Scanning negatives

[bill_taylor2]

<p>I know this have been done to death, but I just went through the exercise myself. What scan should I use for a given negative and print combination. So I went back through first principles to work out the (semi) optimum scan pitch for a given final print. <br>

The question is this: for a given assumed viewing distance, a given assumed visual acuity (the smallest thing you can see), and various negative/transparency sizes and final print sizes, what scan pitch (dpi) should I use?<br>

The assumption for the spreadsheet is viewing from 10 inches out. This is basically critical viewing right on top of the print. Kodak uses a 14 inch viewing distance, I've soon others use 8 inches. The spreadsheet number can be changed easily enough. If you are making a billboard, you might use 20 feet (240 inches). It is your application. <br>

Visual acuity is just how good your eyes are. I used 2 minutes (2/60 degrees) as a pretty strict limit. That really is eagle eyes for humans. You might want to relax it to 3 - 5 minutes.<br>

From there you work out the dot pitch on the print, then back through the magnification to find the dot pitch on the negative. From 10 inches, 2 minutes works out to about 172dpi on the print. A lower pitch (e.g. 160dpi) would be coarse enough to see individual dots. A higher pitch (e.g. 200 dpi) would be fine enough to visually blend. The magnification is just how big you are blowing up the negative. I assumed the whole negative is being printed, which is often not the case, but you have to start somewhere. Digital sampling theory suggests that, for accurate reproduction, you need to sample at twice the frequency of the highest frequency of interest. That just means there is an extra factor of 2 involved in the sampling. Do the math, and you come up with the pitch the scanner should run at in order to get a good printable scan file.<br>

Keep in mind that film has an ultimate resolution, usually in the 50 - 64 lp/mm range. That works out to 5000 - 6500 dpi, way higher than most people ever get to. So generally speaking, the film has more information than the scanner can capture. But on the other hand, you probably will never print anything that will show that detail, unless you are just making extreme enlargements.<br>

I also estimated the file size, based on a 24 bit scan. Your mileage will vary.</p>

<p> </p>

[kelly_flanigan1]

<p>A billboard is often at 200 feet; only 20 feet if you are the installer.<br>

<br /> Considering prep work to do scanning; many of us just scan "at what our scanners will do".<br>

<br /> A criteria I have used for about 2 decade is 300 pixels per inch at 1 foot; that is like:<br>

<br /> 150 ppi at 2 feet<br>

<br /> 100 ppi at 3 feet<br>

<br /> Your billboard at 200 feet needs 1.5 ppi<br>

<br /> Thus for a 12x48 foot billboard; one needs a 216x 864 pixel image; a Blackberry will do this. A 35 dollar Walmart Bubble pack Digital exceeds this.<br>

<br /> Most extreme enlargements like Billboards require better than a 16mm Hit camera; a Minox is OK; 35mm is total overkill.<br>

A disposable 35mm camera from Walmart with Walmarts Frontier 1300 dpi scan is *TOTAL* overkill for a 12x48 foot billboard.</p>

[bill_taylor2]

<p>It is a minimum, not a maximum. If the scan doesn't look right, you can go with more.</p>

[kelly_flanigan1]

<p>With a 10 to one error in viewing distance;</p>

<p>one has a 10 times the number of linear pixels; or a 100 times error in megapixels.</p>

<p>Thus the mighty 216x864 pixel required becomes 2160x8640 pixels;</p>

<p>ie Barbiecam/Screen capture to My 50 Megapixel Phase One scan back; or drum scan of MF.</p>

<p> Thus One has a huge amount of confusion because folks cannot nail down the viewing distance.</p>

<p>Hockey dasher boards I print have glass at 85 feet away if across the glass; maybe 50 to 60 feet if one looks diagonally by a goal if one has front row seats. The average observer are 2 to 3 times these distances. </p>

<p>Thus with 300 ppi for 1 foot; 50 feet is the closest dasher board observer and one needs 6 pixels per inch, For a 3 by 12 foot dasher one needs a huge 216x864 image; the same as the 12x48 foot billboard.</p>

<p>A high end scan from my Pentax A110 slr is good enough; 1/2 frame 35mm is overkill. </p>

[john_shriver]

