Challanging assumptions about scanning B&W

[Ed Richards]

I have been working on optimizing scanning of 4x5 black and white on a Canon 9950. I do not do color and am not addressing color issues.

Conventional wisdom, reflected in recent View Camera articles, is that you should do a lot of processing in the scanner software, and that many people think Silverfast, with its many bells and whistles, is a good scanning choice.

My prejudice is for scanning once to generate a master file with as much info as possible. I quit using Silverfast and moved to Vuescan - which gives the most control over the scanner and the output file. I oversample - I scan at 4800, even thought this does not really improve resolution much over 2400, because it gives me 4 data points to average to reduce noise. I output the raw file - the linear output from the CCD. I set Vuescan to generate the data from the green channel because that is the sharpest one on my scanner. You can change this based on your scanner's characteristics.

Unless your scanner supports different exposure values, the raw file contains all of the possible info the scanner can put out. I downsample the raw file in Photoshop or PWP, using bicubic sharpening. I shoot for about a 4 fold reduction in file size, from about 770 megs for 16 bit B&W to 180 megs. This averages out a lot of the noise and gives a sharper and smoother image than does a 2400 dpi scan that is used at native resolution. Once I have resampled, I rotate the file and flip it to get the orientation right - I scan emulsion side down, so I get a mirror image. I then invert the image and save this as my master file.

You use levels to set the white and black points and use the gamma slider to set the midtones. You can do this as a layer if you want to tinker with it later. At this point you can see the detail so it is the time to do spotting. You then use curves to fine tune the contrast.

Vuescan has a free trial - give it a run. I think this gives a very good result and makes the best of the technology in the better consumer scanners. I am sure a drum scanner does a better job, but the costs are high, unless you want to take the time and trouble to run your own. I will be doing a comparison at some point to determine what print size it takes to really see the difference. With properly exposed negatives I do not think the better handling of dmax with a drum scan will matter much, nor should the blooming you see with chromes be as much of a issue with negatives.

[tim atherton]

Ed, I basically do this but use NegPos to convert the RAW file once it's in PS (for both B&W and colour neg) - I find it does an exellent job

Bit of a learning curve - it's rather like early Vuescan - but you really need to read the manual and docs with it as it's slightly counter intuitive (to me anyway...) - as a friend said about it; " if you are a Machead who expects an iCute version, you'll have a shock, but if you thought win 3.1 was a great operating system :-) you'll be fine!"

http://www.c-f-systems.com/Phototips.html

http://www.c-f-systems.com/ColorNegs.html

http://www.c-f-systems.com/Docs/NegPos103.zip

http://www.c-f-systems.com/Docs/NegPosManual103.pdf

and if you hunt around there are some very good docs/theories on scanning color on there too

[Denis Pleic]

Ed, Vuescan does not work with my scanner (Microtek ScanMaker 6100, which is a USB scanner).

What I do is rather similar to your process, but I'm using Silverfast AI. I use "16-bit Greyscale HDR" setting, which is basically a RAW scan. I open the scan in PS and invert it (it scans as a positive when you use HDR), and start from there.

Gives perfect results - or, at least, the results with which I'm quite happy :)

Silverfast's various presets and "adjustments" never yielded the best possible results, and I realized that the way to bypass any in-scanner processing is to use the HDR scan setting in Silverfast AI. The cheaper Silverfast version, which usually comes bundled with new scanners, is the "SE" (i.e. "Lite") version, and does not provide 16-bit grayscale HDR scanning. :(

BTW, the HDR scan (TIFF) of a 4x5 B&W neg on 1600 dpi is about 100 MB.
I could use higher resolution setting (goes up to 6400 dpi), but I think it's an overkill for Web display and sending JPG scans by e-mail to friends, which usually end up being about 100-150 Kb in size.

If I were to prepare a scan for lab printing, I'd probably use the highest resolution possible (6400) in combination with HDR scan to produce a large high quality JPG.

Denis

[Kirk Gittings]

I have extensively compared Silverfast and Vuescan. SF controls are far superior to VS. The only place VS shines is in its single pass multisampling. Does that mean that you can't get a good scan with VS-of course not, but for me I need SF to get what I need from a scan.

Ed I agree with you about your workflow, but only on drum scans. With scans from say an Epson 4990 (or a Canon 9950f which I have tested extensively too) or an Imacaon, I approach this differently. I do allot of adjustment in the scan, because I want the sampling to be done with my final output in mind. For instance a flat negative that I know I want to print slightly contrasty (like with a steep midtone curve), I will do the scan with the tones pushed "in the direction" of the final out put. Because I manipulate tone and contrast allot. This approach gives me fewer artifacts in the transitionary tonal areas of the final file. This is especially true in the sky areas that I want to "Burn" in allot. The transition areas of tone pick up allot of granularity if you start out with a low contrast scan and then apply a steep curve. Look at the last example in the Scan Around and you will see what I mean about applying a steep curve in low end scanners. This is especially true with scans from like Epson 4990, Canon 9950 f or even an Imacon vs. a drum scan.

