Comparing "real-life" image quality of various film formats and 2 digital cameras

Nederlandse Versie

Purpose of these experiments

The experiments described here were designed to answer the question: what can we expect from photographs taken on several film formats and with digital cameras in terms of image quality and resolution. Scanning digital backs for medium and large format are not included in the test, since they do not allow universal photography due to their slow scanning action.

The tests were performed in a standardized environment with as an endproduct a digital file that can be viewed on a high quality screen or printed, preferably with a high quality printer, such as a Durst Lambda. It should be stressed here that it is a real-life test, irrespective of what manufacturers give as specifications of their products. Therefore the results may be disappointing when one tends to use these specifications as real-life hard numbers. The endresult of the tests are in terms of ultimate number of Mega pixels and printsizes that appear critically sharp to the human eye at reading distance.

These tests are completely independent of any manufacturer of cameras, scanners or beer brands.

The cameras and formats

Gandolfi

Gandolfi Variant 8x10 inch
 Linhof

Linhof MT 2000 4x5 inch
Plaubel

Plaubel Makina 670 6x7cm
 M6

Leica M6 35 mm
300D

Canon EOS 300D 6.3 Mpixel
 S50

Canon Powershot S50 5 Mpixel



Experimental conditions

With the exception of the two digital cameras used, all cameras have top quality optics mounted. They were not limiting to the resolutions measured, neither in the centre, nor in the corner of the image at the apertures used. This was established with a 50x microscope looking at the air image of a USAF chart, which indicated a resolution of 100 linepairs/mm or more in the image. The lens of the digital reflex camera used was not top quality but had a resolution of more than 2.5 times (at full opening and on axis) offered by the whole camera, so here the lens was not limiting either. All cameras were mounted on a very stiff tripod and head. The shutter was always operated by cable release or self-timer. Focussing in all four conventional cameras was checked on reproduciblity and accuracy. Focussing (standard deviation of 0.02 % of focal length) in the large format cameras was well within the depth of focus at the used aperture (using an extremenly critical circle of confusion of 0.01 mm in the 4x5 camera). De shift of focus due to stopping down was within the accuracy of focussing and within the depth of focus. Film flatness in all film based cameras was within the depth of focus. All format films had the same emulsion and developed with the same computerized developing machine (Jobo ATL 2000), using the same procedure. The digital cameras were used in RAW mode, after which the files were processed using the camera plug-in of PhotoShop CS to correct for chromatic error in the lenses. Both digital cameras were used in the factory setting of sharpening, while no other effects were used.

The overall picture


Overview

The beerbottle at the left and the USAF testchart left of the middle, indicated with red rectangles, will be magnified below.
The real life scene was 2.4 meter wide. The USAF test chart had a finest grid of 3 linepairs per millimeter in the 1/6 pair of testblocks.
That implies that under ideal conditions 14400 pixels are needed in the width of the image to resolve these finest lines and that would boil down to 311 Mega pixels. In the real world more pixels are needed due to aliasing or interference between the grid of pixels and the testobject and due to the fact that pixels do not form the limit of resolution because of scanner defocussing and limitations and image processing inside digital cameras.

The film used was Fuji Provia 100F.

The cameras, formats, exposure and scan information:
  1. Gandolfi Variant 8x10 inch with 240mm Apo Sironar S, f22 at 2 seconds, scanned @ 3200 dpi with Epson 4870
  2. Linhof MT 2000 4x5 inch with 150mm Sironar N, f16 at 1 seconds, scanned @ 3200 dpi with Epson 4870
  3. Plaubel Makina 670 with 80 mm Nikkor, f11.5 at 1/2 second, scanned @ 3200 dpi with Epson 4870
  4. Leica M6 with 35 mm Summicron, f8 at 1/4 second, scanned @ 3200 dpi with Epson 4870 and scanned @ 5400 dpi with Minolta 5400
  5. Canon EOS 300D, Canon EF-S 18-55/3.5-5.6 used at 30 mm, RAW, ISO 100, f8 at 1/4 seconds
  6. Canon Powershot S50, RAW, ISO 100, f8 at 1/4 seconds

