LaserSoft Imaging

I'm new to this so my terminology may not be correct but I will try to explain my issue as best I can.

Using SE+ Demo Win XP, HP 4890 flatbed, Elements 4.0. sRGB colorspace through to printer.

Using negafix and controls I can get a balanced prescan that looks as good as I need it to look. My issue is thus. Depending on the resolution (size) of the scan I make I get varying degrees of color cast and contrast changes.

Example: I set up a negative of Kodak MAX Versatility 400 using the Fuji Superia 400 profile, which looks best to me, in Negafix and a few tweaks and I get a nice looking prescan. I used no filter, AACO, SRD and ACR are all off, single-pass and I'm scanning a 48->24 tiff to Elements 4.0.

As a trial I scanned at each of the optical resolutions of my scanner those being 50, 100, 150, 300, 600, 1200, 2400 and 4800 at 100% scaling each resulting in increasingly larger files and pixel count. I used the exact same settings for each scan only changing the output resolution and pushing the scan button between scans.

As compared to my prescan the lowest resolution (50dpi@100%) shows a distinct magenta cast and underwhelming contrast, as I move up the scale each preceding scan becomes closer to how I set up the prescan, at 600dpi@100% the scan and my prescan become closely matched, as I scan at resolutions above this the scans take on a green cast and becomes increasingly contrasty and blown out until at 4800dpi@100% (my max optical) almost all the highlights are blown and everything is green.

One thing I noticed as the scans become larger and increase in pixel count the percent of pixels taken up by the demo watermark becomes less. In the smallest scan the watermark covers fully 1/4 of the scan and in the largest scan they are barely visible. Could this somehow be affecting the balance of the scans? I also examined the scans in in the Windows picture and fax viewer and they had the same issue.

This program is wonderful and once the conventions are learned it becomes intuitive to use, at least at a basic level as I used it. It scans excellent prints and I am amazed how good they look. I love the multi-scan, the shadows look great! It's like having a new and better scanner.

I know I haven't paid yet so I am not entitled to support but I will buy it if I can somehow get this negative problem ironed out. Well I have to buy this program anyway because the software that came with the scanner is really bad. I intend to use Silverfast SE+ to scan family prints and negatives if the prints are lost or in poor shape so I can share them with family on recordable media. I don't expect to have to do too many negatives.

I thank you and if someone can help here I thank you more.

What happens when you prescan at full optical resolution and then downsample the final scan ? This could be a sampling issue - if you are prescanning at 600 that means the scanner is averaging a 8x8 set of pixels to arrive at an average sample value and then at the other resolutions different averaged blocks of pixels are getting the adjustment applied. It is not unusual to see small defferences when prescanning and adjusting based on a limeted set of pixels, but not the extremes you are seeing. Does this happen when you use the HP software ?

Simple workaround ? - prescan and adjust at the resolution you intend to use for final scan.

another idea - download the Ai demo and see if you get the same result. Could be an SE issue.

It's NOT Silverfast, it's my scanner and only negatives. The HP software does it too. Slides and prints look very near the same at all pixel counts/file sizes.

I spent the better part of today messing with this. Here is what's happening although I don't know why.

When doing negatives there is one scan resolution that looks like the prescan, it doesn't matter what combination of dpi and scaling I use. It's seems totally related to pixel count/file size i.e 600dpi@100%, 300dpi@200%, 100dpi@600% all give the same results as they are the same size of course. If you go smaller than the sweet spot shadows black out and you get one type of cast, if you go larger highlights get blown and you get the opposite cast. The type of emulsion changes the size of the sweet spot and the colors and strength of the cast but it still behaves the same. The Kodak Max Versatility 400 is very sensitive while Kodak Gold 100 gives more leeway but with either you go to far from the "sweet" resolution with this scanner and you get color casts and contrast changes. So I'll just have to deal with it.

Now slides are sometings else, awesome is the only word I can think of. I scanned some Ektachrome I took of my kids 25 years ago and they were beautiful. Colors were saturated but not blown, contrast was perfect, low shadow noise and decently sharp and this was single pass without GANE. I needed very little messin' around. I made some prints from the slide scans and they look pretty darn good, not art level but better than some snapshot prints I have.

The same with prints, they are as good as can be expected and better than I thought this scanner was capable of and the slides are also much better than I thought I could get.

Silverfast makes THE difference.

I'm sold.

Thanks for answering.

Maybe they are being scanned as positives by both ? just a wild guess. Maybe you could scan them as positives and then invert and cancel the mask with curves ?

I reinstalled the HP stuff and it's still there. I downsampled and upsampled same thing. When the negative scans reach about 3.5 megapixels or so in size all of a sudden I get a large increase of pixels to the right on the PS histogram, the left and center flatten out but don't go away.

I made a large scan of the negative as a positive, inverted it, took a scan of the emulsion mask from the strip edge, made a color fill layer using that color, reduced the opacity, set gray and white points and it came out better than those overblown scans. I'm surprised it worked since I don't know squat about PS Elements..yet. I just started digital photo and scanning last month. So scanning negs as positives doesn't show the problem.