[Ed Richards]

Kirk

My point with Vuescan is that I can set it to do nothing but the scan and output the file with the unscaled CCD data. There is really no reason to treat the raw output any differently for the 9950 than from the drum scanner. There is no intrinsic contrast to the scan at all if you use the raw output. (Again, I could never tell whether Silverfast was doing anything with the HDR, so I cannot speak to its use.) For Vuescan, it is better to not have the software do anything to the file. The fewer times you do adjustments, the better. Using photoshop on the raw file skips one level of processing, and using levels and its gama control will let you combine the processing you are doing with scanner software and the processing with Photoshop.

[Bruce Watson]

There is really no reason to treat the raw output any differently for the 9950 than from the drum scanner.

There might be. It depends on the scanner and the software of course. But... I suspect that you'll get better results if you at least set your black and white points to reflect the density range of the film in question, and your contrast as well, in the scanning software.

The reason for this is that you want to give the software the best use of available resources. If you only have 12 or 14 bits of real data capability, you want to use it all - you don't want an output file that is half zeros.

If your scanner/software generates a raw file that "has room for" the full dynamic range the scanner can generate, say the density range of 0-3.4, and it scans a B&W negative with a density range of say 0-1.4, then your file is more than half zeros. The file's density range from 1.5-3.4 will be zeros. Looked at another way, you could say that you are effectively scanning at 5-6 bits or so (no, I'm not going to sit here and do the math and give you and exact answer!). This is sort of like craming all your film's information into the toe of the response curve. It's not a good thing.

I think Kirk is right on this one. Tighten down on the film's density range as much as you reasonably can at scan time, erring on the side of caution to avoid leaving any actual image information behind. Then clean up the scan in an image editor.

FWIW, this is what people get when I drum scan film for them. I wouldn't necessarily consider it "raw" output, but rather output that is ready for image manipulation.

[Henry Ambrose]

Ed,
I'm with you on getting the full info from the scanner into Photoshop and doing so at as high a bit depth as supported by your machines and software. You are absolutely working in the right direction but Kirk and Bruce do have solid ideas that you can add into your mix. Setting black and white points just outside of your film's scene information range is a good idea for sure! Beyond that I suggest that you continue to reserve most of your adjustments for Photoshop. The less you rely on the scanning software and the more you use PS the better off you are in terms of controlling your files and not having to learn and re-learn each time you buy a different scanner.

[Kirk Gittings]

You guys know that I am not really technically minded. If things work i don't really care why.

My point is based soley on extensive anecdotal experience from Epson 3200, 4870, 4990, Canon 9950f, Imacon 848 and dozens of drum scans for comparison and Silverfast and Vuescan and the native software of each scanner. I am currently doing about 1250 4x5 and 6x9 scans s a year about 2/3rds of which are NPS color negatives and the rest FP4. All the b&w pigment prints for my current show were done on either a 4990 or an Imacon. These are the kinds of tonalities I am talking about. The b&w files for this portfolio were taken directly from the files used for the show prints.

www.portfolios.com/profile.html?MyUrl=KirkGittings (http://www.portfolios.com/profile.html?MyUrl=KirkGittings)

If you apply a steep curve to a flat unclipped, unsharpened, downsampled "raw" scan from those mentioned scanners (say in a sky section that you want allot of drama in-less so in the Imacon in b&w mode) you WILL get allot of granularity (similar to noise) in the transition areas between tones that are being stretched or compacted. This does not happen in good drum scans. If you scan with those scanners closer to the tonalities you want in the final output the sampling is done closer to those final tonalities, less expansion and contraction is done to achieve the final results and less granularity appears. I have struggled with this issue allot and decided that Ed's workflow will not work for my imagery.

[Ed Richards]

Bruce

Let's black box this. We have one box that is the sensor. The output from the sensor is 12/14 bits - as you put it, a lot of zeros. We can process it through a second black box inside the scanner that scales it by spreading the data over the entire range from 0 to 32,000, instead of, say, 0 to 6,000. All it is going to do is multiply the data with some function to scale it. We then take that data to our second black box (Photoshop) and multiply it by another function to scale it a little more accurately, then perhaps twiddle the contrast by applying yet another function. Each time we run it through a function we lose data with rounding errors. From a computer perspective, there is nothing that the computer in the scanner box can do that cannot be accomplished by the editing software - we are combining the two functions into one function and thus removing one set of rounding errors. (We just combine our Turing machines, for the fossil geeks out there.) I have also run this past Ed Hamrick, who writes Vuescan and knows a bit about the innards of scanners.

