Most Compelling B&W Scans on 4990 w/48 bit Scan and PS Conversion?

[Kirk Gittings]

I'm sure that works well for you.

Another point of view. Any scanning software worth its salt applies curves and clipping adjustments etc. on the raw file delivering a full histogram for import into PS. Silverfast talks about this in their manual. Who knows about Epson scan? If you wait till you are in PS to do your clipping, steep contrast curves etc., you are throwing away allot of information right off the bat vs. allowing the software to apply these adjustments to the raw file in a good scanning software. Depending on how much you work your files this may not be important. If you work the files allot it does.

[walter23]

I think you are making this too complicated and wasting time. Scan in 16 bit and move on to manipulating the result to look the way you want. If you have to jump through many hoops to get a good scan from your B&W negatives you might want to re-visit how you expose and develop your film. If you present a good piece of film to your 4990, it is capable of making a good scan within its optical and mechanical limits. Make your film fit your scanner. Just like making the film fit your paper in the darkroom. After that its setting black and white points and a curve.

I mostly agree. However I have noticed a slight spatial shift between the R, G and B channels in my scans (at least, at very high resolution, like 4800 DPI) on a V750. This doesn't result in anything that would really show up in most normal print sizes, but, for optimum sharpness of a B&W image selecting one of the channels (probably whichever one happens to have least noise) eliminates this intrinsic offset and can give you a sharper image than just blending the three channels into a single image.

[sparq]

Any scanning software worth its salt applies curves and clipping adjustments etc. on the raw file delivering a full histogram for import into PS. Silverfast talks about this in their manual. Who knows about Epson scan? If you wait till you are in PS to do your clipping, steep contrast curves etc., you are throwing away allot of information right off the bat vs. allowing the software to apply these adjustments to the raw file in a good scanning software. Depending on how much you work your files this may not be important. If you work the files allot it does.

I am skeptical about that. Does anyone have a solid proof and a good technical explanation? Does it mean that the scanning program works with more than 16 bits per channel when applying curves? I don't think so, I am not aware of any consumer scanner that supports more than 16bpc in HW. What other information might get lost during the encoding of an in-memory image to a 16-bit per channel tiff file?

[Henry Ambrose]

About the only thing I see wrong with the Epson software is that due to a bug it limits the file size you can produce on a 4990. As far as I know Epson never bothered to fix that. For B&W EpsonScan works just fine and its not too bad for color. SF does give lots more control and is quite good with color negative film. But you don't have to have it. Being able to pick out a channel is a nice feature y'all found in SF - assuming its really needed.

All this picking channels and special workflows have been kicked around for at least 10-12 years now. They're not new. While I suppose there may be some advantage to working around your machine and software flaws, I'm still of the opinion that given the possible mechanical and optical quality of a consumer flatbed you're not gong to make it into a drum scanner by some high-wire act. Feed it good film and you can get good scans. Develop really good PS skills and you can make fine prints from these scanners and you don't have to spend lots more money for scanning software.

[Henry Ambrose]

Some scanners use a less than 16 bit analog to digital converter. Often 12 or 14 that might be then sent out as 16. There's lots of claims made up for marketing. Read the specifications closely and with a doubting eye.

Does the scanner work on the analog signal in any fashion? Does the software work in 16 bit if its passed info from a 12 bit converter? What is the difference in 16 bit adjustment in PS or in scanning software? Is there any? About the only way to get real answers to those questions is a long conversation with the engineering team who did the particular machine.

[Kirk Gittings]

Sparg,

I can't directly answer your question on consumer scanners. SF of course makes software for high end drum scanners (most of whom are actually 14 bit I understand) and makes the same claim on consumer scanners. I lent my SF book out but it states that major adjustments are best made in the scanner software where the mosst raw information is available. It is very easy to show the benefits of this workflow on 8 bit files. You can test this easily with an 8 bit scan. Scan one flat and unclipped into PS. Then clip it and apply a strong curve. Do the same adjustments to one while in the scanning software. See how the histogram breaks up when the same adjustments are done in PS? A bit harder with 16 bit, because it takes allot to break the histogram, but if you work files as much as I do, it becomes most obvious far down the workflow by some noise enhancement (noise is less in the transition tone areas if you make the major gross adjustments in the scanning software). Scanning in 16 bit lessens the necessity of this workflow but doesn't eliminate it and with drum scans or high end flatbeds, that generate very little noise, it is even less necessary.

