Bit depth and banding (stepping)

[alanrockwood]

I've been using a V850 Pro with both Epson Scan and SilverFast, started with Epson until I got the time to explore SilverFast, which I really like.
However - In scanning some 4x5 negs, black and white, which have large open sky areas (no clouds, the yellow filter kind), which attempting to deepen in value, I get banding. Some info on what I'm doing:
I adjust the scan settings to very close to final tonal values
Saving in Tiff 8bit grayscale (SF doesn't allow 16bit, for some reason)
Open the Tiff in raw, if any other adjustments are needed.
Open in PShop and make edits if needed.
I'm working on one with a large sky area which I want to deepen just the sky, doing so with a gradient mask, in PShop, to the sky area only. When I saw the banding the first time, I reopened the Tiff in Raw, deepened the whole thing, opened a second copy of the file, copied the darker image, and pasted it into the original file as its own layer, applied a gradient mask to that image layer, just to hide the part I don't want to see. The effect is right, but the banding is there anyway, just because the sky has it's own gradation, I guess.
I haven't had this happen with smaller formats, maybe because grain masks or prevents the banding.
I'm thinking this banding would be minimized in a 16 bit file, but SF won't allo that (it's selection is 16->8bit).
Epson Scan allows a simple 16 bit choice in their popup menu. (I've used it before)
Any ideas about this?

A couple of thoughts: What dpi are you using? In principle if you want to reduce the risk of banding it is best to use higher dpi. Interpolated dpi doesn't count. Just real hardware limited dpi. The explanation of why this is so is somewhat involved and rather subtle, but I will explain it if anyone wants to hear the reason. I realize that the scanner resolution is limited to about 2400 effective dpi, so a lot of people will scan at low dpi. However, there are other factors in play here that makes it desirable to scan at higher dpi if you want to reduce the likelyhood of banding.

The other thing is that if you scan in 8 bit mode then before you do any processing of the stored file you should convert to 16 bits.

Those two things should reduce or eliminate the risk of banding.

Also,

 

[PhilBurton]

I use Lightroom extensively - I just need to figure out why it won't sync my photos properly . The reason for saving RAW from SilverFast is that the less expensive versions of SilverFast do not allow you to save as 16-bit TIFF files. Scanning in RAW is a work-around for this problem.

+1

 

[PhilBurton]

So your belief is that general purpose software is just as good as purpose-written, optimized software?

That's an... interesting viewpoint.

Yes, you probably can do the same sort of editing with Lightroom or Photoshop-- but you'll work considerably harder at it.

@grat

Do I understand you correctly? Are you saying the same sort of corrections are easier to do in Silverfast or Vuescan than in Lightroom or Photoshop?

IF, IF, IF my understanding is correct, another point of view is that the editing features in the Adobe software are much better than the ones in any scanning software. And, you want to keep all your edits in one place.

 

[PhilBurton]

A couple of thoughts: What dpi are you using? In principle if you want to reduce the risk of banding it is best to use higher dpi. Interpolated dpi doesn't count. Just real hardware limited dpi. The explanation of why this is so is somewhat involved and rather subtle, but I will explain it if anyone wants to hear the reason. I realize that the scanner resolution is limited to about 2400 effective dpi, so a lot of people will scan at low dpi. However, there are other factors in play here that makes it desirable to scan at higher dpi if you want to reduce the likelyhood of banding.

The other thing is that if you scan in 8 bit mode then before you do any processing of the stored file you should convert to 16 bits.

Those two things should reduce or eliminate the risk of banding.

Also,

@alanrock

If you scan at 8 bits, then even if you store and save as 16-bits, you only have 8 bits of data to work with. I think it's necessary to scan at the max bit depth allowed by the scanner.

 

[grat]

@grat

Do I understand you correctly? Are you saying the same sort of corrections are easier to do in Silverfast or Vuescan than in Lightroom or Photoshop?

IF, IF, IF my understanding is correct, another point of view is that the editing features in the Adobe software are much better than the ones in any scanning software. And, you want to keep all your edits in one place.

No. I'm saying that the software and algorithms inside SilverFast are specifically designed for scanning. Take negafix-- If you ever get a chance to look at the "expert" mode, you'll see it's applying not just level adjustments, but actual color curves and some mapping. All of these features are available in Lightroom, but I can't click "Fuji --> Pro --> 400h (6x6)" in Lightroom.

