Cinematography Mailing List - CML
    advanced

HD Scanned Film In Theatres


I have a film coming up for which I would like to have telecine like control of colors, and contrast. Surprise surprise, it's a scary movie. 2k scans may be too much for our budget. I had heard that an HD transfer then put into a non linear system might be a cheaper option. It had been explained that if you scan the whole frame at 2k, the actual 1.85 area of the scan is similar in resolution to the 16x9 of an HD transfer. I have asked this to a few labs, but the rep always starts in telling me that color space is more compromised and that there is more artifacting, of course, that's why it's cheaper. I just want to see an example projected to quantify the difference in my own head, see if it's an option.

Thanks.

Kirk Douglas.


Talk to Technicolor Services in NYC. I believe they've been experimenting with HDCam transfers as opposed to 2k data to see the difference.

Mitch Gross
NYC DP


>Talk to Technicolor Services in NYC. I believe they've been experimenting with HD-Cam transfers as opposed to 2k data to see the difference.

I have seen test results of the same scene done 2K Cineon DI and also HD data (to disk at 1080 24P 4:2:2) DI. Resolution is definitely not the big issue. Colour is. Descriptions of images is hard but the 2K had a richness in the colour that the HD did not. Whilst this colour difference is not huge it is noticeable. I will take a punt and say the HD was 90% of the quality of the 2K. This difference depends on your material and if your project was to be de-saturated the difference may be minimized but you never get nothing for free.

HD comparisons can also be difficult as there are many variations with HD. Uncompressed 1080 24P 4:4:4 data would be superior to heavily compressed and sub-sampled 8 bit HDCAM*. HD should be considered a poor mans process to be only used when its the only alternative.

*Of course HD-cam SR may change this balance again.

Many Thanks
Tom Gleeson
www.cinematography.net


Tom Gleeson wrote :

>have seen test results of the same scene done 2K Cineon DI and also HD data (to disk at 1080 24P 4:2:2) DI. Resolution is definitely not the big issue. Colour is. Descriptions of images is hard but the 2K had a richness in the colour that the HD did not.

I have a short test of super 16 to 35 done through data and through d-6 HD. The difference is really apparent even in this. The Data looks like film and the D6 looks more like it has been through video in some way because of the limited highlight range as well as the more muted colors. Going from film through HD 1080 -24 4:2:2 will no doubt yield a big difference qualitatively.

The question then becomes what is the extra quality worth to you? As one rep from a post house put it to me, "if they never saw the original, then they won't miss what they never saw."

Mark Smith


Mark Smith wrote :

>I have a short test of super 16 to 35 done through data and through d-6 HD .

Mark :

Was the HD 10 bit, 10 bit log, or 8 bit? My guess is that 4:4:4 HD, 10 bit log, would yield a pretty nice DI (essentially it's going to be almost the same as 2K).

Jeff "interesting developments await" Kreines


Kirk Douglas wrote :

>It had been explained that if you scan the whole frame at 2k, the actual 1.85 area of the scan is similar in resolution to the 16x9 of an HD transfer.

Yes - the 1:1.85 area of a scan is similar to the 16x9 of an HD transfer - 16x9 being 1:1.78.

However if you scan the whole frame at 2K, chances are it will end up as a 2048x1536 scan if it was shot Super-5 which has an aspect ratio of 1:1.33. Or as 1828x1332 if it was shot as academy - which has an aspect ratio of 1:1.37.

Either way you're going to have a lot more image top and bottom to play with and re-rack for 1.85 delivery than if you just start with 1:1.78.

>I have asked this to a few labs, but the rep always starts in telling me that color space is more compromised

It depends on how much you want to push your grading - you are going to be able to push it further with less artifacts if you've scanned it as RGB data than if you transferred it as "YUV" video and graded it afterwards.

Similarly you will preserve shadow detail better if you scan at 2K log space than if transferred in video "linear" space - and grade it subsequently.

You can of course get perfectly acceptable results from an HD transfer - it all depends on what the material is and what you want to do with it.

>I just want to see an example projected to quantify the difference in my own head, see if it's an option.

It's the only way!

Regards

Martin Parsons
http://www.moving-picture.com


Jeff Kreines wrote :

>Was the HD 10 bit, 10 bit log, or 8 bit? My guess is that 4:4:4 HD, 10 bit log, would yield a pretty nice DI (essentially it's going to be almost the same as 2K).

Notes have been misplaced so I am going from memory here...It was HD recorded to D6 so I am thinking 10 bit but probably not 10 bit log.

Any way there was a noticeable difference, highlight/contrast range/color saturation wise between the data path and the D6 path with the d6 path being the lower quality version.

I agree that the 4:4:4 10 bit log would be a nice pathway as DI but it isn't quite happening yet.

Mark Smith


Jeff Kreines wrote :

>Was the HD 10 bit, 10 bit log, or 8 bit? My guess is that 4:4:4 HD, 10 bit log, would yield a pretty nice DI (essentially it's going to be almost the same as 2K).

To my knowledge, the only device currently available that creates 4:4:4 HD in 10 bit log space is the Viper camera. I could be wrong, but as far as I know, no telecine is currently set up to do this. Not by the manufacturer, anyway.

Mike Most
VFX Supervisor
IATSE Local 600
Los Angeles


Michael Most wrote :

>To my knowledge, the only device currently available that creates 4:4:4 HD in 10 bit log space is the Viper camera. I could be wrong, but as far as I know, no telecine is currently set up to do this. Not by the manufacturer, anyway.

I assumed that the Spirit could do this, since Philips/Grass/whatever they're called this week embraces 4:4:4 10 bit log with the Viper, but obviously I assumed too much.

There will soon be a telecine that does 4:4:4 10 bit log, but I'm sworn to secrecy. (It wlll also do 2K.)

