Cinematography Mailing List - CML
    advanced

2k 4-4-4 Spec Check

 

I would like to check the information a post house gave me recently about scanning a S16mm negative at 2k for a DI to 35mm print, doesn't quite sound right to me.

- There is no such thing as a 4:4:4 monitor, the best currently available is 4:2:2.

I simply don't buy this. I thought Spectre and IQ were both fully 2k RGB 4:4:4 systems from start to finish.

- The human eye cannot distinguish between 4:4:4 and 4:2:2. 4:2:2 is only necessary for effects work and not primary and secondary colour grading.

Again, this just doesn't sound right at all.

- The minimum physical size of a 4:4:4 monitor would have to be at least 37".

I didn't think monitor size had anything to do with resolution or colour space.

- Grading (allegedly) 2k 4:4:4 data on a HD 4:2:2 monitor presents no problems.

But if I'm only working with half the colour information then there must be a significant problem. What is happening to the rest of the data that I cannot see?

- LUT's are only necessary for print stock and not for negative.

This may be true in that the negative is scanned at 10bit log and so in a sense is negative independent as all the information is digitized regardless of stock variation. Correct?

The flat data looked about right but when their 2383 LUT was applied the ungraded image seemed extremely contrasty, akin to a skip bleach print on 2393. I was told that what you see is exactly what will be printed and the LUT’s were correct.

The director has expressed a clear preference for 2393 but the post house only had the 2383 LUT, it was then suggested that what you see will never be the same because of the difference between log and linear data(?), variations in projection bulbs and from theatre to theatre. So that we would have to account for the increased contrast and saturation gained from printing on 2393 while (supposedly) grading for 2383 by eye as we graded. I likened this to being asked to tolerate the Avid EDL with a +/-24 frame error throughout the film, which would of course be unacceptable. I was told that "on AIM" prints would eliminate any abberations.

I'm a bit stumped. I thought I'd done my homework but I wonder if my facts are wrong, in which case I'll happily apologise for my questions and proceed. But this does sound a bit odd, doesn't it? My trust in this particular version of DI has gone out the window in an instant I'm sad to say, when my previous experience at 2k on a Spectre last year was nothing but impressive. I don't feel looked after at all. I'm now into post-DP mode, I'm simply trying to protect and grade the images that everyone worked hard to create and get the best out of our DI route.

Any feedback would be grafefully received.

Kind regards

Shane Daly
London DP


Shane,

I don't know who told you this, but you need to get away from these people as quickly as possible.

Some answers below...

I would like to check the information a post house gave me recently about scanning a S16mm negative at 2k for a DI to 35mm print, doesn't quite sound right to me;

>There is no such thing as a 4:4:4 monitor, the best currently available is >4:2:2.

> I simply don't buy this. I thought Spectre and IQ were both fully 2k RGB >4:4:4 systems from start to finish.

[Lucas] Are we talking about monitors or post systems?

------------------

- The human eye cannot distinguish between 4:4:4 and 4:2:2. 4:2:2 is only necessary for effects work and not primary and secondary colour grading.

>Again, this just doesn't sound right at all.

[Lucas] Yeah, that's total and complete crap. Take a look at a good key shot on HDCAM and HDCAM SR, and you can pretty clearly see the difference.

> - The minimum physical size of a 4:4:4 monitor would have to be at >least 37".

>I didn't think monitor size had anything to do with resolution or colour >space.

[Lucas] Huh? I'm trying to figure out what they think they're saying here, and I can't make sense of it. But whatever they think they're saying, it's incorrect.

- Grading (allegedly) 2k 4:4:4 data on a HD 4:2:2 monitor presents no problems.

But if I'm only working with half the colour information then there must be a significant problem. What is happening to the rest of the data that I cannot see?

[Lucas] (exhaling loudly) It presents all kinds of problems. But a good engineer, and a good colorist working on a well calibrated system can make a lot of those problems go away. But saying something so vague and general is just plain silly.

>- LUT's are only necessary for print stock and not for negative.

[Lucas] Uhhh... what kind of LUT, and for what purpose? Again, a gross generalization, and absolutely wrong. Get away from these people.

---------------

Shane, you're wise to question these statements. A DI post path is something that can go in a dozen different directions depending on what post systems are used, who the talent is, and what the distribution method will be.

