URSA Mini 4.6k The Original™ Magenta Cast Issue

John, may I ask you which lens you did use for picture no. L1000574?

Jeff

The Leica 50mm F2 APO. You seem to have had the exact same reaction as Kholi when he saw that photo a while ago

JB.

Thank you. This is a fantastic lens. Complete different rendering compared to Zeiss.

full raw 5600k in camera adjusted -2 stops with adobe

1.jpg (116.01 KiB) Viewed 22136 times

here is my test. I followed in the footsteps of Adam, ND on window, white paper, no lens, camera was approx 10 cm/5 inches away from the white paper

So please tell me if i am doing something wrong hereI opened
I opened a single dng in photoshop 6. reduced the exposure by -2 stops. applied global correction toward green to get it as natural as I can and it looks like I get some magenta cast on the upper left corner.

am I correct?

here is my test. I followed in the footsteps of Adam, ND on window, white paper, no lens, camera was approx 10 cm/5 inches away from the white paper

That does not look to have any colour cast in the corners. Seems uniform.

Also I have the mount of the camera pressed against the paper. The mean there is no chance of light coming from the side. All the sensor can "see" is a flat white surface.

So please tell me if i am doing something wrong hereI opened
I opened a single dng in photoshop 6. reduced the exposure by -2 stops. applied global correction toward green to get it as natural as I can and it looks like I get some magenta cast on the upper left corner.

am I correct?

It does appear to have a slight magenta cast to the top right but it also have some heavy vignetting on the bottom left. This leads me to suspect you did not shoot a flat evenly light surface.

You also need to make sure that none of the image is clipping in camera.

How did you shoot the test?

So please tell me if i am doing something wrong hereI opened
I opened a single dng in photoshop 6. reduced the exposure by -2 stops. applied global correction toward green to get it as natural as I can and it looks like I get some magenta cast on the upper left corner.

am I correct?

It does appear to have a slight magenta cast to the top right but it also have some heavy vignetting on the bottom left. This leads me to suspect you did not shoot a flat evenly light surface.

You also need to make sure that none of the image is clipping in camera.

How did you shoot the test?

Same as you - nd taped to window, white 8x10 paper taped to the ND and window, camera at 5 inches away no lens

Same as you - nd taped to window, white 8x10 paper taped to the ND and window, camera at 5 inches away no lens

Press the mount of the camera against the paper. This way there is no change of light from the side causing the test to have unwanted light.

Got a similar result Eli, also just bump the sat a few notches, it becomes clearer. Just like you would with your normal footage.

... Not to mention 4K DCI 4:1 Raw is only around 75Mb/s how awesome is that lol...

Brandon, I calculate DCI 4K raw 4:1 including audio would be a little under 80MB/s at 24 fps. Perhaps you meant about 75MB/s (megabytes), not 75Mb/s (megabits).

without a lens pressed against

5.jpg (307.04 KiB) Viewed 22788 times

ok so here it is with the camera pressed against the surface and with the 18-35mm f 1.8 at a distance of 3 inches from the window

5.jpg

ok so here it is with the camera pressed against the surface and with the 18-35mm f 1.8 at a distance of 3 inches from the window

...not too bad, but not very good either...

ok so here it is with the camera pressed against the surface

Got a DNG so we can check the scopes in Resolve?

5.jpg

ok so here it is with the camera pressed against the surface and with the 18-35mm f 1.8 at a distance of 3 inches from the window

That doesn't look too bad man. I wouldn't worry about that, personally. On an image, it doesn't seem like it would do much. But that's my take on it. I guess you'd have to check actual footage and see how bad it is.

5.jpg

ok so here it is with the camera pressed against the surface and with the 18-35mm f 1.8 at a distance of 3 inches from the window

That doesn't look too bad man. I wouldn't worry about that, personally. On an image, it doesn't seem like it would do much. But that's my take on it. I guess you'd have to check actual footage and see how bad it is.

Mine is a bit worse, and if you apply some contrast and saturation like you would do with your footage it's getting a bit more noticeable.

But indeed, if you don't see it in you real life footage, then don't bother.

5.jpg

ok so here it is with the camera pressed against the surface and with the 18-35mm f 1.8 at a distance of 3 inches from the window

That doesn't look too bad man. I wouldn't worry about that, personally. On an image, it doesn't seem like it would do much. But that's my take on it. I guess you'd have to check actual footage and see how bad it is.

