Preserving Grain Structure in Film Scanning

Bill DiPietra · June 6, 2015

As it looks like I will be getting either a 2K or 4K scan for my 16mm short film (and not doing a traditional A/B negative cut,) my main concern now is preserving the grain structure that I achieved in-camera.

I went for a specific look with this film and pushed 7219 2 stops, creating a nice amount of grain. As I read and look online and different clips on Vimeo & YouTube, I see a lot of 2K scans that look too pristine. Not all, but a lot. So my question is will I still be able to retain that kind of gritty look with a 2K scan, or will it defeat the purpose of the original look I was going for?...

Bruce Greene · June 6, 2015

You should be able to keep your grainy look through DCP.

But, highly compressed vimeo/youtube will be more difficult. The grain detail may well overwhelm the long GOP compression giving it a blocky grain look. Compression is most effective with clean images.

Kenny N Suleimanagich · June 6, 2015

You’ll definitely keep some grain in a 2K scan. Especially with your +2 process. I would even go for 4K, if you can budget for it, since compression will definitely make too much grain look blocky and blotchy.

David Mullen ASC · June 6, 2015

2K for 16mm is the same as 4K for 35mm, you should capture the grain just fine. I believe the Spirit Datacine that some people use though doesn't use a point source for the light it shines through the film and thus grain is less hard-edged, less sharpened, so you may want to avoid that device. Some people preferred 16mm transferred on a Spirit for that reason.

Also, make sure that your scanner / D.I. techs aren't employing any noise reduction in the process, thinking they are helping you out.

Bill DiPietra · June 6, 2015

You’ll definitely keep some grain in a 2K scan. Especially with your +2 process. I would even go for 4K, if you can budget for it, since compression will definitely make too much grain look blocky and blotchy.

Many thanks, everyone. Yes, I was told that 4K will look grainier simply because there is more detail being captured. I can afford it, but it also seems like overkill now, so the only reason I would go 4K would be to maintain the grain structure.

Also, I calculated the amount of data I would being getting if I had everything scanned into DPX files. I'm still trying to find a colorist and I might need to just use one at whatever lab I wind up bringing the neg to for the scan. Obviously, I would have everything on a large drive, but how easy are those files to work with on an external drive?...

David Cunningham · June 7, 2015

For proper grain reproduction I find the smaller the gauge the more it maters. For example, I have found a remarkable increase in the perceived sharpness of an image from both super16 and super 8 even as high as 5K.

There is certainly not enough "resolution" in either format to justify this high resolution a scan. However, properly resolved grain becomes that much more important when that much more of the image is grain.

These days I actually overscan my super 8 and super16 at 5K which is immediately down sampled to 2K. I edit and reframe my image at this resolution and output to 1080P. I find little thing like distant street signs, grass and leaves on trees just appear a bit sharper when scanned this way.

Bill DiPietra · June 7, 2015

These days I actually overscan my super 8 and super16 at 5K which is immediately down sampled to 2K. I edit and reframe my image at this resolution and output to 1080P. I find little thing like distant street signs, grass and leaves on trees just appear a bit sharper when scanned this way.

Is that kind of like doing a 1.85:1 reduction print from a VistaVision camera negative?...or am I oversimplifying?

David Cunningham · June 7, 2015

Kinda but more importantly you are accurately resolving the whole film grain. As the grain gets larger lower resolution scans tend to make it look more chunky due to grain aliasing. This softens the image and kind blocks up the grain.

Carl Looper · June 17, 2015

Normally you would scan film (regardless of gauge) at the highest sampling rate possible.

And then filter it and re-sample it at half the frequency of the scanning frequency (or even less if your release format is smaller)

Scanning creates a "false grain" (called grain aliasing) which always makes the transferred image grainier than what would otherwise project in a film projector.

But it depends, because by "gritty" you might actually mean this false grain.

C

Bill DiPietra · June 17, 2015

Normally you would scan film (regardless of gauge) at the highest sampling rate possible.

And then filter it and re-sample it at half the frequency of the scanning frequency (or even less if your release format is smaller)

Scanning creates a "false grain" (called grain aliasing) which always makes the transferred image grainier than what would otherwise project in a film projector.

But it depends, because by "gritty" you might actually mean this false grain.

C

I'll know when I see it, Carl. Thanks.

Carl Looper · June 18, 2015

I'll know when I see it, Carl. Thanks.

Yeah for sure. Your eyes are your best instrument.

I'd recommend 4K anyway because grain aliasing (if that was ever desireable) can always be reconstructed from such.

There's not much online information on this phenomena called 'grain aliasing'. The earliest online insight into such appears to be the following article (Sept 2000), in which the term "grain aliasing" could very well be the first usage of the term (at least online):

http://www.photoscientia.co.uk/Grain.htm

It's certainly a visible phenomena and discussed in many forums, but it appears to be under-theorised. For example there's no Wikipedia article on such a thing (and you'd imagine there would be, even if a badly conceived one). Simple solutions exist - which follow on directly from conventional aliasing theory. But optimum solutions based on more insight into grain aliasing in particular would be preferable.

As previously mentioned, in conventional aliasing theory you would sample (or encode) a signal at twice the display (or decoding) rate, with a filter stage in between. But for grain aliasing this may not be optimum. To minimise grain aliasing you might need to sample at more than 2X. But how much higher? 4X? At the moment, it seems just the higher the better.

The relationship between gain aliasing and what might be called the "native grain" of the film remains to be determined. Indeed it remains uncertain whether "native grain" can actually be disentangled from grain aliasing (ie. as an independant concept).

In other words the choice of sampling rate (and looking at the results) appears to be the only way of currently determining what kind of grain a result will have.

It would be good if there were some more research and theorisation available in this area. Would certainly help to optimise film to digital workflows.

C

Edited June 18, 2015 by Carl Looper

Robert Houllahan · June 18, 2015

Most new scanners scan at 1.5-2x the 2K resolution internally and down sample for 2k output which resolves grain better in the final scan. I would avoid the Spirit SDC2000 because it's 2K output is actually slightly uprezzed because the line arrays are 1920 pixels so you won't get the advantage of over-sampled resolution.