After the release of my article about the camera calibration, the feedback that I received from the HD100 forum was very positive. Some members pointed out that the color matrix looked good but the dynamic range was not optimal. In particular Tim Dashwood mentioned the knee settings and he kindly provided a link to an article, published by Panasonic, about the use of the knee function and the design of a specific gamma function (CineGamma) that provides more latitude than the traditional settings for video cameras. In fact the article points out that, for the longest time, video cameras were not expected to go beyond the boundaries of video. With HD we are basically trying to use a video device for film-like work.

I spent another few days designing and performing tests aimed to verify the effects of different Gamma functions, knee, black stretch and dynamic range. What follows is the result of these extensive tests followed by my conclusions. Keep in mind that my target is to shoot movies at 24fps in a controlled environment. Run-and-gun interviews in broad daylight might require different adjustments. Also, these settings will work for tungsten lamps and daylight but will need to be adjusted if you use HMIs. Lastly, I wanted to mention that the frame grabs presented in these two articles show just a distorted picture of what happened. We can't rely on compressed JPEGs displayed on a computer monitor. In order to verify and appreciate the work described here you really need to shoot your own charts on a calibrated monitor.

My starting point is my "TrueColor" configuration as presented in part 1 of this article. I switched to "CineLike" gamma and set the value to -1. This gives me a nice even display of all the grayscale targets positioned approximaly where they should be. For reference, here are the frame grab, WFM chart and histogram. Knee and black stretch are off. Iris is set a f4. The chart used is the DSC FrontBox Pro HD. The white clips at the bottom right is actually brighter than the one at the top left of the grayscale pattern.

After taking these measurements I emailed DSC in the hope of getting some advice about the effect of the knee function. DSC is really one of those companies that care about customers. They were so pleased with my article that David Corley, president of the company, called me and graciously spent 20 minutes on the phone discussing the fine points of camera calibration. The good news is that he reviewed the article and found it accurate. There are no bad news :)
Corley basically agreed that it's a delicate matter when it comes to knee settings. There is a point where the function starts chopping off too much data and could become detrimental. Of course this has to be taken in context. Shooting style and expected post work play a role in deciding the acceptable threshold.

In the following clip I marked the points where I changed the knee setting. The superimposed WFM chart helps in understanding the effect of the knee on the video signal with the chart the frame grabs and WFM charts of the application of the knee function at 95%, 90%, 85% and 80%. While we don't see any side effect below the 80% region, it's quite obvious how the affected highlights flatten out, an effect that reminds me of the black compression when observed on the WFM. This clip should be viewed on a calibrated monitor, otherwise you'll miss the point where the knee separates the two brightest chips. This happens at 90%, keep an eye on the white chips at the bottom right of the chart. The scene is overexposed on purpose, the idea is to see how the knee function handles highlights in extreme situations.

Click here to download the clip (H.264, 51Mb)

For my taste, knee at 90% is just right. It gives me more room at the top, probably 1/2 to 1 f-stop without affecting the signal too much. Remember, my goal is to preserve those precious bits as much as possible. Whatever I preserve in acquisition will help me during the data-loss phases of color correction and compression to the final codec.

Finally, I measured the latitude of this configuration. To tell the whole truth I didn't know exactly how to do it. I mentioned this with David Corley and he suggested to talk with Scott Billups. I thought about it and then decided to email him with my story and see if he could look at my current analysis of the problem and if he could suggest a procedure for measuring the latitude.
Mr.Billups replied in about 1/2 hour (0:30am) and confirmed that the article was on track. To measure the latitude he suggested the following: "As for checking the latitude; light your chart so that it's blown out at your widest exposure and then keep stopping down one f-stop at a time. When you run out of f-stops throw in ND in one stop increments and then go to shutter speed. In the end, you want to take the image down to where you can barely make out the image. Count your stops (probably 6) and there you go."

So I did. The final number is close to 7 stops, a number that I believe is pretty good, especially for a camera in this price range. In order to get to the last bit I switched the shutter speed to 1/100, twice as fast as the default of 1/48 used for 24fps. You could probably further than this but I believe that would start compromising the image quality elsewhere.

So, in conclusion here are my revised settings for accurate color reproduction and wide latitude.

I'm confident that this configuration will improve the look of your footage compared to any of the stock settings. Keep in mind, though, that any two HD100 might have slightly different reactions and you should calibrate your own HD100 with a chart.

One note about the use of these settings. I spent a considerable amount of time to get an engineering approach to color calibration. If you use these settings or configuration derived from them in a professional production I would appreciate if you credit the source. Also, if you have the chance, drop me an email to let me know which features/shorts/commercials have been shot with the HD100 calibrated as described here.

Ciao!

--
Paolo
PS: sorry about the watermark. I just found out that some people on a Japanese blog are copying the about table and not only not mentioning the source but crediting it with their name!