updated 8 Apr 2007
DI - Scanning
Digital Intermediates - Scanning the film before grading
go to ...
Before you can grade the film as a Digital Intermediate
you need to scan the original 35 mm or 16 mm negative
and create a "digital negative". This is scanning.
The Digital negative
Assuming that you've read the page on "DI - Preparation" you know that after the film is edited, information on recreating this edit in the form of EDLs, cut lists, and tape transfers is transmitted from the editing facility to the DI facility.
The DI facility takes all this information and needs to create a "digital negative" to grade and create a "look" for the film. This digital negative consists of files on a hard disk. One file for each frame. Each and every frame in the film, which for a 3 hr film can be as many as 2,00,000 frames.
Each scanned frame is a file in a format called .dpx or DPX file. DPX stands for Digital Picture Exchange. This format stores data in a manner easy for a grading system to read very rapidly. Remember, the DI system needs to open and close 24 files each second.
The DPX file format stores not just the picture, but the time code and the Keycode of that frame and other parameters. This is the basis of the Digital Intermediate process and this property of DPX is what makes the DI process even possible.
You do the math.
There are 24 frames i a second. And 60 seconds in a minute, 60 minutes in an hour.
OK enough on DPX. How does the film which is in the form of a negative or OCN (original camera negative) get into the form of a DPX sequence, or simply, a 'digital negative'?
To convert the actual movie film to a 'digital negative' series of files, one needs to scan the film negative. This is done in a "scanner" or a 'datacine'. So, the second step in the DI process is scanning.
The scanner is a rather large device which can run film in up to 2000 feet rolls, light it up with a precisely calibrated light and convert each frame into a digital picture. It does this with as much as 16-bit sampling depth. Simply put, it can convert all the individual pixels in each frame into digital pixels with a range of 2 raised to the power of 16 or 65,536 levels of brightness. Think of this as 16 stops.
And it does this for all three primary colours red, blue, and green so it can resolve many billions of colours. In effect you end up with each film negative frame as a digital picture which is faithful to the original.
How faithful? Is the digital neg the same as the original neg? Are there losses? All this is still hotly debated and will very likely be till the end of time.
If you really want to 'do DI' on your film, you better start by believing that it is, and there are no losses really. And that this process will actually enhance the look of your film and be worth the money you're spending.
And to help this belief, I strongly recommend taking out a piece of your negative, going to the DI place you are considering, making them pass this through their circuit, and taking out a neg and a print and watching it in a theatre. Most likely you'll be amazed at how good it is.
2k or 4k
Once people start 'believing in DI' the next thing that's hotly discussed in '2k or 4k'? Before getting into a final answer, let's see what these numbers mean.
The scanner that scans the negative, 'images' it into a digital picture or a DPX file as earlier said. This imaging happens at a resolution commonly called 2k. Or at a higher resolution called 4k. So 2k and 4k are resolutions.
2k means each 35mm frame is 2048 pixels wide by 1556 pixels tall. Which puts it at about 3 million pixels total.
So the scanner takes your negative, scans it and spits out digital pictures in a form called DPX. One DPX file per frame.
These DPX files which are about 12 MB per frame (at 2k) are then stored on a hard disk. And each such frame contains a timecode, and a keycode. And they're all organised into folders and neatly classified by camera roll or lab roll so further work on them is not too difficult. If scanned at 4k the files are 50 Mb per frame.
Scanners are typically rather slow. Some work at 2 sec per frame, some do 2 frames per sec, and datacines work at 24 frames/sec. This speed of scanning doesn't affect your sound sync in any way, since scanners scan each and every frame and put out exactly as many frames as there are in the film. You just need to run them at the proper speed to sync with your sound. Think about this carefully and it makes sense.
But back to 2k vs 4k. Exactly how many 'k' is needed to faithfully represent the original negative? Depending on who you ask, 2k is 'good enough'.
So far, no Hindi or other Indian film has been 'made in 4k'. Even in Hollywood, Spiderman was one of the few films made in 4k. So 4k, while being better than 2k for sure, is too challenging for systems to do reasonably. And with file sizes being larger, systems will work slower at 4k, almost a quarter as slow. So if a 2k DI would take a month, leave aside 3-4 months for a 4k DI.
