Ursa Mini 4.6k Blue LED light looks Magenta

I have encountered a very strange behaviour last week with the Ursa Mini. While shooting with RGB LEDs i noticed that the deep blue was captured as magenta by the URSA Mini. here is a photo i took with my smartphone:
https://www.dropbox.com/s/r69hzctkqa0mr ... D.JPG?dl=0

and here is a raw file captured by the ursa:
https://www.dropbox.com/s/f66f7b9uroirr ... 6.dng?dl=0

I have imported the Raw Clip into Davinci and the blue light looks also magenta. In Adobe RAW however the light is totally blown but it is blue.

I have also tested this with a Zylight RGB Headlight with same results the deepest blue that the light creates looks Magenta when Captured by the URSA.

It would be very helpful if someone owning a RGB Light and a Ursa-m 4.6k could test if this is a problem with BM-Colour Science or if it is a problem of my unit only. I have also reported this to Support but had no answer yet.

Pulled it into Resolve using ACEScct and the blue looks blue. I'm not sure what you are using to interpret the raw file, but from what I see in ACES, there is no obvious problem with your camera.

Screen Shot 2016-12-05 at 1.49.56 PM.png (997.22 KiB) Viewed 10273 times

Lights like this are tough as their spectrum is so narrow which can have odd interactions with Bayer pattern sensors. I'd say the Ursa Mini did pretty well, though it would be interesting to point some other cameras at the same scene to see how they compare. Many Sony cameras (such as the A7S FS5 and FS7) have particular problems with saturated blue lights.

For lights that bright, the shot is overexposed if you don't intend for them to clip. Close down your aperture and they won't clip as badly.

If you are displaying the shot using BMD's Rec709 LUT or Resolve Color Management to get it to REC709 most shots will look crap, especially in the highlights. For whatever reason, BMD's own REC709 interpretation is just ugly. You could switch to ACES for a better starting point for REC709 grading, or stay in DaVinci YRGB interpret the RAW as BMD Film and grade from scratch.

Mixinglight.com has a wonderful free 4 part series of videos about how to grade using ACES in DaVinci Resolve:
https://mixinglight.com/portfolio/getting-know-aces/

Also, if you are going to bother to shoot RAW and are concerned about capturing good saturated blues, I'd shoot full sensor rather than the UDH crop as you have here so that you can benefit from oversampling, especially in the blue and red channel which get half as many photosites on a Bayer pattern sensor.

EDIT: I also noticed in your phone shot of the camera's LCD that you've got your white balance set to 5200. For indoor/stage lighting you should be around 3200K (unless the location is lit with a daylight balanced LED or HMI)

Thanks Jamie

I really do appreciate taking the time! and thanks for the ACES workaround Tip seems i have to check out ACES.

Reading your lines one might think this is caused by a user mistake, but sadly this is not a user mistake, the camera shows deep blue light as magenta light, as the subject of the thread says. The workaround would have been possible when we had shot raw but we shot the day with prorez UHD.

Consider being on location that one day, seeing the blue lights in Magenta on all monitors, having to tell the lighting guy to change all blue lights to a more cyan colour, while most situations he had to change the programming wich took quite a while, im not sure if you would be as realaxed about this. This alone was a huge pain. Not to speak about the client.....

Guess i have to check ACES seems worth checking out. I use RAW workflows since back starting with the RED1 and to be honest i always take the appropriate application and import the raw files use Luts or create a look myself and go on from there. There has not been a need to interpret the footage before import and i hope there will not be a need in the future to do so. allthough as said ACES seems to be worth it (:-).

We shot 5200 because tungsten lights are quite out of fashion these days in modern circuses and theaters and stages, they use special HMI Lights and LEDs.
UHD was shot because we shot Prorez UHD all day and just for the test i switched to RAW to check what this would look like in Davinci. We do this job every year since back to 2011, we have shot this with, Red Epic, 5Ds, Sony FS700, Red Dragon, we have not encountered anything alike with these cameras.

these rgb leds are often used, you can also see them fe. on jimmy fallons tonite show, when the band performs.

It would be really great if someone could test his Ursa-Mini with a RGB light to see if this is the general behaviour of the camera or if this is my unit only.

I am sorry of hear you ran into trouble with a client watching. I hope they were forgiving.

Also, I did not mean to offend. All I had as data is the DNG still you shared and your statement that the lights should be red, green and blue. Setting the raw tab at 3200K for that image in Resolve, the light on the left is blue. At 5200K it skews toward magenta. So to me, that pointed to a white balance error as the cause of the magenta. However, this single still isn't a good test from which to make any conclusion as there is no concrete reference in it.

Shooting a color chart in sunlight with the camera color temp set to match would help determine whether your camera has a color issue.

