URSA Broadcast G2 - Setup for Rec.709 broadcast legal

Hi there,

what's the proper way to make a URSA Broadcast G2 (FW8.5) put out straight broadcast legal 1080p59.94 in Rec.709 on it`s main SDI output? Chip cropping needs to remain at UHD/4K or higher though to at least reasonably match the lenses focal lengths.
Dynamic range is set to video.
Turning all LUTs off as per the manual looks weirdly washed out, and turning on the Gen5 to Video LUT on the output does not look right either. Gen5 to 2020 seems to go in the right direction, but that can't be correct...
Mixer is an ATEM HD8 ISO if that helps; though I can't find anything pertaining to color space in it, so I assume it just passes it 1:1.

The video setting with no LUT is most correct but theres a pretty big difference between 709 and “video”. It’s not illegal in any way but video is less saturated and includes fairly aggressive higlight compression.

Good Luck

The video setting with no LUT is most correct but theres a pretty big difference between 709 and “video”. It’s not illegal in any way but video is less saturated and includes fairly aggressive higlight compression.

Good Luck

It's *a lot* less saturated, to the point of being almost unusable for a green screen.
So what would be a better setting for non-HDR live production? Extended Video? Can the HD8 handle and display it correctly on the MV?

Maybe there is something else going on. I'm using the G2's pretty frequently these days and I'm not noticing anything crazy vs. typical broadcast setups that use a knee. Definitely don't use Extended Video as that matrix specifically desaturates green and magenta.

Here's a comparison of 709 v BM Video. For live work I definitely prefer BM Video but I can see how in the hyper controlled environment of a GS shoot 709 might be preferred. It's actually pretty simple to create a true 709 LUT in Resolve that can be applied to the SDI out of the G2.

Good Luck

709 v bmv.jpg (83.2 KiB) Viewed 5916 times

It's actually pretty simple to create a true 709 LUT in Resolve that can be applied to the SDI out of the G2.

Is there a resource for obtaining a 709 LUT for a G2? I don't use Resolve or other editors.

Thank you.

Keep in mind that for most applications BM Video is a really good choice. A true Rec709 matrix only holds about 2+ stops of dynamic range over mid grey where BM Video is holding 5-6. Just 2 stops over one can see how the highlights in 709 are gone, but BMV is holding things together nicely. The difference in the color matrix between 709 and BMV is slight, the difference is primarily a function of the tone mapping inherent in BMV.

Enough evangelizing, here's a link to the 709 LUT. The camera should be set to BM Film and the LUT should be applied to the SDI output or wherever else you might need it.

https://www.dropbox.com/scl/fi/1lu6kqjx ... qbelw&dl=0

Good Luck

2over.jpg (271 KiB) Viewed 5789 times

That's great, much obliged!

Here's a comparison of 709 v BM Video. For live work I definitely prefer BM Video but I can see how in the hyper controlled environment of a GS shoot 709 might be preferred

Very interesting Howard. Forgive my naïve question, since I'm not a multicam guy; and I've never hardly switched from Film/Lut workflows on my UBG2 - do you then further colour correct in Atem? Since that is pretty far off Rec. 709 on those charts? And did the OP do so, to get 'a lot less saturated than that' possibly?

Since that is pretty far off Rec. 709 on those charts?

It's not. A Chroma Du Monde references SMPTE color primaries (and secondaries) at 50% saturation which is why Resolve's scopes have a 2x vector gain. DSC Labs asserts that the dyes can't be printed at 100% saturation which is why the chips don't look like SMPTE bars.

According to the scopes this is about as close if not closer than most other Rec709 cams which include the Grass, Sony, or Panasonics that I use regularly.

All you can do in the Atem CC is primary correction (white balance), you can't do any matrix correction without affecting the WB.

Good Luck

709.jpg (117.38 KiB) Viewed 5699 times

Thanks. I'm only really used to EBU bars with 75% and 100%. Pretty irrelevant for me now; especially since the UBG2 can't record them. Are you saying then that the difference in your 1st image is down to a vector line up distinction. Again forgive my naivety here and perhaps not following you fully.

