12k lf banding

Hello,

I'm planned to buy a 12k lf so I get all information about this camera.
I downloaded the two samples on the "Blackmagic URSA Cine 12K Gallery" page. (sample with a girl and the big wheel)

When I import this samples on resolve, I can see banding (if I set the timeline to 8k or 12k).
This is outside images so this is not because of light.

Does someone see the same problem or I make a mistake in this so simple workflow ? or the 12 lf generate banding everytime ?

Can you post some examples of what you're seeing?

What software platform are you viewing these files in?

JB

My guess is your viewer shows some odd percentage of the clip like 24%.
Set to 25/50/100 and all wil be fine. If the banding is still there. You might want to set the scaling to crop.

That clip has no banding issues what so ever.

@John This is a sample with banding.

I use the studio version 19.1.3 build 7 (last version) and the last version of the bm raw.

@Bunk Thanks for this explanation. You are right this banding appear only when percentage are not 100 or 50 or 25 or 12.5
So the banding seems to be an artefact of the resize algorythme. I don't understand why.
If the clip is 12k and I want to resize to cine4k the coef is 33.33%
I tried to export as cine4k resolution from a 12k timeline and the banding doesn't appear on the rendered clip but it appear on the timeline at 33% or 34%.

There are others that can speak to the best ways to work in Resolve but my instinct is to at you shouldn’t have a 12k timeline if you’re rendering to 4k for starters.

Have a 4k timeline and let resolve handle the scaling. You also have options on render for scaling options and image quality.

If it’s not rendering it then it’s just a set up issue.

JB

For my experience, if I need to render a full hd, it's better to work with a timeline in 4k (if source is in 4k).
This is why I use a 12k timeline but perhap's it's not a good idea for 12k ...

I make another test with another clip.
I export the result in 8k uhd and when I play it in mpchc, there are banding (as you can see in attachement) so I think that camera generate banding.
I generate the 8k uhd from a timeline 12k and from a timeline 8k uhd and the result is the same. There are banding in the two cases.
When you set the timeline to native resolution (12k) and set the display to 100%, you can see a very small banding from 2 or 3 lines.

Hi Didier,
can you provide a link to the footage from the screenshot. I would like to take a look. Thanks

here is the link (remove the space after https:) :
https: //drive.google.com/drive/folders/1ydtj9OKs1PAem2gL19MgdB1DmLsgrbNI

I make another test with another clip.
I export the result in 8k uhd and when I play it in mpchc, there are banding (as you can see in attachement) so I think that camera generate banding.
I generate the 8k uhd from a timeline 12k and from a timeline 8k uhd and the result is the same. There are banding in the two cases.
When you set the timeline to native resolution (12k) and set the display to 100%, you can see a very small banding from 2 or 3 lines.

The viewer you're looking at here is an mp4? What does it look like if you render to a delivery / mastering codec?

JB

here is the link (remove the space after https:)

Thanks. The footage looks absolutely fine. There is no banding in the sky what so ever. I can't even break it. I think whatever the problem is, it's with your setup. Maybe the codec you used for this render, like John suggested?

Mixmaster looks superb in 8K HDR, 12k center crop, envious. I couldn't break it either.

Thanks for your response.
I exported the clip in Quicktime DnxHR 444 12 bits. (10Go) and the banding is visible as you can see in the attachement.

I explain what I did and if someone can explain me what in my workflow create the banding, I will be happy to understand :
1 - open DVR
2 - import the original 12k clip
3 - add the lut (Gen 5 to extended video) on the node in the color page
4 - set the timeline to 8k uhd
5 - export clip in Quicktime DnxHR 444 12 bits
6 - play the clip with mpchc

You can do the same thing if you want because I understand if you don't believe me.

I can't understand how the banding can appear with that workflow except if the banding is in the original clip (or a bug in the raw dematrice or because of the rgbw sensor) ?