Hi Jamie

No Problemo and thanks again for helping, very appreciated.

Yeah the client was not too happy but they understood and to be safe we tried to balance the blues into cyan on the lights themselfs as a workaround.

I do think the problem only occurs with leds ive done the colour chart, looks o.k. i have not encountered strange colours in real life situations either. But as mentioned with my RGB Headlights its reproducable. As you mentioned the narrow spectrum of the leds is probably the cause of this. Maybe this is a glitch in the colour science. It looks like a "bad" secondary grade (;-). when i find time i will do a test with gel and tungsten and hmi lights to see how they behave.

Support answered this morning. They have past it to the developers.. i should wait until the next firmware comes out, seems to have major improvments. Lets see.

I shoot mainly in these kind of situations (clubs, festivals,..).

Blue looks blue on mine, I had absolutely no issues.

Valentin, are you shooting raw or ProRes in those situations? What setting do you use for the temperature and tint typically?


Sent from my iPhone using Tapatalk

Indeed, I should have mentionned

Prores, default daylight settings (5600k, Tint 10).
I find it to be a great balance for all the different colors.

I do think the problem only occurs with leds ive done the colour chart, looks o.k. i have not encountered strange colours in real life situations either. But as mentioned with my RGB Headlights its reproducable. As you mentioned the narrow spectrum of the leds is probably the cause of this. Maybe this is a glitch in the colour science. It looks like a "bad" secondary grade (;-). when i find time i will do a test with gel and tungsten and hmi lights to see how they behave.

Support answered this morning. They have past it to the developers.. i should wait until the next firmware comes out, seems to have major improvments. Lets see.

I appreciate you posting and raising the question with BMD. It will be very interesting to learn whether future firmware makes a difference. That unfortunately seems to be the default position we find ourselves in with this camera.

I don't have that type of colored LED light to test with my own Ursa Mini, but it really has me curious to find one. I'd like to know how my Ursa Mini responds before I find myself shooting at a concert or similar event that uses them. I don't want to have to find out too late that I should have created a monitoring LUT specific to the situation ahead of time.

On top of the color questions, the way the extreme saturation of those lights responds in your cDNG is certainly not easy to grade. The saturation clips and reverses into negative values in odd ways. Like a bad secondary, as you noted, or the quantization and clipping from a poorly matched LUT.

You mentioned shooting the same location on other cameras in years past. If you have ungraded shots under those same lights that you can share, it would be a useful comparison.

Indeed, I should have mentionned
Prores, default daylight settings (5600k, Tint 10).
I find it to be a great balance for all the different colors.

Very useful to know. Strange that the raw clip from Thomas appeared to have more accurate color at a tungsten balance. If you have any clips you are able to share a download link for, it would be great to take a look at them in Resolve in comparison to the clip from Thomas.

Sure, here's two printscreens, from a Prores 422 Film file.




There's red, green, blue, magenta,.. All look great in my opinion ^^

Only thing that caused a problem with my WB setting was magenta sometimes, just when it its white skin. I could have improve it by changing the Color Temp but honestly, this color is so ugly and rarely used, it's not worth it ^^



But this problem is common, and the thread's op problem is different (even tho I don't think his dng looks off by checking Jamie's screen. It only looks blown out to me ^^).

EDIT

Well, I opened it myself in Resolve and indeed, it's not blue ^^
Didn't change the WB.



With Rec709, it looks blue.

Even on your smartphone picture it looks weird, blue above, and magenta on the ground.

EDIT 2

Okay I did the test myself at home and he's right. I'll post the result in the next post.

I wonder if setting the white balance to 5600K would have brought that violet back to blue?


Sent from my iPhone using Tapatalk

I did a comparison with my a7s (slog2). Same Color Temp by the way (5600K), and it's really obvious ^^

The blue is totally off.

That is at 5600K, with 3200K it's normally blue on the UM.

But still, that's one hell of a difference compared to the a7s at 5600.

I think the reason I didn't notice this in my previous experiences is that the blues I had to deal with weren't that extreme.

Doesn't help us that manufacturers are notoriously poor at meeting their published colour temperatures and an accurate meter isn't cheap.


Sent from my iPhone using Tapatalk

Thomas, you mentionned that it's going to be improved in the future firmware, did BM answer that to you ?

Thomas, I'll get an Osram Kreios RGB LED this week, then I can do some tests with my URSA Mini 4.6k.
Did you use any filters like IR/UV or polarizing?
Once I forgot a polarizer on my Canon XF305 and it turned the blue light of a projector into magenta.

I'm using an IR filter, just did a test without it and it made no difference.

4BM to REC709 V3 Lut

But without a rec709 conversion, it's magenta.

