Index of this page

• Test setup
  • Equipment used
  • Films used
  • Equipment and film choice
  • Test chart description
  • Sharpening digital images and presentation in this test
  • The definition of a "line pair" and resolution testing

Equipment used

- Minolta Dynax 7 (analog version)

- Sony Alpha 900

- Minolta 50mm 1.7 prime / standard lens

- Manfrotto 190B tripod with 141RC head

- Minolta remote release

- Minolta 5600Xi and Minolta 5600HS flashes & umbrellas

- Imacon Flextight 848 scanner

Films used

- Kodak Ektar 100 (Color negative)

- Kodak Portra 160VC (Color negative)

- Kodak TMax 100 (B&W negative)

- Fuji Velvia 100 (Color positive / slide - RVP100 emulsion)

Equipment and film choice

Why the choice for this equipment? Well, mainly for practical reasons. My Minolta Dynax 7 has been a long time companion and the 1.7 prime lens has given excellent results in the past. Being virtually distortion free and a fixed focal length prime lens, this lens is ideal for such a test where quality of the lens is of utmost importance. In addition, the 1.7 maximum aperture allowed for a bright viewfinder and accurate focusing. For a decent review of this particular lens, see this reference: Minolta AF 50 mm f/1.7- Legacy test report. As for the choice of films, these were the films I had in stock at (about) 100 ISO, and it is only useful to compare films at roughly the same ISO.

Lastly, the scanner choice is the only practical option too. I realize that some of you might argue it might have been better to use a drum scanner and it's sophisticated options for "hiding" grain and it's superior dynamic range, even compared to the high end quality of the Imacon. However, having film drum scanned is both expensive, and a not very common service here in the Netherlands, whereas I can simply hire the Imacon scanners for a very reasonable price per hour at the same professional lab that develops my color film. Having these images drum scanned would probably have cost me 10 times the amount of money...

To show you the difference between a professional scanner and a consumer grade combined flatbed / filmscanner, I hereby include an 4800 ppi scan of my consumer grade Canon 9950F. Although Canon boasts this scanner to have "4800 ppi" resolution (ppi = Pixels-Per-Inch, so the number of pixel rows or columns a scanner or digital camera captures per inch), in my experience it doesn't do more than an actual optical resolution of 1200 ppi, good enough for large format 4x5 inch photography, but not 35mm. Above that value, it just adds blurry redundant pixels, that do not add to the actual image detail.

To show this last aspect, I have also made scans at 1200 and 2400 ppi that are subsequently upscaled in Photoshop to 4800 ppi using a bicubic interpolation. The small insets in the 1200 and 2400 ppi upscaled scan samples represent "actual pixels" or 100% viewing crops of the original 1200 and 2400 ppi scans. As you can see from these images of the 9950F, the straight 4800 ppi scan doesn't look any better than any of the upscaled versions, even compared to the low resolution 1200 ppi scan, meaning the actual optical resolution of this scanner is no more than a measly 1200 ppi... Do notice though, that at 1200 ppi, the "actual pixels" inset looks pretty good and reasonably sharp, the reason why I still consider this scanner suitable for large format scans. Even a low 1200 ppi scan of a 4x5 inch negative results in a 4800 x 6000 pixel image size, so an almost 30 Mpixels scan...

To make a fair and clear comparison with the Imacon scans, I downscaled the Imacon scan from 8000 to 4800 ppi. This loses some detail, but as you can see, the Imacon still beats the Canon 9950F by far!

And yes, the difference is REALLY that big and shocking!!! All Canon 9950F scans, as with the Imacon, without sharpening applied. All sample images in this comparison from Fuji Velvia 100.

And now some of the pure digital folks are going to argue that the left Imacon image still looks crap in terms of sharpness and detail compared to their 6 Mpixel digital shots. Well, you have to remember:

• This image is completely unsharpened and more or less equivalent to a RAW output digital shot, whereas JPEG output on a digital camera usually constitutes full in-camera sharpening
• At 4800 ppi, you are looking at a tiny crop of a roughly 4400 x 6400 pixel image (28 Mpixel). At this detail level of scanning, we are stretching 100 ISO film to it's absolute limits and start to reveal the grain or dye clouds that form the image. The dye clouds of a color negative or slide are inherently a bit fuzzy, and will never produce the clean signal of a sensor's pixel.

Imacon Flextight 848 8000 ppi scan downscaled to 4800 ppi, bicubic interpolation	Canon 9950F 4800 ppi straight	Canon 9950F 1200 ppi scan upscaled to 4800 ppi, bicubic interpolation	Canon 9950F 2400 ppi scan upscaled to 4800 ppi, bicubic interpolation

If you want to know more about Imacon scanners and their sophisticated design including the autofocusing mechanism, visit the Hasselblad website and read the following good review by Giorgio Trucco: Imacon Flextight 949, A review of the grapes, as the fox sees them!

A note to Sony Alpha users:

Since I do not own a Sony Alpha 900 yet (nor have the money for it at the moment), and just got to the opportunity to shoot a few test shots with it in far from ideal conditions, I must beforehand apologize for the appalling quality of the flash exposure in this image. I made this photo in the "passport-photo" corner of a (camera-)shop. The test chart was hung on the wall where normally a person would sit, the test chart oriented vertically, with a softbox to the right, and studioflash with umbrella to the left. Unfortunately, both of the flashes were bolted to the ceiling in such a way that I had no possibility to re-position them, resulting in a very uneven flash exposure and severe reflections in part of the image. Sorry for this, I was just glad to be able to shoot with it for comparison. Luckily, the most important part of the chart (the center) was largely spared from reflections and still usable.

