8K on 12K sensor - How does it work?

Is there a good description or any tests showing how the 8K mode of the 12K sensor functions? ie is it skipping information, down sampling, etc? Does noise/aliasing/moiré increase?

I shoot almost exclusively at 120+ fps (Ursa Mini G2) which means it would be 8K mode for me on the newer sensors, unless I want 2.4:1 of course.

Cheers!

The readout in 8K 120 is much faster so it would seem to be skipping or binning but whatever it's doing, it's as clean as clean can be, no artifacts.

Following are some info:
https://patents.google.com/patent/US20190306472A1/en
other patents by the inventor:
https://patents.google.com/?inventor=Carsten+Buettner
and a forum topic of which there are several others:
viewtopic.php?t=117952

I have not found the exact math of the algorithms used, likely proprietary.
However, I do suspect that something of a binning concept is used. Although this gives good results for 8K, it does not work so well for 4K (IMO.) After shooting for some time with my URSA Mini 12K I suggested that the algorithm be improved as I personally feel the sensor is capable of even more.

It would be interesting to have someone compare results of the 12K with the new Cine 12K in 12, 8, and 4 modes. Has the the algorithm been improved? ...or is it just a sensor with larger photosights?

I would also like to see BMD use a 6K super16 (cut down of the 12K sensor) in a new pocket camera.

How does it work?
???
It works great... noticeable color improvement from the previous Bayer filter sensors incorporated by BMD. However, I still think the 8K and 4K modes can work even better.

Do note that there are some differences from using the full density (12K super35 and 6K super16) and 8K or 4K. Some color shifts are documented here:

That's a great test!
Interesting compression comparison and finding about the color shift.

The 12k sensor uses a 6x6 CFA. 8k uses a 4x4 cut and 4k uses a 2x2 cut. Pixels outside the cut are still used to derive the Braw pixel values for those inside.

Good Luck

Based on Howard’s summary, it seems clear that recording 12K will give the best results where that is feasible to deliver 8K or 4K.

Then 8K recording still looks like a very good option and may be frequently used if media and data wrangling are sufficient; delivery in 4K by downscaling in camera may produce very good detailed images.

Recording in 4K is still interesting but possibly only as a last resort or it’s dictated by high frame-rate or Super 16 lens selection.

The reason for using 8K is faster readout or 120 fps, actually 160 fps in 2.4:1.

I didn't see any problem with 8K mode at all. On the other hand, 4K while not bad, you could still see a small to moderate degradation.

8k is outstanding. It’s really hard to pick between 12k and 8k.


4k does not do as well in my view.

JB

I just happened to run across this video I shot a couple years ago in 8K-120. I had lost track of it.

I stopped using 120 fps at weddings, just too slow, has no purpose, 60 fps is more than slow enough for weddings.

This is presented in HDR. I think it probably looks okay in SDR. I usually color on the conservative side with HDR, and the down-conversion to SDR by YouTube is usually treated kindly.

These days I stick to 12K but 8K looks just fine to me echoing John's comment.

The color shift may be due to the fact that at 4K the cam is using 2 (non symmetrical) filter kernels from the diagram below. In 12K it uses all 6, and at 8K it's four. I believe the debayer has changed since release so this may not be accurate as of the latest firmware.

Good Luck

chrominance.jpg (394.23 KiB) Viewed 919 times