Session Digitally Generated Color Separation Masters

Presenter

 Phil Feiner
President, Pacific Title and Art Studio 


ABSTRACT

The latest film-to-digital-to-film post-production process for motion pictures, the Digital Intermediate, represents a new problem for preservation, and perhaps also a new opportunity.  The data files which source these intermediates are often fugitive and difficult to recover or deploy by the time the film has entered the “library” phase of its existence.  Data files captured on LTO or DFT2 data tapes are sitting on vault shelves of production companies or in archives in untested form.  But these files do offer the possibility of recording out perfectly matched and registered black-and-white color separation masters at the same resolution as the digital intermediate, and these synthetic elements may comprise a part of a preservation regime for the Digital Intermediate.  The test material to be screened was scanned at 4k, processed at 2k and recorded back out to color separations at 4k, then recombined and printed conventionally

PRESENTATION

Good morning, ladies and gentlemen. Thank you very much for this opportunity to go through this new process that we’ve nicknamed Rosetta, for making digital YCM masters directly from the data from which the DI (digital intermediate) is made.

The first question that may occur to you is “Why did we come up with this process?” The answer is that this method of recording out color separations directly from the original DI data produces a result that is superior in quality to color separations generated by the traditional method -- that is, by the photochemical separation of the DI negative. This process is superior because it allows for complete control of timing in the pre-recording stage and eliminates a generation in the film printing stage. And it contributes significantly to the preservation cycle of the DI process because it provides an independent check-path for validating the integrity of the DI data tapes. When the digitally generated color separations have been recorded out, recombined and printed, the recombination check print constitutes proof of the integrity of your DI data.

The Rosetta process produces pin-registered YCM masters. The traditional way of producing masters from the negative is with a continuous contact printer, which allows a certain amount of film movement, and thus the three images of the recombining separations don’t exactly overlay one another, resulting in softer, less accurate images. Because our process is controlled in data and skips a generation in printing, it offers an increased scale range that more accurately preserves the cinematographer’s creative intent.

This graphic [cite PowerPoint slide] illustrates the DI process including our digitally generated color separation phase. The process will be basically the same whether you scan film or ingest HD files. You go through a normal DI color correction stage with the cinematographer, and once you get finished data files you send them off to a laser film recorder that produces a DI (digital intermediate) negative.

That negative may be used for theatrical release printing. In some cases, multiple DI negatives will be recorded. In other cases, when the distributor doesn’t want to pay for multiple record-outs, only a single DI is recorded and then an interpositive and a printing internegative are made from that DI negative.

Then, in the traditional process, color separations (YCMs) are made from the DI negative by means of continuous-contact printing. Each of these YCMs is individually processed in a developer, with the speed of processing determined by the gamma requirement of the individual separation.

Once the YCMs are produced, they are continuous-contact printed to a recombined internegative, and this internegative is check-printed to verify that the separations have been correctly made and are free of printing (wet-gate) or developer problems, and matched against a studio guide print to establish that the separations have correctly rendered the look of the original release. Studios follow this practice so that they can be sure that what they vault as the ultimate protection element for their film is in fact a viable source for long-term conservation of the asset.

The process that we have developed records out the data itself directly to YCM color separation masters, bypassing the DI negative altogether. It is a full generation closer to the original data source than the traditional method. Our process will ingest 2k data files and up-rez them to 4k files. This stage reduces substantially the problem of aliasing. As demonstrated in our test film, the 4k output renders a much more accurate image than a 2k output. Once the digitally generated YCMs have been recorded and developed, we recombine them by pin-registered contact printing (which is considerably more accurate than continuous-contact printing) used in the conventional scenario, to an internegative. From this internegative, we manufacture a check print, which we compare with a studio guide print to insure that our process has in fact successfully rendered the intent of the filmmakers.

So in comparing the two processes, we manufacture a preservation element (YCM color separation) directly from the data, whereas traditionally, the preservation element will be made from a 2k DI negative. Our 4k record-out process eliminates aliasing and some forms of digital artifacting that may have been interposed into the 2k negative. This means that our process is a full film generation closer to the original source, and considerably closer to true film resolution. In addition, our recombination of the preservation element uses pin-registered contact printing rather than the less accurate process of continuous-contact printing. We will be testing this process with resolution charts to quantify the advantages of our process.

