Suggestion to RED

[Matthew Greene]

I think some people have tried to use Bayer in one exposure per frame to scan with, the J&K scanner seemed to work that way, maybe some others. Not all labs let you into their scanner room to see what they are using, and there is pressure to run everything as fast as you can, so no commercial scanner is going to make 27 exposures per frame, even if you can see a difference.

I think some 2 1/2 sensor backs have been used in some labs for special projects, and those must be Bayer?

I was talking with a guy in a LA lab who wanted to use a Nikon DSLR and start a scanning lab, so people have the idea that one exposure is enough. My tests show that you can get an images, but the color is not as good as separations even with extra software processing, like they say, "garbage in, garbage out".

The one good thing about the Canon XTi for making seperations is that it has good dynamic range on the brackets, and I can hold both ends of most dense negatives in three exposures per color. But its noise level is way to high and with 27 exposures per frame it will break after a few thousand feet scanned if that much. So DSLRs are not a good idea for anything other than testing your projector, lens, and light house, at some point getting a good monochrome camera will probably be on the purchase roster...

First off I want to clarify I'm discussing scanning negative stock below, (not reversal). There are numerous issues when using a Digital SLR, one is the lifespan of the mirror shutter, you'll be lucky to shoot 100K frames before the DSLR mechanical shutter dies (unlkess you use a system with a shutter lock), another is inconsistency between frames due to the temperature buildup on the sensor which will cause noise and color variations between frames. One more issue is speed... yes DSLRs can burst several FPS but then have a wait before being able to burst again. I've been there/done that. I've also been lucky enought to either do some dissection and/or maintinance on the Spirit Datacines and Rank telecines and even the old Kodak Cineon systems in the past and took in all the good and out filtered the compromises. Start with the perfect light source, get as much usable image mapped on the sensor. By this point the image should look good. All post processing be it on a DaVinci, Pandora or Software based system should be limited to minor cleanup and artistic decisions as far as I'm concerned.

I want to reiterate that multipe exposures for capturing the full latitude of a negative are not needed in negative film... The whole latitude of the film lives in a very shallow density range. The magic comes when you try and expand that to data world without leaving banding and gaps in gradations hence the desire for high bit depth imagers and processing. If using color separations on Monochrome images, yes, you want at least 3 exposures for the primaries plus perhaps the option of some additional IR/UV (plus a special pass I've been playing with) exposures for masks as post processing masks for artifact removal. When talking about a S8 or 16mm scan from a bayered sensor such as RED's I don't feel that there's much compromise in even doing a single pass when it comes to color rendition and latitude. There are other reasons I would do multiple passes in those scenarios but I'm keeping them to myself at the moment till I experiment some more.

P.S: I personally find the 27 exposure approach a bit eccentric although maybe it's necessary in some situation I've yet to come by.

 

[Dan Hudgins]

A2d...

A2d...

P.S: I personally find the 27 exposure approach a bit eccentric although maybe it's necessary in some situation I've yet to come by.

Because the A2D converter in the camera is not giving you real data on all bits, if you make several exposures while ramping the lighthouse brightness you can get closer to real data when you average the several exposures.

What camera are you using that you would not be able to see or measure an improvement in by averaging exposures?

As was mentioned if the pixel values from each scan vary below the 6th or 8th bit you are not getting 12bit data.

In compressed files the data deltas are restricted by a threshold that "acts like noise reduction", this throws off the real data values for each pixel, so if you make multi exposures on a compressed camera the average may be no better than a single exposure, unless you vary the brightness of the light source to add chaos into the compression thresholds.

Unless you properly compare the results side by side for one or more exposures averaged and done with useful chaos and bracketing you cannot say that there is no improvement.

27 exposures would be 9R,9G,9B, with 9 being the point where there is not much more visable improvement, you might get by with 5R,4G,10B being better since the noise level in the blue may be worse.

Bracketing the exposure is not so much to extend the tonal range if the sensor has enough dynamic range, it is to move the tones in the film up and down on the Digital Camera's transfer curve, so that irregulatories in the A2D steps and other sensor issues such as bloom and noises get averaged out.

We scan black and white negatives and other odd stocks and they can have
densities that are very high. Bracketing can also alter the printing lens flair issues if you roll off the HDR exposures before fusing them.

When viewing the scans if you boost the contrast, saturation matrix, and sharpen you can see the multi exposure improvements easer. Since scaned images will get more contrast, saturation, and sharpen later by the end user, images that can hold up better to those may be worth making, even if you can get by when looking at them flat and soft on the grading monitor. Also you need to look at them 200% size since the monitor cannot reproduce 1:1 resolution, so display 200% on the monitor and step back, but not so far as to have your eyes not be able to see single pixels...

Anyway, if you can say what camera you are using to scan with I can look into what advantages it might offer.

 

[Dalibor Fencl]

Hey guys don´t forget all the billions feet of film that was shot in last hundered years. Those will all be scanned into 4K one day.

