On occasion I read in messages here references to scanners' actual resolution. I'm wondering how you make these tests? I own an Epson 4990 and I've been curious about its actual max resolution, so I decided to see if I could figure out a way to test it myself.
To test the scanner, I thought of the 1951 USAF test chart, and after hours of searching the 'net for a suitable example, I finally located a pdf of it. Actually, the pdf is not an actual copy of the 1951 USAF chart, but it's pretty close. I found it here:
After wasting a bunch more time trying to get the .prn translated to a file my image processing software could read without prompting me for dpi parameters, I finally just gave up and printed it out with my Epson Workforce 600. I also printed it out on my HP Laser Jet 1200. Both have 1200 dpi output, and as it turns out, the Epson does a slightly better job. But it wasn't nearly good enough. The printer output fell apart long before the scanner's resolution did. So now I know at least that my 4990's claimed 4800 dpi is way better than 1200 dpi.
It occurred to me then that I wouldn't necessarily have to be able to read the smallest area of the chart that was supposed to be there, but that I could make an accurate estimate based on the smallest feature I could see in the scanned page, and measure that against one of the groups in the chart.
Then a little light went off in my head, and I realized I didn't need the chart at all. The image processing software I use -- Paint Shop Pro X2 -- has an image scale that can be toggled on, which gives a readout in pixels. I'm sure PS has this same ability. So, who needs any sort of test chart, I ask? So, I loaded a 4800 dpi scan of a fine-grained slide into PSPX2 and had a look.
But then things become very subjective to interpretation, I'm finding. I'm looking at this scan, which is of a Kodachrome 64 slide, and with only a bit of contrast and color adjustment done in the way of post processing. In the scan, the 4990 is resolving grain particles, or textures in the pigment, whatever you want to call them, that are 3 to 5 pixels in size. So, resolving an object of 3 pixels in a 4800 dpi scan would tend to make the argument that the effective resolving power is about 1600 dpi. But if this is grain, then, it would seem that the scanner can do even a bit better. And it's pretty clear that it does much better, which is obvious by looking at 100% scans of the 1200 dpi USAF test target, which is only 400 dpi away from the test target scans. As it turns out, at much higher levels of magnification, there are other objects in the image in which a continuum of coloration is present, and the coloration changes from pixel to pixel in an even, non-random manner. This strongly argues for a full 4800 dpi to me. What I seem to be encountering is a few things: resolution limits in the film caused by grain, resolution limits of the image caused by the sharpness of the camera lens, sharpness limits caused by the function of the photographic gear (taking into mind camera shake and the like), and image characteristics that are not as dependent upon resolution limits as they are upon image density and color rendering accuracy, and finally somewhere in all this, resolution limits of the scanner itself. But perhaps even more important from a practical standpoint than resolution limits is resolution detail. It just seems to me my 4990 should be able to render sharper images than it does.
I also have scans of this same slide done by an HP S20 (2400 dpi dedicated film scanner) and an Opteka slide duplicator used with my 10.1mp DSLR (effective 2400 dpi resolution). Both of these images compare favorably to the 4990 scans until one begins to enlarge them. The two lower resolution images quickly begin to pixelate before image detail is exhausted. With the 4990 scan, it's just the reverse. Image detail tends to become exhausted before pixelation occurs.
These are at very high magnifications, though. Way above 100%. The reason why I have scans of this same image done by all three devices is for a comparison I did. At 100% image size, a slight nod can be given to the Opteka/DSLR combo in terms of sharpness, while the HP S20 and the Epson 4990 are practically dead even when looking at the S20's scan at 100% and the Epson's at 50% (because it's 4800 dpi vs. the S20's 2400 dpi). You can read it here:
So, in a way, there's not much difference to be offered between 2400 dpi scanned output and the 4990's 4800 dpi output. But to me that's only a part of the picture. Sure there isn't much difference in many instances. But when one wishes to enlarge their image to a larger size than the others can support, then a usable 4800 dpi shines. And the 4990's is usable, IMO. I do wish it was a bit sharper though.
So, while I still can't point at a hard number with much confidence, and say, "Well, my scanner's effective resolution is yadee yadee dpi," I can at least say, hey, it's actually better than I expected.
To test the scanner, I thought of the 1951 USAF test chart, and after hours of searching the 'net for a suitable example, I finally located a pdf of it. Actually, the pdf is not an actual copy of the 1951 USAF chart, but it's pretty close. I found it here:
The 1951 USAF test chart will indeed work for testing the effective resolution of scanners. However, you need a very good target and these are not inexpensive. Edmund Scientific sells chrome on glass 1951 USAF targets in a couple of sizes, the least expensive of which is about $150.
