Scanhancer: Color Cast

<p>I find that the Scanhancer is essential for Nikon LS-5000 and LS-4000 scans.</p>

<p>Why?</p>

<ol>

<li>It virtually eliminates 'pepper grain'... to bring it near the performance of the LS-9000</li>

<li>It allows one to use Anti-Newton glass to hold 35mm film flat, as long as you place the AN glass (etched side against film base) in the path between the LED light source and the film base (but not between the film and the CCD).</li>

</ol>

<p>The only problem?</p>

<p>The Scanhancer, at least the one I have, introduces a significant red color cast to all scans.</p>

<p>Let me illustrate:<br>

<img src="http://staff.washington.edu/rjsanyal/ScanningWoes/ColorCastIntroducedByScanhancer.jpg" alt="" width="800" /><br>

<a href="http://staff.washington.edu/rjsanyal/ScanningWoes/ColorCastIntroducedByScanhancer.jpg">Link to Full-Size Image</a></p>

<p>I know all about color management, so trust me on this one :)</p>

<p>Basically the above scans did not receive any treatment from Vuescan... it's about as RAW as an output as you can get from Vuescan. Raw RGB values were spit out, and so let me share with you the R,G,B values for Dmin (white) on the two scans:</p>

 

<ul>

<li>Without Scanhancer: Dmin = 249, 245, 242</li>

<li>With Scanhancer: Dmin = 246, 213, 194</li>

</ul>

<p>Clearly, something's going on with the Scanhancer. I have many more examples of scans, but let me post even more convincing evidence that the Scanhancer seems to be letting through, preferentially, red light:</p>

<p>Here's a shot of my halogen lamp bulb, taken with a little digital camera, with and without a Scanhancer placed in front of the lens:<br>

<img src="http://staff.washington.edu/rjsanyal/ScanningWoes/HalogenLampThruScanhancer.jpg" alt="" width="800" /><br>

<a href="http://staff.washington.edu/rjsanyal/ScanningWoes/HalogenLampThruScanhancer.jpg">Link to Full-Size Image</a></p>

<p>Consistent with my scanning results, the diffuser seems to have a heavy penchant to letting more red than green & blue light pass</p>

<p>Has anyone seen this with their Scanhancer?</p>

<p>The color cast is so bad that even proper profiling isn't entirely able to remove this cast.</p>

<p>I'm simultaneously going to contact Erik de G.</p>

<p>Thanks,<br>

Rishi</p>

<p>Dear Rishi,

<br />

Essentially your findings are correct. The Scanhancer is supposed to lineary diffuse blue frequencies most, while red frequences are nearly not diffused and IR frequencies aren't diffused at all. (Your test with the halogen lamp shows this clearly.) The idea behind this is to suppress the grain grittyness optimally since it is mostly caused in the blue and green channels. At the same time you want ICE dust removal to work fine, so diffusing the IR light of the scanner would not be wished for. It is a particular property of the Scanhancer material that it diffuses some frequences more than others. A normal diffuser would not do it. On a light box Scanhancer will appear neutral, while under collimated light it may seem to introduce a color cast. In reality it does not block any frequency, but since blue and green light are diffused more away from the receiving lens it appear as if more red light comes through. (Also see the introduction page of Scanhancer 5LE: http://scanhancer.com/index.php?art=15&men=15 ) If you read the information available on the Scanhancer website you would find that I suggest to used manual exposure settings of the scanner as to expose the blue channel longer than the red one and the green channel should be set somewhere in between. See the information on this page for example: http://scanhancer.com/index.php?art=42&men=3 . Especially check out the image of the exposure control of the Minolta Dimage Scan Multi Pro scanner. Whether this same sort of exposure correction can be done with your particular scanning software I don't know, but in order to properly profile your scanner it would be best to linearize the RGB-channel output as much as possible before you make a master scan of your target. In future scanning with the created profile you should always use the found RGB channel settings that gave you the best linearization.

