DIGITAL SLR vs. FILM SCANS
Including comparison images from
Canon 1Ds - 11mp, Canon 1Ds Mark II - 16.7mp, Canon 10D - 6mp,
35mm Provia F ISO 100 and 4X5" and *6X9cm Velvia
Current Revision -11/21/05
Steve Hoffmann's Nature and Landscape Photography
If you are in a hurry, click here to go to the image comparison page links. If not, find them at the bottom of this page.
Some of the major camera manufacturers are now making DSLR (Digital Single Lens Reflex) cameras that use the lenses designed for their 35mm SLR cameras. These DSLR's have 4 to 16.7 megapixel arrays and are available for between $900 to $8000.
|I was a staunch proponent of using film cameras and desktop
film scanners for digital image creation until I got my Canon 10D DSLR. My experiences with
the 6mp 10D and most recently with my new Canon 1Ds Mark II 16.7mp
DSLR have changed the way I look at digital camera capabilities. After 2
years of comparing DSLR camera images against scanned 35mm and 4X5 film, I
decided to go with a pure digital workflow. I no longer own any film based
Canon DSLR color renditions using auto white balance seem to be very accurate. The Canon DSLR's dynamic range (how much detail can bee seen in deep shadows before bright areas start to lose texture) is equal to or better than my Provia F ISO 100 film scans from my Nikon Super Coolscan 4000 ED. 6mp DLSR image quality can equal or exceed 4000 dpi 35mm film scan quality. The Canon 1Ds Mark II 16.7mp DSLR image quality may equal or exceed 6X9cm film scan quality. Read more about resolution and image quality comparisons in the last section of this article called 'my opinions' and on the image comparison pages.
A film scan is a second generation image. The first generation is camera lens to film and the second generation is film to scanner CCD. Digital camera images are first generation with the camera lens producing the final image. Each step in the process of image creation introduces the probability of some image quality loss. It's become very apparent from my comparisons between 4000 dpi film scans (19mp) and 6mp DSLR output that many small cumulative problems detrimentally affect the final image in the film and scan digital image workflow. These problems might include some of the following: light dispersion in the film emulsion during exposure, halation (light reflected back from the film pressure plate from bright areas in the image), lack of critical film plane flatness in the camera and/or film scanner, film grain aliasing during scanning (apparent magnification of film grain), digital noise incurred during scanning, scanner lens and focusing issues, inaccurate color interpretation by the film dyes and/or by the scanner. Also, film has significantly less resolving power in low contrast situations. As an example Fuji Provia F ISO 100 resolves 140 lines per mm in high contrast situations and only 60 lines per mm in low contrast situations. Many lenses are capable of 80 to 120 lines per mm or more.
Digital camera technology has come a very long way in the last four years. One of the most significant improvements has been in reducing or nearly eliminating digital noise artifacts in 'prosumer' level DSLR cameras. Canon has done an especially good job of noise reduction with the 10D and all the Canon DSLRs introduced since the 10D. 10D ISO 400 and 800 images are very clean when compared to the same ISO output in film. ISO 400 output from the 10D looks much like film grain in ISO 100 Provia F. Canon's newest 'in camera' image processor (DIGIC II) produces high ISO images that have even less digital noise. DIGIC II is incorporated in the new 20D, Digital Rebel XT and the1D and 1Ds Mark II cameras. Digital noise in images from the 1Ds Mark II at ISO 400 is nearly non-existent. Digital noise artifacts in some digital cameras are as noticeable and problematic as high ISO rated film grain. For those of us who like clean looking images film grain and digital noise artifacts limit the upward scalability of an image.
