URSA Mini 4.6k The Original™ Magenta Cast Issue

Roberto's sample without lens looks like it just has a tint and I was able to correct it in AE perfectly neutral. But the second sample with the cars has a magenta corner in the upper right and slightly weaker in the lower right.
Initially I was trying to use his white wall sample as a calibration sample for canceling out the magenta cast on the car sample. But that did not work, obviously - because they do not share the same edge color cast. I was able to remove the vignetting but the cast in the sky remained.

And after grading the car sample the magenta cast in the corners look even worse.

Roberto's sample without lens looks like it just has a tint and I was able to correct it in AE perfectly neutral. But the second sample with the cars has a magenta corner in the upper right and slightly weaker in the lower right.
Initially I was trying to use his white wall sample as a calibration sample for canceling out the magenta cast on the car sample. But that did not work, obviously - because they do not share the same edge color cast. I was able to remove the vignetting but the cast in the sky remained.

And after grading the car sample the magenta cast in the corners look even worse.

Hi Robert, I did several tries too and I concluded that Ursa's footage is all different clip from clip, I can't just grab a still of my correction in resolve and apply in batch.

That green image with magenta vignetting perfectly illustrates my point. I'm just not explaining it well enough. Bear with me...

I see that you and many others are making adjustments to make the issue much clearer, that's understandable. I'm suggesting that one cannot make those wild adjustments without causing the issue, because it changes the intensity of the vignetting in a fashion that is not equal in all three channels, unless the image is perfectly white balanced.

Consider this:

1. Make a perfectly gray image in a paint program. Like this (RGB values are 199,199,199):
   
   
   
2. Add vignetting and some noise to it. We now have a perfectly gray image with vignetting. No color cast. True?
   
   
   
3. Here's what the image looks like in a scope, perfect!
   
   
   
4. Lets introduce a severe, uniform green cast by setting the Tint Control to -100:
   
   
   
5. Green cast scope:
   
   
   
6. Now, if you try to fix it, with for example the Offset color wheel (affecting the whole image) instead of with the tint control, the darkest corners of the vignette will no longer align:
   
   
   
7. Try the same with a curve (affecting the whole image) and it's even worse:
   
   
   
8. Let's add some saturation (again, affecting the whole image) so that everyone will really be able to see it. The corners are now a totally different shape:
   
   
   
9. What does this image look like? You guessed it, magenta corners. Without using any "ovals" or anything fancy.
   
   
   
10. White balanced:
    

Moral of the story; If you have a uniform cast (corners or no corners), use the tint control, it works.

Of course I too want BMD to fix it so that we don't need to fiddle with the tint control, that goes without saying. But it seems to me that in this case it's not the sensor that is to blame.

Edit: For peace of mind, added the white balanced image nr10.

After WB the center (which has no magenta cast), I tried fixing the magenta corner cast in this image using only the tint adjustment in Capture One Pro:

Blackmagic URSA Mini_1_2016-04-17_1520_C0016_000023 1.jpg (905.76 KiB) Viewed 8540 times

By bringing the lower right magenta corner back to neutral, I get a green overall image:

Blackmagic URSA Mini_1_2016-04-17_1520_C0016_000023 4.jpg (927.32 KiB) Viewed 8540 times

By selectively "painting out" the magenta in Capture One Pro, I can neutralize it and get an overall balanced image, but a GLOBAL TINT or GLOBAL BIAS adjustment does not work with this problem. A mask of some type MUST be made. The LCC tool in Capture One Pro does this more precisely, but it can be done manually:

Blackmagic URSA Mini_1_2016-04-17_1520_C0016_000023.jpg (906.93 KiB) Viewed 8540 times

Here is the original DNG: (I'd love to see how this can be fixed in Resolve. Even Power Windows would be welcome at this point!)
https://www.dropbox.com/s/2kd8418iyy8k3 ... 3.dng?dl=0

Moral of the story; If you have a uniform cast (corners or no corners), use the tint control, it works.

amen, this is essentially the point. Pretty basic and simple to understand. not to mention that a camera should work correctly without any adjustment.

