Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

What was tested?

PaulClements

PaulClements

Well-known member
Joined
Dec 28, 2006
Messages
1,995
Reaction score
0
Points
36
Age
47
Website
www.pclements.co.uk
Can anyone post a list of what was tested please? Like many here I'm still a bit in the dark as to what actually occured at LART. I've read the odd bit here and there but a complete roundup of what was tested/shot would be interesting.
Cheers
Paul
 
Yeah, this what many of us want to know.
Every time we see that Post-LART thread, they show us some girls and juicy stuff, we forget why we were there.

We want to know precisely, the item(be it lens,battery,monitor) working or not, if working the ratings(good,bad,ugly etc).

Cheers,
Siva
 
Thanks Paul for speaking up. With the LART post build up, it certainly seemed to be platformed as the most definitive and exhaustive test environment for the RED ONE. After the numerous reschedules, the event finally happened--Only now, the event results are 'CONFIDENTIAL'.

I'm certainly anxious and excited to hear and read about the collective results.
 
Since everyone at LART, (Including myself) is under NDA we cannot make mention yet as to what went on in detail, until authorized to do so.
That said, tests like this take time and with everyone's busy schedule, it makes it that much more difficult to process and evaluate the results. There were a lot of different moving pieces of the puzzle those two days and we are sorting out all of those pieces in our spare time.
Results have been promised, along with a full description of those test results coming very soon to Vuze and here at Reduser.
There were quite a few different lenses, rigs, etc present which will be touched on further by Sharky and Gibby after the results are posted.
We are in the Conform/Grading stage of our post pipeline for the first couple sequences shot on day 1 & 2.
I can say, the shots look great and were a lot of fun to edit!!! We will try to get you all some graded footage from Scratch as a nice Christmas present if time permits.
We have not forgotten about you all, and are working hard to bring results, ASAP!
 
I found it interesting how over at cine . com that one of the LART members posted a portion of his NDA stating that no one from LART could publicly report anything bad. Why this measure? I own a Red, I know the bad, and it aint that bad, so what's up? Was there fear that certain types would descend into hyperbole?

It seems odd that we still haven't seen the goods. If the cause is still 'lack of time', then why wouldn't a few shots have trickled out?

Perhaps there is talk of a second LART replacing this LART will all results of this deep-sixed since Jim has announced so many improvements?

Not trying to be negative or disrespectful (since I have an ENORMOUS amount of respect for everyone that i know was at LART) but rather curious about what's going on.
 
Shawn,

I don't know who else said what where, but I was never told we couldnt post anything negative? Shots haven't been posted yet for many reasons. First, shots need to be sorted through, tested, graded, meta data consolidation, encoded, and uploaded. Also, we are editing actual sequences together rather than just posting single shots to start. There is a method to the madness that not everyone may agree/be happy with, but makes sense to be able to fully evaluate the footage. No one is hiding anything. Sharky has already said they will be posting some R3d files, so what is the point in posting ungraded low resolution quicktimes?
Trust me, everyone is giving as much effort and time as they can while traveling, and trying to communicate with different time zones, jobs etc. Bottom line, these things take time and we want to make sure that everyone gets to see the footage BIG and Beautiful!
Please don't assume there is some mad conspiracy theory behind LART. That is obsurd. The whole point of LART is to share with the RED community the results to help further the advancment of the camera.
As for the next LART, no information has been said about another test anytime soon.
 
Thanks for chiming in Mike, and I realise that everyone is under an NDA which makes it very difficult to discuss the tests in detail. If a second test like this does happen I hope the organisers will only NDA the specific products that require it (Such as prototypes) and allow for free flowing discussion afterwards on all other aspects. And whilst I look forward to seeing graded Red footage I'd very much like to see the original R3D files if that's at all possible. With as many people visiting RedUser and access to RedCine, Scratch and other software we'd be far more likely to highlight areas of problems or excellence in the footage than just the small team of individuals, or individual, grading the footage might notice - Large scale scrutiny is a good thing, but once graded and conformed is much harder to scrutinize.

Having said that, my original post was merely to get an idea of exactly what was tested, even if that's just a list that Ken or Gibby could post if no one else knows.

I also do appreciate the hard work of those involved and certainly recognise that the results take time to publish when post work is being applied to it. I remain confident that some interesting footage will be released and look forward to seeing it. I was just looking for a bit more clarity on what was tested is all :)

Paul
 
There is nothing in the NDA that says you can't discuss negative results, only that Gibcor controls the footage and the release of information.

