7219 Red Comparison

[Zach Gray]

I wanted to share a comparison test a colleague of mine shot comparing 16mm 7219 to 2k Red footage.

This comparison was not done to measure dynamic range, grain structure, sharpness, or the relative merits of either format, but rather to study the differences in color temperature and flesh tone handling.

We have had an ongoing discussion of the different qualities of the Red image, but flesh tones were of particular interest. We had noticed a trend in the vast majority of red footage that we had shot or reviewed from others at all levels – gray or dull flesh tones.

There seems to be a fundamental difference in how the colors roll off as the value decreases. With film, the colors seem to glow and shift hue as they drop off, where Red footage drops into gray. Carefully compare the core shadow, (the fast roll off into shadow after the highlight), and you’ll see a subtle but significant difference in the luminance and tone of the skin.

The other point of interest is how color temperate variations are expressed. The film seems to render higher temperatures closer to white and be generally more sensitive to subtle color, while Red appears to reflect truer object color. (Made more pronounced in this clip by halation?)

There are some great minds here on RedUser that can probably articulate both the aesthetic and technical differences much better that I could hope to, but I’m guessing that these image qualities may be related to fundamental differences between silver and silicon. I wonder if silicon will have the ability to capture some of these subtleties as improvements are made in dynamic range?

We shoot and own Red, and love it. It’s very good at what it does. Again, this is to study the differences, not to crown an overall champion. There are many factors that must be balanced when choosing how the image will serve the story and production.

It would be great if the discussion/interpretation could focus on color rendering as opposed to sharpness, clarity, dynamic range.

Hope to hear what you guys think of this imperfect and subjective test.

93MB H.264 7219_RED_Comparison.mov

Technical Info:
Aaton XTR 7219 Ziess Super Speed
RED 2k Build 16 Ziess Super Speed
I wasn’t there for the shoot, so I don’t know any details on exposure. Speaking out of ignorance, I think the film may have been a tad under exposed on the clip with the white background.

EDIT: The light is a kino miniflo (not diva, oops) with a 3200 and 5600 tube.

Film transferred on a Spirit thru a daVinci directly to ProRes 422. Mild SW grain reduction applied, but untouched color.

REDCine to RedSpace/RedLog OpenEXR32 with no adjustments. Rough match in AE using prolost logLin settings and curves. I didn't make any individual channel adjustments. I know I didn't match the blacks, I just wanted to get the images pretty close, and I'm not convinced that would cause a huge shift in what we're looking for in the image.

There wasn’t an effort made to preserve highlights, but get an interesting range of values and colors, so please don’t gripe about clipped highlights.

 

[hunterrichards]

I'm guessing the 7219 is on the bottom?

 

[Zach Gray]

Forgot to mention...

Film, Vision 3, is top, Red Bottom

 

[David George]

I was there and helped facilitate the shoot. I haven't confirmed this with Zach, but pretty sure the top images are the 7219. These tests confirmed some things that I had been feeling for a while. For sheer cleanness and resolution the RED easily beats the 500 speed S16. There is something immediately identifiable that many people would say is "better" about the RED material, but when you look closer (and this is subjective) there is something clearly more expressive about the film. Skin tones are just more pleasing, and shifts in color temperature of light are handled with more grace. BTW they are mini-flo lights, not divas...divas would be way too big to use like this! I've shot relatively little 35, and the main thing that being involved with this test and seeing the results did for me more than anything is remind me that, money aside, it's what I would still choose. 16 is too grainy depending on the application, RED feels sterile to me - although I still love shooting 16 and RED...

What do you guys think of the images? BTW - they look so different on an LCD computer display than on a reference quality CRT! The film actually looks fantastic on an HD CRT, but the LCDs computer displays really make the grain pop out!

 

[Nir Shelter]

Interesting test as we are talking only about the colour drop of from highlights to dark. These images should reflect the max potential of the footage from the camera in either RC28 or 36. That said what were you shooting in?

Footage out of a camera like Red is lit, framed and shot to a look intended to be brought out in the grade. In this case increasing the variance of hue with in a colour from the brightest to the darkest. Where as film will have this inherent in the image regardless of the grade. There for the test would be more conclusive when shooting RC36 and then passing it through a grade of the R3Ds. Specifically intended to accentuate the aspects of colour we are discussing.

I would really like to see the result of such a test as I am often faced with the question: 16mm or Red?

 

[David George]

I believe it is RC36...

 

[D Namir]

A picture worth a thousand words.

 

[David George]

I don't know how obvious it is from the images, but the lighting is a mixture of tungsten balanced kino (from the tubes on screen) and daylight balanced kino - just to see how the relative shifts in color would be handled...

 

[Nir Shelter]

That's good now it needs to be graded off and exported from the R3Ds to fit the needs of the experiment...

 

[hunterrichards]

Thats really interesting- I haven't shot the '19 but I haven't seen that much blue grain in '18 ever unless it was extremely underexposed.

