Gadgets
Epson Perfection 1640SU Photo Scanner Review
About this Review
My scanner review, tests, and examples emphasize the capabilities of the "Photo" option of the Epson Perfection 1640SU Photo scanner. I concentrate on the ultimate performance of the transparency adapter and its ability to deliver good scans of film. I don't mention the scanning of paper media, because this capability is just fine in this scanner. I also don't talk too much about the design of the scanner or scan speeds, because these aspects are also just fine and have not caused me to complain. I hope that this section is useful to you in your quest to buy a "film scanner."
Summary
The popular notion to 35 mm film scanning is that a flatbed will produce "terrible" results. Contrary to this belief, the Epson Perfection 1640SU Photo scanner can actually produce very good results, depending on the quality of the image being scanned. For an average image that has lots of mid-tones and highlights, the Epson Perfection 1640SU Photo scanner produces crisp, detailed scans. Under ideal conditions, the scanner has a top resolution of about 24.8 lines/mm. The scanner is challenged by shadows and dark areas, however. In such a case, scanning a print may produce better results.
Although the default Twain 5.0A settings introduce strong color casts to scans, Epson's Twain 5.0A driver gives users significant control over the color and tonal adjustments pre-scan. Proper calibration of the scanner results in accurate scans that require only minor post-scan adjustment. Updated versions (Version 5.53A discussed below) of the Epson Twain driver deliver excellent results, with better color accuracy, automation, and device color management.
The Epson Perfection 1640SU Photo scanner is a great choice for someone who may be on a budget and cannot afford to buy both a flatbed and a dedicated film scanner. Versatility is an important aspect to this scanner, which has the ability to scan 35 mm and medium format film, paper media, and three-dimensional objects.
Resolution Up to 24.8 Lines/mm
I scanned a 35 mm negative that contained an arrangement of USAF resolution charts. See the Scan Samples section for more information on how I did this or to download the full file. The raw scan was very soft. For the purposes of presentation on this site, I used Adobe Photoshop 5.0 to apply unsharp mask to the scan. The resulting "sharper" image did not differ significantly from the raw scan in terms of resolvable triplet line pairs.
|
|
|
| Lower Left Corner Chart | Center Chart | Middle Right Chart |
Furthermore, although the overall image was unevenly lit, the ability of the scanner to resolve triplet line pairs was about the same among the individual charts.
|
|
| Center Chart Enlargement (Raw Scan) |
Center Chart Enlargement |
Conservatively, the highest resolution attainable was the "-1,3" triplet line pair. More liberally, the highest resolution was at the "-1,5" triplet line pair. By doing some measurements and simple algebra, I determined that the resolving power of the Epson Perfection 1640SU Photo scanner at 1600 dpi is about 24.8 to 31.7 lines/mm. Not too bad!
This negative represents a nearly ideal scan target, however. It consists primarily of high contrast blacks and whites. Most photographs do not consist merely of black and white lines. Rather, they are black, white, and a range of grays (or colors) in between. Comparing some real life photographs is, thus, useful.
Practical Resolution Is Less
I scanned a 35 mm color negative containing a picture of a Steller's jay in a mist net. I also scanned a 4x6 inch print of that negative. For more information on how I did this or to download the full file, go to the Scan Samples section. I cropped the shoulder area for comparison. The fine detail of Steller's jay feathers is not captured too well in the negative scan. Despite some tweaking with Photoshop for the negative scan, the raw print scan is still superior!
|
|
|
| Raw Print Scan | Raw Negative Scan | Sharpened Negative Scan |
Keep in mind, however, that obtaining good print scans depends very much on the ability of your photo lab, or you, if you do your own darkroom work, to produce good prints. Furthermore, even if the resolution of the print scan is better, the other aspects of the picture, such as color and contrast, may not be. For my tastes, the print was made a little too dark and contrasty. Nothing that I do in my image editor can regain any of the shadow detail that was lost in the print. The negative scan, on the other hand, retains more shadow details.
Poor Color Accuracy but Good Color Adjustment (TWAIN 5.0A)
This section title seems contradictory, but it does describe this scanner well, as I will explain. The default settings of the Epson Twain driver produce very saturated scans with moderate color casts. For some purposes, this is not bad at all, because the tendency is toward warm tones. The added warmth and saturation can improve dull pictures and make them "glow" a bit. For someone who wants to represent their photos accurately, however, this means lots of time spent on post-scan color correction.