One more thing that I find fascinating with working with the raw output from Vuescan - you realize at once how much info is in the negative. Most negatives look pretty dark, meaning that the data is compressed into fewer zones. Some are full range, showing the data is spread across more zones. Those are going to produce better scans. What I have not yet done is work out whether the ones that look good are much higher contrast than you would want for silver printing. I had originally assumed that the best scanning negative would be relatively flat, but after scanning a step wedge I realized that the scanner could sort out much greater densities than I had assumed.

[Henry Ambrose]

Kirk,
You may not have the time or inclination for this but I'd love to see a straight "raw" scan and one using your current method of working on the same negative.

I think there are lots of ways to get to the same place. I'd most likely start using layers and mode: multiply and layer curves to get the hard worked sky effects if they were not there in the straight scan. Obviously you do it quite differently.

This is not a contest............ I'm just curious.

Also I'm headed to Chicago this weekend for the marathon. Do you have any pictures up there?

[Jeffrey Sipress]

Very interesting. I'm scanning a lot of B&W lately on my 4870, both from negs and from NPS color negs that I will convert in PS. I'm sure we have all noticed that some images just seem to scan better than others, with others often engaging in a fighting match with you. Say you've got a wonderfully captured B&W neg that contains a good eight to ten stop range of tonalities. How much of that will be noticeably transferred to your digital file at scan time? Sadly, not much. Nobody ever relates the scale of the film to the scale of the device. Hell, no one really knows the scale of their scanners because manufacturers don't divulge that information, and feed us some other unit of measure called dmax. What exactly is that? How does it relate to stops? When a nice soft scene (on a color tranny, color neg, or B&W neg) that's easy to expose and typically has a three to five stop range comes along, it will scan beautifully, and be truer to the original. That's because the scale of the device is more closely matched to what the film is giving it. I scanned the Kodak Q-14 greyscale test strip, which goes from 0 to 100% black in 20 increments, and measured the values of each area in PS and output them to excel to create a graph of the curve for the resulting scan. It was dismal! This is all why I think Kirk is on to something using color neg film for B&W work. The scale of that film better matches the scale of the scanning device. Sometimes scanning a color transparency for conversion to B&W is better than anything else. It simply matches the limitations of the system. No fighting!

[Kirk Gittings]

Henry,
Sorry I don't have any up in Chicago now. I will next summer at the Schneider Gallery, but i don't know the dates.

I wish i had time to do the side by side tests that you require. I don't right now. My workflow for FP4 negs. is this with the 4990 (or Imacon when I am in Chicago). Adjust overall tonailties so that they are in the ballpark for final print with plenty of shadow and highlight detail. Scan at 720 DPI RGB (Imacon gives a better scan in Grayscale) at largest size I will ever print (20x24). Discard all but the blue channel (least noise and sharpest). Spot the file at 100% and do some capture sharpening. Down size to 360 DPI. Tweak levels and curves in adjustment layers. Do burning and dodging via Snapshots and history brush. Do test print. Adjust levels and curves again via layers. test print. Final sharpening via Optipix Safe Sharpen 1X in a duplicate layer. Final output downsizing (usually to 11x14 or 16x20) done by Imageprint RIP. Final Print.

[Henry Ambrose]

Jeffery,
What we're talking about here is not new, its been beat about for years now on other forums and by other individuals. Lots of folks shooting 35mm use color neg and cheap commercial processing to do their "B&W" work. The density range of the color negative film will fit the range of most any desktop scanner so it works great for digital B&W printing. And I'm referring to inexact exposures and wide ranges of scenes on one roll. But you can fit your "real" B&W film to your scanner so you don't have to retool your process completely. After all, you do have the luxury of adjusting exposure and development on each individual sheet. But still, what makes the best picture makes the best picture.

Where we really get lost is in not knowing what the scanner is doing even when we think that we've "turned off the software". As far as I can tell they behave differently across brands and I think this accounts for some of the differences we are discussing here.

Ed, next time you scan a negative and bring it into PS at 16 bit, look at the histogram to see if you got it all wadded up instead covering nearly the entire available range. This does make a difference when you start adjusting the file -- or it might be just the interpretation you wanted. The problem is that the second scenario occurred by accident.
; >)

Kirk,
Your workflow is not all that different from mine. I do think its funny that you use the blue channel, Ed uses the green channel and I use all three!

[Paddy Quinn]

Jeffrey

It sounds more like problems with limitiations in either the scanner you are using, the software or operator skills.

I have few problems capturing the full range of a B&W negative via scanning (and scanning my Stouffer 21 step transmissive wedge gives a full range of steps with nice seperation).

Scans usually show a greater range of detail from shadow to highlight than I ever got in a darkroom print (and of course translating the scan and file into a "digital" print presents a similar challenge). Scanning to give your 8-10 stop range isn't impossible at all .

