Welcome to our community

Be a part of something great, join today!

Register Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

RED TECH - REDCODE RAW

Jarred Land

Jarred Land

Fire Chief
Staff member
Joined
Dec 27, 2006
Messages
10,644
Reaction score
33
Points
10
REDCODE is one of the most beneficial technologies that you get when you shoot RED.

Nate goes over why its important and how to use it.

 
Great video! Love these RED Tech videos! Makes it easier to explain to people why to go with RED!

And at 4:20 at "Competitors RAW" that made me chuckle lol :devil:
 
I think this is the most "techy" one to date and I actually think that's a great move. Nicely laid out and executed for sure.
 
Phil Holland said:
I think this is the most "techy" one to date and I actually think that's a great move. Nicely laid out and executed for sure.
Click to expand...
+1

Fantastic/great!

VERY clear concise and useful AND most of all very practical.

Nate and the format reminds me of Haas "Tip of the day" lol.*

The table of data rates and capture resolutions and compression schemes/methods is very useful indeed.

It's really cool that Nate/RED TEAM chose to stress the point that with REDCODE that the larger the image size in term of sensor K the more efficient that REDCODE becomes. I think that's a critical point for the future as I never understood how 4444 and totally RAW dynamic capture schemes have not already been crushed under their own weight. Having to bring data storage systems on set the size of a small coffin can't be sustained forever IF sensor sizes and number of photo sites increases in the future. So mathematically (right now) wavelet type compression schemes and REDCODE seem to be the only viable way forward in the future I think. RAW capture for high resolution sensors above 8K (and similarly with 4444) just seems a mathematically intractable problem and new hard drives and higher bandwidth circuitry is not going to fix that any time soon. GOOD to hear and learn the REDCODE becomes increasing relevant in the future to better handle progressively larger image sizes in a disproportionately more efficient way.

So this "take-away" was good...

Paraphrase/slightly sloppy quote;

"[Taking ... a Weapon 6k RED DRAGON @ 5:1 (compression ratio on REDCODE) gives rise to a data rate of 146 MB/s whereas a Helium 8K @ 8:1 has a data rate of 162 MB/s; That's only an 11% increase in data rate but a 78% increase in pixel count...".

REALLY LOVE that resolution target comparison... That speaks volumes right there.

FANTASTIC! :-)

If I was a multi armed deity that would be six thumbs up. (Excellent video; really appreciate you guys reaching out with such clear and direct material)

_________________________________________________________________________________________________________________________________________________________________________________

* Haas "Tip of the day" https://www.youtube.com/watch?time_continue=19&v=HxPjH4v5iEg
 
Hey Jarred

We bought Reds simply because no-one does Raw better.

My old business partner used to describe our 4 Reds as Codecs with cameras attached. Nothing comes close to the efficiency of Red Raw.

I better watch this video now.
 
I like these videos. Good stuff!

Is it just me or does Nate's skin tone have a very distracting pink hue to it? It's jumping out at me. I noticed that on some of my stuff when switching to IPP2 in RCX.
 
Great stuff, as always. And it does drive home the point that shooting REDcode is actually more efficient than ProRes, a point that is often lost on many people.
 
The pixel Data in raw Bayer is one third of 4:4:4. Adjusting for the data differences the compression ratio on the list shot, is a portion less for ProRes, as is expected between wavelet and prores's type of compression.

A more practical pixel format for filming is 4:2:2, to capture the underlying single sensor, or 3 chip data. Even 4:2:0. Mathematically Bayer is less data than 4:2:0 which is 50% more when uncompressed. If 3 chip didn't have it's prism issues, these could be used better. But fortunately, the way the real world is structured, it's relatively, easy to fill in the data somewhat by estimation. So, you get something somewhere between Bayer and 4:4:4 per pixel. You just have to have a computer spend the time processing it. Frankly, when dealing with a single chip to 4:4:4/4:2:2/4:2:0 (as many are) doing the conversion in post may be better and more reliable than in camera. But for mission critical in feild broadcast, or low budget, it shouldn't matter that much (except I imagine the broadcast standards probably have a different opinion).

I haven't look into pro res technically. Does it use the latter overlapping dct compression to get reduce artifacting at similar rates to wavelet? Wavelets losses at high compression show up differently to this.

6k 5:1. Imagine that because of the upres for 4k deliver it probably requires extra detail for pixels crossing target pixels in the interpolation handling scheeme. On 8k, the pixels share even boundaries with the 4k or 2k target. But you see also, that you are using extra datarate to record 6k Bayer to get 4k output, but 4:X:X has the flexibility to down res in camera and record the target 4k directly raw. So the gains of Bayer diminish.

So, the mag would be full in around 80 seconds, so at 8:1 it will last 640 seconds. How big do the mags go?