Since the Epson 4990 does not focus you will also have to establish the point of best focus by scanning the target at difference distances from the base glass, one the glass, at 0.5mm over the glass, at 1.0mm over the glass, at 2.0mm over the glass, etc. There is a guide that comes with the target that gives you a direct reading of lp/mm based on which element/group your scanner can capture.
I have tested my own Epson 4990 and a couple of other high end scanners that I own. The maximum resolution that I get from the 4990 is about 35-40 lp/mm, which is about 1800-2000 ppi. I found very little increase in actual resolution scanning at 4800 dpi instead of 2400 dpi.
By contrast, my Leafscan 45 (vintage 1994) tests with a real resolution of about 4500 dpi when scanning at 5080 dpi, and my Eversmart Pro (vintage 1997) tests at 3000 dpi when scanning at 3175 dpi.
Yeah, I found plenty of for sale listings for this test chart, ranging from $300 to $700. Didn't see the $150 at Edmund Scientific, though. Doesn't matter. I've got a lot of more important things to spend $150 on than a test chart.
Actually, I have done tests something like you mention by removing a slide from its mount and then using scraps of negative film as shims. I didn't record my results anywhere than I recall, but as I do recall, there was some depth of field and the sharpest results were just about exactly where they would be if a neg was mounted in its holder or if a slide was sitting flat on the glass. No surprise there, I guess.
1800 ppi still seems to be on the low side to me, though. Especially since I can visually resolve detail at significantly greater magnifications than what would be indicated by 1800 ppi.
Yeah, I found plenty of for sale listings for this test chart, ranging from $300 to $700. Didn't see the $150 at Edmund Scientific, though. Doesn't matter. I've got a lot of more important things to spend $150 on than a test chart.
Actually, I have done tests something like you mention by removing a slide from its mount and then using scraps of negative film as shims. I didn't record my results anywhere than I recall, but as I do recall, there was some depth of field and the sharpest results were just about exactly where they would be if a neg was mounted in its holder or if a slide was sitting flat on the glass. No surprise there, I guess.
1800 ppi still seems to be on the low side to me, though. Especially since I can visually resolve detail at significantly greater magnifications than what would be indicated by 1800 ppi.
The purpose of using standard targets is for all of us to get objective results and be on the same page. The targets permit testing of resolution up to 225 lp/mm, well beyond what most scanners are capable of resolving. And the results are what they are -- they don't lie, even though there is some element of subjectivity in determining line discrimination, and most scanners generally do better in the direction of the movement of the CCD than in the other direction.
Over the years I have owned several Epson flatbed scanners, including the 2540, 4870 and the 4990. I tested all of them with the resolution target and none of them gave over 2000 ppi in real resolution. I have also tested the V700 and the V750 and these scanners are incrementally better in that they test at about 2200-2400 ppi.
There is in fact a lot of depth of field with the lenses of the Epson flatbed scanners, which is one of the problems since they never focus exactly anywhere.
I am attaching three files of tests that I made with my own scanners to determine how much effective resolution they are giving. One is with the Epson 4990 scanning at 4800 ppi, another with the Leafscan 45 scanning at 5080 ppi and the third is with the Eversmart Pro scanning at 3175 spi. I have done these tests several times, always with about the same result.
Sandy ... that eversmart does quite a good job, that'd even be usable for 35mm
Cooltouch ... worth keeping in mind also is that if you are using 4x5 (or even 120 in a folder) you may not actually get even that much available on the film. I count on getting 2000dpi from my Epsons. Printing at 180 dpi for big enlargements is a x10 enlargement which makes 50 inch wide prints. Well within what I want 90% of the time. And yes ... I find that the prints do present quite sufficiently at 180dpi ... and no I don't put my face 15cm from them as much as I do 5x7 inch prints ;-)
Thanks for the responses, guys. I value your views and appreciate your doing a bit of pedagogy with respect to this subject. Sometimes I'm concerned that, as a layman, I'm probably misusing terms that I'm not even aware of misusing, and at other times, I'm searching for a term whose name I don't know, and whose properties I may be only partially aware of.
Sandy, I appreciate you attaching your test charts. Even though my scanned printouts fall apart by group 4, they seem to agree more or less with what you've achieved. By that I mean this: with your image, blurriness has set in by group 4, and I'm just beginning to see signs of this blurriness at group four in my scanned printout.
