2.39 output blanking is failing QC

Using Davinci Resolve 17.4.5 Build 7.

Universal QC, Amazon QC and Netflix QC have flagged blanking issues on three different projects and we can not figure out what is happening.

When applying the 2.39 blanking preset in 3840x2160, the output sizing blanking reads...

Top: 276.653
Bottom: 1883.200

QC says the measured lines are 276 x 1882 but they should be 277 x 1883.

So we type in those numbers to override the preset. Next QC report says it is now 277x1882. Fixed top but not bottom.

We are checking this before export and then after export and the video is always cut off by the blanking as expected. I have no other tool to measure the blanking in Davinci.

What is really weird is that when we incrementally change the blanking amount by decimal places (.200, .400,.600 etc), the visual blanking line does not change on the whole numbers. For instance from 1882 to 1883 the actual blanking edge does not move until we get to 1882.400. The blanking moves next at 1883.600

Why are we getting fractional readings even when typing in whole integers? It appears to be rounding down. Why would it not just count the lines? We can't display 2/10ths of a line.

QC is not playing nice about this. Any idea what is happening and how we can fix this?

1) dont use the preset. the preset is wrong (rounding error)
2) go to the timeline sizing and punc the pixels manually.

3) that is PRECISELY why I'm not asking for more "presets" but the ability to simply punch two numbers in teh blanking custom: 1920x804. Done. And it will pass qc.

Thanks Walter.

If you check my first post you will see that I am punching in the numbers, exactly as QC is asking for. And still getting failed for bottom line off by 1 pixel. When we manually change the numbers the actual matte line does not move on whole numbers.

This is why I'm confused as to what is happening.

QC says the measured lines are 276 x 1882 but they should be 277 x 1883.

So we type in those numbers to override the preset. Next QC report says it is now 277x1882. Fixed top but not bottom.

I will context that QC report. We do 1920x804 that multiply by two is 3840x1608, that is an accepted blanking.

we do the proportion:
2048:858 (DCI spec)= 1920:x

that yield: (1920*858)/2048= 804.373 (rounded to the nearest even number 804). this does not get rejected.

for 4k/UHD
4096:1716 (DCI spec)= 4096:x
(3840x1716)/4096=1608.75 --> rounded to the nearest even number 1608 (double 804).

this pass QC every time.
forget about 2.39, the actual DCI spec for cinema (and tv...) is

2048/858=2.3869463869463869463869463869464/1.... a hefty not even number....

that is, (2160-1604)/2=278 (top blanking)
2160-278=1882 bottom blanking.

1882-278=1604 active lines.

this is how we set it up AFAIK.

Thanks again, Walter.

According to QC, Universal Studios has a very specific line count required for 3840x2160 with a 2.39 matte.

QC insists that 276 top and 1883 bottom are the correct matte lines.
So the active lines are 3840x1607. Not 1608.

The spec sheet they sent us says this...

16x9 Letterbox (2.0:1, 2.35:1, etc.) – Letterbox versions are to be consistent with theatrical exhibition aspect ratios. Matte settings should be set according to spec CM-001C. Universal will not accept any line variance from those stated in spec CM-001C for the first and last line of active picture. Matte lines should not fluctuate within the feature. If delivering a theatrical version and an extended or unrated version, then matte lines must be consistent between versions.

Of course they did not provide the spec sheet for CM-001C, so we are going off of the QC report. I have requested a copy of it but so far no one has responded to the request.

My issue is when I type in those specific numbers, the bottom line still comes up 1 short at 1882. It has happened twice now. As described in an earlier post, if I nudge the value by .100 decimal values, the matte line does not change on the whole number. It changes at 1882.400 and them again at 1883.600. So must I type in 1884 to get a true line at 1883? Is there a software tool any one can recommend that can read the active lines of video so I can check this before resending it again? I may have to send them a test clip.

Seems very strange and not very intuitive.

Thanks again, Walter.

According to QC, Universal Studios has a very specific line count required for 3840x2160 with a 2.39 matte.

QC insists that 276 top and 1883 bottom are the correct matte lines.
So the active lines are 3840x1607. Not 1608.

