Weird banding on my Epic-W

[Luca Orlandi]

at this point a public beta to solve the problem for the rest of us would be appreciated
As you say, the problem is irreparable in post production.

It's not a debayer issue. I was reproducing the banding straight out of RCX at full debayer. But I guess since this is catching some traction again I'll go ahead and post the results of an alpha firmware fix.

This is with saturation jacked to +100 in photoshop to exagerate the improvement and the remaining quibbles (slight magenta shift now which can be corrected with a slight curve adjustment in your grading application). The color shift was annoying but really the banding was the irreperable problem before and that's fixed in the current WIP firmware, so they're really close to pushing out a solution.

 

[Jason Han]

It's not a debayer issue. I was reproducing the banding straight out of RCX at full debayer. But I guess since this is catching some traction again I'll go ahead and post the results of an alpha firmware fix.

This is with saturation jacked to +100 in photoshop to exagerate the improvement and the remaining quibbles (slight magenta shift now which can be corrected with a slight curve adjustment in your grading application). The color shift was annoying but really the banding was the irreperable problem before and that's fixed in the current WIP firmware, so they're really close to pushing out a solution.

Maybe you need to bring it in Davinci for full debayer, I know 100% that this improved the issue for me and several people. But that firmware is looking freaking amazing! Really great news

 

[Stacey Spears]

Maybe you need to bring it in Davinci for full debayer, I know 100% that this improved the issue for me and several people. But that firmware is looking freaking amazing! Really great news

Gavin's issue is not a debayer issue and is a different issue than yours. He has a private build from Red to test.

 

[Jacob Ryan]

That's great. Thanks Jason.

In Davinci:
Project settings > Camera Raw > RED > Decode Quality: Full Res. Premium
And when you render go to advanced and tick "Force debayer to highest quality"

I only use davinci now so I can't say Premiere does the same, perhaps try setting playback quality to full and drop it in to see if it decodes it.

Graeme might be able to chime in on the Red Cine X, as he tested it with my R3D's that I sent weeks ago. We found that opening them in software full res fixes most of this problem

 

[Aaron Lochert]

In your preferences in Red Cine-X, you can set snapshots to be full-res tiffs. Jason, are you saying this not a 1:1, full quality demosaic? Because if not, then the whole motion-to-stills workflow would be totally flawed.

But just for the hell of it, here's an entire folder: https://drive.google.com/open?id=0B_l7sR0CCsLdQzM0ODU5REExSDg

Stills in the Davinci folder were exported via the stills export function, there's a DPX, a JPEG, and a TIFF. Preferences in Resolve were set to highest quality debayer, 16 bits. The video is an h.264, forced highest debayer.

Stills in the Premiere folder were exported via the export frame function, there's a DPX, a JPEG, and a TIFF. For the video export, it's an h.264 with highest bit depth and maximum quality checked. I tried to do an 8K HEVC but it kept stalling on me. There's also a DNxHR HQX 10 bit 4K file in there just for the hell of it.

Stills in the Redcine-X folder were exported via the snapshot function. There's a DPX, a JPEG, and a TIFF. There's also a single r3d frame via the r3d snapshot function. In all cases, export resolution was set to highest quality/debayer. The video is a ProRes 4K 422 HQ.

I still see a magenta/green shift happening on the top image. Near the top-middle of the frame, it's green on the wall. Closer to her head, it shifts pink. Another example is around the door handle. It's pink, but where the light reflects off the handle and lights up the door, it goes green.

This image still has banding which I would not consider acceptable. Look around her neck. The transition from shadow to highlight is really telling. She also has a narrow green band which goes through the middle of her forehead, past her nose on the camera-left side, all the way down to her chin.

Gavin, that new firmware looks pretty promising. When you have time, can you post a shot of a face, where you purposely put the skin in the same range where the banding used to happen? For me it was right between stops 11 and 12 on Gio Scope. I don't know if that's where it was happening for you, no one else has answered my question about that yet.

 

[Gavin Greenwalt]

Gavin, that new firmware looks pretty promising. When you have time, can you post a shot of a face, where you purposely put the skin in the same range where the banding used to happen?

So here it is color corrected to remove the slight magenta shift where the green used to be (See before/after image above of firmware update). It was just a slight curve operation done in Photoshop quickly as a proof ofconcept. I also ran it through my DragonColor/Kodak LUT.

And the 'raw' jpeg with the slight magenta shift smack dab in the middle of the highlight range (DragonColor/RG3) where the green shift used to show up badly.

 

[Jason Han]

This image still has banding which I would not consider acceptable. Look around her neck. The transition from shadow to highlight is really telling. She also has a narrow green band which goes through the middle of her forehead, past her nose on the camera-left side, all the way down to her chin.

I know, this is ProRes with green shift baked in. Not a raw 1:1 res. I was just saying higher CRI lights will have a better result.