Mine is a bit worse, and if you apply some contrast and saturation like you would do with your footage it's getting a bit more noticeable.

But indeed, if you don't see it in you real life footage, then don't bother.

I'd definitely have to see real footage, but I mean, JB's Leica grab is pretty shocking how much color shift there is... but... look at his photos. They're elfin' STUNNING.

here you go.
2 DNG's
one with lens on one without.https://spaces.hightail.com/receive/jQp ... 9ucy5jb20=

[quote="PaulDelVecchio

I'd definitely have to see real footage, but I mean, JB's Leica grab is pretty shocking how much color shift there is... but... look at his photos. They're elfin' STUNNING.[/quote]

I posted some in one of my previous posts.

[quote="PaulDelVecchio

I'd definitely have to see real footage, but I mean, JB's Leica grab is pretty shocking how much color shift there is... but... look at his photos. They're elfin' STUNNING.

for me there is to much color shift of the sensor. Does not fit to the images at all....

I am in touch with BMD and send them the 2 dng's to asses.
I really love this camera and hope all is well.

April 1 or not, I would not announce a camera unless I am sure about it. And they announced it after April 1 2015.
See, I have been used to BMD with their post-production products. They are brilliant. I am quite used to their production products too, they are not. I have faced my problems with my BMCC BMPC and after this UM46, I am not sure whether I'll ever trust a camera from BMD which they build from scratch.
My suggestion to BMD would be that you take over some camera company like Red or Canon and build your camera based on their expertise. The way you did with your DaVinci Resolve or Fusion. That would be more practical!

... Not to mention 4K DCI 4:1 Raw is only around 75Mb/s how awesome is that lol...

Brandon, I calculate DCI 4K raw 4:1 including audio would be a little under 80MB/s at 24 fps. Perhaps you meant about 75MB/s (megabytes), not 75Mb/s (megabits).

I definitely meant MB! when typing and using the Shift key I always miss the second letter that needs to be in CAPs on abbreviations. Yea 80MB/s is not bad thats about 86 mins of footage with two 256GB cards. Pretty frugal on bandwidth if you ask me.

here you go.
2 DNG's
one with lens on one without.https://spaces.hightail.com/receive/jQp ... 9ucy5jb20=

Paul I agree with your statement above, this doesn't look too bad....in fact looks pretty good actually.

For everyone who is wondering about their own camera, BMD's or otherwise....I would follow Adam's advice from a testing protocol perspective and capture a RAW image with flat light.

Then import into Resolve and do a quick gain balance and once balanced, then check for consistency.

Go back a page in this thread and have a look at the RAW Alexa RT image that John Brawley posted and look at the before and after. As John said the image looked a tad Magenta in pure RAW but RAW is RAW guys, that is the whole idea and it requires a little bit of human effort.

What everyone should really be checking for is consistency....so Eli, with a quick gain balance across the color channels your image looks pretty good.

Falloffs at the edge aren't too bad...after all we are talking pseudo scientific tests so am thinking these are well within expectations.

EliBalanced.jpg (515.05 KiB) Viewed 22628 times

EliParade.JPG (66.59 KiB) Viewed 22628 times

Shooting RAW, means you gotta get used to going through a simple quick process before applying them LUT's and messing around with colors and all those other toys we get access to in the super cool editing tools..... and it really is quick and easy to balance an image.

Then you always have a known starting point before messing with knobs and dials.

What you should be precious about is that you have a sensor that produces consistency across the frame for each of the base colors, which from what I can see this 4.6K camera seems to do....maybe there are a few out there in the wild that need a check but most of what I see ends up not being a problem.

Having been one of the people who initially made a bunch of noise about Magenta, it is worth me maybe reiterating what all my emotion was about.

It wasn't that I was thinking the camera was crap, nope...what had me beside myself was all the footage being put up online that was basically flawed from a grading 101 perspective (purple heaven) and everyone banging on about how amazing it was and bang on they did.

So if you are keen to check your camera, follows Adam's test advice, use the Paradey Scopey Thingos in Davinci Resolve (that's why they are there) and then at the minimum, just balance the gain across the colors and maybe keep this as a preset or save it in your memory for when you are grading real footage.

