Canon 5D MarkII v Mamiya 7

Digital Versus Film Revisited

As a landscape photographer, I have more than a mild obsession with detail.  Many years ago, I visited the Ansel Adams Gallery in Yosemite Valley and witnessed first hand the endless depth and richness of the late master’s prints.  The clarity and detail in Adams’ original prints had me absorbed in a way not unlike standing before nature herself.  Unfortunately, the ensuing obsession came with a high price tag, both financially and physically. The large format equipment needed to produce such images is expensive, heavy and slow. 

Eventually, I settled on medium format as a compromise between image quality on one hand and speed and portability on the other.  Though I've worked with several different systems, the Mamiya 7 has been my tool of choice for many years.  With exceptional optics, a 6x7 cm image and light weight, the Mamiya allowed me to get to locations where larger formats dare not tread.  Its speed and ease of use made it possible to capture the moment instead of standing by a tripod waiting for the moment to happen.

With the transition to digital photography, however, the effort required to process the film, scan the film, remove the invariable dust spots and process the file started seeming laborious and anachronistic compared to the all digital workflow.  Moreover, without the benefit of instant feedback, film based cameras didn’t afford the opportunity to make useful changes on location.  Only days or weeks later would you know what you got.  Unfortunately, DSLRs still couldn’t compete with the image quality of larger film formats, while medium format digital was and remains in the realm of the financially gifted.  But how quickly things change.

Late in 2008, two cameras were announced that promised to change the game.  The Sony A900 and Canon 5DMkII both herald a resolution and price tag that are commensurate with the desires and budget of many landscape photographers.  Moreover, being based on the traditional 24x36mm format, they offer a wider range of lenses and options than any other format or technology.  The large image circle required for medium format lenses makes design difficult and expensive and producing large image sensors pushes the price into the stratosphere.  Smaller formats, such as APS-C and compacts, suffer from the inevitable noise degradation of small pixel sites.  The question arises, are the new breed of high resolution, full frame DSLRs the new sweet spot for landscape photography on the move and on a budget?

Without the benefit of  anything but a spec sheet, I pre-ordered my Canon 5DMkII way back in September on the faith that it could replace my trusty Mamiya 7.  For many this is bound to be an academic question, having abandoned film to the dark ages of memory.  No doubt that many of you have been happily shooting landscapes with a D2x, 5D, D300, 40D, K20D, and even a handful of medium format digital systems, for some years.  There are many very good DSLRs and even a few good compacts that offer great image quality for smaller enlargements.  But the smaller file sizes from 12 or so megapixel sensors do not scale well beyond about 16x24 inch prints.  The 6x7 cm film format, on the other hand, produces wonderful 24x30 inch and larger prints.  So how does the 5DMkII compare?

At the risk of beating a dead horse, I’ll start by saying the resolution of film and digital sensors are not directly comparable.  Different output media and sizes will benefit one or the other technologies.  As a landscape photographer, my benchmark is set by the largest print size that I regularly create.  That happens to be around 24x30 inches for the 6x7 format, with the occasional 30x38 finding the light of day.  Therefore, I chose 24x36 and 28x36 prints, for the 5DMkII and Mamiya 7 respectively, as the reference point for comparing image quality.  I found the results to be a bit surprising.

The methodology used was an effort to get the maximum detail from each camera.  The shots were taken simultaneously with the cameras tripod mounted about a foot apart, employing both the self-timers and mirror lockups.   For the shots included in this comparison, I used the Mamiya 7 with an 80mm lens at f11 and Fuji Velvia 100 film.   The Canon was mounted with the 24-105mm lens set to 40mm and f8, shooting RAW at ISO 100.  This produced roughly equal fields of view and depth of field given the different format sizes.  Also, both lenses were set near the peak of their respective performance.  If anything, the Canon was at a slight disadvantage owing to being a zoom lens competing with an extremely sharp prime. The images below show the full frames and the cropped areas.

