Hunting Magenta: 10 different lenses and 2 URSA Mini 4.6k’s

Hi Everyone,

I just received my Ursa Mini 4.6k less than a week ago and Benton Collins had likewise just received his camera back from BMD after RMA’ing it, specifically for the magenta corners issue. We figured out that we're neighbors and decided to run a test together. We went with real world conditions (in the very real world of Red Hook) with a variety of lenses on both cameras, side-by-side, to see what (if any) differences there were. This test is only concerned with color uniformity. You may notice slight variations in exposure and that’s because the Roki’s are rated in T-Stops and the Zeiss in F-stops. Again, this test is only for uniformity. The RAW DNG’s and all source files are at the end of the post so you can process them yourself. Give this next section a read; it’ll probably answer any Qs you might have about the test:

Methodology/Notes:
-As a control, we also shot with a Canon 6D in both raw and h.264 on four of the lenses and on a BMPCC on the 28mm. These cameras are “known good” and are great to compare against.
-We ran the test without IR filtration for the majority to keep the amount of moving parts to a minimum. We then put a Hoya IR Cut on the 35mm (both Zeiss and Rokinon) to see how IR affected things.
-In order to get a reasonable exposure on a sunny day, we shot at T8, ISO 200 and, not worrying about motion cadence, 11.25 as the shutter angle.
-We ran the tests side by side with the framing as close as possible and ran through his Rokinon lens set and my Contax Zeiss set. Just for fun, we also shot with Benton’s Rokinon 8mm fisheye and 16mm.
-My camera is listed as “A” and Benton’s is “B”. Remember, the B camera just went through the RMA process.
-The 28mm test is actually between a Nikon 28mm and the Zeiss 28mm.
-I kicked out some JPGs from ACR. The lenses render differently enough that I didn't try to make everything match. This test is only for uniformity.
-Benton's son is our talent for this and is in training to be a ninja.

Here’s a list of the lenses used:

Contax Zeiss:
28 f/2.8
35 f/2.8
50 f/1.7
85 f/2.8

Rokinon
8mm T3.8
16mm T2.2
35 T1.5
50 T1.5
85 T1.5

Nikon
28 f/2.8

I can't add more than 3 images here, so you'll have to download the dropbox file to see the full results. Here is a sample using the 28mm (note that I added a bit more saturation than usual here. Around 20 in ACR):

Zeiss 28mm
A

FahnonUM46_1_2016-04-17_0532_C0001_000061.jpg (744.66 KiB) Viewed 6194 times

B

Blackmagic URSA Mini_1_2016-04-17_1520_C0016_000023.jpg (797.43 KiB) Viewed 6194 times

6D

28mm_MG_8951.jpg (796.46 KiB) Viewed 6194 times

LINK TO RAW FILES
https://www.dropbox.com/sh/sx88uc0wgaj7 ... -_iqa?dl=0


Where things get interesting is when you throw the 6D into the mix. I think you can see a clear issue with color uniformity in this test and it becomes especially apparent when you directly compare both UM46’s, which react DIFFERENTLY in the same scenario, to the 6D. As you can see, both 4.6k cameras have issues with uniformity in the corners when using the same lens and settings. Benton and I were discussing this and both agreed that we LOVE everything about this camera, except for the glaring flaw both of ours seem to have. This is especially true after today’s awesome announcement of firmware that addresses every goddamn thing I’d want in the software. BMD has now the best camera for the money and for double the money (a working Raven is $10k). I’ve bought the original BMCC and have been using it on both professional and personal jobs for years. I have a Pocket sitting on my shelf right now that I love. I also really love this camera other than this issue and I don’t want this post to sound too negative; I simply need a sensor that performs to spec. I hope BMD sees this and responds with a solution.

Thanks for doing these tests ..now both you and Benton have some more work to do. Download software 3.2 with the new color science, and re-test. I'd love to see how much of a difference it makes. I'm in Brooklyn as well and would love to come over and spa with benton's son since I'm Training in Brazilian jiu-jitsu. You guys can record me getting my butt whipped.

Thanks for handling and uploading all the RAW files Fahnon! This was certainly an eye opening test. I for one after having received a new unit that had been throughly checked out by BM, was hoping that the magenta cast issue was behind me. Unfortunately, it's like it never left. At this point there are only two conclusions that I can make:

1. This is simply the way the new 4.6k behaves and I have a unit that functions within spec in a way that ALL OTHER 4.6k cameras will also exhibit under the same circumstances.
2. I got another lemon.

Either scenario is not a great one.

