Bright Speckled Noise in Shadows

Hi!
I’m having an odd issue with grading my footage at the moment, I posted in the ACESCentral forum but am yet to receive a reply. Here is a Google Drive folder with Screengrabs detailing the problem.
https://drive.google.com/drive/folders/ ... sp=sharing

I'm using ACEScc 1.1 with C200 RAW footage into Rec.709, but when previewing and exporting, there is bright speckled noise all over the shadows. I tried disabling each colour node one by one, and bypassing all effects but in both cases the noise persists. In order to check if the problem was with the camera, I switched Resolve’s colour management to Davinci YRGB and applied a LUT- and the noise disappeared.
One strange note is that when the footage managed by ACES is playing (with no proxy mode enabled) the noise completely disappears and the shadows look smooth.

I’ve run out of ideas, so anything you guys know that could help would be much appreciated!

Regards,
Max

I'm very curious about this as well! Did you ever find a solution? I have a whole project with this problem, which mainly appears in footage with noise reduction. Not during playback or cacheing, only render. I've graded in ACEScc, but when I switch to ACEScct, it goes away in the renders. I just would have to adjust all the dark shots to switch to ACEScct at this point, and the transform built in doesn't make the switch seamlessly. Any input would be greatly appreciated.

Those speckles are likely negative code values that mess up the log() function.
Using a project level DCTL to clip (or better, roll off the negative code values) should do the job.

That would be amazing if there was a fix that simple. Do you know where I could get a DCTL that could help with this, Cary?

Save this in a .DCTL file:

Code: Select all

__DEVICE__ float3 transform(int p_Width, int p_Height, int p_X, int p_Y, float p_R, float p_G, float p_B)
{
    return make_float3(p_R < 0.0f ? 0.0f : p_R,
             p_G < 0.0f ? 0.0f : p_G,
             p_B < 0.0f ? 0.0f : p_B);
}


Unfortunately Resolve does no use the 'Input DaVinci CTL' at the project level for ACES Color Management.

But it does work if you select YRGB Color Management and use the ACES OFX!

Thanks for the response! I'm not very familiar with how to make a DCTL file. Do you know if there's one already made with this code that would work, or where would I go to find out how to make one?

Just open a text editor and put everything in the code block into the editor, then save it as a file with a DCTL extension.

Negative code values are not impossible if you consider the noise floor at 0. For Resolve to properly work it should either have a noise floor parameter or do the log calculations in the complex (principal) plane.

Hi Cary,

Thanks for that info again. I put the code you entered into a .DCTL text file and successfully imported it into Resolve such that it is recognized in the CTL LUT dropdown menu. However when selecting it, I received a warning "Error Processing Davinci CTL" with the path and filename I used. Is there anything in the code that might be causing it to fail?

Also, where would be the best place to apply this CTL? On a Timeline node, or just at the end of the problem shots' node tree, or someplace else?

Again, so appreciative of your help on this!

Hi Cary,

Thanks for that info again. I put the code you entered into a .DCTL text file and successfully imported it into Resolve such that it is recognized in the CTL LUT dropdown menu. However when selecting it, I received a warning "Error Processing Davinci CTL" with the path and filename I used. Is there anything in the code that might be causing it to fail?

Also, where would be the best place to apply this CTL? On a Timeline node, or just at the end of the problem shots' node tree, or someplace else?

Again, so appreciative of your help on this!

You can use the DCTL before you apply the ACES OFX.

So setup a project YRGB non color managed.

Then use the following node structure:

aces.jpg (34.83 KiB) Viewed 3384 times

The first node cuts the negative values. The second node uses the ACES OFX to go into ACES mode. Then use one or more processing nodes. The last node is again an ACES OFX but now converting ACES to the destination color space.

For the error, check the log file, under the support folder there should be a logs directory. Check the davinci_resolve.log file and look for the last set of DCTL error messages (the file can be quite large, but we only need the latest DCTL errors). Cut and paste it so we can take a look at it.

To OP: i've seen this as well, with a very similar setup, but with r3d footage. Also with DCTL. I notice you also work with RAW. After playing around, i got rid of the noise when i set the debayer settings to the highest quality.