The reason that Adobe (or Affinity) is a better suite of editing tools, is that SilverFast isn't an editing package. It's a scanning package, that allows you to get a head start on your image editing-- which, as you say, should be done in an editing package. Scan with scanning software, edit with editing software-- I'm not sure where that became confused.

To me, the most useful parts of SilverFast are:

• negafix: I wish they'd come out with some more profiles, but it's nice to have your negative inversion "just work". By the way, trim out the white borders around the images for best results.
• histogram before scanning. I tweak the histograms to 1) avoid clipping and 2) set the histogram to match the dynamic range of the image. If I'm scanning 8 bits per channel, I want 0 to be the darkest pixel, and 255 to be the brightest: If it's 16 bits, then I want the brightest spot to be 65535 (check with densiometer function).
• multi-exposure: For negative film especially, it should give a better dynamic range. I need to pull out one of my "trouble" negatives and re-scan to verify that, though.

Everything else? iSRD, USM, WTF, SRDx, GANE, AACO, BBQ**... I turn those off. I used GANE (noise reduction) once on a very, very bad negative, and it probably helped.

After that, all the work is done inside the editor of choice, which for me right now is Affinity.

** It's possible I made some of those up.

 

[shutterfinger]

** It's possible I made some of those up.

I think you left out MT compensation. (Miller Time but may be Bud compensation)

 

[alanrockwood]

@alanrock

If you scan at 8 bits, then even if you store and save as 16-bits, you only have 8 bits of data to work with. I think it's necessary to scan at the max bit depth allowed by the scanner.

If the noise is comparable to the ADC step size in the scan then you won't see banding if you follow the procedure I mentioned (i.e. converting to 16 bits before processing the image further.) For the purposes of this discussion, film grain is a form of noise, and so is sensor noise.

It might be useful to look at post #53 in the link https://www.photrio.com/forum/threads/grayscale-or-rgb-for-b-w-negs.176653/page-3.

This is the same concept as dithering in signal processing, which is a well-known method of getting rid of the ADC step size problem in converting analog signals to digital signals. The tradeoff is that dithering can introduce noise, but if the signal already contains noise (e.g. grain and/or sensor noise) then you don't need to add any noise to do the dithering because it's already built in, and you haven't actually lost anything .

 

[Alan Edward Klein]

So your belief is that general purpose software is just as good as purpose-written, optimized software?

That's an... interesting viewpoint.

Yes, you probably can do the same sort of editing with Lightroom or Photoshop-- but you'll work considerably harder at it.

It's an viewpoint based on practical experience and results. I've scanned 35mm, 120, and 4x5 film with both V600 and V850 Epson scanners and worked with Epsonscan for over ten years. And I've used Lightroom and PS Elements in post. The editing in all scan and post processing software changes colors, exposure values, black and white points, etc. Lightroom in particular is more powerful. PLus it has the added advantage of not changing a flat scan made by the scanner. Changes made by Epsonscan in other than the original capture functions just duplicate Lightroom's functions with less finesse. Plus, if you get edits wong during the scan, you have to rescan. I've never used Silverfast or Vuescan. But they edit the scanner's output as well except for special scans like multiple scans which I don't do or need. While I never used these, most people who have reported that SF in particular is a very difficult program to use, Vuescan a lot less so.

Why do you feel its harder with Lightroom or Photoshop?

 

[Alan Edward Klein]

No. I'm saying that the software and algorithms inside SilverFast are specifically designed for scanning. Take negafix-- If you ever get a chance to look at the "expert" mode, you'll see it's applying not just level adjustments, but actual color curves and some mapping. All of these features are available in Lightroom, but I can't click "Fuji --> Pro --> 400h (6x6)" in Lightroom.

The reason that Adobe (or Affinity) is a better suite of editing tools, is that SilverFast isn't an editing package. It's a scanning package, that allows you to get a head start on your image editing-- which, as you say, should be done in an editing package. Scan with scanning software, edit with editing software-- I'm not sure where that became confused.

