Target densities for scanning

[Per Bjesse]

Hi all,

I recently had an interesting discussion about film development for scanning that provoked a lot of thoughts for me, and I would love to hear more input on the matter. Here goes: Assume you are developing bw film for optimal scanning. Standard development settings aim to get a normal SBR scene to around 2.0 maximum density. This is slightly less than 7 stops, so not even 8 bits of data. Now, modern scanners can easily deal with a DMAX of 3.0 or more, so at least 10 stops. So: What is a good target density for highlights for optimum 16 bit grayscale scanning?

One could argue to that the best approach is to use the full density range the scanner can read. However: This will mean increased grain, and there is also the possibility of more noise from the scanner in the densest regions. So where should one aim?

Second thought: If you develop for scanning, what is the value of compensating developers like extreme minimal agitation pyro (like Steve Sherman) or two bath developers? All that does is to distort the response curve to compress the highlights. Given the the scanner has plenty of headroom in terms of maxium headroom, it seems like this distortion will not be needed, and if it is desired it can still be imposed in post.

Third thought: If you accept that compensating effects from developers become relatively unimportant when developing for scanning, it seems like the only developer effects you care about really is acutance and fine grain vs coarse grain. And if you scan large format film, for reasonable size viewing, then really all that matter in terms of developer effects is acutance. Is there something I am missing in terms of developer effects?

I would love to get some input on this.

Regards,
-Per

 

[Ted Baker]

Assume you are developing bw film for optimal scanning. Standard development settings aim to get a normal SBR scene to around 2.0 maximum density. This is slightly less than 7 stops, so not even 8 bits of data. Now, modern scanners can easily deal with a DMAX of 3.0 or more, so at least 10 stops. So: What is a good target density for highlights for optimum 16 bit grayscale scanning?

This first statement is not really correct, so you may wish to revisit after considering that many epsons flatbeds, coolscans, and certainly a DSLR allows you vary the exposure time to the sensor. A sensor often has linear relationship for light against voltage. So a Dmax of 2 can bit spread over all 16bit by extending the exposure time. Actually Its not actually as simple as this, but this is example of why you starting point is wrong.

 

[Mainecoonmaniac]

If you want to have target densities, that mean you have to have a densitometer to read those densities. I woudn't worry to much in trying to recreate the wheel. If your interested in film processing, Flicker has a bunch of great folks that will tell you what ASA the film is shot at and how it's developed. You could also checkout FilmDev.org. Also, the Massive Dev chart is good too. Don't get too wrapped up into reading densities. Shoot and have fun!

 

[Per Bjesse]

This first statement is not really correct, so you may wish to revisit after considering that many epsons flatbeds, coolscans, and certainly a DSLR allows you vary the exposure time to the sensor. A sensor often has linear relationship for light against voltage. So a Dmax of 2 can bit spread over all 16bit by extending the exposure time. Actually Its not actually as simple as this, but this is example of why you starting point is wrong.

Ok. So my question then remains: If you have complete freedom to chose the density range that you render your scene for scanning, what should one aim for in order to get optimal information in the scan? For wet processing, it is about 2.0 maximum density due to the exposure scale of the paper. But for scanning, it seems better to encode things in as wide of a density range as possible (within reason) so as to make it easy for the scanner to get as much information as possible out. Sure, the scanner can vary exposure, but if you compress the information into a very narrow range then it seems clear that will be suboptimal. So why not use the full range?

 

[Per Bjesse]

If you want to have target densities, that mean you have to have a densitometer to read those densities. I woudn't worry to much in trying to recreate the wheel. If your interested in film processing, Flicker has a bunch of great folks that will tell you what ASA the film is shot at and how it's developed. You could also checkout FilmDev.org. Also, the Massive Dev chart is good too. Don't get too wrapped up into reading densities. Shoot and have fun!

My intention here is not to recreate the wheel, but to figure out how to control my exposure and development for optimal scans.

It is easy to do BTZS style film testing to make sure that your developed negatives covers the whole printable range for papers regardless of the subject brightness range of what you are photographing (and yes, I have gone through this exercise in order to control my wet printing process). My question is: For a hybrid flow, why set your development to match the exposure scale of paper, and what advantage could there be to use the full range the scanner can use? Again, too dense negs will be grainy and possibly have scanner noice in the highlights so just going all the way to the DMAX of the scanner does not seem reasonable.

 

[Mainecoonmaniac]

My intention here is not to recreate the wheel, but to figure out how to control my exposure and development for optimal scans.

It is easy to do BTZS style film testing to make sure that your developed negatives covers the whole printable range for papers regardless of the subject brightness range of what you are photographing (and yes, I have gone through this exercise in order to control my wet printing process). My question is: For a hybrid flow, why set your development to match the exposure scale of paper, and what advantage could there be to use the full range the scanner can use? Again, too dense negs will be grainy and possibly have scanner noice in the highlights so just going all the way to the DMAX of the scanner does not seem reasonable.

