FS100 & Atomos Ninja Tests
I had the chance to try out the Atomos Ninja with the Sony NEX FS-100. It seems the jury is out as to whether adding the near $1000 device to the camera kit is worth the purchase or not. There are three areas I wanted to test with the FS100:
- Is recording to the less-compressed Prores 4:2:2 look better or worse than the AVCHD in terms of overall picture quality, judging:
- Compression artifacting
- Noise Level
- Is recording to the less-compressed Prores 4:2:2 improve results when post color grading,
- Is recording to the less-compressed Prores 4:2:2 improve the ability to chroma-key in post.
OBSTACLES OF THE NINJA/FS100 PAIRING:
I wanted to run down some of the obstacles when using the Ninja with the FS100. The Ninja only records 1080i 50fps or 60fps. The interlaced and high framerate does not remove any information when shooting in 23.98 or 29.97, but rather sort-of duplicates the information to fill up the other frames. This requires you to remove pulldown in post. Also the LANC control doesn’t work with the FS100 nor can it read the camera’s timecode.
- OVERALL PICTURE QUALITY. It seems there is no difference in compression artifacting, INCLUDING when there is high movement or quick pans…which is where you usually see this phenomena. Obviously you can’t see the result of movement in the photos below but believe me, I whipped the camera around all over and it’s amazing how well the AVCHD held up. I couldn’t see any difference at all whatsoever. Now there is a noticeable difference in the noise level, but surprisingly not in Prores’ favor. The noise amount appears to be the same, but the noise structure in the Prores version seem larger and more like video noise than the AVCHD which is smaller and more grain-like. At first I interpreted the footage incorrectly in Premiere and it appeared that the Ninja Prores had severe aliasing…but it turns out that was my own mistake and I see that there is no difference in the “stair-stepping” between the two codecs.
- I dropped equal “fast color corrector” settings on each clip, mainly pushing the contrast way past acceptable levels to really see what happens to each image. This SHOULD bring out the worst in any 8bit codec an this should be no different. As seen below in the images, there is no banding in the gradients on the walls nor sky (sorry my sensor was very dirty…I cleaned it right after this test!). If this was DLSR footage there would be horrible banding occuring so this is a great testament to the AVCHD coded and Prores alike…but unfortunately there is no additional benefit to Prores in this test as they both equally handle contrast equally.
- I did a simple chroma key on a greenscreen using something I think would pose issues…a poinsetta with red and green leaves. Even the foil wrap is green. I pulled it into Premiere and applied the EXACT same settings using ultrakey and both keyed perfectly fine with no difference at all. Even as I blow it up 500% and look at the edges and detail, I can see where the foil got keyed out in the shiny parts in the Prores file but it’s very very minuscule and could be easily adjusted using the settings in the ultrakeyer but I wanted to keep the setting exactly the same to show how close the two clips are as far as how easy it is to key.
I really really really REALLY wanted the Ninja, and more importantly the Prores less-uncompressed codec, to hold up better in terms of the criteria I outlined above. But in every way, the AVCHD held up much better if not the same. I will say, however, that the Prores seems to play nicer with Premiere, but that’s another topic for another time! All in all, I will probably be returning The Ninja. It may very well be great for other cameras or applications, but I don’t feel in my opinion that it improves footage from the FS100…in fact I believe it does the opposite.
CLICK HERE for the screengrab .tif files so you can see a non-compressed image. Warning: these files are very big. Each files is marked as “Shot A, B, C” and the codec is marked “AVCHD or PRORES.” There are also the two chromakeyed screengrabs comparing the two codecs.