4k downscale 2k 3k 1k in 4:4:4:4 ?

[Madjid El ayari]

......................

 

[D Fuller]

Is it technically possible to have a firmwire wich will downscale the footage from 5k to 2k ,or 3k etc in 4:4:4 red code in the scarlet process ?

No. Redcode is Raw. 4:4:4 is RGB. So what you are asking is not possible.

 

[Matt Johnston]

so it makes 4 bit per chanel rgba , isn t it ?

No, it is RAW. Not RGB.

 

[Erik Franzén]

If you are asking: "can Scarlet-X technically convert to RGB, put it in a 4:4:4 container, and downscale to any of its many available resolution formats (1K, 2K, 3K etc.)" then the answer is "no" with the addendum being "it lacks the necessary processing power to perform the required number of operations in realtime."

If you are asking if Scarlet-X technically could scale .R3D files, then the answer is again "no". That isn't how RAW works.

I am confused by your question. What exactly would any of this accomplish? You can achieve a superior debayer on a workstation rather than in-camera, so that's not it. This would take more processing power, so additional frame rates can't be it. Keeping DOF consistent by scaling instead of cropping perhaps? But still, it seems daft since the required extra number of cycles would hammer performance and remove extra frame rate options...

Maybe I just don't get it...

 

[Gunleik Groven]

Well... You could allways record the SDI tap...

 

[ChrisLyon]

Once upon a time there was a DSMC module that would transcode footage to a set of sizes and color spaces, but I can't remember if that got cancelled, delayed, or what. But Scarlet cannot do it in camera. And you wouldn't really want to... kind of defeating the purpose of shooting high resolution and RAW - even for things that will get final mastering at a lower res. But since you're asking, tapping the SDI feed would be the fastest option to date until (and if) a RED conversion module is produced.

 

[Tom.Wong]

444 is just a container where color information as a linear broadcast term. a lot of people confuse term, generally just thinking it's the maximum amount of color information possible, period.

RAW is non linear sensor data. meaning the information hasnt been combined together yet, the image has not been de mosaiced yet. a gamma curve has to be applied, the the de mosaic determines what pixel is what color. there is no such thing as 444 with RAW, but RAW can become 444. 444 is just a container, with still a certain number of values for color. It's still scaled, and generally comes from broadcast use.

so to answer your question, REDCODE itself is not 444, it's RAW. when you process the footage via software, it can become 444, or 422, or 420, or whatever you so decide, depending on what container you put it in. now the amount of color and how much you can manipulate once it hits a linear container is all about what the sensor actually captures, and the de mosaic algorithm used. but the only processing done in camera is encoding to RAW. it does not do any kind of linear format other than what you see coming out of the hdsdi or hdmi.


as for the apple documentation, again it's more of 444 generically used as "the most color you can possibly have in a digital medium" but it will still require the RAW to become a linear format before you can even hit that point.



edit: and to add to it, additional processing like making a linear container would have broken what RED has been trying to do with the camera. more processing means more parts, more parts means a bigger camera, more heat, more bits and components that can go wrong. even with current generation and next gen competing cameras, look at how they are accomplishing raw and a linear container. RED is the only company able to do internal recording as raw. every other camera system that does RAW right now or soon to come needs a external unit to do that kind of processing. cost would increase too. it would be much more cost effect as gunleik sad to just get it off the hd sdi since the camera is capable of doing a live de mosaic and scaling on just the uncompressed tap.

 

[Tom.Wong]

Exactly . THx

Can the scarlet could used the all sensor to get 2k 4:4:4 in redcode or 4:2:2 apple prores, dnxhd,.......... with a firmwire ?

ok i give up.

 

[Gunleik Groven]

Hi... slowly

The Scaret/Epic does a 1080 4444 downscale in realtime out the SDI

Which can be recorded to prores or DNx through an external deck/recorder


There's no way RED will allow for direct RGB recording as the main option, though....

OTOH a proxy-module has been discussed a lot. many want a 1080 prores/DNx proxy module.
But RED hasn't really confirrmed that. They have wanted to give us a h.264 module for dailies only

What you ask for is achievable through a KiProMini or similar.

 

[Matt Johnston]

Sorry I think i wasn t really clear : i just think that if it would have been possible to get 2k from the full processor like the 5k ( full res ), not a crop from the 5k to get some 2k , that would have been useful .