Mamiya 7 80mm f11 / Fuji Velvia100 / 4000dpi Scan / No Grain Reduction or Sharpening

Canon 5DMKII 24-105 @ 40mm f8 / ISO100 RAW ACR Default / No NR or Sharpening

Because of the notoriously inaccurate Mamiya view finder, only 83% coverage, it was a bit of guess work getting the exact same framing in each camera.  Of course, this has no effect of the image quality of the crops.  It just shows a point in favor of the Canon as a photographic tool with a 98% view finder and 100% live view.

The Fuji Velvia was processed normally, then oil mounted and scanned on a dedicated film scanner at 4000 dpi, 16 bits, without sharpening or noise (grain) reduction.  The finished image size is 11024x8819 pixels, resulting in a 580 mb TIFF!  Compare that to a 21 mb raw file from the 5DMkII that is interpolated to a 16 bit TIFF for a finished file size of 126 mb.  You would think there’d be no comparison in printed resolution.  Think again.

The 5DMkII raw file was processed with Adobe Camera Raw at default settings, except sharpening and noise reduction were set to zero.  Only a slight amount of chromatic aberration was corrected.  Finally, the white points of the two files were matched to better compare their dynamic ranges.  The Fuji Velvia, true to form, is more saturated than Canon’s default output.  If desired, this would be easy to match in the 5DMkII file.  Below you’ll find crops of both files before further processing was applied.

Mamiya 7 Fuji Velvia100 / 4000dpi Scan / Unprocessed / 100% Magnification

Canon 5DMKII ISO100 RAW / Default Settings, No NR or Sharpening / 100% Magnification

Canon 5DMKII ISO100 RAW / Default Settings, No NR or Sharpening / 200% Magnification

As you can see, the image from the Mamiya is about twice the size of the image from the Canon.  If we compare the Mamiya to the 200% crop from the Canon, we see that the scanned film contains subtly more fine detail.  Surprisingly, the 5DMkII 200% crop shows better acutance (edge definition) than the much larger scanned film image.  I say surprisingly because the film based image has every advantage, an exceptional prime lens and a larger format by a factor of two.  As you will see, the acutance of the digital image is a significant player in the end game, the print.

The images, at this point, were processed and optimized for the final printed output size.  Since I was primarily interested in resolution and to a lesser degree dynamic range, I didn’t bother with color, saturation, or other artistic considerations.  For the Mamiya image, processing included some noise reduction, mainly to minimize the film grain in the sky.  As such, NR was applied only in the blue color range and the detail in the foreground was largely untouched.  The image was also sharpened and very slightly down sampled from the scanned resolution of 11024x8819 to 10800x8640.  This amounts to going from 36.75 to 36 inches on the long side.  The Mamiya crops below show a section of sky before and after noise reduction.  Note that the 5DMkII crop, also resized and sharpened for printing, had no noise reduction of any kind.  Remarkable!

Mamiya and Canon Crops Demonstrating Levels of Grain or Noise

The Canon image needed to be up sampled significantly to match the Mamiya image’s 28x36 inch 300 dpi resolution.  Having tried many different forms of interpolation, I’ve concluded that up sampling in sequential, small increments is as good as any and better than most in retaining detail and edge definition.  By comparison, using Genuine Fractals to scale the image rendered a more processed look. Once up sampled to 24x36 inches at 300 dpi, the image was then sharpened exactly the same as the Mamiya image.  Below are crops of the print ready files. These crops, depending on your monitor, will be 3-4 times the size they are in print.