|There are a few things you need to be aware of before jumping on the DSLR bandwagon. Since the imaging sensors in most of the consumer level DSLR cameras are somewhat smaller than a 35mm camera's film area, your lenses don't produce the same field of view in sub 35mm sized sensor equipped DSLR than they would when used with a 35mm SLR. The effective focal length multiplier for most sub 35mm sized sensor DSLR cameras ranges from 1.3 to 1.7 depending on camera model. While this field of view difference is commonly called a focal length multiplier it could more accurately be called a crop factor. There is no actual increase in focal length.|
|The field of view in a DSLR camera with a 1.6X
multiplier is equivalent to what you would see with a 35mm SLR and lens
with a focal length 1.6 times longer. Using the image example above you
can see that a DSLR with a 1.6X multiplier and
a 24mm lens crops the lens covering area down to the same view you would
see using a 38mm lens on a 35mm film
camera (1.6 times 24 = 38.4). This is not a big deal for some people,
especially for those who use zoom lenses or do a lot of work with telephoto
lenses. A DSLR with a 1.6X multiplier and a 200mm lens gives the same
viewfinder image as a 35mm SLR and a 320mm lens. A DSLR with a 1.6X
multiplier and a 17-40mm zoom would give the same viewfinder image as a
28-64mm zoom mounted on a 35mm camera. If you buy a DSLR with a
sub 35mm size sensor you'll also have to buy an ultra wide angle lens (20mm
or less) if
you are a fan of true wide angle view. The popular 35mm 28-105 zoom range
becomes the equivalent of a 45-165 zoom on a DSLR with a 1.6 focal length
multiplier. A DSLR with a 1.6 focal length multiplier would require a
lens to equal a 25mm lens field of view on a 35mm film camera.
There currently is no Canon ultra wide lens solution for Canon's legacy sub 35mm size sensor equipped DSLR cameras. Canon has recently brought out a new line of lenses called EF-S, including a 10-22mm zoom. The EF-S lenses will only fit on certain current DSLR models (Digital Rebel and 20D). Sigma has introduced a 12-24mm EX lens that is available in both Nikon and Canon EOS mount. Nikon has a 12-24mm lens that can be used only on their sub 35mm sensor sized DSLRs.
|A DSLR camera imaging sensor can be capable of very
high resolving power. A 6mp by 22.7 x 15.1 mm sensor resolves 3436 ppi .
The math for this deduction is as follows: the long dimension of the sensor is 22.7mm, which is .894 inch. The long dimension of a full resolution image
from this sensor is 3072 pixels. Divide 3072 pixels by .894 inch and you get
3436 ppi (ppi = pixels per inch). The Canon 1Ds Mark II resolves 3565 ppi.
This small difference in ppi between these two DSLRs does not equate
to only a small difference in image quality. Projected image component size
on the sensor is a huge factor in overall image quality. I will discuss the
image component magnification on the sensor issue in more detail two
You will need to use pro quality zooms or prime lenses to get maximum advantage from a DSLR sensor's ability to resolve very fine detail in an image. As mentioned above, due to focal length magnification factor you'll need to use a wide angle lens on your sub 35mm sized sensor DSLR to get a normal focal length field of view and super wide angle lenses for wide angle views. Wide angle lenses, due to optical design constraints, are not quite as sharp as normal focal length lenses. Sharpness from the center to edge of the image is not as consistent with wide angles as it is with normal and telephoto focal length lenses. Also, image component magnification on the sensor is less with shorter focal length lenses regardless of the type of camera body they are mounted on. So, the actual projected image components are smaller on a sub 35mm sized sensor equipped DSLR and a 28mm lens than they would be with a 35mm film SLR or 35mm size sensor DSLR using a 50mm lens.
Let's say we are photographing a scene with an old barn in the middle of the picture. With a sub 35mm sized sensor equipped 6mp DSLR using a 28mm lens the image of the barn projected by the lens on the sensor is 1/4 high by 1/4 long (.250 inch). With a 35mm film SLR using a 50mm lens the projected image of this barn on the film would be a little larger than 3/8 high by 3/8 inch long (.390 inch). This would allow for more detail to be seen in the 35mm SLR/50mm image since the image of the barn on 35mm film is about 1.6 times larger than the image in the sub 35mm size sensor DSLR with the 28mm lens. However, both the 28mm and 50mm lenses would produce approximately the same composition and 'normal' field of view in their respect camera's viewfinders.
