|A word about our scanner testing philosophy: Some publications have taken the position of scanning everything using the scanner's default settings, believing this to be most fair, neutral methodology. The problem with this approach is it may show unacceptable results for an otherwise perfectly usable scanner. (Most users are willing to engage in some tweaking of the scanning parameters to get the best result.) For our part, we believe the most accurate representation of real-world performance is to allow for a reasonable level of twiddling of the scan parameters. In the interest of objectivity though, we also show scans performed with default settings, to provide a completely neutral reference point. |
Also note that all images here have been JPEG compressed for compatibility with 'web browsers. This will degrade image quality somewhat, but we used a very conservative compression setting ("9" in Photoshop) to minimize this.
| ||"Musicians II" image: (272k) The main image here was scanned at 1098 x 765 pixels (800 dpi), and minor tonal adjustments were made, using the tone curve controls. (Mainly pulling a little blue out of the image, since the auto adjust left quite a bit of that color in there.) Here (235k) is a version scanned with the Dimage Scan Multi Pro's default settings, which shows somewhat heavy midtones and a rather warm cast. This shot (256k) shows the effect of the scanner's auto-adjust option, which cleaned and brightened things, but left a way too much blue in the image. Even the unaltered image shows surprisingly good color accuracy, tonal range, and saturation though. |
(NOTE that this is NOT the identical "Musicians" image as used in our digital cameras test! It's very similar, but the models are different, and the digital-camera version is a couple of reproduction generations removed from this particular version.)
| ||"Musicians II" detail clip: (260k) Wow - That's a lot of detail... The Dimage Scan Multi Pro has the highest optical resolution (4800 dpi) of any scanner we've tested to date (September, 2001). There's more to resolution than just dpi though, since the scanner optics and even the light source affect how much detail is ultimately captured. |
It does appear though, that the other elements of the Dimage Scan Multi Pro (DSMP from here on, to save space) are up to the task. This maximum-resolution scan of the Musicians slide revealed exceptional detail, right down to the film grain. Truthfully, this slide may not be the best test of detail, as it's most likely a dupe from the original. Looking at this image, we felt that the resolution of the slide itself was limiting the performance of the scanner somewhat, but the bottom line is there's an enormous amount of detail in this image. The DSMP's light source appears to be highly collimated, which helps it pick up very fine structures on the film, but by the same token is very unforgiving when it comes to film grain and surface dirt.
Speaking of surface dirt, our Musicians slide is getting a little grubby from several years use now, so you can see some fairly fine-grained dirt on the image here (612k). Compare this image (607k), scanned with the Digital ICE defect-removal feature engaged. The dirt and a few minor scratches completely disappear, with remarkably little cost in sharpness. (In fact, there's so little loss of sharpness we're hard pressed to find it.) This is an excellent illustration of Digital ICE in a more practical example than with our heavily-damaged negative film used in the main review. This level of dirt is pretty typical of what you'd find when dealing with older film in a production environment. If this were a job for pay, Digital ICE would have saved us a good 30 minutes or more of careful spotting in Photoshop.
| ||Kodak Royal Gold 25 "House" detail clip: (855k) This is a detail clip from the same negative used to produce the original "house" poster for our digital camera tests. (Now superseded by one shot on 4x5 transparency film.) It was shot on Kodak Royal Gold 25 film (sadly, no longer manufactured), which is extremely fine-grained, but which has very different color characteristics from most normal color negative films. Most scanners we've worked with have difficulty with RG 25's color balance, and the DSMP fell prey to this. The default scan (431k) was quite washed out and flat, although the hues were generally correct. A few tweaks in the Curves control panel extended the tonal range and cleaned up the shadows a bit, producing this scan (846k). |
Consistent with what we observed with the dirt on the Musicians slide, the DSMP showed every bit of the tiny chemical flecks or emulsion defects present on this negative. (Most scanners show tiny white specks across this image.) They were very prominent, to the point that the scan would really be unusable without considerable retouching or other processing. When we engaged Digital ICE though, they almost completely disappeared, producing the result you can see here (855k). - It appears that scanners with highly collimated light sources are much more susceptible to film defects and artifacts of this sort, but that Digital ICE is also much more effective with such scanners. The net result is that the Digital ICE-processed scans from scanners with either collimated or diffuse light sources end up about the same. (We do see a few artifacts from the slight scratches on the negative, most visible along the sloping gable roof, to right of center though, and there are very visible artifacts on the white vanes of the vent at the top of the gable, where Digital ICE substituted some of the nearby black for the white trim.)