We believe the color separation masters produced by our process are a true representation of the original digital data. Because of the archival nature of our product, we exceed the ANSI recommended standards for elimination of residual hypo – sodium thiosulfate, with less than one unit of hypo per square centimeter in our YCM separation masters. To substantiate our claim regarding the absence of residual hypo in our process, we sent some of our film samples back to Eastman Kodak in Rochester, and they graciously analyzed the film. We were a little surprised when the film was rated at zero micrograms of residual hypo. Residual hypo in your separation masters may result in increased base density over time, which may introduce flickering in the images produced by the recombination of these separations. We presume that our hypo-free separations have a shelf life in excess of a thousand years. You can consult our bibliography, which contains some of the Kodak estimates that corroborate this claim [cite bibliography]. For example, one paper suggests that polyester-based film with silver will last in excess of fifteen hundred years if properly stored under recommended archival conditions.

I want to thank Grover Crisp and Sony for supplying us with the digital intermediate files on “Stuart Little 2,” that we used for the test and this demonstration. This film was a unique opportunity for our test because the color separations had already been made conventionally and accepted by the studio, and we were able to compare the check print from the conventional recombination to one made with our process. You’ll notice several differences. The conventional print has more weave, and the 2k recording process introduces aliasing or “jaggies.” In the print from our process, we have used proprietary algorithms to up-rez the 2k data to 4k and to eliminate the aliasing artifact. [cite comparison images -- PowerPoint slide]

You may notice some of the other advantages. There is no dirt printed into the YCM masters. In conventional printing, every time you print the negative you tend to accumulate dirt, which is why we see the characteristic red, green, or blue spots in elements derived from traditional separations. The YCM elements produced in our process are pristine, and thus yield a virtually dirt free recombined negative from which to print. The digital YCM masters that we’ve developed are processed at one gamma, not three, which makes recombination printing simpler.

Our process produces a 4k separation master, so it is natively 4k compliant. Our process is engineered to accommodate DI technology when it migrates to the 4k level, as we believe it will.

+ + ++ ++ ++ ++

What we’ve tried to do is create a “future-proof” solution to the problems of preserving the images of a digital intermediate. Last night Mr. Silverman gave a very eloquent speech referencing many of the preservation issues relating to the digital intermediates that are the basis of many theatrical productions now. And the digital intermediate, as a finished program, exists in the form of a data tape which is something like the traditional original negative in that it is the highest resolution source for the reproduction of all derivative printing and protection elements for a given feature.

Film and acid-free paper can, with proper care, last for centuries, but the magnetic tape systems we’ve developed, whether for video or for data, will last only a few decades. Some of our colleagues such as Rob Hummel and Bradley Hunt have pointed out the fugitive and complex nature of electronic systems by comparison to film’s simplicity – “in 1500 years all you’ll need to do to read back the YCMs is use a magnifying glass with a light bulb.” Milt Shefter provided a quick overview at NAB of the 60 or 70 now-obsolete tape formats [website or PowerPoint citation] that have supported video over the last fifty years. When we talked about this within out IT group, we realized that in addition to obsolete video formats, there are also another 70 or so computer tape formats that have also been used to store electronic data. And in excess of 95% of these formats are obsolete, which means that playback equipment is no longer being manufactured, the formats are incompatible with current equipment, and the tape medium itself is aging and deteriorating. We won’t have to wait 1500 years to encounter problems with these systems. So we think the only viable preservation medium for the digital intermediate system is polyester based silver film.

I’d like to go on to the film presentation. What you’re going to see is you’re going to see will be a comparison of a film sample from a DI negative, separated to traditional YCMs (by a reputable lab in Hollywood, which did an excellent job), and recombined and printed. And following that, we’ll show you the same segment from the DI data, up-rezed to 4k and recorded out a new negative and printed to match the supplied original guide print. Among the visible improvements you may note reduced granularity and an extended tonal range.

The prints were made using the same emulsion batch, processed through the same lab, printed on the same printer, so that we achieved an even playing field for purposes of comparison. We attempted to match our print to the contrast of the supplied release print, but you will notice more contrast in that guide print.

[Film demonstration #1: DI data to DI negative to print at 2k]: This is what the audience saw, the first generation print from the DI negative.


[Film demonstration #2: DI data to DI negative to print at 4k]: We matched the contrast of the original. Look at the scale in the flesh tones. We matched the contrast of the original to a very great extend, but you’ll notice more contrast build-up in the print manufactured by traditional methods. This contrast is difficult to control in processing, but we believe the contrast values in the print from our negative are superior. In the scene with the large moon in the sky, you can see the superior grain characteristics (less grain) in our print. And we want to minimize grain as much as possible because the process of recombining the YCM separation masters tends to intensify the appearance of grain.