However, someone did tell me later that it was mentioned on the day that we would bring any problems first to the attention of the RED team for their feedback & explanation before discussing it publically.

I would catagorize the tests as being more "real world" than being explicitly scientific. That's partly the nature of the facility we were shooting at, and mainly because people explored aspects of the camera that interested them.

In my two-thirds of a day there, I only had time to set-up a few lighting tests in an office. I shot a high-contrast day interior scene to see how the camera handled exposure generally, then I upped the contrast to see how the camera handled overexposed areas at different ASA ratings, to see noise vs. overexposure latitude. I shot a test subject lit to daylight and to tungsten. During these tests, Matt Uhry was also trying out different lenses that we had with us.

I sort of wish we had a reference grade lens there like a Zess Master Prime, something we know to be the sharpest on the market, but instead we had an odd collection, the RED zoom, some old Cookes and Ang. and some Zeiss Super-Speed primes, which were probably the best lenses available.

I don't know if the LART people are doing any post work to my tests.

For me, the most informative one was the one where I tried out different ASA ratings. My rough conclusion just looking at the footage on Lucas' Scratch station was that the image looked best in the 250 to 400 ASA range, like most people have found. There was better bright highlight detail at the high ASA ratings but at the expense of more noise, so it's a trade-off.

The only issue I thought was worth discussing with the RED team was how RED Alert! converted the files at different ASA ratings.

Here's the thing. A DP may decide to shoot the RED at 500 ASA, let's say, instead of 250 ASA, because he wants that extra stop of bright highlight information to be recorded in RAW.

So he sends his RAW footage to some post person who quickly (without making color-corrections) converts/processes the footage for dailies and editorial at 500 ASA.

Unfortunately, because each ASA setting in RED Alert! just boosts the overall signal, the clipping will look the same at 250 ASA as it will at 500 ASA, even though on the RAW file, there is more information in the highlights at 500 ASA. In fact, based on my tests, you won't see any improvement in highlight detail even at 1600 ASA, even though there definitely is more info there on the RAW file.

Now obviously you could open up the high ASA files at a low ASA setting and color-correct from there, boosting the brightness while keeping the brightest areas from clipping and thus see the improvement in overexposure detail when using the higher ASA ratings.

But then, it sort of negates the point of having automatic ASA ratings in post for processing files if the image that is created is inaccurate. It seems to me, and I don't know how they should do this, if the point of ASA ratings on the camera stored as metadata and being opened in RED Alert! at the same ASA rating is for automatic dailies or something, some sort of gamma curve or knee compression, black stretch, etc. needs to be applied automatically so that whatever ASA rating you choose to process the image at, the full exposure range of information is visible in the image. Otherwise, it's misleading - at high ASA ratings, you get the false impression that there is more clipping going on than there really is.
 
Thanks very much David. Interesting stuff. Do you know if the issue was put to Red or not?
Paul
 
"However, someone did tell me later that it was mentioned on the day that we would bring any problems first to the attention of the RED team for their feedback & explanation before discussing it publically."

Why?
The camera is very public, why not its drawbacks?
 
Professional courtesy... plus sometimes talking first to the RED team may save you some public embarrassment if the fault was due to something stupid on your part.

I sent an email to the RED team regarding my observations over the ASA rating conversion, but after waiting awhile for some feedback, I decided that enough time had passed to bring it up here. Besides, it's not a question of a mistake or something bad about the camera, I just think that selecting different ASA ratings should still display the available dynamic range when the file is converted, not clip information that exists on the RAW file.
 
David I remember you working with Lucas at LART and have been trying to figure out a good solution ever since that is both flexible and powerful and wonder what you think about the following...

The best solution I've come up with to date actually borrows from a 'very very old' system from the prehistoric times of tape recording: Superblacks and Superwhites.

If the RED sensor only records 11-12bits of useful data and every respectable uncompressed RGB file format operates in 16 or 32bit color space we have a whole extra 4 bits going to waste.

What seems to be happening now when exporting a 16bit (half float) or 32bit (full float) image is REDCine scales the dynamic range to fit. White is the maximum value of the color space and black is 0.

