Odyssey7Q + FS700 - 4K Modes

brazzam said:
I very much agree. Dan, how are the engineers planning to obtain these QuadHD recordings? By simply cropping the 4K image down to 3840 or are they going to resample the 4K signal?
Click to expand...

Hi Alan (and everyone contributing to this thread)-
We (Convergent Design), very much appreciate the discussion and feedback on this forum. Based upon the discussion, we will implement 4K first, followed quickly by QHD. QHD is derived by cropping pixels from the 4K (not rescaling).

Just for clarification, on the 2K/HD recording, we plan to use 4K RAW to create 2K/HD for frame rates up to 60fps. This should yield beautiful images, as the 2K/HD is basically supersampled from 4K. 2K/HD at 120/240 fps will be derived from the 2K RAW, which has lower resolution, since the pixels our subsampled from the 4K sensor (line and pixel skipping).

Mike Schell
Convergent Design
 
Hi Mike,

Thanks for coming in and for posting. So let me take a stab at boiling this down:

The Odyssey can record either video, or raw signals. The video that it records can either come from a regular SDI or HDMI input, or it can be sourced from the raw and converted to video internally in the 7Q.

There are two potential raw readouts from the FS700. There's 4K raw, which is a true, uncompromised, raw readout of the entire sensor. Then there's 2K raw, which is a compromised, pixel-skipped version which maintains the full field of view, but compromises resolution and image fidelity, introducing aliasing and potential moire.

The Odyssey has the ability to see both those streams, but cannot record 4K raw from the Sony, because of the limitations of a licensing agreement with Sony. So the Odyssey can record FS700 4K as compressed 10-bit video, but not as raw. (on the other hand, if using a Canon C500, Canon's licensing agreement is more favorable and the Odyssey7Q can indeed record 4K raw from the Canon). But 4K raw is not possible from the Sony, due to the licensing restrictions from Sony.

The Odyssey can record the FS700's 2K as raw, but only from the compromised 2k raw stream that the FS700 transmits.

So, for an FS700 user, as a RAW recorder, the Odyssey can't record the pure 4K raw because Sony won't allow it, and the raw that it can record (the 2K) is pre-compromised by Sony. However, it can record that pre-compromised raw from Sony at up to 240 frames per second. It should provide higher quality than the AVCHD-compressed 240fps, but it still won't be ideal because Sony has compromised the feed before the Odyssey even gets to see it.

On the other hand, the Odyssey can record great high-quality video in 4K, 2K, and HD. The Odyssey can use the 4K source to debayer and record a pristine 4K image, at up to 30fps. And it can internally downconvert and scale down that camera-original 4K raw into a very high quality 2K, or into a very high quality HD 1080p signal, at up to 60fps and 10-bit. So all recordings (4K, 2K, and 1080p) start from the original, high-quality raw signal, and are de-bayered and scaled to the appropriate size, resulting in high resolution and great image fidelity.

In fact, I'd dare say the downconverted 4K->2K recorded to ProRes/DNxHD in the Odyssey7Q will almost certainly look better and be better resolution and have better image fidelity than the 2K raw will. The 2K raw may have a couple more bits of color depth, but the downconverted footage should be much more robust in terms of resolution and avoidance of aliasing.

If I have all this right, then I think I can further boil it down to:
For an FS700 user, you're not likely going to bother recording raw very much. You can't get 4K raw, and the 2K raw is compromised. Simply put, Sony won't give Convergent the good stuff, they've reserved that for their own recorder.
As a ProRes/DNxHD recorder, the Odyssey 7Q should be superb. It should be flat-out superb, and offer a substantial image quality boost over the onboard AVCHD recording.
For the highest-quality slow-mo, the Odyssey 7Q should offer a significant bump up in quality since it will be bypassing the AVCHD compression.
And, besides all this talk of compression and recording, let's not forget the Odyssey7Q is also a first-rate monitor... :thumbsup:
 
Not to sound greedy or anything (well.. Maybe I am), but since the 4:2:2 10 bit is converted from the 4K RAW, is there any reason why we can't get 4:4:4 over the 4:2:2.

Would it make at much of a difference anyways btw 4:2:2 and 4:4:4?
 
gene_can_sing said:
Not to sound greedy or anything (well.. Maybe I am), but since the 4:2:2 10 bit is converted from the 4K RAW, is there any reason why we can't get 4:4:4 over the 4:2:2.
Click to expand...
In 1080p or 2K, there should be enough color information to create a proper 4:4:4, so I guess it depends on the codecs they're recording to. If the codec allows it, then it would be reasonable to expect that we should be able to ask for it...

Would it make at much of a difference anyways btw 4:2:2 and 4:4:4?
Click to expand...
Yes and no. Yes, in that 4:4:4 is literally twice the color of 4:2:2, which in itself is literally twice the color of 4:2:0 (as used by the native AVCHD codec). In practice, it's not that substantial a difference to the naked eye. The jump up to 4:2:2 does yield better color fidelity and better keying, but most don't really notice it unless you have highly contrasting colors in the same area, such as a diagonal red stripe across a blue background; in that case the 4:2:2 would render substantially sharper than the 4:2:0, and, accordingly, the 4:4:4 would render sharper than the 4:2:2 would.

It makes for better keying and compositing, that's for sure.
 