I don't understand :/

Because you have to process the footage and put it into the correct color space. That's how all cameras that shoot LOG work. It's grading 101 with LOG cameras. This isn't a Canon DSLR.

I suggest you dive into a few hours of online tutorials about processing not only BM cameras shooting RAW but RAW grading in general. Then it will be very clear what to do. There's absolutely nothing wrong with your camera.

I'm not shooting raw, I'm shooting ProRes.

A7s can shoot Log, why is the Blue correct there then ?
I could plug an external recorder on it to get a proper ProRes 10bits, blue would still be blue as it should be.

Here, the camera just shows a completely different color.

I'm not shooting raw, I'm shooting ProRes.

A7s can shoot Log, why is the Blue correct there then ?
I could plug an external recorder on it to get a proper ProRes 10bits, blue would still be blue as it should be.

Here, the camera just shows a completely different color.

At this point, you're on your own. You really need to learn more about these cameras instead of hop on here and argue with everyone telling you what your error is.

Good luck.

Hey, not arguing here, just looking for answers

I just don't get why other log cameras show blue like blue, and why the 4.6k shows something completely different.

For example, I just converted my footage to the Arri Colorspace and Gamma, now it looks fine with it.

I mean, I shouldn't do that additional step. Blue should look blue and that's it, right ?

I can understand that there are differences between color spaces, but here it's quite a lot.

EDIT: of course this is a really extreme case and we won't bother 99% of the time, but for Thomas, myself and many other people who shoot in these kind of situations, it's too bad that we don't get a proper blue out of the box instead of having to convert to another colorspace.

So, Thomas, to help you with your issue, here's a workaround:

In Resolve, create a Node and apply the "Color Space Transform", then use these settings



You'll still get a log image, but with a proper blue this time.

Also I just noticed that if you output to the BMD Film colorspace instead of the 4.6k one, your blue will be correct (but still a LITTLE BIT more magenta than the Arri one)

This is the out of the box BMD 4.6k color space



Now, with the old BMD Film Output



So yeah, there is definitly something wrong with the 4.6k colorspace in this kind of situation !

Hey, not arguing here, just looking for answers

I just don't get why other log cameras show blue like blue, and why the 4.6k shows something completely different.

For example, I just converted my footage to the Arri Colorspace and Gamma, now it looks fine with it.

I mean, I shouldn't do that additional step. Blue should look blue and that's it, right ?

I can understand that there are differences between color spaces, but here it's quite a lot.

My (admittedly non-expert) guess is that it has something to do with how BMD has chosen to roll off overexposure/overstauration in their log processing yet provide a more saturated and contrasty image than you get in Arri LogC or even from the log files from the original BMCC as Valentin showed in his post. One selling point of the Ursa Mini 4.6K that many have commented favorably upon is that the log files are much easier to grade because they aren't as flat as earlier BMD cameras. Perhaps this is the cost?

In Resolve I took the sample DNG from Thomas and used a Color Space Transform node that changed the clip to Arri LogC and, as you found, the color of the light on the left goes from magenta to blue.
Then, I put one node before the Color Space Transform and another node after it. If I decrease contrast or decrease saturation in the node AFTER the Color Space Transform, the saturation simply decreases with no color shift. If I decrease contrast or decrease saturation in the node BEFORE the Color Space Transform the blue vector rolls toward magenta. The same thing happens if instead of a Color Space Transform node I use the BMD 4.6K Film to Video v3 LUT in its place. I recommend taking a look for yourself in Resolve and watch the vector scope as you decrease contrast in the node before versus the node after, it's quite an interesting experiment.
(BTW: in ACES with the ODT set to REC709, the light is blue and there is no need to fuss with LUTs or Color Space Transform nodes and having to grade before versus after them. EDIT: This applies as long as the IDT is set to BMD 4.6K Film not BMD 4.6K Film v3.)

Based on this I think it could be that BMD have made some type of trade off in the log color space of the 4.6K by making it easy to grade to a nice looking image with most typical scenes and light sources, but the down side of that choice shows up with these kind of highly saturated LED light sources where BMD's 4.6K log math rolls the blue channel toward magenta when it's close to clipping.

Perhaps a relatively simple solution might be for BMD to provide more color and gamma options in the Camera Raw tab (and in the camera), similar to the larger range of color and gamma options that you get with Red raw.

EDIT #1: So this is very interesting. I tested the Color Space Transform going from the current BMD 4.6 Film v3 color space to the original 4.6K Film color space. And guess what? It's still BMD log color space, but the light is now blue instead of magenta. So, whatever BMD changed to create 4.6K Film v3 is what is causing the magenta look compared to version 1 and REC709 (and, well, pretty much everything else). Another way to achieve this is in ACES by setting the IDT to BMD 4.6K Film rather than BMD 4.6K Film v3.