The 3/8X3/8" barn in the 35mm film image when scanned at 4000 dpi would contain about 1550X1550 pixels. The 1/4X1/4" barn in the 6mp 1.6 crop factored DSLR image would contain about 890X890 pixels and in a 1.6 crop factored 8mp DSLR the barn image would contain about 1015X1015 pixels. A reasonable conclusion made from this math would indicate that the barn in the 35mm film scan would win hands down in image quality. After all, the35mm scanned image has three times as many original pixels to recreate the image detail in the barn. Surprisingly, 35mm film scan superiority just doesn't seem to be the case in most of my 35mm film scan vs. 6mp DSLR image comparisons. Sub 35mm sensor size DSLR images that have been resized up to equal 4000 dpi 35mm film scan pixel dimensions compare very favorably to those 4000 dpi 35mm film scans. With the Canon 1Ds Mark II we have about 1388X1388 original first generation pixels for the barn part of the photograph. When comparing apples to apples (10D to 1Ds MK II) 2.5 times as many first generation pixels gives us a quantum leap in image quality.
Under certain circumstances image magnification at the film plane and wide angle lens optical quality issues may reduce or eliminate the advantages of the sub 35mm size sensor DSLR's grainless and noiseless first generation 3436 ppi image over 35mm film scans (see the Santa Ysable Farm image comparisons). The projected image magnification issue is further demonstrated by comparing a full size sensor (35mm size) Canon 11mp DSLR 1Ds image shot at 62mm against the same image shot at 38mm. with a 10D. The 1Ds image is noticeably better than the 10D image and the 4000 dpi scan even though the 1Ds sensor, which resolves 2955 ppi, actually has less resolving power then the 10D at 3634 ppi or the 4000 dpi film scan. See the Oceanside Harbor comparison page. Review the third paragraph of this article for some possible explanations of why a film scan doesn't always compare favorably even when a film scan's resolution is equal too or higher than a digital camera's sensor resolution.
One problem related to digital cameras that merits serious consideration is moiré effect. Moiré is caused by repeating lines or patterns in the image composition that are at a certain frequency to the camera sensor's resolving power. Moiré pattern is displayed as a 'rainbow' type blur that destroys detail in this type of image content. Fence lines, brick walls, patterns in weaved fabric and other repeating pattern type content are all prone to moiré degradation when the pattern magnification is very small on the sensor. See the first Carlsbad Flowers image example set for an example of moiré. Moiré artifacts can be mitigated by a strong 'in camera' anti-aliasing filter and in software post processing. However, strong in camera anti-aliasing filters can soften the entire image so much that some of the finer image detail is lost. Excessive sharpening in post processing can not restore that detail. With software moiré removal solutions the rainbow colors are removed but the detail in the area that was affected by the moiré is usually destroyed by residual artifacts from the moiré effect. Camera manufacturers need to put their engineers to work and figure out how to design sensors that eliminate the possibility of moiré without sacrificing image sharpness. Professional photographers who pay $3000 to $8000 for a camera should expect to see flawless imaging performance with any type of subject matter. Moiré is not an issue in a film and scan workflow because scanner imaging sensors are designed differently than camera sensors. Scanner sensors do not need to make a full frame instant capture.
|Canon and Kodak make DSLR's with full size 24X36mm
sensors (35mm film size). The Canon 1Ds Mark II has 16.7 mp ($8000) Kodak
has the SLR/n for Nikon mount lenses ($4999 and based on a Nikon body) and
the SLR/c for Canon EOS lenses ($3499 and based on a Sigma body). Both of
these Kodak cameras have 14 mp sensors.
While I haven't worked with either of the Kodak cameras, I have done some testing with a rented Canon 1Ds (11mp) and my own 16.7 mp 1Ds Mark II. These full size sensor equipped cameras eliminate the focal length multiplier issues. The only possible downside to full size sensor cameras, besides their current high cost, is that the full covering area of the lens is used and image corner sharpness can become an issue just like with 35mm SLR film cameras.