This is a small banding (but I can see it). If you set the blur radius in color page to 0.42 the banding will be very visible (This is not what I did usually. In several case I set blur radius to 0.47 whith my other blackmagic camera and I never see banding)

Hello,

I'm planned to buy a 12k lf so I get all information about this camera.
I downloaded the two samples on the "Blackmagic URSA Cine 12K Gallery" page. (sample with a girl and the big wheel)

When I import this samples on resolve, I can see banding (if I set the timeline to 8k or 12k).
This is outside images so this is not because of light.

Does someone see the same problem or I make a mistake in this so simple workflow ? or the 12 lf generate banding everytime ?

No banding at all looking at it on my monitor

Thanks Ben,

I don't understand what's happen with my configuration

I'm happy to said that I found where was the problem :

I haven't banding if I remove the checkbox "preference/decode options/use GPU for Blackmagic RAW decode"

I have a RTX 4090 and my drivers was version 560.94
I tried version 572.16 (studio and game ready) but result is the same so I need to let the checkbox unchecked.
The problem is a bug in the hardware decoding of BM raw with my graphic card.

Sorry to disturb you and I hope this will help some of you.

In my case, all is ok for this camera and I just wait the availability to buy one body version.

Thanks so much for the update, good you found what works.

Does this forum is the good way to report this bug to Blackmagic ?
If yes, I think the "Davinci Resolve" is the good thread, right ?

Great to hear you found the culprit.
More or less the same question was asked in the Resolve section
viewtopic.php?f=21&t=215816&p=1120909&hilit=bug+report#p1120548

Don’t expect a solution soon.

...I haven't banding if I remove the checkbox "preference/decode options/use GPU for Blackmagic RAW decode"

I have a RTX 4090 and my drivers was version 560.94
I tried version 572.16 (studio and game ready) but result is the same so I need to let the checkbox unchecked.
The problem is a bug in the hardware decoding of BM raw with my graphic card....

Make sure your graphic card is set to 10-bit and not 8-bit.

try these settings
In NVIDIA Control Panel under Change Resolution #3 Use NVIDIA color settings:
Desktop color depth = Highest (32-bit)
Output color depth = 10 bpc
Output color format = RGB
Output dynamic range = Full

and under Adjust desktop color settings check box Override to reference mode

and maybe check the 3D settings

The control panel for AMD is much better and more professional than NVIDIA but I think those setting may help.

Thanks Jeffrey for this informations.

I tried your settings but the banding remain with theses settings if I activate the GPU Raw decoding

I have the 4090 and also 65 inch 8K display, with GPU Raw decoding checked but no banding. I also tried a 12k center crop (1:1 pixel perfect).

I am viewing the BRAW image in Resolve Studio. I am also a user of MPCHC. It is a fine player but it has many, many filter options. I would suggest resetting that program to its defaults or looking carefully that you don't have an option interfering there.

Screenshot 2025-02-07 113321.jpg (248.31 KiB) Viewed 5754 times

Screenshot 2025-02-07 113226.jpg (294.15 KiB) Viewed 5754 times

Thanks Tom,

It seem that your parameters are exactly the same as mine. I don't understand why I have problem and you don't have.
The problem is not mpchc because I see the bandin in resolve before render and if I remove the gpu raw decode, then the file is ok on mpchc.

Thanks Tom,

It seem that your parameters are exactly the same as mine. I don't understand why I have problem and you don't have.
The problem is not mpchc because I see the bandin in resolve before render and if I remove the gpu raw decode, then the file is ok on mpchc.

But still not okay on Resolve's BRAW playback before render?

This is OK on resolve braw player. No banding.
I don’t know if resolve braw player use gpu decoding ?

Sorry, I meant Resolve Studio itself. You are saying that unchecking the GPU decode box in Resolve Studio preferences fixes the banding problem for you in the (full screen) Edit and Color page viewers, with the native BRAW clip on a timeline?