To date, the official stance from Kristian Lam of BMD, is that the magenta corners issue is basically normal for this camera and sensor: " ...the magenta corners exhibited in some of the samples posted here. This is not a sensor or hardware issue. Some of you have already correctly pointed out that this has to do with how the light rays are hitting the sensor. This is especially true for lenses that are not telecentric by design. You can find many examples of this happening on many other camera systems and also a myriad of ways to deal with this. The reason there seems to be cameras that exhibit this phenomenon more than other cameras is to do with the combinations of different types of lenses, different aperture or focal length, and even variances between the same model of lenses. " viewtopic.php?f=2&t=46112&start=500 (bottom of page 11)

Please post your single most favorite Magenta Corner fix below:

If you don't have a solution, but think this would be helpful, please reply with a +1 to keep this post on the front page and the possibility of finding a solution closer at hand. Any software/post solution should not be seen as letting BMD off the hook for what could very well be a hardware/manufacturing problem which would need to be addressed and fixed. ( I personally believe this is the case since there are clean cameras with no magenta issues reportedly out there) A short term software fix would at least provide a way to fix the problem immediately with the tools we have at hand and back up the claim made by Kristian Lam that there are "a myriad of ways to deal with this".

The people at BMD try to do their best (I hope), but I guess they are in a situation now, that they cant handle....hundreds of cameras coming back, and maybe no good sensors from the sensor manufactur available. Much money in the game...they just want to get the cameras out there...defective or not...

Here is the original DNG: (I'd love to see how this can be fixed in Resolve. Even Power Windows would be welcome at this point!)
https://www.dropbox.com/s/2kd8418iyy8k3 ... 3.dng?dl=0

Unfortunately I would classify your DNG as the real deal, sorry. It's not the same issue as the typical "magenta corners because of off-tint on a vignetted image". In fact, if I white balance the car as best I can, the tint slider ends up at 0 (temp at 5020). It looks sort of ok, but still has a magenta cast in the bottom corners, mostly to the right.

Do you think a matte box would have helped? If the issue is stray light hitting the sensor at extreme angles, maybe it can be controlled.

Roger, your example has nothing to do with the BM issue.

You were confusing enough in your post to hide your actual proposition so that people didn't debate it.

You said that if you have a tint issue (affecting highlihgts more than shadows) and attempt to use a global offset to correct it, that you would be over-correcting darker areas and produce a magenta corners because the corners are darker due to normal vignetting.

That is true, but it does NOT describe what is happening in the camera.

The corners are not magenta because they are darker. They are magenta because the angle of incidence of the light is being affected in the green channel differently than the others.

Most if not all cameras and lenses exhibit vignetting, and yet they do not usually have this kind of problem.

Now today you are acting like an authority and telling people whether their cameras are borken or not based on your incorrect paradigm, and that is unhelpful to those of us who have broken cameras that need a fix.

Consider this:

Most lenses vignette more when the aperture is wider, for lower f numbers. And yet, this magenta corner issue is worse around f/8. That's because of the non telecentric aspect of these lenses, which has to do with the angle of the light coming onto the sensor.

Also, I see a lot less of it in shorter focal lengths, like 18mm, and yet they still vignette just as much.

Therefore, your theory does not apply here, and you should not be telling people their cameras are fine based on a faulty premise.

Here is the original DNG: (I'd love to see how this can be fixed in Resolve. Even Power Windows would be welcome at this point!)
https://www.dropbox.com/s/2kd8418iyy8k3 ... 3.dng?dl=0

Unfortunately I would classify your DNG as the real deal, sorry. It's not the same issue as the typical "magenta corners because of off-tint on a vignetted image". In fact, if I white balance the car as best I can, the tint slider ends up at 0 (temp at 5020). It looks sort of ok, but still has a magenta cast in the bottom corners, mostly to the right.

