Setting Optimal Exposure on Scanner (Coolscan)

[Rider]

Although I had been using scanners for years, I did not really realize that you can set a scanner's "exposure" much the same way as a camera's. So, even if you have software (such as Vuescan) that can produce a "raw" scan, it's still crucial to get the exposure right the first time.

How do you go about setting the optimal exposure on a scanner? The one I am using is Nikon Coolscan 5000. In Vuescan, the value starts at "1" and goes up.

My guess is that the tradeoffs are (i) noise generated by inreasing the exposure vs. the noise generated by cranking up the shadows in Photoshop and (ii) of course the highlights vs. the shadows, since I am quite surprised that even slides have more exposure latitude than a scanner.

So when scanning, do you expose for the shadows and let the highlights go, or do you expose for the highlights and then bring up the shadows in Photoshop? And what does "exposing for the shadows" mean in this context--how do you know when you've exposed for any thing?

I hope my question is at least somewhat lucid, or am I completely off the scent here?

[Peter J. De Smidt]

Hi Rider,

You're not off base. With scanners there are "hardware" adjustments and "software" adjustments. Hardware adjustments effect what info is captured, whereas software adjustments manipulate the data captured. It unfortunate that most scanner software makes it very unclear as to which is which.

Some hardware adjustments are:

Analogue gain (Nikon software), (I'm not sure what exactly this amounts to.)

Scanning speed, i.e. how long the sensor reads from a specific area,

Multi-sampling (taking more than one reading of a specific area of what is being scanned. (Multi-reading, i.e. taking multiple readings of an area at one time, are preferable to multi-passes, which involve scanning the whole negative multiple times. The problem with the latter is negative movement and resulting softness.

Source brightness. (I don't think that this is really variable on many scanners.)

Sensor gain (perhaps what Nikon means by analogue gain. This would be equivalent to upping the iso on a digital camera.)

Drum scanners have some more options, such as aperture size...

With digital sensors, such as in digital cameras, you want to give as much as exposure as possible such that clipping is avoided. With a digital camera, this means exposing so that the date represent by a histogram is close to, but not touching, the right edge of the graph. Doing this will minimize noise and maximize tonal separation.

In my experience, the same thing is true with scanners.

I really wish that scanning software would be available such that allows very precise and easy optimization of the hardware adjustments available. In my experience, most of the other features are detrimental. This ideal software would have a very good histogram feature, along with instant previews along the lines of the better raw processors. I've not found such software, although the professional versions-Oxygen, DPL, Color Genius--come closest.

I used Vuescan for a number of years. I certainly liked the idea of one software for many scanners, and Mr. Hamrick is very responsive. But it does include way, way too many options, and it's not-immediately obvious what some of them do.

Hopefully someone will be able to chime in saying what Vuescan's hardware adjustments are and how to use them.

[Rider]

Thanks Peter. I feel better knowing I'm not alone. It seems that scanner manufacturers always managed to keep the distinction between hardware and software controls somewhat fuzzy. Does moving that slider change the bits coming off the scanner or the way the bits are processed? Often unclear. One of the reasons I am using Vuescan is because it's possible to scan in "RAW" and do the processing later (the other reason of course is that it works under 64-bit Windows). That's what made me focus on "What are the minimum settings I need to set to make a RAW scan?" The Exposure setting seems to be the biggy. I think it's basically the length of time the CCD is exposed, like shutter speed. Aperture is fixed, and "ISO" possibly is controlled by Gain (Can anyone confirm this?).

In the Vuescan manual, it says that there is rarely any reason to use a greater exposure setting than the program chooses automatically. However, it does not say (as far as I could find) how this automatic exposure setting is chosen. Unfortunately (or fortunately, because it let me to question it), in the very first slide I tried the Exposure value was much lower than I would have used myself. None of the other sliders rescued the slide, until I rescanned with a longer exposure valie. In many other slides, the automatic determination was *ok*, but I would still like a methodical way to set an optimal Exposure value.

Other hardware settings I've identified with the Coolscan 5000 are more obvious: Crop, resolution, bit-depth, multi-sampling, multi-exposure, infrared cleaning. There is also an Infrared exposure, which I don't anticipate using since the IR algorithm is a bit of a mystery to me. There is also a gain setting for each of the R/G/B channels, which I don't anticipate needing except in extreme situations where a slide is devoid of a wavelength for some reason (maybe shooting under a funky aritificial light). {Edit: is Gain something like ISO?}

It appears that Vuescan does have a historgram, which could be set to display the "RAW" data; is that what you use to optimize the exposure setting?

Any help would be greatly appreciated.

Regards.

[Peter J. De Smidt]

'Hardware' changes should change the "RAW" data, and so if you get a histogram of that, preferably in RGB, then it could be a good way to evaluate exposure.

