Getting RAW Video

[leegong]

If the sensor outputs 10bits in video mode , for 10bits 3x3 binning , the maximum value after binning is 0x3FE * 3 * 3 = 0x23EE ,
if so , it's impossible to explain why there is highlight overflow in 14bits :
viewtopic.php?f=2&t=2316&start=100#p17528
viewtopic.php?f=2&t=2316&start=100#p17534

i tend to think that single pixel ADC is 12bits , then 9 pixels are 3x3 binned (added) to 14bits , so highlight overflow may occur .

[leegong]

Another possibility is that there is no binning at all , only 1/9 of total numbers of pixels without pixels binning
are sent out from sensor , meaning the worst video solution .

[a1ex]

I can check the binning pattern if you send me a full-resolution image and a LiveView image of a resolution chart (both raw).

The only caveat - there must be absolutely no motion between the two test images (more than pixel-perfect alignment). Shutter motion might be a problem.

On Canons, I've got the test images with a script that takes a LiveView frame, then zooms in, then takes a LiveView frame of the zoomed display, which is 1:1 crop, similar to a full raw image.

If you don't have a way to implement a similar script, using mirror lockup (and maybe a remote release as well) might help.

[astronomer]

OK, after talking to someone who works with large format Sony sensor, I got a much clear understanding.
Basically there's no vertical binning on these devices, only H Bin. Thus this fits the expectation of vertical color moire.

The output is 10bit result from H Binning and it cannot increase bit depth. End of story.

[Diegoureva12]

what does it means? there is no way to make it work?

[max]

[matoomba]

[nikkor]

So this is the reason for lineskippingmoiree and bad highlight rolloff? Anyway maybe I'm getting this wrong but, if that's the way nikon produces the video files it's still useful because after all, that's what you have been recording anyway. 2.2K 10bit raw would be great.

Oh and this might be total nonsense so please excuse me if I'm saying something stupid, but what if you skip the lower bits and then sum up, I mean who gives a damn about the lower bits

[astronomer]

Yes, now the problem is how to save this 10bit raw stream to sd card. Because the way of packaging it into 14bit for transmission, the downstream has to shift data right by 4 bit, which is not possible by the FR80 due to its very low CPU clock rate. Also there's problem not able to liveview the data. Currently this is where we stuck.

Photo mode is not possible due to the time it takes to read out an entire frame.

Anyway, I think the IMX071/094/128 etc are designed primarily as a still sensor optimized for image quality, not video. There's not a lot of circuit for video binning on chip like the small format or Exmor RS which has an entire DSP stacked beneath the sensor array.

[hamed]

Still hope to see the raw video hack!
agree with that 12bit is the best, but 10bit raw also is way better than the highly compressed 8bit video.

dear Nikonhacker guys! you have worked very hard on this and we appreciate your work and its not important if the raw video is not 12bit
you are AWESOME!