URSA 12K banding

Hi guys!

I have a strange situation with my Ursa 12k. Haven't noticed that before (maybe wasn't looking right). I get heavy banding in the sky. Resolutions were 8K & 12K, compression was 8:1 and 5:1.
The strangest thing is that the same shots on my Ursa Mini Pro 4.6k are completely clean.
There are some examples. Note that I deliberately pushed saturation to the max (on both cams) to see the problem better.


Is that sensor problem? It's not ND filters problem since on 4.6k I used the same ND stop and it's clean.

Your thoughts?

Thanks!

Doesn’t look right at all.

What’s your post workflow?

JB

Doesn’t look right at all.

What’s your post workflow?

JB

Hello John,

in this case I wasn't doing anything special. I had some 6-7 nodes. The 2nd one is CST (DWG input - Gen 5 Film gamma). At the end there's Kodak 2383 D55 LUT and before that node with (rec 709/gamma 2.4 to rec 709/cineon).
As I said, I pushed saturation all the way up when I saw this banding to emphasise the problem.

But the main thing is that the same was pasted on Ursa 4.6 Pro and it's completely clean...

Linear or Tetrahedral LUT interpolation?

Hi Robert,

linear...

But you have pushed the sky saturation much more in the UMP12k footage.

Does changing the interpolation mode to Tetrahedral solve the issue?

Otherwise - could you post the BRAW sample for others to investigate?

It somewhat looks more like aliasing, but given there is some quite heavy chromatic aberration, it may be connected with your lens - were both cameras using the same lens? CA is present in both images, worse in the 12K, but they are too low resolution, to get a clearer impression.

It’s not the lens. I tested it with 2 lenses. The same thing happens on both.
Tetra didn’t change anything. Yes, sat is pushed all the way but jost for the purpose of showing that 4.6k survived that same thing without any issues.

Otherwise - could you post the BRAW sample for others to investigate?[/quote]


Yes, I’d like to, but here I can upload only 1mb files!!

It’s not the lens. I tested it with 2 lenses. The same thing happens on both.
Tetra didn’t change anything. Yes, sat is pushed all the way but jost for the purpose of showing that 4.6k survived that same thing without any issues.

I can see that you pushed the saturation for both the same - but when you compare the skies you can clearly see that the UMP12k version got more blue depth to start with and you are pushing it over the edge.

Again - please post the sample. If you got no upload space, you can upload it to my WeTransfer account:
https://laufbildkommission.wetransfer.com

Okay Robert, I’ll upload those BRAW files this evening.
Just wanted to say, banding was visible before sat push.


Thanks!

You can export a single BRAW frame from the RAW tab

JB

You can export a single BRAW frame from the RAW tab

JB

I thought you were going to say "Get this new camera and the problem will be gone"

You can export a single BRAW frame from the RAW tab

JB

Will do that, John!
Cheers!

You can export a single BRAW frame from the RAW tab

JB

I thought you were going to say "Get this new camera and the problem will be gone"

What like RED?

JB

Hey Robert!

I just uploaded BRAW frames to your wetransfer account!

Downloadlink here:
WeTransfer

I've done some testing and indeed there is banding in the sky.

NodeFlow.png (92.45 KiB) Viewed 3278 times

UM46k

1.1.1_1.1.1-50%.jpg (1020.5 KiB) Viewed 3278 times

UMP12k

1.3.2_1.3.2-50%.jpg (1021.01 KiB) Viewed 3278 times

But there is something you can do:

UMP12k banding almost removed

1.3.1_1.3.1-50%.jpg (932.98 KiB) Viewed 3275 times

The trick is this setting at the start of the flow:

NR_settings.png (21.49 KiB) Viewed 3275 times

You can see that the red channel has already banding even without boosting the saturation:

Red channel - without saturation boost

R_channel.jpg (937.18 KiB) Viewed 3262 times

Is this the Ursa 12 without the OLPF?

Whilst there is banding as Robert has shown, I am not seeing it anywhere near as severe, even with a quick and dirty sat and color boost pushed right up. This is in my work online suite with a 10bit grade A reference monitor. The whole chain has to be considered when you are faced with artefacts: this is coming from 8K/12K down to HD, and then here the stills are further compressed to get below 1mb. What is the bit depth, sampling rate and codec of your timeline/output?

Since however, neither the 4.6K nor 12K camera, with 12bit Braw should exhibit banding, this is why I suggested aliasing. Aliasing can occur through the concatenation of codecs in post and it can often look like bit starved banding. I see it quite often in my job as an online/grading editor in UK broadcast; especially from very high rez acquisition sources, to HD, that we still broadcast in here. So therefore, one explanation here could be that the 4.6K Ursa is producing significantly less ultimate aliasing over codec transition than the much higher rez 12K. And could be why the 12K needed an OLPF before Netflix acceptance; and the 4.6G2 didn't, for example

Timeline 1_01_02_13_19_2compressed.jpg (955.7 KiB) Viewed 3239 times

Thanks guys for analysys!