Regarding 2K RAW: pixel and line skipping occur at 120FPS and up. This is always the case, no matter what. Sony is not compromising the signal- that's just how the camera works at those speeds. 2K RAW at up to 60FPS could use a super-sampled 4K RAW Bayer array (which should actually reduce aliasing). Again, the C300 averages (low pass filters) two green Bayer grids with excellent results; it should be possible to do the same for red and blue (maintaining the Bayer array structure), whereby the final high-quality deBayer can take place in post. HFR will always have artifacts due to the line+pixel skipping in those modes (http://provideocoalition.com/awilt/story/high_speed_and_low_light_with_the_nex-fs700/). If Sony is providing 2K RAW at < 120FPS via supersampled 4K RAW, that could actually look OK if processed with a high-quality deBayer algorithm (Adobe ACR, etc.).
 
i might be the few who want the bump in the quality of my 240 and even 480 frame rates , and any chance 480 1080p for the Q7. something other that avchd ? this is why I chose the fs700 . getting good images now but a step up in these modes would be great!!!!
 
Hi Barry-
You have summed it up extremely well. We actually have 2K RAW @ 240fps working very well and today our engineers demonstrated 4K Raw debayered, scaled and displayed on our 1280x800 OLED monitor. It looked great.

We will continue to post progress updates, as appropriate.

Mike Schell
Convergent Design
 
gene_can_sing said:
Not to sound greedy or anything (well.. Maybe I am), but since the 4:2:2 10 bit is converted from the 4K RAW, is there any reason why we can't get 4:4:4 over the 4:2:2.

Would it make at much of a difference anyways btw 4:2:2 and 4:4:4?
Click to expand...

We are definitely planning to support 2k/HD 4:4:4 compressed recording, supersampled from 4K RAW, up to 60fps (12-bit Linear data). This will come after 2K/HD and 4K compressed 4:2:2 support, but it's on the roadmap.

Mike Schell
Convergent Design
 
Mike Schell said:
We are definitely planning to support 2k/HD 4:4:4 compressed recording, supersampled from 4K RAW, up to 60fps (12-bit Linear data). This will come after 2K/HD and 4K compressed 4:2:2 support, but it's on the roadmap.

Mike Schell
Convergent Design
Click to expand...

Yowza, I would love 2K 444 out of the FS700. Can't wait to get the 7Q.
 
@ Mike or Dan,

This is a little off topic, but with the 7Q, will / or can the FS700 still record to the internal SDHC card simultaneously? Would be nice just as a bit of insurance.

As for 4:4:4, that's is just amazing... just amazing.....
 
Mike Schell said:
We are definitely planning to support 2k/HD 4:4:4 compressed recording, supersampled from 4K RAW, up to 60fps (12-bit Linear data). This will come after 2K/HD and 4K compressed 4:2:2 support, but it's on the roadmap.

Mike Schell
Convergent Design
Click to expand...

This is really great news. I would love 2K 4:4:4. Any way of getting 4k 4:4:4 "Popular Codec " since we can't get RAW? That would be incredible for keying work!
 
4k could never be 4:4:4. At the most it would be 4:2:0. It is only through scaling down to 2k or 1080p that the potential of 4:4:4 becomes possible.
 
vcfilms said:
Oh, I see. So the compressed 4k signal would be 4:2:0?
Click to expand...

The meanings behind 444 and 422 are a little muddied with bayer sensors.

There are two issues here, one is the recording codec. If that codec is 422 then the two colour channels will be sampled at half of the luma, no matter what the source. If the codec is 444 and you feed it 422, the colour is still half, it's still 422, you don't get an upgrade.

Lets assume that the popular codec has a 444 mode.

The bayer sensor reconstructs colours using the neighbouring pixels, so technically it is producing a 444 image because each pixel *is* unique. However many say, because pixels are reconstructed based on neighbours that it isn't a true colour pixel. You could argue each way.

So you could take the 4K RAW and put it into a 444 container and indeed depending on the quality of the debayer you could well see a visible difference between a 444 and 422 recording.

But the benefit of supersampling the 4K RAW down to 2K is that you're getting *real* quality colour information for each pixel and that would benefit from 444 recording.

The *only* benefit RAW recording would get you is that you could choose the debayer process in post and use a method that works better for your scene - low light, high con etc,. There are many debayering options and a post debayer is usually better than a real time one. But at this point you're splitting hairs.

2K downsampled from 4K out of the O7 is perfect for 99% of my uses. I'm very happy to hear Mike confirm that and can't wait to get my hands on one :)

cheers
Paul
 
Just read a bunch of papers on various deBayer algorithms- the higher quality versions can indeed reconstruct almost full resolution for every pixel (ACR is the best I have seen). Not clear how much CPU is available for these embedded systems, but one of the faster versions used about 80 arithmetic ops per pixel (still with a very high quality result). 4K 444 is possible unless there is some other restriction.
 
Thanks Paul for the detailed info. Very helpful. Let's see what CD says about this possibility on the 7Q. It would be nice to have the possibility of 4K 444 (if of course it's possible) for greenscreen work. I'm so excited for this device!
 
vcfilms said:
Thanks Paul for the detailed info. Very helpful. Let's see what CD says about this possibility on the 7Q. It would be nice to have the possibility of 4K 444 (if of course it's possible) for greenscreen work. I'm so excited for this device!
Click to expand...
Dan already answered this question - i asked the same in another thread.

Dear Stefan,

Thanks for posting.

We have 4K photosites to deal with. 4096 x 2160, each in 12-Bit Linear Raw.

One cannot obtain real 4K in 4:4:4 from this source. To do so, requires 6K or 8K photosites.

But, we can obtain 4K in 4:2:2.

The destination codec, in 4:2:2 is 10-Bits, thus this will be very high quality 10-Bits, obtained from a true 12-Bit Linear Source.

I hope this helps.

Respectfully,


Dan Keaton
Convergent Design
Click to expand...
 
Ok, thanks for the info Stefan. Well, at least it's 10-bit 4:2:2 :) Still better than what I'm getting now out of my FS700 at 1080p ;)
 
You must log in or register to reply here.
Back
Top