Epson Sharpness relative to very specific dpi setting

[newindustar]

<p>I have read so many of the posts claiming there is nothing else to be had above about 2400 dpi and I have always struggled with getting enough out of the V700 and 4870.</p>

<p>Of course I have done optimal height adjustment on the V700, which came out at max height on factory holders and at stock height on 4870.</p>

<p>I decided to test by 100 dpi increments all the way from 800 to 3200.</p>

<p>I found a very specific best sharpness at 1900 dpi with V700 with noticable fall off on either side such as 1800 or 1900. I also compared to the standard multiples of 2400, 3200 4800 and 6400 and the 12800. to make it even more of a challenge for the 1900 I upsampled it to all the higher reses listed in PS and 1900 still won. Of the standard multiples 2400 and 4800 were the best but softer than 1900. 3200, 6400 and 12800 were poor.</p>

<p>I used Silverfastt SE latest version, and Epsonscan latest version and the results still held. resampling was done in CS3 bicubic smoother for enlargements.</p>

<p>On the 4870, 2500 was the sharpest dpi and I followed all the protocols above. I used Epson Scan and Vuescan for the 4870.</p>

<p>The 4870 was close or a just bit sharper than the V700 at their relative sharpest DPIs.</p>

<p>So I have joined the camp of all those who say there is nothing above 2400, many say the low 2000s.</p>

<p>The good news is I was able to do better than the standard multiples for some reason.</p>

<p>Anyone care to take the 100 dpi test? I'd be fascinated to hear the results.</p>

<p>Any comments on my test results would be very interesting to hear as well.</p>

[peterbcarter]

<p>I have a Plustek 7200i and an Epson 4490 with Silverfast AI on both. I will do some testing when I get home tonight.</p>

[newindustar]

<p>I'd be very interested to see your results Peter.</p>

[dan_max1]

<p>What you did was very clever! You defined the scanner's resolution. What people do not realize is that the scanner's "resolution" above the real resolution, -let us say 1900 dpi, is used by marketers as a way of hyping the specs, by the stepping motor stopping many more times to take a sample. Now, stepping motors are not optical devices, and can not achieve higher resolution, this is why scannning above the native resolution did not improve results. One problem with the tests you did is that the film plane is uncertain due to film curvature and that all determinations of plane of focus take in film curvature into the calcuilation. That can be corrected by fluid scanning where you have a totally flat film plane.<br>

Dan Max<br>

</p>

[newindustar]

<p>Thnaks for the feedback Dan.</p>

<p>Well I carefully ran the test again on the V700, hours worth of work and got different results. It was a different image. The first set of tests were on two different images but the results held with Silverfast and EScan on both the 4870 and V700 so it was pretty convincing.</p>

<p>This second test was first run with the latest Vuescan demo, different image however. I did not get the blurring this time. I was able too hold my 1900 res sharpness results all the way up to 6400! 12800 was obviously interpolated. There was no loss of sharpness and no pixelation like I have seen with EScan at higher reses.</p>

<p>I was very surprised and thought ah, "Vuescan knows how to talk to this scanner". I thought I had arrived at the answer, it was software limitations!</p>

<p>Being the dedicated tester I am, I reran the test on the same image without disturbing the holder with Silverfast SE and EScan (latest versions) and both gave similar results to VueScan, i.e holding sharpness up to 6400.</p>

<p> I was confounded. Nothing worse than getting non-repeatable results after much work!</p>

<p>I need to be clear, this was not extracting more detail from the film as I went higher than 1900 but it was able to scan optically it appeared with out losing the sharpness and detail it had at 1900.</p>

<p>Actually 1900 was irrevalent the second go around. It was just a pretty linear increase in filesize with no increase in detail after about 2400 but no <em>loss</em> in detail all the way to 6400. I know it was scanning this in as I compared all the steps by resampling in Ps. This time the PS resamples were softer than the scans which was not the case the first time.</p>

<p><br /> <br />Dan I am at a loss to explain the difference in the tests, and they were done very carefully. This is where your comments and abouts flatness may enter in. This scanner has so little depth of field it seems likely the variation is coming in there. Maybe if a particular piece of film is right at the edge of the depth of field it becomes resolution sensitive. Iam running my at max height on factory holders.<br>

I don't know about the stepping motor resolution. The fact that I was able to use the higher reses successfully on at least one test may mean the scanner is capaple of higher res even if it cannot get more detail if this makes more sense. They are not necessarily the same issue.</p>

<p>It is possible the film often does not have much more to offer.</p>

<p>Detail was very consistent after about 2400 but it is very important to be able to scan higher res without artifacts and pixelation. This is a real film advantage digital folks like to gloss over.</p>

<p>This is the problem with Digital cameras, interpolation is the only way to get big. I resampled a D700 raw file from 35 meg tiff to 130 meg tiff (enough for a 20x30 at 288) and it was horrible even at iso 320, iso 3200 was disgusting. Even if there is not more real detail being able to scan bigger is still an advantage over software interpolation.</p>