Fahnon and I went through 10 different lenses from Zeiss, Rokinon and Nikon rotating through on both of our cameras pointing at the same scene and there is a magenta cast on every single frame we took. The magenta cast ranges in severity from not quite, but almost acceptable from the Rokinon 50mm to seriously abysmal from the Zeiss and Nikon 28mm lenses.

So if we give the Rokinon 50mm a D- passing grade on this test, that leaves 9 very popular lenses, that are at best, headaches just waiting to happen on this camera. This is a situation that I am frankly dumfounded by! If this magenta cast was a known issue as it has been reported in past posts, why was this allowed to be released to the public without at least a clear cut solution?

Much of my work also involves medium format still life work in which I use a PhaseOne P45 back on an Arca Swiss technical camera. Getting magenta lens cast in an image is routine with many of the necessary swings and tilts that must be employed if a sharp image through a product is desired. The software provided by PhaseOne is Capture One Pro and it has had the ability for years to create custom LCC (Lens Cast Correction) files which effectively neutralize all uneven casts for each lens setup. This type of approach has always gone hand in hand with shooting medium format and is an expected part of the process toward a final pristine image. IF Blackmagic simply couldn't get rid of the magenta cast in the camera, they should have at least had a post fixable solution like the creation of an LCC file at the ready with the release of this camera.

I haven't given up yet and am still willing to fight and pursue a fix to this problem wether it be hardware or software, as this camera, if the magenta problem can be fixed, is simply and utterly amazing and an incredible value! (but only if it works dependably) I so desperately want to love this camera, but she is putting up quite a fight!

Thanks for doing these tests ..now both you and Benton have some more work to do. Download software 3.2 with the new color science, and re-test. I'd love to see how much of a difference it makes. I'm in Brooklyn as well and would love to come over and spa with benton's son since I'm Training in Brazilian jiu-jitsu. You guys can record me getting my butt whipped.

Already did a test in that same location today with the new firmware and it makes no difference at least in regards to the magenta issue at all.

Blackmagic URSA Mini_Magenta corner crop comparison.jpg (714.37 KiB) Viewed 6091 times

I created the composite above from the Zeiss 28mm test to more easily illustrate just how much cast appears in the sides and corners. This was from my newly replaced camera.

Nikon 28mm AIS no cast correction

Blackmagic URSA Mini_1_2016-04-17_1431_C0008_000052.jpg (791.84 KiB) Viewed 6091 times

The above is the Nikon 28mm AIS without any cast correction.

Nikon 28mm AIS with a custom LCC created and applied

Blackmagic URSA Mini_1_2016-04-17_1431_C0008_000052 2.jpg (792.41 KiB) Viewed 6091 times

The above is the same Nikon 28mm AIS image with a custom LCC file created and applied.

Thanks for doing these tests ..now both you and Benton have some more work to do. Download software 3.2 with the new color science, and re-test. I'd love to see how much of a difference it makes. I'm in Brooklyn as well and would love to come over and spa with benton's son since I'm Training in Brazilian jiu-jitsu. You guys can record me getting my butt whipped.

Already did a test in that same location today with the new firmware and it makes no difference at least in regards to the magenta issue at all.

Wow Benton I'm real sorry to hear that. Maybe you'll have to send it in again. That custom file works really well.

BTS shot with an iPad by my son Griffin. It was great meeting and working with Fahnon, we had a lot of fun hunting the magenta beast of the 4.6k, but missed out on having a slice of Steve's Authentic Key Lime Pie, who's shop was just to the left side of us, because they closed before we wrapped up! That should tell you just how dedicated we were! If you're ever in Brooklyn, it's worth a trip to Steve's who makes the absolute best Key Lime Pie in the world and a real Red Hook fixture!

IMG_0718.JPG (461.58 KiB) Viewed 6058 times

FullSizeRender 2.jpg (474.01 KiB) Viewed 6058 times

FullSizeRender.jpg (515.64 KiB) Viewed 6058 times

I have a bellyfeel that I understand the physics of the magenta process. Unfortunately I have no Ursa Mini 4.6k, so I can only recommend the experiment to be produced by somebody else.

Take a piece of cardboard (or maybe some black plastic box) and produce 3 pinhole "lenses" with diameter of the hole, say, 2mm, with hole to sensor distance, say, 15mm, 50mm and 100mm. Use the juice tetrapack cardboard and duct tape to avoid particles that may appear if you use a simple paper cardboard. Be EXTREMELY careful not to touch the sensor with cardboard.