The spec sheet they sent us says this...

16x9 Letterbox (2.0:1, 2.35:1, etc.) – Letterbox versions are to be consistent with theatrical exhibition aspect ratios. Matte settings should be set according to spec CM-001C. Universal will not accept any line variance from those stated in spec CM-001C for the first and last line of active picture. Matte lines should not fluctuate within the feature. If delivering a theatrical version and an extended or unrated version, then matte lines must be consistent between versions.

Of course they did not provide the spec sheet for CM-001C, so we are going off of the QC report. I have requested a copy of it but so far no one has responded to the request.

My issue is when I type in those specific numbers, the bottom line still comes up 1 short at 1882. It has happened twice now. As described in an earlier post, if I nudge the value by .100 decimal values, the matte line does not change on the whole number. It changes at 1882.400 and them again at 1883.600. So must I type in 1884 to get a true line at 1883? Is there a software tool any one can recommend that can read the active lines of video so I can check this before resending it again? I may have to send them a test clip.

Seems very strange and not very intuitive.

sorry we deal with universal and whoever i telling you that the blanking is not correct is too stringent.

really nobody here send a master where teh active count has odd lines count....

besides, there is no technical reason whatsoever that it cannot be any of the lines we talked about, it will invariably end up in teh middle of the frame no matter what.

the even count will riproduce a better HD scale, the odd will not....

in any account, try to fight once more as "intended approved cinematografic blanking" then have it their way if really are that ****..... (not putting the real wording there...)

I checked with my online editor at Company3.... we do even number 3840x1608. no rejection.

Thanks for following up with that additional information.

TBH, we have always used the preset and have never had a fail. Like you said, it's still within the middle of a 1.78 container, so why does it matter? If we manually type it in, we come up with the same math that you do. 3840x1608. (276 top, 1884 bottom) So when they flagged it as wrong and said it should be 276 and1883, that really confused me. Which is why I asked for the spec document. Still waiting to get that.

We are dealing with Universal in London and that is where the QC company is.

If they push back again on this, I will mention what you have offered and see how they respond.

I still need to understand why the matte line doesn't move on whole numbers, but instead moves on decimal points in between. That is very confusing.

What do you expect after Brexit?

Seriously, any odd line number is suspicious, never read about that in any QC report.

What do you expect after Brexit? Seriously, any odd line number is suspicious, never read about that in any QC report.

Yeah, I'm with Walter -- no odd lines for us for blanking. We usually go for 1920x800 or 3840x1600, but we're dealing with old movies shot film, which have splice-lines and glue-marks a lot of the time. I'd rather do that than the old 2.35 or the newer 2.39 standard. 2.40 works with everything pretty well, and I have yet to have one kicked back from QC, at all. And that includes Canal Plus and Network 10, who are super-OCD and crazy.

Going with 1920x804 and 3840x1608 could also work just fine. I'd rather have an even number of blanking top and bottom, just because the math makes more sense.

We've delivered these for creating DCPs and never heard a peep from the distributors... so "I assume" it was all fine.

What do you expect after Brexit? Seriously, any odd line number is suspicious, never read about that in any QC report.

Yeah, I'm with Walter -- no odd lines for us for blanking. We usually go for 1920x800 or 3840x1600, but we're dealing with old movies shot film, which have splice-lines and glue-marks a lot of the time. I'd rather do that than the old 2.35 or the newer 2.39 standard. 2.40 works with everything pretty well, and I have yet to have one kicked back from QC, at all. And that includes Canal Plus and Network 10, who are super-OCD and crazy.

Going with 1920x804 and 3840x1608 could also work just fine. I'd rather have an even number of blanking top and bottom, just because the math makes more sense.

We've delivered these for creating DCPs and never heard a peep from the distributors... so "I assume" it was all fine.

I'm with Mark: for old movies 1920x800 or 3840x1600 (2.40:1) is safer. The 1920x804 and 3840x1608 is for elements that come from DCI specs theatrical (4096x1716 - 2048x858).

