A quest in the debyering/exporting of RED footage

[Evangelos Achillopoulos]

For many of you, am the guy that is chasing Jim and Graeme about RED latitude…

 

It’s not exactly that, instead, am an engineer that has above all, the axiom to be objective.

 

So in order to do my job as a DIT/DI supervisor/Production engineer, I have to test the cameras and find the possible short fallings in order to try solutions and protect my fellow cinematographers in such a way, that when we will try to put on film (or digital) the collective work of so many people that are involved in the production of a movie, do not have surprises.

 

Of coarse I don’t like marketing blah blah… because when the problem will arise I will be the one that has to solve it or workaround it.

 

But let’s get to the chase…

 

As always, when a new camera falls into my hands I do two things first I measure real resolution in a chart that has 2000 lines resolution like the ISO12233 to know, what the effective resolution is, and second I shoot a transmitive step chart from Danes Picta that has 28 steps of 0,15 neutral density chips a total of 14 f-stops.

 

On the ISO chart, RED has surpassed my upper limit easily with 1550 lines of horizontal resolution. Measured with a 50mm Arri Distagon 2,1 in 4 aperture. I can upload the .R3D file for this measurement, but I thing the resolution that this camera has, is toping even 4K scanned film resolution, so it’s almost irrelevant to continue the discussion in resolution.

 

Another thing that was a very pleasant was the absence of chromatic aberrations in the corners of the resolution chart…

 

On the second thing, the step chart, I shoot it in a darkened room as its being evident from the photos, with a back light source of 3200K diffused. I was using 3200K in order to measure the worst case scenario and also it’s the same light that am using in all my camera tests. When we ware looking the chart with our eyes the darkest step was having a very easily noticing difference from the totally black carton board background. This is ensuring that the chart was correctly lightened. Also on the photos is easily understood that the camera wasn’t have any stray light coming from the board. The step chart is calibrated with error less than 5% on every step.

 

Photo taken with flash

 

 

Photo taken without flash

 

 

This test is for evaluating what is happening into the deep dark areas, for that reason ignore the first 8 bright chips, in which the first it was just clipping.

 

As you all know there’s nothing that can be adjusted on the camera in order to affect the exposure of the RAW footage. I only measuring released versions of firmware, so the camera was having the 13 version, which is the current release firmware.

 

I know what beta means, that’s why I don’t want to play with beta versions…

 

I used the 1,0,4 RedCine to export a tiff and a dpx in order to check out the headroom of the camera (Dynamic Range).

 

That, to my surprise, was an adventure, every change that I made, even cropping or new white balance, I was getting different results, not to much different but different…

 

After a while playing with RedCine and then with Photoshop and Brightness/Contrast in order to see what was visible on the step chart, I realize that I have to experiment heavily with the Debayering software in order to maximize the results and avoid the tweaking of every shot in RedCine, which is a nightmare.

 

But I have to test the settings that are looking good to real footage and see the resulting real life image. Just in case, to avoid plastic effects…

 

Since I don’t have the luxury of time to play that much and I don’t have the diversity of footage to do such thing, I decided to let in public the R3D RAW file of the reference step chart shoot and let all of you, that want to play with REDcine or REDalert to share with the forum your findings…

 

A proposed work flow could be:

 

• Use RedCine or RedAlert and specific settings that you thing its best for maximum latitude export a 16bit Tiff or whatever you thing its best.
  
• Import the resulted image to Photoshop or any other software that will allow you to judge the latitude and noise levels (with Photoshop you can use Brightness and Contrast to expose noise and ND steps) Emphasies in the dark chips and ignore the first 8 brightest.
  
• Note that every one ND is ½ stop the chart has 28 steps (a total of 14 stops is available on the step chart) and the outer edge of the deeper step is in between the two bright (over and under) dots and in line with the right one
  
• After many trails find the best setting and try it to your real footage and see how it looks like.
  
• When you are getting in same conclusion post it in the forum to be shared with the rest of us.
  

I hope the combined work of so many professionals will allow us to find a way through, to the new tool that is called REDONE and maximize its performance to our common benefit.

 

I reserve my personal opinion for the time being.

 

The link to the .R3D file the size is 67Mbytes .R3D 3200K TS28D step chart

 

For reference am posting the links for three popular HD cameras that has being shoot the very same step chart in exactly the same conditions.

 

Sony CineAlta F900R -3db HG3

Panasonic Varicam F -3db FilmRec Dark Compression

Panasonic HVX200 0db Cinelike-D

 

Any comments or suggestions are welcome.

 

With respect to community,

[Gary McClurg]

The link to the .R3D file the size is 67Mbytes .R3D 3200K TS28D step chart

 

The link doesn't work...

[Evangelos Achillopoulos]

For same weird reason the FTP does not allow .R3D files to be downloaded?

So I add in the filename at the end a .zip extension?

 

After you have downloaded the file remove the extra extension .zip from the filename.

 

The new link is A001_C004_080314_001.R3D.zip

 

Sorry for the inconvenience.

[DJ Joofa]

Evangelos,

 

Since you have the infrastructure, and are well placed technically, may be you can do an experiment to reverse engineer the sensitivity curves of the Red's Mysterium sensor. It may not give exact curves, but is doable, and is instructive, and might require stuff such as Wolf Faust's targets, or equivalent. PM me and I shall send you some information on how to do this.

Edited March 19, 2008 by DJ Joofa

[Evangelos Achillopoulos]

Joofa thanks for your kind words, but I thing there is no point to reverse engineer RED original curves since am interested on how I can use the RED RAW converting to something useful to our post production and DI process.

 

If you want to do that your welcome from me to use the .R3D file in order to make that research?

 

Any one with some results from the charts??

[Michael Morlan]

Hi Evangelos,

 

Thanks for taking the time to shoot a step test. I'll play around with your .R3D file in the next few days and see what I can discover. I'm still learning a lot myself. Your REDUser thread is quite interesting as well. Besides subjective evaluation of noise in blacks, what other methods would you recommend for ascertaining usable dynamic range?

 

Best,

 

Michael

[Evangelos Achillopoulos]

I can dare to suggest www.imatest.com . You can download the trial version and use the step chart function, export .Tiff 16bit and import it to step chart analyzer, select the area of the steps in the image, don?t forget to select 0,15D step or TS28D Danes Picta chart on the dialog, and let everything else to auto.

 

The best range is on High quality and mid high or 0,1 f-stop noise level or 0,25 f-stop noise level.

 

It?s a nice tool can be helpful, for reference I have put on my thread a set of popular cameras stills with the exact same chart.

 

Have fun.