 

[Dan Hudgins]

Who.. How many... How much...

Who.. How many... How much...

Hey guys don´t forget all the billions feet of film that was shot in last hundered years. Those will all be scanned into 4K one day.

Film sits well in cold dry storge.

For marketing a "end user film scanner" one needs to ask:

1) Who would want to scan their own film other than a few labs and service companies?

2) How many units would that be, 10, 25, 100?

3) How much would it cost to make the lowest cost scanner that would produce results that would be so good than any scanner costing more would not produce a visable advantage?

In asnwer to 1) I do not think many people want the hassles of doing their own scanning and clean room editing to prep rolls for the scanner, if it is just to convert archives will they make any money off the Digital versions, if not then just scan when asked and leave the 98% of the other stuff to rot in time.

In answer to 2) with few new filmmakers, and in 10 years maybe none, only a few film archives and labs would need to digitize film, like Quadruplex videotape is now, maybe just one scanner in the world left running in a back room somewhere.

In answer to 3) it is probably possable to make a film scanner that runs at 4 to 8 frames per minute for between $5000 and $10000 msrp in volume of 100 units or more, but the profit would be small. And could you sell 100 units at that price?

And a bigger issue is that if you scan billions of feet of film at 4k what media do you archive that data on that does not cost more to store than the film you scaned? Most data formats are less stable than 35mm film, so someone needs to copy, back-up, and update the data archive, and that would cost more than the old film could ever be worth, so who would pour money into a black hole like that?

Most film is less than HD resolution and not 3D, so in the future it will look more like silent movies do to us today, and would only be if interest to a small audience of historians.

I am happy to help anyone who wants to design a low cost portable scanner at no cost, but you get to pay for manfufacturing contracts...

If someone wants to set up a service using a RED ONE (tm) to scan at 60fps they might be able to reduce the cost of scanning to get some volume business and recover their cost, but they would need to address the archive storage media for the digital files. REDCODE is not tuned for color negative scans, so that would make for some more issues with the compression artifacts...

The dyes in color negative films do fade over time, not as much as older prints, but some, so if some films are going to be saved they will need to go to digital at some point. Black and White films (separations) can be washed and stored cold and dry for maybe 400 years, so they are in better shape, but it depends on the quality of the original lab work how well they will keep.

What future film will have is murky now, as the sales volume falls a point is reached where the last lab will close and the last emulsion for dupe negative and master positive film is coated. The day will come when you cannot make a film to film print any longer, probably not for 10 years but sometime within the life of many now on REDUSER perhaps. I would like the quality of the digital versions to be as high as cost would allow (uncompressed 4K TIF 48bpp), but don't see it going that way for most of the film that has been shot in the last 100 years...

 

[Matthew Greene]

What camera are you using that you would not be able to see or measure an improvement in by averaging exposures?

I have used many cameras from Panasonic DV cameras to Machine Vision to RED with the best results I've had being from a "borrowed" Hasselblad with a (I think Phase One) digital back followed by the Panny Varicam.

Even though it's not common practice, I don't argue that averaging multiple exposures wouldn't produce more accurate scans. In my playing around I just haven't had/felt a need for it. Most times I feel the results look pretty good but I haven't approached the subject with the intention of getting the most technically accurate scan nor have I done film outs of the material with the intention of preservation. My uses have been more in the likes of a telecine for SD or HD.

 

[Dan Hudgins]

pixel+pixel=

pixel+pixel=

I have used many cameras from Panasonic DV cameras to Machine Vision to RED with the best results I've had being from a "borrowed" Hasselblad with a (I think Phase One) digital back followed by the Panny Varicam.

Even though it's not common practice, I don't argue that averaging multiple exposures wouldn't produce more accurate scans. In my playing around I just haven't had/felt a need for it. Most times I feel the results look pretty good but I haven't approached the subject with the intention of getting the most technically accurate scan nor have I done film outs of the material with the intention of preservation. My uses have been more in the likes of a telecine for SD or HD.

If you are using a 2 1/4 back to go down to SD then you are probably averaging several pixels anyway?

I have been wondering about how the RED ONE programs down res the 4K to 2K and 1920, if you use real numbers or scale to 32bit for the average of the 4K pixels you should get more than 12bits in the result, like about 14bits, going out to 16bit TIF you would then have less "Posterisation". If they limit the bit depth of the output then you do not get the full benefit of the down res.

When you see side by side single vs. 3+average you will probably see some improvement after S curve, contrast, saturation matrix, gamma adjustments, sharpen, all of which tend to make the noise come out. I have looked at single bayer vs. multi averaged separations and there is a clear improvement in the cleanness and smoothness of the images when viewed side by side.

Film uses different peak wavelengths than digital cameras, so when you use the film seperation filters to shoot through you get more film like results, sort of. Wratten 99, 98, and 70 are a nice set with 2E+2B and IR block and dichroaic bandpass, heat absorbing glass, and maybe a XB30 (or XB29?).