Question on a different approach on digital negatives

Hi all, I am very interested to know if anyone has a different approach about generating digital negatives for Pt/Pd. Since I think that such drastic curve adjustment on the positive generates a huge amount of noise and thus a degradation of image quality.

Why would you think this, once you calibrate your input values to your output values you should see no noise or degradation

Raúl Aragón said:

Hi all, I am very interested to know if anyone has a different approach about generating digital negatives for Pt/Pd. Since I think that such drastic curve adjustment on the positive generates a huge amount of noise and thus a degradation of image quality.

Click to expand...

Maybe it would help if you described, in detail, your process for making digital negatives.

It might also be useful if you attach one of the files you have used to print a problematic negative and the resulting print.

The usual approach is used many folks, including myself, to make perfectly acceptable negatives. Thus, I suspect that the problem may be specific to your working method rather than a general issue.

Carnie Bob said:

Why would you think this, once you calibrate your input values to your output values you should see no noise or degradation

Click to expand...

hi Carnie, the extreme curve creates video noise (electronic grain) because when you apply such an aggressive curve to any digital file (even sound file) what you are basically doing is amplifying it and that inevitably augments the noise, that's a fact.

fgorga said:

Maybe it would help if you described, in detail, your process for making digital negatives.

It might also be useful if you attach one of the files you have used to print a problematic negative and the resulting print.

The usual approach is used many folks, including myself, to make perfectly acceptable negatives. Thus, I suspect that the problem may be specific to your working method rather than a general issue.

Click to expand...

I'm not having problems with my prints, it's just that I'm searching for innovation in the process of creating DN and avoiding aggressive curves that inevitably create video noise due to the amplification of the signal.

Yes, there is a method that uses the linearization curves to control ink loads within QTR instead of the image file itself.

It was primarily for producing silver gelatin prints with digital negatives. I'm sure the same principles apply for pt/pd.

EDIT: Actually, now that I recall it better I think this was only possible with RC coated inkjet papers (higher dmax hence silver only). Would have to test for alt printing with OHP film and potential ink pooling.

Yes, I use a different approach for the DN creation that's developed exactly for addressing this sort of problems. I called it "Soft-Proofing Calibration" and a few year ago I presented the first steps of development here and here. I have since adjusted and further developed it and I am using it regularly.

Raúl Aragón said:

hi Carnie, the extreme curve creates video noise (electronic grain) because when you apply such an aggressive curve to any digital file (even sound file) what you are basically doing is amplifying it and that inevitably augments the noise, that's a fact.

Click to expand...

I'm afraid you misunderstand. A dramatic correction curve does not necessarily amplify pre-existing noise in the source in the final print. It may/will increase contrast of noisy parts of an image in certain densities, depending on the correction curves, in the intermediate negative. But if the calibration (linearization) was done correctly, this will not translate into increased noisiness in the final print produced from that negative.

One potential problem with dramatic correction curves is the introduction of posterization artefacts. In practice, this is virtually entirely avoided by working in 16 bit space in the digital stage of the process. This is effective in combination with selecting printing parameters that deposit only as much density as necessary to print paper white in the target printing process. Posterization problems will/can emerge in parts of the tonal scale if you fail to do this and attempt to linearize negatives to an inkjet/digital printing process that creates excessively contrasty negatives. In this case, you would effectively end up using only a small part of the dynamic range of the negative and squash the entire tonal range of the final print in that color space. This will introduce artefacts, but as said, it's usually fairly easily avoided by selecting appropriate color channels and ink densities for your target printing process. Tools like QTR are great for this, but this is limited to Epson printers; for printers of other brands, you usually have to work with whatever functionality you can unlock in the printer driver dialogs. Even so, the problem I address here does not relate to increased noise that exists in the source file.

Another possible issue related to 'noise' in alt. process prints is the dither pattern of the printer that permeates into the final print. This effect can show up fairly strongly especially if the source file is already very noisy or grainy to begin with. In this scenario, the key problem is that interference effects between the printer dither pattern and the existing luminance noise in the image itself result in a relatively coarse-looking, gritty print. This is actually somewhat common with hybrid prints. Cutting down this effect relies mostly on optimizing the inkjet printer stage; again, selecting the most opportune ink channels and densities to be used is key, as well as more fundamentally selecting a printer that offers good performance to begin with (e.g. smaller droplet size/higher effective resolution). Again, this is unrelated to noise effects that are allegedly generated by the curve adjustment itself.

Finally, in some alt. processes, grainy/noisy prints are caused by the actual medium on which the print is made and process choices; e.g. brushed-on processes like cyanotype can end up somewhat grainy on some papers as a result of disturbance of the paper surface while applying the sensitizer. Some processes also can become 'grainy looking' as a result of chemistry choices; e.g. try increasing the amount of tartaric acid in a Van Dyke Brown sensitizer and see how the image ends up looking decidedly grainy. These inherent process parameters of course have nothing to do with adjustment curves.

The TL;DR of all this is that if you're trying to solve a problem with noisy prints, then start by thoroughly diagnosing the nature of the problem. I understand that at present, you have no particular problem to solve, but only a vague intuition of a potential source of a problem. That's not really a good position to start looking for solutions; simply put, without understanding what problem you're dealing with, there's not much you can do. So far, the problem that you do imply, does not seem to exist in the way you think it does, and I'd suggest not trying to solve it at all.