Yes, many think that it would be useful. Which is why all the talk of external recorders like the Samurai or Pix.

It is highly unlikely that would ever happen in firmware, as it requires lots of processing.

 

[Gunleik Groven]

Sorry I think i wasn t really clear : i just think that if it would have been possible to get 2k from the full sensor like the 5k ( full res ), not a crop from the 5k to get some 2k , that would have been useful .

I am Wondering if this is possible , with a new firmwire ?

Simple answer:
Nope
That is not possible

The processing IS there, as that is what gives the SDI signal, but RED does not want us to work that way...

Happy?

 

[Andreas Mendritzki]

And in addition to the external recorder option - redrocket transcodes can give you pretty much whatever the heck you want, in real-time.

 

[Luis Caffesse]

Hi... slowly

The Scaret/Epic does a 1080 4444 downscale in realtime out the SDI

Not to complicate this thread more - but just to clarify this point, the SDI output is actually 10bit 4:2:2, not 4:4:4 - which would need dual link.
(a limitation of the SDI protocol, not the Scarlet itself)

That doesn't change anything else Gunleik was saying though - what the OP is asking for is not possible.
Like everyone else has explained it would require additional processing power, which means additional components, which means a larger camera and (I assume) a higher price tag.

 

[D Fuller]

Raw Red code is 16 bit ? so it makes 4 bit per chanel rgba , isn t it ?

Anyway im sure you have understand my real question ! Is it technically possible to have a firmwire wich will downscale the footage from 5k to 2k ,or 3k etc in red code in the scarlet process ?

http://documentation.apple.com/en/finalcutpro/professionalformatsandworkflows/index.html#chapter=11%26section=1%26tasks=true :The REDCODE color sample ratios are 4:4:4 RGB

The Apple document is not quite accurate. (Useful for practical purposes in FCP7, but not exactly what's going on.) Redcode is raw Bayer-pattern sensor data. It produces 4:4:4 color when debayered and rendered to RGB, but Redcode itself is compressed raw bayer matrix sensor data.

16 bit does not mean 4 bits per channel rgba, it means 16 bits per channel RGB (for a total of 48 bits). There is no alpha-channel.

Sorry but , Everyone one know that raw is a container !!!

And In fact my question is : If it is thecnically possible s to get some 2k footage with redcode or something else out of the scarlet without cropping the sensor , using the full sensor to get some 2k footage, with a firmwire ?

Here is a source of your confusion. Raw is not a container. "Raw" refers to recording the raw sensor data. Redcode is a codec that encodes the raw sensor data with with a variable amount of compression. If you scale the image, you would have to demosaic it to RGB first, so it would no longer be raw, and Redcode would no longer be a useable codec. This is why what you ask is not possible.

As others have said, you can record the 1080p SDI output if you like, using any of several devices on the market, but not using Redcode.

 

[Marden Blake]

Ok D Fuller. thx for the clariifaction .

But what about getting 2k from all the sensor ( full res )in redcode ( not scalling then )? Does it require more power , processings like Matt Jonhston said ?

I would have tought that 5k redcode(full sensor ) require more processing than 2k redcode ( full sensor )?

Also i was thinking that the red scarlet are enough powerful build to do a lot more things that what she already does( soon hdrx wich require a lot of power and a good program ) , and i was thinking a lot more for the epic .

Currently not possible due to processing power from what I heard. Lots of things are possible through firmware updates, but for now appreciate and deal with what RED has done with the Scarlet in it's current condition. Any future updates is just an added bonus.

 

[Philippe Van Horebeek]

Madjid,

What you are asking is getting compressed raw bayer sensor information (=redcode) in 2K ??
Well that would basically require another sensor, a 2K sensor !

Otherwise you can't do anything other than to debayer the data (no raw data anymore) and use algorithms to rescale the resulting image.

 

[Erik Franzén]

Ok D Fuller. thx for the clariifaction .

But what about getting 2k from all the sensor ( full res )in redcode ( not scalling then )? Does it require more power , processings like Matt Jonhston said ?

I would have tought that 5k redcode(full sensor ) require more processing than 2k redcode ( full sensor )?

Also i was thinking that the red scarlet are enough powerful build to do a lot more things that what she already does( soon hdrx wich require a lot of power and a good program ) , and i was thinking a lot more for the epic .