What I feel the best solution would be to be able to work with white balance and exposure controls as I do currently but use the extra 4 bits of data to store the data which spills over what would be 'white' at say 640ISO and not crush the data in the shadows which falls below 'black'.

This way you could batch export at the ISO you set in camera. When loaded into Davinci, Lustre, Flame, Scratch, Iridias or whatever finishing tool you wish simply load the "RED LUT" and see what you would expect. This RED LUT would be nothing more than moving up the black point and moving down the white point to clip the data where your shooting settings defined.

If you need some of that highlight data or some of that shadow data you still have it and you don't have to return to REDCine to Debayer the clip all over again.

Do you think a solution such as this would alleviate some of the challenges and confusions of the RAW paradigm or further muddy the waters?
 
Not sure if that would work or not. It seems to me that what you need to do is apply some sort of gamma curve so that as you brighten the underexposed image by processing it at higher ASA settings, no recorded detail gets pushed to the point of clipping.
 
I guess it would become something of a philosophical debate on where the destructive curve should be applied.

I would personally argue that it should happen at the colorist during the grade and the objective in REDCine should be to ensure he or she receives the most data possible to work with, while simultaneously maintaining a rough visual representation of what the DP intended during the shoot.

12bitLUT.jpg


The idea of course with products like Cineform is you stay inside the codec through post adding VFX, editing and conforming without ever leaving the codec and without losing the LUT, looks or metadata. However I'm not sure how reasonable of a concept that is unless you have absolute control over the post pipeline from ingestion to print (unlikely).

It would be trivial however to tell your colorist "All footage had a onelight export with the black point at .2 and the whitepoint at .8". The colorist could then use that as a guideline but still have wiggle room to make the falloff gamma more appealing or apply a stronger Scurve to the data.
 
Well, I was hoping one could eliminate the colorist until final color-correction -- you'd think if you were rating the camera at 1000 ASA for some reason, like a candlelight scene, and simply opened up the file in RED Alert or REDCINE at that ASA rating, the clipping in the converted file wouldn't be the same as if you had exposed it at 320 ASA.

It would add a lot of expense if all the dailies on a RED feature shoot had to go through a color-correction step just to see the highlight detail.
 
ISO_Comparison.jpg


You could do something like this. But I think rushes would be the only time you would want it as it has to distort the luminance pretty heavily. Someone like mmost would be able to comment for sure which would be preferrable but my gut tells me if you do end up going to a grade/conform it'd be better to only have linear modifications made to the image.


Edit:
Whoops it looks I accidentally deleted the label for this image:
Top image is 1200 ISO
Bottom is custom curve.
 
Dave,

I have all of the lightmeter data in a word doc I could email you, if you'd like. I was hoping to put some of you tests up with the metadata so people can see what you and Matt tested. Mike Burton and I are knee deep in getting the other footage conformed, graded and uploaded for everyone to see.
 
ASA setting

ASA setting

David Mullen ASC said:
you'd think if you were rating the camera at 1000 ASA for some reason, like a candlelight scene, and simply opened up the file in RED Alert or REDCINE at that ASA rating, the clipping in the converted file wouldn't be the same as if you had exposed it at 320 ASA.

It would add a lot of expense if all the dailies on a RED feature shoot had to go through a color-correction step just to see the highlight detail.
Click to expand...

Why not leaving the camera setup at 320 ASA for processing in REDALERT/CINE and shooting at 1000 ASA on your meters, having in mind the underexposure? So you know the data will be best in respect of the specificity of the sensor and hold your highlights details once processed.

Isn't it a common "expose for highlights" problem?

I'm afraid if you have to put some Scurves or other color corrections you still will need to apply your predefined LUT for (accurates) dalies.

Sorry for my bad english.

Patrick
 
Tresch said:
Why not leaving the camera setup at 320 ASA for processing in REDALERT/CINE and shooting at 1000 ASA on your meters, having in mind the underexposure?
Click to expand...

Because I want the offline editing to use images that don't look underexposed (or clipped inaccurately either), even if the final color-correction will go back to the original RAW files.

You have to understand that I'm coming at this from a feature background where you may shoot a hour or two of footage daily for a month or two. For a 2-day commercial or something, you can just convert and partially color-correct the originals before you start editing them.
 
If it's not going to need to be manipulated later it does seem like a simple drop down to select between

ISO Adjusts: [White point], [Gray Point]

would be a simple enough addition.
 
You must log in or register to reply here.
Back
Top