Get rid of this post facility, and find someone who will answer your questions rationally, and not give you the "we know what's best. Don't question us" routine.

Lucas Wilson
Assimilate, Inc.
Los Angeles


Shane Daly wrote :

>- There is no such thing as a 4:4:4 monitor, the best currently available >is 4:2:2.

This is somewhat true in that monitors are basically 8 bit. That's one of the reasons you must use a lookup table when displaying other formats.

>I simply don't buy this. I thought Spectre and IQ were both fully 2k RGB >4:4:4 systems from start to finish.

They are, but that has nothing to do with monitoring. The deeper color depth is significant in signal processing, particularly keying and color correction, because there is more information to work with. In keying this allows fewer edge problems, in color correction it prevents banding and provides a wider color pallette for the film out, even when displayed on an 8 bit monitor.

>- The human eye cannot distinguish between 4:4:4 and 4:2:2. 4:2:2 is >only necessary for effects work and not primary and secondary colour >grading.

While I agree with the human eye comment, I think you misunderstood what might have been said. I think you have it backwards: 4:4:4 is useful for effects work and color grading, compared to 4:2:2.

>- Grading (allegedly) 2k 4:4:4 data on a HD 4:2:2 monitor presents no >problems.

That's true, if a proper lookup table is used. That's what the lookup table is there for.

>But if I'm only working with half the colour information then there must >be a significant problem. What is happening to the rest of the data that >I cannot see?

See above.

>- LUT's are only necessary for print stock and not for negative.

I don't quite understand the comment. Lookup tables are usually used for display purposes in order to emulate the print stock that will be used for the final film release. It has nothing to do with the origination medium.

>The flat data looked about right but when their 2383 LUT was applied >the ungraded image seemed extremely contrasty, akin to a skip bleach >print on 2393.

If the "flat data" looked correct to you, the scan was most likely not done correctly. Log format data should look washed out in almost all cases, it is the nature of log encoding. When the lookup table is applied (i.e., the material is "de-logged), proper display values are obtained.

>My trust in this particular version of DI has gone out the window in an >instant I'm sad to say, when my previous experience at 2k on a Spectre >last year was nothing but impressive.

It sounds to me like the facility you were in is, quite frankly, being a bit too honest with you and telling you far more than you need to know, which in turn is making you nervous because in your previous experience, they simply did their job and kept you blissfully unaware of the problems. DI is not an exact science, and practically everything you've mentioned here is indeed true, or at least partially true.

The curse of those in a technical business is that they constantly ride a fine line between what they should tell the clients, who really don't need to know about technical issues and their attendant quirks, and what they should simply handle internally and let the client judge the final result. It is, to some degree, a decision between total disclosure or salesmanship.

I would suggest judging by the reputation and results rather than whether the technical explanations jive with your understanding.

Mike Most
VFX Supervisor
IATSE Local 600
Los Angeles


Lucas Wilson wrote :

>[Lucas] Yeah, that's total and complete crap. Take a look at a good key >shot on HDCAM and HDCAM SR, and you can pretty clearly see the >difference.

Whoa, that's not just comparing 4:2:2 and 4:4:4 -- remember HDCam is filtered down to 1440 and sampled at 3:1:1, and more compressed than HDCam SR. So you are comparing apples and pomegranites.

Jeff "well, it's more complex than an orange" Kreines


Lucas Wilson wrote :

>Llucas] Yeah, that's total and complete crap. Take a look at a good key >shot on HDCAM and HDCAM SR, and you can pretty clearly see the >difference.

Keying is not something "seen" by the human eye. I happen to agree with the original statement: the human eye generally cannot differentiate between the two in terms of what they look like, especially on an 8 bit monitor. What something looks like and how it responds to processing are often two completely different things. If you need more evidence of this, I offer the DV format - which, when shot through good glass with good lighting, can look great, but just try to color correct it or extract mattes from it. That's when it falls apart.

>But saying something so vague and general is just plain silly.

Why? Because it's honest? What was said and what the poster is inferring from it are not necessarily the same thing. What was said is generally true. What was inferred was based on expectations and assumptions which were not necessarily accurate.

>- LUT's are only necessary for print stock and not for negative.