</p>

<p>Here's my question- if it does introduce a cast, does it matter? Is it beyond what ICC profiles can correct?</p>

<p>Also, is the cast removable with analog gain? If so, dial down red and you're probably good.</p>

<p>Here's my example:<br>

The first is a sRGB jpeg of a Wolf Faust Provia 35mm IT8 target. Up top it's scanned with my fluorescent FS4000US with all settings to none. I converted it to sRGB for display.<br>

<img src="http://www.jingai.com/phototests/scanhancer%20comparison.jpg" alt="No profiles applied" width="500" height="675" /></p>

<p>For ｔhe second image I assigned the proper IT8 profile for both the non-scanhancer (top) and scanhancer scan (bottom). Both ICCs were made in Scarse. I can't really see much difference in the results beyond the slight exposure differences (scanner only has 6 exposure steps and the scanhancer blocks a little less than 3 steps worth of light). I converted this to sRGB for the web. <br>

<img src="http://www.jingai.com/phototests/scanhancer%20comparison%20profiles%20applied.jpg" alt="ICC profiles applied" width="600" height="810" /></p>

<blockquote>

<p>In reality it does not block any frequency, but since blue and green light are diffused more away from the receiving lens it appear as if more red light comes through.</p>

</blockquote>

<p>Hahaha. Scanhancer models a sunset :)</p>

<p>All joking aside, I have yet to read through both your posts... so I'll be back.</p>

<p>Thanks!<br>

Rishi</p>

<p>OK, some quick thoughts:</p>

 

<ol>

<li>Like most men, I can't follow directions. Or, rather, choose to ignore them. :)</li>

<li>Erik, thanks. Blazing fast response. Instantaneous almost.</li>

<li>Roger, re: your question of 'is this so bad that an ICC profile can't remove the cast?'. My answer: yes, for the LS-5000 and Velvia film. I have to dig up some comparisons, some of which may reside on my hard drive at school, which'll have to wait till January. But, even after the ICC profile (I used an IT8 target from Wolf Faust), the Scanhancer scan was warmer and more pinkish than the LS-9000 scan. <strong>However</strong> , this was <em>without any R, G, or B analog gain adjustments.</em> So, next step: I will try adjusting the analog gains. This has been on my 'To-Do' list for some time now :)</li>

<li>On one LS-5000 unit, the only one I tested this on, Erik Krause's Super Advanced Workflow failed for scanning negatives using the Scanhancer. Why? Because no analog gain adjustment got rid of the red. This problem may just be worse for negative film, since it already has that orangish/red tone. Again, I've only tested this on one scanner, so it could've been a bad one? I don't know. I'm going to test it on my own LS-4000 tonight or tomorrow... to see if I can hike up the green and blue gains so as to get rid of the orange mask. Yes I'll be careful to adjust the overall exposure such that nothing blows out.</li>

<li><strong>Erik:</strong> question for you. Nikon probably ran into this problem also when putting the diffuser in their LS-9000... because ICE still works, so it must also have the same issue of red cast. Do you think they inherently have higher 'analog gains' for the green & blue channels?</li>

<li><strong>Erik:</strong> Scanhancer + ICE on the LS-4000/5000 introduces artifacts, unfortunately, that require you to set IR removal to 'Heavy' to not have the artifacts. It seems that image information is being subtracted when IR exposure is set to '1' or when it's set to 'Light'. IR autoexposure usually results in somewhere between 1 and 2 without the Scanhancer. With the Scanhancer, the IR autoexposure results in ~13. But this still subtracted image data, especially around the edges of the frame. Setting IR exposure to 16ish helped, as looking at just the IR channel, I saw dust show up as black, but image data hardly show up at all (which is good). At IR exposure ~30, of course, the entire IR channel was white, as not even the dust was picked up (too high signal for dust's absorption to introduce any detectable signal)... so, any ideas why this is happening? I show the artifact below for negative film... note the dark blotches in the skin on the left (IR: Light, with Scanhancer); and the lack thereof on the right (IR: off, with Scanhancer). For slide film, the opposite happens -- one sees lighter blotches.</li>

</ol>

<p><img src="http://staff.washington.edu/rjsanyal/ScanningWoes/Scanhancer-IR_Artifacts.jpg" alt="" width="800" /><br>

<a href="http://staff.washington.edu/rjsanyal/ScanningWoes/Scanhancer-IR_Artifacts.jpg">Link to Full-Size Image</a></p>

<p>Cheers,<br>

Rishi</p>

<p>Dear Rishi, </p>

<p>I'll try to answer your questions 5 and 6, but please bear with me as I do not use this scanner myself and I deliberately named the Scanhancer for Nikon V/5000 a "beta" as to make sure that users would understand that some home experimenting with it could be needed under some circumstances.</p>