As we have seen above, even RAW images from a $70,000 camera can use a bit of love.



Step one is to make sure you have the right RAW settings but don't touch anything else, until you have balanced gain as you will only get lost in the grade if you start friggen around with tint and temp when the basic sensor characteristics havn't been sorted.

BalanceGain2.JPG (41.39 KiB) Viewed 22573 times

Then balance gain with the bars or just remember the numbers and as you can see, there is stuff all in the percentage change but our eyes see color like we are hawks and slight differences in the numbers make a massive difference in the colors we see.

BalanceGain.JPG (45.19 KiB) Viewed 22576 times

For some perspective....

Thing is that I've been using this camera to shoot GREAT photos.

...the point I make is that with the fault above, I captured the photos below, and I'd say it's pretty hard to pick it.

JB

Some very relevant pictures you posted John. You are a master of images & light. Very impressive captures, your shots are just stunning.

Noam is a filmmaker and one of the beta testers for the 4.6K. He says he's been shooting with it for a few months now.

He also made the statement that he's using the 4.6K as his A camera on paid work (his Dragon now becoming his B cam).

That is a very bold statement Noam makes Ivon. I had signed up to attend the session too, but completely forgot about it after I got involved in my current project (which was shot on the RED Dragon).

Considering the RED Dragon has such a well established workflow, and produces a very beautiful image (its IR pollution notwithstanding) am I am intrigued to know what would have prompted him to make that decision. URSA Mini 4.6 with its cDNG files is not exactly the easiest workflow compared to the RED?

Hello,

I wanted to contribute to this debate by saying that when working with Raw files, one must know perfectly his tools. When using a new camera that Adobe hasn't referenced, you will (if adobe allows you to open the file!) have under the camera setting tab a preconfigured base setup (processus).
This influence greatly the result of the color debayering process. It is the same as using the correct color science in resolve raw tool.

Therefore, with the Ursa mini 4.6k, until Adobe hasn't came up with an camera Raw upgrade including the Ursa Mini 4.6k; no test shall be correctly done in their softwares...

Please have a look at the attachment to better understand.

(And as we are on the 1st of April: if you don't like vignetting, you can paint the inside of camera, around the sensor, with a very glossy white paint...or green for the magenta cast

Noam is a filmmaker and one of the beta testers for the 4.6K. He says he's been shooting with it for a few months now.

He also made the statement that he's using the 4.6K as his A camera on paid work (his Dragon now becoming his B cam).

That is a very bold statement Noam makes Ivon. I had signed up to attend the session too, but completely forgot about it after I got involved in my current project (which was shot on the RED Dragon).

Considering the RED Dragon has such a well established workflow, and produces a very beautiful image (its IR pollution notwithstanding) am I am intrigued to know what would have prompted him to make that decision. URSA Mini 4.6 with its cDNG files is not exactly the easiest workflow compared to the RED?

I certainly can't speak for him, Ashok, however, the statement came as he was comparing two shots he recently took at a factory with both the Mini 4.6 and Dragon. The Dragon shot had a very noticeable green tint -- a known quality of Dragon's. Noam said the Mini looked more like what he was seeing with his eyes and therefore he preferred that image. This seemed to be the basis for his decision in choosing the Mini as his A cam. He did not discuss the difference in workflow, however he did mention a few times during his presentation that he shoots a lot in ProRes so he may not be as concerned about a workflow using raw files.

Noam -- are you around the forum and care to comment?

Hello,

I wanted to contribute to this debate by saying that when working with Raw files, one must know perfectly his tools. When using a new camera that Adobe hasn't referenced, you will (if adobe allows you to open the file!) have under the camera setting tab a preconfigured base setup (processus).
This influence greatly the result of the color debayering process. It is the same as using the correct color science in resolve raw tool.

Therefore, with the Ursa mini 4.6k, until Adobe hasn't came up with an camera Raw upgrade including the Ursa Mini 4.6k; no test shall be correctly done in their softwares...

Please have a look at the attachment to better understand.

(And as we are on the 1st of April: if you don't like vignetting, you can paint the inside of camera, around the sensor, with a very glossy white paint...or green for the magenta cast

To illustrate my point with one of the dng from Eric Ackman having "magenta cast issue" using Camera Raw... I use his methodology; And tweeking the camera profile and the vignetting, i could obtain a almost perfect white...