Do you think a matte box would have helped? If the issue is stray light hitting the sensor at extreme angles, maybe it can be controlled.

...what about putting a circled cut out LEE 1/4 green filter over the front of every lens ....but seriously, its not the job of the custumer to make this camera work....

It wont work because the problem is not an overall cast or tint. It has to do with the angle of light hitting the sensor and that changes with focal length and aperture, and varies lens to lens. It's an interaction.

Roger, your example has nothing to do with the BM issue.

You were confusing enough in your post to hide your actual proposition so that people didn't debate it.

You said that if you have a tint issue (affecting highlihgts more than shadows) and attempt to use a global offset to correct it, that you would be over-correcting darker areas and produce a magenta corners because the corners are darker due to normal vignetting.

That is true, but it does NOT describe what is happening in the camera.

The corners are not magenta because they are darker. They are magenta because the angle of incidence of the light is being affected in the green channel differently than the others.

Most if not all cameras and lenses exhibit vignetting, and yet they do not usually have this kind of problem.

Now today you are acting like an authority and telling people whether their cameras are borken or not based on your incorrect paradigm, and that is unhelpful to those of us who have broken cameras that need a fix.

Consider this:

Most lenses vignette more when the aperture is wider, for lower f numbers. And yet, this magenta corner issue is worse around f/8. That's because of the non telecentric aspect of these lenses, which has to do with the angle of the light coming onto the sensor.

Also, I see a lot less of it in shorter focal lengths, like 18mm, and yet they still vignette just as much.

Therefore, your theory does not apply here, and you should not be telling people their cameras are fine based on a faulty premise.

All valid points, even if a bit harsh. But the "no lens, paper against the window" test almost always seemed correctable and pointed to another issue, didn't it? Anyway, if I have expressed myself in an authoritative manner I apologise, since I'm no authority on the matter. That's why I tried to formulate my posts mostly as questions or subjective thoughts. I've been writing technical instructions for offshore consultants for the past two weeks and maybe that "matter of fact" style rubbed off here. I'll shut up now.

Guys my original image with magenta corners has nothing to do with lenses, since I shot it with no lens, body only in front of my studio large white backgroud, on calibrated lights.

Roger, I have no problem with you personally, you seem like a nice guy.

But as BM figits about what they are saying about this issue, misinformation that leads away from their responsibility to fix it is unhelpful in my view. I'm angry about the situation, and certainly not at people who are legitimately trying to help as you are.

Still, I'm trying to clear the fog for the truth, and that's where my intent is. I really do apologize if I seemed too harsh, I meant to be direct, and nothing I said was intended to be about feelings or personal.

I'm sure we all want good cameras, and to go out and shoot already.

Here is the original DNG: (I'd love to see how this can be fixed in Resolve. Even Power Windows would be welcome at this point!)
https://www.dropbox.com/s/2kd8418iyy8k3 ... 3.dng?dl=0

Unfortunately I would classify your DNG as the real deal, sorry. It's not the same issue as the typical "magenta corners because of off-tint on a vignetted image". In fact, if I white balance the car as best I can, the tint slider ends up at 0 (temp at 5020). It looks sort of ok, but still has a magenta cast in the bottom corners, mostly to the right.

Do you think a matte box would have helped? If the issue is stray light hitting the sensor at extreme angles, maybe it can be controlled.

The sun is behind me and a bit to the left side. There is no direct light hitting the lens, but I will conduct more tests using a matte box, choking down the light entrance as much as possible. I also really want to see if I can get the problem to show up greater on the left side. So far the problem always shows up more on the right side, which to me indicates some sort of sensor predisposition, which if the problem is simply lens cast, should not be the case.

Lets break this down in a clear way....

Lens Falloff: Lenses do show falloff and do vignette, for instance some wide angle lenses with a image circle designed for a 35mm sensor will show vignetting on a super 35mm sensor like the ursa mini. Some falloff is normal and varies from lens to lens. This has nothing to do with the issue we've seen.