There should more development work on the current separation film emulsion to improve the grain characteristics.

[Film demonstration #3 – traditional YCMs, that is, DI negative to black-and-white YCM color separations to recombined negative to print]: This print was produced by the traditional process – that is, by color separation of the DI negative, recombination of the separations to a new negative, and then printing that recombined negative.

We generally don’t like to show the single stimulus like this, we like to show the film samples side by side so you can see both images on the screen the same time for comparison purposes. The differences are subtle but when you project them side-by-side you see the differences more clearly.

[Film demonstration #4 – digital YCMs, that is, DI data to rendered directly to black-and-white YCM color separations to recombined negative to print]:

Look at the tonal range in her flesh. And you’ll see less grain in the neutral areas as well. Now, when analyzing the film clips under a microscope you really see the sharpness difference. Look at the grain of the 2238 film emulsion. In order to get the correct gamma values you have to boil up that film in the development process, and this unfortunately accentuates the grain.

If, as we believe, the majority of all feature film production is going to digital intermediate, we’ll be facing this problem with nearly every feature when we attempt to manufacture preservation elements. We saw this problem in the evolution of the DI, and we began to develop this solution about two years ago. We have established all the necessary steps of this separation process to fit into our laboratory regime. We have already produced price quotes for several studios to move forward with the rollout of digitally generated separation masters. If you want to preserve DI – generated films, you’ll need to produce a new class of preservation elements. The digital color separation master process that we have introduced solves to specific preservation problems. First, it verifies that the DI data that you will be vaulting is complete, error-free and capable of rendering the feature back to film or across to other media, and second, it provides the highest quality film elements possible for future film work or for scanning to other digital media. If you are in the business of protecting contemporary motion picture assets, this is a process you should consider.

Q & A

Tom Scott (SMPTE and Visual Data): One of the biggest problems our industry faces in gearing up for this kind of fundamental change in technology is the speed with which these things can be accomplished. I’m wondering what you can say about how long it takes right now to accomplish this kind of work, and what you see in the future to speed up these processes, given the tremendous volume of material that has to be archived.

Phil Feiner: The technology in our industry has already changed. Films are already being produced with the DI (digital intermediate) process. I believe 40 went through last year, and probably this year there will be 60 to 80. So the conversion to DI is well under way. Digital separations are based on the same technology as DI. We have seven Arrilaser recorders for DI work, and we use the Arrilaser to record out separations as well as for making feature negatives. As far as the pipeline for the separations, we’ve got the developer configuration and the printing process set up to accommodate the requisite black and white output. As an example, this test we screened today went through the entire process from record out to developing to recompositing the print overnight. From receipt of a purchase order until delivery, the process for a complete feature takes about six weeks, and today it is about 33% more expensive than traditional separation methods.

Rodney Vilnuff: What do you do with the digital data?

Phil Feiner: We don’t do anything with the digital data in the sense of changing it. The studio supplies the digital data and we load the data to our drive, buffer it and then record it out to film. We use the same data source that was used by the DI vendor to produce the DI negative. But one of the important parts of the process is that it validates the data.

Interestingly enough, we’ve done a test for another studio with DI data provided by an outside facility, and we got a corrupted print. The studio thought that their LTO tapes contained all the images of the film, but when we read them in we found that there were several bad frames in the data. This should underscore the importance of checking the data you are going to put on the shelf for years or decades. We’re currently working in a non-standardized context where there are almost as many types of data and media as there are facilities, so with all these “flavors,” how will you know if your data is good, if it is rewritable or re-recordable, without this validation process?

Rodney Vilnuff: I agree that your process represents an intelligent approach in terms of producing the production support (DI) as well as the archival support (color separations) more or less simultaneously. But what does your client do or expect to do with the digital data? Are the data tapes discarded right after the DI is produced or do they have a process or strategy for the digital archives?

Phil Feiner: That question should be addressed directly to the studios themselves. Mr. Roper or any of the other asset managers from the major studios who are here will be able to speak to that issue since these data tapes are generally the property of the producing studio. It’s a daunting question.

Unidentified Speaker 3: The rule is to save everything, so we do retain the digital data tapes. Most of us regard the DTF2 or LTO drives as temporary but archivally necessary, and we expect to migrate the data contained on these media. Ten years ago we migrated Exobytes to a new generation of storage media and we’ll continue to do this as long as there are improved formats to migrate to.