Hi Cary,

I'm actually in ACEScc color space already for the project I'm troubleshooting. So the node tree you mention wouldn't apply per se, but I could put the DCTL node at the start of my tree for the problem shots.

I dug up what I think are the relevant errors from the log file. Here's that code:

-----------
[0x700012691000] | SyManager.DeckLink | INFO | 2020-10-14 21:57:00,482 | TranslateCfgToSDI 4235, pulldown 0.

Width = 3840, Height = 2160, FPS = 24
m_frameDuration = 1001, m_frameTimescape = 24000
[0x12cda4dc0] | IO | ERROR | 2020-10-14 21:57:00,596 | Braw HRESULT = 0x80000008
[0x12cda4dc0] | UI | WARN | 2020-10-14 21:57:00,597 | Failed to find value '0' in combo-box
[0x12cda4dc0] | SyManager.Signals | INFO | 2020-10-14 21:57:03,874 | Main view page is changed to 3
[0x12cda4dc0] | IO | ERROR | 2020-10-14 21:57:04,500 | Braw HRESULT = 0x80000008
[0x12cda4dc0] | UI | WARN | 2020-10-14 21:57:04,501 | Failed to find value '0' in combo-box
[0x12cda4dc0] | IO | ERROR | 2020-10-14 21:57:04,511 | Braw HRESULT = 0x80000008
[0x12cda4dc0] | UI | WARN | 2020-10-14 21:57:04,511 | Failed to find value '0' in combo-box
[0x12cda4dc0] | IO | ERROR | 2020-10-14 21:57:04,713 | Braw HRESULT = 0x80000008
[0x12cda4dc0] | UI | WARN | 2020-10-14 21:57:04,714 | Failed to find value '0' in combo-box
[0x12cda4dc0] | IO | ERROR | 2020-10-14 21:57:05,116 | Braw HRESULT = 0x80000008
[0x12cda4dc0] | UI | WARN | 2020-10-14 21:57:05,116 | Failed to find value '0' in combo-box
[0x12cda4dc0] | IO | ERROR | 2020-10-14 21:57:09,339 | Braw HRESULT = 0x80000008
[0x12cda4dc0] | UI | WARN | 2020-10-14 21:57:09,339 | Failed to find value '0' in combo-box
[0x70000f365000] | DVIP | ERROR | 2020-10-14 21:57:21,285 | DaVinci CTL compilation failed.
[0x70000f365000] | DVIP | ERROR | 2020-10-14 21:57:21,285 |
DVIP Exception: Kernel build failure
- API: Metal
- Call stack:
0 Resolve 0x0000000107005314 _ZN4DVIP14DataComparisonItEEbPT_S2_i + 96020
1 Resolve 0x0000000106fd4e18 _ZN4DVIP35c1_65d0bb02f29d7515f6f706d5a5b66519ItEEviiiiPT_S2_ + 21560
2 Resolve 0x0000000106fd5844 _ZN4DVIP35c1_65d0bb02f29d7515f6f706d5a5b66519ItEEviiiiPT_S2_ + 24164
3 Resolve 0x000000010712a8f4 _ZNK4DVIP11DeviceImageILNS_7APITypeE2ENS_15GPUResourceTypeILS1_2EEEE14GetNativeQueueEv + 21428