Sensor global casts: Some have mentioned that sensors lean towards a cast and that Alexa leans green, (I think that isn't actually true ie, it's more of a lens color or ND and less of a sensor cast with alexa) but in any case if a image is leaning a few points green or blue, because of a lens or more likely because of ND it's very easy global fix and shooting raw is useful as the tint control makes this fast and simple. The Ursa mini 4.6 from what I've seen posted shows global casts towards magenta that vary from camera to camera and are correctable using raw corrections.

Sensor area specific casts: THIS is the problem. This issue is misrepresentation of the color is specific areas. The sensor thinks the road in the right corner in this image is magenta and the raw data is reading the wrong color in that part of the image. I'm leaning towards thinking this is a sensor color shading problem caused by some part of the sensor array being misaligned with the lens during the manufacturing process.

Is it fixable in resolve with RAW?

No. Using raw tint and temperature corrections to balance out the image will only clean up part of the image and will leave a maganta color in the affected areas. Cleaning up the magenta areas will push the rest of the image green.

Is this fixable in resolve with power windows?
No. A power window just pushes color or luma into a image in a specific area, and can be softened around the edges of the correction to blend in. A regular vignette like a lens vignette can definitely be helped with a power window that pushes a bit of light into the corners, this is normal and happens all the time. This issue is completely different, in that the sensor reads color incorrectly in the right side and corners. A power window pushing some green into those areas to balance the magenta would need to match the sensor pixel by pixel only in the area affected as the sensor moves over different color, and shifts both in the area affected and in intensity with different light and this is impossible for a power window.

What about mapping the sensor and applying a inverse to correct it?
From what I've seen the issue presents differently with different focal lengths and lens designs and within a lens with different T-stops. For this to actually work a white or grey card would need to be shot with every lens or lighting change and the camera would have to remain locked relative to the light.

Anyways I've balanced the DNG posted above with Raw corrections only, you can see the One Shot chart hitting the vectorscope and the color is very far away from ok.

Screen Shot 2016-04-29 at 1.55.50 PM.jpg (226.77 KiB) Viewed 8449 times

Screen Shot 2016-04-29 at 1.59.30 PM.jpg (65.04 KiB) Viewed 8449 times

my ursa vignette even without the lens mounted, and it also has more magenta cast on corners, and not simmetrical too..

The definitive test that remains to be seen in a controlled way, is testing a UM4.6 that is known to have the Magenta Corners issue (mine) against one that has never exhibited those anomalies, side by side with the same lenses under the same conditions similar to the test that Fahnon Bennett and I conducted here: viewtopic.php?f=2&t=46898&p=273698&hilit=hunting#p273698

If anyone in the New York Metro area has a magenta corners free EF mount UM4.6 camera that has tried at least one of these problem lenses listed here: viewtopic.php?f=2&t=47465 and are willing to meet up to conduct this test, please reply here and also send me a PM. As the owner of a clean camera, you actually hold the key in getting to the bottom of this issue. Evidence needs to be shown conclusively that there are actual differences between cameras (good vs bad) with regard to this issue OR we will show that ALL OF THEM will show this problem to some degree under the same conditions with the same lenses.

Since anyone that already has a good camera may not have much incentive to donate their time and equipment for this monumental and epic test of all tests, I will throw in a couple of beers and at least one good joke at a nearby watering hole. PLUS! Ashok Vardhan also a forum member, has donated food too! But before I hand over those beers, food and joke, you'll need to pre-qualify your camera with exemplary examples of magenta freedom that have confounded less fortunate cameras! Sounds like a fair deal?

Benton, you are my hero. Right the _____ on.

Benton, you are my hero. Right the _____ on.

Thanks Kyle!

And Mine!!!!

Man I think we already know who's the next ultimate beta tester for BMDs next camera

You sir are doing us all very good... if that phrase makes any sense... i'm not english native.

And Mine!!!!

