2, element 5, is clearly resolved, but is starting to saturate for the medium-contrast detail.

1/4 second at f/11.  The DSLR image has completely lost it at this point, for the medium-contrast detail.

1/2 second at f/11.

1 second at f/11.

2 seconds at f/11.

4 seconds at f/11.

8 seconds at f/11.

15 seconds at f/11.  Group -2, element 5 is resolved in the film at this exposure, but not the next.

30 seconds at f/11.

By the way, you may be wondering whether lens resolution is limiting in these tests.  It's not.  Using the photomacrography setup to observe the aerial image cast by the Olympus 135mm lens at f/11, we see that the lens delivers over 125 line pairs/mm, well over twice the resolution captured by the film even at optimum exposure.  There is always the possibility of focusing error, of course, even though I focused the 35mm camera using a 7X monocular looking through the eyepiece.

Here are the optics.

Here is the capture of the aerial image.  Note that Group -2, element 5, gives 20 line pairs/mm with the Olympus 135 mm lens.

Here is the corresponding scan of the negative from 1/30 second exposure.  Note that the film resolves one full group less than appears in the aerial image, even though all the optics are identical.

This comparison also addresses another issue raised by one reader of this page: "Were the results affected by the ZX-30's lack of mirror lockup?"

The answer to this question seems to be clearly "No."  In the ZX-30, both mirror movement and shutter movement occur in a vertical direction.  So barring some pretty weird interactions, mirror and shutter vibrations would introduce vertical but not horizontal blurring.  Looking very closely at the above photo, and across all photos in the most problematic range, e.g. 1/125 to 1/8 second, there is at most 1 chart element difference (6'th root of 2) between what is resolved vertically and horizontally, and from one shutter speed to another.  The lack of vertical blur and the consistency across a wide range of shutter speeds indicates that vibrations due to mirror and shutter movement were not significant issues.

Discussion and Conclusions

See above.  There's not much to add at this point.

Related Reading

Regarding dynamic range, Ben Kreunen writing at http://www.path.unimelb.edu.au/~bernardk/tutorials/360/technical/hdri/ claims to show 15 f-stops for Fuji Reala color negative film. 

On the other hand, Roger N. Clark writing at http://www.clarkvision.com/imagedetail/film.vs.digital.summary1.html claims that "print film shows about 7 stops of information". 

The difference is easily explained.  Kreunen measures the maximum range that can be achieved by exploiting the film's tolerance for overexposure.  This corresponds to the old advice, "when using negative film, expose for the shadows".  Clark, meanwhile, measures the range manifested by much less exposure, essentially "exposing for the highlights".  (He's very explicit about this: "the three data sets are registered at the bright end" -- http://www.clarkvision.com/imagedetail/dynamicrange2/, about 20% of the way down.)  Their goals and results are different.  Kreunen's approach gives maximum dynamic range but sacrifices highlight gradation to get it.  Clark's strategy gives excellent highlight gradation, at the cost of losing several f-stops of dynamic range.

Referencing This Work

This web page is a personal publication of work performed by the author.  You may link to this page, but please do not reference in archival publications; contact the author instead.  All images and text are copyright Rik Littlefield, 2007-2009.

Change History

Minor edit February 11, 2009, addressing a reader's concern about vibration due to mirror & shutter movements.