"Hell, no one really knows the scale of their scanners because manufacturers don't divulge that information, and feed us some other unit of measure called dmax. "

It is D(ensity)Max - and goes with D(ensity)Min to give you the Density Range of the scanner. however accurate or not the manufacturers measuring of it may be for a particualr scanner so the general principle is the density range of a negative (hypothetical numbers) might be say 2.9 while that of a transparency is 3.6 - the difference due to the offset caused by the different types of film base (clear v's cloudy).

Most decent flatbed transparency scanners can do a pretty good job of capturing the full range of a negative (colour or B&W) - they can be challenged by the higher density range of transparency film

The problems comes with how you capture that information - i.e. how good the software is and how good the operator is at using it and then how well that data is dealt with in Photoshop or whatever the imaging programme used. Much scanner software does a rather bad job at dealing with both types of negative film, clipping lots of the information and having problems inverting the image correctly. Most also go about dealing with the orange mask incorrectly. In addition, Photoshop's "Invert" command also doesn't go about inverting the image the way you would think it does and this itself introduces all sorts of problems

It sounds like you have probably put a lot of time and effort - perhaps many years - into developing your zone system skills. Digital skills have an equally steep learning curve and take at least as long to develop to a level of expertise.

[Kirk Gittings]

"Kirk, Your workflow is not all that different from mine. I do think its funny that you use the blue channel, Ed uses the green channel and I use all three!"

Henry, I think it really varies by the scanner and that is why everyone has to test this for themselves.

On my 4990 the Blue channel is clearly sharper and less noisy than the others. If I remember correctly my Canon 950f (I returned three of them for streaking and finally gave up) had less noise in the green but was sharper in the blue. The Imacons I have tested are very very sharp in all modes but have allot of noise in RGB. So I can get the best b&w scan from the Imacons in the grayscale mode. I have only tested Vuescan on Epson scanners. The only way it surpased Silverfast was in the singlepass multisampling which gives a sharper result than the SF multipass sampling. The SF eliminated noise better but the many passes resulted in a less sharp image than the Vuescan.

No one should be loyal to their equipment because they spent allot of money for it. The Imacon I use cost $16,000.00 but is only better than my $500.00 4990 in terms of sharpness and that only minutely shows up in a 16x20 or larger!TEST!

[Jeffrey Sipress]

Thanks, Henry and Paddy. Perhaps my wedge test was useless, since it was a reflective scan,and doesn't prove what a transmissive test would yield. I need to get hold of that Stouffer strip. I've never heard or it.

My images are coming out better than the test I did should allow! I hope you're right about my scanner capturing all of what's on a B&W neg. I'd just like confirmation.

[Ed Richards]

> Ed, next time you scan a negative and bring it into PS at 16 bit, look at the histogram to see if you got it all wadded up instead covering nearly the entire available range.

Of course it is all "bunched up" - the point I am making is that "unbunching it" is a computer calculation that is about the same as applying the levels command with gamma adjust in photoshop or PWP. The advantage of doing it in the editor is doing it once saves rounding errors over doing it twice. There seems to be a notion that what is going on in the scanner computer is different from what is happening in the computer on your desk. Once you have the CCD data into digital form out of the sensor, it does not matter where you scale it, except that you get more control by doing it all in the editor. This is esp. true if you are doing the scan once approach - what I save as the master file has no scaling applied at all, so I am not locked into white and black point and gamma choices made when I originally scanned the negative. This is useful because black and white points and gamma interact with curves - it is very useful to have both on layers you can play them off against each other.

As for the invert command - at least on B&W I do not see it doing anything unusual. The histogram seems to a mirror of the histogram before inverting, i.e., it is flipped about the midpoint, with data being bunched at high end rather than the low end. What are you seeing that is unusual? I may be missing it. I usually invert in PWP anyway - it is much faster than Photoshop for resizing and inverting

[Frank Petronio]

Just a vote for VueScan with basic tonal editing done in the preview and very little else. It works with all of my scanners and is pretty nice software.

[Paul Butzi]

The advantage of doing it in the editor is doing it once saves rounding errors over doing it twice. There seems to be a notion that what is going on in the scanner computer is different from what is happening in the computer on your desk. Once you have the CCD data into digital form out of the sensor, it does not matter where you scale it, except that you get more control by doing it all in the editor. This is esp. true if you are doing the scan once approach - what I save as the master file has no scaling applied at all, so I am not locked into white and black point and gamma choices made when I originally scanned the negative. This is useful because black and white points and gamma interact with curves - it is very useful to have both on layers you can play them off against each other

I'd agree that, in the abstract, it's better to perform all the scaling and curve adjustments exactly once, if all other things are equal. I even agree that it's quantization/rounding errors that increase if you repeatedly perform the calculations.

That said, I would disagree with the idea that this is significant.