Jeff "hush, hush" Kreines


Jeff Kreines wrote:

>but as far as I know, no telecine is currently set up to do this. Not by the manufacturer, anyway.

>There will soon be a telecine that does 4:4:4 10 bit log


Old news.

CReality (Cintel) was probably the first machine capable of doing 10 bit log scans. We built a room back about SIX YEARS ago at the now defunct Hollywood Digital facility. It could run in 2K scan mode at about six frames per second. Pictures looked great. We did the color correction on a daVinci 2K. The whole thing was a bit less than idea 'cause it was bleeding edge technology. But it worked. In fact, that's when and why I wrote my proprietary DPX viewing and manipulation software. We needed to accomplish certain things that couldn't be done with existing tools.

I think Dave Walker's Cineglyph has been able to do 10 bit log since inception. I first saw the machine four or five years ago.

There's also a Thomson machine that can do 4:4:4 scans and, obviously, output 10 bit log.

Arri's scanner?

Imagica? etc.

Martin Euredjian
eCinema Systems, Inc.


Martin Euredjian wrote:

>It could run in 2K scan mode at about six frames per second.

I think one of us is misinterpreting the original post. My assumption was that what Jeff was talking about was 4:4:4 HD coded in log format - which would be a video format, not frame scans. The Viper does just that in HDStream mode - it produces an image that is in log format as far as the image goes, but is presented in an HD video signal (4:2:2, of course).

I believe it can do the same thing in 4:4:4 using the dual link output for recording on an HDCam SR. This is very different than film scanners, or telecines doing the work of scanners, which in either case would be producing a series of individual frame files, not video.

In any case, my comments were in reference to whether any telecines are currently set up to do log format image output in a 4:4:4 video stream.

I still think the answer is no.

Mike Most
VFX Supervisor
IATSE Local 600
Los Angeles


Mike Most wrote:

>In any case, my comments were in reference to whether any telecines are currently set up to do log format image output in a 4:4:4 video stream. I still think the answer is no.

Understood. I'm fairly certain that either CReality, Cineglyph or Thomson's 444 machine can do it. I think the reason you don't see it has a lot more to do with workflow than anything else : Where do you record it? How do you use it?

"Data" everyone knows how to use. Dual link HDSDI is a different animal, even though all of it is data...the line between video and data is blurred.

HDCamSR is just starting to deploy. As tools become available facilities might inclined to make the telecine output happen in that format. Right now, data is much more usable, by far. Slow. But you can use it.

Martin Euredjian
eCinema Systems, Inc.


Michael Most wrote :

>My assumption was that what Jeff was talking about was 4:4:4 HD coded in log format - which would be a video format, not frame scans.

Yes, Mike, that's what I meant.

Obviously any scanner, recording individual frames as frames, can be jiggered to do whatever bit depth and sampling one needs. Storing that as 4:4:4 HD in log format is a bit different.

Jeff Kreines


To clarify further. Nearly all telecines I know run in the range of 12 to 14 bits linear internally. This is in real time, say, 30fps. Making HD (single or dual link) out of that requires a certain type of processing.

Turning that linear data into 10 bit log is almost academic in terms of the hardware involved and most of these machines would/should/could have it in place. The distinction between running "data" at 6fps or "video" at 24+fps, is, as I've always understood it, not necessarily related to how the data stream can be manipulated.

I placed "video" and "data" in quotes because, in reality, it's all data. If we take "data" as meaning RGB (linear or log) then we can call them distinct.

I insist that the only reason you are not seeing machines installed to product 10 bit log data output at normal video frame rates isn't that they can't do it but rather that there's very little you can do with it. Building a room to run in this mode would be a just about useless due to the lack of support equipment. In sharp contrast to that, a data room knows how to play with the likes of Inferno, Fire, etc. without many issues.

Martin Euredjian
eCinema Systems, Inc.


>In any case, my comments were in reference to whether any telecines are currently set up to do log format image output in a 4:4:4 video stream. I still think the answer is no.

I could not find one either. I have been investigating just this as an improved workflow. Best solution I have found is 10 bit log scan in data to a server and play back video directly off the server. This can be done on the fly, without render times. All devices after the server treat the video as linear, but benefit from the log scan/record. It is preferable to stay 4:4:4 RGB but not strictly essential Thomson Specter, Quantel IQ, DVS Clipster, Discreet Stones with Backdraft, and da Vinci SIO can all achieve this "trick".

The principal advantages are that the log scan captures greater (unchanged) dynamic range than the linear video scan, and the distribution of bits across that range is closer to the characteristics of film and the human eye. The server technologies also add some work flow enhancements such as list management (virtual neg cut or even auto-conform) non linear access and minimal film handling. And this approach is all real time after the initial scan, and more economical in terms of storage and infrastructure than an all data DI ( I believe there are still benefits in 2K and above, but they come at a huge cost).

Some telecines may have a log video output, but I believe they apply a log LUT to 10 bit linear video (ie too late in the signal path) so the damage is done and benefits are not gained.

Empirically I am convinced this workflow produces better images in both film outs and video. I would be very interested in hearing from anyone who has tried or thought about this approach.

Kevin Shaw
Director of Training
The da Vinci Academy
Total Image Solutions


I have used telecines to output HD 444 log 'video' a number of times for lower budget DI projects. Setting the telecine up is an issue, but not impossible. It's actually easier with Cintel's Millennium and DSX and will get a boost soon as Cintel are looking to make this a simple user setup, exactly for the reasons this thread has been talking about (cost).

The dynamic colour range equals that of 2K/4K log data, about 2.0 density based on the .CIN specification, so full colour grading is possible in post as for true DI data.

This is a very interesting approach for obvious reasons.

Steve Shaw
Digital Praxis



Copyright © CML. All rights reserved.