<p>Remember also that grain aliasing is minimized at full optical resolution of the scanner, especially if it has an optical low-pass filter. So, for example, I never run my Nikon Coolscan IV at anything but 2900 ppi.</p>

 

[swilson]

<p>

<p>I think your visual acuity is perhaps off just a bit, I believe 20/20 is when you can make out the parts to and E that is 5 minutes of arc in height, and since there are 5 bars in a capital E I figure at a minimum you would need 5 pixels across those 5 bars, or 1 pixel/minute of arc.<br>

Go down to the forth page in this document for a referance.<br>

<a href="http://www.cit.iit.bas.bg/CIT_03_en/v3-2/25-33.pdf">http://www.cit.iit.bas.bg/CIT_03_en/v3-2/25-33.pdf</a></p>

<p>But viewing distance is really going to be the key, not many people are going to view even a 12x18 inch print as close as 10 inches.</p>

</p>

[bill_taylor2]

<p>Just to double check, I reread my entry. Sure enough, it does say you can change the viewing distance number. And you can in fact enter a different number in that cell. And the math is correct. So, if you desire to view your prints from 200 feet off, feel free to adjust accordingly. 200x12 = 2400 inches. <br>

Scott, it is more of a line pairs thing. If there were 3 black bars AND two white ones (to provide contrast) in a height of 5 arc minutes, that is 1 arc minute per line, but 2 arc minutes per line pair. Without the pairs, you would have five black lines, which you could see (as a block), but not distinguish.</p>

 

[dave_redmann]

<p>AFAIK, the most widely-accepted limit of normal human vision is one-half minute-of-angle. That corresponds to 382 ppi at 18 inches, which is what I regard as a normal viewing distance for, say, an 8x10-inch print. (The math is ppi at limits of vision = 1/(viewing distance * tan(0.5/60)) where the trigonometric function is set to work in degrees.)</p>

<p>So if, say, you're making a 16x20-inch print, and regard minimum normal viewing distance as 3 ft, the print needs 191 ppi, or 3056 x 3820 pixels. If you're scanning one of my 4x5 negatives, which are about 96 x 121 mm, you'd need 3056 pixels in 96 mm, or 809 ppi in the scan.</p>

<p>But I would suggest that, if you're going to the trouble and perhaps expense to get a good scan, it makes little sense to scan at less <em>or more</em> than the effective resolution of the negative. Color transparency materials have less potential resolution, color negatives more, and B&W negatives potentially more still. Your lenses, technique, depth of field, and other things all can have major impacts. But realistically, your usual scan range is about 2000 to 4000 ppi, depending on particulars, and assuming your scanner can really deliver it.</p>

 

[jim_jones3]

<p>I scan at the maximum effective resolution, edit, resize to the desired print size at the printer's native ppi, and sharpen. The edited image can be resized and sharpened to any other practical print size. This does mean that the edited image takes a lot of storage space, but the scanning and editing usually has to be done only once for all size prints.</p>

[bill_taylor2]

<p>Eyeball resolution: all / most / some is revealed here:<br>

<a href="http://blogs.discovermagazine.com/badastronomy/2010/06/10/resolving-the-iphone-resolution"></a> Discover Magazine, Resolving the iPhone resolution</p>

 

[bill_taylor2]

<h2>For some reason it won't let me link that. so here it is in plain</h2>

<h2><a title="Permanent Link: Resolving the iPhone resolution" rel="bookmark" href="http://blogs.discovermagazine.com/badastronomy/2010/06/10/resolving-the-iphone-resolution/">http://blogs.discovermagazine.com/badastronomy/2010/06/10/resolving-the-iphone-resolution/</a></h2>

 

[swilson]

<p>300 ppi at 10 inches works out to 1/3 mr, which is a number that I have been using since the 70s. I belive from an earlier referance that I gave that 20/20 vision is defind as closer to .29 mr instead of .333, but the two are close enough not to make much of a differance.</p>

<p>

<p>On a practical note if the pixels are really sharp then most people will not notice a difference between 200 and 300 ppi, at 10 inches. Of course pixels from scanned film can be all over the place, going from very sharp to very soft depending in large part on how many ppi your are trying to get out of film. The 300 ppi limit at 10 inches is for very high contrast detail, when you start to put film scans past a certain point you get more detail but it is low in contrast and will not be visible in the print.</p>