<p>Now I'll have to rerun the test with 4870, not soon.</p>

<p>I also got the MultiPro working under Vuescan now so that 'll be interesting on the same frame the second V700 test was run on.</p>

[newindustar]

<p>I am going to add to this as I discover more.</p>

<p>I did a spot check on an image I was working on and found I was back to the sharpest around 2000 again. I found what I expected, the film on one end of the frame I was working on was not supported by the holder and it was bowed. My original finding was supported, resolution sensitivity is dependent on focus/depth of field.</p>

<p>Next I rigged up a glass holder and height adjusted it for optimal sharpness on the same image. The sharpness resolution dependency went away and not nid not lose sharpness all the way up to at least 6400, I forget about 9600.</p>

<p>I did find something interesting however. I had an image with a nice sharp alloy wheel in it with distinct bright to dark edges. I found the V700 began pixelating above 3800 on the diagonal. I could see there was more res in one direction than the other based on pixelation. It is much less obvious on less distinct areas. This was a better indicator of resolution and stepping motor than just sharpness. I don't know how repeatable this will be.</p>

<p>I did determine for myself the V700 needs a glass holder, the depth of field is just to shallow to avoid it.</p>

<p>The 4870 seems to produce good results without glass holders and I am able to get 4800 and maybe 6400 with no loss in sharpness or pixelation although 4800 seems optimal.</p>

[newindustar]

<p>Thanks for the feedback Les. Scanning posts seem to get less interest than they used to. I realize I put out some seemingly contradictory info here but on my last post I'd hoped to clarify. These flatbeds are often not very repeatable and all scanners are rather sensitive to the particular film and even frame.</p>

<p>There are a couple of particular points I feel are consistent however.</p>

<p>1. from my post;<em> "I need to be clear, this was not extracting more detail from the film as I went higher than 1900 but it was able to scan optically it appeared, with out losing the sharpness and detail it had at 1900".</em><br>

<em>It was obvious that at 100% viewing the scans higher than close to 2000 were blurring. I found the cause for this to be related to film flatness. If the film was in the end position thus held on three sides the sharpness was consistent up to 6400. The resolution sensitivity was only happening typically on frames near the unsupported ends. My V700 is very sensitive to flatness. My 4870 is less so and combined with the 4870 holders which I prefer over the V700s, the 4870 can be as sharp on average. I think the V700 has a tad more if the frame height is just right.<br /> </em></p>

<p><em>2. "I found the V700 began pixelating above 3800 on the diagonal. I could see there was more res in one direction than the other based on pixelation. It is much less obvious on less distinct areas. This was a better indicator of resolution and stepping motor than just sharpness. I don't know how repeatable this will be"</em><br>

<br /> On later tests the pixelation and stairstepping often is not evident up to 6400, and it requires a sharp line of distinct color distinction to see it anyway. Interaction with monitor resolution has to be wtched as well. For this reason I believe the scanner has the resolution required to not pixelate not interpolate up to 6400. There is a very fuzzy point regarding the difference in sharpness vs resolution. It seems to me these scanners may have the resolution while lacking sharpness. Normally they are considered one in the same by I am not so sure, just as a dedicated scanner which has the real resolution can be out of focus.</p>

<p><em>3. </em> On many frames the slight lack of sharpness is almost a blessing on negs for the the much prettier and smoother tonality provided. If sharp scanners could just deliver film grain sans noise.<em><br /> </em><br>

<em>4</em> . To concur with your point I agree the film often has no more info to deliver or it is obscured by the grain and noise. However if the scanner is actually blurring that is a problem.<br>

If one is looking at the grain structure amidst the noise, even if not very sharp, one can tell if there is more there or not. It is not as if these scanners cannot show grain. I actually have a better appreciation for these scanners after all this. I like what they do with color better than the couple of film scanners I have.</p>

<p>5. I have found scanning negs as HDR in Silverfast gives less noise and a little better sharpness after inversion and correction. Negafix seems to introduce a lot of noise at times.</p>

 

[newindustar]

<p>Regarding your examples I appreciate the difference of the film scanners over the flatbed. The problems with flatbeds is it not often possible to sharpen them adequately post with out raising up artifacts espicially ICE induced ones. Sommetimes on an absolutely optimal flatbed scans I prefer them over dedicated despite the softness. Your film scanners are better than what I have to work with however. I have a 2900 dpi Coolscan IV. That scanner came off pretty well in one of the scanner bakeoff and it is quite good. I often consider if I should trade it in for a 4000 or 5000 or a Minolta 5400 however. My other is a MultiPro not known for excellent neg scans.</p>

<p>I often consider selling off all my scanners and just getting a 9000. Any advice on this quandry would be appreciated.</p>

<p>How do you feel these days regarding film scanning ve the newer digicams?</p>

[newindustar]

<p>I just got another medium format, a Mamiya 645 Pro TL so am compelled to get something I like out of these scanners as well.</p>

<p>I just have Silverfast AI which has the option to scan HDR. Scanning HDR sort of tosses out the Negafix benefit but the scans do seem cleaner if haeder to balance post inversion. I really like the color fromm the flatbeds. Some very bad things happen with ICE from time to time however.</p>