Shoot an uniformly lighted white sheet of paper. Look at results.

If it goes as I expected your 15-mm pinhole should give a field with white middle and magenta edges. 100mm should give you an uniform field. Also, if you remove the lens altogether and shoot a white sheet it should be uniform too.

I'd be happy if my expectations fail.

This is a shot from the Canon 6D with a Zeiss 50mm lens, the exact same Zeiss 50mm that caused a magenta cast in the 4.6k. The 6D is a full frame 35mm still camera that has a sensor that is MUCH LARGER than the S35mm sensor on the 4.6k. It is a well known fact that as sensor size increases, so does the likelihood that lens ray angle cast becomes much more of an issue. So why does the 6D produce a virtually cast free image and the URSA Mini 4.6k produces a cast using the same lens? Not only is the sensor in the Mini smaller, which should give it an advantage toward combatting lens angle casts using virtually any lens over the 6D, but it really should have no trouble at all here with the 50mm, since it is also cropping and using just the sweet center portion of this lens! I'd very much appreciate if Blackmagic can explain this counter intuitive evidence.

_MG_8947 2.jpg (861.16 KiB) Viewed 6000 times

Above: Canon 6D with Zeiss 50mm

Blackmagic URSA Mini_1_2016-04-17_1502_C0014_000004.jpg (798.13 KiB) Viewed 6000 times

Above: URSA Mini 4.6k (B camera) with same Zeiss 50mm

Hi Fanon & Benton,

great initiative...

I was personally very interested in Benton's Camera being that it had been RMA'd.

I downloaded and tested Benton's 50mm Rok (I Believe).

I could not tell on the composition of that Image where exactly Benton's Camera would exhibit any Manufacturing Flaws.

What I DID experience was that the Camera's Matrix is setup so that achieving a neutral image by adjusting temperature and tint only is impossible. There is always going to be a 'pink' flare to image due to the green channel's representation. However, after setting temp. and tint to 'as close as possible' further manipulation the green component is fairly simple...

That said, I do not believe that any image coming off any camera should be neutralizable by adjusting temp. and tint controls ONLY. If not, the Camera Matrix needs a tweak and I am sure we're going to get one via a firmware update... Perhaps 3.2 is already delivering that.

Anyway, I would be interested in hearing where exactly you believe that the camera (Benton's) is falling of wrongly..

-- Best Morten

nuCAMB.jpg (746.49 KiB) Viewed 5867 times

This is a shot from the Canon 6D with a Zeiss 50mm lens, the exact same Zeiss 50mm that caused a magenta cast in the 4.6k. The 6D is a full frame 35mm still camera that has a sensor that is MUCH LARGER than the S35mm sensor on the 4.6k. It is a well known fact that as sensor size increases, so does the likelihood that lens ray angle cast becomes much more of an issue. So why does the 6D produce a virtually cast free image and the URSA Mini 4.6k produces a cast using the same lens? Not only is the sensor in the Mini smaller, which should give it an advantage toward combatting lens angle casts using virtually any lens over the 6D, but it really should have no trouble at all here with the 50mm, since it is also cropping and using just the sweet center portion of this lens! I'd very much appreciate if Blackmagic can explain this counter intuitive evidence.

Hi Benton,

perhaps I am stating the obvious... But a cast has the potential of going in 4 directions. Red, Blue, Magenta, Green. Or temperature, Tint.

Most casts on most cameras are resulting in a pleasing red-blue cast. Perceived as too warm or too cold. And would not cause folks to raise a flag. If the cast is producing an unpleasant magenta-green color folks tend to raise a flag, as they are not used to that. The cast regardless of color is a cast.

If the cast is not naturalizable by adjusting either or both the temperature/tint sliders - the matrix describing the colors is off. BUT adjustable via a firmware upgrade.

With all the images from my camera which was broke and all the ones I have tested on this forum from both broke and perfectly working cameras I am seeing two issues... One being the major issue of non-uniformness. Which is a manufacture issue. Two being something which is inherent in all images - a matrix issue which makes it impossible to achieve a neutral image by simply adjusting the temperature section sliders.

People should not have to resolve to advanced grading to remove ANY cast regardless of color.
Such a cast is most visible by looking at the neutral colors in the image but ALSO when looking at the colors in an image. I.e. ALL natural occurring magenta colors are oversaturated and SUCH is NOT an indication of wrong white balance but a strong indicator of a matrix not describing its colors correctly.