Once again, a technically savvy person could read what was said and understand it. I'm not sticking up for the facility (I don't even know who it is), but the fact is that lookup tables are not generally used in transfers from negative to tape for tape release. They come into play when film or electronic projection is the intended final product. So the statement, while confusing, is not necessarily silly or wrong.

>Get rid of this post facility, and find someone who will answer your >questions rationally, and not give you a "we know what's best. Don't >question us." routine.

It sounds to me as if that's the approach the first facility (the one where he was more confident) took - we'll only tell you as much as you need to know. This one was a bit more forthcoming and a bit more honest about the flaws in the process, even if their answers turned out to be a bit nerve wracking.

Mike Most
VFX Supervisor
IATSE Local 600
Los Angeles


Shane Daly wrote :

>- There is no such thing as a 4:4:4 monitor, the best currently available >is 4:2:2.

Most HD monitors are only 4:2:2. If they support a dual-link interface they are likely to support 4:4:4. Most computer monitors are 4:4:4.

>The human eye cannot distinguish between 4:4:4 and 4:2:2. 4:2:2 is >only necessary for effects work and not primary and secondary colour >grading.

With the caveat, at a great enough distance (usually about 4 screen heights) it is hard to see a difference. All the colors are there in 4:2:2 they are just spread out. Note that there is still a difference between 8bit systems and 10-bit systems which sometimes both get referred to as 4:2:2.

4:4:4 is good for effects work and is a benefit in color grading as well.

>The minimum physical size of a 4:4:4 monitor would have to be at least >37".

At a certain distance from the screen, how big would the monitor have to be so you could see the difference between 4:4:4 and 4:2:2. This is their answer. Has nothing to do with the making of the monitor.

>Grading (allegedly) 2k 4:4:4 data on a HD 4:2:2 monitor presents no >problems.

No problems is not the right answer here. But 4:2:2 isn't the source of the problems. 4:2:2 just spreads the colors out but can represent the same number of colors.

The real problem is that HD has a certain color gamut which is less (usually) than that of film. Some colors will not respond the same way on the monitor as they will on the film. This is a mental adjustment you have to keep in mind at all times. Typical suspect colors are saturated yellows, cyans, purples. Low value colors in the green and cyan are are particularly subject to color shifts. Bright blues and reds on the monitor usually come out more subdued on the film.

A well-designed 3D cube lookup table can help reduce some of these problems, but won't eliminate them completely. There are several hardware solutions being sold that provide these 3D color cube corrections. Find out if they are using one.

>LUT's are only necessary for print stock and not for negative.

LUTS are most necessary to show the final projected film look of the image when displaying it on a gamma calibrated monitor. The LUT has to build in a standard assumption about the negative as well as the print stock. Cineon files for example always assume that the scanned log data represents values from a film negative with a 0.6 gamma.

>This may be true in that the negative is scanned at 10bit log and so in a >sense is negative independent as all the information is digitized >regardless of stock variation. Correct?

True

>The flat data looked about right but when their 2383 LUT was applied >the ungraded image seemed extremely contrasty, akin to a skip bleach >print on 2393. I was told that what you see is exactly what will be printed >and the LUTs were correct.

Hard to know the answer to this without knowing how there system and monitor is setup. Could be true. Easiest way to find out is to get a test scene shot to film and see what it looks like.

Note that it is possible that the data is being "double-corrected". If for example, there was some kind of contrast setup adjustment on the Spirit when the data was scanned, it could be building one contrast correction on top of a contrast correction in the LUT. Also, the monitor could perhaps be incorrectly setup, but if they are right and it looks that way on film, be sure to see the monitor image in a dark surround environment, as there is a necessary contrast increase needed for display of images in a dark surround (that is built into print stocks BTW).

>The director has expressed a clear preference for 2393 but the post >house only had the 2383 LUT,

It isn't that hard to make a LUT for 2393. Ask for one. It wouldn't be a good idea to use an '83 lut for premiere stock. It might mostly look OK, but the higher contrast on the 93 might pull out some details in the blacks (or cause black levels to float scene to scene) in a way which you couldn't preview on the monitor and will thus be surprised by.

I have to separate the below to have them make sense to me...

>it was then suggested that what you see will never be the same >because of the difference between log and linear data(?), variations in >projection bulbs and from theatre to theatre.