Screen Shot 2022-12-13 at 12.15.42 PM.png (114.05 KiB) Viewed 2591 times

Thanks for all the feedback, guys.

This is what they sent us. Redacted for privacy.

They just sent me the Universal spec for aspect mattes. It appears the QC operator is incorrectly counting the last line of active picture as the first line of blanking. Mystery solved on that part of it.

The spec even has guidance on Davinci Resolve based on Walter's first response about the presets not being correct. Red outlines were added by me for clarification.

Thanks everyone.

Screen Shot 2022-12-13 at 12.31.24 PM.png (156.92 KiB) Viewed 2586 times

They just sent me the Universal spec for aspect mattes. It appears the QC operator is incorrectly counting the last line of active picture as the first line of blanking. Mystery solved on that part of it.

The spec even has guidance on Davinci Resolve based on Walter's first response about the presets not being correct. Red outlines were added by me for clarification.

Thanks everyone.

Screen Shot 2022-12-13 at 12.31.24 PM.png

not surprised!!!!

Hey guys, what a timely discussion. we've had this discussion now for a while. which way to get the right matte.. We have multiple Resolve Editors that share projects, so we need a fool proof way to pass from one operator to another.

I had read somewhere b4, possible Dobly's site, about the mattes needed to be Even numbers. and logic dictates not having decimals (pixels vs Film)..So, thank you for all the info, and "rules" and the CM-001 (eric thank you).

So here's the bug in Resolve 18.1.1, in the output blanking window, i punch in these numbers, and it looks great. Save as a preset as we can now, and to check it, reset my settings, and reload it..except now, the numbers are gone.. but yet, the blanking/matting, still shows up on the image. tried it 3 times..

Save as a preset as we can now, and to check it, reset my settings, and reload it..except now, the numbers are gone.. but yet, the blanking/matting, still shows up on the image. tried it 3 times..

Yeah this bit me the other day too. It's super annoying and caused a bit of confusion when I made the preset in online, and then later in the color suite my colorist had no idea why there was blanking when the numbers said there wasn't.

This should be fixed, but it's not a critical issue now that we know if happens.

I just use an adjustment layer with crop, the numbers automatically adjust when switching timeline dimensions too. Alternatively save this out as a TIFF/PNG with alpha. Either can be named to show what it is and the dimensions involved.

Having a clip on the timeline is a good visual indicator of exactly what has been done and makes things very clear for anyone who should work on the project. Can easily be toggled on/off too so you can see what is being cropped out.

It appears the QC operator is incorrectly counting the last line of active picture as the first line of blanking.

There are two difficult problems in Computing Science.

They are Naming Things, Cache Invalidation, and Off-by-One Errors.

This should be fixed, but it's not a critical issue now that we know if happens.

Unfortunately, this issue is the bane of our existence, constant QC failures because our sources arent following industry standard numbers, and of courses, no way to know the OB is on or off.
So super important to our operation..

I just use an adjustment layer with crop, the numbers automatically adjust when switching timeline dimensions too. Alternatively save this out as a TIFF/PNG with alpha. Either can be named to show what it is and the dimensions involved.

Tom, So you see that as more of a benefit, than making the mattes in Photoshop first, and bringing them in, like your Tiff/Png suggestion?

Thanks guys,
marc

Tom, So you see that as more of a benefit, than making the mattes in Photoshop first, and bringing them in, like your Tiff/Png suggestion?

You can do Photoshop if you find that convenient, though if you use the crop method it will dynamically adjust to different timeline resolutions (provided the aspect ratio matches at least) and you'll know there won't be any resizing artefacts.

I'd not do any of the overlay techniques recommended here. If you do any sort of timeline grading it will affect the letterbox aswell. Grain, blacklevel.. all that stuff can make you fail QC.

Also depending on the resizing filter, outputing to a lower res delivery could introduce soft edges on the letterbox.

BTW, let me say that I've dealt with Visual Data in Burbank before as a QC company, and they're generally pretty good and reasonable. They understand that a lot of QC issues boil down to "interpretations" of the rules. We famously had a feature about 4-5 years ago where we initially got a 7-page list of problems, and after 2 phone calls and an email, I got it down to a single page and 8 problems, most of which were related to title placement. I consider that a victory.

