Review: JVC GY-LS300, Part 2: Image Quality and Deliverables

In the outdoor shots, both the highlight and the overexposure tests showed me that the LS300 acquits itself well in terms of latitude and image quality. The sensor did not flare out, there were no bad artifacts, and when underexposed, I was still able to lift up a very clean exposure for the face without any difficulty. In fact, I have to say that the JVC color science is a bit better than my GH4, which tends to skew yellows and reds in odd ways, leading the industry to develop specific color correction LUTs to fix the Panasonic image.

When I tested the JVC against my GH4 in extreme grading tests, the GH4 bested the JVC. But I have to admit that I’ve tweaked my GH4, which I’ve owned and used for over a year, to optimize its color performance. By contrast, I’ve been testing the JVC for only a couple weeks and it has considerable image-tweaking capabilities--none of which I used in this test. So it may do a lot better when the image parameters are adjusted by someone who knows their way around JVC’s image parameters and has logged as many hours with it as I have with my GH4. But keep in mind, this is an extreme test to see what compression artifacts the cameras have in the darks. You should not lift up the shadows this much with regular video.

Moreover, JVC has a Firmware 2.0 Update (Figure 3, below) that offers a LOG profile to better preserve the image before compression. Because this part of the review already had so many tests, and there are two videos and the HD video is already 12 minutes long, I opted to not include LOG in this part of the review. We’ll address this and other relevant elements of the Firmware 2.0 update in Part 3.

Figure 3. JVC’s Firmware 2.0 update for the LS300 offers a LOG profile, among other useful new features.

Sharpness vs. Aliasing

Another area of particular interest was how well the camera is able to interpolate all the various virtual sensor sizes down to 4K and HD resolutions. So I set up a camera chart (Figure 4, below), made it fill the frame with one good M43 zoom lens (Lumix 12-35mm/f2.8), and jiggled the camera a bit while I recorded 20 seconds of video at each virtual sensor size.

Figure 4. Yes, I made a very big test chart, to ensure that there was no moire in the chart itself, and so wide lenses could be tested too. Click the image to see it at full size.

You can see the results in the two videos embedded at the bottom of this page, one at true 4K if you have a 4K monitor, and another separate video recorded at HD. These two videos let you assess how well the camera scales the sensor at the various virtual sensor sizes. My conclusion: It varies.

There were some virtual sensor sizes that are markedly sharper/clearer than others. For instance, 89%, 83%, and 80% were dramatically clearer. 92% was notably softer than others. The sharper sizes demonstrated more moire and aliasing in the final video. You can see this for yourself in the 4K video.

I tested this by rendering a high-bitrate 4K video and watching it on both a high-end consumer 4K monitor- playing from my USB stick. Then I also played it back on Apple’s 24" iMac with the 5K “retina” screen (Figure 5, below). Seeing it play pixel-for-pixel and still have tons of space around it on the computer display was interesting. It really demonstrated to me how different VSMs are sharper than others.

Figure 5. Assessing all the 4K, pixel-for-pixel (no overscan or “image enhancement”) on a 5K iMac. Click the image to see it at full size.

Is “soft” bad? No. Within reason, it’s not. You may think you want the sharpest, most clinical image possible, but then you have an older face in front of your camera and the question is whether you want to accentuate every wrinkle and blemish in this woman’s face, or whether you want the viewer to just listen to her message. You realize that clinically sharp is not always the right tool to use. This is the reason I started looking into older glass. Because it has a “charm,” a “look”--and part of that distinctive look is not being hyper-sharp.

But you don’t want your image to be overly soft either. So selecting the specific virtual sensor size effectively becomes a creative choice that you can leverage to your subject matter. Interview? Softer VSM. Detailed product shot? Sharper VSM. Etc.

For you to assess this critical test, I have made two videos. One video is 4K, and all 4K-originated material is presented at 4K so you can assess it on a 4K monitor of your own. A second video is HD, with the critical moire/aliasing tests recorded at HD resolutions in the LS300. There are smaller sizes and more VSM choices when recording HD in the LS300, so a separate test was required to properly assess it.

HD Video:

4K Video:

In HD, 95%, 92%, 80%, 72%, 67%, and 54% were clearly sharper, and thus more susceptible to aliasing and moire. Super 35, 97%, 89%, 86%, 83%, 76%, 63%, and 59% were softer, with little-to-no moire or aliasing. The three smallest virtual sensor sizes--52%, 47%, and 43%--were particularly softer than the rest.