The beerbottle


8x10beer

8x10, scanned @ 3200dpi, 
Pixelwise magnification: 0.25 x
Enlarge
 4x5beer

4x5, scanned @ 3200dpi,
Pixelwise magnification: 0.5 x
 Enlarge
 6x7beer

6x7, scanned @ 3200dpi
Pixelwise magnification: 1 x


300Dbeer

EOS 300D, from RAW
Pixelwise magnification: 3 x
 s50beer

Canon S50, from RAW
Pixelwise magnification: 3.5 x


35beer

35mm, scanned @ 3200dpi
Pixelwise magnification: 2 x
 35mmbeer5400

35mm, scanned @ 5400dpi
Pixelwise magnification: 1 x


Conclusions from these magnifications

    1. There are slight differences in exposure of the images, which is not taken into account in these evaluations. The beerbottle was a very difficult object to render due to the silver label. Also there are some color differences, which are not taken into account either.
    2. There is a clear ranking in image quality, which can be explained by the number of pixels given
    3. However, the 35mm scan @ 3200 dpi with the epson scanner is not better than the EOS300D image, this is due to limitations of the scanner
    4. At 5400 dpi the 35mm scan is superior to the EOS300D image. Compare for instance the year indication "1744" in both images.
    5. The image from the S50 has much more image noise than that from the 300D (due to size of the image sensor) demonstrating that Mega pixels is only one measure of image quality
    6. The scanned image from 6x7  @ 3200 dpi is slightly better than the 35mm image scanned at 5400 dpi, which is largely due to the grains in the film of the 35 mm image, which becomes limiting. Also 3200 dpi on 6x7 cm produces more pixels than 5400 dpi on 24x36 mm.
    7. The film-based images have a higher contrast due to the limited range of the film used (4.5 stops). The digital cameras have approximately 5.5 stops with a rather long toe in the shadow range.
    8. The 8x10 image is truly twice as sharp and contains 4 times as much information as the 4x5 image and so 8x10 should be considered when the ultimate in image quality is desired.


The USAF chart


8x10USAF

8x10, scanned @ 3200dpi, 
Pixelwise magnification: 0.25 x
Enlarge
 4x5USAF

4x5, scanned @ 3200dpi,
Pixelwise magnification: 0.5 x
 6x7USAF

6x7, scanned @ 3200dpi
Pixelwise magnification: 1 x
35mmUSAF

35mm, scanned @ 3200dpi
Pixelwise magnification: 2 x
 35mm5400

35mm, scanned @ 5400dpi
Pixelwise magnification: 1 x
 35mm5400detail

35mm, scanned @ 5400dpi (detail)
Pixelwise magnification: 2 x

Here is space for an image of a digital
camera that equals 35 mm scanned at
5400 dpi
 300DUSAF

EOS 300D, from RAW
Pixelwise magnification: 3 x
 S50USAF


Canon S50, from RAW
Pixelwise magnification: 3.5 x

Based on these charts, the following resolution and image content in resolved Mega pixels:



Camera or Format resolution in
filmplane in dpi		 resolution in dpi
 relative to 35mm		 resolution in
linepairs/mm
in filmplane
 effective number of
resolved Mega pixels
canon powershot S50 7200 1300 142
 4
canon EOS 300D 2600 1600 51
 5
35mm scanned @ 3200 dpi 2400 2400 47
 8
35mm scanned @ 5400 dpi 3400 3400 67
 15
35mm original* 3800 3800 75
 19
6x7 scanned @ 3200 dpi 2200 4200 43
 29
6x7 original* 3500 6800 71
 83
4x5 scanned @ 3200 dpi 2200 7400 43
 86
4x5 original* 3200 10800 63
 209
8x10 scanned @ 3200 dpi 2200 14900 43
 375
8x10 original* 3100 21500 61
 836


In graphic form:

Graph


In graphic form with logarithmic scale:

GraphLog


* The resolution of the original transparencies was determined with a microscope at 50 x magnification.
  These are given here with reference to conventional prints under ideal conditions or projection with high quality optics or with the use of better or future scanners

Conclusions from these magnifications:

