30 Oct Matching the Pocket 4K & GH5

I’ve been shooting Panasonic’s GH line since the GH2. The GH3, GH4 and GH5 have each been my main cameras in their day. Even though the Pocket 4K will become my A-CAM for most work, I’ll be keeping the GH5 in rotation for two reasons:

1. It has features, like IBIS, that the Pocket 4K lacks.
2. Who knows when I’ll be able to get my hands on another Pocket 4K.

As such, it’s likely I’ll need to intercut the Pocket 4K and the GH5.  Since I’ve had a similar need each time I’ve upgraded in the past (GH3 > GH4, GH4 > GH5) I’ve come to appreciate the wisdom in evaluating two cameras’ differences – and building a corrective LUTs – before needing to match them in a production environment. Be prepared!

Methodology

Scene

I set-up a test shot in my office with natural daylight as my key light, bounced for fill. In addition to a Datacolor SpyderCheckr and Spyder-Lens-Cal, I drafted two subjects with differing complexions. I think it is important to test with actual live subjects as the skintone patch on a color chart has little to do with how actual skin looks, and the vast majority of the time, how real people look is the most important aspect of my work. Additionally, I included windows and some edison bulbs to evaluate not only how the two cameras handle highlights, but also the different color temperatures.

Set-up

Both cameras were locked to tripods positioned as closely to each other as possible. Their lens mounts were the exact same distance from the chart, certainly within a millimeter. I used the Panasonic 25mm f/1.7 on both cameras for a couple of reasons:

1. I had it handy.
2. It’s a pretty neutral, high contrast lens that I knew I could trust focus.
3. I wanted to see the impact of Panasonic’s in-camera lens correction on the GH5 compared to the Pocket 4K which does not do anything of the sort.

I metered the scene with my iPhone 7 Plus wearing a Luxi light meter running the Cine Meter II app.* It read:

Lux: 750 WB: 4850k Tint: +3 magenta With the both cameras’ at their native ISO400, I set the lens aperture to f/4, and manually dialed-in white balance to match**.

Results

Straight out of the camera

The most striking difference between the two cameras is how much wider the field of view is from the Pocket 4K. (I suspect this difference wouldn’t be quite as dramatic against the GH5S.) I was not expecting such a big difference – this alone could be the reason for choosing one camera over the other. The next difference to note is their respective “LOG” profiles’ gamma curves. While Panasonic’s V-Log L utilizes less than 70% of the available bandwidth in order to match the same curve as all of their cameras (including the Varicam, with much more dynamic range) Blackmagic Design instead optimizes the gamma curve for each camera. The purpose-built “Pocket 4K Film” curve clearly uses more of the signal. Since it’s hard to judge exposure and color from flat log images, I’ll save further comment until we “normalize” the images for viewing on a REC709 display.

Default LUTs

I can’t say I think much of either manufacturer’s default REC709 conversion LUT, but Blackmagic’s “Extended Video” LUT does a better job maintaining correct levels while at least attempting to tone map the highlights into something usable.  Panasonic’s generic LUT is pretty much unusable, in my opinion.

Manual adjustment

For these two images I exclusively used Davinci Resolve’s curves tools (monitored with the vectorscope) in this order:

1. Adjust middle grey to 50%
2. Adjust shadow patch to 25% and highlight patch to 75%
3. Adjust specular highlight (the steel ball on the cube) to just barely clip
4. Adjust the cube’s black hole to noise floor
5. Adjust each memory color’s hue to align
6. Adjust each memory color’s saturation to balance
7. Adjust each memory color’s value to balance
8. Adjust highlights to hold as much detail as possible

Now, it should be noted that I was overly aggressive in tone mapping the extended highlights into a REC709 image. This was done on purpose simply because I wanted to see how each camera would react. Overall, both cameras handle this scene quite well. I noted a few differences:

1. At the same white balance setting, the GH5 has a global green cast, whereas the Pocket 4K is effectively spot on.
2. At the same ISO, the GH5 is about 1 stop overexposed, whereas the Pocket 4K is within 10% of the measured rating.
3. The Pocket 4K’s largest deviation from any memory color was 4%, with an average deviation of 1.7%. The GH5’s largest deviation was 23%, with an average deviation of 14%. (That’s actually still very good.)
4. While absolute dynamic range is very similar, the Pocket 4K has more usable latitude. It holds color more accurately into both the shadows and highlights, allowing more usable range to reasonably be recovered.
5. It surprised me to see that the Pocket 4K actually resolves a tiny bit more real detail, and actually looks a bit sharper. Due to the GH5’s in-camera sharpening, I expected it to at least look sharper even if it was false ringing.)
6. The Pocket 4K holds saturation over a greater luminance range. This is a good thing for most images, but it can cause very bright colors to oversaturate to an unnatural state. (Note the exit sign) and in extreme cases, even cause color channels to prematurely clip.

Matching

Given that the P4K is so close to the metered readings, and so close to the chart values, it makes the most sense to match the GH5 to the P4K’s balanced/normalized values. To match, I did the following:

1. Small shifts to white balance and tint
2. Small adjustments to gamma curve
3. Small increase to midtone saturation
4. Small hue adjustments to memory colors.

That’s it. With more work, the match could be made more exact, but I feel that even this amount of correction yielded results that could be seamlessly intercut. Satisfied, I baked these and the previous manual adjustments into a “GH5 > P4K” LUT.

Take away

Two very good cameras. We’ve arrived at a point where the image quality differences from camera to camera – at least for a reasonably well exposed scene – are pretty minor, and easily corrected. This is good news, as it allows an image maker to choose a camera based on a whole host of other factors that more materially impact how they work. For myself, I’m now comfortable using both cameras on a typical shoot knowing that a basic match is a LUT away. 4K PNG’s can be downloaded here.

---

*I know that this combination is not technically as accurate as dedicated hardware, but in the past I have had the chance to compare its results to that of a Sekonics C700, and with only minor adjustment it was able to consistently match quite well under a variety of lighting conditions.

**The GH5 doesn’t allow as precise WB adjustments as the Pocket 4K, so it was set at 4800k. It also doesn’t use a traditional tint selector, so I nudged its colorwheel the appropriate amount towards magenta.