Hey Raphaël,

I had some time to try Adam's method of testing by taping ND6 gel to window, taping a piece of white paper over that, and then snagging some frames with the lens housing pressed directly onto the paper itself. 5600K balanced, ISO 800.

Andrew, I followed your instructions from the previous post and balanced only using the primary bars. Once I added saturation and contrast to the BMD 4.6k film gamma to make the discrepancy clear, this is what I ended up with:




The resident Resolve experts here who are more talented than I, have a look at the DNG:

http://pigeon.to/1X4Eoi3

At any rate, I've been shooting with 4K DCI in the meantime which has cropped out the issues I've been experiencing. As Brandon pointed out the bandwidth is great as well as the faster RS readout speed. Also I've been loving ISO 1600, looks great!

Hey Raphaël,

I had some time to try Adam's method of testing by taping ND6 gel to window, taping a piece of white paper over that, and then snagging some frames with the lens housing pressed directly onto the paper itself. 5600K balanced, ISO 800.

Andrew, I followed your instructions from the previous post and balanced only using the primary bars. Once I added saturation and contrast to the BMD 4.6k film gamma to make the discrepancy clear, this is what I ended up with:




The resident Resolve experts here who are more talented than I, have a look at the DNG:

http://pigeon.to/1X4Eoi3

At any rate, I've been shooting with 4K DCI in the meantime which has cropped out the issues I've been experiencing. As Brandon pointed out the bandwidth is great as well as the faster RS readout speed. Also I've been loving ISO 1600, looks great!

Hi Eric,

I had a look at your dng. In order to white balance it I had to push the color temp to 8580K ... and the tint to -12.35 and here is the result PUSHING SATURATION TO 100 and CONTRAST to 100 (which is A LOT in real scenario raw setups!!)
Even in Raw, the color temp is a bias. Which means that the sensor has an optimal color temp and if your lighting is very far from it (8580k is a very strong color temp shift), the sensor will react poorly.
Did you use a uncalibrated computer screen as a light source?

Hey Raphael that was my mistake as the window wasn't facing DIRECT sunlight, but ambient shade so your adjustments are correct. My bad. Please disregard my resolve screen grabs folks, I've been drinking.

So putting aside that the color temperature was off because I'm a hack, and after balancing it in your opinion would you say everything appears normal?

Did you use a uncalibrated computer screen as a light source?

Or perhaps a tinted window?

Eric, this is definitely magenta on the left and green on the right as shown by the corners and the different slopes of green versus red and blue on the scopes. I think that needs BMD's attention.


Sent from my iPad using Tapatalk

Hey Raphael that was my mistake as the window wasn't facing DIRECT sunlight, but ambient shade so your adjustments are correct. My bad. Please disregard my resolve screen grabs folks, I've been drinking.

So putting aside that the color temperature was off because I'm a hack, and after balancing it in your opinion would you say everything appears normal?

I can see the problem immediately...it's not Single Malt so therefore is inconsistent in the color range and this is my take using 'Triple Filtered Vodka'.

Jokes aside I am guessing what you are seeing in the falloffs is actually consistent with the falloff of the light coming in the front of the camera.

Given the pseudo science, paper....pointing out a window etc....when the three curves are added together they overlay pretty well.

SingleMaltBalanceda.jpg (746.06 KiB) Viewed 22439 times

SingleMaltBalancedb.jpg (71.84 KiB) Viewed 22439 times

Eric, this is definitely magenta on the left and green on the right as shown by the corners and the different slopes of green versus red and blue on the scopes. I think that needs BMD's attention.


Sent from my iPad using Tapatalk

+1


Try another test, with water this time and if the same magenta top left/green bottom right comes up, you should ask BM techs for advise or return policy... But this makes me remember the Canon 5D and it's very variable sensor! (Try to make a grey background photoshoot for hours with a 5D and see the face of the postproduction guys!!! )

Taping a piece of white paper over an ND gel still may be a questionable technique for two reasons that come to mind. The ND will likely be very consistent edge to edge, but could not it introduce some colour bias which as we know can even vary from one ND to another both in terms of various brands of filter and strengths within a brand? So you would have a starting adjust meant for every filter and every strength?

