Hi, does anyone know what exactly the “auto grain size” does, as far as scaling goes? What does it reference to? The sensor size, the resolution?
Does it scale itself so that when I choose “auto: 24x36” it applies a grain layer as if the sensor was a 24x36 film negative regardless of the sensor resolution? (Basically getting the same grain size on a 16MP sensor as on a 26MP sensor, both being APS-C)
When I look at 100%, what am I looking at in terms of print size then?
Or is the grain layer scaled taking resolution into account, like Lightroom does (pixel based)?
PS please no LR vs DxO digression, thanks in advance…
Yes, the “grain size” corresponds to what would be visible on the original film of the given format: 24 x 36, etc. It is therefore independent of the resolution of the sensor, and therefore of the size in Mpix of the file.
This is consistent since “historical” films were generally available in all three size categories: 24x36, Medium format, Large format.
We must not think in terms of a 100% screen view, but in terms of printing on a paper of fixed dimensions.
For example, whether your sensor is M4/3, APS-C or 24 x 36, the grain obtained on an A4 print will be identical if you choose “Auto 24-36”. This will simulate the grain that would be visible on the enlargement of a print of the same format from 35mm (24 x 36) film.
“Auto” simply means that FilmPack calculates the grain size according to the format of the sensor to give an identical result, including if the image has been cropped by checking “Adjust size to current crop”
OK. I thought there would be some sort of ponderation taking sensor resolution into account when setting this to “auto”. I do check at 100% screen view, because first of all I know what print size this relates to (I do indeed edit for prints). The problem I encountered not doing so is that my sensor resolution and lens sharpness can produce results that are impossible with film emulsion: there’s an amount of sharp detail to be seen that can’t logically exist with the grain size that was applied.
So IMHO the “auto” idea is cool, but needs to be refined. We are/will be seeing very high pixel count in small sensors these days, and the current “auto” algoritm should evolve to follow that trend and produce a realistic grain-to-resolution ratio when printing large.
I think that the main problem I’m facing is the very high resolution of my source files when shooting low-light photography, which I do a lot. The fine details I get out of my Fuji X-H1 26 Mp sensor with the XF50mm f/1 lens are not proportional to the grain size of the film simulation I would have shot with in such situations (in my case: Kodachrome 200). On well lit daylight scenery, when I would have used a 64 ASA film or (or 25) the proportions are correct when using the equivalent film simulation, as long as I switch off any sharpening.
So, having put a lot of thought in it, I came to the conclusion that DxO filmpack does a great job but within certain limits. Very high resolution sensors and lenses are not well suited for realistic results when using the 24x36 auto setting, which is what I went for naturally, using APS-C. Selecting MF ot LF auto is better for that, or using a diffusion filter on the lens to tone down the razor-sharp quality of the resulting file a bit.
When producing files for social media it’s not much of an issue, but when you print for exhibition-grade results this must be taken into account. Like: either you don’t use grain at all, or you should get it perfectly right. I think I will, from now on, save a final tiff with grain, and another without grain when archiving.