    1. Again there is an obvious ranking in image resolution, which is due to the number of pixels available.
    2. The on film resolutions in the 4 formats measured represented the 20-25% modulation area of the MTF curve published by Fuji for the Provia 100F emulsion. This supports the statement that the optics and focussing were not limiting in the conventional cameras.
    3. The 35 mm scan @ 3200 dpi seems less sharp than the 300D image, but has better resolution. In fact the 300D shows some strange artefacts, due to the sharpening algorithm. In the -1/5 block of stripes, the stripes appear to be rotated by 90 degrees!
    4. The modulation transfer of a digital camera is very different from a film-based camera: The modulation of a digital camera remains maximal up to the maximal spatial frequency is reached. The modulation of a film-based camera gradually drops until the maximal resolution is reached and thus gives a softer image
    5. The chart was never limiting for the 8x10 and 4x5 images on the original film, but the scans are limited to about 2400 dpi, which is only half of what the Epson 4870 is supposed to resolve.
    6. Surprisingly, the images from the EOS 300D and from the S50 are limited to 2500 and 2200 pixels total width of the image, which is respectively 4 and 3.5 Mega pixels, both below the claimed 6.3 and 5 Mega pixels
    7. The EOS 300D has reached the level of a 35mm shot scanned at 3200 dpi with a flatbed scanner. However, the original film still contains 4 times the amount of information, of which 15 Mega pixels appears to be recoverable with a dedicated 35 mm film scanner: the Minolta 5400. 
    8. When digital cameras become available with 20 Mega pixels, apart from the optics and from the process of using classical cameras, there will be no more advantages to 35 mm cameras. When scanning is used, this limit shifts to 15 Mega pixels.
    9. However, 8x10 inch cameras  will only become obsolete when digital cameras will be produced with 836 effctive Mega pixels
    10.  4x5 inch cameras will be outrun by 209 effctive Mega pixel cameras.
    11. 6x7 will become obsolete when 83  effctiveMega pixel cameras arrive.

Remark on modulation of digital images:

As can be seen in the image of the USAF chart taken with the Canon EOS 300D, there appear lighter ghost lines around the black stripes. These are due to the sharpening algorithm in the camera and can be expressed as a modulation of higher than 1. For some viewing distances, those where the human eye's resolution is slightly better than the spatial frequency in the image, this can be favourable for the impression of sharpness. For spatial frequencies well below the resolution of the eye, the image looks brutal and artificial and has a distinct "digital" look. The modulation as a function of spatial frequency in an image is very important for the sense of realism or beauty in an image. Therefore unsharp masks should be used with great care both in digital cameras and in the scanning process or postscanning operations.
As a rule of thumb one can use sharpening algorithms up to the point where ghost lines in high contrast parts of the image start to appear and then reduce the sharpening process a bit. In this way the modulation is increased for all spatial frequencies without producing unwanted artifical lines in the image. The same holds for the setting in digital cameras, which tend to oversharpen the image.

Final conclusions

  1. Using convential cameras, the order of increasing weakness of the links forming the final image in the present experiments was: 1) lens, 2) film, 3) scan. In other conditions the lens may become the weakest link, for instance at large apertures or defocussing, or the film when large grain film is used.
  2. All film formats, 35 mm, 6x7, 4x5 inch and 8x10 inch are superior to digital camera's with 6.3 Megapixels or smaller
  3. However, when a flatbed scanner is used at 3200 dpi, such a digital camera makes 35 mm superfluous (apart from the optics and feel and use of a classical camera). This is still not the case when a dedicated scanner for 35 mm is used.
  4. Digital cameras are wonderful devices for quick evaluation of light and image impact, and so have acquired an important place in photography.
  5. In non-civilized conditions (extreme heat, cold or absence of current sources), film based classical cameras have advantages

Final remarks

It is understandable that photographers have flaming debates on the choice between film based cameras and digital cameras. This always happens during marketing of new products with enormous sales figures. The word "digital" has become a buzzword in the consumer world. One day when I was setting up my Linhof 4x5 camera, a man came over in admiration and remarked: "that must be a digital camera!" The camera firms obviously have done an excellent job in hammering down this insight. However I feel that a camera is an instrument used to make (interesting) pictures. It consists of a lens, a shutter, some control of exposure and a sensitive surface. This sensitive surface can be either a chemical emulsion or an electronic matrix.

The revolution is about this surface and, as can be seen from these tests, these surfaces have different properties. This can be used to the advantage of the photographer: every photographic opportunity has an optimal camera. In order to make the choice, the photographer needs to know the properties of the cameras. That is one of the purposes of tests like the one I've presented here.

Most likely we should not speak of the digital revolution, but of a gradual change of photography as an art form and human activity. In the end, the image is appealing or not, is innovative or "old hat", is an enrichment of the collective photographic heritage or just a waist of film surface or disk storage. That is what is essential. I do not feel that a choice needs to be made to use either film based photography or digital photography. Either can be used successfully, as is obvious from the fantastic flow of creativity on the net, for instance on www.photo.net with an input of thousands of images each week, some of them truly inspiring. When looking at some of them, I would love to be able to see the originals, either on an excellent screen in high resolution, on a light table or as a large print hanging on the wall, well illuminated.


Bert Otten


artwork                                     Printsizes