This seems to be a subject that receives very little knowledgable discussion in the many books that I've been reading recently. Some writers sort of "hand-wave" about it, but I don't recall seeing any real in-depth discussion.

I know that part of the problem is with current vendors of printer drivers pretending that the way their drivers work should be a Deep Dark Secret. In the good ol' days, well-documented printer interface languages allowed a programmer or educated user to understand how printer resolutions worked. In a former life I wrote printer drivers for a living. Alas, those days are over now, and apparently manufacturers consider it to be a competetive secret to hide this information from users.

I know that I have to use a resolution somewhat less than the printer's physical capability so that colors can be dithered at the final resolution. And I presume that an image resolution that is evenly divisible by an integer (eg, 300 dpi for a 1200 dpi printer) should provide a smoother rendering by the printer.

Most printers these days will casually increase an image "to fit the page" regardless of original size or page size. And a number of people have told me that the printer driver's up-rez is "good enough."

I'm interested in any thoughts on this subject. For example, if I want to print an image on 13x19 paper with .25" margins, should I be calculating the image size and pixel resolution in PhotShop for 12.5x18.5 inches at some particular PPI value? Does it even make a difference? I suppose I could print out a whole bunch of test prints at different sizes and resolutions, both with and without using the driver's sizing capabilities, and then get out the loupe and see what it tells me. Has anyone done this? And I imagine it would be different between printer manufacturers.

Epson printers have resolutions that are divisible by 360. Canon and HP printhead resolutions are divisible by 300. You're absolutely correct that a printer with 1200dpi resolution does not need a file at 1200ppi. It takes multiple dots to render a single pixel.
Within reasonable ranges the printer drivers by and large do an excellent job of resampling. I certainly woudn't take the time to change an image from 343ppi or 240ppi to 300 or 360 (depending on the printer). If I'm going to print small, I would downsample a 10 megapixel capture for a 4x6 print, but that's for the smaller file size more than anything else. Now if I'm going to upsample an image (to print it bigger than it's native output allows) I would target the lowest ppi that is a multiple of the printer resolution.
So unless you are way over in image resolution and want a smaller file size and spooling time or have to upsample anyway, I wouldn't spend time matching an even multiple of the printer resolution.

** My math downfall began in 7th grade Algebra, so I may very well be using "multiple" and "divisible" incorrectly, but I think you know what I mean**

Heh. My math downfall didn't occur until 10th grade. And by then I knew how to program computers so I didn't care.

But my question has less to do with resampling and more to do with how printer drivers work. And the dearth of information available from manufacturers. Seems to me that all image resizing should be done in PhotoShop (or whatever tool of choice) where I understand how it is being done and can make choices about the outcome, not in the printer driver where no one will tell me how it's happening. For example, my shiny new ipf5000 has a check box for "print to roll width", which is very convenient as a way to get the largest possible prints while conserving paper. But if it has to up-res or down-res the image while printing, I wonder what potential damage it is doing to the image. If they up-res, are they applying some sharpening after?

Part of my curiosity on this subject comes from the fact that I used an 8X loupe for the first time yesterday to examine the output from my new ipf5000. And I was disappointed to see how square some of the pixels appear at the edges of a highlight in the image. I'm in pursuit of the highest-quality image at larger magnifications, trying to make wallpaper from photographs. Unlike a highway billboard, these will be observed from inches away.

OK. That's a much different question. Don't let the driver scale your images. Don't let the PS print dialog scale your images. Doing all of your image sizing in PS is the only way to have the degree of control for which you are looking. THere is always a lot of black box, proprietary stuff in printer drivers that the companies are never going to share. HP has, by far given the most detailed info to end users about what their media settings do, but that is very rare.

Using an 8x loupe can be fun but really, the issue is how does it look with the naked eye? Few people will press their noses against wallpaper, and no one would pull out a loupe, so while I understand your quest for ultimate quality, there comes a time when you just have to put away the loupe. Having said that, Epson still has the best dithering of them all. Their alogorithms combined with variable drop sizes for all channels (Canon and HP have fixed drops per head) give it the edge from a technical standpoint. This benefit you can clearly see with a loupe. Does this mean that Canon and HP cannot deliver wonderful, detailed prints? No. And that's the goal, right?

I just read in "Fine Art Printing for Photographers" that some RIPs will scale and up or down-size the image for you and that the algorithm that the RIP uses produces sharper looking images than if you resize the image in Photoshop. Its noted however that not all RIPs will automatically dither, scale or resize for you, but I'm with you I'd like to see my final prints with as little loss in quality and resolution as possible. It sounds to me that the idea is if my digital image has a native resolution of 180 and i'm trying to output to 11x14, then the RIP may choose to upres the image to say, perhaps, 300 or 360 for optimal output. I'm actually really excited to try this out, and find out just how far I can push the limit when I allow the RIP to choose the correct resolution for the output size.

I'm curious to know what others experiences with letting RIPs scale and resize are, and if it truly results in sharper looking images.

I believe this issue is why Qimage was created. It automatically handles the up/down rezzing of the image file depending on what printer is used and what print size is desired. It also provides different algorithms for this process...some faster some better in quality.

See this link to a discussion on Qimage output quality:
Qimage Quality

I have been using Qimage to "Print to File" to generate the optimum resolution output...which I then open in QTR for final toning & hard copy printout.

Works very well.