First, the calculations done in the scanner/scanner driver may well have more data to work with than the calculations done in photoshop. That is, the scanner may well have, say, 18 bits per channel to work with - those 18 bits are then run through the scanner scaling and curve calculation, with the results being expressed in properly rounded 16 bit values. You're also assuming that the scanner cannot adjust exposure and amplifier gain based on the setting of white point and black point. While that might be true for some very cheap scanners, I know it's not true for at least two of the scanners I've owned - an Epson 1640SU and a Microtek 1800f, because both of those scanners scan more slowly when scanning negatives when the white and black points are set up higher (to scan negatives with high base density).

Only if ALL the data that was available on the scanner side was present in the 'raw' scanner format would Photoshop have all the data. I don't know, but I expect that the 'raw' scanner output is not this wider data format, but instead just the same scanner software calculation run with the black point and white point set to the extremes, and with the tonal distribution being set to some default curve. So, in the end, the calculations get done twice, no matter what you do.

Also, we're talking about truncation/rounding introducing noise in the image. That means that, for any given pixel, the maximum error introduced by such rounding will be a single bit error in the least significant bit - a relatively small error.

And, if it's true that the scanner can adjust exposure and gain based on the settings of white and black point, AND the scanner works internally with a wider data stream than is passed to Photoshop, then you'd definitely be better off (in the theoretic sense) setting the white and black points carefully in the scanner software (thus increasing the number of bits available to express the data, and getting more resolution passed to photoshop) and taking the minor single bit hit when you do further adjustment in photoshop.

All that said, I have a hard time believing that it's possible to discern the differences if you have a workflow which is not narrower than 16 bits between the scanner and Photoshop.

[Kirk Gittings]

"The bottom crop is a shadow performance test suggested by Paul Schilliger. It examines how much detail the scanner is able to pull out of deep shadow, and how much noise is in there. To see what was happening in the shadows, I used Photoshop to apply the steep curve that you see on the right." From this sites scanner comparison.

Check out the bottom image in the scanner comparison and compare the Creo, Imacon Prec. II, and the Epson 4990. These all have a steep curve applied in PS as I noted above that I apply to my cloudy skies. Notice the increased noise in all but the drum scan.

My experience suggests that if this curve was applied in Silverfast (or whatever software) originally in the scan, when the sampling is done, that this noise would is greatly reduced, because this increased noise is a result of expansion or clumping (or something) of the pixels in the transition areas between tones. Therefore...... similar to what Paul stated above, it is better to do this tonal expansion as much as possible in the original scan to minimize this granularity.

I have struggled with this problem for over a year on everything but drum scans and the solution that seems to work is to do as much of the tonal manipulation in the original scan as possible.

If my thinking is wrong here I would like to hear why? How is my experience inconsistent with the scanner comparison's examples?

[Ed Richards]

Back to my point on workflow. If you do adjustments in the scanner, you have to rescan to adjust the values. If you do it in Photoshop/PWP, you can save a master file that can be readjusted in the future without rescanning. Whether rounding errors matter or not, it is very useful to be able to readjust the B/W points and gamma as you work with the curves and local contrast, which you cannot do if it done in the scanner.

Kirk - I would not use a curve at all but use levels and gamma to get into the ball park. When I tried converting raw scans with curves, I got exactly your problem. As for noise, that drum scanner is going to beat flatbeds no matter what we do to optimize them.

Paul - I would be really surprised to find that the scanners have a wider internal data path. If so, why would they internally limit it to 12-14 bits as almost all do, and not at least broaden it to 16? I doubt that the sensors have that much dynamic range. Some scanners do change exposures when reading denser negatives. The two you have might be doing that. They might also just be scanning slower because the software is doing more calculations and they are bogging down on data transfer - I do not think they have much in the way of a buffer, and I know they dynamically adjust scanning speed to match data output rates. As for the insignificance of the rounding - in the worst case, which is pretty common, the original data in areas such as the sky is spread over a very narrow range and is being scaled out over a much wider range. If the entire range of original data is spread over just a few bytes worth of range, then rounding errors become much more significant as they compound.

[tim atherton]

" I would not use a curve at all but use levels and gamma to get into the ball park. When I tried converting raw scans with curves, I got exactly your problem. As for noise, that drum scanner is going to beat flatbeds no matter what we do to optimize them. "

I'd really urge anyone who is making raw scans and then converting them in PS to try out the NegPos plugin I detailed above (+ read the various documents he has up - especially if you are into the math of the stuff... he has a LOT of detail on that).

I have found it generally does by far the best job of either doing by "hand" (levels, gamma etc) or Silverfast Ai, or any of the other scanner software I have tried such as Vuescan

[Bruce Watson]

Ed,

It occurs to me that this discussion isn't going to help you much. What you need to do to resolve the issue is to run some tests on your own equipment. As the wise man said, "Why guess when you can know?"

