"True Colour" Scene File

[Tim Dashwood]

Paulo Ciccone, one of our members and HD100 contributors, just emailed me a link to the results of his tests with a DSC Labs colour chart.
He has successfully calibrated for chart and created a "true colour" scene file recipe.

Here's the web page with the full article: http://www.paolociccone.com/hd100-calibration.html

A few people have emailed me asking about ITU-709 colour space match settings, and I haven't had the time or resources to create such a file. I imagine Paulo's scene file will fill that gap.

It is also interesting to note the results of colour cast on the default matrix settings. Some others have mentioned it before, but Paulo clearly demonstrated what is going on with the Cine matrix. It looks like it has a very warm response compared to the calibrated one.

Thanks to Paulo for the great article. You've inspired me to write up articles in html format!

[Steven Thomas]

Thanks Paulo for taking the time to perform this calibration.

I'm looking forward to trying these settings.

I'm glad you spent the time to maximize color space. I like the idea of leaving color correction for post production. For those who perfer not to, the JVC offers a lot of internal control.

I might have to break down and buy a DSC chart. You have to wonder how closely our cameras are calibrated.


Also, Tim thank you for this link.

[Paolo Ciccone]

Originally Posted by Tim Dashwood

Thanks to Paulo for the great article. You've inspired me to write up articles in html format!

Hey Tim.

Thanks for the sticky. Initially I tought about posting the article in the forum but it became soon to complex in formatting, with all the pictures and links. HTML was the logical choice. I used Adobe Go Live to format the page, my first attempt to use the program. I'm actually quite impressed with the flexibility of this application.

Take care.

--
Paolo

[Paolo Ciccone]

Originally Posted by Steven Thomas

Thanks Paulo for taking the time to perform this calibration.

You're welcome Steven. Hope it helps.

--
Paolo

[Miklos Philips]

"A director must be a policeman, a midwife, a psychoanalyst, a sycophant and a bastard."
- Billy Wilder

[Earl Thurston]

Excellent work, Paolo.

One thing I found myself wondering: With the IRE range adjusted to the extremes of the DSC chart, just how "white" and "black" are those are corresponding targets on the chart? Wouldn't a scene with lighter or darker elements get clipped? For example, the black target on the chart compared to a piece of black velvet.

[Paolo Ciccone]

Originally Posted by Earl Thurston

Excellent work, Paolo.

One thing I found myself wondering: With the IRE range adjusted to the extremes of the DSC chart, just how "white" and "black" are those are corresponding targets on the chart? Wouldn't a scene with lighter or darker elements get clipped? For example, the black target on the chart compared to a piece of black velvet.

Thanks.

DSC has some detailed documentation on their website. Particularly interesting is a paper presented by their founder at a SMPTE conference. He's more qualified then myself to explain the fine points of colorimetry and calibration. My understanding is that their charts are calibrated to reproduce pure white and black and so, when properly lit and exposed, the chart should reach the maximum available for the camera.

Once you know that the camera is calibrated at the limits of its range you should set the zebras to warn you well before you reach those limits. Of course this could be problematic in a run-and-gun situation. I have my zebras set at 85-95%. Also, DVRack, the program that I use for monitoring, has zebras for light and dark areas.

--
Paolo

[Tim Dashwood]

Originally Posted by Earl Thurston

One thing I found myself wondering: With the IRE range adjusted to the extremes of the DSC chart, just how "white" and "black" are those are corresponding targets on the chart? Wouldn't a scene with lighter or darker elements get clipped? For example, the black target on the chart compared to a piece of black velvet.

That's a good question, and it should be noted that Paulo's scene file does not utilize knee or black stretch, so the dynamic range is much more limited than the cinewide setting. So yes, with knee set at 100%, the whites will clip much earlier than with knee set at 80%. There will also be less shadow detail than if black stretch were increased.

[Nate Weaver]

I've chosen to implement Paolo's color matrix, saturation, and gamma settings while retaining Tim's black stretch and knee tweaks.

I'll see how it goes.

[Paolo Ciccone]

Originally Posted by Tim Dashwood

That's a good question, and it should be noted that Paulo's scene file does not utilize knee or black stretch, so the dynamic range is much more limited than the cinewide setting. So yes, with knee set at 100%, the whites will clip much earlier than with knee set at 80%. There will also be less shadow detail than if black stretch were increased.

Tim, my understanding of the knee function is that it generates a smoother transition between the highlights when clipping happens. In doing this the knee circuitry actually reduces the amount of data that is recorded. It has a nicer look but the range is more limited with the knee active.

In a run-and-gun situation you might want to use my configuration with a 90-95% knee in order to stay on the safe side. In a controlled environment it actually gives you more range to disable the knee and keep the levels in check.

Of course I could be wrong :)

I decided to re-run my test and verify what really happens with the knee enabed.