Using the default settings and autoexposure mode of the Twain driver, I scanned a Kodak Q-60 Test Target. The color casts are readily apparent in the gray tones. The Eyedropper tool in Photoshop shows that neutral gray areas have an excess of red (about 12%) and blue (about 5%). Although the color casts are hidden somewhat in the color swatches, they are very visible in the woman's face, which appears very red.
Despite the poor color accuracy of this scanner right out of the box, Epson provides in their Twain driver the tools necessary for bringing scans back to neutral. Specifically, there is a Tone Correction tool for adjusting the global and individual RGB curves. In addition, a Color Adjustment tool can change the saturation and gray balance. With some trial and error, I was able to create some curves that made most of the grays neutral. RGB values were no more than about 1% different from each other.
The Autoexposure button, used in conjunction with a custom-made curve, adjusts the tonal range of the scan. The automatic settings tend to clip the highlights a bit, giving the true white and the next lightest swatches 255,255,255 values. This is no big problem, however, because the Image Control tool can change the gamma and black and white points, and expand or contract the tonal range. Although a histogram tool would have been nice, its absence doesn't necessarily prevent good results, as long as the user is willing to experiment. Despite some trial and error, I wasn't able to achieve 0,0,0 values for the true black swatch. This was probably due to random CCD noise. Finally, when I was able to come up with settings that satisfied me, I could save all of them to a profile in the Settings tool. The crop box size and position are even saved. Once I load up a setting, I can drag the crop box around and do a final scan without needing to zoom into individual frames.
I have found that once I came up with a scanning profile based on the Kodak Q-60 target slide, all my subsequent scans of various film types needed very little color correction. So, although the out-of-the-box color accuracy of this scanner is poor, very accurate scans are still possible if some effort is dedicated to calibrating the scanner.
Updated TWAIN Drivers Deliver Better Results
Twain 5.53A to be discussed.
Dynamic Range and CCD Noise Shown by the Kodak Q-60 Test Target
To try to assess the dynamic range of this scanner, I looked at my corrected Kodak Q-60 Test Target slide scan. I cropped the grayscale swatches and looked at its histogram.
The frequency values for luminosity show that the first 16 swatches, going from true white towards black, are well separated have relatively little CCD noise. This scanner is fairly accurate in representing each swatch as a single gray value, with a small amount of deviation. From about swatch 17 to 20, the pixel values begin to spread and deviate much more from the ideal value. The last few darkest swatches, 21 to 24, are barely separated from each other.
A visual inspection, based on an inversion of the grayscale (swatches 15 to 24), of the shadow areas shows that the swatches are represented as distinct, despite the high CCD noise. I also increased the brightness and contrast of the grayscale. The CCD noise is evident. There is also some line artifacts in the 23rd swatch. Not surprisingly then, the shadow area to the right of the woman's face, when brightened and contrast-increased, shows a large amount of CCD noise. A casual inspection of the individual RGB channels for the whole slide show that the green and blue channels contain much of the CCD noise. The clarity of shadow areas in scans made by the Epson Perfection 1640SU Photo is not too great. It may be useful, though, to compare the performance of this scanner against other (film) scanners that have been evaluated in a similar fashion.
Practical Dynamic Range Challenged by Darks
The 35 mm slide that I chose for my Scan Samples was actually not too challenging, but rather an ideal image for the Epson Perfection 1640SU Photo scanner. The slide has lots of medium tones. Everything is just about bright enough for the details to come through. Only a small amount of tweaking in Photoshop can turn out a great image.
On the other hand, the Epson Perfection 1640SU Photo scanner does not handle dark areas too well. This was already demonstrated by examining a scan of the Kodak Q-60 Test Target slide. For a more real-world test, I scanned, at 1600 dpi, a 35 mm slide of a rather high contrast scene of snow, forest, and mountain. This image was taken with a Canon 24 mm FD lens, at f/22, and on Fuji Provia 100F RDP III film. The aesthetics of the picture is not too good, but the slide presents truly challenging content for scanning. Specifically, the middle area of the slide consists of a dark colored forest, with dark green pine trees and brown leafless oaks. I cropped the area around the mountain peak for comparison. I reduced the size for presentation on this page. You can view the forest detail image crops at 100% size.
|
|
| Raw Slide Scan | Sharpened Slide Scan |
The Epson Perfection 1640SU Photo scanner renders the dark forest detail as a murky mess. It is possible to clean up the scan a bit and regain some detail with some levels adjustment and unsharp masking. The results will never be magic, but could be considered pretty good for a flatbed scanner.