What I recommend is doing some low res test scans, keeping track of various hardware changes, as well as watching the histograms in Vuescan. It shouldn't take too many tests to see if this is worth playing with.

In my case, I use Screen's Color Genius software, as that's what I have to use to run my scanner. It doesn't have histograms, but it does have very precise numerical readouts, which can be edited, for each channel. They show both the pre-adjustment and post adjustment values. Unfortunately, I don't think these changes affect the capture of the data.

[SW Rick]

I think there's a new book by Sascha Steinhoff, The Vuescan Bible, just out- you might take a look at that to see if it covers the subject.

[Peter J. De Smidt]

See: http://www.lumika.org/gear_nikon_scan_analog_gain.htm

Also see: http://www.imaging-resource.com/SCAN/CSIV/C4A.HTM

In particular:

"One of the most basic enhancements was to recognize that simply increasing the brightness of the light source would increase the amount of light making it through to the sensor. This is accomplished via the somewhat mis-labeled "analog gain control" function, which increases or decreases the brightness of the LEDs providing the illumination. (This is another welcome carryover from the earlier LS-2000 and the Super Coolscan 4000 ED design.) While this increases overall brightness and depth into the shadows (on slides, that is: in the highlights on negatives), the less-dense areas of the film can suffer from a light overload. Thus, this control's usefulness will depend somewhat on the particular image being scanned. It will be most useful on film or slides that are dark overall."

[Rider]

Now, I am really confused.

It seems that what Vuescan calls "Exposure", Nikon Scan calls "Gain". Vuescan routinely ups the Exposure on dark slides whereas Nikon Scan never automatically increases Gain unless you tell it to.

I've taken a quick look at Sascha Steinhoff's book. I also looked at his website. On his website he actually recommends Nikon Scan (as of 2009) for negatives (he even calls it possibly the greatest scan software of all time or something to that effect), less so for slides because it does not support IT8 calibration (I am not sure I would ever do that; B&H does not even carry IT8 targets for 35mm). Now he has written a book on Vuescan.

I've reinstalled Nikon Scan. Now that I have Vuescan installed, Nikon Scan works under Windows 64-bit. I will be curious to see how the two compare when scanning the same slide. I would prefer to spend less time comparing and learning multiple programs, but I am curious to know whether Nikon Scan can tease out data out of dark slides without increasing the Gain. Maybe Nikon knows their equipment better than third parties.

Back in the day when I used Nikon Scan (although I had no idea what I was doing), I recall getting decent scans without much effort. It may be an overly rosy memory, unclouded by having subsequently looked at thousands of grainless nicely exposed Nikon D700 images. We shall see.

[Ken Lee]

So when scanning, do you expose for the shadows and let the highlights go, or do you expose for the highlights and then bring up the shadows in Photoshop? And what does "exposing for the shadows" mean in this context--how do you know when you've exposed for any thing?

One approach that seems to work well, is to perform as many adjustments as possible during the scanning phase, rather than as a series of "down-stream" adjustments in an editing tool. This is an extension of the approach which prefers optimal exposure and development, rather than corrections in the darkroom. (It's easier to do with B&W than Color film of course.)

It may be an extreme view, but every adjustment is a destructive adjustment, and analog adjustments are preferable to digital ones. Therefore it's best if we can treat every negative individually, and try to accommodate the SBR of the film once and for all, when the conversion is made from analog to digital: it's the last chance we have to interpret the smooth analog data.

In cinematography, where everything is very expensive, and the medium is rather finicky, they try to make as many corrections as possible before the exposure. They call this... "Lighting".

So to address the question: when scanning, don't just accommodate the shadows or the high values: try to nail the whole thing, right then and there.

Although it's about Epson drivers, you might find this article helpful. (I've been meaning to write a similar article with Vuescan, but keep procrastinating )

[Rider]

I've read the Vuescan Bible, and it was very helpful. Now I have a much better idea of I don't know how to do--namely decide when to use Multisampling, Multiexposure and Fine mode, and when to override the automatic exposure setting when I don't like.

I've scanned a handfule of rolls--all exposures, good and bad, to try to learn more.

Ken--I am fascinated that you can actually try to nail down the whole right then and there? That is so old school!

I am reading your article next.

[Peter J. De Smidt]

I agree with Ken that ideally we want a well exposed and processed negative or slide that requires as little adjustment as possible to get what we want. This is the way to the highest quality. We also want the hardware settings on the scanner to give us the best info that it can capture from the film. But for software adjustments Photoshop and some other imaging tools provide much better options for manipulating the captured data than most scanning software. In particular, there are better previews, finer adjustments available, masking, layers.... Sure, I can apply a curve in Color Genius, but it is much less user friendly than doing so in Photoshop. But this is something that we should each test for ourselves and our particular equipment.