So the 1:1 full res didn't help you at all? For me it takes a lot of the problem out to a point where you can correct the remaining color shift, but it doesn't completely fix the issue. If that didn't help you, then there is nothing you can do except wait for the firmware I guess

 

[Jon Dishler]

Similar findings of green on white wall -- this is Epic W, natural light, white balanced, full debayer with Nikon 14-24 at 24 f/6.3 (no chromatic) DC/RG4. The areas that go green or purple are very close, here at IRE 65 and IRE 75. No help with wide gamma.
The only thing that I find helps is FilmConvert which brings everything back closer to white as shown.
I hope Red can fix this.

 

[Jason Han]

So here it is color corrected to remove the slight magenta shift where the green used to be (See before/after image above of firmware update). It was just a slight curve operation done in Photoshop quickly as a proof ofconcept. I also ran it through my DragonColor/Kodak LUT.

And the 'raw' jpeg with the slight magenta shift smack dab in the middle of the highlight range (DragonColor/RG3) where the green shift used to show up badly.

That's looking very nice Gavin! Can't wait for this firmware, really hope it's coming soon. So is the banding gone completely? Rolloff improvements? Any of that hard line remaining between highlights/midtones?

 

[Aaron Lochert]

I know, this is ProRes with green shift baked in. Not a raw 1:1 res. I was just saying higher CRI lights will have a better result.

Ah, gotcha. My mistake.

So the 1:1 full res didn't help you at all?

No, because I've been doing a full quality debayer the whole time. From my perspective, full debayer is bad, and 1/4 is worse. Not 1/4 is bad, and full is slightly better, if that makes any sense.

Basically my shooting practices right now are to avoid putting any skin at 50% IRE in DC/RG4 @ ISO 800. That's right where the banding happens on my sensor. I don't know if you guys have it the same spot as me, but again, stops 11-12 on the Gio Scope tool are right where I have a magenta/green band happening. If someone else can chime in, that'd be helpful.

So that's yet another reason I'm rating the sensor at 2000: it pushes the banding up at 75% IRE and outside my typical skintone level. Luckily Helium has enough noise floor and latitude to make this compromise.

 

[Luca Orlandi]

I will try... while wait for a definitive solution in the next firmware upd.

Ah, gotcha. My mistake.

So that's yet another reason I'm rating the sensor at 2000: it pushes the banding up at 75% IRE and outside my typical skintone level. Luckily Helium has enough noise floor and latitude to make this compromise.

 

[Luca Volpe]

I did today some tests you can see here: https://www.facebook.com/groups/1816671215236474/

 

[Gavin Greenwalt]

Just as an update. We shot three 30s spots with the new alpha firmware and there was probably only one shot where the "10% left to fix" will need to be graded out. And that was in a very wide variety of skintones and lighting conditions. Graeme says that they've already 2 more builds beyond what I was shooting with and it's even better. So I think it's safe to say that this problem will be licked as soon as the new build drops. (Also a side note, I had some initial instability for some reason right after updating to the new firmware but across the 3 spots we never had a single crash or boot error. It ended up being more stable than I find most Epic-Dragon builds.)

 

[Aaron Lochert]

Just as an update. We shot three 30s spots with the new alpha firmware and there was probably only one shot where the "10% left to fix" will need to be graded out. And that was in a very wide variety of skintones and lighting conditions. Graeme says that they've already 2 more builds beyond what I was shooting with and it's even better. So I think it's safe to say that this problem will be licked as soon as the new build drops. (Also a side note, I had some initial instability for some reason right after updating to the new firmware but across the 3 spots we never had a single crash or boot error. It ended up being more stable than I find most Epic-Dragon builds.)

Thanks for the update.

Conveniently (not really but I'm looking on the bright side), my Epic-W had to go back to RED for a 0x0000001d error I kept seeing. So since I'm already playing the waiting game, hopefully it'll come back home with the new build in it.

 

[Luca Orlandi]

oh... sad

... my Epic-W had to go back to RED for a 0x0000001d error I kept seeing. So since I'm already playing the waiting game, hopefully it'll come back home with the new build in it.

 

[Anton Esteban]

Once you do a 1:1 debayer, the other good thing to do, in Da Vinci Resolve, is to use the Saturation vs Saturation curve. Simply use the eyedropper on something like skintones and then move the left most point (least saturated part of the image) down and it will remove any color cast on things like walls etc but still keep parts of the image that should have color intact.

 

[Derrick Cohan]

I own a Weapon Helium and noticed these banding color shifts using Cooke, Sigma and Canon lenses even after the firmware update from today.

 

[Derrick Cohan]

So are there any in-camera settings that can make this banding issue go away?

Any ideas of when RED is going to offer a solution to this problem?

 

[Brad Grimm]

So are there any in-camera settings that can make this banding issue go away?

Any ideas of when RED is going to offer a solution to this problem?

I am actually holding off on sending my CF in for its Helium upgrade because of this issue.. but an educated guess as to when it will be fixed... this month.

 

[Phil Holland]

Any ideas of when RED is going to offer a solution to this problem?

Alarmingly soon.