Normally, we've found that Digital ICE softens the image noticeably, if only by a little bit. In the case of the Dimage Scan Multi Pro though, we were hard pressed to find any evidence of this. The DSMP software does have a very flexible unsharp masking function built in, which many publication-oriented users will doubtless find very useful. We tried it out in this scan (1137k), with the controls set to 100%, 1 pixel radius, 0 threshold, and "dark protection" to 10. The result is nice and crisp but certainly didn't appear to be needed to correct any sharpness loss from Digital ICE.
| ||"Train" Shot (Extreme shadow detail): (780k) This slide is an extraordinarily tough test of scanner dynamic range: The slide contains areas of moderately bright highlight, but the shadows are exceptionally dense. |
This is indeed a very tough piece of film, and the Dimage Scan Multi Pro handled it very well. Compared to other top scanners we've tested recently (the Nikon Super Coolscan 8000 ED is the obvious competitor), we found interesting similarities and differences. Both the DSMP and 8000 ED produced scans with very low noise, and in fact were only slightly improved by using the 16x sampling option. It's a very close call, but we felt that the 8000 ED just slightly edged the DSMP in noise performance, but the Dimage held its own, and in fact won out in two areas. First, the Nikon 8000 ED has a special "one line scan" mode that you need to use on scans requiring extreme tonal adjustments, to prevent banding. The DSMP had no such requirement, and the scans were remarkably clean and uniform in this respect. Second, we were surprised to see that some of what we'd interpreted as lens flare on the film itself may in fact be flare in the scanner optics: Looking around the front truck on the locomotive, there are areas where brightly sunlit earth is juxtaposed to the pitch black of the underside of the locomotive. There's very visible flare in the Super Coolscan's image in these areas, and also in the shadows to the right of the front truck, obscuring details in the rails and ties of the train track. Looking at the same area on the scan from the Dimage Scan Multi Pro, we see much lower flare, and the details in the train track are much more visible. This looks to us like a significant advantage for the DSMP. Judge for your self though, here are links to the best-case scans (with some subsequent Photoshop adjustment) from both scanners, the Dimage Scan Multi Pro (780k), and the Super Coolscan 8000 ED (492k).
One area where the DSMP was a bit weaker though, was in the ability to complete correct for color casts in the extreme shadows, using the scanner software alone. Try as we might, we couldn't avoid a reddish color cast in the shadow areas of the slide. On the other hand, for images like this, your best bet will almost always be to scan at 16 bits/channel, and plan on doing the finer tonal adjustments in Photoshop.
Phew, lots of variations here, see the table below for all the links... First, we have the default scan, which came out quite dark (no surprise there). Next, we tweaked the tonal adjustments and did an 8-bit scan. This wasn't at all bad, but had a pronounced reddish cast we couldn't completely eliminate with the scanner software. A little work in Photoshop with the levels dialog cleaned it up pretty nicely, as seen here. We then tried a 16-bit scan, which showed the same reddish cast, and likewise cleaned it up in Photoshop, as you can see here. (Note that these images have all been converted down to 8 bits after the scan, so they could be displayed here as JPEG images.) We then experimented a bit with the multi-sample scanning, using the 16x option. Given the very low noise levels in the single-pass scans, we weren't surprised to find that there was only fairly modest improvement with the multi-sample scanning. (Although we did observe that the blue-channel noise was noticeably reduced.) The table below contains links to the various files we created from this target:
| || |
"New Train" Shot (Extreme shadow detail): (997k) As the name suggests, this is a new "train" shot. We made it because the old shot was a one-of-a-kind slide, meaning we'd be stranded if it ever got lost or damaged. We shot a full roll of photos (with exposure bracketing) of another locomotive, so we'd have spares for the future. We'll gradually transition over to this new slide, but in the near term, we'll scan both targets for backward compatibility with our previous reviews. Overall, we found very similar results with this slide, very low noise, but difficult removing the reddish cast, using the scanner controls alone Here's a table with some of the same shots as above:
| ||Q60 Color Target: (298k) Kodak's "Q60" color target (formally adopted by the ISO as part of the IT8 color standard) is a good test of color accuracy and tonal rendition. We first scanned this target with the scanner's default settings, with the "Auto Expose for Slides" option enabled. The result was this scan (322k), with a somewhat reddish cast. (For some reason, the DSMP seems to really like red in its images.) Despite the red bias, the overall default scan was very good, much better than we're accustomed to seeing. A few tweaks of the tone-adjust controls though, gave this (298k) nicely-balanced scan, with very good color. |
Some folks on the internet have settled on using a crop of the woman's face in the upper righthand corner of this slide as a reference for detail and resolution. To help with people making comparisons with scanners we haven't reviewed yet, we offer this crop (485k) of that area, captured by the Dimage Scan Multi Pro at its maximum 4800 dpi resolution. (Consistent with our other results, really excellent detail here.)