A picture is worth 1000 words so here are several snapshots of the WFM from my config file with knee at different values:

Knee 100%
Knee 95%
Knee 90%
Knee 85%

We can see a pretty strong flattening (clipping) of the highlights as we increase the knee value.

I also took some snapshots of the black stretch. The knee is left at 100% while increasing the stretch value. Use the "Knee 100%" image as a reference for black set to "Normal".

Stretch 1
Stretch 2
Stretch 3

It seems to me that the stretch causes a whole shift to the upper range and that's why I didn't use. It does make the black chip look more round, but it affects a lot of other things too. Again it is perfecly fine as a look but my goal was to get as close as possible to 1:1 representation.

BTW, Tim, big mistake on my side about the color shift on the Cinelike and your CineWide config. I did't realize that switching to another config changes the white balance. I balanced at the beginning of the test and never changed it, assuming that the camera kept the 3200K value. My apologies, I'm gonna retake those samples after setting WB at every switch. I'll update the article asap.

Take care.

--
Paolo
http://www.paolociccone.com

[Paolo Ciccone]

Originally Posted by Nate Weaver

I've chosen to implement Paolo's color matrix, saturation, and gamma settings while retaining Tim's black stretch and knee tweaks.

I'll see how it goes.

That's cool Nate, I'm eager to see your conclusions.

--
Paolo

[Stephen L. Noe]

Originally Posted by Paolo Ciccone

BTW, Tim, big mistake on my side about the color shift on the Cinelike and your CineWide config. I did't realize that switching to another config changes the white balance. I balanced at the beginning of the test and never changed it, assuming that the camera kept the 3200K value. My apologies, I'm gonna retake those samples after setting WB at every switch. I'll update the article asap.
http://www.paolociccone.com

I'd noticed that and wondered about the white balance. How about white balancing and then getting a reading. Now place the reading on a preset and rerun the test for the cinelike gamma. The colors won't punch as much as the standard gamma but they transition more smoothly. Excellent work and I like your scene file. If you get a moment can you load my scene file and take a look at if for me on your chart?

Click here for scene file or Here for screenshots.

Thanks,

[Tim Dashwood]

Originally Posted by Paolo Ciccone

BTW, Tim, big mistake on my side about the color shift on the Cinelike and your CineWide config. I did't realize that switching to another config changes the white balance. I balanced at the beginning of the test and never changed it, assuming that the camera kept the 3200K value. My apologies, I'm gonna retake those samples after setting WB at every switch. I'll update the article asap.

Yeah, I would suggest using preset 3200K if you already confirmed that your lights are exactly 3200K. The way the scene files record EVERYTHING, including the value set in the A & B white balance is frustrating. I didn't realize it at first either.

Originally Posted by Paulo Ciccone

Tim, my understanding of the knee function is that it generates a smoother transition between the highlights when clipping happens. In doing this the knee circuitry actually reduces the amount of data that is recorded. It has a nicer look but the range is more limited with the knee active.

Not quite. The dynamic range of the CCDs is much greater than the dynamic range of the video data encoded to tape or of a standard video monitor. Your chart represents probably only about 4 stops of exposure. You should set it in front of a window on a bright day and compare the knee settings. You'll find that the knee brings the wider dynamic range of the CCD into the effective range of what we can record to tape. There is a cost however, and you've shown it in your knee tests. Values that were previously above the knee point get shifted down. Therefore the white chip is no longer 100% white, but you can now expose detail in white values brighter than the white chip. For the most part the knee should be able to extend the dynamic range at least 1 stop, maybe 2. This gets us closer to the way film negative handles highlights, (but not that close.)

Black stretch works essentially the same way, but on the other end. It takes a bit of the lower mid range to use in the shadows so that the blacks are not entirely crushed. Black stretch should not affect any values in the mids or highlights.

[Paolo Ciccone]

Originally Posted by Tim Dashwood

The way the scene files record EVERYTHING, including the value set in the A & B white balance is frustrating. I didn't realize it at first either.

Talk about potential for embarrassment :) I reshot everything and updated the web page. Instead of using the preset at 3200K I switched to a different config and white balanced every time. Much better now. Sorry for the mistake.

Interesting stuff about the knee and black stretch. I'm still not convinced that the knee doesn't cause data loss but I'll defintely try to get some real life test as you suggested.

--
Paolo

[Paolo Ciccone]

Originally Posted by Stephen L. Noe

If you get a moment can you load my scene file and take a look at if for me on your chart?

Click here for scene file or Here for screenshots.

Thanks,

Hi Stephen. I'll be glad to look and test you scene file. I'll probably do it tomorrow, today I had to re-shoot the whole thing ad I need some rest :)

I'll post an addendum to the article with your scene file asap.

--
Paolo