Both Ursas have OLPF installed.
Don’t know what to say. Aliasing or banding this should be happening on such camera.
The saddest thing is that cam is not under warranty so
i can throw it in the bin…

Is this the Ursa 12 without the OLPF?

With OLPF.

It can still happen with an OLPF, Mario, under certain conditions, with all cameras. Really it didn't look that bad here on a broadcast reference monitor. You pushed it quite heavily too; any codec will break down eventually. Yes I know the 4.6 didn't but I wouldn't have said the camera's goosed from what I'm seeing. You could just get it checked by BMD/an engineer, rather than throw it in the bin

It was already at BDM service two times. Faulty usb-c board. I really have no patience nor I think I should pay for this even out of warranty. I bought this camera (and payed nice sum money) to work with it. In 2 years it was almost 10 months out on service. So, what exactly did I give my money for??

If the banding is getting worse down the line, adding very mild grain should help. Sometimes, footage out of the cameras is just too clean, and any kind of compression (even BRAW) is reducing the information in very subtle transitions.
You don't mention using a polariser, which I'd suggest for this situation, to get deeper blue from the very start.
And then, there's the redistribution of values in the camera, going from linear to log, since human perception is differentiating better in the shadows than in the highlights. For that much of a saturation boost I'd expose lower.

Bildschirmfoto 2024-04-15 um 09.36.33.png (347.65 KiB) Viewed 3086 times

Oh, and if you want to dump the camera, rather send it here. I'll pay for transport

I loaded those shared BRAW frames from the 12K, and see the same issue when isolating the red channel. For what it’s worth, the Resolve Deband OFX qualified to the sky eliminates that banding. This method will easily run real time, and is much less processor intensive than using noise reduction.

Whilst there is banding as Robert has shown, I am not seeing it anywhere near as severe, even with a quick and dirty sat and color boost pushed right up. This is in my work online suite with a 10bit grade A reference monitor. The whole chain has to be considered when you are faced with artefacts: this is coming from 8K/12K down to HD, and then here the stills are further compressed to get below 1mb. What is the bit depth, sampling rate and codec of your timeline/output?

Steve - I can clearly see the banding on my calibrated 10bit UHD EIZO monitor, in an UHD timeline.
It can be seen in the 16bit TIFF files I rendered out.

As I have shown, the banding is inherent in the Red channel of the BRAW footage.

Since however, neither the 4.6K nor 12K camera, with 12bit Braw should exhibit banding, this is why I suggested aliasing. Aliasing can occur through the concatenation of codecs in post and it can often look like bit starved banding. I see it quite often in my job as an online/grading editor in UK broadcast; especially from very high rez acquisition sources, to HD, that we still broadcast in here. So therefore, one explanation here could be that the 4.6K Ursa is producing significantly less ultimate aliasing over codec transition than the much higher rez 12K. And could be why the 12K needed an OLPF before Netflix acceptance; and the 4.6G2 didn't, for example

I don't think any of these valid points is here the case.
It looks like a problem created by the partial de-mosaicing of the RGBW sensor in camera, an edge case for the algorithm. I think we should point Kristian Lam and Hook onto this for further investigation.

Whilst there is banding as Robert has shown, I am not seeing it anywhere near as severe, even with a quick and dirty sat and color boost pushed right up. This is in my work online suite with a 10bit grade A reference monitor. The whole chain has to be considered when you are faced with artefacts: this is coming from 8K/12K down to HD, and then here the stills are further compressed to get below 1mb. What is the bit depth, sampling rate and codec of your timeline/output?

Steve - I can clearly see the banding on my calibrated 10bit UHD EIZO monitor, in an UHD timeline.
It can be seen in the 16bit TIFF files I rendered out.

As I have shown, the banding is inherent in the Red channel of the BRAW footage.

Since however, neither the 4.6K nor 12K camera, with 12bit Braw should exhibit banding, this is why I suggested aliasing. Aliasing can occur through the concatenation of codecs in post and it can often look like bit starved banding. I see it quite often in my job as an online/grading editor in UK broadcast; especially from very high rez acquisition sources, to HD, that we still broadcast in here. So therefore, one explanation here could be that the 4.6K Ursa is producing significantly less ultimate aliasing over codec transition than the much higher rez 12K. And could be why the 12K needed an OLPF before Netflix acceptance; and the 4.6G2 didn't, for example

I don't think any of these valid points is here the case.
It looks like a problem created by the partial de-mosaicing of the RGBW sensor in camera, an edge case for the algorithm. I think we should point Kristian Lam and Hook onto this for further investigation.