Screen Shot 2016-12-06 at 11.34.50 PM.png (68.09 KiB) Viewed 9927 times

Screen Shot 2016-12-06 at 11.34.58 PM.png (587.68 KiB) Viewed 9927 times

EDIT #2: So I checked some daylight images of people from my camera switching between BMD 4.6K Film and BMD 4.6K Film v3 and the skin tones are much much better in v3. Again, this points to some kind of trade off where the price for better colors in normal shooting scenarios comes at the expense of that blue LED light.

thanks everyone for the input, very cool indeed!

The reason why the blue light is also "redish" on the floor on that Mobile-Phone shot is because the floor surface is red.

Im in a hurry and cant do any testing at the moment, i have 2 big shoots this week, tomorrow we'll be in the mountains 2000 Meters and i dont even have the gear ready amd my clothes together.... brrrrrr.

Try to do a little Recap:
Chances are high that all current 4.6k Cameras are affected. right? Lets also see what Roberts results are with his LED Light.
It seems that the latest BMD Film V3 Colour-Space is causing the shift, as mentioned by Jamie and V1 is showing correct colours, also interesting in that screenshot it seems like the blues are not as blown out, they are similar in intensity as the other colours.
There are workarounds for DNG Sequences as shown by Valentin, but when shooting Prorez the only safe thing would be to shoot in "Video"Rec709, right?

About the mail from Support, that is what they wrote:
"As commented on the forum by the product manager, I assume that the team is working on a new build which would be released in the future. Therefore this build might carry significant improvement to you camera picture quality without the need of returning the camera.."

There are workarounds for DNG Sequences as shown by Valentin, but when shooting Prorez the only safe thing would be to shoot in "Video"Rec709, right?

Our workarounds work with ProRes as well, I did the test with my footage.

So yeah, with v1 showing the correct colors, now we know it's a little glitch with v3.

(So, I suggest some people should dive into a few hours of online tutorials about being nice and not condescendent )

+1 to Valentin. It works with ProRes.

Using Resolve's TrimLUT function, I created a monitoring LUT you can load into the Ursa Mini 4.6K that will change 4.6K Film v3 to 4.6K Film v1.

https://drive.google.com/open?id=0B1bnfGXKHqpcSVhBQ1p4NV9jRWc

Although you could bake this LUT into your ProRes capture, I would recommend just using it to monitor. Then, in Resolve you can either work in DaVinci YRBG and use the Color Space Transform node, or work in ACES and set your IDT to 4.6 Film (not 4.6K Film v3).

Based on my testing so far, I would only use this workflow if you are encountering problems with very saturated blue lights as Thomas was shooting. For all other shooting under normal daylight or tungsten, 4.6K Film v3 will yield better results, especially for skin tones.

If anyone has time to test whether there is a difference between tungsten lights gelled to deep blue versus a blue LED, it would useful to know if there is a difference.

Probably this shift is due raw interpolation. I find this glitch when played with BMMCC test files and color spaces too.
Try to convert your raw to prores (bmdfilm/gamut) and put it aside to original raw file and you'll see that there will be a difference in color in some situations.
if you didn't set timeline color space as bmdfilm, or if you didn't set input color space for RAW file (right click menu-input color space) and you just put a color space transform or input LUT with bmdfilm/gamut input to it, it will produce shifted wrong colors for unknown reason
You can use LUT or Color Space Transform OFX as input only for ProRes files but not for RAW.

example with bmdfilm to video v2 lut applied:

Screen-Shot-2017-01-19-at-10.23.30-PM.jpg (438.22 KiB) Viewed 9663 times

Screen-Shot-2017-01-19-at-10.23.18-PM.jpg (433.81 KiB) Viewed 9663 times

Screen-Shot-2017-01-19-at-10.22.54-PM.jpg (415.22 KiB) Viewed 9663 times

same goes to test file shared in first post. if you set project settings input and timeline color space as bmdfilm.v3 but check bypass input color space in right click menu in raw file, the colors are shift to magenta.
p.s. and no, it is not a simple WB shift, its something different and more complicated.

project-bmdfilmv3.jpg (62.63 KiB) Viewed 9655 times

bypass.jpg (57.91 KiB) Viewed 9655 times

This has happen to me a number of times with bright blue leds. I'll try to find the screenshots


Sent from my iPhone using Tapatalk

Sony cameras like the F3, A7S or A7R II tend to do very strange things if there are strong sources with a certain wavelength of blue:



Graeme Nattress from RED says that such blue LEDs are out of gamut for many color spaces. What he does is mapping them back into a regular color space without too much of a shift with IPP2, have a look at this thread:

http://www.reduser.net/forum/showthread ... view-Build