Mamiya 7 Velvia100 / 28x36 Crop / Processed For Print / 100% Magnification

Canon 5DMKII ISO100 RAW / 24x36 Crop / Processed For Print / 100% Magnification

While these 100% print ready crops don't look all that great, keep in mind that at this magnification the whole image is about 6x9 feet! That said, the Mamiya 7 image still shows subtly more fine detail and has a less processed look.  By contrast, the Canon image looks coarser and a little, well, interpolated.  In terms of resolution, at 100% viewing on a monitor, I still give the edge to the old film format.  There are, however, other key considerations, namely acutance, dynamic range and noise or grain as it were.  Here the Canon pulls ahead with noticeably better edge definition, more open shadows and shadow detail, and much less noise. Since this comparison is intended as more than an exercise in extreme pixel peeping and is ultimately interested in the finished product, a fine art print, I’ll get to the point.

The files were then printed on Ilford Gallerie paper with an Epson Stylus Pro 9600 at the highest quality settings.  In terms of perceived image quality, the 24x36 inch print from the 5DMkII is surprisingly better than the best I could squeeze out of the Mamiya 7 and Velvia film.  Peering at the prints from a few inches away, I cannot see any more detail in the Mamiya 7 print. The 5DMkII print, by contrast, appears sharper because the details have better edge definition as well as shadow and highlight detail separation. The subtle advantage in absolute resolution retained by the film format, it appears, is lost beneath the resolution threshold of ink on paper and human vision. It is also overshadowed by the heightened edge definition found in the Canon 5DMkII files.  Additionally, the shadows in the Velvia image are blocked up and create the illusion of muddiness and areas without detail, further softening the overall print appearance.  Finally, the lack of noise in the Canon image is exceptional and may, in part, account for the better acutance.

Since analysis can get long winded and still not make the point, I've included two files for you to download and print if you'd like to see for yourself. The files are print ready, 8x10 inch crops of the full 24x36 or 28x36 originals. The crops were saved as maximum quality JPEGS from TIFF originals and are available for download here and here. All in all, I was surprised not only to see a 21 megapixel digital camera hold its own at this print size, but to best my reigning champ by a noticeable margin.  I’ve printed images from a few other DSLRs, including the original 5D, and none until now have matched or exceeded the best from the Mamiya 7.  Perhaps at print sizes of 30x45 and above, the 5DMkII images would start to fall apart.  That’s a question for another time.

When you read the various internet review sites and photography blogs, you get everything from film is long past dead to get a second mortgage and buy medium format digital to achieve the detail and tones still available with film.  By my own calculations and prior experience, I assumed it would take about 40 megapixels to achieve parity with my Mamiya 7 and Velvia film.  It seems, at 100% on screen magnification this may still be true.  But in large 24x36 inch printed output, even viewed from 6 inches, the better acutance, dynamic range and grainlessness of the 5DMkII image creates the impression of greater sharpness and detail.

Is the Canon 5DMkII then the new sweet spot for landscape photography on the move and on a budget?  I’m excited and hopeful that it may well be.  While I haven’t had the chance to test a Sony A900 or the new Nikon D3x, I suspect the results will be similar or even a bit better owing to their higher resolution sensors.  But being budget minded, the Canon is difficult beat.  Moreover, the added functionality of shooting digital along with the flexibility of a DSLR system makes the comparison with any film format smaller than 4x5 a no-brainer.

While this exercise in pixel peeping was designed to answer only a narrow set of questions, and not as a complete camera review, I hope it was still useful to some of you.  Do I have gripes about the 5DMkII?  Sure, a few:  I’d like the ability to turn off noise reduction for low ISO jpegs to retain more fine detail.  The difference between low contrast detail in jpegs and well processed RAW images is dramatic.  There are a couple of ergonomic foibles as well, but nothing I can’t get used to.  All in all, the camera is a pleasure to use and the output, if properly processed, is phenomenal.  To achieve noticeably better high end print quality, you need either a large format film or medium format digital camera.  The former is too large, heavy and slow for the work I do, though still a viable option for many.  The latter has some real limitations, including system flexibility and most of all price.  The portability, flexibility, price, and above all image quality, of the 5DMkII have indeed made it my new tool of choice.

The first portfolio images from the 5DMkII will be posted soon.