| ||"Davebox" test target: (381k) This is our official "weirdness of color negative film" test target. The Dimage Scan Multi Pro's default settings (345k) didn't do too badly, although the image did show the customary washed-out look we've seen from most scanners with this target. Surprisingly, the auto adjust had little effect, producing this shot (357k), also rather washed-out. A few tweaks of the tonal adjustments produced a very nice scan though, as seen here (381k).|
Black/White Negative Target: (220k) I've had a lot of requests to look at black/white scans with the scanners I test, and I finally got around to shooting some Tri-X to play with. What I found with the Dimage Scan Multi Pro was illuminating, and agreed with what I've heard from casual emails with others in the digital photo community: To get the best results from a black/white negative scan, scan it as a positive, possibly even as a color transparency, then invert the results in Photoshop or another imaging program. The film I scanned here is rather "thin", with not a lot of density to it, especially (!) in the dark foliage. I scanned it three ways with the Dimage Scan Multi Pro: First as a b/w negative, then as a b/w positive, finally as a color transparency. - The latter two I inverted in Photoshop to produce black/white positives. The results were interesting, to say the least. The table below holds cropped samples of these scans. (Click on any image to see a full-sized version.)
The two positive-scanned images are fairly similar: Perhaps just a bit better tonality and detail in the dark shadow areas, but it's close to a toss-up. The sample scanned as a negative has a much harsher tonal curve though, and loses a lot of detail in the deep shadow areas. What's significant here is that there appeared to be no setting in the scanning software that allowed me to recover the shadow detail when scanning in negative mode: The darkest areas of the image were pushed all the way to black, no matter what. (So it isn't just a matter of my having chosen different tone adjustments for the two positive-scanned versions.)
As noted above, I've heard in casual emails that film scanners often do a poor job with black & white negatives, but this was the first time I'd managed to check it out for myself. The good news though, is that scanning the b/w negs as positives brought out significant detail in portions of the negative that appeared to the naked eye to hold very little image information.
| ||WG-18 Resolution Target Horizontal Clip: (35mm, 4800dpi) (native) (192k) |
WG-18 Resolution Target Horizontal Clip: (6x7 cm, 4800dpi) (interpolated) (324k)
WG-18 Resolution Target Horizontal Clip: (6x7 cm, 3200dpi) (native) (151k)
WG-18 Resolution Target Horizontal Clip: (35mm, scanned with 6x7cm lens, 3200dpi) (native) (56k)
Whooee! While it's hard for longtime 35mm stalwarts like ourselves to admit it, medium-format film certainly captures a LOT more detail, as evidenced by the Dimage Scan Multi's maximum-resolution results on this shot!
The full WG-18 resolution target is very large (see below), so we cropped-out these clips to show the scanner resolution on this familiar target. There are two different targets here. The 35mm one was shot on Kodak Technical Pan black & white negative film. This film is extremely fine-grained, with perhaps the highest resolution of any commercially-available 35mm emulsion. The target was shot with a Nikon 50mm, f1.4 lens (a notably sharp lens), at an aperture of f8. Thus, while not a "laboratory" grade target, this represents about as much detail as you'll ever see in a conventional 35mm film image. The downside of this target is that the Tech Pan emulsion is a little "thin," lacking density. It is thus difficult to set scanners properly to produce adequate contrast to separate the finest details without losing critical information.
The 6x7 target was shot on Fuji Velvia transparency film, a very fine-grained color emulsion. We didn't record full details on that one, but do recall that it was shot with a Mamiya RZ-67, using a Mamiya lens of around 200mm focal length. (250mm?) It was also stopped-down a fair bit, to hopefully be within the lens' optimum aperture range. We're less familiar with medium-format lenses, so can't say how this compares to the ultimate attainable with a 6x7 camera. The amount of detail captured, relative to what we're used to seeing in 35mm scans is certainly impressive, however.
The Dimage Scan Multi Pro did a really exceptional job with these targets, producing incredibly crisp, sharply focused scans. Resolution is incredible, with detail easily going beyond the 2000 lines per picture height limit of the targets. By comparison, most 4000 dpi scanners we've tested run out of steam somewhere between 1800 and 2000 lines.