Let me explain this real slowly...
You cannot -- CANNOT -- scale RAW bayer data -- by that time it is in RGB land. That is why we have cropping -- it doesn't take any additional resources since there is no additional processing done to the RAW image, nor is there any conversion from RAW. Getting the full sensor at 2K would require some form of processing, hence not RAW anymore. Also, how would we get "full res" (I assume you mean the full sensor really) without scaling? 5K requires more data throughput than 2K, sure, but you're not comparing apples to apples -- 2K scaled would require more processing power than 5K unprocessed; yes. 2K cropped naturally does not require more cycles.

Anyone knowledgeable is free to correct me on this since I ain't sure as hell the RED expert around here.

 

[Luis Caffesse]

Canon manage with the c300 wich is a 4k sensor to scale the data to 2k mpg2 4:2:2

As Erik said:

Let me explain this real slowly...
You cannot -- CANNOT -- scale RAW bayer data -- by that time it is in RGB land.

The Canon may use a 4K sensor, but it must convert the RAW sensor data (which we never have direct access to in the C300, by the way) into RGB in order to scale it down.
The Canon is scaling a 4K sensor down to 1080p 4:2:2 8bit MPG based footage.

You could get a similar output, in fact a BETTER output, from the Scarlet or Epic by recording the SDI output at the full 10bit 4:2:2 signal (once the SDI output is not bordered by black bars).

But what Erik was saying is that it is impossible to take a 5K sensor and scale the RAW sensor data and somehow wind up with 2K Redcode Raw.
The scaling process would require converting the RAW sensor data into RGB.

 

[Sulekh Suman]

Ok D Fuller. thx for the clariifaction .

But what about getting 2k from all the sensor ( full res )in redcode ( not scalling then )? Does it require more power , processings like Matt Jonhston said ?

I would have tought that 5k redcode(full sensor ) require more processing than 2k redcode ( full sensor )?

Also i was thinking that the red scarlet are enough powerful build to do a lot more things that what she already does( soon hdrx wich require a lot of power and a good program ) , and i was thinking a lot more for the epic .

If I understand him right, he wants to know why it isn't possible to change the resolution(3K, 2K) without “windowing,” crop factor or change of effective focal length.

 

[Alexander Waldman]

Hopefully I'll be able to clarify...

Yes, the C300 can do that. Yes, the Scarlet and Epic have significantly more processing power than the C300, so in terms of pure computing power, the Scarlet or Epic **technically** could record scaled sensor data in mpg2 4:2:2.

HOWEVER, recording in mpg2 4:2:2 defeats the entire purpose of shooting RED. This is for (at least) two reasons: First, recording scaled images directly from the sensor defeats the purpose of having a 4k or a 5k sensor. Yes, the C300 does it, but doing this has many disadvantages. Mainly, oversampling (recording in 5k or 4k and then downconverting in post) produces a much nicer image that is less susceptible to moire and other issues with recording at lower resolution. Second, all RED cameras record REDCODE RAW. REDCODE is not a container, merely a compression scheme that compresses the RAW sensor data. It is not possible to even view REDCODE RAW images unless debayering and processing takes place. As a result, the strength of REDCODE RAW lies in post production. RED cameras do have more processing power than a C300, but this is simply used to record the RAW data coming off of the sensor at a high bitrate (~50MB/s), not to process it. Using a computer or system that has SIGNIFICANTLY more processing power than the camera to do all of the image processing in post (debayering, colorspace assignment, transcoding, etc.) allows you to get a higher quality image than one that is achievable in [any] camera, because of the amount of flexibility it affords. It's like shooting a digital negative.

The C300 does all of the debayering, colorspace assignment, and transcoding in-camera, and at a much lower bitrate than the Scarlet (~50mbit/s or around 6MB/s), but at the expense of high resolution and the flexibility of RAW in post production.

If you want to record a 1080p image from a 4k sensor directly in camera, I suggest you get a C300.

I knew that with sdi or hdmi recorder it is possible !

It is not possible with an SDI or HDMI recorder to record RAW. Only compressed video in RGB space.

There is absolutely no way to record scaled REDCODE RAW. This is simply a technological impossibility. It's like saying "I want to build a miniature Eiffel Tower using full-sized parts." It just doesn't make any sense.

It is technically possible for the Scarlet to record a mpg2 4:2:2 scaled 2k image, but RED has stated over and over again (and for good reason) that this will not be implemented in camera, as it defeats the entire purpose of shooting RED--namely, REDCODE RAW.

Hope this helps!