In terms of color there’s no difference between 75% SMPTE bars and 75% EBU bars. In RGB the primaries are all 192 (in 8 bit) and the secondaries are combined, cyan for example is 0, 192, 192 in both SMPTE and EBU.

The colors on the chart aren’t meant to hit 75% SMPTE or EBU levels, more like 37.5%. The VS 2x gain is so that those 37.5% saturation chips properly align with the 75% targets on the scope.

The difference in the first pair of images illustrates the difference between Blackmagic Video and a mathematical Rec709 transform. The early BM LUTs were often criticized. Folks would use instead the Log C to 709 LUT because it featured tone mapping, wheras the BM LUT did not. BM (Gen 5) Video is a pleasing transform, not an accurate one.

Are you clicking and or scrolling on the images?

Good Luck

Hi Howard,

Thanks for your informative posts. Can you explain to me if I've misunderstood, the best combinations of method we should be using to match a UBG2 to a Sony 2500 or similar running in SDR?

Thanks,
Ben

Are you clicking and or scrolling on the images?

Scrolling. Is that my error?

Are you clicking and or scrolling on the images?

Scrolling. Is that my error?

I think if you click on the image it will expand and what I’m explaining will be more explanatory.

Good Luck

Hi Howard,

Thanks for your informative posts. Can you explain to me if I've misunderstood, the best combinations of method we should be using to match a UBG2 to a Sony 2500 or similar running in SDR?

Thanks,
Ben

That’s a tough nut to crack. Sony’s and most broadcast cams have weak cyan and green channel saturation. Ostensibly because they want skin tones to look good so they desaturate the opposing colors.

Matching cameras in a vacuum is an impossible endeavor. Additionally the UBG2 doesn’t have a user adjustable matrix so all secondary correction must be acheived via LUT. You could probably create a 709 LUT in Resolve that only had ~2 stops of knee and it would likely be fairly close. The reverse option is to match the Sonys to the UBG2s which is a lot easier.

Good Luck

Related to this: what is the actual signal flow inside the UBG2 with regards to Color space?

There is the basic "dynamic range" switch (Film/Extended Video/Video) and there are the three outputs (LCD, Front and Main SDI) that can be assigned LUTs.

Does the "dynamic range" switch actually change the raw signal from the sensor to the outputs (recording, LCD and SDI) or does it simply insert an LUT there...?

Also, has any one here already tried to match a DSLR (Nikon Z8) to the UGB2s this way? Needs to be live obviously and with as little correction on the DSLR as possible...

N-Log uses a Rec2020 gamut with the N-Log transfer function (gamma). To match to BM Video use a Color Space Transform that uses 2020/N-Log as the input and BMV as output color space/gamma with tone mapping enabled. To match to 709 use a 2020/N-Log as input and 709 as output/gamma with tone mapping disabled.

With on the wire formats WYSIWYG. For the most part there is no difference between Prores and 10 bit on the wire. Raw is always raw regardless of LUT so there is a small difference between raw and log, but a massive difference between log and 709.

Again it's pretty easy to create a conversion LUT within Resolve. Use the Color Space Transform OFX to create the conversion then right click the clip and Generate LUT. Of course it won't be a perfect match but it will get you in the ballpark quickly. Additionally the CST is using maths and will be less destructive than trying to manually drive the conversion.

Good Luck

I think if you click on the image it will expand and what I’m explaining will be more explanatory.

Thanks Howard. Do you know, I never knew a simple click works? I've been downloading images before; there's a 1st world problem of mine for you.

I can see the colours are more or less close in the 1st image; I was referring to the difference in luminance and white point. I'm also aware BMs internal color corrector is no matrix; but I meant the need to match cameras in a multicam environment. Presumably without recourse to 3D luts one has to use Atem for this, and from what I gather those facilities can work:

So therefore I was asking; since I don't use BMV nor extended, or work in multicam/live broadcast, what's your common approach to this? You say you much prefer to use BMV over traditional Rec.709, I can see why now; just interested in the balance aspect between cameras. I'm no engineer unlike you and only work in single camera setups, although I have done a lot of multicam as director, where I have hired a vision mixer/engineer then. So your knowledge here, genuinely interests me.