After we posted the first version of this review, several sharp-eyed readers emailed to point out that the Dimage Scan Multi Pro doesn't really scan medium-format film at 4800 dpi, but rather interpolates up from the 3200 dpi that represents the raw sensor resolution. Embarassed by our gaffe (we completely missed this in the DSMP's docs), we re-scanned the res target at 3200. Also, to get a better measure of the actual resolution at 3200 dpi, we mounted our 35mm target in the 6x7 film carrier and scanned it. The ratio between the narrow dimensions of the 6x7 frame and 35mm frame result in an effective multiplier of 2.28 for the numbers on the resolution target when scanned in this way. (The numbers represent lines per picture height, a measure that related resolution back to the captured frame, rather than the film area being scanned.)
Scanning at the 3200 dpi resolution setting, the DSMP still extracts a remarkable amount of detail from medium-format film. Looking at our 35mm target scanned in this mode, the target lines remain distinct from each other (albeit just barely) all the way out to the 2000 line/picture height limit of the test. This translates into about 4570 lines across the 6cm dimension of medium-format film. (We're going to try to get this 35mm target scanned by other medium-format scanners as well, so we can report back the results for those models as well. Stay tuned...) (But don't hold your breath, we have to get the target shipped around, and get people to scan it for us...)
Another side note here: We scanned this target as a color slide, then inverted the scan to get a positive image. We felt that the purely monochrome "b/w negative" scanning option lost some of the tonality of the image. In the past, we've often seen scanners have trouble with this target, producing color artifacts at the highest spatial frequencies when we scanned it in color vs b/w mode. No such problem with the Dimage Scan Multi Pro!
| ||WG-18 Resolution Target Vertical Clip: (35mm, 4800 dpi) (native) (155k) |
WG-18 Resolution Target Vertical Clip: (6x7 cm, 4800dpi) (interpolated) (260k)
WG-18 Resolution Target Horizontal Clip: (6x7 cm, 3200dpi) (native) (168k)
WG-18 Resolution Target Horizontal Clip: (35mm, scanned with 6x7cm lens, 3200dpi) (native) (58k)
Here's the corresponding vertically-oriented clip of the WG-18/Kodak Tech Pan target.Essentially identical to the horizontal clip above.
|Full-Size ISO-12233 ("WG-18") Resolution Target (35mm only): (3,324k!) For the real masochists, here's the full-size ISO-12233 target, scanned at the maximum resolution of 4800 dpi. (Duplicate scans at 3200 dpi in medium-format mode.) A side note: We didn't explicitly set up a test for frame coverage by scanners, but our ISO-12233 shot goes right to the edges of the 35mm frame, and we found that the Dimage Scan Multi Pro covers a bit more than a full frame horizontally, but just a bit less vertically. (Of course, if you wanted to get all the way out to the film sprockets on a 35mm frame, you could always just mount the film in the 6x9 glass holder, but you'd be limited to the 3200 dpi native resolution of the scanner in medium format mode.) |
Another quality note: The Dimage Scan Multi Pro produced very sharp images corner to corner here, with only the slightest softening at the edges. It also showed very low chromatic aberration in its optics, with only a couple of pixels of color around the target elements in the corners, decreasing steadily toward the center. (We did feel that we saw more "coma" in the medium format lens of the scanner, particularly toward the edges/corners of the frame.) Overall though, it looks like the DSMP has a great set of optics.
WARNING: This JPEG expands into a 91 megabyte file, which will almost certainly crash your browser if viewed directly! To view it, you must first download it directly to your hard drive (right-click in Windows, click & hold in Mac Netscape), then open it in an image-editing application. here (2654k) is the link to the RAW JPEG IMAGE. (No surrounding HTML file, 2.5 megabyte download.) (For the real masochists, here's a copy of the full-frame 6x7 res target scan, a 9.7 megabyte download scanned at 4800 dpi (interpolated), which expands into a 262(!) megabyte file. Finally, here's a link to the 6x7 target, scanned at the 3200 dpi native resolution of the scanner on medium-format film, and here's a link to the 35mm target, scanned at 3200 dpi when mounted in the 6x7 holder.)
| ||USAF 1951 Resolution Target: (252k) (Elderly technoids only ;-) Old-line lens and film testers will be well-familiar with the "USAF 1951" resolution test target. (1951 is the year it was created, giving you an idea of what we mean when we say "old-line".) This was scanned at the maximum 4800 dpi from a laboratory-grade target (chrome on glass slide) before being cropped down, and generally gives an excellent view of the scanner's ultimate capabilities. On this target, the Dimage Scan Multi Pro set a new record, for the first time showing us a cleanly resolved view of group 6, element 1, both horizontally and vertically, at 64 line cycles/mm (1626 line pairs/inch). The short extent of the USAF pattern targets doesn't permit the sort of visual interpolation our eyes do naturally on the more extended ISO-12233 pattern. As a result, the USAF target generally gives much more conservative resolution numbers.|