Sorry for my misunderstanding. English is not my native langage.
Yes, inside resolve studio, when I uncheck the GPU decode box, there are no banding in playback (on full screen or window) for a native braw file from 12k lf.
I have the same problem (banding if I check the GPU decode box) with braw file from blackmagic gallery page and from another user in internet (cars on night).
I hope my answer match your question

"If banding" is not a widely used term. Are you referring to "interlaced/field?"

Perhap's banding is not the good term.
I d'ont know the perfect term for this problem. You can see the problem on the two first pictures of this thread (display picture at there full size to be sure to see the problem). Perh'aps that someone can confirm that correct term is "interlaced/field" instead of "banding".

I can see faint horizontal lines in both of the posted screenshots, but not the BRAW file that you linked for download. I would caution, not to confuse this with the stripes on the green road signs for 'Dallas' and 'Galveston', as those are laminations inherent to the assembly of the road sign itself.

For the screenshot of the girl, more visible in the tree cover, I can see horizontal lines that resemble interlacing artifacts, which should not be there, and in fact were not there in the case of the BRAW camera file (for cars at night).

This is why my suspicions are with video memory and the screen capture itself, or the Windows configuration of screen buffers used for the screenshots.
I would suggest running the Windows DirectX Diagnostic tool.

1.) Win +R
2.) dxdiag
3.) Save All Information

Post the generated file contents here for review.

At this point, nothing points to the BRAW camera file as containing banding artifacts. I also am not seeing banding artifacts when using the Resolve Studio viewing windows, at any scale or resolution up to 12K 1:1 (100%).

Banding is identified by seeing the transition in light levels across an area that ideally should appear to be so gradual you can’t see where the light levels change. The opposite of banding would be a smooth area that differing brightness. The usual example can be in a clear blue sky in daylight or a lit painted wall. The sky near the zenith is a different shade than the sky near the horizon naturally, but you don’t see the boundaries of the transitions in the illumination. Banding occurs when you can see exactly where an illumination change goes from a slightly brighter to a slightly darker shade. It’s unmistakable and unnatural. It usually occurs because the sensor has insufficient latitude to record a smooth transition. At one time sensors may have only used 8 bits to describe up to a 256 shades of one colour. But when you record in BRAW, you have 4096 shades of one colour. Monitors usually only allow you to see an image with 10 bits of one colour or 1024 shades. There are a lot of colours at your disposal though so these numbers got into the millions and billions when you consider all possible digital hues mixing together.

Banding might occur when you are processing the image data. Resolve uses 32bit data internally in its calculation of colours to avoid creating banding. But if your deliverables are a lower quality such as 8bit, it might show some banding. ProRes 444 uses 12bit colour and ProRes 422 uses 10bit colour. As mentioned BRAW uses 12bit log derived from a 16bit linear value and some other cameras record higher values.

Unfortunately Rick, banding is a generic term that can be anything. What you are describing is just one type, false contours. Other types include scan lines, CMOS smearing, striping, screen tearing, interlace lines to name a few. When people try to describe artifacts that they see that others are not, we have to use more precise terms to be sure we are looking for the same thing.

Good, Tom.

It looks like a player scaling artifact to me. I had a similar issue with the MPHC in clips with added filmgrain (Dehancer) - the scaling created horizontal stripes. Changing the scaling algorithm and/or changing the playback video renderer did resolve this.
Could be that bicubic scaling would introduce artifacts for example as the algorithm has a sharpening effect due to its nature.

I would also try different scaling algorithms in Resolve. Lanczos for examples does sharpen even more than Bicubic. Catmull-Rom might be better suited, or a softer algorithm like Gaussian or Bilinear.

@Robert
I think that if the problem was just in the player, the problem must remain when I uncheck the GPU decode raw in resolve before rendering. Are you agree with that ?

