DRAGON vs EPIC MX Noise - NOT GOOD

[Peter Majtan]

Jarred - maybe I am going a bit crazy here - but I know you guys have been playing around with a modular "behind-the-lens" ND solution...
It would be amazing to have the OLPF modular (user-echangeable) as well. This way folks can have several OLPFs at hand and chose the one (with appropriate calibration) to suit the particular needs. This could even extend to IR and other applications...
I am sure this could be "walk in the park" for you guys... ;o)

:sifone: Peter

 

[Adam Eden]

Jarred i sent back my Epic M #850 to get dragonised. Remember you owe me a free DRAGON hat as promised from last year!

 

[Jarred Land]

Jarred - maybe I am going a bit crazy here - but I know you guys have been playing around with a modular "behind-the-lens" ND solution...
It would be amazing to have the OLPF modular (user-echangeable) as well. This way folks can have several OLPFs at hand and chose the one (with appropriate calibration) to suit the particular needs. This could even extend to IR and other applications...
I am sure this could be "walk in the park" for you guys... ;o)
:sifone: Peter

your not crazy.. that was always the intention, user replaceable OLPFs. We have about 6 different ones already that we have made along the way ( IR PASS , no LP, etc. ) so it is a cool option.

Problem is that as easy as it seems for our engineers to do it, for normal people like me it is kind of a pain in the ass to change the OLPF without the proper custom tool.

I guess we could just make a bunch of tools....

 

[Isaac Marchionna]

... or more specifically some red colored noise ( which is much more noticeable than monocromatic noise as the old OLPF has ) in the RED channel, which Graeme is fixing.

And right there is why this thread is worth it's weight in electrons, problem identified, and problem being solved.

 

[Peter Majtan]

your not crazy.. that was always the intention, user replaceable OLPFs. We have about 6 different ones already that we have made along the way ( IR PASS , no LP, etc. ) so it is a cool option.

Problem is that as easy as it seems for our engineers to do it, for normal people like me it is kind of a pain in the ass to change the OLPF without the proper custom tool.

I guess we could just make a bunch of tools....

I can foresee a small peli case (ala the one for the mounts) with a set of OLPFs and a custom tool (or a bunch of tools) becoming a hot-selling item in the REDSTORE - just make sure you package it with the Arctic Butterfly cleaning system... ;o)
The question then remains whether the brain can "sense" which OLPF is in at any given time - or will this be user responsibility to chose the right OLPF calibration from the (newly added) menu in the UI...?
Either way - I think most will be happy with that!!!

:sifone: Peter

 

[Mark Crabtree]

I can foresee a small peli case (ala the one for the mounts) with a set of OLPFs and a custom tool (or a bunch of tools) becoming a hot-selling item in the REDSTORE - just make sure you package it with the Arctic Butterfly cleaning system... ;o)
The question then remains whether the brain can "sense" which OLPF is in at any given time - or will this be user responsibility to chose the right OLPF calibration from the (newly added) menu in the UI...?
Either way - I think most will be happy with that!!!

:sifone: Peter

This would be fantastic. I'm in!

 

[Mark Wilkinson]

Hey Jared,

Since you popped by for a visit, any thoughts on the benefits and/or possibilities of shooting Dragon at very low ISO - like 50?

Would the sensor benefit from that or is there a theoretical limit to more like=better?

Thanks,
Mark

 

[Meryem Ersoz]

Does RED need any additional footage samples, or does Graeme have everything he needs to diagnose and tackle the issue?

I started processing some of the footage that I shot last week and scrutinizing it all more closely on my new 4k monitor by jove, it's all there, just as described, excess noise at 800 and a busy red channel.

Thanks to Eric for shining a light on it. Thanks to RED for re-calibrating the sensor to fix the problem.

There is so much that is good about the new sensor. The color and the texture are dreamy, and when it's shot clean, it's very very clean. I hope Graeme can dust things up a bit between 800-2000 ISO, that would be flipping amazing. It doesn't seem like it is going to be a huge deal when you look at the individual channels, specifically the red channel, which is consistently overly intense in the RGB parade. Seeing the problem in my own footage actually gives me a lot of reassurance that adjustments to the sensor to clear up the issues aren't going to be too monumental. I have witnessed RED tackle many tougher problems successfully.

If more footage is needed as example, I can provide it.

 

[Marc Wielage]

I will climb out from under my chair to observe that I don't think the noise in the gootage is that bad, and in truth some mild noise reduction (even if it's red-channel-only processing) will solve it. Coming from a film background, my observation is that film grain was 10 times worse than this, especially at 800 ASA.

Back under chair...

 

[Medavoym]

I will climb out from under my chair to observe that I don't think the noise in the gootage is that bad, and in truth some mild noise reduction (even if it's red-channel-only processing) will solve it. Coming from a film background, my observation is that film grain was 10 times worse than this, especially at 800 ASA.