 

[Dan Hudgins]

red-green crosstalk, adjust the chroma matrix values

red-green crosstalk, adjust the chroma matrix values

I have been doing some grading on both film scans of Vision T200 and some RED ONE images, and found that I could get the RED ONE images to look more like Eastman color by adjusting the color matrix to have more mixing of the RED-GREEN. Eastman color tends now to render flesh toward the Orange which helps keep flesh colors close to flesh color and not so much going Green or Magenta.

If you do negative matrix between RED and GREEN the flesh tones tend to go more Magenta in places, with positive matrix they go more tan color.

The sensor needs IR cut-off to prevent far red from exposing the blue pixels, doing a negative matrix cannot remove this issue, you need to filter the far red light out. Since Film is OK with far red and the blue layer does not pick up red at all, film and Bayer sensors will render flesh not quite the same. If you have a three chip camera you can filter the light to render more like a film, but because of the type of prism used in most 3 chip cameras that is not one of their goals.

The chroma going low in the shadows is a give away for some Digital Cinema camera footage and seems to be the fault of the de-Matrix software using filters to reduce the chroma noise, perhaps. I do not like that artifact and would like to have the true RAW footage to work with, but that does not seem like it will happen from here, maybe tests on the Acam dII can tell more on this issue since it will have true RAW data.

Using color enhancing matrix to get brighter colors does not always make flesh colors better if you remember what ECO or EF looked like you will know what I mean! Sometimes the placement of the peaks and blending of the colors can make people look better even if test charts don't. Kodak makes their movie negative films work better in mixed light and off K light since that is what people shoot under now. They were working on a "Type G" film that would look good under very off K light for home movies, I think part of the research ended up influencing todays Vision films.

Your film scans are way clipped, if you can send me a few frames I can try to scan them and see what I can do to show the range film can hold...

 

[pensound]

IMO, REDONE will fit to low-ISO at the dark scene. (not UP, Down to ISO200)
I hope evolution of the CCD.

 

[John Tissavary]

The Red footage needs a little more grading attention... I do not believe Red footage can be adequately assessed without a really good grade by someone with experience.

I do think tests like this are a move in the right direction. I'm very interested in knowing the distinct differences (advantages & disadvantages) both objective and subjective, of film and digital medai.

In tests I've seen where Red footage is graded to be as identical to film as possible I get the feeling that it could have been graded to better.

cheers,

JT

 

[Zach Gray]

Dan,

I think you are right, the most telltale sign is low chroma in the shadows. I'll post the .r3d files when I get into work tomorrow and see if someone with more grading experience can get better life in the skin. My hunch is it may get closer, but won't matter much, especially without doing secondaries.

Your work with the color matrices seems very interesting. I may be interpreting your post incorrectly, but it sounds like it would be difficult to apply this processing as a LUT. Does it need to happen during the de-bayer?

(I'll have to check to see if we still have the negative from the test.)

 

[Dan Hudgins]

LUT issues

LUT issues

Dan,

I think you are right, the most telltale sign is low chroma in the shadows. I'll post the .r3d files when I get into work tomorrow and see if someone with more grading experience can get better life in the skin. My hunch is it may get closer, but won't matter much, especially without doing secondaries.

Your work with the color matrices seems very interesting. I may be interpreting your post incorrectly, but it sounds like it would be difficult to apply this processing as a LUT. Does it need to happen during the de-bayer?

(I'll have to check to see if we still have the negative from the test.)

Yes the chroma matrix is done as part of the de-bayer, but in the case of the de-bayer used for the RED ONE the issue is more complex. Because the wavelet compression can cause some high frequency artifacts in the fine chroma the filtering of the chroma needs to be somewhat aggressive. I am not saying that this is where the disparity between the chroma and luma happens, but it can happen there depending on the bit depth for the chroma, it may also build up in the other programs used to grade the images.

My CC is all RGB, so there is no chroma vs. luma in DANCAD87.EXE (tm) at this time.

I have a chroma matrix in my CC that can be used after the chroma matrix in the de-bayer. I have been working with some TIF frames from R3D that were saved in REDLOG type. I do not recomend using the CC in REDCINE, rather save the images with the controls in REDCINE at neutral. In my program do not lower the white clip point (in Levels 1), except if the white balance is off, since you get more filmlike highlights if you use the Curves 1 to rase the shoulder but not to crush the highlights flat.

My CC system was based on the notion that the 35mm film negatives would be scanned using color separation filters, so a chroma matrix is needed to correct for the lack of mask in some negative stocks, and to adjust the mask in negative stocks that are masked.

After much testing I came up with an order that seemed to work for images from any source, Levels 1, Curves 1, chroma matrix, Curves 2, and Levels 2. Then filters can be used after the basic grading steps. Levels adjust the black and white points, and the gamma to set the middle gray to the right value. Levels 2 can be set a little tighter than Levels 1 if you like, and can adjust the over all gamma which means midtone brightness. Curves are used to adjust the head and toe of the tones in ways that just the gamma cannot so that all the boxes on the gray step are balanced and the right brightness. Curves 2 is used after the chroma matrix to correct for some crossover or color shifts that can show up, like off color shadows or highlights. The chroma matrix has nine values and can be used to adjust the color purity, saturation, and to make the two strip look, or various monochrome filter effects.