Grover Crisp: As one of those studio asset managers you just mentioned, I’d like to indicate that one of the reasons for doing a test like the one Phil Feiner just showed us is to verify the integrity of the data on the tapes. At the moment, there is no standard for the DI process, either in terms of the production throughput or the conservation of the confirmed final data. Everyone is taking the DI data on various data format tapes and no one is actually testing the data, or even proving that the data is all there. This kind of test is one way to validate that all the data has been captured, to demonstrate that conserving on data tapes works and the data is retrievable, at least in the short term.

Michael Pogorzelski (Academy Film Archive): I have a question about the film stocks that you’re using for the process. Are these off-the-shelf or does the process require special order stocks?

Phil Feiner: We’re using a special order stock from Eastman Kodak.

Leon Silverman (Laser Pacific and Kodak): You receive some data files that have been used to produce a DI. How do you insure that the color that you record into your YCMs (separation masters) is going to match the original intent of the filmmakers who created the data files?

Phil Feiner: As you know, most DI’s are recorded out by an Arrilaser film recorder, and the data that is recorded out has been processed in different ways. What we want to do is match the final film product exactly without compromising tone scale, or increasing granularity. So what we’re doing is matching to that client-supplied original release print. We’re doing nothing more and nothing less.

Leon Silverman: But do you have to go through any additional color process to achieve that match?

Phil Feiner: No.

Leon Silverman: What if the original record out had been created using a modified look-up table, as opposed to having all of the color information in the original data source that records out through the standard look-up table?

Phil Feiner: Then we would start with the target image as represented in the original release print and reverse engineer the LUT data and then “bake” the correct look of the film into the YCMs.

Unidentified Speaker 4 [Feiner]: One of the things that the film industry hasn’t been doing very well is colorimetric calibration. It’s difficult to keep colors the same as the film is created through the chain of processes from production to post-production to final print and archival elements. In the digital cinema world, standards are evolving and we are seeing proposals to use systems such as CIE XYZ colorimetric tools that allows you to calibrate color throughout the chain.

It seems that the industry has already taken some steps toward the implementation of this approach to color control. If we develop this approach as an industry-wide standard, our film scanners need to be calibrated to the XYZ scale, and film recorders will eventually also be calibrated to this scale. In fact, it would be a refinement of this process to do colorimetric calibration of the original camera negative so that the values of the camera original can be tracked correctly through the system. That isn’t really what we’re doing today. Currently, we’re tracking to the DI as it was made for presentation to the audience.

Unidentified Speaker 5: And you’re assuming that the DI was printed at some nominal area LUT.

Feiner: Yes, and whatever lab did that particular DI used a nominal area.

Unidentified Speaker 5: But if it wasn’t printed in a nominal area look up table (LUT) you’d have to go back in and do some color recalibration.

Feiner: Yes, that’s likely. But our separations method closes the loop with the DI process so that whatever the DI rendered, we’re exactly recreating as separate colors on 5242 film stock.

Unidentified Speaker 5: And the problem is that there is no such thing as a DI process that is calibrated in the industry.

Feiner: That’s correct. Each lab has its own way of doing things.

Milt Shefter (Miljoy Enterprises): I have a question following up on Leon’s. We’ve been told that each one of the DI labs has their own proprietary software that Leon describes as “secret sauce.” Obviously, you’ve found a way of reading the data, but are you losing any qualities that may be imparted by these proprietary methods?

Phil Feiner: No, each DI facility has different processes using different hardware and software. But the Arrilaser reads Cineon files. So we take the Cineon files and map them to black and white.

Milt Shefter: Without loss of information?

Phil Feiner: Yes, without loss of information. And I think I’m out of time. So I thank you very much, and if there are any further questions feel free to contact me. Thank you.

_________________________________

SPEAKER BIO

Phil Feiner is the President & CEO of Pacific Title & Art Studio. He has been with the Company for the past 28 years. Besides creating its Digital Imaging division in 1989, Pacific Title remains one of the most successful Post Production facilities in the Motion Picture Industry taking note of its 85 years in the business. Phil has been involved in some of the most prominent restoration projects from the theatrical re-release of the Wizard of OZ in 1998, through the Star Wars trilogy re-release, The Exorcist, and Touch of Evil, to name a few, not including all current feature releases.

Phil is a member of the Academy Motion Picture Arts & Sciences’ Scientific & Technical Committee, a member of The Technology Branch of the Academy, an Associate member of the ASC and involved with the Technology Steering Committee of that organization, affiliate member of the ACE (American Cinema Editor's), SMPTE, and a member of Local 600 as a Director of Photography.