As an aside my old JVC HD101E has SMPTE bars as opposed to EBU, they were quite useful. Although in the digital age bars were less important than on analogue BetaSP cameras, I always recorded 30 seconds at the head of a tape. I'd still like that in the UBG2, although it's unnecessary; but I used them too for wild VO sometimes; black can be confusing. But that's just me and it is unimportant now. But I wonder also why BMD cameras have EBU bars and not SMPTE?

I can see the colours are more or less close in the 1st image; I was referring to the difference in luminance and white point.

That is interesting. I'm not seeing any of the tone mapping in the video that I see in my Micro G2s, maybe BMV is handled differently between the 2 cams or it could be a firmware rev. The OP's observations with the UBG2 reflect what I'm seeing with the Micros. Ultimately the tone mapping is a good thing because the cams don't have a knee. I run our LDX86Ns with a powercurve (knee) that starts at 70% and has a max input of 600% which is little over 2 stops. It's not as extreme as BMV but it means that if (when) the lighting shifts dramatically the shader has a little protection against skin tones blowing out and the highlights falloff a little easier.

The video is really good, but you'll notice that at about the 30 minute mark to get the vectors properly placed he still relies on a LUT, he does a little tweaking of color prior to that but just messes up the WB that was meticulously set prior. You can also see how close the UBG2 is right out of the box versus the Panasonic which starts miles out. If the UBG2 was that far out there'd be no way to get it close without a matrix or LUT. Even when the Panasonic is dialed you'll notice the green/cyan is not fully saturated. We get around this by pushing the saturation prior to the matrix adjustment so that we don't have to push cyan/green so hard and instead reel in the red/blue. his observations about extreme signal manipulation causing noise are spot on. He does make a couple errors regarding placement of the grey and black values which are not 50 or 0 but overall it's a great vid.

Overall I'm not obsessive about the cameras matching a chart perfectly, often pushing them too hard creates more problems than it solves. Then there's the effects of the knee and how it affects saturation. I can have the Panasonic robos and the Grass cams perfectly matched on a 6 stop chart but how the colors react in a 10-11 stop scene will be wildly different. Obviously the Grass cams handle it much better. One of the nice things about the CDM charts are all the skin tone chips, you can setup skin detail in cam without needing talent in front of the lens. I'd consider skin detail a more important feature update than a matrix if anyone at BM is listening.

I mostly direct Multicam, and program servers now, but I've been an EIC since the late 90's. Colorbars used to be far more important (mandatory) in the days of analog. If your deck/cam was referenced you'd need to adjust the timing, SC phase, setup, gain, and sat in the proc amp just to get it looking correct on the scopes, then do the same for your monitors. With digital, bars are always perfect, subcarrier phase and line loss don't exist anymore, black is black, white is white, signals are relocked, the signal either makes it or it doesn't.

Good Luck

Very detailed and very interesting Howard. Yes I'd forgotten he used a 3Dlut, as well. I remember having to line up analogue bars too in Avid mostly. Great insight.

Hi Howard,
Please could u provide more details on the UBG2 setting.. I'm using Canon HJ17 HD lens
When I put the dynamic range to Extended Video with no LUT there is some green on the image. I'm not familiar with the scops and the color grading. So I will set my camera with your setting and LUT. But please provide more details my camera is attached with fiber converter to the Atem constellation 8k with the CCU for live broadcast with external ingest recorder.
I need HD resolution 50p what's the best quality and codec to use along with the dynamic range "Film" and your LUT you provided earlier?
What's the setting for RED, Green, and blue in the Atem software?

I've been using these cameras regularly over the last couple years and I just don't think there's any need to jump thru hoops to get a nice picture out of them. Set the DR to video, set the temp, set the tint, done.

It's common for monitors to have a cast which makes it hard to dial out color casts with any precision. In lieu of scopes I'd use a split wipe to compare the camera white source (paper, card, towel, sheet, etc.,) to a generated white background. Generated white is white regardless if it looks blue, green, or magenta on the monitor. Match the camera to the generated white and you'll be in the ballpark.

Remember it's easy to send things off into space and start chasing your tail, less is more.

Good Luck

Something to add that is important, turn off the gammut compression to have a good range of colors in ideal lighting situations.