In the chroma matrix you have positive and negative values for red, green, and blue that map to the output red, green, and blue. Equal positive values give you a monochrome output, greater than 100% of the primary with negative values of the other two primaries increase saturation but in a more film like way than using a chroma/luma saturation filter that can give muddy results.

I have a vector wave form display that helps when making adjustments once you learn how to read it. There is also a video type wave form display that show all three colors so you can adjust the signals that way also, and use the "density" probe to just look at part of the image on the waveform displays.

If I was going to scan the negative I would make several exposures to capture the full range in the highlights and shadows, then grade the image to hold some of that detail. With the RED ONE footage I would save redlog TIF and also try to hold some highlight detail. If you shoot RED ONE with a 80 series filter you can get maybe a stop or more more dynamic range under tungsten lights. If you shoot film with off K light it reduces its dynamic range also since one layer is over exposed and another under exposed. I scan movie film to 48bpp TIF files now, and even then try to FUSE several exposures to get better results. If you scan with just one frame the digital artifacts and noise become part of the images, but if you scan several times and fuse the images together, then you loose much of the digital artifacts. Maybe with a better digital camera in the scanner taking several exposures would result in less of a gain, but some gain should result.

Doing the chroma matrix in a 3D LUT would take 65535 cubed addresses, which is quite a bit of memory.

The point I was making was about the magenta or green skin tones you can get sometimes with digital cameras, and that if you use positive values in the chroma matrix for red-green you can get the flesh colors more tan looking, if you force the colors apart you may find parts of the face look OK, but other parts go pink or greenish, that can come from mixed light, or other tracking problems in the source images, so you have a choice to make the color errors greater or less, or at least shift them around.

If you pull parts of the chroma to "on color" points you can make shifts in other near by colors, so the best way to get good color is to balance the lights used to the right K value and use optical filters to white balance the camera to keep the source images close to correct to start with... GIGO.

You can see an example of the chroma matrix being used and abused at my site, that LAD film I scanned is in sad shape the gray patch does not match the gray steps for hue... so there was no setting that could balance the whole frame with its uneven shifts...

 

[Zach Gray]

R3d Uploaded

R3d Uploaded

I've uploaded one of the .R3Ds. If anyone can grade to improve the shadow chroma and color temp, that would be great to see. (and learn!)

Light Background 898 MB

Dark Background 1644 MB (5hrs left UL, slow outgoing connection)

 

[Dan Hudgins]

Light background frame 0467

Light background frame 0467

I've uploaded one of the .R3Ds. If anyone can grade to improve the shadow chroma and color temp, that would be great to see. (and learn!)

I did a quick CC on Light Background frame 0467, tried to have it look a little like T500 maybe? Without the film of that frame it is just a guess.

Maybe someone else can have a stab at that frame also with some other software?

The RED sensor values seemed clipped in the REDLOG TIF I made, so I had some problems with the hair over the white background, GREEN and BLUE had a little more in the histograms at the top(?). Maybe some fill, an eye light, and a half stop less on the background would have given a better result?

[sample is for viewing on natural CRT gamma about 2.2 to 2.4 you may need to adjust the brightness or contrast on your monitor to just make out the shadow details in the black shirt, otherwise the color in the faces may look too strong (too dark setting) or the lens flare may wash out too much (too light setting)...]

 

[Sanjin Jukic]

Nice and interesting to show some film freaks.

 

[Dan Hudgins]

Light Backgrouund frame 1206

Light Backgrouund frame 1206

I've uploaded one of the .R3Ds. If anyone can grade to improve the shadow chroma and color temp, that would be great to see.

I had a few minutes free and did a quick CC on Light Background frame 1206, with a little different settings than on frame 0467. I gave this one a cooler softer "frosty" look, than frame 0467. See the Dark Background frame 1478 CC on the next page also...

I lowered the "white" clip for the Green and Blue a little to match the slight burn out in some reflections on the faces, the Red was a little too hot to hold as much as I would like to have to work with for a filmout.

When you click on the thumnail you get a larger view on a light background, if you cover the white screen with black paper or black-wrap foil it looks more like a movie, if you cover the white parts of the frame you can see the shadow detail that is there but hard to see with all the white in the frame...

[see not above about monitor gamma/etc. adjustments]

 

[Zach Gray]

You have definitely been able to improve the shadow chroma in those images.

What makes this test interesting is the changing color temperature of the light source. Do think it will be possible to create a filter or process that can be applied to RED footage generically? It looks like a single frame or lighting condition can be finessed, but can the color handling and color be improved globally? So if it looks good with 3200, will it fall apart in 5600?

If there were a way to process red footage so it was more sensitive to shadow chroma and color temp handling overall that would be amazing. You could start your grade much further ahead.