4 Resolve 0x000000010712b818 _ZNK4DVIP11DeviceImageILNS_7APITypeE2ENS_15GPUResourceTypeILS1_2EEEE14GetNativeQueueEv + 25304
5 Resolve 0x000000010710c301 _ZN4DVIP13DeviceContextILNS_7APITypeE2ENS_15GPUResourceTypeILS1_2EEEE23GetOpenGLTextureBlitterEv + 28641
6 Resolve 0x0000000106626cb6 _ZNK4DVIP16LUTApplicatorImpILNS_7APITypeE2EtE7RunDCTLERKNSt3__112basic_stringIcNS3_11char_traitsIcEENS3_9allocatorIcEEEESB_RKNS_12BtVariantMapERKNS_12DCTLUIParamsERKNS3_6vectorINS3_4pairINS_9LUTParamsENSI_INS3_10shared_ptrINS_12DeviceMemoryEEENS7_ISN_EEEEEENS7_ISQ_EEEERKSM_RSM_ + 150
7 Resolve 0x00000001065ebebc _ZNK4DVIP16LUTApplicatorImpILNS_7APITypeE2EtE17RunDCTLApplicatorERKNSt3__112basic_stringIcNS3_11char_traitsIcEENS3_9allocatorIcEEEESB_RKNS_12DeviceMemoryERSC_PKNS_12BtVariantMapE + 2540
8 Resolve 0x00000001068d14a9 _ZNK4DVIP15RGBCorrectorImpILNS_7APITypeE2EtE15RunRGBCorrectorERNS_12DeviceMemoryES4_RKNS_18RGBCorrectorParamsEPbRKNSt3__16vectorINS9_10shared_ptrIS3_EENS9_9allocatorISC_EEEEPS3_RKNS9_12basic_stringIcNS9_11char_traitsIcEENSD_IcEEEESP_RNS_10DCTLStatusERKNS_9CLFParamsERKNS_12BtVariantMapE + 21257
9 Resolve 0x00000001068aa414 _ZThn8_N4DVIP15RGBCorrectorImpILNS_7APITypeE2EtED0Ev + 1572
10 Resolve 0x000000010597d5c4 _ZN15MallocArrayPool12s_FreeMemoryEPv + 111668
11 Resolve 0x00000001057ffd1c _ZN3OFX17InteractMainEntryI40InteractiveWarpOverlayInteractDescriptorE24overlayInteractMainEntryEPKcPKvP20OfxPropertySetStructS8_ + 626796
12 Resolve 0x00000001052d9735 _ZZN22SyBMDKeyboardDeviceMac10DoPostInitEvENUlPviS0_15IOHIDReportTypejPhlE_8__invokeES0_iS0_S1_jS2_l + 234181
13 Resolve 0x000000010532c710 _ZZN22SyBMDKeyboardDeviceMac10DoPostInitEvENUlPviS0_15IOHIDReportTypejPhlE_8__invokeES0_iS0_S1_jS2_l + 574112
14 Resolve 0x00000001052ecf57 _ZZN22SyBMDKeyboardDeviceMac10DoPostInitEvENUlPviS0_15IOHIDReportTypejPhlE_8__invokeES0_iS0_S1_jS2_l + 314087
15 Resolve 0x00000001052eb3cb _ZZN22SyBMDKeyboardDeviceMac10DoPostInitEvENUlPviS0_15IOHIDReportTypejPhlE_8__invokeES0_iS0_S1_jS2_l + 307035
16 Resolve 0x00000001052e9789 _ZZN22SyBMDKeyboardDeviceMac10DoPostInitEvENUlPviS0_15IOHIDReportTypejPhlE_8__invokeES0_iS0_S1_jS2_l + 299801
17 libsystem_pthread.dylib 0x00007fff6bd6d109 _pthread_start + 148
- Build log:
Failed to load metal library.
Metal Error Code: 3 Domain: MTLLibraryErrorDomain
Metal Error Message: Compilation failed:

