Thoughts and Tips...

[Jannard]

The new Mysterium-X sensor.

This showcases just how good the new sensor is "shooting the hard stuff". Thanks to Michael Cioni of LightIRON for the grading and posting.

Jim

http://www.lightirondigital.com/Ligh..._4K_STLLS.html

[Jannard]

FLUT. What?

FLUT (or Floating Point LUT) is a science developed by Graeme that efficiently allows you to balance your mid-greys in the center of the histogram without worry that you're pushing highlight detail over the edge.

FLUT works to smoothly bend in any highlights that might get clipped when going to a higher ISO, working like film's smooth highlight roll-off.

As a user, you really don't need to know anything about how FLUT works... just that it should be your primary brightness control.

Jim

[Jannard]

Words from Graeme...

This will get updated over time.

Regarding clipping and FLUT Color Science...

"Clipping can be a tricky subject - there's strict mathematical clipping and then there's practical clipping.

FLUT doesn't allow anything that is not actually clipped (ie 12bit value = 4095) to actually clip. They will get crushed up towards 4095, might actually reach 4094, might in extreme circumstances get quantized up to 4095, but that's it.

In practical terms, if you crush the highlights way way up, any detail left in them, assuming they are un-clipped and have detail to begin with, will get diminished, just like what happens in film's soft clip knee. The histogram and right clip meter in the histogram are "practically" oriented. There are 9437184 pixels in your 4k image, and if just a few are clipping, then the bar on the histogram that represents that will be very very short, and probably not visible, even though it does exist.

With visual display of data over such a range of values, some manner of scaling should be used to ensure you can see a reasonable representation of that data, which means unless clipping is "significant", and the white bar at the right is to help you see that, you may have small amounts of clipped image that the histogram doesn't reflect. This is perfectly correct and normal as if it works the other way, you end up vastly under-exposing the image to save those odd pixels and ruin the over-all effect.

Hope that explains the histogram monitoring methodology for you."

Graeme
__________________

[Jannard]

Mysterium-X News.

The 1st project shot on the new Mysterium-X sensor is "The Social Network", directed by David Fincher... cinematographer- Jeff Cronenweth. The movie is being shot with two RED ONE cameras modified with the new sensor upgrade. David and Jeff were kind enough to shoot "The Match" (DiCaprio) for us a week before RED Day One.

The next project to be shot with the M-X sensor (RED ONE upgrade) is "Knockout", directed by Steven Soderbergh. Shooting begins in two weeks.

Interestingly enough... Fincher is using Soderbergh's original cameras that were used to shoot "Ché 1&2", "The Informant!" and "Girlfriend Experience"... so we had to prepare new RED ONE M-X cameras for "Knockout".

The 1st round of customer M-X upgrades have begun.

Here is a reminder of what the images can look like from this sensor. This is ISO 2000 at T 1.3. No noise reduction whatsoever.

Jim

[Jannard]

White Balance.

So just how important is WB? Check out the two grabs below.

The top one's WB is from metadata (about 4200K, -6 tint) left over from a previous in camera Auto WB (I forgot to do a new one in different light). The bottom is the same clip but with a picked WB in REDCINE-X (about 5100K, -2 tint). Seemingly not a ton of difference... but look at the histogram of each. The 1st is clipped in the blue channel, the 2nd not even close. If I had exported the 1st clip to an RGB space for grading, I would never have been able to bring the clipping back in.

(Note the RED logo on Force's car)

Jim

[Jannard]

MYTHS.

"RED doesn't do good skin tones".
Last weekend, Tony Pratt of Park Road Post showed footage from The Lovely Bones with film intercut with RED footage. All the scenes were "skin tone scenes". He challenged the audience to tell which was which. Between the presentations, he confided that whenever he actually gets an answer to that question, most get it wrong.

So if Park Road Post can get great skin tones from Build 15 and Build 16 footage, why can't others? Here are some possible answers:

1. You committed to RGB space for grading without a proper white balance first. If you are off a significant amount in WB, it will make it very difficult to get back where you want. This has been discussed in Post #2 of this thread.
2. Early RED color science was not as exact as it is now, which meant that some work was necessary (as opposed to none) to get great skin tones. Today's new FLUT Color Science (see Post #11 above) makes it extremely easy to get proper color, including skin tones.

"RED footage is soft".
Sorry, nothing could be further from the truth. Peter Jackson has compared RED 4.5K-RAW footage to 65mm film as recently as one month ago. If you have actually seen projected RED 4K footage, you realize how incredible this statement is. So why do some people say this?

1. They actually didn't see RED footage but heard someone else say this.
2. They shot 4K footage and did a half-res debayer to 1080P (half-res throws away 78% of the recorded information) and did not "un-sharp mask" the RAW footage because _________ (insert another myth here).
3. They shot RED 2K footage (1/4 of the resolution) and expected it to be the same resolution as 4K.
4. They watched a bad film print.

RED 4K-RAW footage has 256% more measured resolution than a 1080P camera (assuming it actually measured the captured resolution). RED 4K-RAW footage has more measured resolution than a 4K scan of S35 film. Resolution is not something that the RED ONE is deficient at. Shoot a zone plate and see for yourself.

Jim

[Jannard]

REDcolor vs. Camera RGB.

Camera RGB is the native color coming off the sensor. REDcolor is corrected Camera RGB matched to industry standards. The punchline is... use REDcolor.

Jim

[Jannard]

REDundead.

REDundead is a command line tool that recovers R3D files from drives and CF cards that have been formatted to the file system that has been corrupted. It does a great job of recovering the R3D files unless the drive was erased and then recorded onto again... but it might recover the remaining files.

Filesystem corruption usually happens when a drive is unplugged from the computer when the filesystem is being written to.

REDundead has saved many a project. Thank Deanan.

Jim

[Jannard]

RED Rocket...

Once you use it, you will wonder how you ever lived without it.

Jim

[Jannard]

TIPS...


CAMERA FIRMWARE
Perform frequent firmware upgrades. We are improving builds on a regular basis. Take advantage of that. If you experience anything outside of normal camera operation, re-load firmware and email log files to RED for evaluation. Do a black shading calibration after EVERY firmware upgrade.


HEAT
If you're not recording audio, then set the fans to "Hot" to avoid ever even getting close to over heating. Don't tape off/block the ports on the side of the camera since heat escapes out of those ports. If fans get noisy, replace them. Professional cameras need maintenance.

Jim