As for determining resolution, it appears that you stop counting as soon as you can no longer discriminate line pairs in either a horizontal or a vertical direction. It certainly appears that the 4990 has a lower resolving power in the horizontal direction than the vertical direction. So using that criterion, I get 1828 ppi. If I'm allowed to count line pairs vertically, however, I can read all the way to 5.6, which results in a resolution of 2896 ppi.
The way I see it, this higher resolution capability should not be completely ignored. Also, the property I mentioned in my first post, namely that of an even gradation of coloring at max optical resolution, should not be ignored either. Here's an example of what I mean. The following images are of the slide I mention in my first post, scanned by the Epson 4990 at an indicated 4800 dpi. Note what appears to be a large spec of dust on the slide above the roof, left of center. That's actually a seagull.
Here is a 1000% enlargement of the seagull. (The easiest way to display an image this size was to do a printscreen):
Perhaps I should have increased the magnification even further to more clearly illustrate my earlier point, namely that in the highlighed area of the seagull's head and upper torso, there is a smooth pixel-by-pixel gradation of color. Despite the fact that PSPX2 forced me to convert from 16 bits per pixel to 8, the gradation remains. This isn't an isolated incidence, either. I can show other areas from this same slide which indicate the same sort of smooth color gradation.
So, I ask myself, is this a function of the Epson Scan software? Maybe, I suppose. But I'm not using any interpolation that I know of. I realize that this smooth gradation does not translate to the bottom line when resolution-based sharpness is being measured, but it does seem that it would have an impact in overall image quality nonetheless.
Hey Pellicle, I'va actually been very happy with the quality of my medium format scans with my 4990. Even my old 3170 did a great job with medium format. And typically I scan these images at 2400 ppi indicated. And yes, I am aware that large format optics generally have considerably lower resolving power than 35mm ones.
I haven't yet had the opportunity to do any large format scans. I have a few 8x10 slides that I'd like to try, but I'd have to make a frame for 8x10 in order to do so. Still, I'm sure that, if I like the way medium format came out, I should really like 4x5. I know a couple of guys who shoot large format and who use their Epson 4990s as part of their digital darkroom. And they love it.
Hey Pellicle, I'va actually been very happy with the quality of my medium format scans with my 4990
....
Still, I'm sure that, if I like the way medium format came out, I should really like 4x5. I know a couple of guys who shoot large format and who use their Epson 4990s as part of their digital darkroom. And they love it.
well I'm one of the people who use large format and like my 4990
I hope I didn't give the impression that I think that the 4990 is not a good scanner for MF ... personally I think its a good scanner for any film up to about x10 enlargement
I've even used mine for 35mm and have an A4 print of this image
My self (and other people also have said) that the foreground pot looks so good its like looking at it ... but this scan is not the best I've done from the slide (and why would I put that up for the image theft)
I hope I didn't give the impression that I think that the 4990 is not a good scanner for MF ... personally I think its a good scanner for any film up to about x10 enlargement
Oh, I wasn't thinking that at all -- I was just sort of adding to what you mentioned earlier, not having any direct experience yet with scanning large format.
As for determining resolution, it appears that you stop counting as soon as you can no longer discriminate line pairs in either a horizontal or a vertical direction. It certainly appears that the 4990 has a lower resolving power in the horizontal direction than the vertical direction. So using that criterion, I get 1828 ppi. If I'm allowed to count line pairs vertically, however, I can read all the way to 5.6, which results in a resolution of 2896 ppi.
The way I see it, this higher resolution capability should not be completely ignored.
Most CCD scanners have higher resolution in dpi in the direction of the movement of the sensor. The 4990 has stated resolution of 4800 X 9600, my Eversmart Pro has resolution of 3175 X 8200.
Most people count the higher number as interpolated resolution so I base my resolution on the ability to discriminate both the horizontal and vertical lines. I think it is fine if someone wants to count only the highest or lowest lines so long as that is stated up front.
I certainly don't discount the higher figure. In fact, I often look at the fine detail of a negative before scanning it to figure out if it might be possible to take advantage of the higher resolution by scanning in a certain orientation.
Sandy ... that eversmart does quite a good job, that'd even be usable for 35mm
One thing to keep in mind is that scanners like the Leafscan 45 and Eversmart Pro are extremely well built and cost between $20,000 - $40,000 USD new. In fact, the lenses on these scanners cost more new today than an Epson V750. For example, the lens on the Leafscan 45 is a 75mm f/4 Rodenstock Apo Rodagon D. That lens costs around $700 USD new from B&H in New York today, and lists for a lot more.
LinkBack URL
About LinkBacks
Adv Reply
Originally Posted by cooltouch 