Lenses are just not as sharp at the edges and in the corners of the image area as they are toward the middle. The sub 35mm sensor equipped DSLR crops out the side and corner areas of the lenses covering area. So, most of your image is being produced from the very center of the lens's covering area where it is at its sharpest. If you want to make technically superior quality big prints from a full size sensor 16.7mp DSLR, you better have good glass. The only focal length range that this is a significant issue is Canon's super/ultra wide angle lenses (24mm and under). Canon just doesn't have lenses in this focal length range, zoom or fixed, that remain decently sharp in the corners. If you own a Canon DSLR and you are a super/ultra wide angle fan you should encourage Canon to make some decent lenses in the 24mm and under focal length range. For the wide angle photography fan there isn't going to be much advantage to adding more megapixels to Canon DSLR cameras until there are lenses capable of taking advantage of all that resolution.
The purpose of this article was to see how the Canon 1Ds, IDs Mark II and 10D cameras' output compared to film scans. I shot the same scene with each camera. Care was taken to focus at the exact same point in all cases. I used f8 for all DSLR and 35mm SLR photos and F16 & 22 for the 4X5 photographs. I tried to use a focal lengths for all of the cameras that would give approximately the same field of view. The 4000 dpi Nikon Super Coolscan scans used in these comparisons were saved in 48 bit color and Camera *RAW images were processed into 48 bit color.
The 4X5" and 6X9cm scans shown in this article were scanned with my Epson 2450. The 4X5 scans were done at 1200 dpi and in 48 bit color. This Epson scanner doesn't pick up much, if any, more image detail scanning at higher resolution. Scanning with the Epson 2450 above 1200 dpi just makes bigger files. After all, this is a **consumer level scanner. I did buy and return an Epson 4870 after tests determined that it didn't resolve even as much detail as my 2450.
All DSLR cameras were set to ISO 100. While these test shots were carefully done with the mirror locked up in all of the SLR and DSLR cameras, I don't claim that the results produced are scientifically accurate comparisons. Differences in lens performance at different focal lengths, subject magnification and other variables should be taken into consideration. We were looking for an approximate benchmark for each different camera's capability.
Since I express my opinion frequently in this article, you might be interested in the criterion I used to judge image quality. I've always considered proper focus and sharpness to be the first thing I look for in an photograph. Secondly, I like to see clean smooth tonality with no visible film grain or digital noise. Color fidelity is last on my list because it is so hard to objectively quantify.
|My opinions -
after doing these comparisons:
Output from the 6mp 10D in most photographic situations is as good or better than my 4000 dpi film scans from 35mm Provia F ISO 100 film. Images from the10D resolve as much or more detail than 35mm Provia F ISO 100 film scans when the same lens, f stop and focal length setting is used (see Oceanside Harbor #2 image comparisons). The 10D's 22.7 mm (about 7/8 inch) length imaging sensor, which has 3072 pixels, produces an image equivalent to a 3436 dpi 35mm film scan. Since there is no film gain and digital noise is almost non-existent in this camera’s output at ISO 100, images can be resized up to 5500X3600 pixels and beyond and the resulting prints are beautiful. I printed the Santa Ysable Farm 35mm film scan and the same 10D image resized up to 5500X3600 pixels to 12X18 on my Epson 2200 printer. Even with very close inspection both prints looked virtually identical. In my opinion the lack of visible film grain gave the visual quality edge to the 10D print.
