Redrocket or CPU Debayering

[Kemalettin Sert]

im doing transcode for some company they want 2K DPX files all the footage shot on Epic and post production company sent email saying this

we would suggest a low to medium OLPF compensation and we wouldnt suggest to use Redrocket for debayering as there several disadvantages compared to the CPU based debayering

can anyone explain this to me please?

 

[Bob Gundu]

This can be argued but there is a difference in quality when using the software debayer. There are a couple of options not available when using the Rocket to debayer, like DRX, Sharpen, Denoise (I think). In any case, the hardware debayer seems sharper and the software smoother. If I'm planning on doing upscaling in my NLE I prefer software debayer and apply sharpening based on the amount of scale.

 

[Mike 'Fireman' Ross]

I have heard it said... that the software debayering used when no RR is installed is... sub optimal? Not *entirely* as good as it could be yet?

I've certainly heard from more than one person that material debayered in hardware by RR looks subtly different - better, sharper - than the same material done entirely in software.

Anyone else care to comment?

Mike

 

[Tehben Dean]

Yep there is a difference. In my experience for stills, software is better. I doubt you will be able to tell for video and the RR will probably appear sharper right off the batt.
If you zoom in and pixel peep the RR leaves and interesting mosaic.
Here is an example of each: http://www.reduser.net/forum/showthread.php?73869-Strange-RR-sharpening&highlight=

 

[Bob Gundu]

BTW, I believe Graeme prefers the software debayer.

 

[Robino_J]

Good thing I have seen this thread since I have a Rocket on order and this makes me rethink.

I have some questions about the RR debayering.

For my current workflow, when going down to 1080p from 4K I debayer using 1/2 Premium or FULL. These two debayering settings give a nice smooth image FULL being the smoothest of course with no weird sharpening and noise is VERY minimal. After some testing I found that debayering with "1/2 Good" was sharper and VERY noisy. I don't like that look at all and stay away from this debayering setting unless it's for dailes etc..stuff I would not use for finishing.

After looking at the screenshots posted by Theben Deanhttp://www.reduser.net/forum/member.php?21976-Tehben-Dean, the RR debayered images look like the result I get when using 1/2 Good. They are very noisy and not pleasing to my eye. Of course that's a personal taste and I'm talking for myself here.

Is this how I can expect the RR debayered footage to look like??

I can edit R3Ds with Premiere Pro at 1/4 already so I don't see any advantages of using the RR for either editing or final render if it looks all noisy and too sharp.

 

[Kemalettin Sert]

thank you guys for the answers.but i still dont get software means debayering without RR right? wouldnt take days debayering footage without RR?

 

[Thomas Koch]

It will take longer without the rocket, but that's a trade off for better. Charge by the hour.

 

[Les Dittert]

I predict the software debayer will improve late summer to keep parity with 2.5k outputs from a different cheapo camera.

 

[Michael Notar]

i like a little grit to the image, helps sharpening bit on to something. my opinion of course, coming from a photography world tho. unless that is generally not a good idea for video/film. i am used to sharpening the capture file a little and then at final output. is this not how video is usually done...

 

[Mike Lary]

thank you guys for the answers.but i still dont get software means debayering without RR right? wouldnt take days debayering footage without RR?

Yes, software means without RR, and yes it will take much longer. Graeme has talked about the difference between the two. He says there's a sharpness difference (RR-assist being sharper).

 

[Axel Mertes]

thank you guys for the answers.but i still dont get software means debayering without RR right? wouldnt take days debayering footage without RR?

Thats relative. On a single computer - yes, it will probably take longer. However, the very latest dual i7 machines can beat a single ROCKET. Its another story when you use multiple ROCKETs. Who really has that and can afford it?

We have written transcoding software that runs as network rendering and can utilize 100% CPU cores of all networked machines. We run actually faster than a ROCKET can, using the IMHO better software deBayer from Graeme.
The software core is also being using in our FinalDCP encoding software, its the same RenderNode.

Our workflow is as follows:

1. Shoot 4K
2. Edit using RED files natively or use create proxy files and edit them.
3. Based on EDL or XML perform trimmed clip transcoding into preferably CineForm or DPC (aka DPX-C).
4. Post production, VFX, Grading etc. using the CineForm/DPC files in full resolution.
5. Render Master (typically again a CineForm file).
6. Create the distriution master, e.g. DCP or BluRay or MXF OP1a etc.

Fast, streamline, proven to work well.

Exchange point 1 for any other camera you might use. We just play with F65 and it works too. The same probably with a Penelope Delta etc.

Regards,
Axel