Vuescan/Canon 9950/Dmax

[Ed Richards]

Paul has put some great info on his site about the scanning software and Dmax:

http://www.butzi.net/articles/scannersoft.htm

Since Dmax is as important as resolution for B&W, I wanted to see what a new generation consumer scanner could do. To this end, I scanned a 21 step Stouffer step wedge - same density range as the 31 step, just steps of .15 rather than .1 - multiply my step numbers by 1.5 to compare with Paul's.

http://biotech.law.lsu.edu/stepwedge01.jpg

I used Vuescan and took the raw data file and adjusted the gamma in Picture Window Pro, then slightly adjusted the low end contrast with curves to better separate the lower steps. (Remember, this is a positive image, so these would be the highlights on a print from a negative.) This was scanned at 4800 DPI, then sampled in Vuescan by 6, to give a final file of about 20 megs. No other manipulations.

I think I have clearly useful data to at least 18, and perhaps 19. While you can see 20, I am not sure how good the data is. Since there is no detail to resolve, it is hard to say what this means for highlight detail, but it is clear that the scanner can resolve more in the highlights that I was assuming. You lose a few steps by letting the software scale the data. I am sure there is noise, but highlight noise on 4x5 is not a big issue to smooth out. (I did the same with Siverfast and the results look about the same - while I think the Vuescan file has a little better Dmax, that might be related to the extra steps I have to go through to process a Siverfast file since is not flat tiff.)

This is easy to do, and since there is no dmax test on the scanner test page, perhaps others would like to add their scanners. The Photographer's Formulary (www.photoformulary.com) sells the step wedge for a lot less than anyone else I found - $26.

[Paul Butzi]

I thought I had posted a response but it isn't here.

This is very interesting. I'll have to try similar adjustments on my 1800f to see what sort of results I get.

By the way, the cheap way to buy the step wedges (both the 21 steps of .15 density that Ed uses and the 31 steps of .1 density that I use) is to get them directly from Stouffer Graphic Arts (www.stouffer.net). They want $21.50 for the 21 step version and 31.20 for the 31 step version.

[Ed Richards]

One caveat - I am not sure how to do the adjustment in Photoshop. PWP has much more powerful tools for handling exteme files like raw scanner files.

[paulr]

Are you talking about scanning negs or prints?

I've found d-max to be a non-issue with scanning b+w negs. Even my densest negs are well within the range that an epson 4870 can handle. Maybe if you have negs that were developed for printing out processes it would be a different story. But I just scanned some negs dating to when I had no idea how to process film (some of these are pretty opaque!) and had no problem getting all the highlight detail.

b+w prints are a different story. The shadows can be very dense, and my scanner doesn't see all that far into them. And of course chromes are also a bigger challenge.

[Ed Richards]

Chromes are certainly a bigger problem. It might be worth trying a 48bit raw file for a dense chrome , it might give you a little more range. If I can dig up a dense chrome, I will give it a try.

[tim atherton]

A scanner doesn't need near as much Density Range for neg film as for trannies. The clear base of the tranny means the scanner has to capture a wider Density range than it does with negative film. The DMin on the clear base is obviosuly much lower than the coudly base on neg film (while the DMax is obviously roughly the same - black is black) - so the range between DMax and DMin is less for negs than trannies

In vuescan on even a moderately good flatbed film scanner, you can see in the histogram you can nearly always capture the full range of neg film with room to spare on either side, whereas with tranny film, it's oten squished up to one side with no tail off

[Ed Richards]

That is why I used to the step-wedge, it is just the chrome from hell as regards Dmax. The histogram looks as you describe but the raw file does have enough data to sort out that tail.

[Brian Ellis]

Ed - I hate to reveal my ignorance but I don't see what your step wedge tells us about dMax. It obviously tells us that you can obtain excellent tonal separation with your scanner and software plus a little tweaking but absent a densitometer reading what are we learning about dMax from it?

[Struan Gray]

The key measure for scanning negatives is signal to noise. As Paul said, there are no scanners designed only for negs, and any scanner that can handle the on-film densities of a slide can easily cope with any reasonable negative. The problems crop up when the scanner and/or its software increase the on-neg contrast, either via an analogue gain in the CCD readout circuit, or after digitisation. If the scanner has poor analogue noise performance, or digital algorithms designed for speed or cheap processing, you start getting artifacts in your scans - speckles, posterisation, blotchy patches etc.

One clue is to look at the histogram of the step wedge scan. In an ideal world, with a step-wedge made from clean, homogeneous ND filters, you would see a progession of spikes, one at each of the step-wedge density values. In real life you'll seee a progression of sugar-loaf peaks. Their width, shape and how much they overlap can tell you a lot about the scanner's noise performance, and the reasons for any departures from the ideal.

[Ed Richards]

> I don't see what your step wedge tells us about dMax

That it probably does not matter as much for scanning negatives as I thought.

> The problems crop up when the scanner and/or its software increase the on-neg contrast ...

I would love to see a drum scan of a step wedge to see how close it comes to ideal - it would be interesting to know what we are starting with.

My interest in what density range a scanner can handle is based on getting the best signal to noise ratio. Ideally your data will be spread across the entire range of the scanner's linear response, which will minimize the artifacts from crummy electronics, dirty glass, cosmic rays, etc. The more compressed your negative is, within that range, the more accurately the scanner needs to differentiate the data, and the more the scanner errors affect your output.

Looking at my histogram, I get about 17 good spikes and then the densities start to run together at the dense end. While I can visibly separate another step or so, I doubt that fine detail would resolve there.

[Struan Gray]

Looking at my histogram, I get about 17 good spikes and then the densities start to run together at the dense end. While I can visibly separate another step or so, I doubt that fine detail would resolve there.


That's what I would call my system 'toe'. If I can, I expose to put shadow detail above that, and see where the highlights end up. Use N- development or low-contrast film if they end up with too much density / halation / colour shifts. I've yet to end up in the latter situation.

[Ed Richards]

> That's what I would call my system 'toe'.

I think we are reversed. I get sharp peaks on the shadow (less dense negative) end, and bunching at the highlight (more dense negative) end. If there is enough exposure to put some silver down in the shadows I can get a good read of it. It is the dense highlights I cannot separate as cleanly.

[Struan Gray]

The peaks bunching together is a sign that your scanner firmware or imaging software is applying a gamma function to the linear scan. This is a whole can of worms in itself, and by no means necessarily bad, but if you have a problem image where you are losing needed highlight detail at scan time, the solution is to scan in as many bits as possible, and as linear as possible, and only convert to an 8-bit gamma-coded file once you've got the detail how you want it.

In my post above I was thinking about other issues: mostly how signal to noise gets so bad in the shadows that you lose the whole LF grainless look. Sorry if unclear.