Scanning 4x5 Negatives

Hi there...

I'm quite new at scanning. I'm hoping that someone with more experience than I can offer advice on the best ways to

scan 4x5 B&W negatives to allow me to print from 4x5 inch up to 11x14 inch. What would the optimum dpi for such

sizes be? And once scanned, how should I save the files for best & sharpest printing on my Epson printer? I am

using an Epson 4870 scanner and am using Epson Scan and PhotoShop Elements. Any tips/suggestions would be

greatly appreciated. Thanks in advance for any help.

<p>You normally want enough pixels to have 300 pixels/inch at print size. So you want to scan at 1200 dpi on the scanner. Work and save in lossless formats, such as TIFF and PSD.<br>

When you resize to print size at 300 dpi, you want to sharpen (say unsharp mask) to make up for the resampling losses.<br>

Study posts in the Digital Darkroom format.<br>

You may find that buying an adjustable height filmholder from http://www.betterscanning.com for you 4780 results in significantly sharper scans.<br>

You may find that Epson scan is so hard to use that you need to use something more professional, such as SilverFast or VueScan.<br>

Download Ctein's book on printing. Even though it's more focused on wet printing, there's a lot to learn about digital workflow there as well.</p>

<p>Hi George and welcome to photo.net. I understand completely how complex the issues you raise in your question are to a nubie. Your questions are broad and cover a lot of ground. My best advice would be to start with your scanner manual. Next go to amazon and search on the words film scanning and see the multitude of focused books you will find. The library is another source. There are many dedicated websites dedicated to these questions. Google is your friend. There are many newsgroups and fourms on the web dedicated to scanning best practices.</p>

<p>Do some foundational work and come back with some more specific questions and I think you will have a more productive result.</p>

<p>The best scanning method for a desktop scanner is to fluid mount the film. This holds the film perfectly flat and ideally in a plane parallel to the scanning bar path and at the right distance from the bar so that you get the ful lresolution of the film.<br>

Ideally you should scan atthe maximum Optical resolution for the scanner (usually 4000dpi) and at 16 bits per channel in the very large Pro Photo RGB color space.<br>

You want to set the clipping points in the scanning software so that you get the full range of shadow detail (if a transparency set the black clipping point to 0, if a negative, 255) and the white clipping point only about 5 points above the end of the histogram so you aren't wasting dynamic range on information that isn't there. </p>

<blockquote>

<p><em>You may find that buying an adjustable height filmholder from <a rel="nofollow" href="http://www.betterscanning.com" target="_blank">http://www.betterscanning.com</a> for you 4780 results in significantly sharper scans.</em></p>

</blockquote>

<p>I agree fully with that recommendation. </p>

<p>This is a good start for scanning 4x5 sheet film with an Epson scanner. <br>

http://www.kenleegallery.com/html/tech/scanning.php</p>

<p>Ellis's ideas are a good start-point, although I doubt the figure for optical resolution.<br>

Assuming for the sake of argument that you want your 14x11" print to arise from an image at 300ppi (reasonable for the size of print and expected viewing distances), that's a 2.8x enlargement of the neg, so, double it again for Nyquist reasons, and you're looking at about a 1680spi scan.<br>

Given that, you might as well go the whole hog and aim for the optical resolution of the scanner, which (unlike Ellis's comment) is likely around 2000-2400spi. This gives you the best archive-quality scan from which you can scale down to your print-size of choice any number of times in the future.<br>

I would also second the recommendation for Vuescan, myself. Turn all the sharpening filters off, get a good high-resolution b&w scan with all the pixels well-exposed (probably fit all the dynamic range used into the full width of the histogram unless you know the scene better), saved as 16-bit greyscale TIFF, and then play with it in Photoshop (Elements?) to your heart's content, adjust levels first then curves then spot the dust off it with a healing or clone brush (depending on healing algorithm, you might find a slightly fuzzy clone brush interacts with the film grain pattern better). Apply a dash of wide-radius (10 or 20px) unsharp-masking for subtle shadow-edge enhancement, and/or use high-pass filtering for sharpening (google these terms if need be). At this point you should have an archive-quality image; don't lose it, but save a copy on the side as 8-bit sRGB JPEG at highest quality (12) for posterity (this is good for uploading to photo-sharing / selling sites).<br>

To make a print, go back to the pre-JPEG image, scale it down using bicubic (maybe bicubic-sharper) interpolation such that it's 14x11" at 300dpi ie 4200*3300px. Now use unsharp masking with a tiny radius - maybe 0.85px - and whatever amount is required to make every individual pixel just sharp enough (try some extreme values such as 1, 100 and 250 to get a grasp of the consequences, then build it gradually up).<br>

Printing is a yet further dark art: you need to get an ICC profile for your favoured paper (or calibrate it yourself), and then whilst looking at a profiled preview, apply whatever levels and curves are required to give a desirable result (either pure grey (hard to get right) or simulate selenium (cool) or sepia (warm) toning). *Then* hit the magic print button, sit back and prepare to repeat the process until it looks good in suitable light.</p>

<p>I hate Vuescan but it might be better than Epson scan. If you are serious about scanning my recommendation , despite it's non-intuitive inerface and too amny controls for photographers, is SilverFast AI Studio from lasersoft.com<br>

I agree that the actual maximum resolution may be lower than the claimed optical resolution but have also seen that wet mounting the film does increase resolution over using the standard Epson film holders where the height isn't optimum and the film is not held flat. I would start at 4000dpi and then try lower resolution settings. </p>