Test chart description

The test chart has been created by a German company called IE Image Engineering and features a 60x90 cm large test chart image containing both resolution test lines and a few pictures. The small pictures to the right of the center are repeated both in low and normal contrast for evaluation of (low) contrast rendition of digital sensors. The numbers denoted on the chart represent lines-per-picture-height, where height is the short site of the image, so in case of a regular 35mm negative or full frame digital camera, it's 24mm. Actually, the number on the chart should be multiplied by a factor of 100, so "20" is 2000 lines per 24 mm. In my definition of line pairs, this is equivalent to 1000 line-pairs per picture height, as IE counts both black and white lines.

IE Image Engineering test chart (actual size of test chart area 60x90 cm) , Fuji Velvia 100

Shooting setup

Sharpening digital images and presentation in this test

Please note that almost all of the images presented in this test represent 100% images ("Actual pixels" in Adobe Photoshop), unless otherwise stated, so each captured pixel is truly visible. Some of the images have been sharpened, others not, besides maybe a very modest default amount of sharpening applied by the scanner or camera outside the control of the operator. All images have been converted to the sRGB colorspace for proper display of digital images in webpages.

For those people less familiar with digital images and processing, I hereby include some samples of an image sharpened to different amounts, to show you the effect of what sharpening does with a digital image. Most scanned photos or photos made with a digital camera, actually need a bit of sharpening. Sharpening enhances edge contrast, thereby making an image appear sharper. However, it should generally be applied with caution, as it can ruin an image if improperly applied. For professional digital cameras like digital reflexes, manufacturers often choose to keep in-camera sharpening to a minimum, so as to leave the amount of sharpening up to the photographers choice. The Alpha 900 does appear to do a moderate amount of sharpening though in it's JPEG output. The Imacon Flextight scanner can be tweaked to skip sharpening fully. The high resolution scans and images generated by the Imacon and the Alpha 900 actually leave little room for sharpening. Looking at the results, I am pretty much convinced that for most printing applications, whether a true press type application like a photo in a glossy magazine, or printing a single photo on a professional photo printer, additional sharpening could be left out of the equation altogether for the Alpha 900, whereas only a moderate amount of sharpening would be required for the Imacon scans, as the large images created by both sources already contain huge amounts of detail that will print fine on most regular print sizes.

However, as you can see from the image crops below from the Imacon film scans and the Alpha 900, a certain amount of sharpening does make the eyes "pop" and stand out more. I should note though, that a low radius is an absolute necessity, as even the slightest increase starts ruining the image. There is that much detail captured by the Imacon and Alpha 900 with proper focusing... Actually, the Fuji Velvia 100 scan is on the brink of being oversharpened at a sharpening amount of 200%, but although the left unsharpened image may look more pleasing in terms of "smoothness", the right image that starts to show the grain, does reveal more detail, for example in the lips. And the grain "specks" visible in the sharpened image are not just "artifacts" of the sharpening process, but are actually visible in the left image as well, if you look carefully.

Unsharpened Fuji Velvia 100	Sharpened Fuji Velvia 100 Unsharp Mask Photoshop settings: Amount: 200% Radius: 0.7 Threshold: 2

Unsharpened image Alpha 900	Sharpened image Alpha 900 Unsharp Mask Photoshop settings: Amount: 80% Radius: 1 Threshold: 2 Notice the first signs of sharpening artifacts at the cheeks rim, even with this modest level of sharpening

The definition of a "line pair" and resolution testing

There is lots of discussion about resolution testing of lens / film or lens / digital sensor combos. Unfortunately though, many authors fail to acknowledge the need to define the actual scale at which the resolution is being measured. A common measure is "line-pairs-per-millimeter" where millimeter is a millimeter of the film negative's width or height, or similar for the digital sensor in a digital camera. The value that is assigned to this measure usually defines the maximum number of line pairs that can be distinguished when looking at a print or digital image. It is usually tested by photographing a test chart containing a set of closely spaced black and white lines, often in the form of a set of slowly converging lines as in the test chart used in this test, but regular spaced patterns are also common.

Now, what is a "line pair"? In this test, I assume a line pair is a combination of one black and one white line, as in the figure below:

Line pair definition

Now, while this maybe rather obvious, and the definition of a line pair not to ambiguous, things start to become a little bit more confusing when talking about a scale like "lines-per-millimeter", as opposed to "line-pairs-per-millimeter"... because, which lines do we count? Only the white or black lines, or both? If we just count the black lines, the number will be the same as in my definition of a "line pair", however, when counting both black and white lines, the figure will be double that. This ambiguity and the failure to define the scale at which resolution is measured, gives rise to a whole bunch of misconceptions and misunderstandings about resolution measurements of lenses, and lens / film, or lens / digital sensor combinations.

Worse though, some authors that abbreviate the scale (like "lpmm" for "lines-per-millimeter" or "lppmm" for "line-pairs-per-millimeter) even manage to intermingle these measures in one and the same article, often due to typing errors, adding to the confusion. So you might find "50 lpmm", "50 lppmm" and "50 lp/mm" in one and the same article referring to the same resolution figure. And even worse, typing errors like "lpppm" and "lpmmm" are common too...

Now you tell me what the actual resolution figure must be in such an article!

Back to my definition: I use "line-pairs-per-millimeter" and I will (at least try!!) to consequently abbreviate this as "lp/mm"