Managing files from B&W 4x5 scans

[Tadge Dryja]

Hi,

I'm doing a project where I will be shooting a lot of black & white 4x5 sheet film & working digitally. It seems like a good way to work would be to make contact sheets of everything, and scan the ones that look good. I want pretty huge scans (~100M pixels?), and I'm worried that I might run out of space. What file formats to people who have done this kind of thing use? I probably want to stay away from lossy compression like JPEG, and I also might want to stay in 16-bit grayscale. I'm assuming 16-bit grayscale makes your files twice as big, is this worth it?

Obviously you can't see a difference on the computer monitor between 16 and 8 bit (since every graphics card I know of does 8 bits/channel) but is it important for the final output, be it inkjet, lightjet, whatever?

Also, what is a good file type to use? I'm going to be working in photoshop CS; are PSD files compressed in any way? The only choices there seem to be when working with 16 bits/channel are TIF, PSD, RAW and PNG. When you go to 8 bits / channel, there are approx. 1.7 zillion file types.

Any advice or tips from people who have experience with this would be great.

Thanks!

-Tadge

[Bruce Watson]

I routinely scan 4x5 Tri-X on my drum scanner. I believe in a "scan once, use many" approach. Since some of my prints reach mural size, I scan to large file sizes in 16 bit grayscale. The resulting filesizes are around 550MB, which I downsample as required.

The only file type I use is TIFF. The scanner can write it, Photoshop can read and write it, my RIP can also read it, and it's non-lossy. I don't use any compression, because LZ compression can implode with small bit errors, and 550MB files have too many bits to take that chance.

When I get done with an image, I copy the workfiles to DVD. I know it's not archival, but it's the best I've got, and if I have to I can rescan the negatives, which are archival.

As to 8 bit vs. 16 bit, this can be a religious issue. Really, it comes down to what kind of work you are doing in Photoshop. If you are make large adjustments or effects, your 8 bit image can posterize. The same adjustments with a 16 bit image likely will not posterize.

Another way to look at it: If I take a scanner file, work on it, and send it to the printer, and take the same file, convert it to 8 bit, work on it, and send it to the printer, sometimes I can tell the difference. OTOH, if I take a finished 16 bit image file and send it to the printer, then change the mode on it to 8 bit and send that to the printer, I can't tell which is which.

It all depends on what kind of work you are doing in Photoshop. Of course, YMMV.

[Ralph Barker]

I concur with Hogarth's scan-once approach, and generally scan at the optimum output DPI for the largest print I'm likely to make from the scan. After spotting and basic levels and curves adjustments, I then store the "master" scan in PSD format. Any further adjustments are made to that file, and then stored separately with a different file name. I use a file-naming scheme that denotes location, date, subject, and sheet/negative number in abbreviated form, and that corresponds to the edge marking on the PrintFile sleeve in which the negative is stored. That way, I can tell what the image is by looking at the file name, and find the corresponding negative easily if I need to. So, a scan of the Emmigrant Pass shot on the bottom-right 4x5 negative on the third sequential sheet of images from a trip to Death Valley in February of 2003 might look like: DV0203-03D-EmPass. A B&W 600 pixel (long side), reduced-size version of that image for Web display would be DV0203-03D-EmPass-600bw.

Whether to scan and save as 8-bit grayscale, 16-bit grayscale, or 16-bit RGB depends on which pew in the church one occupies. And, while disk space is cheap, backups on DVD or CD are essential.

[Geoffrey Swenson]

Just for the record;

PPI = Pixel per inch for scans and video display

DPI = Dot per inch for printers

[Brian Ellis]

I thought it was pretty well accepted that 16 bit is important when you're making adjustments (levels, curves, etc.) but that a 16 bit file printed won't look any different from an 8 bit file. In other words, the higher number of bits doesn't translate to a different looking print, it just minimizes posterization and artifacts when making adjustments before printing. There was a long, heated discussion about this in the comp.periphs.scanners group a couple weeks ago. In any event, if that's correct then the idea would be to make as many adjustments as possible in 16 bit then convert to 8 bit for things that can't be done in 16 bit and save the file as an 8 bit file. However, if someone who knows more than I do(most everyone when it comes to digital) says I'm wrong I wouldn't argue with them.

[David F. Stein]

One advantage of Vuescan or Silverfast is that you can generate a max resolution, 16-channel output TIFF and an output RAW file. There are 2 advantages to having that RAW file: as these programs improve in their quality or features, you can go back and "rescan" from the disk file (plus other scanner software doesn't get updated that often); and can go back and "slice off" a less than maximum size scan from the RAW file on disk at any time. Always working or starting with a file many times larger than you may need for a specific output gets tedious. The best archival storage right now is high quality, large capacity, large buffer ATA drives with a Firewire bridge system. The current cost of hard drives makes this feasible. A 120 GB, 8 MB buffer ATA drive is less than $200. External cases are inexpensive enought to even make one case per drive cost-effective or there are hot-plug systems. Now, for backup, you can easily have redundant hard drives. Even at the current state of DVD, they don't hold that much and are expensive. I consider them, like their CD predecessors, as best for distribution of stuff. GOOD LUCK.

[Ralph Barker]

To make things even more confusing, one must be cautious with the context in which the "8-bit" and "16-bit" descriptions are used. Within the context of an RGB scan, both terms generally refer to either 8 or 16 bits per channel - meaning either 24 or 48 total bits per pixel, previously called "true color". Within the context of grayscale, however, 8-bit often means that there are only 8 bits total for each pixel, resulting in only 256 different shades. Depending on the subject matter, this can result in what is called banding, where subtle tonal differences in an otherwise plain field (e.g. a seamless background) are not handled well by the palette of available tones.

[Michael Rosenberg]

Todge,

Why scan the contact proofs? I proof my negatives by scanning them on my Epson 1680 w/transparency lid using VueScan software. I have it set for the scans to be low contrast, similar to a contact sheet. I then use PS to explore cropping and contrast. Using the PS file browser I add information on each negative. All my negatives are sorted numerically, and backed up to a CD or a second hard drive. The price of this scanner should be low with the 3200 out and the new Epson due out soon. It does a decent job. I have VueScan set for 800 dpi and it is a pretty decent scan. OTOH, it does not compare to the more expensive higher quality drum scans (I guess quality before price is more appropriate).

I also recommend getting a decent backup software like Norton Ghost or Acronis TrueImage. I use the latter and it has saved me from hard drive crashing and my system files corrupting.

Regards,

Mike