Looking Deeper into Log [Archive]

Lee Powell

March 17th, 2016, 02:59 PM

Thanks for the detailed background on REC-709 and LOG gamma curves. As I see it, the LS300's J-Log1 has two main advantages over ITU709:

* Linear exposure of highlights in the 100-109% superwhite range, no highlight knee.
* All Camera Process settings standardized, leaving only exposure and white balance.

With ITU709, all Camera Process settings are adjustable. What I do is set the highlight knee to compress the 100-109% superwhite range and use Gamma and Stretch settings (+3 to +5) to boost shadow and midrange exposure. I then adjust highlight exposure to ETTR level. This gives me a gamma curve that's a decent match for viewing directly on an HDTV monitor with a standard 16-235 REC-709 curve. If I want more highlight detail, I can recover the superwhites with an inverse knee in After Effects/Color Finesse, and tweak the shadows as desired.

With J-Log1, I can still preview footage on an HDTV monitor, as the log curve's pedestal gives me a suitable black level. However, the uncompressed superwhite range makes the midrange look underexposed. The LS300's 1D-LUT compensates well enough to make exposure look normal, though saturation still looks washed out. A proper 3D LUT would balance it all in post, but so far, JVC has not delivered on that promise.

The great thing about a LOG curve is that it's linear. That allows you to apply a pure gamma curve in post to adjust it to any gamma level you like (e.g.REC-709 gamma 2.2). A practical LOG curve, however, is linear only in the highlight and midrange regions. To handle the shadows, a knee is used to roll off shadows down to a raised pedestal level. For a pertinent example, check out the gamma curve on page 3 of Panasonic's V-Log Reference Manual for its Varicam line:

http://pro-av.panasonic.net/en/varicam/common/pdf/VARICAM_V-Log_V-Gamut.pdf

With V-Log, 18% gray level is at EV+0, and the LOG curve's linear range extends from about EV-4 to EV+8. Below EV-4 are four stops of shadow range down to black at EV-8. A proper 3D LUT would need to use a non-linear curve tailored specifically to match the V-Log knee in the shadow region. LUTs for the V-Log L profile on the Panasonic GH4 must take this into consideration. (The GH4's V-Log L is almost the same curve as the Varicam's V-Log. The difference is the highlight range extends only up to EV+4 on the GH4).

With J-Log1 on the LS300, JVC has not yet disclosed where the shadow knee lies on the LOG EV scale. Until that's measured and documented, it's anyone's guess how to make a proper 3D LUT for the LS300. In addition, color saturation level is unknown as well. I have tried JVC's downloadable LUT package for J-Log1 and IMHO, the color saturation level is garish, i.e. it's wrong.

What we have so far with J-Log1 is already far better than the crippled V-Log L on the GH4. To fulfill its potential, however, J-Log1's shadow knee and color saturation level must be properly documented and accurately modeled in a 3D Reference LUT. Until then, we're all just winging it with subjective grading tools.