[Kirk Gittings]

Henry,

I'm still of the opinion that given the possible mechanical and optical quality of a consumer flatbed you're not gong to make it into a drum scanner by some high-wire act.

If you can find one person here suggesting anything close to this I'd be very surprised. In fact the opposite has been stated so many times that people are tired of hearing it. However, having taught scanning for many years now, I will guarantee that a careful work flow, paying attention to details, such as those discussed here, produces significant better final results than the alternative. These are the kinds of questions that people concerned about craft ask.

Don't I remember you saying you were going to buy a high end flatbed scanner because you were tired of apologizing for your scans?

BTW your site looks great. Allot of strong work there.

[Henry Ambrose]

Kirk, Thank you for the kind compliment.

I guess what I'm thinking when I write about this stuff is that some folks are going to find this thread and think they have to buy lots of extra software and learn exotic techniques or they're doing something wrong. I find these threads to be a little like "is the XXX Schneider really better than a YYY Rodenstock?". One could almost draw the conclusion that just because they now own the software or because they now use a more complicated workflow their pictures are somehow better.

I have full Silverfast and really enjoy it for some things - its just great for color negatives. I also use EpsonScan for my 8x10 B&W because its simpler and faster and I have no apologies to make for the results. The 3X reproductions are glorious.

I've been scanning since desktop scanners have been generally available and I've owned and used lots of them. I did my first flatbed test against a drum something like 12-15 years ago. I've used plenty of drum scans and owned and run scanning cameras making thousands and thousands of image files. I've been around the block a time or two.

And yes, I'd love to have a high end flatbed like an iQSmart. But its not that my 4x5 Epson scans lack anything for general commercial purposes. They are already gross overkill for anything that is printed by offset such as a magazine. I'd like it almost as much for the speed of workflow as for the rare occasions when I need to make a really big reproduction. Now I go pay for a drum scan and it'd be great to keep it in-house to save the drive to the lab and the time. But from an economic stand point I can't quite make sense of the cost.

If you pixel peep an Epson scan and a drum scan you'll see things that distress you. But if you print them on an inkjet at a size that is within the capability of the flatbed you won't see any difference. Push the limits and then you start wanting to jump through hoops.

Anyway, I only want to present an argument for simplicity in the midst of this thread. Or perhaps offer an alternative view of getting the work done while not taking any shortcuts. Or at least no more shortcuts than you take by buying and using a consumer flatbed. And I don't mean to say anyone is bad. If I've given that impression I am sorry. I'm just here for the discussion.

[Tyler Boley]

This conversation seems very familiar, I believe we had it recently on another forum. My findings after extensive testing were that making as many moves as possible in the scanner software is only advantages in terms of loss when you are having it export in 8 bit.
In fact, when exporting in hi bit, I found much less loss saving out totally raw and untouched, and doing all moves in Photoshop, with B&W negs. That loss may or may not be relevant to different users. Since files come in from clients occasionally with missing levels, and all our efforts here to be well calibrated print those tonal breaks nicely to paper quite visibly, I think this is an important issue for B&W digital workers. With color negs, they are so wacky that if the scanner software has good color neg setups it's more convenient to make the big moves there. With transparencies I happily adhere to a color managed workflow with good profiles, there it is important to leave the controls alone, as they will make the profile less relevant. Edits should be made after conversion to a perceptually uniform working space.
Making initial big moves in the scanner software would be very important if exporting 8 bit, as those moves will be made in the higher bit depth of the scanner space. Also there is a convenience factor with all of this that must be considered, and some of this stuff is just conversational.
Tyler

[John Berry]

I usually scan my pyrocat negs in 48 bit RGB. I use channel mixer mostly. If you like one aspect of one channel and another part of another,Drop the channel on as a layer and mask till you are content. I'm thinking of one shot I had that had detail only in the blue channel in the highlight of bleached wood. Lot's a ways to play with this toy.