program_source:1360:1: error: expected unqualified-id
{\rtf1\ansi\ansicpg1252\cocoartf2513
^
program_source:1371:2: warning: missing terminating ' character
\'a0 \'a0 return make_float3(p_R < 0.0f ? 0.0f : p_R,\
^
program_source:1396:1: error: expected '}'
^
program_source:1355:31: note: to match this '{'
namespace DAVINCIIP_NAMESPACE {
^


Urgent message: Error Processing DaVinci CTL||ACES/BlacksRollofff.dctl
-------------------

To OP: i've seen this as well, with a very similar setup, but with r3d footage. Also with DCTL. I notice you also work with RAW. After playing around, i got rid of the noise when i set the debayer settings to the highest quality.

Hi Rick, good to know I'm not crazy. Are you referring to the R3D debayer? I'm working with Arri ProRes4444 footage, so that doesn't apply to me. I've got "Force debayer to highest quality" and "Force sizing" checked at render, which is the most I can apply I believe.

Hi Cary,

I'm actually in ACEScc color space already for the project I'm troubleshooting. So the node tree you mention wouldn't apply per se, but I could put the DCTL node at the start of my tree for the problem shots.

The problem is that Resolve has no pre-ACES handle for a DCTL.

program_source:1371:2: warning: missing terminating ' character
\'a0 \'a0 return make_float3(p_R < 0.0f ? 0.0f : p_R,\
^
program_source:1396:1: error: expected '}'
^
program_source:1355:31: note: to match this '{'

Perhaps some line formatting error?

This is exactly what should be in the DCTL:

dctl.jpg (28.85 KiB) Viewed 3361 times

To OP: i've seen this as well, with a very similar setup, but with r3d footage. Also with DCTL. I notice you also work with RAW. After playing around, i got rid of the noise when i set the debayer settings to the highest quality.

Hi Rick, good to know I'm not crazy. Are you referring to the R3D debayer? I'm working with Arri ProRes4444 footage, so that doesn't apply to me. I've got "Force debayer to highest quality" and "Force sizing" checked at render, which is the most I can apply I believe.

Turns out it was the ''decode quality''

Schermopname (19).png (458.13 KiB) Viewed 3355 times

This is what it looks like on my end in TextEdit. I've been saving as a Rich Text Document (.RTF) and then changing the extension to .dctl after. I've also tried Pages and saving as a .OTD (open text document) from Text Edit. All have the same result.

Screen Shot 2020-10-15 at 2.47.08 PM.jpg (66.39 KiB) Viewed 3331 times

Placed in the Davinci Resolve > LUT > ACES folder as Blacksrolloff.dctl

This is what it looks like on my end in TextEdit. I've been saving as a Rich Text Document (.RTF) and then changing the extension to .dctl after. I've also tried Pages and saving as a .OTD (open text document) from Text Edit. All have the same result.

Screen Shot 2020-10-15 at 2.47.08 PM.jpg

Placed in the Davinci Resolve > LUT > ACES folder as Blacksrolloff.dctl

Do not use rtf, use a regular text.

Do not use rtf, use a regular text.

Hey that fixed it! I got the DCTL working without error! That's the good news. The bad news is, within the color space my project file is in, the DCTL doesn't fix the issue no matter where I place it in the node tree.

Do not use rtf, use a regular text.

Hey that fixed it! I got the DCTL working without error! That's the good news. The bad news is, within the color space my project file is in, the DCTL doesn't fix the issue no matter where I place it in the node tree.

As I explained before you can't use it within ACES, it's too late, you need to use it before you get into ACES.

And for Resolve that means you need to use the ACES OFX as Resolve has no pre-ACES handles.

As I explained before you can't use it within ACES, it's too late, you need to use it before you get into ACES.

And for Resolve that means you need to use the ACES OFX as Resolve has no pre-ACES handles.

I see. I could be wrong, but it seems to defeat the purpose of ACES if you're not using it within Color Management with IDT's at your disposal. Or at least it creates more confusion for me! I think the moral of this story for myself is to just never use ACEScc in Resolve, but only ACEScct! A lesson hard won.

Thanks for your support, Cary! I very much appreciate your patience in helping me work through this.

sorry, duplicate posting.

As I explained before you can't use it within ACES, it's too late, you need to use it before you get into ACES.

And for Resolve that means you need to use the ACES OFX as Resolve has no pre-ACES handles.

I see. I could be wrong, but it seems to defeat the purpose of ACES if you're not using it within Color Management with IDT's at your disposal.

Resolve has two shortcomings wrt ACES in my opinion:
1. No pre-ACES handles
2. No custom IDT support

Thanks for your replies everyone!

It was fortunately one of those serendipitous problems that dissolved with an update.
The node-based (Davinci YRGB) ACES workflow is a great suggestion, Cary! Not only for the DCTL node but more precise qualifiers in pre-ACES space (among others).

Capture.JPG (37.46 KiB) Viewed 2961 times

I used to see this working with Sony footage, probably all footage, it seems to change every time they do an update.

My easy fix was just to lower the Lum V's Sat curve just a little on the left and BOOM! problem solved.

I believe it's a Saturation in shadows issue as well.