This won't be hard or take long. Toss a favorite negative in the scanner. Scan it twice. Once in your "raw" mode. Once pulling in your white and black points and setting the constast curve. Process both in your favorite image editor making the best image you can each time. Make a print from each (why? because this is about prints, not what you see on your monitor) at your maximum size. Compare the prints side-by-side under the same lights.

If you like one better than the other, that's the method to use. If it's a tie, do it the way that easiest for you. Let us know what you decide.

[Ed Richards]

> NegPos plugin

I have read their documents and they seem consistent with what I am doing. Their plug in is designed for color - are you using it for black an white?

> It occurs to me that this discussion isn't going to help you much. What you need to do to resolve the issue is to run some tests on your own equipment. As the wise man said, "Why guess when you can know?"

I have been doing that for the past month:-) What I found is that you get a much better adjustment using levels and then curves, rather than curves alone, even in PWP which has a much better implementation of curves than does Photoshop. I think this is what people are doing when use the scanner software to preprocess their data by setting black and white points and gamma. I am betting if they try the same thing in Photoshop with levels and gamma before curves on a real raw file, they will get the same results they are getting with scanner processing. The fact that it reduces rounding errors is interesting, but probably only philosophical.

The real advantage I am finding is the iterative use of levels and curves together, using prints for fine tuning. You cannot do this iterative tuning if you preprocess in the scanner. I also change my mind about how an image should look, so I like to be able to readjust it without rescanning. Now this may just be me - I am sure the experienced folks know exactly where they want to be in the final print and what B/W points and gamma they need to get there before they scan the negative.:-)

[Kirk Gittings]

Ed your final point may be the real key. I am an old Zone System guy who "previsualizes" my images. I pretty much know right where I want to take the print before I ever expose the film. An old analoge way of workig that I have brought with me into scanning and priting.

[Chuck_1686]

Trying to understand this I'm confused about some scanner basics. I'm assuming the scanner sensor feeds some type of A/D converter. This converter might be adjustable by software to use the maximum range based on the negative density. If this is true it would seem like you would want to do scanner adjustments for each negative. I suppose that if this stage of the scanner was not adjustable and scanner software worked on the digital data then doing the adjustment later would be better.

[steve simmons]

There is a long and extensive article on scanning black and white in the Sept/Oct 05 issue of View Camera if anyone wants some additional info

steve simmons

[Paul Butzi]

Ed your final point may be the real key. I am an old Zone System guy who "previsualizes" my images. I pretty much know right where I want to take the print before I ever expose the film. An old analoge way of workig that I have brought with me into scanning and priting.

Kirk, are you saying that you don't do any tonal adjustment post-scan at all?

Wow. If that's a workflow that works for you, I'm impressed.

I'm like Ed - the very first thing I do to an image is add a 'levels' layer, adjust the overall levels (usually to be just a teensy bit 'soft' relative to where I want) and then add a 'curves' layer, which I use to adjust to overall tonality.

[Kirk Gittings]

Paul,

"Kirk, are you saying that you don't do any tonal adjustment post-scan at all?" I wish that were true as things would go allot quicker.

But, I do allot of post-scan "fine" adjustment. The original vision of that image never changes for me, but that original vision may "previsualize" allot of computer work to realize that original vision. As for my workflow issue it is just that if you scan the tones in the ballpark of where you want them in the final output then you create allot less artifacts in the overall process as you manipulate tones and contrast.

[Struan Gray]

I'm no expert, but I love playing one on the internet.

Scanners work by taking a voltage from a PMT preamplifier or a CCD cell readout amplifier, scaling it with an analogue gain circuit and then digitising it with an analogue-to-digital converter chip (ADC). There may be some oddball ADCs that adapt their scaling on the fly, but all the standard ones in the catalogues have a fixed input gain and a fixed output bit depth.

The voltage is a linear measure of the amount of light that passes through the film. You can't have negative light, so that voltage will vary between zero and some maximum value. If you are scanning a transparency the maximum value will be set by the clear film of a blown-out highlight and there will be a miminum that is close to zero - often within the noise level - from the densest shadows. If you are scanning a negative the maximum will be from unexposed portions of the film (i.e. whatever base+fog lets through) and the minimum from the densest highlights.

The linearity of the light measurement means that there is no benefit in terms of image data quality from applying a gamma curve or a custom tonal curve at scan time. There may well be workflow benefits, but the data doesn't care if it gets scaled in the digital domain inside the scanner or inside the computer. One caveat: often cheap scanner firmware is optimised to be small and fast, so tonal manipulations in the computer can in principle offer you more control, less noise and fewer artifacts.

If you are scanning transparencies things are simple. Almost all films have similarly clear bases, so there is no need to fine tune the analogue gain before digitisation. Just choose a value that puts zero absorbtion by the film at the maximum of the digital dynamic range and go. Your bit depth and noise performance will then determine the darkest shadows in which you can reliably detect detail.

