a7s 4K Downscale Test and Request for Input

[tonydtv]

Apologies if this is old news, but the "GH4/a7s 4K downscale to 1080p" discussion got a bit too academic for me to follow, so if there has been a conclusion on this, it hasn't been in easily searchable laymen's terms.

So. I figured I'd give it a shot.

I picked up a 4K scaler off newegg.com for $30 (Coboc HM-SPL-4KUP-1X2 2 Ports 1 in 2 out 4K2K HDMI Scaler splitter, http://www.newegg.com/Product/Product.aspx?Item=N82E16812117500) and borrowed a buddy's Atomos Ninja 2. Did three shots:

1) a7s 4K 24p -> Coboc 4K Scaler, downconvert to 1080 24p -> Ninja 2, ProRes 422

2) a7s 1080 24p -> Coboc 4k Scaler, set to bypass -> Ninja 2, ProRes 422

3) a7s 1080 24p, internal XAVC S

The clip below shows the three shots, then again with the saturation jacked up to 100, then a third time zoomed 200%.

 

To me it seems clear that the 4K downconvert is cleaner and sharper, esp in the shadows. But, if anyone else has any input on what's going on here or if I am missing something or misinterpreting something, please chime in.

Like what's up with the red noise in the non-4K shots?

I'll be filming more footage this weekend (assuming I can make this setup portable) and will post.

[hmcindie]

Contrast levels are different = shadows look different.

The output of the slog through HDMI is different to internal encoding so you must set the levels in your editor to same before comparing (16-235). It's hard to gauge sharpness as everything is blurry.

[tonydtv]

True, but there's still a noticeable difference between the 4K out and 1080p out, specifically that sea of red that appears over the 1080p footage. And assuming I fix the blacks on the XAVCS footage, it'll still look just as blocky in the shadows.

And I misspoke saying its sharper, I meant the noise looks sharper (if that makes sense), or better yet, the noise looks less blocky and compressed, especially in the shadows. Like the difference between watching a movie on bluray vs HD cable compression.

Seems like I'll have more play in color correction with the 4k downconvert (vs the 1080p out or the internal 1080p) before the image breaks down into a JPG from 2002.

Again, I'll have to test more, but for now, it looks promising. Especially considering it's a $30 box that gave me this much more latitude.

[dhessel]

It is kind of a given that down scaling UHD to HD in post is going to give better results than HD straight from the camera on the A7s even though that is exactly what the camera is doing internally anyway. The reason for this is simple, the A7s needs to read the data off the sensor, de-bayer it, downscale it, and compress it in real time at up to 60 fps were down scaling in post the footage only needs to be scaled and it doesn't even have to be real time during the rendering process. This means that better more computationally expensive scaling operations can be used compared to those used in camera.

Here lies another advantage of cameras that can shoot raw, all of these operations can be done on more powerful machines at higher quality when done on a computer in post rather than in camera. 

[tonydtv]

Just to clarify, the processing here is happening while recording, not in post. That's what the $30 4K scaler is for

What I'm surprised by more than the difference between the 4K output scaled and the XAVC S internal is the difference between the 4K output scaled and the 1080p output. Colors seem cleaner and less saturated (RED!).

And, sure, a RAW camera would be that much better, but I think most of us doing indie work ride that line of quality vs the lint in our pockets. For 30 bucks, this seems worth it to get a better image.