They left out how different release prints are and how much you lose going from the answer print to the release print. I don't get what they meant by the log/linear comment either, but could think of a few things it might have been. They would have to supply more information about how there system is setup so that you could understand that comment better.

All anyone can ever do about this is to get it as right as you can in the film-out from the place that you are doing the DI negative. (Don't rely solely on any monitor for what is right or not).

>So that we would have to account for the increased contrast and >saturation gained from printing on 2393 while (supposedly) grading for >2383 by eye as we graded.... I was told that "on AIM" prints would >eliminate any abberations.

AIM prints aren't going to eliminate the contrast and saturation difference. Black levels can be a problem because you can't see them on the monitor. If you are looking at '93 prints from your filmouts, you should be able to guess the differences on the monitor needed to get the look you want.

>I'm a bit stumped. I thought I'd done my homework but I wonder if my >facts are wrong, in which case I'll happily apologise for my questions >and proceed. But this does sound a bit odd, doesn't it?

Many post-houses that use a 4:2:2 work flow bring up a lot of issues to prove why it is acceptable to use an HDTV based system for film grading. Many pictures have been done on these types of systems. The issues that you raise above are the kinds of statements that are brought forward to justify using HD for grading. If your film-outs look OK, then you needn't worry too much.

It is not in my opinion however the best way to get the most of a film negative for a DI, but then I have a company bias there.

Jim Houston
Head of Digital Mastering
Pacific Title & Art Studio


I've seen grading sessions for 2K done on HD screwed up from the start because of LUT mismanagement, horrible color transforms, etc. With your experience, you know that there are a lot of ways to screw it up, and a statement like "Grading (allegedly) 2k 4:4:4 data on a HD 4:2:2 monitor presents no problems."

Is not at all correct. Of course there are problems with it. They can be dealt with by a good crew, but there are lots of steps in the path that can make things go wrong.

My favorite is a job I worked on where 2K material was downrezzed onto HDCAM full-frame, with black on the sides. The color session was done tape-to-tape HDCAM-HDCAM, then blown up for a filmout. Now tell me that doesn't present problems...

The answers that Jim gave were much more to the point, more clearly understood, and -- if I were the client -- would give me much more faith that that facility knew what they were doing and how.

Lucas Wilson
Assimilate. Inc.
Los Angeles
www.assimilateinc.com


Shane Daly wrote :

>- There is no such thing as a 4:4:4 monitor, the best currently available >is 4:2:2. I simply don't buy this. I thought Spectre and IQ were both fully >2k RGB 4:4:4 systems from start to finish.

CRT monitors can certainly be 4:4:4 given standard definition signals, but as the pictures approach HD, 2K and higher this becomes a true statement. The Sony evaluation grade broadcast monitors in 24" and 36" CRT sizes don't have a fine enough dot pitch to display more than about 1000 lines across the useable face of the tube. That's acknowledged in Sony's "1000 TV lines" specification for these monitors.

You might say that these are 2:2:2 monitors.

>- The human eye cannot distinguish between 4:4:4 and 4:2:2. 4:2:2 is >only necessary for effects work and not primary and secondary colour >grading.

Whether it sounds right, it is actually true. At normal viewing distances, the human eye detects less detail in the colors that make up the half resolution samples. It's no accident that the standards take advantage of this fact.

>- The minimum physical size of a 4:4:4 monitor would have to be at >least 37". I didn't think monitor size had anything to do with resolution or >colour space.

Size matters. The CRT size has quite a lot to do with the full bandwidth RGB resolution you can see on it. The ability to manufacture the aperture grille with those little dots very close together _and_ enough of them across the width of the screen sets the upper limit on the resolution that a CRT can display. Based on the numbers I see in the Sony CRTs, I'm not sure the glass wouldn't have to be 50" or larger to display "4:4:4" resolution of HD or above.

All of the above comments are based on evaluation grade CRTs. I'll let Martin tell you about what is possible with LCD displays.

David Tosh
Telecine Engineer,
Southern California


Thanks very much to all who replied on & off list, very informative and helpful. The test print looked awful and there is currently some discussion as to how to proceed. I'd rather not say anymore until a decision is taken and then give some detailed feedback.

Thanks again

Shane Daly
UK DOP



Copyright © CML. All rights reserved.