Man I think we already know who's the next ultimate beta tester for BMDs next camera

You sir are doing us all very good... if that phrase makes any sense... i'm not english native.

Thanks Francisco!

Yes, Benton. Thanks for taking this on. There are many of us that will be very interested in your results if you can find a good camera in your area.

What's a beer party with food to match? I am willing to pitch in along with Benton, to buy the person who comes forward food/snacks when Benton takes you out for the beer. Benton I am based in LA but can always share the bill through PayPal. Curious to know as well.

What's a beer party with food to match? I am willing to pitch in along with Benton, to buy the person who comes forward food/snacks when Benton takes you out for the beer. Benton I am based in LA but can always share the bill through PayPal. Curious to know as well.

Thanks for your fantastic offer Ashok! That definitely sweetens the pot!

I certainly hope I get someone who comes forward with a clean camera. And now that there's a real incentive, I'm really going to want to see some tests of open pavement with the sun to the rear and left side using one of the problem lenses. Whatever camera I test next to as a control, has to already have been put though the magenta corners stress test.

The sensor thinks the road in the right corner in this image is magenta and the raw data is reading the wrong color in that part of the image.

The sensor doesn't see colors. It measures photons in shades of gray. The CFA breaks it down. Colors are then generated or blended if you will in debayering. The RAW data isn't reading data - it IS the data which in return is made into what you perceive as colored pixels.

Trying to explain this in simple terms. What's your read on the magenta vignette cause?

+1

+1

Thanks Subrata! I have a feeling that getting multiple viable fixes for the Magenta Corners Issue using software is not as bountiful as Kristian initially stated. I'll settle for just ONE approach right now. But I do have hope that this will be solved one way or another.

That'd be great. I went back and looked at the DNGs from the short you shot for Blackmagic and its a flawless image edge to edge. So I'm curious to see if there's any variation with the other bodies you have.

Yes I have a few now for the new show. My crew did a quick QC on them and nothing "obvious" came up but I haven't really gone looking. I'll be sure and take a look on Tuesday. And for sure I'll hear about it if it goes' bad on set on Tuesday too

I'll try and do Benton's backlit scenario, though it's Melbourne heading in winter...soo....

jb

To JB: Were you able to test the new cameras? Are they all working well on the new show?

Just out of curiosity... How many nodes on average, using Resolve would everyone here consider "normal" to process a raw DNG? On any camera for that matter. I corrected one of Alessandro's magenta DNG's in three nodes and one might consider his camera as functioning normally and put it down to Resolve operator error...

Hi Aaron,
Maybe you worked on the white wall no lens test.
My camera (now at BM's UK dept for RMA) showed just uniform magenta cast + vignette when shooting without lens.
Easy correctable.

Issue is that with every lens I tried I got magenta corners.
The worse is when using a 35mm stopped down above f8.0
Let me know how you would deal with an image like this


I'll be happy to share the whole dng sequence if you have some tips to correct it to an acceptable level

Hey Allesandro,

I definitely wasn't suggesting you weren't using Resolve correctly, there is definitely an issue with these cameras.

I was trying to imply that it's possible BMD are trying to defend their faulty product by being able to say it's recoverable in Resolve if the operator knows how...

PM me you email address and I'll send you the DNG that I graded.

+1

Sent from my FP1U using Tapatalk

Hey Allesandro,

I definitely wasn't suggesting you weren't using Resolve correctly, there is definitely an issue with these cameras.

I was trying to imply that it's possible BMD are trying to defend their faulty product by being able to say it's recoverable in Resolve if the operator knows how...

PM me you email address and I'll send you the DNG that I graded.

...thats why they added tint and temperature controls

+1

+1


Sent from my iPhone using Tapatalk

www.ryanhamblin.com
www.brainstem.tv

Benton, given how non-reponsive BMD has been regarding this issue, are you beyond your return period for the URSA Mini 4.6K? Can't you return to your dealer to get a refund?

Benton, given how non-reponsive BMD has been regarding this issue, are you beyond your return period for the URSA Mini 4.6K? Can't you return to your dealer to get a refund?