<p>I am not sure that I have ever seen the Nyquist issue come into play on modern scanners, printers or cameras, so I don't know that I would worry about it. (When they did occur at higher dpi's, the problem was generally resolved by moving from an 8bit scan to a 16 bit scan).</p>

<p>But what you should consider is whether you want to work a file over and over if you decide to make different sized prints. If you are near being a beginner, you may only want to scan to the size you think you would print now. Once you have honed your skills shooting and in post processing, I would suggest that you scan the largest file that is optically possible and work it. I can spend a full day working an image and it wouldn't be prudent to only scan for a single print size when I know I will use it in various ways.</p>

<p>But, if you want to scan a give piece of film for a maximum prints size and keep your file size modest, just divide the film size into the print size and multiply by the desired resolution of the print. For Epson printers I have always preferred to use 360 dpi, which is a direct multiple of their resolution capability (360/720/1440/2880) while consumer printers. like at Costco, are generally optimized at 300 dpi. In the case of a piece of 4x5 film, an 11x14 inch print is 2.8 times as Tim mentioned. Taking into consideration that 4x5 isn't actually 4x5 inches, using a 3x factor would give you a resultant 900 dpi scan. 16 bit is very important and I have seen issues scanning b/w white on older scanners, moire, when the 8 bit setting has been used--as mentioned above.</p>

<p>But I also suggest that you find out the optical scanning points for your scanner. Most scanners will scan at various dpi's, set by you, but their true optical scans might be at intervals, like 300/600/1200 etc. Scanning a bit smaller file or a bit larger file, but at the optical resolution of the scanner, will produce a better result overall. Small up sampling or down sampling of your file will be less destructive than working a scan that has already been up or down sampled. I always like a bit bigger file than a smaller one.</p>

<p>Grain aliasing is all about Nyquist issues. It's mostly an issue with B&W film scanning, and it can be a problem. Another subject worth investigating on the web.<br>

The actual spacing of the pixels on the CCD device on the Epson 4870 is 4800 dpi. Scanning at any resolution over that is silly, you're just generating interpolated data, nothing real. But the reality of the optical path on this scanner is that you probably can't really get more resolution than 1600 to 2000 dpi. So I'd say scan at 2400. Now, if you get grain much more violent that what's really on the film, you may need to experiment with 4800 dpi or 1200 dpi.<br>

I generally don't like the scan at resolutions that aren't an even factor or multiple of the native CCD resolution of 4800. Otherwise you may get odd interpolation artifacts.<br>

I agree with Ellis on VueScan -- either you love it or hate it. I find SilverFast infinitely better, but many folks swear by VueScan. Both get the job done well, it's a matter of how they treat the user.</p>

<p><em>double it again for Nyquist reasons</em></p>

<p><em>I am not sure that I have ever seen the Nyquist issue come into play on modern scanners . . . .</em></p>

<p>Um, do we maybe have a misunderstanding of what the Nyquist theorem says? Nyquist applies to any system that records something by sampling its value(s) close enough (in space and/or time) to capture all the changes. Basically, it says that if you want to resolve X distinct things, you need to take > X samples. How much greater is not some fixed amount, but generally it's on the order of 10%. Take as an example the sound on a regular audio compact disc. In order to provide a high frequency response limit of > 20 kHz, that is, resolve the high and low points of a waveform cycling 20,000 times per second, you need two samples per cycle plus about 10%. Two samples per cycle means 2 x 20000 = 40000, and add 10% plus a little more and you get to the CD standard sampling rate of 44.1 kHz (that is, 44100 Hz or 44100 samples per second).</p>

<p>Now applying the same to scanners: So if a piece of film contains 50 lp/mm of real image detail, per the Nyquist theorem, to resolve that you need a scanner that can give you a true resolution of about 55 lp/mm (2800 ppi). (Yes, I am well aware that film does not have a fixed resolution like that, and typically the MTF response curve rolls off slowly.)</p>

<p>Scanning to deal with grain and how grain and image detail interact is kind of a separate issue.</p>

<p>But in the end, if you want to print 11x14 inches at 300 ppi from 4x5 film, your scan only needs to deliver about a real, effective, optical resolution of about 1000 ppi* to capture all of the detail that the print can contain. I'm not suggesting that there's no benefit to scanning once and capturing all that's there, or that achieving a true 1000 ppi may not require scanning at a higher resolution.</p>

<p>*The rectangular opening on my film holders is about 3.78 x 4.76 inches, so 11x14 is a 2.94x enlargement. At 300 ppi, that means a scan capturing 882 ppi. Add 10%+ for Nyquist issues and you get 970+ ppi.</p>

<p>You guys are a tough crowd, but I am not a scientist, however, I have been scanning and working with scans for 20+ years. Black & white film can be problematic and my suggestion is that when I have seen issues like moire and other "scanning" issues, at any given resolution, 16 bit scanning seemed to solve these things. I use Imacon/Hasselblad scanners, so the issues might be different with other types.</p>

<p>The problem with scanning a 4x5 at too high a resolution, for many, is just the size of the file you have to work with. A 16bit file scanned at even 900 dpi is going to be 92mb and one at 2400dpi will be 660mb--which is too large for most people's equipment or needs.</p>

<p>Some compromises do have to be made to fit your equipment, software and needs.</p>

<p>George,<br>

You have asked this question on at least two differnent forums & at the Large Photography Info forum. Why don't you do search?</p>