16-Bit B&W Printing: Worth the Trouble ?

Now that Photoshop CS supports much more image correction in 16-bit mode, I wonder if it is worth the trouble.

The good news is that while 8-bit B&W gives 256 shades of grey, 16-bit gives 256 * 256 shades of grey: quite a lot, a "Zone System" with well over 65,000 zones. The downside, of course, is that files are much larger, slower to work on, and more resource-expensive to save.

Up to now, I have always scanned at 16-bit, but eventually "down-sized", saved and printed as 8-bit files. They arleady look pretty good.

Upon close examination, I can see the difference on screen, between the 16 and 8 bit files, but wonder if the extra storage requirements and processing slowness result in an image which is appreciably finer in quality when printed.

Any experience with this ? Thanks

There was a long thread about this question in comp.periphs.scanners a couple months ago in which some knowledgeable people like Wayne Fulton participated. I confess to not understanding some of the technical stuff that was bandied about but FWIW I came away with the impression that editing in 16 bit is a good thing but once that's done and you're ready to print and save you can covert to 8 bit with no loss in quality, i.e. that there's no observable difference between two otherwise identical prints, one from an 8 bit file and the other from a 16 bit file.

"Upon close examination, I can see the difference on screen, between the 16 and 8 bit files"

... I have trouble beleiving that it's anything other than an illusion. The DACs on video cards don't have more than 8 bits per channel. *mayyyybe* there's some crazy video card out there that can do 10 or 12 bits per channel on the DAC, but I've never heard of it.

If you're using windows, you can confirm this in the display properties->settings->Color Quality menu. 32 bit color is what most people use, and that's gives you 256 shades of gray coming off your video card to your monitor.

So I don't think you can see the difference between 8 and 16 bit on your monitor.

That said, I work with 16 bit files a bit. Generally what I do is scan them at 16 bit, and then do levels, curves, and other such global adjustments when working in 16 bits/channel. When I'm done with that and the "exposure" looks about right, I convert to 8-bit so I can work on files faster, touch up dust, or whatever other photoshop things.

8 bits is definetly good enough for a print. 16 bits are nice when you are manipulating the curves or histograms of the image, because it will remain more continuous in tone even after repeated manipulation.

Also, I wonder how the printers out there deal with 16 bit/channel CMYK data. I don't really know, but I'd suspect that most printers out there can't actually make use of 64 bits for a single dot. The computer engineer in me has a hunch that most printers would just throw away a lot of the least significant bits they're sent at 16 bits/ channel. If anyone knows specifically how printers' "DAC"s work (do they call them that?), I'd love to know.

Tadge, the last I checked the printer driver spec in Windows requires 8bit color, so all programs must down sample to 8bits for printing. I also doubt that you can see the difference on the screen - I sure can't.

Well, I can certainly see the difference *after editing*. 16 bits can make quite a difference on a "severe" edit; for example, a considerable contrast expansion (usually by curves for me). I think once the big edits are made, there will be very difference between 16-bit and 8-bit. This is my opinion, but you will find quite a few "experts" which state essentially the same thing.

YMMV.

doh, that should be "very little difference"....

I'm not at my photo-editing computer, but when I get to it, I will send a sample screen where one can see a difference between 8 and 16-bit images. The difference shows up at > 100 % enlargement - in other words, the pixels change.



If, as you say, the monitor, card, and OS itself can only describe 8 bits, then perhaps it is some artifact, which results from the conversion back to 8 bits.



In any event, this article nicely illustrates some of the differences you can expect if you scan at only 8 bits.

An interesting point made in the article, concerning the Ansel Adams negative:

"Actually, in our original 16-bits grayscale scan of the Tetons, there were only about 572 tonal values..."

Hmm... yeah, I believe that you can sometimes see pixel differences with 16/8 bit, but I'm not sure how meaningful they are. I don't know how photoshop works well enough to know why, but I know that if I have a greyscale image (only one channel), and I convert to an RBG file in the image->mode menu, the image changes! It gets a little darker, sometimes. It shouldn't do anything, since all the R G and B values are the same and should be equal to the gray values. It's a subtle shift, but I can keep hitting ctrl-z and see it get brighter and darker. Here is a little part of an image that does it pretty noticably, at least when I try it in photoshop.

There's lots of weird stuff like that in photoshop. I usually try to ignore it.

I think if you have a really good, spot on scan in 8 bit, you're probably fine. I usually scan pretty quickly in 16 bit and then make it look right in photoshop afterwards... a lot of people have to share the imacon scanner, but I can run photoshop as long as I want back in my room. If you set your black and white point too far out, it's not really a problem to bring them back in post-scan, but if they're far enough in that they start cutting out parts of the histogram, then that's data you don't have, and you'll probably want to re-scan.

I'm no expert on scanning or anything, but I'm getting the hang of all this digital stuff in the class I'm in now. Then again, it's so much more expensive, I think I'll stick to making prints in the darkroom for a bit longer, until I can go pick up a lightjet for a few hundred dollars.

The eye can detect around 256 levels of gray. It varies with induviduals, but much more than 256 is not seen as a smother transition.. Any digital editing that alters the contrast and brightness will leave some channels empty. At 8 bits these drop outs are noticeable. Editing at 16 bit and then dropping down to 8 bits before printing fills in the gaps and gives the smoothest output.

Matrox makes three video cards that run at 10-bits. They are the P650, P750 and Parhelia. PC only though. They definitely make for a smoother look in areas prone to banding (ie:sky's and fog) when working on 16-bit images of course. But in the end you have to stare close to notice much of a difference.