Of course, if you have an underexposed tranny it would be nice to be able to turn up the analogue gain. This way you match the dingey highlights to the maximum of the digital dynamic range, and all the lower zones get expanded up in proportion. There is a noise penalty associated with this, but that can be somewhat counteracted with multisampling, and it's usually more acceptable than the results of a purely digital scaling of the values. This is why setting the white point can be a good idea from the viewpoint of the quality of the scanned image data. Some scanners change the analogue gain when you set a white point, others (like the Nikon 35 mm Coolscans) make it a seperate control. Some simple experiments with a dense slide will tell you if your scanner is one of those that makes this worthwhile.

Note that setting a tranny scan black point is almost certainly of no use at all. In theory you could subtract the black point equivalent voltage in the analogue domain and then scale only the black-to-white range of light/voltage values onto the dynamic range of the ADC, and thus gain some more image quality. Laboratory instruments do this routinely, but I don't know of any photographic scanners that do, and certainly none of the consumer flatbeds. If your scanner does turn out to have an analogue offset combined with its gain circuit setting a black point will enhance scans by supressing noise in the shadows. The benefit will be greatest with scenes that have plenty of light but little contrast such as high key fashion and foggy seascapes.

With negatives the issues are inverted: setting a black point will help in a scanner that adjusts its analogue gain to match; setting a white point probably does little good; setting gamma or curves at scan time is of no benefit. Note that traditional darkroom concentration on base+fog is less of an issue here since digitisation is intrinsically linear - there is no paper toe to avoid. However, base+fog will exacerbate the problem of there being no analogue offset, so expect problems with low contrast originals.

Also, negatives are inherently less dense than transparencies. This makes the issue of analogue gain more acute because if you simply rely on scaling in the digital domain, you've thrown away most of the dynamic range of your ADC and you become heavily reliant on good noise performance. The very cheapest scanners keep a fixed gain for both negs and transparencies, which is usually a disaster for the former, but doesn't bother reviewers who are fixated on Dmax and scanning slides. Most consumer and professional scanners will increase the analogue gain (or equivalently, the exposure time on the CCD) to compensate, which is why scanning negs as transparencies and inverting in photoshop is usually a bad idea: the switch to negative mode in the scanner software/firmware is what ups the gain.

Finally, colour negatives are orange. Good scanners - even in the consumer realm - will use different exposure times or gains for the different channels, and this can affect their tonal quality, especially with respect to noise.

So, all that said, what are the consequences for scanning B+W negative? The most important thing is to find a way to force your scanner to adjust its analogue gain to match the negative stock and your particular image. That almost always means scanning in a negative mode, not positive, and it sometimes helps to set a black point. Some scanners don't adjust gain for B+W negs, but do for colour ones, which is why scanning B+W as colour negative can be a good thing. In colour neg mode many scanners will have much better noise performance in the red channel. Naturally, they are usually the ones that are sharpest in the blue channel: your choice :-)

Finally, low base+fog is a good thing. N+ processing for low-contrast scenes is also a good thing. N- processing is redundant.

I use TMAX100, expose for the shadows, and let the highlights end up where they will. I will push a neg of a low contrast scene if tonal seperation is important to me - i.e. if I don't want a low-contrast final print. On my Epson 3200 I get best results scanning in colour negative mode and taking the red channel, but that's because I'm more interested in avoiding noise than maximising sharpness. 16 bit is always worth it, even though it's really only 12 bit on the Epson and I often reduce the bit depth very soon after scanning. YMMV.

[Bruce Watson]

Basically, I meant what Straun actually said. Nice explanation Straun.

[Struan Gray]

Ah. But you were brief :-)

[Chuck_1686]

Struan, thanks for the education. I'll probably only have to read it a couple more times but I can see how things are working.

[Kirk Gittings]

Ed read Michael Mutmansky's contribition to the new thread "Ideal film/developer-drum scanning B&W Landscapes", which explains my approach better than I can.

[Ed Richards]

> The linearity of the light measurement means that there is no benefit in terms of image data quality from applying a gamma curve or a custom tonal curve at scan time. There may well be workflow benefits, but the data doesn't care if it gets scaled in the digital domain inside the scanner or inside the computer. One caveat: often cheap scanner firmware is optimised to be small and fast, so tonal manipulations in the computer can in principle offer you more control, less noise and fewer artifacts.

I think that is where I started.:-)

The second point is that b/w points only matter if your scanner does change the gain. I wonder how many consumer flatbeds really change the gain in the hardware, as opposed to the scanning software? For example, the Canon 9950, which is a pretty good consumer level scanner, does not change the hardware gain. It would be interesting to know how many really do - Ed Hamrick would probably know, since he lets you change the gain/exposure in Vuescan if the scanner allows it.