This is a shot from the Canon 6D with a Zeiss 50mm lens, the exact same Zeiss 50mm that caused a magenta cast in the 4.6k. The 6D is a full frame 35mm still camera that has a sensor that is MUCH LARGER than the S35mm sensor on the 4.6k. It is a well known fact that as sensor size increases, so does the likelihood that lens ray angle cast becomes much more of an issue. So why does the 6D produce a virtually cast free image and the URSA Mini 4.6k produces a cast using the same lens? Not only is the sensor in the Mini smaller, which should give it an advantage toward combatting lens angle casts using virtually any lens over the 6D, but it really should have no trouble at all here with the 50mm, since it is also cropping and using just the sweet center portion of this lens! I'd very much appreciate if Blackmagic can explain this counter intuitive evidence.

Hi Benton,

perhaps I am stating the obvious... But a cast has the potential of going in 4 directions. Red, Blue, Magenta, Green. Or temperature, Tint.

Most casts on most cameras are resulting in a pleasing red-blue cast. Perceived as too warm or too cold. And would not cause folks to raise a flag. If the cast is producing an unpleasant magenta-green color folks tend to raise a flag, as they are not used to that. The cast regardless of color is a cast.

If the cast is not naturalizable by adjusting either or both the temperature/tint sliders - the matrix describing the colors is off. BUT adjustable via a firmware upgrade.

With all the images from my camera which was broke and all the ones I have tested on this forum from both broke and perfectly working cameras I am seeing two issues... One being the major issue of non-uniformness. Which is a manufacture issue. Two being something which is inherent in all images - a matrix issue which makes it impossible to achieve a neutral image by simply adjusting the temperature section sliders.

People should not have to resolve to advanced grading to remove ANY cast regardless of color.
Such a cast is most visible by looking at the neutral colors in the image but ALSO when looking at the colors in an image. I.e. ALL natural occurring magenta colors are oversaturated and SUCH is NOT an indication of wrong white balance but a strong indicator of a matrix not describing its colors correctly.

Thanks for your insight and input Morten! I agree with you 100% that people should not have to resort to advanced post correction for this problem, plus it looks like that isn't even an option at this point in time. Creating a LCC file is the only way I know of dealing with this non-uniformity cast issue and that only exists in Capture One Pro at this moment, which is a stills only program. Here's the bottom line: If a larger sensor camera like the 6D delivers uniform color straight from the RAW file, so should the URSA Mini 4.6k. It can clearly be done.

Below is a shot taken the next day after the tests using the updated camera firmware 3.2

Firmware 3.2 Nikon 28mm AIS cast un-corrected

Blackmagic URSA Mini_1_2016-04-18_1744_C0025_000092.jpg (812.15 KiB) Viewed 5734 times

Below is the same shot with a custom LCC file created and applied within Capture One Pro: This color uniformity is what the camera should be delivering straight out of the box!

Firmware 3.2 Nikon 28mm AIS cast corrected with custom LCC file

Blackmagic URSA Mini_1_2016-04-18_1744_C0025_000092 3.jpg (811.22 KiB) Viewed 5734 times

That's disappointing, Benton. I'm also going to re-test with FW3.2 to see if it makes any difference in my camera.

Just a quick update. I called BMD to RMA the camera yesterday. They were super nice and said they're aware of the issue, are waiting to hear back from engineering on it and that they'd like to see the DNGs I have. So basically, they're wisely not taking my camera back because they don't know what's wrong with them yet. That was actually a relief because the company line here was blaming lenses entirely and one other user over on bmcuser posted an email of them saying it was just the "look" of the camera.

So I sent all the raw files from this thread over to them. Will post again when they've had a chance to look everything over and get back to me...

That is fantastic news that they are aware and working on it! 90% of the frame looks so dang good... I'm confident that they will get 100% looking great. Perhaps this will be similar to the the BMCC flange issue of 2012... and will be forgotten soon.

As one of many customers that have pre-ordered the Ursa Mini 4.6K after the NAB 2015, I am pretty confident, that by NAB 2017 they will have some progress regarding the magenta issue. That is about 1 year from now. Until then - FS7 will be the better camera

Just bumping this thread because we've got a lot of new blood here who are asking questions that are answered by this round of testing. Key takeaways:

-Most importantly, the Canon 6D and BMPCC right next to the affected Ursa Minis had NO ISSUES, even when using the same lenses and apertures.
-IR Cut doesn't make a (noticeable) difference.
-RMA'ing the camera was not successful for Benton.
-Focal length didn't seem to matter, but it's totally possible that aperture did.
-The two cameras actually behaved slightly different under the same circumstances.