High detailed, high contrast (clear images) images, and talking heads on simple backgrounds. A better test is to shoot a normal artistic dof controlled scene with plenty of motion/movement. These are harder to compress due to the movement and blur not conforming well to compression schemes. They have to handle data from multiple locations aggregated into pixels, rather than having single pixels represent a single location to map out image shape between pixels (I'm talking about sensor pixels rather than delivery). High contrast detail suites single chip compression better performance. Debayering scheemes use estimations that may lower contrast (but that is from experience years ago before Red) but really the difference between pixels, making it easier to compress, and colour channels where detail is reduced in instance based estimation. High contrast images make detail stand out. But at 6k to 2k or fhd delivery, you have plenty of detail anyway, as well as 8k to 4k/2k/fhd. You get a quality product here.

-----------

Now some misconceptions. David Newman developed modern raw Bayer wavelet compression. I was there when he announced publically that he had done so over the previouse night or so. I've had a number of valued breif correspondences with David in those years. I didn't think of it. I was more concerned with the sparseness of data and detail in debayering at the early stage. But it proved to be much more reliable, so I joined the crowd. I had been more interested in three color multi layer or prism (though prism has real issues with wide aperture I was informed by one of the engineer's). But X3 went virtually nowhere as far as high res video shooting use went (but I still keep an eye out to use non-foveon alternative three color video sensors). I was instead devising a lossless scheme that turned out to be similar to the original lossless estimation based codec jpeg from the JPEG group, but more sophisticated, along with an idea which Jaun, who since was working at Red and is familiar with my publicly expressed compression ideas (not the better private ones), came up with to simple compress neighbouring frames by difference (inter). I actually would advocate that simplified inter frame as a potential simply scheme to get a bit more compression in raw here. I've advised BM in the past on that too. I forget wherever they implemented it.

Storage capacities should go up exponentially. The Intel/Micron 3D solid state storage technologies had/has great promise. But it was introduced charging like a bull at a gate. But there are other 3D storage schemes that could come on line (many developments over the years) even tape/film. I have a scheme for more than 10 years for a very long archival grade ultra dense tape mechanism. But the reality is unheard of densities could be coming over the next ten years. Even enough to film and store three point 3D 8k in three large dots, which I mentioned in my other thread. So, much potential for camera developments.

Now, higher frame rates reduce temporal movements increasing compressibility. 3 color increases compressibility relavent to data. Higher bit depth increases compressibility compared to data as well. This works due to the way compression works. Reducing temporal movement makes the data more predictable intra frame. 3 color makes the data more consistent for compression. Higher bit depth increases compression because it cost little to increase the bit height of a wave estimating it compared to a more shallow wave, then you add the detail difference ontop of that.

Now, more pixels increases compressibility, but how much I don't know, as your lens resolving definition will determine that too. If the lens reduces the differences between pixels due to not being able to resolve, you should be looking at increased compressibility. It would be an interesting experiment to put 2k, 4k, and 8k resolving lenses on three 8k heliums through a precise mirrored prism setup (to give the same image to each) to see how much difference this makes. But the reality is that you have to test for the strengths of each setup against each other (Bayer, and 3 color) with normal hard and hard scenes, to see how each performs against each other's best on closely equivalent terms. While I can tell you Red Helium cameras will offer you about the best overall on the current market, I can't quantify the image to a high degree without tests like these.

In the meantime, to increase the shooting capacity of storage with increasing resolution with very high quality, really requires compression routine development. There is one benifit to higher resolution though, the smaller 8k details on average should matter less. So their fidelity can be sacrificed a bit. Younger and fewer older people might be able to see the highest resolution, and only in a tiny portion of our visual feild that our brains use to flick around a scene sampling details many times a second, at that, but most of use are 4k and below. The 4k is probably acceptable to 8k sighted people anyway. 8k single chip shoots better 4k, and 8k is better for stills. So 8k shooting for 4k delivery is something worth doing. My advice to ambarella, published on dvinfo about the original JVC 720p vertical camcorder when released (where I pointed them) was a list of how to systematically break down an image to maintain maximum quality for the given datarate (what to sacrifice and what to give more data too). For increased compression against visual quality, Red could do this sort of research for higher compression ratios, and consumer compression products. Present 13:1 compression could be enhanced by sacrificing datarate of the least substantial part of an image to redistribute that data portion to enhance the most substantial parts of the image. At 8k holographic, significant savings could be had in future. Ambarella was a leader in the feild at compression at consumer rates.

Thanks.
 
The only drawback of REDCODE is that you need a beefy processor like an AMD Threadripper 1950x that will cost you around 1% off the cost of a Monstro.
 
You must log in or register to reply here.
Back
Top