@tom
I did the dxdiag but there are 1839 lines. Not possible to post this here and not possible to join file because this is .txt extension
You can download it from here : https://www.swisstransfer.com/d/7f24d9e ... d685eb99af (available for one month)

I think that if problem was in video memory and the screen capture itself, or the Windows configuration of screen buffers used for the screenshots then uncheck the GPU decode raw in Resolve must not fix the problem.

For me all is ok now after unchecked the GPU decode raw in resolve so I don't need more explanation. I suspect a bug in gpu decoding. I don't understand why you have not the same problem with same hardware ...
I hope a new version of resolve or raw library will fix this.

@Robert
I think that if the problem was just in the player, the problem must remain when I uncheck the GPU decode raw in resolve before rendering. Are you agree with that ?

It could be that GPU decoding will introduce slight artifacts which get than amplified in the player.

Yes Robert, I'm agree but in this case using gpu decoding is not acceptable for me. I prefer slower playback and final rendering without artefact even if theses artefacts are visible only on some player at some window size.
I think that my worflow will be to use the gpu decoding when I work on resolve and at the end just before rendering, I will uncheck this checkbox (I will put a post it in my forehead to be sure to don't forget )

It looks like a player scaling artifact to me… a softer algorithm like Gaussian or Bilinear.

In my tests, I found Gaussian gave me the best results.

@tom For me all is ok now after unchecked the GPU decode raw in resolve so I don't need more explanation. I suspect a bug in gpu decoding. I don't understand why you have not the same problem with same hardware ...
I hope a new version of resolve or raw library will fix this.

Do you set your Resolve project settings to 5120 x 1440 ?

Monitor Model: Dell U4919DW
Monitor Id: DELA10D
Output Type: Displayport External
Current Mode (NVIDIA): 5120 x 1440 (32 bit) (60Hz)

No.
5120 x 1440 is the resolution of my main monitor.
I use resolve in full screen on this monitor.
I set the project setting to differents values from 12k to 4k and the banding appear with 12k and 8k but not in lower definition.

Mac or PC?

Mac or PC?

As he is using a RTX 4090 it’s a PC.

Thanks, I didn’t notice that. These days, I just look at the sig for such information.
So, there goes one theory of mine.

As he is using a RTX 4090 it’s a PC.

Right

No.
5120 x 1440 is the resolution of my main monitor.
I use resolve in full screen on this monitor.
I set the project setting to differents values from 12k to 4k and the banding appear with 12k and 8k but not in lower definition.

I assumed you were referring to the Dell 49 inch monitor for the banding, not the Zalman 22 inch gaming monitor.

You said in an earlier reply to Bunk Timmer

@Bunk Thanks for this explanation. You are right this banding appear only when percentage are not 100 or 50 or 25 or 12.5
So the banding seems to be an artefact of the resize algorythme. I don't understand why.
If the clip is 12k and I want to resize to cine4k the coef is 33.33%
I tried to export as cine4k resolution from a 12k timeline and the banding doesn't appear on the rendered clip but it appear on the timeline at 33% or 34%.

Have you tried setting your project settings equal to the Dell monitor's 5120x1440? That's the way I would do it anyway, and then switch the project settings to cine4k for the render. Regardless, I couldn't provoke banding with your file on my display with any project setting or viewing scale.

And one other note, if you're running dual monitors simultaneous, there's different scaling for each, different aspect ratio, different screen resolution, different ports, Display Port and HDMI. That's a lot of variables uncertainty. If so, I would disconnect one of them and try again.

Yes, 5120 x 1440 is the resolution of the Dell Which is my working monitor.

If I set the timeline resolution to 5120x1400, I will get an image with bad ratio. This is not what I want.
I don't know if disconnect the zaman monitor will fix the problem but I need this second monitor so this is not an option for me.
Thanks for your help but for me the problem is fixed. I just need to uncheck the GPU decode raw option before rendering. This a good workaround for me.

Now, I just wait for bm 12k lf arrive in france shop.

Good news this morning. The camera is available so I will get it in some hours.
I will can make tests by myself