If the banding is getting worse down the line, adding very mild grain should help. Sometimes, footage out of the cameras is just too clean, and any kind of compression (even BRAW) is reducing the information in very subtle transitions.

If you look at the samples you can see they are far from being too clean - in fact the colored noise is quite strong.

As I have shown, the banding is inherent in the Red channel of the BRAW footage.

It's in all 3 channels, not just the red.

It looks like a problem created by the partial de-mosaicing of the RGBW sensor in camera, an edge case for the algorithm. I think we should point Kristian Lam and Hook onto this for further investigation.

So you'd assume this is inherent to the de-mosaicing, just showing too much with the specific scene and nothing is wrong with that camera?

As I have shown, the banding is inherent in the Red channel of the BRAW footage.

It's in all 3 channels, not just the red.

That's correct - but the Red channel is the only one where you can see it immediately without further enhancement.

I have now found another way to deal with the banding:

2 nodes with their color space set to LAB and AB channels (channels 2 & 3) active only:

Deband_setting.png (117.38 KiB) Viewed 2929 times

NR_Faster_setting.png (21.5 KiB) Viewed 2929 times

Watch in a 300% view what happens to the color noise.

And also remove the Chromatic Aberrations:

Chromatic_Aberration_settings.png (14.79 KiB) Viewed 2929 times

Steve - I can clearly see the banding on my calibrated 10bit UHD EIZO monitor, in an UHD timeline.
It can be seen in the 16bit TIFF files I rendered out.

As I have shown, the banding is inherent in the Red channel of the BRAW footage.

Fair enough Robert, I can only tell you what I see on my calibrated monitor I use for my job. I didn't doubt there is banding, I said so; and if there is a fault with this particular camera then of course it's good to get to the bottom of it and make BMD aware. But everything else I said also pertains and even the original 4.6K G1 failed EBU tests for tier 1 UHD acquisition, from a purely sensor resolution perspective, for the reasons I outlined.

Furthermore if you look at the small rooftop between the 2 bigger buildings, there is clearly colour aliasing on all the images, which points to this as well. An OLPF is never a catch all in any camera but one that has to work across many different resolutions may have an even harder time too; though I am no expert on sensors; only I know what to look for in my job and through QC for UK television.

Thanks guys for all of your suggestions and views on this matter.
Robert, I'll try your node workflow but seems to me we forget the main thing here, which is
THIS CAMERA SHOULD BE ABLE TO HANDLE THIS SCENARIO! Right?

Also, it was mentioned here a pretty much of chroma noise. That's true. It was @ ISO 800, it was middle of the day. How come this "magnificent" (until 12k CIne released) camera has so much noise at those values???
Am I missing something here? I mean I've been shooting BMD since Ursa Mini 4.6k, then I got me UMP and Pocket 6K Pro and 12k 2 years ago. Color science is great but some other stuff not so. Now, when I see their new stuff announced, I can't help but have a grin on my face.

It would be fair if they say, "Okay, there's something kinky with your camera (after 2 service trips of more than half a year and now this) and we'll replace it with a new one, courtesy of us!"

It would be great to make BMD aware of this somehow but....I wish....

Phew!

Thanks guys for all of your suggestions and views on this matter.
Robert, I'll try your node workflow but seems to me we forget the main thing here, which is
THIS CAMERA SHOULD BE ABLE TO HANDLE THIS SCENARIO! Right?

I agree. I do not own an UMP12k, only have the sample footage from John Brawley and others. So I have no first hand experience from shooting myself what to expect from the camera and what makes the footage break.

I could be - and this is just wild speculation - that your lens with its strong chromatic aberration could make things worse because it shifts colors around and messes with the RGBW de-mosaicing/compression.

Also, it was mentioned here a pretty much of chroma noise. That's true. It was @ ISO 800, it was middle of the day. How come this "magnificent" (until 12k CIne released) camera has so much noise at those values???

Bear in mind that when you increase saturation that drastically, you will increase the ratio between color noise and luminance noise too. You can throw a lot of noise reduction onto the UMP12k material without doing harm - much more than with the Bayer pattern sensors.

Am I missing something here? I mean I've been shooting BMD since Ursa Mini 4.6k, then I got me UMP and Pocket 6K Pro and 12k 2 years ago. Color science is great but some other stuff not so. Now, when I see their new stuff announced, I can't help but have a grin on my face.

It would be fair if they say, "Okay, there's something kinky with your camera (after 2 service trips of more than half a year and now this) and we'll replace it with a new one, courtesy of us!"

It would be great to make BMD aware of this somehow but....I wish....

I would contact either Hook or Kristian Lam via PM and make them aware of this thread.