To me, the most useful parts of SilverFast are:

• negafix: I wish they'd come out with some more profiles, but it's nice to have your negative inversion "just work". By the way, trim out the white borders around the images for best results.
• histogram before scanning. I tweak the histograms to 1) avoid clipping and 2) set the histogram to match the dynamic range of the image. If I'm scanning 8 bits per channel, I want 0 to be the darkest pixel, and 255 to be the brightest: If it's 16 bits, then I want the brightest spot to be 65535 (check with densiometer function).
• multi-exposure: For negative film especially, it should give a better dynamic range. I need to pull out one of my "trouble" negatives and re-scan to verify that, though.

Everything else? iSRD, USM, WTF, SRDx, GANE, AACO, BBQ**... I turn those off. I used GANE (noise reduction) once on a very, very bad negative, and it probably helped.

After that, all the work is done inside the editor of choice, which for me right now is Affinity.

** It's possible I made some of those up.

I just noticed this post after posting my last one. Some questions on your three points.
1. Negafix. I now have a V850 with SF 8 SE Plus and X-rite ICC color software. Epsonscan allowed me to set up the ICC profile which I can select when scanning color film. So it apparently can work with Epsonscan. I'm not sure how it works and what I have to do. Any instructions?
2. Epsonscan also allow you to select the black and white points during the scan. I often do that but frankly, you can set it at 0 and 255 and scan flat and change it afterwards in your post processing program. I haven't noticed any practical difference using either method. SOme people have claimed that setting prior to scan gives more data between the two points in the file results. But I've haven;t been able to see that and confirm that point.
3. Multi Exposure. I haven't tried it. I believe that scanner's are limited to their inherent designed dMax. You can;t get blood from a turnip. However, I'm open to experiment. Once I finish shooting my 4x5 Velvia 50, I'll be able to try SF. THe other issue I have with the SF version versioned with the V850, is that I have to buy an upgrade to get 16 bit processing rather than 8 bit. I think that's a ripoff after spending $1100 on teh scanner.that came with a crap version of SF that can;t be used without spending more money.
4. See my last post for question about ICC Negafix and using Epsonscan for same thing. HOw does it work and how do you do it?

 

[PhilBurton]

@ everyone who comments on my recent posts about specialist software. I think we are in strong agreement here. I certainly believe that it is better to use scanning (specialist) software rather than say Photoshop, for the actual scanning process. So I'm sorry that I caused some confusion.

Photoshop can "acquire" an image from a scanner, but Lightroom cannot. The latter is designed to post-process a digital image. So in a sense, both scanning software and Lightroom are specialist software.

 

[Adrian Bacon]

I've been using a V850 Pro with both Epson Scan and SilverFast, started with Epson until I got the time to explore SilverFast, which I really like.
However - In scanning some 4x5 negs, black and white, which have large open sky areas (no clouds, the yellow filter kind), which attempting to deepen in value, I get banding. Some info on what I'm doing:
I adjust the scan settings to very close to final tonal values
Saving in Tiff 8bit grayscale (SF doesn't allow 16bit, for some reason)
Open the Tiff in raw, if any other adjustments are needed.
Open in PShop and make edits if needed.
I'm working on one with a large sky area which I want to deepen just the sky, doing so with a gradient mask, in PShop, to the sky area only. When I saw the banding the first time, I reopened the Tiff in Raw, deepened the whole thing, opened a second copy of the file, copied the darker image, and pasted it into the original file as its own layer, applied a gradient mask to that image layer, just to hide the part I don't want to see. The effect is right, but the banding is there anyway, just because the sky has it's own gradation, I guess.
I haven't had this happen with smaller formats, maybe because grain masks or prevents the banding.
I'm thinking this banding would be minimized in a 16 bit file, but SF won't allo that (it's selection is 16->8bit).
Epson Scan allows a simple 16 bit choice in their popup menu. (I've used it before)
Any ideas about this?

8 bit is the issue. If you don’t want banding, you need at least 10 bits, more is better.

 

[George Collier]

Yes, I've done some comparative scanning and working in Pshop with both psd and tif formats, and have come to the realization that I need to upgrade the SF software to the AI version (as someone suggested earlier). I'm annoyed that it doesn't come bundled with the V850, but it's only $79 so worth it.