Back under chair...

The noise in the gootage is of no consequence. It's the noise in the footage which is actually causing all this ruckus.

Just kidding my friend

 

[Matt Ryan]

The noise in the gootage is of no consequence. It's the noise in the footage which is actually causing all this ruckus.

Just kidding my friend

:lol:

 

[Gavin Greenwalt]

If I had a choice it would be an OLPF with similar color rendition but a slightly weaker OLPF with slightly increased aliasing. I feel like 6k in of itself is pretty good for minimizing moire.

 

[Robert Ruffo New]

I don't doubt that it's the better option! I am so excited to see that noise cleaned up. I was looking at some match lit footage with the old OLPF and it was so clean. If Graeme gets it that clean and keeps the extended highlight range I think people are go to flip out about how good the sensor is. I know I will.

Me too.

 

[Gunleik Groven]

My apologies if someone else has raised this already...

BUT WOULD IT BE POSSIBLE ?

To calibrate the dragon sensor so that at lower ISO's, let's say 400 and below we have the maximum DR ? (Dynamic Range)

Normally we would use lower ISO when shooting outdoors in bright sometimes contrasty less controllable light, so this is when we need all the DR we can get.

Then as we increase the ISO above 400 we start to loose some DR but we keep the grain cleaner and smoother.

Could this be a realistic trade off ?

I'd happily lose a stop or more or DR if the images were much cleaner.

Just a thought.......

Is this even possible or realistic ?

Practically you allways have "all" DR available at "all" ISOs...

 

[Gunleik Groven]

Thanks Jarred for chiming in on the status and resolution.

 

[John Marchant]

I can foresee a small peli case (ala the one for the mounts) with a set of OLPFs and a custom tool (or a bunch of tools) becoming a hot-selling item in the REDSTORE - just make sure you package it with the Arctic Butterfly cleaning system... ;o)
The question then remains whether the brain can "sense" which OLPF is in at any given time - or will this be user responsibility to chose the right OLPF calibration from the (newly added) menu in the UI...?
Either way - I think most will be happy with that!!!

:sifone: Peter

I like your crystal ball

 

[Raphael Varandas]

Nice thread.
Thanks Jarred.
Great as always Hans.
Cheers

 

[Alexandros Tsoupras]

Yes, but needs some tweaking regarding noise. IMHO, its the colourfulness of the noise that is the problem. It's a colour science subject. And thus, I'm pretty confident that Jarred and his team will resolve this issue in a few weeks.

But Dragon won't become an ISO 2000 camera. Very unlikely. And I never expected it to be. But noise-free ISO 800 is what I expected.

Hans

+1 on that.

 

[Alexandros Tsoupras]

I don't doubt that it's the better option! I am so excited to see that noise cleaned up. I was looking at some match lit footage with the old OLPF and it was so clean. If Graeme gets it that clean and keeps the extended highlight range I think people are go to flip out about how good the sensor is. I know I will.

+1 on that .

 

[Björn Benckert]

We got the new olpf now I found the old one amazing in low light and you had to be a bit stupid to make it clip in the highlight... The new olpf has mad amount of highlight protection as you can see in this test where I exposed for a shadow side of a wall and then aimed straight into the naked sun at 1 o clock in the after noon... to me, yes then any camera should clip like crazy but the new olpf dragon just rolls of nicely so to me that is, sound strange to say, a bit more highlighy protection than I possibly ever need, especially since we got HDRx. The only issue with the old olpf, to me was the olpf flare. I though it would be no problem but allready on our two first projects with the camera we got those olpf flares all over the place and I really had to use my worst saies tricks to push those flairs through the client approval. So I think sticking with the old olpf is not an option.

As I understand these "olpf Calibrations" are more or less 3D luts, no? I mean the sensor capture the light and the rest is processing that info. And the calibration is more or less a lut tagged on so what is shot is looking as it should when viewed and different olpf's needs different luts to represent colors correctly. This is ofcourse just me speculating and trying to simplyfy what is going on. But personally I would really want the camera to just leave a linear image the "post processing dynamic rounding stuff" I think should be left until later. As it is now RCXP in linear color space does not behave linear at all, as the highlights are rounded off when brought up to towards the clipping point. And the iso is not a mathematically correct gain/offset is a dynamic offset. making it impossible to match for example two shots one with ND.3 and one without One would think that if the histogram is not clipping then you could just bring up the iso to double for the ND shot to match the without ND. But thats not the case as the image is squashed towards the highlights when moving the iso slider, a.e it's not a linear push towards the right in the histogram so the image is not just getting brighter it also dynamically changed... to me that's a bit mind bothering, and does not feel mathematically right at all, even though it might look pleasing to the eye.