Rethinking Compression w/RED's 8K Helium Sensor

[Marcus Friedlander]

Huge +1 for Edge mode!

I do a lot of DP/OP gigs, and when I'm shooting handheld, I like to pull my own focus too.

Ill map one of the EVF buttons to Edge Focus, and the other to Exposure Check, and fly around the set.

I've gotten so used to looking at Edge Focus, that I'll shoot most scenes with it on in the EVF

So unbelievably helpful for nailing focus on these larger sensors!!!

 

[Hrvoje Simic]

+) I am guessing that Increasing Resolution reduces Noise.

It doesn't. It does the opposite.

You are looking at two different sensors and latter has a lower noise floor so allows smaller pixel pitch.

 

[Brendan H. Banks]

Very interesting. I have been moving around from 8:1-12:1 at 7K/8K and been trying to figure out a sweet spot. This is super helpful and picked up a few things to think about. Looking forward to future parts of this research.

 

[Anton Shavlik]

It doesn't. It does the opposite.

You are looking at two different sensors and latter has a lower noise floor so allows smaller pixel pitch.

Increased resolution (i.e. increased number of samples) does reduce noise. Shoot 3200 iso in 2k prores 4444. Use 2k crop and then 8k crop. Napkin math says 8k should look about 4x less noisy per pixel even though you're at the same iso and recorded resolution.

 

[Rob_Austin]

Is there a thread discussing the effect RWG/L3G10 has on Dragon footage?

I have just begun to play with it and, whoa, there's this apparent ~1 stop of
luma loss, desaturation that happens right off the bat.

Apparent in Mark's chart comparison from this thread, anyone have info?

 

[Alexander Thomson]

Is there a thread discussing the effect RWG/L3G10 has on Dragon footage?

I have just begun to play with it and, whoa, there's this apparent ~1 stop of
luma loss, desaturation that happens right off the bat.

Apparent in Mark's chart comparison from this thread, anyone have info?

RWG/Log3G10 is meant to be viewed through a transform LUT like the Medium Contrast, Rolloff 4 REC709 LUT. That brings everything back to normal.

 

[Björn Benckert]

All of this makes perfect sense as long as you're not shooting material for a client that specifies "compression ratio no higher than 6:1."......

Then you a could, whithout lying, argue that 8k 24:1 = 4k 6:1 or even better, if you are mastering in 4k/ UHD.

As the 4 to 1 downsample after compression is in effect.

 

[jake blackstone]

It just occurred to me, that with 4k deliverable down from 8k sensor you nearly double oversample the 4k image, which means, that technically speaking, you may no longer need the OLPF in order to prevent aliasing. Can anyone test this theory?

 

[Sabovic Adis]

+1!

 

[David Battistella]

Pixel pitch will definitely have an effect.

The 8K VV sensor is based on Dragon sensor pitch. Thus the frame is larger (even though it is 8K) than the helium sensor. Just to put in into perspective the Helium 8K frame is slightly smaller than Dragon 6K frame.

When it comes to compressing the finer pixel pitch of the helium sensor (and lower noise floor) it might be slightly easier to compress those pixels. I'm interested to see more VV footage in terms of noise as the noise profile should be pretty similar to what we are seing in dragon (but there are certainly hardware improvements on the camera side that will help take care of that.

In any case. Mark is absolutely correct.

Helium is a brand new sensor and needs to be tested and treated the same way you would a new film stock. Just as Dragon was a whole new thing over the MX sensor and performs completely differently.

Thankfully we have a unified color science in the post process which makes all off RED camera's cut together quite nicely. I am still working on projects that have mixed sensors (through evolution) and as I go back through the ols stuff it mixes really well with the new, especially since the glass and the cinematographer (me) have been consistent.

Still.

Test the shit out of everything and make sure YOU know how it responds.

David

 

[PatrickFaith]

I find the compression setting has to do a lot with using the first two bars on the gio scope, if I am using them I basically can't compress the noise so need to keep the R3D setting low. if I can increase the ambient light evenly, and also keep the blue channel also out of the first couple of gio bins, then I can really crank the compression up on a weapon dragon. In many cases it's hard to light and not be down in the lower gio scope bins, I normally keep the compression the same across a entire project, so setting the compression high can lead to some unusable scenes. Another thing is with anamorphics it can be dangerous to crank up the compression past 5 because I have not oversampled horizontally.

 

[Joachim Hoge]

Excuse my ignorance, but could one assume that Helium 8K 16:1 would equal (or better) 8:1 at 6K Dragon. Or is it more complicated than that?

 

[M Most]

And my point is ... THAT SPECIFICATION comes from experiences based on a different sensor, sampling from a different resolution.

I have more to post still - specifically with respects to chroma & compression - but I can not say this enough .... shooting on "RED" is NOT the same animal - Dragon vs. Helium.

Clients ... and users ... need to rethink some things. It's a new sensor - totally different design - you would think that would be obvious

Perhaps. But until they do, one still needs to adhere to delivery specifications, regardless of whether one believes they're outdated. And particularly when the client is the one we both know I'm referring to
:wink5:

 

[Nick Morrison]

It just occurred to me, that with 4k deliverable down from 8k sensor you nearly double oversample the 4k image, which means, that technically speaking, you may no longer need the OLPF in order to prevent aliasing. Can anyone test this theory?

Fascinating question.

 

[Aaron Lochert]

It just occurred to me, that with 4k deliverable down from 8k sensor you nearly double oversample the 4k image, which means, that technically speaking, you may no longer need the OLPF in order to prevent aliasing. Can anyone test this theory?

How do we test? The actual "low-pass filter" is affixed to the sensor glass this time in DSMC2. Even though the optical accessories we can buy are still called OLPFs, they're really just "OFs".

 

[Marcus Friedlander]

How do we test? The actual "low-pass filter" is affixed to the sensor glass this time in DSMC2. Even though the optical accessories we can buy are still called OLPFs, they're really just "OFs".

I guess you would have to remove and replace the cover glass as well?

Wonder if that'll void your warranty