Ah. I get it. I used to do the same thing for printing in silver gelatin printing but you want to optimally process your film for scanning. For film, there's a rule of expose for shadows and develop for highlights. I would find out what ASA your film is with your chosen developer. Then make a series of Zone VIII exposures (lightest zone with detail) then creep up the processing time to find your optimal density. Zone VIII is chosen because Zone IX and X will have no detail (No noise). Maybe fellow posters can verify my methodology?

 

[Per Bjesse]

Ah. I get it. I used to do the same thing for printing in silver gelatin printing but you want to optimally process your film for scanning. For film, there's a rule of expose for shadows and develop for highlights. I would find out what ASA your film is with your chosen developer. Then make a series of Zone VIII exposures (lightest zone with detail) then creep up the processing time to find your optimal density. Zone VIII is chosen because Zone IX and X will have no detail (No noise). Maybe fellow posters can verify my methodology?

I know how to do the testing once a target maximum density has been chosen; I have gone through this process both for paper and film trsting. The question I am asking here has to do with what the best maximum density for a pure scanning flow would be.

 

[bernard_L]

Again, too dense negs will be grainy and possibly have scanner noice in the highlights so just going all the way to the DMAX of the scanner does not seem reasonable.

Precisely! You answered your own question. If the highlights reach D=2.0, and the least significant bit in the sensor output corresponds to D=3.0, then the highlights are digitized with only three significant bits. You should start to definitely see posterization. Even worse if you would do as you propose:

"For wet processing, it is about 2.0 maximum density due to the exposure scale of the paper. But for scanning, it seems better to encode things in as wide of a density range as possible (within reason) so as to make it easy for the scanner to get as much information as possible out."​

Actually, it's a little more complicated. The presence of grain introduces what is called in another context dithering, effectively spreading the decision levels of the quantizer (analog-digital converter), and mitigating teh appearance of posterization.
Another way to cope with a substantial density range is dual exposure, as is possible with vuescan with some scanners at least.
And, your numbers might not be totally representative. Typical scene would span zones 0-8, a range of 2.4 in logE. Developing for what seems to be the standard C.I. of 0.55, the density range will be 2.4x0.55=1.32, not quite the 2.0 that you assume, and would like to push up to 3.0. And, that is fortunate, allowing the scanned output of even the highlights to contain a decent number of significant bits.

 

[Alan Johnson]

IMO, it's possibly the grain,I don't show it here but the edge effects with different films and developers are retained on scanning, eg Rodinal needs less sharpening than Xtol..
Scanning at 3600 dpi the differences between developers using MDC times are quite apparent .You could do a test like this to definitively answer the question re grain.
https://www.photrio.com/forum/threads/scans-retain-developer-properties.156243/

 

[Per Bjesse]

IMO, it's possibly the grain,I don't show it here but the edge effects with different films and developers are retained on scanning, eg Rodinal needs less sharpening than Xtol..
Scanning at 3600 dpi the differences between developers using MDC times are quite apparent .You could do a test like this to definitively answer the question re grain.
https://www.photrio.com/forum/threads/scans-retain-developer-properties.156243/

Thanks for this interesting read, Alan. Different developers clearly give a noticeably different look.

 

[ced]

You could do worse than applying the old Kodak AMB method using a reflection grey wedge which were designed for reproduction photography as used in the graphic arts test/tune to your intended satisfaction:
https://www.kodak.com/motion/Produc...Separation_Guides_and_Gray_Scales/default.htm

 

[Per Bjesse]

You could do worse than applying the old Kodak AMB method using a reflection grey wedge which were designed for reproduction photography as used in the graphic arts test/tune to your intended satisfaction:
https://www.kodak.com/motion/Produc...Separation_Guides_and_Gray_Scales/default.htm

Interesting. Could you elaborate? I have a step wedge and a densiometer, but the question at hand is what the density requirements/contrast index of the perfect negative for a scan would be. How would the AMB method leverage these to answer this question?

 

[MattKing]

I expect that it would vary with every scanner model and every version of the scanner's firmware. And the result would be further complicated by the relationship between the scanner's firmware and your computer's software.
You certainly could experiment with your scanner to get an answer - just be sure to use software that truly does offer an option for a "raw" scan.

 

[ced]

To make a test take a photograph, a painting or any still life and include the wedge in the scene.
Take a light reading for your correct exposure on the film you are using. (make sure that the wedge is not giving off glossy reflections, alter the angle of the wedge if needed.
Process your film and put the result into your scanner or enlarger. If scanning try the values used here:
http://ulib.hamilton.edu/using-reference-strips
If using the enlarger you aim to get as close to the original densities on the wedge.
Hope it helps.

 

[Lachlan Young]

In short, aim for densities that deliver a wet print that you like. If your scanner can't handle the range, get a better scanner. Consumer flatbeds are largely not up to the job.