Dynamic Range: A Second Look Reveals a Capable Scanner (Update Dec 2002)
I wrote this scanner review some short time after I bought the scanner in November 2000. Since then, I've gotten better at scanning, color correction, and understanding film. At this point, I don't believe that the above slide scan shows anything meaningful about dynamic range. The crop displays too much of the tree areas (which aren't really that dark at all) and too little of the dense areas between trees. Resolution is also a confounding factor. The Epson Perfection 1640SU Photo scanner isn't capable of resolving every single pine needle, and even using a 15X loupe, I'd be hard pressed to pick out any detail in the trees.
So, I found a slide with significant shadows and density that contained detail that would be visible under a loupe and resolvable using the scanner. This view of Stonewall Peak was taken using a tripod, a Canon 50 mm f/3.5 macro FD lens at f/22, and Fuji Provia 100F RDP III film. The image was scanned at 1600 dpi, color corrected, and the black and white points set. No sharpening or gamma (mid-point) correction was done. You can view the shadow detail image crops at 100% size.
|
|
| Raw Slide Scan | Levels-adjusted Slide Scan |
|
|
| Raw Slide Scan | Levels-adjusted Slide Scan |
The results are suprisingly good. Noise is present, but at an acceptable level. The amount of detail revealed is good. A little image-editing wizardry can make decent results.
Miscellaneous Comments: Somewhat Clumsy Handling
Epson does not provide any way for the transparency adapter to attach to the scanner, the way the normal lid does. The adapter just rests on the glass. Sometimes this presents itself as a pain because lifting the light lid sometimes causes the whole adapter to shift out of place. If you are too rough, there is a risk of scratching the film that you have nested in the adapter. The film carriers are ok. Some people consider them flimsy. They are, but I don't have a huge gripe with them.
Download High Resolution Scan Samples (*.tif files)
I provide a few 1600 dpi scans of 35 mm slide and negative film for those people who wish to see what this scanner can do. I have written my own evaluation of the scanner, but you don't have to take my word for it. Tweak the image files in your image editor and print them out to determine if the scanner meets your standards.
I didn't put any text onto my images so that you can view the whole, unobstructed scan. Therefore please respect my image copyrights. By downloading these images, you agree not to use the images for profit or commercial gain. You may not distribute or display these images or any derivative form of them without written permission. Thanks!
Resolution Test (3.2 MB).
Film: Kodak T-MAX 100
Lens: Canon EF 50 mm f/1.8 II
Settings: f/11, 1.5 sec
Methods: Tripod, mirror lock-up, remote release
Light: Tungsten (common household bulbs)
Scan: Twain 5.0A, 1600 dpi, 14-bit black and white mode, image size 100%, unsharp mask off
Notes: USAF charts and instructions from http://wdn.com/~johnchap/lenstest/testlens.htm. Thank you to the site author for providing the charts.
Slide scan (17.7 MB).
Film: Fuji Sensia II
Lens: Canon EF 50 mm f/1.8 II
Settings: f/11
Methods: Hand-held
Light: Daylight
Scan: Twain 5.0A, 1600 dpi, 42-bit color mode, image size 100%, unsharp mask off
Negative scan (20.1 MB).
Film: Fuji Reala 100
Lens: Canon EF 50 mm f/1.8 II
Settings: f/8
Methods: Hand-held, fill flash
Light: Daylight
Scan: Twain 5.0A, 1600 dpi, 42-bit color mode, image size 100%, unsharp mask off
Notes: The dust that you see on the Steller's jay's head and neck is not an artifact of scanning. It is actually feather dander, as this picture was taken during a change in feather molt.
Print scan (20.5 MB).
Film: Fuji Reala 100
Lens: Canon EF 50 mm f/1.8 II
Settings: f/8
Methods: Hand-held, fill flash
Light: Daylight
Scan: Twain 5.0A, 1600 dpi, 42-bit color mode, image size 24% to match file size for negative scan, unsharp mask off
Notes: The dust that you see on the Steller's jay's head and neck is not an artifact of scanning. It is actually feather dander, as this picture was taken during a change in feather molt.