If it's a hardware problem why do slides and prints work so well? Do you know if Silverfast uses the HP drivers at a low level. maybe it's something there if the same drivers are used?

Thanks again.

slides and prints are positives. It sounds like the software is not doing the removal of the emulsion base correctly.

check out this guide for SE and negatives

http://download.silverfast.com/misc/Dav ... Eguide.pdf

and Ian's

http://www.computer-darkroom.com/sf5-ne ... egafix.htm


The appearance of pixels on the RHS of the histogram at a fixed number of pixels is suspicious to say the least. Almost sounds like a mismatch in data sizes between the programs. i vaguely remember some problems a while back in early version 6 that did this, but it was fixed. Does the same thing happen at 24 bit scan 24 bit output in negative mode ?

Submit a bug report to SF with this forum topic as a reference.

another possibility :

" However, there is one thing you should be aware of: When scanning, the SilverFast- settings are carried over to the next image. So if the colors in the first image were accurate, and all following images were purple, it might me that a setting from the first scanned image got carried over. To avoid this, please click the small red trashcan- symbol in the pre-scan window before you do the next scan. This resets the program and should get you accurate colors for the next scan."

and another :
"I know just what he's talking about. I tried to scan some negatives with my Espon 3200 perfectiona, Mac G5 dual, and Photoshop or Silverfast interface. The negatives/image had a heavy orange cast, grainy, and lots of defects showing up everywhere.

The problem? I was scanning in Reflective mode with the white background of the flatbed causing a problem with white balance. (I suspect the defects I was getting were from the white backing for the flatbed scanner)

The solution?
Remove the white backing of the scanner as you would for doing slides. Use Silverfast SE (other whatever) in transparency mode, use the negative setting, and bingo, the negative scan worked prefect.

Ok here are some links to small jpeg's of the scan and the histogram, I get the change when the scanner jumps from using 600dpi optical, as determined by holding Ctrl while mousing in the dpi window, to 1200 optical.

What I did is do a series of scans using the same negative just changing the pixel width and let the program figure out the rest sizewise. I started at 100x??? and increased by 100 for each succeeding scan. I was looking for the point when the scans went from being close to what I set up to being blown out and color cast. I did this several times to account for variations and errors. Any thing above the changover point only gets worse by degrees as the scans get larger.

Here is a mini-jpeg of a 800x507 scan, it is close to what I set up, look at the histogram.

http://marauder.millersville.edu/~cmsharwa/800x507-2.jpg

Now here is 900x571, boom the histogram has moved right and everything went nasty. The only difference in setting from the above is me adding 100 to the pixel width and when Ctrl holding in the dpi window the above said 600 optical and the below said 1200 optical.

http://marauder.millersville.edu/~cmsharwa/900x571-4.jpg

Note: it does this only with negatives and with BOTH the HP program and Silverfast, so either it's a hardware issue or a low level driver issue or so I am leaning. Slides and reflective media show none to very slight variations across the entire optical range of this scanner, HP Scanjet 4890, and look very good using Silverfast.

i would definitely submit a problem report.

Dear DonS,
that sounds like a strange Problem to me...

This obviously could be a hardare Problem, do you have the chance to try this out on a different 4890 Scanner ?
(a colleague will try it out on our 4890 Scanner asap ),

IF we can reproduce it on another 4890 Scanner then this is certainly a first time appearance of this kind of "flaw", and we have to think about if/how we can compensate for the color drift in SilverFast.

IF it cannot be reproduce it, it surely is a Hardware defect...

Are you 100% sure that it only occurs in Negative Mode ??

My line of thought would be that it might be also there in positive transparency mode, but that the ( very steep ) negafix curves amplify the problem to an easy visible scale...

Well, i keep you posted after we gave it a try, until that it would be nice if someone else with a 4890 could verify that too...

Greetings,

Nils Heidorn. R & D, LaserSoft Imaging AG

Nils, here is something you might find interesting.

If while scanning negatives I use the auto-correct functions of the HP software it overcorrects in the exact opposite direction of the problem I have above, that is, the prescan and scans below 1200 optical look underexposed and magenta, therefore scans in the range where I have the problem look correct. This is one reason I never used auto-correct with the HP software, the prescans looked awful.

I had wondered why the HP software auto-correct function was so far off when scanning negatives and just chalked it up to the usual so I quit using it. Now I also see that when scanning slides and prints the HP auto-correct is in the ballpark, not perfect by any means, but a useful starting point.

If this can be reproduced on another 4890 it just may be that HP is aware of the color and lighting shift when scanning negatives and is correcting for it in their software but not at a driver level. Since for most uses you don't get a useful size from a negative scan at the the lower resolutions and wouldn't scan that small of a size they could do that. I don't have another 4890 to test this myself.

Are you 100% sure that it only occurs in Negative Mode ??

Yes in both scanning interfaces the difference between low and high resolution scans using slides and prints scan is negligible. In thumbnails it can't be seen at all and at larger sizes only a direct side-by-side shows very slight differences.

I want to thank you and degrub for the responses.