D4000 operation

[Rich14]

Hi all,

I've operated D4000s for years using Howtek's Polaris software on a G3 under OS8. I've just moved a D4000 to a G4 where I'm running OS X and Lasersoft's Silverfast Ai software. In doing so, I'm recalibrating a lot of parameters and I'm wondering about some of the controls that any of the various Howtek control software has on the scanner.

Does anyone know whether the "gamma" settings or other color or sensitivity "adjustments," either for all 3 RGB channels simultaneously, or each of the individual color channels effect the actual sensitivity of the hardware, or are these controls simply applying tone compression/expansion to the raw scan data itself after it has left the analog to digital conversion.

In other words, is the photomultiplier circuitry or the analog to digital circuitry controlled by the software in any way?

I prefer to use scanner control software to do nothing but mechanically operate the scanner (spin and position the drum, focus, etc) and get the raw scan data into Photoshop as a 16 bit TIFF, where I apply all necessary image manipulation (ICC input profile asignment, highlight and shadow, etc., etc).

If scanner software can control the scanner in terms of RGB balance and "gamma" before the formation of the image, I would much prefer to set up the scanner that way. But if the software simply applies adjustments to the raw san file before it spits it out to Photoshop, I'd rather accomplish that in Photoshop.

Does anyone have any info on this?

Rich

[Bruce Watson]

In other words, is the photomultiplier circuitry or the analog to digital circuitry controlled by the software in any way?

In fact, the control software should be making inputs to the log amps, setting their range. The point is for you to define your black and white points for each channel. The software tells the scanner to scan specifically for this density range, and spread this density range over your 12 bits of A/D output. If you set your black and white points properly, this should maximize image quality and show up in your final prints as smoothness of tone and especially smooth transitions of tone. This is a good thing.

If you want to boot the control software out of this picture, you are basically setting your black and white points as the limits to what the scanner can do (a density range of what, 3.6 or 3.8 maybe?). If your film doesn't actually cover that range, lots of your bits will be empty, effectively compressing your image into a subset of the available bits. If the compression is severe, say your image only fills half the bits, you've got the effect of a 6 bit scan and you'll surely see it in the final print. This is a bad thing.

Clearly, this is going to matter much more to negative scans than it will to tranny scans due to the lower density range of negatives, particularly B&W negatives. So, if all you are scanning is trannies what you suggest will probably work just fine. If you are scanning B&W negatives however, I suspect that you'll be less than pleased by the results.

[Larry Gebhardt]

I use the levels tool in Silverfast for the D4000 and get good results. I adjust each channel's black and white point to make the film edges black and the highlights to balance the image, usually just backing off a bit from the clear part of the drum. I then fine tune this in Photoshop.

Aztek claims Silverfast doesn't adjust the gains on the analog to digial circuit (as I understood them on the phone), but I seems like it must because you get good tonality out of both slides and negatives. I really haven't noticed much difference between the quality of Silverfast and DPL for my way of working.

[Rich14]

Bruce,

Thanks. I had hoped the controls were adjusting the hardware as you describe. I wonder if there is any way to test? I agree that setting highlight and shadow points in the scanner hardware results in the theoretical maximum 12 bit depth being assigned only to the image's dynamic range, not to the scanner's.

No argument, whatever, that starting with 4096 gray levels per channel is a good thing and one doesn't want to knowingly throw any of those bits away. But there are several other realities to consider.

One is that input profiles are valid only under the specific scanner settings for which they were made. Don Hutchinson, for example, advocates profiling the scanner set to a gamma of about 2.6, but otherwise, "wide open." That is, with the highlight and shadow sensitivities at their respective limits. Scans are then made under identical conditions, the profile is assigned, then H&S are set.

If one is going to set the scanner's upper and lower sensitivity controls uniquely for each scan, to be very accurate, one really needs to have a profile for each of those potential scanning situations - pretty impractical.

My practice is to set the scanner H&S controls near, but not too close to the ends of the image histogram. I bring the "aproximate" scan into Photoshop for accurate final setting. I just don't like the crude (compared to Photoshop) scanner controls. That's what I meant in my first message about using Photoshop to actually set H&S.

I have six sets of such scanner H&S placements and ICC profiles for each:

0, 255;
32, 255;
0, 240;
32, 240;
0, 224;
32, 224

In most cases the scan's histogram is much closer to the shadow end of the window than the highlight end.

I am not sure of this, but I believe your "math" regarding loss of bit depth is wrong about a scan made at the scanner's extremes of H&S and the image data not filling the histogram window. If the image fills half the window, the scan contains 11 bits, not 6 bits.

The window is a linear representation of 4096 levels = 12 bits. Image data occupying half the window represents half of 2048 = 2048 levels = 11 bits. An image bit is a terrible thing to waste, and I never would throw away 2048 levels, but a scan made under these conditions certainly does not contain only 6 bits = 64 levels per channel.

I realize I am not getting all 12 bits using the technique I've described, but I am getting almost all 4096 levels and I believe the profile information is more accurate this way. But I'm not sure which is more important. Obvioulsy I lean in the direction of the profile.

Rich

[Rich14]

Larry,

That's disturbing Aztek would say that about Silverfast.

Disturbing that they would lie to malign other's software.

Disturbing if it is true, as that would mean Silverfast throws away data.

Rich