bit depth of Hell S3900 drum scanner

[Bill17]

Hi everyone

I recently have a chance to scan my positive film on a Heidelberg S3900 drum scanner.
The result is great and has really fine detail.

but I find a problem.
It seems that S3900 can only scan with color bit depth of 8 bit while most of recent scanner (drum or flat) can scan at 16 bit.
The S3900 operator told me that 8 bit is more than enough for most of application since printer can't printer finer than that.
Another person told me that I can transfer the result to 16 bit in Photoshop but I am almost certain that this doesn't sound right.


my problem is, what difference does it make? 8 bit and 16 bit.
I understand that 16bit color depth offers better color resolution.
but does it matters in real life application or just something looks different on number.


also, I look more into the scanner, it seems that Hell S3900 is 36bit.
Doesn't it means it has 12bit per color(RGB)?
Is there a way I can scan it at 12bit(if it is really better than 8 bit)?



Thanks

[Daniel Stone]

IF you can scan @ 12/bit per color channel, then that should go into a 16bit workspace in PS.

Yes, you are correct, most later model flatbed and drum scanners scan directly into a 16bit file.

"Finer detail", not sure about being able to resolve more, but having more "bits" per color channel allows one to have more range of tones and color to reproduce. Scanning @ 8bit is like cutting your neck before you even start editing IMO. Only downsample to 8bit before you print it, or let the RIP(software which controls the files for printing) and/or printer handle it. Many pre-press operations would scan for CMYK output, which has a much smaller colorspace than Adobe RGB 1998. So this might be what they're referring too. I'm not sure if the 3900 is in the generation where its still doing color separations like referred to by Dean Collins in the 80's, IDK. The pre-press industry has their standards, and can usually differ from what us photographers need and/or want in terms of final "image quality"...

I'm just getting started with the funny-language digital/PS business, despite being 23 and having grown up with computers all along ;)

Honestly, its part blatant ignorance of accepting the digital workflow as an option for me, but owning one's own drum scanner sure affords one the ability to experiment much more, at a drastically lower cost than farming out to a lab/service bureau!

[Ben Syverson]

At 2400 DPI or above, 8 bits are actually overkill. Film grain conspires to reduce the gross SNR of film at large magnifications, and SNR tells you effective bit depth. At 4000 DPI, color negative film is something like 5 bit. Positive film is maybe a hair more—maybe 6 bits.

The grain itself is quite coarse, but because there's so much area in LF, we see a completely smooth image. So dropping to 8 bit (or even 5 bit) won't affect the subtle color gradations in your image. Unlike a digital file, it's actually the noise (grain) in film that allows it to represent so many colors.

Somewhere on my HD I have a 2400 DPI 8x10 scan which I saved as a 3 bit GIF file. It has 100% of the tonal subtlety of the original scan, and you could get a great looking 16x20" print out of it.

I should really do a full article about bitdepth, along with the confusion between "dynamic range" and "latitude." It has to be the aspect of photography most misunderstood by photographers.

[SCHWARZZEIT]

I agree that at higher resolutions the tonality of most films can be described accurately at very low bit depth. But the data needs to be optimized to use the lower bit depth properly. On most scanners there is very little you can do to the signal before it is digitized in the A/D converter. Thus the raw data represents the density of the film in the full density range the scanner is able to capture plus the noise inherent to the scanner. I'd like to add that on most drum scanners the signal goes through a log amp that boosts the higher densities on the analog stage before digitization. So the ADC output is 12 bit log data and not linear.
If you do your color correction in the scan software based on a prescan than I would assume that your software uses the full 12 bit/channel data, applies your corrections and then writes the data in an 8 bit file which should be perfectly fine. But if you do raw scans you should definitely try to get all the available data from the ADCs into your file. Especially on color negs that use only about half of a drum scanner's range, and also because the neg's huge capture range is compressed in a smaller density range a lot of the raw data is thrown away while the meat you've left is pushed and stretched a lot in the grading process to get to a pleasing image in the desired output medium. The point is you should throw away as much of the data you don't need for the image as you can before you step down into 8 bit territory. Carefully choosing the black and white points is essential in that workflow.

-Dominique