As @koraks suggested, working in 16 bit is a must.

I can see why one might think extreme curvatures can be problematic - it’s the damn Photoshop (if that’s what one if using) that does not allow close spacing of points. So where one might need to put more points, i.e. at the two extremes, it becomes rather difficult be accurate.

One way to solve this problem is split the calculated curve into 2 or more curves. So instead of say doing 10 (input) to 60 (output) on one extreme curve, do 10 to 20 on first curve, followed by 20 to 40 on 2nd curve and 40 to 60 on a third….and so on. Sum total of the series of curve should be the same as the original curve.

Another way to circumvent extreme curvature is first apply a gentler, manually smoothened curve (say that is 50% of the way.) Then print with this curve and determine a second curve which will now be much less dramatic than the single full curve. Print the final DN with the second curve below the first curve. This is probably more accurate way to do it.

:Niranjan.

Raúl Aragón said:

hi Carnie, the extreme curve creates video noise (electronic grain) because when you apply such an aggressive curve to any digital file (even sound file) what you are basically doing is amplifying it and that inevitably augments the noise, that's a fact.

Click to expand...

My name is Bob Carnie, here in Canada your response could be considered rude . I have been making negatives for PT PD since 1996 and that is a fact. Its obvious to me you have never made a decent PT PD.

@Carnie Bob I think @Raúl Aragón's way of addressing you is a misunderstanding based on your user name combined with differences in cultural backgrounds. Let's try to keep things respectful, and let's agree that none of us has seen any of Raúl's prints, so there's no way we can know whether they're any good (nor does it really matter for the technical question he asks).

Carnie Bob said:

My name is Bob Carnie, here in Canada your response could be considered rude . I have been making negatives for PT PD since 1996 and that is a fact. Its obvious to me you have never made a decent PT PD.

Click to expand...

I'really don't see why my response could be rude, in Switzerland where I'live people is addressed by their last name. Anyways if somehow I offended you I'sincerely apologize, even that I'don't see a justification for the lack of respect of your answer. The initial question was purely about the digital file, and I'do know a couple of things about digital files, since I've been shooting TV series and commercials as a Director of Photography for the last 27 years

Dan Pavel said:

Yes, I use a different approach for the DN creation that's developed exactly for addressing this sort of problems. I called it "Soft-Proofing Calibration" and a few year ago I presented the first steps of development here and here. I have since adjusted and further developed it and I am using it regularly.

Click to expand...

Thank you very much Dan!

nmp said:

As @koraks suggested, working in 16 bit is a must.

I can see why one might think extreme curvatures can be problematic - it’s the damn Photoshop (if that’s what one if using) that does not allow close spacing of points. So where one might need to put more points, i.e. at the two extremes, it becomes rather difficult be accurate.

One way to solve this problem is split the calculated curve into 2 or more curves. So instead of say doing 10 (input) to 60 (output) on one extreme curve, do 10 to 20 on first curve, followed by 20 to 40 on 2nd curve and 40 to 60 on a third….and so on. Sum total of the series of curve should be the same as the original curve.

Another way to circumvent extreme curvature is first apply a gentler, manually smoothened curve (say that is 50% of the way.) Then print with this curve and determine a second curve which will now be much less dramatic than the single full curve. Print the final DN with the second curve below the first curve. This is probably more accurate way to do it.

:Niranjan.

Click to expand...

Thanks! yes all my work is in 16 bit. I'll try this combined partial curves method and let you know how it went.
Raúl

Raúl Aragón said:

I'really don't see why my response could be rude, in Switzerland where I'live people is addressed by their last name. Anyways if somehow I offended you I'sincerely apologize, even that I'don't see a justification for the lack of respect of your answer. The initial question was purely about the digital file, and I'do know a couple of things about digital files, since I've been shooting TV series and commercials as a Director of Photography for the last 27 years

Thank you very much Dan!

Click to expand...

In my hometown addressing someone like you did is considered rude. I had a Swiss apprentice for three years and he did not exhibit this way of addressing others.
As far as your question, input matching output is the first thing one does to linearize a curve, It is a fact that one uses a steeper curve for Pt Pd but nothing that should cause artifacting, unless of course the original file being worked on is very small and being pushed beyond its limitations, These days modern cameras are producing good original files, back lets say 2002 the cameras we smaller mb and did not handle separation methods well.

koraks said:

@Carnie Bob I think @Raúl Aragón's way of addressing you is a misunderstanding based on your user name combined with differences in cultural backgrounds. Let's try to keep things respectful, and let's agree that none of us has seen any of Raúl's prints, so there's no way we can know whether they're any good (nor does it really matter for the technical question he asks).

Click to expand...

As said below I had a Swiss apprentice out of an Art School he did not show this difference in cultural backgrounds.... Misunderstanding on both parts I hope.

Carnie Bob said:

Misunderstanding on both parts I hope.

Click to expand...

I think that's safe to assume, yes.

Dan Pavel said:

Yes, I use a different approach for the DN creation that's developed exactly for addressing this sort of problems. I called it "Soft-Proofing Calibration" and a few year ago I presented the first steps of development here and here. I have since adjusted and further developed it and I am using it regularly.

Click to expand...

Thank you very much Dan, for sharing so generously, I'll certainly try it