</p>

[kelly_flanigan1]

<p>Look a the issue this way.</p>

<p>With that decades old criteria of 300 ppi at 1 foot; one has the 1.5 pixel per inch criteria at 200 feet for that 12x48 foot billboard; the same 216 by 864 pixel image I mentioned above.</p>

<p>lets us say you are driving down I-10 in nowhere Texas; to get closer you can:</p>

<p>(1) Be as sign installer<br /> <br /> (2) stop the car; jump the fence and ponder the signs flaws while dodgeing bulls.<br /> <br /> (3) One can shoot the bull and then try to see if Daves best case criteria works "AFAIK, the most widely-accepted limit of normal human vision is one-half minute-of-angle. "<br /> <br /> (4) Use binos while the car is parked; look at the billboard while getting ticketed</p>

<p>if you want to goose the numbers then best case everything; use a zillion ppi; make that billboard a contact print.</p>

<p><br /> You really do not put dinky eye chart "E"'s on billboards.</p>

<p>You do NOT design the billboards to be only read by Pilots illegally parked on interstate shoulders during daylight.</p>

<p>The said reality is that few here understand that a Billboard sign has to be readable under practical conditions:<br /> <br /> (A) Day and Night<br /> <br /> (B) Eyes that are Pilot eyes to good eyes; to average and poorer ones too<br /> <br /> © Ponder that ones vision is worse off axis; it drops like a rock a few degrees off axis<br /> <br /> (D) One is in a car driving at 75MPH; looking through a windshield covered with bugs<br /> <br /> (E) Ponder that few folks shoot USAF test targets though car windshields; in cars moving 75MPH</p>

<p>In the 1970's while doing double star work; I never split double stars while driving; or through a windshield.<br /> <br /> A billboard is a very low resolution thing; if you want to waste money; then build you doghouse with lumber to 1/100 inch; cut your grass to 1/100 " with sissors; send giant images to billboards.<br /> <br /> many folks will "never get it"; thus costs can skyrocket on machine parts to images.</p>

<p>Thus clients here that I print hockey dasher boards for that are "experts":) often using dinky text; thus the advert is retarded; ie nobody can read the damn thing. No matter how many samples one does; a bullhead will say do it my way; then one has this crappy sign that is ineffective. Then later the expert gets booted and I redo it with decent text; to fix the mess. </p>

[kelly_flanigan1]

<p>Long ago in the early 1990's scanning cost more.</p>

<p>Thus with a 4x5" trany the scan was about 2 bucks per megabyte.</p>

<p>If one had to scan 100 originals and the project only required a 30 buck scan;</p>

<p>then why pay 300 for a larger scan for a one time use that added no value? an error like that would be 270 bucks per original; a 27,000 buck screwup.</p>

<p>In that era clients cared about waste too; if one was a doofus and was wastefull; you might not get the next job; ie you wasted a mess of cash on scans that added zero value to the end product.</p>

[rodeo_joe1]

<p>Scan "resolution" is mainly limited by the optics of the scanning system, not the output pixels per inch figure. I have personally never seen a true optical resolution any higher than 3600 PPI from <em>any</em> scanner, no matter how expensive. (A drum scanner's true resolution is limited by the size of the scanning spot, and that's usually much larger than the number of PPI produced would lead you to believe). Even so, at 3600 PPI the grain or dye clumping of the film itself is the main limit to detail and IMHO there really is no point in going much higher - all you'll get is empty pixels.</p>

<p>As others have said, a final print output of 300 to 360 PPI is all you ever need. So scanning a 5x4 negative or transparency at a modest 2400 PPI will enable you to produce a print of nearly 32 by 40 inches that'll withstand almost nose-touching inspection. How big do you want to print?!</p>

<p>Thinking back to the dim-and-distant time before we all went digital, it was difficult to get much more than a 10 diameter enlargement from film by chemical/optical means that would withstand really close inspection, so I don't think we should get too hung up on ridiculous pixel numbers. And before you jump down my throat Kelly, I'm talking about normal camera use here, not high-res, high contrast repro stuff!</p>