No issues with color or blanking or shifts or anything like that. I think we had two NR glitches, and both were the dreaded "Overheating Trashcan Mac" era. Two 5-second fixes and we were good.

I'd not do any of the overlay techniques recommended here. If you do any sort of timeline grading it will affect the letterbox aswell. Grain, blacklevel.. all that stuff can make you fail QC.

Also depending on the resizing filter, outputing to a lower res delivery could introduce soft edges on the letterbox.

OK, well the adjustment layer with crop wouldn't be affected by any of that

I'd not do any of the overlay techniques recommended here. If you do any sort of timeline grading it will affect the letterbox aswell. Grain, blacklevel.. all that stuff can make you fail QC.

OK, well the adjustment layer with crop wouldn't be affected by any of that

a blanking on an adjustment layer will generate a failed Dolby XML.....

a blanking on an adjustment layer will generate a failed Dolby XML.....

Can it be removed just for that purpose then? Dolby xmls don’t carry blanking info do they?

Also from a Conforming Artists perspective I dislike working with adjustment clips. I use Place On Top a lot to insert new vfx, titles, etc. wich interferes with an adjustment clip alsways having to be in the top most track.

also deactivating for DoVi sounds like a roundtrip-based workflow with potential user errors. I prefer a solid bullet proof workflow, like setting the timeline resolution to the final delivery, or using Output Scaling for that. Those have other advantages aswell, like less scaling operators and therefore better preservation of image quality.

a blanking on an adjustment layer will generate a failed Dolby XML.....

Can it be removed just for that purpose then? Dolby xmls don’t carry blanking info do they?

Yes they do, that is why it fails. You must use resolve blanking of you want to be correctly representatives of the master real blanking.

If they don't match, you get an automatic red card

Also from a Conforming Artists perspective I dislike working with adjustment clips. I use Place On Top a lot to insert new vfx, titles, etc. wich interferes with an adjustment clip alsways having to be in the top most track.

also deactivating for DoVi sounds like a roundtrip-based workflow with potential user errors. I prefer a solid bullet proof workflow, like setting the timeline resolution to the final delivery, or using Output Scaling for that. Those have other advantages aswell, like less scaling operators and therefore better preservation of image quality.

I think Sven says it very well. We tend to try to go with renders without any letterbox borders when we can, and just render a 2048x858 or a 3840x1608 file, which has no wasted space, and then once that's done, take the same file and just render it as letterbox (no video / gamma / colorspace difference).

Thanks Marc.
for small projects or youtubers it might be more comfortable to do any of the previously stated options. But in a professional environment you have a department for deliveries anyway. You don't render a deliveriable from the grading project.

Thanks Marc.
for small projects or youtubers it might be more comfortable to do any of the previously stated options. But in a professional environment you have a department for deliveries anyway. You don't render a deliverable from the grading project.

Naaa, it's me. We delivered 46 feature films, 6 commercials, 10 promos, and 3 short films last year. We did it exactly the way I describe, and some of them had 2-3 different versions. What helps for us is having multiple Resolve stations in the room.

Most QC places run on a rule of 90% junior operators and 10% those who now what they do (or maybe even worse today).
You give them file normalised to -3dB peak, their software shows -2.99 duet to accuracy/rounding reasons and they mark file as failed. Zero common sense or understanding real meaning of this difference.
Don't ever take QC report values as something 100% true. Always question specially if results are suspicious.
Case from few weeks ago. One of the QC companies failed masters on R128 compliance. It looked suspicious as content owner specifically said masters were normalised. At the end what they done is QC each of 8 tracks as mono, without interpreting them as 2.0+ 5.1 mixes, so getting totally useless results. No comment, specially if this meant to be one of the oldest/most respected QC houses which typically does ok job.

Interesting read!

I'm a QC Supervisor at Visual Data, London.
Please feel free to get in touch in future, might be easier than bouncing reports back and forth via a 3rd party!

Best,
Ben