Thus, unless your scanner changes the gain in hardware, we converge, 37 messages later, on the my original hypothesis that it is better to manipulate the data in the computer than to mess with scanner data manipulation, based on the original assumption that the scanner cannot change the exposure.

[Kirk Gittings]

Your original hypothosis is what I was taught and what is what most people are taught at the two universities that I teach at. It is not new thinking in my experience but old thinking that I have found through my testing creates problems down the road in the workflow at least with imagery like mine which requires tonal expansions.

By the time you got to the thread it had moved well past the point made by Michael that I was refering to. This is not a pissing match. My experience directly contradicts yours and I have tested the hypothosis thoroughly and repeatedly for over a year now with my own work and with my students. Michael simply explained my point in more scientific terms than I can. The section I was pointing you to starts with this statement:

"Kirk,

Yes, it does. That is part of the reason that I believe it is important to get is reasonably close to right in the scan, and not leave it to PS manipulation later on. Every manipulation to the file after it's creation destroys data in some manner, with the exception of some basic rotaional transpositions and other minor things like that.

Curves, Levels, color adjustments, etc. all ultimately are destroying data that was in the original file, which is why it is so important that the manipulations are made as Levels adjustments, and are not applied directly to the background.

In the case of Curves, as you have suggested, the adjustment of the midtones causes a seperation in the steps somewhere in the curve, and a compression of the steps in another part of the curve. When the seperation is drastic enough, it will cause an increase in the apparent 'grain' that is visible. In this case, I mean noise more than grain, but it does ultimately come off looking like digital grain.

This is partially exacerbated by the fact that most people consider a digital file to be a continuum, rather than a stepped data source. Further, PS is apparently not truely 16 bit, it actualy operates at 15 bits. Even more, most B&W files don't have anywhere near the theoretical number of possible levels of gray in them, most have much fewer than 1/2 and high bit files (16 bit) may often have 1/3 or 1/4.

Additionally, the problems most easily crop up in areas where there is a discontunity in the tonal structure of the file, because this is where a large step is already being made in the data.

This can be made worse if the source file is a general scan to cover the entire density range (the raw scan output approach), because then the useful data does not cover the entire range of the image file, and if there were possibility for some interpolated data within the useful data set, it was thrown out in favor of capturing useless data beyond the image density range.

Suddenly, 16 bit doesn't sound like too much, does it? Image how careful you have to be when scaning at 8 bit? My drum scanner has a 12 bit output, and writes a 16 bit file so it is useable in PS, but it only has 12 bits of true data in it. Because of the fine control over the input data, I am able to easily produce good scans that require only minor adjustments to the fie, and these types of scanning artifact problems do not show their head too much. I still have to be careful in open sky areas, though."

The point is simple and confirmed by Michael in his text. I have spent about two years now testing many scanners and software and service bureaus preparing for my current retrospective show, and I came to the conclusion that a workflow as you descirbe was not as effective for imagery like mine as the one I currently use which coincidentally is similar to the one developed by Michael M. and Ed Harris. I am simply sharing this so others do not waste as much time as I have on less "targeted scans".

[Struan Gray]

Kirk, I don't see any contradiction between what Michael has said and what I and Ed are saying. The essential point is the 'if': whether your scanner adjusts gain at the analogue stage or not, and if so, how?.

Most LF-ers are old enough to have recorded to tape. Ed and I are merely saying that some cheap scanners are like tape recorders that don't have a levels setting knob, and that if the knob is missing it is better to control the volume with a quality circuit in your amp than to use the one-cent volume potentiometer in the tape deck.

The biggest problem is that whereas it is usually clear that a tape deck is missing a levels control, scanner software is user-opaque: gain control comes and goes without the user explicitly being told.

Ed: I know that Ed Hamrick has posted a lot of info about which scanners do and do not adjust their gain (or integration time) in the newsgroup comp.periphs.scanners. Or you can email him a specific question. I don't use Viewscan, but I have learnt a lot from studying Ed's explanations of how it works. He's always friendly and informative in person.

FWIW, a lot of the effects that Kirk describes are the classic signatures of roundoff error. If you have 16 bits to play with (or 15) they will disappear, with the caveat that the code is properly written. Unfortunately, even Photoshop seems to get it wrong from time to time, and it's reluctant and grudging support for 16 bits per channel suggest to me that there may be problems under the hood. Note that if you pack a 12-bit number into the top bits of of a 16-bit file, that actually gives you more overhead when compared to the original signal to noise because you can fill those lower zeros with what would otherwise have been rounding errors.

That said, I have run my scans as double precision reals and as 64-bit integers through the package I use for my microscopy (IDL (http://www.rsinc.com/idl/) from Research Systems) and I can confidently say that I am limited by random and pattern noise in my scanner (an Epson 3200). In this case, squeezing every last bit of information out of the scanner helps, even at the expense of some sharpness.