In this corner, we have the Sony F3--episodic TV and corporate/commercial workhorse, best known for its use on season 1 of Key and Peele and the feature film Safety Not Guaranteed. Super 35 2.4K sensor that downscales to 1080p in-camera. 10-bit output up to 4:4:4 RGB.
In this corner, the Sony A7 III--a modern hit, barely out and already one of the most popular mirrorless cameras around. Full frame 24 megapixel sensor that downscales to 4K from both the full sensor and from the APS-C crop. 8-bit 4:20 internally, with the option for 4:2:2 externally.
The Test
I generally light my work, so I thought I'd test with a scenario I often face: talking head interviews. One big soft key, a soft backlight, something splashing the background, maybe a practical. Now, this is a hella lazy, sloppy lighting execution--the backlight is a little too far around, background splash not nearly strong or interesting enough, and there's a huge bit of blank wall amidst the clutter--but it does contain some useful information. More on that after the camera notes. Just keep in mind that it was a rush job and try not to be too critical.
Each camera used the same Leica R 35mm Summicron set to f/5.6. This was to keep color, contrast, and sharpness consistent, as not all my Rs match perfectly and I wanted any differences in the image to be down to the cameras. I did not adjust the tripod distance, instead opting to let precise frame matching go in favor of minimizing inconsistencies. I'll do another test that keeps framing identical.
Since I unfortunately couldn't get ahold of a color chart for this test, I have a few various colored objects in the scene to act as references. The sweater is a deep forest green, the hat is a pale pinkish-red, the UHaul logo is a distinct orange, and the furniture blanket is a nice primary blue. They don't line up with anything specific on a vectorscope, but I find the subtle differences in how these colors, along with my skin tone, are rendered is quite telling.
Yes, that is me in the video. I am not a model, or really comfortable on camera at all. My being in the hot seat was PURELY a matter of desperation, as my roommate was busy and I didn't have another day free to do the test. So you'll have to deal with my pale, blotchy face and greasy hair this time.
The F3 has two sets of clips: one taken internally in XDCAM, one externally in ProRes HQ (10-bit 4:2:2). I don't have access to a 4K recorder, so the A7III files were all recorded internally to the SD card.
If you have any other questions on my methodology please feel free to ask. Even with the studio conditions, it's very possible that something was overlooked.
What to Watch For
Personally, these are the elements I found most interesting.
-The colors of the various objects listed above, as well as their relative brightness. The A7, for example, makes the blue moving blanket much brighter.
-The big patch of blank wall behind me is a great place to watch for midtone noise, compression artifacts, and banding.
-The black portion of the moving blanket, back right corner by the boxes, and the shadow by my left ear under the hair are good spots to look at shadow noise/compression.
-Obviously, the face is there for skin tone reference. I'm pretty pale, and exposed a stop above key to keep noise down, as I generally do in a real interview. This pushed my skintones into the flattest part of the LOG curve, smoothing them out a bit, but in this scenario that's not entirely undesirable. Just something to keep in mind. Additionally, my eyes are a pretty deep shade of blue, with some green flecks. The cameras and color profiles render them with mixed levels of success.
-My hair, beard, and eyes are all good places to look at for fine detail rendering, noise, and moire/aliasing.
-The lightbulb and background splash were an attempt to look at highlight roll-off, but upon review it doesn't seem all that useful. The next test will use a stronger light doing a better scrape for DR/gradient, as well as a bulb with a filament to see how each camera renders hot highlight detail.
The Footage
These clips are straight out of camera/recorder, with no grading or edits. I could've at least trimmed them for brevity/convenience, but I...didn't want to. So. Sorry.
You'll find them at this link: https://drive.google.com/open?id=159ig4Dq7emsxIE6GWpwj8H9SX73XMSQU
Feel free to download them and play around. I can show you my grading results, but I find nothing helps more than massaging footage yourself.
Here are the settings for each clip:
F3
1. 1080p Cine1 F35Like ISO400 172.8* shutter f/5.6 (I don't know if F3 internal profiles suck or I just suck WITH them, but getting a good SOOC result with it still eludes me. Think I'll try the AbelCine stuff next.)
2. 1080p SLog1 SGamut ISO 800 172.8* shutter f/5.6 (ISO 800 is native in SLog, which is why it jumps here. Everything is otherwise identical.)
3. 1080p SLog1 SGamut ISO 1600 172.8* shutter f/5.6 (An attempt to test overexposure. In retrospect, I should've opened up to f/4 instead, but this still shows clipping behavior when cranking up ISO for exposure.)
4. 1080p SLog1 SGamut ISO 6400 172.8* shutter f/16 (High ISO noise test. The levels themselves are pretty good, so this is more a test of how much penalty you pay for cranking ISO to get by with smaller fixtures inside.)
A7III
1. 4K Cine4 SGamut3.cine ISO 400 1/50 f/5.6 Full Frame (with +15 or so saturation. My favorite out of camera profile so far.)
2. 1080p Cine4 SGamut3.cine ISO 400 1/50 f/5.6 Full Frame
3. 4K SLog2 SGamut3.cine ISO 800 1/50 f/5.6 Full Frame (I prefer SLog2 to 3 for banding and skin tones. Like F3, 800 is native for SLog)
4. 1080p SLog2 SGamut3.cine ISO 800 1/50 f/5.6 Full Frame
5. 4K SLog2 SGamut3.cine ISO 1600 1/50 f/5.6 Full Frame
6. 4K SLog2 SGamut3.cine ISO 800 1/50 f/5.6 Crop Mode
7. 1080p SLog2 SGamut3.cine ISO 800 1/50 f/5.6 Crop Mode
8. 4K SLog2 SGamut3.cine ISO 6400 1/50 f/5.6 Crop Mode
9. 4K SLog2 SGamut3.cine ISO 6400 1/50 f/5.6 Full Frame
Again, I'm happy to clarify any questions on my methodology. This was mainly to give me a nice broad spread of information on how my most used settings compare for quality.
Conclusions
I'm more curious what you guys think, so I won't say too much right away. My main take-away is that the F3 is nowhere near obsolete yet...in fact, it's my preferred camera for noise, color, skintones, gradation, and overall look. The A7 has the obvious advantages of full frame, strong internal recording, size/weight, and IBIS.
Next, I want to see how each camera does with fluorescent, LED, and daylight skintones, as well as mixed color temps, motion, compression/bitrate issues, DR, and resolution/detail. If you have any suggestions for future tests, or anything you'd like to specifically see compared, just let me know and I'll do my best to throw it in.
Thanks, and hope you enjoy!