And even if you don't have a lens on the camera, the light falling on your paper has an overall temperature but the light on one area of the paper could vary across the page. You can account for the overall temperature, but the variations across the page leave room for uneven results. I think this needs to be addressed in a more controlled manner, before we all start creating something like a FIRST LUT or project profile for the camera which isn't as accurate as we assumed it would be. After all that questionable effort, it might make more sense just to work with actual footage and adjust that as we normally would.

So an overall cast or a predictable vignette isn't an issue. When we have an uneven distribution of colour or luminosity other than minor corner vignettes, I think that's needing a solution from BMD as that should be considered a sensor flaw, not a sensor characteristic.


Sent from my iPad using Tapatalk

I see in the time it took me to think about my post, others came to a similar conclusion.


Sent from my iPad using Tapatalk

Whilst kicking back I recommend everyone watch this as Bob did everything you see here in a normal everyday plane....which had very few controls, but achieved massive accuracy in everything he did.

Reckon we have too many options with too many levers, switches, dials etc....and we are overlooking the simple things.

Thanks for the responses everyone. I’ll update this thread if I hear back from BM support. I’m going to continue shooting 4K and learn more everyday. The DR on this thing vs my BMPC, wow!

Another test following Adam's methodology. I wanted to take the potential ND IR pollution out of the equation. So, I did a little different set up. Camera up against the white paper/glass on to my front porch with very diffuse lighting.

White paper stuck on glass NO ND
Camera without lens pressed up against the white paper
Camera set to: RAW 3:1, Full Sensor ISO800, 23.976, 22.5º (to bring exposure down to around 768 ire). I don't believe that a fast shutter will make a difference in this case other that lower exposure). WB and minor tint tweak in Resolve

Full Frame Scopes

UM4.6k Full Frame 4608x2592.jpg (878.56 KiB) Viewed 22399 times

Same test but with 4k Window mode
Camera set to: RAW 3:1, 4k window mode, ISO800, 23.976, 22.5º (to bring exposure down to around 768 ire). I don't believe that a fast shutter will make a difference in this case other that lower exposure). WB and minor tint tweak in Resolve

4k Window Mode

UM4.6k 4096x2160 Window.jpg (909.25 KiB) Viewed 22397 times

4k Window Scopes

UM4.6k 4096x2160 Window Scopes.jpg (274.35 KiB) Viewed 22397 times

I heard back from BMD they examined my DNG and concluded that my sensor does not have a problem so I am happy... some (more) knowledgeable people than me looked at my DNG here as well and said the same.

Scott, looking good. RGB consistent levels. Some falloff in the corners open gate but pretty much gone in DCI 4K.


Sent from my iPad using Tapatalk

Scott, looking good. RGB consistent levels. Some falloff in the corners open gate but pretty much gone in DCI 4K.


Sent from my iPad using Tapatalk

I am really please! This camera really is pretty amazing. I have to keep pinching myself.

On another note, I had to send my camera back on Monday to have #2 Cfast card slot replaced - it had a bit of worrisome glitch. Marce, however, got it back to me today for a shoot this weekend! Awesome customer service.

I never really felt my sensor had a serious issue, but I've come to the conclusion that its good to go. I tested this against a white piece of paper backlit by a daylight source. Shutter angle was set to 15 degrees 4K and 4.6K 22 degrees.

MagTestOk.jpg (127.66 KiB) Viewed 22350 times

MagTestOk1.jpg (128.59 KiB) Viewed 22350 times

I never really felt my sensor had a serious issue, but I've come to the conclusion that its good to go. I tested this against a white piece of paper backlit by a daylight source. Shutter angle was set to 15 degrees 4K and 4.6K 22 degrees.

MagTestOk.jpg

MagTestOk1.jpg

Wow, looks excellent.

My suggestion is to immediately use these images and the scopes to develop a baseline which you can use again and again....kinda like a reference starting point.

It seems most people are diving straight into Temp and Tint and the goodness knows what, which might be fine for the few that have perfect colour vision and can happily 'hand mix' from any starting point.....but just means everyone else will get lost in the grade.

Taping a piece of white paper over an ND gel still may be a questionable technique for two reasons that come to mind. The ND will likely be very consistent edge to edge, but could not it introduce some colour bias which as we know can even vary from one ND to another both in terms of various brands of filter and strengths within a brand? So you would have a starting adjust meant for every filter and every strength?