The Canon 1Ds 11mp DSLR outperforms 4000 dpi 35mm film scans by a significant margin and may be competitive with medium format scans of 6X6 and 6X4.5cm films
The Canon 1Ds Mark II 16.7mp DSLR makes images that are very competitive with 6X9cm medium format scans. 1Ds Mark II images may even compete with 4X5 scans when the 4X5 scans are done on a scanner like the Epson 2450/4870 series scanners. For a little hard copy comparison I cropped off the top and bottom of the original 1200 dpi 4X5 scan (5584X4407 pixels) of the 2004 Oceanside Harbor photo (to give it 6X9 proportions). I then resized this image up to 7200X4800. I also resized the 1Ds Mark II Oceanside Harbor image (originally 4992X3328 pixels) to the same dimensions (7200X4800) pixels or16X24" at 300 ppi output resolution). I cropped each of these images in half and pasted each of the remaining halves into a new canvas side by side. Lastly I cropped the center of this new 7200X4800 pixel combination image down to 3600X5400 pixels. This new combo image is the equivalent of the central part of the original content at 16X24" and 300ppi output. I printed the 3600X5400 pixel image to 12X18" (300ppi) and was surprised that I found the 1Ds Mark II side of the combo image to look sharper and cleaner. I thought maybe I was just seeing what I wanted to see so I showed the 12X18" combo print to my wife. She also picked the 1Ds Mark II image as being 'slightly sharper and a little more detailed'.
I believe that if I had drum scanned the 4X5 chrome and printed this comparison at the equivalent of 30X40", the 1Ds Mark II image would probably not fare quite so well. Nonetheless, I was surprised at how well the 1Ds Mark II side of the combo print compared to the 4X5 side at the equivalent of 16X24" at 300 ppi. The bottom line is that 4X5 film, when properly scanned, (drum scanned) is capable of producing more detailed images than any of the DSLR cameras. I suspect that the difference in detail would be noticeable in prints of 20X30" and larger.
To read a more detailed discussion of the merits of a film and scan workflow vs. a pure digital workflow check out the "Film Scanner or Digital Camera?" section on my Digital Darkroom Imaging and Printing Tech Tips page.
Below you can find links to 8 image comparison pages. The images are all full resolution screen captures saved as high quality JPEGs. Have a look, what's your opinion... :^)
*COLOR RANGE ISSUES BETWEEN SCANNERS AND BAYER PATTERN DSLR SENSORS: This is an area that is argued frequently when it comes to the digital camera vs film scanner image quality debate. Digital camera sensors do not record the same range of colors that a scanner CCD records. The DSLR image is constructed with some color interpolation. This interpolation also affects image detail resolving power in the smallest components of the image. I am perfectly happy with my Canon DSLR's Bayer generated images. Personally, I don't "see" enough color range difference between the scans and DSLR images I've presented here for it to be a serious consideration. The degrading of fine detail caused by Bayer interpolation seems to be a non-issue in my comparative examples too.
**SCANNERS: I have had a demo of Microtek's 1800f and 2500f E.D.I.T scanners ("Emulsion Direct Scanning"). These scanners have glassless film holders and separate scanning paths for transparent and reflective media. 4X5 scans done on the 2500f at 1200 dpi actually resolved a little more detail than my Epson 2450 did at 1200 dpi. The 1800f at 1800 dpi didn't even come close to what the Epson could do at 1200 dpi. So, while the Epson is a consumer grade scanner, it is capable of pretty decent output. The Microtek 2500f retails for about $3000 and would, in my opinion, be the minimum investment for a large format photographer who wants to make fine art quality prints.
I realize that drum scans offer the very best image quality. However, Nikon and Minolta make very nice medium format film scanners that may compete with drum scanners. Other than the Microtek 2500f, the 4X5 camera photographer striving for technical excellence who wants to scan at home or small business should look into Imacon's Flextight line of scanners or used drum scanner. It takes a knowledgeable and experienced operator to get optimal results from a drum scanner.
Two of the most attractive aspects of digital photography are independence and immediacy of results. I don't want to pay for film, processing or scanning services or wait for the latter two of these services to be performed.
To print this web page correctly you will need to set your printer's page setup properties to 'landscape' paper orientation
PHOTOGRAPHY AND DIGITAL IMAGING INFORMATION