Scanning negative

[bryancolvinfilm]

Have any of you guys scanned negative? Is there a good workflow for the correct inversion? In a very DIY test years ago I found that I needed to pull the orange cast to white, and then invert, and THEN start grading.

Ideally there’d be some proper conversion that gives a positive image in LOG starting point. A lot has to do with what camera is used of course… Has anyone gone down this road? I’m interested in building a kinograph so i can start shooting a lot more super16 without the big fat scanning cost… No sound needed, its just negative scans for dailies to edit with.

[matthewepler]

I’m interested in hearing others’ experience and expertise on this too. Right now I’m only concerned with positive scanning so we can get to something good quickly. But ultimately Kinograph should be able to scan negatives, too. I suspect this will require some RGB lighting, no?

[bryancolvinfilm]

I think even before RGB lighting there can be a completely software driven conversion with the normal scanning process. I wonder how the expensive scanners handle it…

[bryancolvinfilm]

It looks like the blackmagic scanner does it in software! (I’m assuming because of the following). I own DaVinci Resolve Studio, but there’s a free version as well… can’t confirm whether this LUT is in the free one as well or not… but

This is just a random google image but, here’s a starting point:

Apply the 1D LUT

Then you’re at the place of starting to adjust temperature and tint to get a good starting place

So temperature/tint lift/gain/gamma into something:

Anyway not super precise and just a google image. Not sure what that LUT is expecting to start with exactly, or how LOG profile can be worked into this but… hey I think it shows software route can work.

[bryancolvinfilm]

And i’ll add when scanning negative and the scan shows the negative edges by the perfs. the first few seconds is always flashed white on a camera roll when loading the film, so LOG aside, there is a reference for perfect black and perfect white supposedly. then if a grey card was shot, hey… in pretty good shape for being in the ballpark with gamma as well.

[cpixip]

Let me add a few comments from my experience: negative film stock is very different from color reversal material. The reason is the different usage of these materials: negatives are simply intermediates, while positives are used for presenting/projecting the material.

So a color reversal film will have very dark shadows. The proper exposure for this type of material is to expose to the highlights of a scene - they should be just at the brink of burning out. If they burn out, everything is lost in these highlights. On the other hand, during scanning, even in grossly underexposed image areas some scene detail can be recovered, if needed.

With a perfect exposure setting on color reversal film, the difference of the amount of light passing through dark image areas versus the amount of light passing through bright images areas is immense. This ensures an optimal viewer experience during projection.

In contrast, negative material is nearly always used as an intermediate. The proper way to expose negative material is to make sure that the shadows are correctly exposed. A too short exposure will simply not record any detail in the shadows. On the other hand, even severly overexposed highlights can be (sort of) recovered while scanning negative film stock by an appropriate increase of exposure during scanning.

Because negative film stock is never intented to be viewed, the contrast between dark and bright image areas is usually engineered to be rather dull. This ensures some headway and possibliities for correction in intermediate processing steps (think of the many copy operations involved in the optical printers used during the production of the first Star Wars movie, for example).

So, while scanning color reversal film stock requires a huge dynamic range of the scanning unit, with negative film stock you need to record a much shallower dynamical range. If you do not take this into account, you will just throw away some valuable bits of the dynamic range of your camera when scanning negatives.

Summarizing, a scanner optimized for scanning color reversal stock will show a sub-optimal performance when used for scanning negatives. Such a scanner will lack resolution in the shallow range of film densities a typical negative displays.

A scanner optimized for scanning negative material will show a substandard performance when used for scanning color reversal material. Such a scanner will be overwelmed by the strong contrast of the film stock

[bryancolvinfilm]

thanks for that valuable insight. It does sound like the negative intermediate image is the analog LOG haha. It does seem like HDR is more important when negative scanning to make sure you don’t toss aside valuable data… Something to ponder. Thanks!

[cpixip]

well, more bits certainly help. Our digital sensors are limited by two factors. On the low end of the scale, pixel noise becomes a challenge. With too high illumination levels, the sensors burn out, pretty much similar to the behaviour of a color reversal film.

That’s why you expose digital to the highlights. With negatives, it’s the other way around. This summary of a video comparing the behaviour of negative film and digital sensor might be of interest.

One additional thing a digital sensor is plagued with (in contrast to analog material) is the limited number of quantization steps. If the things you are interested in digitizing are located within a narrow band, only a few of the available digital levels will be actually used. So you might end up with banding artifacts in such a situation. Note that this comment of limited quantization depth applies to your whole digital image processing pipeline - there is a reason HDR and log image formats were developed.

[matthewepler]

Just a note to say thanks and kudos to the contributors on this thread. It’s been very useful and informative!

[friolator]

@cpixip

Summarizing, a scanner optimized for scanning color reversal stock will show a sub-optimal performance when used for scanning negatives. Such a scanner will lack resolution in the shallow range of film densities a typical negative displays.

You can work around this by using RGB lights instead of a white light source. That is, mixing the RGB to create a “white” that’s appropriate for the film you’re scanning. This is how scanners like the ScanStation do it - the lamp house uses Red, green and blue LEDs to mix light that looks basically white when you’re scanning reversal or print. But if you scan color neg, it has a slightly bluish tint. (and there are some tools on the software side to manually override these and adjust the balance as you see fit - at the light source, vs color correction after the image is taken)

Our Northlight has a filter for neg that’s removed when scanning positive film. This is a physical filter that the scanner inserts between the white light source and the film, in the appropriate situations. Our old Imagica scanner had a monochrome camera and used filters on a complicated mechanism using gears and belts and solenoids to put the correct combination in place.

On our 70mm scanner, which is intermittent motion and a mono camera, this all a lot easier. So we built an LED lamphouse that has 4 channels: Red (print), Red (neg), Green, Blue – plus IR in case we decide to implement a dust map feature eventually. Each channel was chosen to best match the sensitivity of the sensor we’re using as well as the film types themselves.

[cpixip]

Let me introduce for reference this document by ARRI, especially chapter 2 were several characteristic film response curves are discussed.

In any digitizing of the analog world, you need to take care that:

• you can cover the whole dynamic range of your signal
• you have enough bits available for all intermediate levels of your signal - otherwise, banding/contouring might occur.

Basically, these two aims (large dynamic range/densely spaced quantization levels) are contradictory for a given hardware. On the other hand, the dynamic range of sensors available today, combined with adjustable light sources and appropriate image-processing algorithms should be able to handle this.