Chuck Guzis wrote: Yes. You can adjust the "quality factor" but I think this would be A Bad Choice. Especially given my original comment that the documents are D-size reduced to B-size before printing (i.e., the text/lines are VERY fine and apt to disappear in the DCT application within the JPEG encoder.)

Jules Richardson wrote: I think his point was "adjust quality factor so resulting image file has a size of ~10KB per sq in of original document". E.g., A size is ~90 sq in so target a 1MB file size. (note that this would correspond to an UNCOMPRESSED TIFF scanned at 300dpi! -- did I do the math right? -- in which case, the whole argument starts to fall apart since a 300dpi TIFF *compressed* would be much smaller! Or, for that given size, a COMPRESSED TIFF could be scanned at a much higher resolution)

der Mouse wrote: I read that as "ten thousand BYTES" per sq in. :>

Bob wrote: Right. *But*, if you take something that was *created* for a large format (e.g., D size) and *print* it at a much smaller format (e.g., B size), then the printing software scales the image down before sending it to the printer (and, ultimately, the marking engine therein). Lines/features get *finer* (of necessity). For example, one of the signals on the drawing I was just now examining is "AddressTerminalCount". 20 characters. It occupies 1 inch (width) on the *original* drawing. Reproduced on B-size media, that's 0.5". Even if you assume a really crappy typeface design (e.g., 5x7), that's 120 "print elements" wide (i.e., 20 * (5+1)). Of course, the actual typeface is much nicer/finer than that but I don't have the time to figure out what exactly how fine-grained it is. :> So, you have to resolve 120 dots in 0.5". That's already pushing the limits of a 300dpi scan if you don't want to end up with just a hieroglyphic blur! :> So, scan at 400dpi or better (in this example) and/or expect the text to look AS IF it was rendered as a 5x7 typeface would have been. :-( In my case, I want to be able to unambiguously resolve the original document's content AFTER scanning. Without having to scratch my head and wonder: "Is that an 'O' or a '0'?" I'd rather scan at 800dpi, e.g., and rely on a compression technology to make that "affordable" than have to live with a "cheaper" scan and some potential ambiguity (which I will undoubtedly only discover MUCH LATER -- after the originals have been discarded!)

Bob wrote: No. There are (at least) 120 dots of information (assuming you could magically transform the actual typeface to a 5x7 rendition -- which seems to be the crudest representation that people can readily recognize) in that 1" of the D size original (which *I* don't have). That was printed at B-size. So, those 120 dots of information are now represented in a 0.5" physical span. Scanning this "240 dpi" data (again, *assuming* you could magically transform the original font to that crude of a representation) at 300dpi is a bit "if-fy". I am saying I would prefer to scan the *B* size drawing (which is what I *have* -- I also don't have a scanner that will handle anything larger than B size :> ) at 800dpi. This would correspond to scanning the original D-size drawing (if it was available in that form) at *400* dpi -- which I consider a reasonable resolution for preserving content. I don't see where you came up with 1600dpi. You *lose* effective resolution, not gain it! (i.e. the image has been reduced prior to scanning) I'm advocating images scanned at ~400dpi IN THEIR ORIGINAL FORM. I.e. I would need to print 8 (?) A-size sheets at 400dpi and paste them together to get the original D-size drawing -- reproduced at 400dpi. The D-size original (with 120 "dots of information" per inch) was originally *printed* in a form that represented those 120 dots in half that space (B-size) -- which is recognizable (hmmm... do the Brits write "recognisable"?) on a *good* 300dpi printer (but quickly turns to crap when those sheets are later photocopied). I am not interested in *printing* D-size drawings. :> Rather, I want to reproduce their content in fine enough detail that I can *view* it on a screen and not see big black "ink blots" (Hmmm... why do they all look like *bacon*? ;-) where there should be *text*. I typically create drawings as D-size. I.e., I set whatever CAD/EDA tool I am using for a "canvas" of that size. It's just too hard to draw anything "substantial" on B size originals (though C is sometimes acceptable). From these, I usually "print" B-size versions as working copies (because it is much cheaper to print B size than D size :> and, because D-size drawings take up a metric crapload of benchspace!). So, those "120 dots of information" per inch, get reproduced on a 300 - 600 dpi printer. That's usually reasonably legible. I usually end up keeping a set of *pristine* B-size drawings in my files (originals go to client, etc.). I.e., I don't keep photocopies (which are much crappier quality). I would like to be able to scan *these* and have "nice" digital copies of them. (I haven't always remembered to keep copies of the drawings in their original "digital" form; nor do I often have access to the *tools* with which I created those files -- even if I did manage to save copies of the files!) I can always tell the software that this is a D-size drawing *after* it has been scanned (by tweaking the document dimensions and letting the resolution adjust inversely). Correct. The issue is not losing *more*. I.e. you can read this text and realize it is reproduced at a greater "definition" (trying to avoid the word "resolution", here) than a simple 5x7 typeface.

On 8/15/2006 at 6:00 PM Jules Richardson wrote: I don't think so--that's 10K per "square inch". Linear resolution would be the square root of that--100 dpi, no? Cheers, Chuck