Besides, if one were not doing a DI for one's film, then one would conform the negative,a nd then optically print it. And if has many prints, one makes dupes of the original neg. It is widely believed that the final print in such a case, has a resolution not exceeding '1k'
So bottom line... 2k is enough for the moment. If you really don't believe this, then avoid doing DI. Really. 4k DI will make you (and the DI house) seriously unhappy with the time it takes.
3 million pixels is also 3 megapixels. And most mobiles now have 2 Megapixel cameras, and many of us have digital cameras that are 4, 5 or even 6 or more megapixels.
And we've all seen a 3 Megapixel camera image on our computer screen. So suddenly 2k doesn't seem like a huge resolution after all.
Lets see other examples.
A normal TV set shows 720x576 pixels. or 0.4 megapixels. Most computer monitors 15-17 inch show 1024x768 pixels or 0.8 megapixels.
Larger monitors show 1600x1200 pixels which is 2 megapixels. And the Apple Cinema display 23in shows 1920x1200 or 2.3 megapixels.
Log and Lin
Another 'angle' to this scanning is log and lin. Short for logarithmic and linear. Almost all scans these days are log scans. Although the scanner itself works and images the nagative in a linear way, it converts and stores it as 'log'.
The scanning process takes place with a bit depth of 16 bits per pixel. To give a sense of what this means, most video work is done at 10-bit (smoke, flame, even FCP work at 10-bit) DigiBeta is 10-bit. Most displays however are 8-bit. DVD and Beta can be considered to be 8-bit. But film has much more latitude so scanning happens at 16-bit.
Back to 'log' scans. This log-lin thing is one of these eternal mysteries even for those involved in film DI. I plan on writing a simple explanation on this, and am trying to make is simple. Till then, All I can offer is an analogy. Not precise but a parallel.
Think of film negative. When we shoot anything with a film camera, the sight before us is... well... 'positive'. All of reality around us is 'correctly coloured' and positive. But when shot on film, it becomes negative. Colours are reversed. But we still consider something correctly shot and recorded if its exposed on film. Even if its negative. Because that negative when again exposed to another film, becomes positive and even 'real' once again. We know that.
Think of log to lin and lin to log as a positive to negative and negative to positive process. Not exactly, of course, but a parallel. (Don't they call this a 'metaphor' or something like that?)
Anyway, the real reason for log is that film's response to light and exposure is not really 'linear' but 'logarithmic'. As you increase exposure, the exposed image on film gets denser or 'blacker' (its negative, remember) But this increase in density is not exactly proportional to increase in exposure. As exposure increases, density increases, but this increase gets slower as exposure increases.
Another analogy for this 'not proportional' behaviour of film. If you drive a car you will observe that as you push the accelerator the car goes faster. More push, more speed. But in the beginning of the accelerator's movement the increase is more and as you push it harder it seems to 'level out' as you reach the floor. So does your car's speed. So its not exactly proportional. Its not linear. Not proportional. It is logarithmic.
Back to film. Film's response to exposure is like that. Its not linear. Logarithmic. So when we scan film in a film scanner, we create a non-linear or logartihmic image. It looks cloudy or non-contrasty. That same image when (mathematically) linearized, looks right. Just like when negative is exposed to another negative we get a picture that's right.
Another reason for working log is that log images are 10bit meaning they take up less disk space. A linear image of the same latitude would take up 14-16 bits. Meaning more hard disk space, more data to transfer.
So log is just a mathematical representation of 'reality'. We have to make it linear for humans to enjoy and appreciate it. But the computer needs it to be log. Log is not right and linear is not wrong. Log is right and even linear is right. But both are needed for specific reasons.
Scanners for DI use a variety of technologies but do the same basic job - to take an original camera negative, refer to an EDL or a cut list, scan only the frames needed, and store them neatly on a hard disk.
To do this, all scanners use the same basic principle. They shine a light through the film. And cast an image on an image sensor like a CCD or CMOS chip. Some use a CRT imager. These imagers 'digitize' the film image and converts it into data.
Then there is a computer system that takes this data and is saves it to hard disks as dpx or cineon or even tif files.
What separates one scanner from another is resolution of scans. type of lighting, type of imager, and speed of scanning.
|Spirit 4k datacine|
go to ...