I'm working closely with my dealer AND BMD. Both are very responsive, but don't yet have a clear solution to the problem. I basically have until the middle of May to return the camera and get a refund if I want, which might be the better path at this point until this issue is sorted out. But if they can assure me that the problem has been identified and they can ship me a good third camera before the middle of May, I will take that route.

A bit depressed now...i don't understand why every new camera released from BM have problems and users have to deal with this problems in post. Really don't understand why. For sure a useless post this, but this is a forum and this is just a vent...

Step right up, step right up, step right up,
Everyone's a winner, bargains galore
That's right, you too can be the proud owner..
of a clean Non-Magenta Corners UM4.6k!

(Thanks and apologies to Tom Waits!)

Come on folks, there must be one good UM4.6k out there in NY land!?

Benton... Will the joke be any good tho?

I need to run your pavement tests... Might get a chance early next week.

Benton... Will the joke be any good tho?

The pressure is high, but I'll do my research and make sure it's a real beer spewer or at least I'll try!
I may even stretch it and have two or three jokes waiting in the wings!

+1

I was planning to buy a UM46. Just stayed back because of the circus going all around. Can anyone clarify the following things?

1. Howsoever BMD ignores, I can see the magenta corners in some cameras. Can they be corrected in post? If so, how?

2. Seems some cameras are magenta free. What's the percentage of those m-free cameras? So that I can work out a probability of getting a camera which is trouble free instead of trying my luck?

3. In case I land up with a camera which is not to my satisfaction, will BMD change the camera for free? (This is troublesome in India for them, but why wouldn't I insist?)

4. I run all my other cameras (BMCC, BM4K, 5D M III) on these lenses. They are:
Rokinon/ Samyang 14, 24, 35, 50, 85
Zeiss Contax (cine modded) 25 (2.8), 35 (2.8), 50 (1.7), 85 (2.8), 135 (2.8)
Canon 70-200 2.8 non-IS
Tokina 11-16 2.8
Which of these lenses will work without any problems on a random UM46 that I buy?

Would be happy if someone from BMD posts something.

+1

+1


Sent - Tapatalk

+1

+1

I was planning to buy a UM46. Just stayed back because of the circus going all around. Can anyone clarify the following things?

1. Howsoever BMD ignores, I can see the magenta corners in some cameras. Can they be corrected in post? If so, how?

2. Seems some cameras are magenta free. What's the percentage of those m-free cameras? So that I can work out a probability of getting a camera which is trouble free instead of trying my luck?

3. In case I land up with a camera which is not to my satisfaction, will BMD change the camera for free? (This is troublesome in India for them, but why wouldn't I insist?)

4. I run all my other cameras (BMCC, BM4K, 5D M III) on these lenses. They are:
Rokinon/ Samyang 14, 24, 35, 50, 85
Zeiss Contax (cine modded) 25 (2.8), 35 (2.8), 50 (1.7), 85 (2.8), 135 (2.8)
Canon 70-200 2.8 non-IS
Tokina 11-16 2.8
Which of these lenses will work without any problems on a random UM46 that I buy?

Would be happy if someone from BMD posts something.

All of your lenses have been know to cause the magenta corners issue. I think it's best to wait to order a camera till after they have identified the problem and can assure that all future cameras will be magenta free. It's still not 100% certain that magenta free cameras actually exist. It could be that the only reason we don't see any magenta cast in some samples like John Brawley's promo which was shot at T1.8 to T2.8. More testing needs to be done, but it might be that shooting wider open may alleviate the problem. Not that this is a fix, but it could be a temporary workaround.

Hi here is another way to prove or disaprove bmd cliam... How about renting a true telecentic lens to shoot a test on your camera Bentonw that would show us a no. Magenta cornets if indeed that is even possible... I really dont think there is a free magenta camera i. Nyc if you ask me since in my opinion all the free magenta cams were only pre production cams .