Hi. Ive followed every rule I can find for DSLR scanning and also done some experimentation. Im using a Nikon D800 with Nikon 60mm Macro lens. (I have also tried with a borrowed Nikon 105mm)
but keep getting really bad noise/grain in the scan. Im using a Lomography Digitalizer Light Panel. Ive tried to de noise via Lightroom but start getting weird artefacts in the scan. Im shooting on Kodak Portra 400 in direct sunlight and in shade but still getting the noise. Ive also scanned on a Frontier scanner which is producing the same colour but doesn’t have the grain. Am I m missing something? Thank you.
you are not saying if this a crop, what are the camera settings (ISO specifically) and what is film format. But it does not look like noise, many people would kill to get reds like yours in camera scan
The frontier scan appears less sharp because it has been denoised. Focus seems to be off to the ventilation grid rather than on the person. NLP reds are more saturated and WB looks different too. Noise (or whatever it is) is more prominent and some noise reduction could help to remove it.
Have you tried to reduce the noise in the converted positive, @Jacqwey?
BTW, many minilabs/scanners are three pass scanners. This should a) raise the impression of sharpness and b) lower noise due to averaging the luminance part of each colour. You could try to do the same with median filtering, something that Lightroom can’t do, but many of us also have Photoshop.
Hi Vlad
It’s a full frame shot, I’ve just cropped to show the noise more clearly.
Ive remained on 100 iso throughout the test scans with various speed and f stop ranges but always the same result. I’ve tried overexposing up to 2/3 stops and underexposing never more than 1/2 but noise remains very visible.
I crop tightly and take the WB reading on the outside of the frame as recommended. I’ve tried scanning both sides of the negative but nothing changes.
There NLP settings are never touched, I leave everything on Zero and have tried all the presets in NLP but although there are colour shifts when using the different presets the grain/ noise is always terrible.
I have made hand prints from the negs in the dark room and can confirm they are perfectly exposed with only typical Kodak Portra 400 grain.
Yes I did actually try a few noise reduction techniques but felt although it did work to a point in also seemed to take away that film look and create a sort of banding in the texture of some of the clothing. I will have to experiment more with this.
And yes you are correct, the WB is different between the Frontier as the reds from the Frontier are off, a little more orange so I was balancing for the correct red in NLP which resulted in a slightly cooler background.
What you don’t see is the model is leaning on the wall and I’m focusing on her face so more inline with the grate of the wall. Well spotted.
…it’s difficult to check and experiment with a cropped RGB image and it usually helps a lot if you can attach the original RAW file or a link to a share containing said file(s). Alternatively, you can use services like wetransfer.com taking care to make the images public for some time. Not all of us have accounts with dropbox, Google or MS drives etc.
Due to current data protection regulations, you might want to ask the models if it’s okay to publish their images.
What are the pixel dimensions of your 2 scans? Your 36MP D800 will give you a maximum of 7360 x 4912 px but I didn’t think that the Frontier could give that kind of resolution.
Hmm, i can only suggest one more rather exotic trick. I am not sure what light source you use, but light with lack of blue component would cause certain elevated noise as orange mask will diminish blues even further and to compensate for that the gain is increased in Blue channel and that may cause noise. So you may introduce something like Wratten Blue 80A filter ( gel placed over light source should be fine) . My experiments with CS-lite and plastic 80A seemed to improved color balance - you would see orange mask much less orange in NLP. Given you have Lightroom you must also have Photoshop, Try Grain 2 pixel plugin (which is free and very easy to install). The latest ver 5.5 gave me very good results. That may give you clues about the nature of noise.
@Jacqwey , @VladS , I’ve tested NLP with light from an enlarger head with dichroic filters set to a) compensate the orange mask and b) to neutral. The conversions I got with these tests were close to identical in respect to colour. I’ve not checked noise though…and I think I’ve deleted the test
I might find a backup of the test captures, but you might be quicker in testing it on your rig. But from what I’ve seen, I’d probably build a discrete rib backlight with more blue lens for increased power.
Thanks, would you be able to give the pixel dimensions of each in that case? In your side-by-side comparison you have made them the same size so one is either enlarged or reduced to match the other.
Thanks Harry but I have found the problem. Ive used the exact same process but changing camera from the Nikon D800 to the D850 and the improvement is remarkably better. So it was the D800 creating the noise all along.
OK, but the despite its age the D800 is also very good camera and sensor, a high dynamic range, and you are using (presumably) base ISO so that shouldn’t be a factor. The D850 has 45MP against 36MP, I’m guessing that the Frontier scan is 20MP at most.
Due to current data protection regulations, you might want to ask the models if it’s okay to publish their images.