When I did a deep dive into the Indvidual color channels, I could spot contouring. Just taking the one 12k BRAW file however, if I were to shoot it differently, (and it looks great the way it is,) I would have given a stop less exposure to the scene in the camera, i.e. 6 ND instead of 4, and F4 instead of F5.6. To get the look I liked in post, I brought down the exposure by two stops. Shooting the way you did made the shadows nice and clean, but availed fewer buckets to contain the sky, hence fewer shades make contouring more apparent, but this is a nitpick quibble because I like the way it graded, no special difficulty, and not the slightest complaint with contouring or chroma from me in the finished image. My preference is natural scene color, not cheap postcard blue skies. Nevertheless, when I pushed saturation, it didn't break apart.

This is how it looked for me, no deband nodes nor added grain. I think it looks great, the way I would have wanted it at least, and I could immerse myself into beautiful Europe from this picture.

https://drive.google.com/file/d/1L8lMJJ ... sp=sharing

Best

Robert, thanks! Now, about the lens(es). It is not 1 lens used here. First is Tokina Cine 11-16mm and second is Zeiss Distagon 21mm. Both of the displayed this "banding".

Don't know what would contacting those guys you mentioned resolve...hmmm..

Tom, thanks for your take on this. Yup, your version looks almost clean. Looks good. I do see some banding but it could be some compression with jpeg google drive file...maybe. I also don't like "kicked to the max" skies, but this I did for the test purposes.
Glad you like our ol' Europe!

This is how it looked for me, no deband nodes nor added grain. I think it looks great, the way I would have wanted it at least, and I could immerse myself into beautiful Europe from this picture.

That's how I would have graded it too and nothing really wrong with that image; and that's how it looked on my monitor initially before messing with it. We were pushing it well beyond it's comfort zone; in order to accentuate any artefacts. I don't think there's much wrong with the camera Mario; but if you'll forgive me perhaps a better lens. Is that Italy; was just shooting there

Tom, thanks for your take on this. Yup, your version looks almost clean. Looks good. I do see some banding but it could be some compression with jpeg google drive file...maybe. I also don't like "kicked to the max" skies, but this I did for the test purposes.
Glad you like our ol' Europe!

It's not jpeg compression, I see it in Resolve with the graded BRAW too, so try again with BRAW Q3 compression.

Regarding the lens, metadata reports Zeiss or Voigtlander. My judgment from the image frame is the lens is of very high quality. A little bit of CA is not uncharacteristic, and I did use minor CA correction with a QFX node.

But an additional point I would make, is that at F/5.6 for small 2.2 µm pixels, diffraction is at play. For landscaping, I would try for F/4.0 or at minimum, F/4.8 and use more ND, i.e. ND6. That and Q3 compression, I think what's left of contouring in the sky will be snuffed.

This is how it looked for me, no deband nodes nor added grain. I think it looks great, the way I would have wanted it at least, and I could immerse myself into beautiful Europe from this picture.

That's how I would have graded it too and nothing really wrong with that image; and that's how it looked on my monitor initially before messing with it. We were pushing it well beyond it's comfort zone; in order to accentuate any artefacts. I don't think there's much wrong with the camera Mario; but if you'll forgive me perhaps a better lens. Is that Italy; was just shooting there

Hey Steve, the first lens is a bit meh (Tokina Cine 11-16 T3), but there was a second lens here also (Zeiss 21 Distagon) which is not so “meh” and it displayed this artefacts also.

Not Italia but close….Croatia.

Regarding the lens, metadata reports Zeiss or Voigtlander. My judgment from the image frame is the lens is of very high quality. A little bit of CA is not uncharacteristic, and I did use minor CA correction with a QFX node.

But an additional point I would make, is that at F/5.6 for small 2.2 µm pixels, diffraction is at play. For landscaping, I would try for F/4.0 or at minimum, F/4.8 and use more ND, i.e. ND6. That and Q3 compression, I think what's left of contouring in the sky will be snuffed.

Hey Tom!
Yea it’s Zeiss.
I thought diffraction steps in at higher f-stops (11 and up).
This was shot on Constant bitrate 8:1.

Cheers!

It depends on sensel size, but f5.6 seems a bit low for me to cause visible diffraction. f8 and beyond for such small sensels, yes. If you had only HD on a 24 x 36mm sensor, it would not show before f11, maybe 16. But who'd buy such a sensor?

I mean in all of those analysis we all forgot 1 thing. How come UMP didn't exhibit such behaviour??

Without all the information only available to BM's engineers, that point is guesswork.

Don't know what would contacting those guys you mentioned resolve...hmmm..

Hook is responsible for the color science and BRAW and Kristian Lam is the product manager for all cameras.

Can someone ask them to join this debate?

Can someone ask them to join this debate?

I'll try to contact them after NAB is over.