<p>5> The diffusing in the Nikon 9000 is not done by a diffuser, but by some changes to the dispersion rod that creates the light line inside from the LEDs that shine into its ends. They might as well have upped the blue and green LEDs a bit in order to compensate on hardware level. I find the term "analog gain" somewhat misleading, by the way. At least when speaking about how it is implemented in the Minolta Dimage Scan Multi Pro, where the separate channels are actually exposed longer to a certain extend. If it was just "analog gain" then the noise would also go up, but this is not the case (as long as you don't correct outside two stops in total). AFAIK the Nikons are also exposing longer when you use "analog gain". Whether this can be done for each channel separately, like in Minolta's, I don't know.</p>

<p>6> I have no experience in playing with IR dust removal in VueScan as I use the original ICE algorithms that came with my scanner's software. The Minolta driver has no options for the heaviness of ICE, so it works or it doesn't. Scanhancer behavior is tuned to work best with ICE the way it is implemented in the Dimage Scan driver. Vuescan uses an algorithm written by Ed Hamrick. If you can find a setting that works fine with Scanhancer I would suggest that you use that. ICE in the Minolta driver does not eat any details other than dust and micro-bubbles (aka pepper grain). The settings for your Nikon scanner in VueScan of IR exposure 12 and IR dust removal to Heavy do seem to work identical and not eat any details either. Maybe you should ask Ed Hamrick if you want to find out the mechanics behind this behavior, but at least it seems to work so I don't know if there is a reason to bother any further if all you want is making good scans.</p>

<p>Erik,</p>

<p>You're right, analog gain just increases exposure time. There are individual red, green, and blue analog gain sliders for the Nikons with Vuescan. So, yes, you can adjust them individually.</p>

<p>Finally, I'm always worried about using a setting of 'Heavy' for ICE... typically this leads to an overall loss of sharpness in the entire image. You're right I should ask Mr. Hamrick how exactly it works. In the meantime, the reason I'm fooling around with different settings is because I've found that upping the IR exposure a little bit (but not too much) with the Scanhancer leads to the IR channel showing less image information 'leaking' into the IR channel scan, yet still showing dust as black...</p>

<p>Here, these results below will explain everything I'm trying to say:<br>

<img src="http://staff.washington.edu/rjsanyal/ScanningWoes/Scanhancer-IR1vsIR16.jpg" alt="" width="800" /><br>

<a href="http://staff.washington.edu/rjsanyal/ScanningWoes/Scanhancer-IR1vsIR16.jpg">Link to Full-Size Image</a></p>

<p>Cheers,<br /> Rishi</p>

<p>Vuescan does not have ICE. It has a proprietary dust-off algorithim that I have never found useful.</p>

<p>I have more to report, but a quick update:</p>

<p>Using R, G, B analog gains on the LS-4000, one CAN get rid of the red cast AND the orange mask.</p>

<p>So, Erik, you should start advertising your Scanhancer for the LS-4000 also... I don't know why you initially said that the light source of the LS-4000 is too weak for the Scanhancer... in my experience it is not.</p>

<p>Clearly the LS-5000 I previously used to try and get rid of the red cast must have had a busted blue LED or something!</p>

<p>Roger, hope this answers your question. So this is encouraging. What's not encouraging is that upon close inspection on some scans, I can still see the etched pattern of the AN glass.</p>

<p>So, how the heck to people who use the glass holder on the Nikon LS-9000 not see the pattern on the AN glass? Is Nikon's AN glass inherently superior to Focal Point's AN glass?</p>

<p>I'll try and start another thread to post specific examples with the AN glass, as this OT.<br>

Rishi</p>

<p>Roger,</p>

<p>Do you use ICE with the Scanhancer and do you see the same artifacts I see (as shown above)?</p>

<p>Thanks!<br>

Rishi</p>

<p>For the sake of posterity, I am able to get rid of the color cast of Scanhancer for both slide & negative scans on the Nikon LS-4000!</p>

<p>Roger does not experience these IR removal artifacts because in his Canon scanner, the Scanhancer is held extremely close to the light source.</p>

<p>I am currently taking measurements to send to Erik in hopes that a redesigned Scanhancer might be perfect for the LS-4000/LS-5000.</p>

<p>Certainly, I find the Scanhancer <strong>indispensable</strong> for these small format Coolscans... but the IR artifacts need to be addressed.<br>

<br /> Cheers,<br>

Rishi</p>

<p>Rishi, et al.<br /><br />Do you have any experience using the Nikon 5000 Scanner with silverfast and the scanhancer? I get dark images. However, there does not seem to be a way to control the gain/exposure settings from within silverfast. There is a lamp calibration feature but this does not seem to compensate for the light losses with the diffuser. I am still waiting to hear from Silverast on this issue</p>