And even if you don't have a lens on the camera, the light falling on your paper has an overall temperature but the light on one area of the paper could vary across the page. You can account for the overall temperature, but the variations across the page leave room for uneven results. I think this needs to be addressed in a more controlled manner, before we all start creating something like a FIRST LUT or project profile for the camera which isn't as accurate as we assumed it would be. After all that questionable effort, it might make more sense just to work with actual footage and adjust that as we normally would.

Rick no one said it was a perfect test. What I was trying to achieve was a repeatable test that was not influenced by other light or camera position.

Pointing the camera at a wall is never gonna get you consistent results. Getting perfectly flat light across the wall is difficult and if the camera is not perfectly square on you'll have inconsistent levels across the frame. Add to that light coming in from the sides make that test very flawed.

Yes the ND could have a cast. So could the paper. However that cast would be pretty consistent across the frame and the area being captured is only about 3cm (vs a few meters of a wall). That cast can be corrected with the global white balance and tint controls in the RAW tab in Resolve.

My technique has much less chance of is not being consistent across the frame and clean enough to show up an issue like magenta/green corners. You would see in the scopes that there is greater shift of one colour. It's a test anyone can setup and it would give you a starting point to raise the issue with BM support if you had an issue.

The only way you could do the test more accurately is in a lab with colour and light measuring instruments. Very few of us have access to that.

Taping a piece of white paper over an ND gel still may be a questionable technique for two reasons that come to mind. The ND will likely be very consistent edge to edge, but could not it introduce some colour bias which as we know can even vary from one ND to another both in terms of various brands of filter and strengths within a brand? So you would have a starting adjust meant for every filter and every strength?

And even if you don't have a lens on the camera, the light falling on your paper has an overall temperature but the light on one area of the paper could vary across the page. You can account for the overall temperature, but the variations across the page leave room for uneven results. I think this needs to be addressed in a more controlled manner, before we all start creating something like a FIRST LUT or project profile for the camera which isn't as accurate as we assumed it would be. After all that questionable effort, it might make more sense just to work with actual footage and adjust that as we normally would.

Rick no one said it was a perfect test. What I was trying to achieve was a repeatable test that was not influenced by other light or camera position.

Pointing the camera at a wall is never gonna get you consistent results. Getting perfectly flat light across the wall is difficult and if the camera is not perfectly square on you'll have inconsistent levels across the frame. Add to that light coming in from the sides make that test very flawed.

Yes the ND could have a cast. So could the paper. However that cast would be pretty consistent across the frame and the area being captured is only about 3cm (vs a few meters of a wall). That cast can be corrected with the global white balance and tint controls in the RAW tab in Resolve.

My technique has much less chance of is not being consistent across the frame and clean enough to show up an issue like magenta/green corners. You would see in the scopes that there is greater shift of one colour. It's a test anyone can setup and it would give you a starting point to raise the issue with BM support if you had an issue.

The only way you could do the test more accurately is in a lab with colour and light measuring instruments. Very few of us have access to that.

Adam is right on the money when it comes to the 'MacGyver' sensor test as A4 Paper comes from 30 Tonne reels and is surprisingly tightly controlled and monitored.

"During the paper making process, whiteness and brightness is computer controlled and constantly monitored for consistency. One sheet of A4 paper may be tested at 20 points across its surface during production. In the event of a colour or whiteness variation, there is a process that scans the sheet as it is running on machine and adjustments are made automatically to pre-set specifications as the system is alerted. This tight colour control is necessary due the natural variations in shade of pulp and ensures we deliver the same level of brilliant whiteness every time."

http://www.reflex.com.au/perfect-paper/reflex-quality

Thanks, Andrew and Adam. As a quick and accessible test, I'm sure it's extremely useful. I'm thinking BMD must have some lab capability to measure their sensors but that would be for spot quality control likely on a production run, not every single sensor. Thanks for your further explanations.


Sent from my iPad using Tapatalk

I would be careful when applying ND gel on a glass surface like a window. This could lead to so called Newton's Rings

I would be careful when applying ND gel on a glass surface like a window. This could lead to so called Newton's Rings

I think you can see it in mine